Bacteria as genetically programmable producers of bioactive natural products

Next to plants, bacteria account for most of the biomass on Earth. They are found everywhere, although certain species thrive only in specific ecological niches. These microorganisms biosynthesize a plethora of both primary and secondary metabolites, defined, respectively, as those required for the growth and maintenance of cellular functions and those not required for survival but offering a selective advantage for the producer under certain conditions. As a result, bacterial fermentation has long been used to manufacture valuable natural products of nutritional, agrochemical and pharmaceutical interest. The interactions of secondary metabolites with their biological targets have been optimized by millions of years of evolution and they are, thus, considered to be privileged chemical structures, not only for drug discovery. During the last two decades, functional genomics has allowed for an in-depth understanding of the underlying biosynthetic logic of secondary metabolites. This has, in turn, paved the way for the unprecedented use of bacteria as programmable biochemical workhorses. In this Review, we discuss the multifaceted use of bacteria as biological factories in diverse applications and highlight recent advances in targeted genetic engineering of bacteria for the production of valuable bioactive compounds. Emphasis is on current advances to access nature’s abundance of natural products. This Review highlights bacteria as biological factories in diverse applications, with an emphasis on targeted genetic engineering for the production of bioactive natural products.

Microorganisms -which include bacteria -account, next to plants, for most of the biomass 1 on Earth.Not only do they occupy nearly all ecological niches 2 but they also form complex associations with other organisms.Bacteria and archaea are prokaryotes -they are microorganisms that lack a membrane-bound nucleus, mitochondria or any other membrane-bound organelle -and represent two of the three domains of life.The third domain of life, the eukaryotes, might have evolved either from a single ancestral lineage via successive mutations or by the symbiotic association of prokaryotic cells with archaea 3 .Microorganisms are defined as microscopic organisms, which are capable of existing in single-celled form.All prokaryotes are microorganisms, whereas among eukaryotes, only protists, fungi and green algae are considered as microorganisms (even though some of the fungi belong to the largest organisms on the globe measured by areal expanse).Although a few species of bacteria and fungi are human pathogens, most do not directly interact with humans but have important roles for our well-being as commensals.
The usefulness of microorganisms to mankind is a result of their evolutionarily diversified metabolism, affording an impressive capacity for biocatalytic transformations 4 .Unlike any other inorganic material or chemical catalyst, bacteria can be considered living biofactories, unique in their ability to undergo evolutionary adaption in relatively short time spans.The distinct exchange of genetic information is the foundation that enables bacteria to fine-tune their chemical production for specialized applications.Beyond the fundamental biochemistry of primary metabolism, which provides evolutionary conserved metabolites that are required for the growth and maintenance of cellular functions, many bacteria produce so-called secondary metabolites 5 .These biomolecules are not essential for the survival of the organism but offer advantages under particular environmental conditions.They are optimized in the course of evolution for interactions with biological targets and are, thus, considered privileged structures 6 , exhibiting diverse biological functions allowing for their application as antifungals and antibiotics, as well as anticancer and immunosuppressive compounds.In the field of natural-product 7 discovery for pharmaceutical applications, major advances in the genetic engineering of microorganisms and the advent of affordable genome sequencing are currently transforming the process of discovering novel microbial secondary metabolites.
The aim of this Review is to highlight the impact of recent conceptual and technological developments on the use of microorganisms in various applications, on the one hand as consumables and living biocatalysts Bacteria as genetically programmable producers of bioactive natural products Joachim J. Hug 1,2 , Daniel Krug 1,2 and Rolf Müller 1,2 ✉ Abstract | Next to plants, bacteria account for most of the biomass on Earth.They are found everywhere, although certain species thrive only in specific ecological niches.These microorganisms biosynthesize a plethora of both primary and secondary metabolites, defined, respectively , as those required for the growth and maintenance of cellular functions and those not required for survival but offering a selective advantage for the producer under certain conditions.As a result, bacterial fermentation has long been used to manufacture valuable natural products of nutritional, agrochemical and pharmaceutical interest.The interactions of secondary metabolites with their biological targets have been optimized by millions of years of evolution and they are, thus, considered to be privileged chemical structures, not only for drug discovery.During the last two decades, functional genomics has allowed for an in-depth understanding of the underlying biosynthetic logic of secondary metabolites.This has, in turn, paved the way for the unprecedented use of bacteria as programmable biochemical workhorses.In this Review , we discuss the multifaceted use of bacteria as biological factories in diverse applications and highlight recent advances in targeted genetic engineering of bacteria for the production of valuable bioactive compounds.Emphasis is on current advances to access nature's abundance of natural products.
and on the other hand for the production of valuable biomolecules.We showcase mainly examples from bacteria, where an extended definition as programmable biofactories 8 comes into play, which pays tribute to their enormous impact in diverse applications [9][10][11][12][13] .The first section thus summarizes the use of the bacteria themselves -the chassis -as bacterial biofactory, for the direct use of biomass or as catalysts for versatile biotransformations.The second part underpins how the utility of these biofactories in many applications is inseparable from their specialized metabolism and emphasizes the role of genetic approaches to reprogram their capability in order to synthesize intriguing and otherwise difficult-to-access molecules.

Beyond living factories
Bacteria combine intriguing metabolic capability, evolutionary adaptivity and the aptitude to respond to external stimuli that makes them important constituents (and, in some cases, the decisive players) in a range of natural habitats, as well as in industrial, agricultural or medical applications.Given their ecological and biological relevance, and to recognize their full application potential, it is perhaps reasonable to expand our interpretation of bacteria from relatively simple life form towards the concept of complex biological machines.Hence, bacteria can be regarded as living biofactories that are able to act as versatile biocatalysts and ready to interact with a plethora of different inanimate materials, other microorganisms, living tissues and chemicals.The wide scope of their application-specific usefulness stems from unique opportunities to fine-tune not only their physical and biocatalytic characteristics but also to modulate their interaction with the environment (Fig. 1).As an instructive example, Miyamoto, Oda and colleagues recently isolated a novel bacterium Ideonella sakaiensis 201-F6 that is capable of degrading polyethylene terephthalate (PET) to use as its major energy and carbon source 14 .Interestingly, this strain produces the required two enzymes for PET degradation only when grown on PET to yield the environmentally safe monomers terephthalic acid and ethylene glycol.This particular case exemplifies the potential of microbes to adapt to rapidly changing environmental conditions.
Complex associations with other organisms show the varied roles of bacteria in their natural habitats.These associations are classified into parasitism, mutualism and commensalism.Parasitism causes numerous diseases in humans and animals, whereas infections of plants are the cause of major agricultural damage.The symbiosis of bacteria (and archaea) in soil with some plant groups (especially legumes) is noteworthy, as it enables nitrogen fixation.This symbiosis is used in the production of biofertilizers -substances containing living bacteria (or their persisting form, such as spores) -applied to seeds, plant surfaces or soil to promote growth of the agricultural plant (Fig. 1).Another use of bacteria in agriculture is biological pest control, in which bacteria related to Bacillus thuringiensis 15,16 are employed to produce pesticides and, thus, prevent crop losses caused by diverse insect pests.Along these lines, a recent study identified cepacin A (1) as a plant-protective, anti-oomycetal natural product produced by the biopesticidal bacterium Burkholderia ambifaria 17 (Fig. 1).An engineered mutant of this bacterium was shown to result in reduced levels of respiratory infection in mice while retaining its plant-protective properties and overall fitness.
A striking example of a tripartite mutualistic relationship -where, at first glance, the role of a bacterium seems less obvious -is presented by the fungus-growing ants of the tribe Attini, living with symbiotic fungi and bacterial symbionts.The ants provide the fungus with leaves as substrate for growth and, in return, the fungus supplies sugars as nutrition 18 .However, several non-mutualistic fungal species that are potentially harmful to the fungus gardens of leafcutter ants can also be found inside the nest 18,19 , including Escovopsis, a specialized fungal garden parasite.Adding to this ambivalent relationship, an actinomycete bacterium lives on the ant cuticle and consumes the nutrients provided by the ant at around 10-20% of its metabolic rate.In turn, this bacterial symbiont provides secondary metabolites such as dentigerumycin (2) 20 and cyphomycin 21 , featuring potent antifungal properties against the fungal parasite Escovopsis (Fig. 1).This example illustrates the extent of complex associations between bacteria, microbes and eukaryotic organisms in the environment.Complex mutualistic associations are also found in marine sponges, which frequently harbour bacterial symbionts to provide the host with chemical defence against predation and microbial infection, as highlighted by the discovery 22 of the bacterial symbiont Candidatus Endohaliclona renieramycinifaciens, which produces renieramycin (3).
The human microbiome is an outstanding example for complex associations between bacteria and eukaryotes, featuring downright personalized microbiota between individuals 23 (Fig. 1).In recent studies, the human microbiome has been recognized as a prolific source of bioactive natural products 24 , such as the thiopeptide lactocillin (4), a potent antibacterial, which was purified and isolated from Lactobacillus gasseri JV-V03, a human-associated bacterium that is a member of the vaginal microbiota 25 (Fig. 1).Further therapeutic applications are connected with the finding that bacteria are symbionts in the human gut flora, for example, of the intestine, where they contribute to gut immunity, provide essential nutrients by enzymatic catalysis and produce small molecules with biological activity.Current therapeutic approaches target the human microbiota diversity by supplementing probiotics, for example, Lactobacillus 26 and Bifidobacterium 27 , to treat gastrointestinal diseases, such as inflammatory bowel disease 28 and necrotizing enterocolitis 29 .Another recently introduced successful therapy involving bacteria as biologically active material is faecal transplantation 30 -the process of transferring faecal bacterial communities from healthy individuals to a recipient, to modulate the immune response in inflammatory disease.
Another extraordinary feature facilitated by intrinsic characteristics of bacterial metabolism is represented by magnetotactic bacteria (MTB) 31,32 , which enable orientation in magnetic fields 33 .Magnetotaxis is dependent on the presence of magnetosomes -intracellular nanometre-sized magnetite (Fe 3 O 4 ) and/or greigite (Fe 3 S 4 ) crystals enveloped by a membrane bilayer 34 (Fig. 1).Applications of MTB include targeted drug delivery 35 , treatment of waste water -particularly for recovery of heavy metals 36 , selenium 37 or phosphate 38 -and even generation of low-voltage electricity based on Faraday's law of electromagnetic induction 39 .Recent studies on MTB concerning the manipulation of chain orientation without impacting cell viability 40 and tuning bacterial hydrodynamics near solid surfaces with weak magnetic fields 41

Genetic manipulation Synthetic biology
Magnetite (Fe

Chemicals kanR
Fig. 1 | Bacteria as living biofactories.Bacteria produce a multitude of enzymes and metabolites associated with the primary or secondary metabolism, which contribute to complex interactions and associations in their natural habitat.A variety of methods are used to access the versatile capabilities of bacteria, including cultivation for biotransformation, strain-improvement techniques, genetic manipulation and synthetic biology to 'reprogram' their properties, leading to engineered bacteria as application-specific, fine-tuned microbial workhorses.Natural-product structures: 1: cepacin A , 2: dentigerumycin, 3: renieramycin E, 4: lactocillin.PET, polyethylene terephthalate.
Rhodospirillum rubrum, may serve as a starting point to enable the tailored use of magnetic nanostructures, which can be further developed for diverse nanotechnological and biomedical applications 42 .The targeted delivery contained within nanoliposomes to the hypoxic regions of tumours using the magneto-aerotactic migration behaviour 43 of Magnetococcus marinus strain MC-1 (reF.44 ) envisions the emulation of artificial medical nanorobots, but requires characteristics and functionalities that are beyond current technological feasibility 35 (Fig. 1).Another important aspect for bacteria-based drug-delivery concepts is the focus on the safety for potential usage in clinics, in particular, side effects such as systemic inflammatory response due to uncontrolled growth of the carrier organism.Hence, a strategy was developed in which the genetically modified organism is grown in synchronized cycles in order that lysis can be initiated at a threshold population density to release the genetically encoded cargo 12 .Further improvement of biocontainment strategies for genetically modified organisms by expanding the genetic alphabet by using unnatural base pairs 45 or synthetic amino acids 46 and synthetic protein design 47 might ease the development of safe biological materials and their adoption into the clinic (Fig. 1).Nevertheless, a recent study has shown that the pronounced exchange of genetic material can lead to the loss of previously introduced safety modifications, thus, pinpointing the importance of precautionary measures for the clinical usage of genetically modified bacteria in these scenarios 48 .
Moving from clinical to industrial applications, we note that many bacteria are employed in the form of whole-cell factories, in which the biological materials themselves are used for their catalytic functions.The applications run the gamut from bacteria that manufacture valuable chemicals, drugs, biogas and products of nutritional or cosmetic 49 interest to construction materials derived from microbiologically induced calcite precipitation 36 (Fig. 1).Corynebacterium glutamicum is one of the major workhorses in industrial biotechnology and is used for the industrial production of around 70 natural and non-natural compounds 50 .Besides the large number of proteinogenic and non-proteinogenic amino acids accessible by the fermentation of C. glutamicum, recent achievements in the bacterial production of l-lysine (120 g l −1 ) 51 , l-methionine (7.0 g l −1 ) 52 and l-isoleucine (31 g l −1 ) 53 reflect the emerging role of C. glutamicum as metabolically engineered microbial factories.To achieve these improved production rates, several adaptions in their biosynthetic pathways were conducted, such as the implementation of feedback-resistant biosynthetic enzymes, interrupting competing pathways or increasing the supply of cofactors.
Metabolic engineering of bacteria is also used for the production of sustainable biofuels 54 .For example, Rhodococcus opacus was genetically engineered and the cultivation conditions optimized to accumulate higher amounts of fatty acids from glucose.This metabolically engineered R. opacus strain could also provide a promising platform for sustainable production of numerous fatty-acid-derived products from glucose and from other unrelated carbon substrates.Not only is metabolic engineering of this well-known microbial chassis foundational for industrial microbiological applications but it has also led to the discovery of new strains with beneficial features.For example, the discovery of the first cellulolytic wild-type bacterium (Thermoanaerobacterium thermosaccharolyticum strain TG57) that can directly convert cellulose and xylan to butanol might support the production of sustainable biofuel 55 .
Not only are the enzymatic products themselves used but the enzymes themselves can be isolated and are extensively used in the textile, food 56 , paper and detergent 57 industries.Recent research has enabled the heterologous production of recombinant-spider-silk proteins (spidroins) in Escherichia coli, which feature identical primary mechanical properties to natural spider silk 58 .In addition, new approaches enabled the functionalization of recombinant spider silk with fluorophores and antibiotics such as levofloxacin 59 .
Another industrial showcase is bioleaching, that is, the extraction of metals from their ores using microorganisms.The process accelerates the oxidative dissolution of the mineral by using Fe 3+ to yield Fe 2+ and elementary sulfur (for sphalerite and chalcopyrite) or thiosulfate (for pyrite).Bacteria continue the oxidation of the ore and regenerate the chemical oxidant Fe 3+ from Fe 2+ (reF. 60 ) (Fig. 1).In contrast to the traditional heap-leaching process, bioleaching enables the isolation of metals from ores with a low concentration of the required element.Using their catalytic capabilities, bacteria are also used for the process of bioremediation 61 , for example, to recycle waste by preprocessing it for reuse by other organisms.The pollutants treated range from organic compounds derived from the petrochemical industry to textile dyes to heavy metals from diverse anthropogenic sources, such as mining, smelting and electroplating 62 .In regard to environmental applications, a noteworthy characteristic of bacterial growth is the formation of biofilms, describing the bacterial community as a multicellular collective instead of the accumulation of individual cells.The formation of a biofilm is based on a self-produced matrix of extracellular polymeric substances and offers bacteria advantages ranging from localized nutrient gradients leading to habitat heterogeneity, improved use of resources and the exploitation of metabolic effects within synergistic microconsortia, as well as tolerance and resistance to antibiotics 63 .Sewage treatment is one example of a technical process using biofilm formation: During the secondary treatment stage, waste water passes over biofilms embedded in 'biosand' filters to extract and digest organic compounds 64 .Bacteria remove mainly organic matter, while protozoa and rotifers are primarily responsible for the removal of suspended solids, including pathogens and other microorganisms.
Beyond the natural formation of bacterial biofilms, the construction of a programmable biofilm-based nanomaterial was described by Joshi and colleagues 13 .In this study E. coli was genetically reprogrammed to control the production of the bacterial extracellular matrix.Not only could this artificial matrix be attached to stainless steel but it could also be employed to immobilize full-length proteins, adhere to a specific substrate and as a template for the assembly of inorganic nanoparticles.The primary proteinaceous structural components of E. coli biofilms, called curli, consist of robust functional amyloid nanofibres with a diameter of approximately 4-7 nm that exist as extended tangled networks encapsulating the cells.Curli are formed from the extracellular self-assembly of CsgA, a small secreted 13-kDa protein 13 .A curli-based system showed multiscale patterning of single amyloid fibres and engineered curli were used for the organization of gold nanoparticles and quantum dots for nanoelectronics applications 65 .Further development of a 2D patterning circuit 66 and incorporation of the aforementioned engineered curli 65 enabled the programmable assembly of extracellular curli fibrils with functional tags into 3D patterns through the usage of pressure sensors 67 .
Taken together, the selected examples discussed above make clear how the ability of bacteria to act as biocatalysts or biological materials with appropriately tuned characteristics benefit a range of applications.Moreover, bacteria are also accurate, sensitive and specific biological sensors for environmental parameters; hence, they can directly respond to changes and are, consequently, involved in the metabolism of a plethora of chemicals in the environment.In their role as biological sensors, bacteria are capable of providing analytical insights, which would remain cryptic with conventional analytical methods based on non-biological materials.Most of these whole-cell biosensors are either unmodified-native or genetically engineered bacterial strains in which the expression of a reporter gene is driven through a promoter, which is inducible by a specific analyte and produces an output signal that is luminometric, fluorimetric or colorimetric in nature 31 .Recently, Freemont, Polizzi and colleagues 68 developed a whole-cell biosensor based on cell agglutination to detect human fibrinogen (a clinically relevant biomarker).The E. coli whole-cell biosensor displays nanobodies on its surface, which bind selectively to a target protein analyte at nanomolar concentrations.The widely used immunoassays based on immobilized antibodies on latex particles (termed latex agglutination tests) require expensive steps to isolate and purify antibodies.In contrast, the use of whole cells as a bioanalytical platform for in vitro medical diagnostics offers the same versatility and sensitivity, without the requirement for the isolation of antibodies.
In summary, these studies highlight how bacteriaboth wild-type and genetically engineered mutants -can be used as living biological machinery.Rationally engineered bacteria as biocatalysts and biofactories in diverse applications are inevitably linked to the advent of modern genetic-manipulation methods and next-generation-sequencing methods, hallmarks of the genomic era.With these tools in hand, the foundation for largely reprogrammed bacteria has been laid, which were previously only adaptable to a certain extent, and this has, in turn, opened up access to unprecedented usage scenarios 69 .Therefore, numerous developments are currently underway to use genetically modified bacteria reaching out even to the creation of engineered biomaterials with application-specific fine-tuned properties, due to the fast growth, evolvability and genetic-engineering abilities of bacteria.To name one particular example, the inexpensive fabrication of microbial biophotovoltaic cells was recently demonstrated by Nixon and colleagues 70 using a commercial inkjet printer to print a 'bioink' of cyanobacterial cells onto paper, which was finally enabled with an additional printed bioelectrode to power a small digital clock or low-power light-emitting diode.There are, in fact, numerous fascinating examples taking advantage of the genetic tools available to engineer bacteria with the prospect to design biomaterials with unprecedented capabilities, including biological computers 11 , self-healing concrete 71 and living wearable devices 72 .
Notwithstanding the fact that the aforementioned examples originate from diverse disciplines and focus on different aspects of bacteria, all of the named applications are based on the use of bacteria as living biocatalysts or as biologically derived materials resulting from their profound metabolic capability.However, access to complex new and altered natural products produced by bacteria plays an equally important role and, indeed, defines a whole area of application of its own (Fig. 1).

Secondary metabolites from bacteria
This section reviews the concepts and methods to program and engineer bacteria for the production and discovery of bioactive compounds.The first subsection introduces briefly the chemical space of natural products and the basic principles of their biosynthesis.The second subsection explains the background of bacterial genetics, the paradigm shift as part of the advent of the 'genomic era' , and the technologies and methods to genetically reprogram biosynthetic pathways.The third subsection is focused on utilizing well-characterized bacteria as 'chassis' , that is, heterologous hosts, to facilitate yield optimization of valuable compounds of pharmaceutical, agrochemical or nutritional interest.The last subsection explains how biosynthetic pathways can be engineered in a rational way.

Chemical space and biosynthesis of natural products.
Biomolecules produced by bacteria include large macromolecules -such as nucleic acids, proteins, carbohydrates or lipids -and small molecules, which are the result of either primary or secondary metabolism, the latter of which are also called natural products.Most natural products fall into the mass range 200-3,000 Da (reF. 73 ),with a few exceptions for smaller compounds such as the earthy smelling geosmin 74 and a few natural products up to 5 kDa, such as polytheonamide 75 (Fig. 2).This mass range overlaps with the mass range (<900 Da) at which molecules can rapidly cross cell membranes to reach intracellular targets 76 .The secondary metabolism of bacteria covers an enormous diversity of chemical structures surpassing the chemical space of synthesized compounds 77 .As a result of millions of years of evolution, the interactions of these molecules with biological targets are highly optimized and the chemical scaffolds are, thus, considered privileged structures 78,79 .Natural products tend to feature sophisticated stereochemistry, polycyclic structures and variable distribution of halogen atoms and nitrogen-containing or oxygen-containing groups 80 .Each chemical scaffold reflects the interaction with several biological entities and, hence, they exhibit diverse biological properties, such as anticancer, antifungal, antibacterial and immunosuppressive activities 81,82 .These clinically useful biological properties are not necessarily related to the 'original' biological target, as shown for the group of cytotoxic antibiotics, such as the anthracyclines, which share similar chemical scaffolds with the tetracyclines 83 .These cytotoxic antibiotics underline the basic characteristic of natural-product evolution that these molecules are often optimized to interact with several biological targets and may feature various modes of action 84 .In addition, many bioactive natural products reveal sophisticated modes of action 85 , such as the antimycobacterial griselimycin (20), which inhibits DNA replication by binding the DNA polymerase sliding clamp DnaN 86 .As a result, numerous libraries are using natural products or their derivatives as leads in drug-screening programmes 78,87 .The Lipinski rule of five 88 is a widely used empirical concept to evaluate the drug-likeness of orally administered, biologically active compounds.It estimates solely pharmacokinetic properties, for example, absorption, distribution, metabolism and elimination, but not pharmacological effects or toxicity.Interestingly, despite the widespread use of natural products or their derivatives as drugs, many violate these rules, thus, emphasizing the special status afforded to secondary metabolites 89 .
The high diversity of natural products with regard to structure, biological function and biosynthesis means that no comprehensive classification system has been established.Distinct groups of natural products can, thus, be classified either by their chemical structures 90 or their biosynthetic origin (Fig. 2).
Many bacterial natural products originate from biosynthetic machineries called modular type I polyketide synthases (PKSs) and non-ribosomal peptide synthetases (NRPSs), forming polyketides, non-ribosomal peptides and their hybrids (Box 1).PKS systems resemble the fatty-acid synthases from primary metabolism and catalyse the formation of a basal polyketide chain by iterative decarboxylative Claisen thioester condensations of various acetate-derived building blocks.Unlike most peptides in living organisms, non-ribosomal peptides are not synthesized by the ribosome and the biosynthesis is independent from mRNA.The modular architecture of NRPSs is comparable to that of type I modular PKSs and generates diverse peptides, including cyclic structures and those incorporating non-proteinogenic amino acids.These large multifunctional and modular enzymes are composed of catalytic domains covalently fused in a linear manner with specific catalytic domains, each catalysing a single reaction step during the assembly of the polyketide or peptide backbone.These domains are bundled into modules, responsible for the incorporation of one particular amino acid or monomeric carboxylic-acid building block.NRPSs can form structurally diverse intermediates through the presence of additional, optional modifying domains that catalyse reactions, including epimerization, hydroxylation, methylation, isomerization and halogenation.One example is given by the PKS-NRPS hybrid natural product epothilone A (11), produced by the myxobacterial strain Sorangium cellulosum So ce90 (reF. 91 ) (Box 1).While iteration events can happen in linear assembly lines, there are different purely iterative biosynthetic NRPSs, type I, type II and type III PKSs, of which type II PKSs are exclusive to bacteria [92][93][94] .These type II PKSs consist of two ketosynthases (KSs), termed KSα and KSβ, and an acyl carrier protein, ACP, to assemble in an iterative manner a poly-β-keto thioester, which undergoes specific ketoreduction, cyclization and other modifications catalysed by discrete tailoring enzymes.Typical polyketides of type II PKSs are aromatic natural products such as the cytotoxic anthracyclines and antibacterial tetracyclines.
Terpenes are the largest class of natural products, with more than 50,000 compounds known from all kingdoms of life.They are constructed from isoprene units and are synthesized from isopentenyl diphosphate and dimethylallyl diphosphate.These activated monomers of isoprene are generated by the mevalonate and deoxyxylulose phosphate pathways, both of which are present in primary metabolism.The fusion of both monomers by oligoprenyl diphosphate synthases to geranyl, farnesyl and geranylgeranyl diphosphate provides the linear precursors for the biosynthesis of monoterpenes, sesquiterpenes and diterpenes.The diversifying step in the biosynthesis of terpenes is exhibited by the conversion of the linear precursor into (poly)cyclic terpene hydrocarbons or alcohols by terpene cyclase 95 .
Another group of peptide natural products that is produced by a complex molecular machinery is the ribosomally synthesized and post-translationally modified peptides (RiPPs).Unlike non-ribosomal peptides requiring large multimodular enzyme complexes to incorporate non-proteinogenic amino acids into a peptide backbone, RiPPs are capable of accessing a similar degree of chemical diversity through extensive posttranslational modifications of a ribosomally synthesized precursor peptide 96 .For example, the marine cytotoxic polytheonamides were supposed to originate from a non-ribosomal peptide biosynthesis 97 ; however, a metagenome-guided study revealed the polytheonamides as post-translationally modified ribosomal peptides 75 .The general biosynthesis of a RiPP starts with the ribosomal generation of a precursor peptide.This precursor consists of an N-terminal leader peptide (in some cases, a C-terminal follower peptide is also employed), followed by the core peptide and an accessory recognition sequence 98 .Post-translational modification enzymes recognize the leader (or follower) and modify the core peptide.Finally, the modified precursor peptide undergoes proteolysis and export, a process presumably guided by the leader peptide 99 .
A recent study presents an alternative and unprecedented biosynthesis of amino-acid-derived natural products such as the glutamate isostere 3-thiaglutamate (23)  and ammosamide A (24) 100 , a cytotoxic pyrroloquinoline alkaloid.Similar to RiPP pathways, the biosynthesis starts with the ribosomal generation of a small peptide; however, this small peptide serves exclusively as a scaffold for further non-ribosomal peptide extension and chemical modification.One specific amino acid is transferred to the C terminus of the peptide through adenosine triphosphate and aminoacyl-tRNA-dependent chemistry.Afterwards, oxidative rearrangement, carboxymethylation and proteolysis of the terminal amino acid yield the final natural product.Proteolytic release of the natural product regenerates the scaffold peptide that enters another round of biosynthesis.The process is, thus, catalytic in the ribosomally synthesized peptide and the overall process is, thus, more efficient than the stoichiometric use of leader peptides in typical RiPP pathways 101 .This finding is an exciting example of the biosynthetic range displayed by bacteria in nature, with a plethora of further similar biosynthesis systems likely remaining to be discovered.
Other groups of natural products can be described as modified primary metabolites, such as the nucleosides derived from nucleobases, glycosides derived from carbohydrates and other diverse groups of natural products previously classified as alkaloids due to their incorporation of nitrogen.Nevertheless, biomolecules that originate from primary metabolism, such as vitamin B 12 (7), also illustrate the importance and complexity of bacterial biomolecules, as the biosynthesis of this essential vitamin is confined to only a few microbial species 102 .

Genetic basis for reprogramming bacteria.
The underlying information on how to form the biosynthetic machineries that produce natural products is embedded in the genome of each bacterial cell.Genes encoding biosynthetic proteins are typically clustered within the bacterial genome, which facilitates the search for the connection between the gene cluster and the produced secondary metabolite (Box 2).With the advent of amenable whole-genome-sequencing technologies, it quickly became apparent, on the basis of genome-encoded pathways for secondary metabolite biosynthesis, that bacteria were underexploited as sources of novel chemistry and Isostere isosteres are molecules or ions with similar chemical scaffold.This means the same number and arrangement of atoms and comparable electronic properties.

Box 1 | Biosynthetic logic of PKS and NRPS natural products
polyketides, non-ribosomal peptides and their hybrids, such as epothilone, are classes of natural products that are synthesized by large multifunctional and modular type i polyketide synthases (pKSs) and non-ribosomal peptide synthetases (NrpSs).each module catalyses a single reaction step during the assembly of the polyketide or peptide backbone.A cycle of chain elongation starts with either the substrate recognition of an acyl-CoA thioester (pKS), or an amino acid following its activation as an aminoacyl adenylate (NrpS).The next step involves covalent binding of the substrate thioester to a carrier protein and then condensation with an acyl or peptidyl residue from the upstream module.The minimal NrpS module consists of an adenylation domain (A), a peptidyl carrier protein (pCp) and a condensation domain (C).Similarly, the minimal pKS elongation module consists of an acyl transferase (AT) domain, an acyl carrier protein (ACp) and a ketosynthase (KS) domain.A variety of optional domains may be present that modify the polyketide intermediate, in particular, reductive domains in pKS modules like ketoreductases (Kr), dehydratases (DH) and enoyl reductases (er) domains.The last module of the assembly line usually contains a thioesterase (Te) domain to catalyse the release of the matured biosynthetic backbone in linear, cyclic or branched cyclic form.compounds.Genomic analysis also reveals that many secondary metabolite genes are not expressed under conventional laboratory conditions.This translates into a silent potential of the microbial world for the production of a number of as-yet-unknown natural products 103 .Hence, the paradigm in natural-product research is shifting from traditional bioactivity-guided fractionation of bacterial crude extracts to genomic investigation and engineering of these bacteria to yield engineered living biofactories, used as production machinery for complex biomolecules.
Our expanded knowledge of bacterial genetics has paved the way for advanced genetic-engineering and metabolic-engineering approaches to produce natural products, which may only be available from the natural source under specific conditions.Several examples have been reported that describe rational pathway-manipulation strategies at the genome level to activate native producers of natural products, which are otherwise downregulated under laboratory-fermentation conditions.The low production rate under certain growth conditions might be explained by negative regulation; therefore, the associated genes are termed silent or cryptic biosynthetic gene clusters (BGCs) 104,105 .Downregulated BGCs can be activated to boost the production of secondary metabolites by empirical approaches, such as variation of growth conditions, co-cultivation 106 , addition of chemical elicitors 107,108 or traces of metal ions 109 , provision of exogenous small molecules to the producer strain 104 and ribosome engineering to alter the translation machinery [110][111][112] .These traditional approaches were primarily carried out in

Box 2 | The collaboration between bacterial genomics and metabolomics
Bacterial natural products are secondary metabolites formed by biosynthetic enzymes encoded within the bacterial genome.The identification of bioactive metabolites can be achieved by classical analytical-chemistry approaches focusing on the metabolites (including metabolomics-inspired approaches) without prior knowledge of the genes and enzymes involved in their biosynthesis.Structure elucidation of natural products is, in general, accomplished by nuclear magnetic resonance (Nmr) spectroscopy or, less commonly, by X-ray crystallo graphy.recent developments have shown the feasibility of cryo-electron microscopy and microcrystal electron diffraction not only for the structure elucidation of proteins 226,227 but also for natural products 101,228 .The advent of amenable whole-genome-sequencing technologies led to a new aspect of natural-product research to become accessible by enabling the investigation of genes encoding biosynthetic enzymes (genomics).The availability of bacterial genome sequences revealed a striking discrepancy between the capacity of genome-encoded pathways for secondary metabolite biosynthesis and the natural products observed in biological samples.The connection between a secondary metabolite in the crude extract and the genes encoding the underlying biosynthetic enzymes can be established by genetic manipulation, such as gene disruption or deletion, induced gene expression or by heterologous expression of all the required biosynthetic genes.Bioinformatic prediction of undiscovered natural-product structures based on biosynthetic genes is a powerful tool because targeted activation or knockout of biosynthetic gene clusters requires bioinformatics analysis to prioritize the selection of the biosynthetic genes.antiSmASH 5.0 (reF. 229 )is currently the prediction tool that offers the broadest range of in silico analysis, whereas the recently developed bioinformatics and chemoinformatics web application TransATor aims to enable de novo structural predictions for trans-AT polyketides, a special case of non-canonical type i polyketide synthase systems 230 . in vitro reconstitution of the biosynthetic pathway with recombinantly produced proteins involved in the biosynthesis can cross link the field of genomics and metabolomics at the protein level.
The widespread availability and improving quality and increasing quantity of metabolomics data has significantly enhanced the dereplication of natural products -in particular using the combination of high-performance liquid chromatography with high-resolution mass spectrometry (mS) -in addition to mS/mS fragmentation experiments in the isolation and structural elucidation of natural products.When combined with comprehensive analytical data sets, it can provide valuable information on the building blocks incorporated in the natural product, especially if used together with feeding studies employing stable-isotope-labelled precursors.emerging computational frameworks like Global Natural products Social molecular Networking set the stage to take the in-depth analysis of microbial secondary metabolomes to the next level by enabling 'global spectral networking' as a tool to relate natural products across multiple producers 231 .

R e v i e w s
former times during the so-called pregenomic era, since the rationale of these traditional approaches is independent of the genome sequence.
Two examples of such untargeted approaches are the discoveries of amexanthomycin and piperidamycin A (31).Amexanthomycin was found following the deletion of the rifA gene encoding the rifampicin PKS in Amycolatopsis mediterranei S699 (reF. 113 ).Amexanthomycin A (30), B and C showed inhibitory activity against human DNA topoisomerase (Topo IIa), albeit at 500 µM (reF. 113 ).The isolation of the antibacterial piperidamycin A (31), meanwhile, was facilitated by mutagenesis of the genes rpsL and rpoB, resulting in the activation of the silent BGC of piperidamycins from Streptomyces mauvecolor 631689 (reF. 110 ).Targeted approaches are based on known genome sequences with a clear focus on biosynthetic 114 , regulatory 115 and resistance 116 genes (Fig. 3).Transcription regulators that are encoded nearby the biosynthetic gene clusters often influence expression of the secondary metabolite 114,115,117 , either as an activator or as a repressor.The exchange of a native regulator for a constitutive or inducible promoter can lead to overexpression (or inhibition) of the controlled genes, and (ideally) a higher concentration of the associated natural product can be achieved.For example, promoter insertion in front of a positive regulator in the genome of S. ambofaciens ATCC 23877 led to isolation of the macrolide stambomycin A (25) 118 ,   119 .Sequestration of repressors by multicopy plasmids derepressed a downregulated BGC and lead to the identification of oxazolepoxidomycin A (28) 120 .Replacement of the native promoter through a strong ermE* promoter in front of the efflux pump gene botT showed a 20-fold increased production of bottromycin A 2 (29) concentration compared with the natively expressed resistance gene 121 .Novel bioactive natural products can be found and correlated with uncharacterized biosynthetic pathways through the analysis of associated genetic markers for host self-resistance.In silico analysis searching for genes that encode functional biosynthetic machinery co-localized with host self-resistance (here, genes encoding pentapeptide repeat proteins) led to the discovery of a silent type II PKS BGC in Pyxidicoccus fallax An d48.Induced gene expression resulted from the insertion of two different promoters in front of two biosynthetic genetic operons of this downregulated type II PKS cluster in the native host and led to the structural elucidation of pyxidicycline A (26) and B, which, indeed, turned out as topoisomerase inhibitors and provided insights into their biosynthesis 116 .
Heterologous expression of this downregulated gene cluster was facilitated through the exchange of promoters, the reorganization of the operon structure and controlled expression of the self-resistance mechanism.
Heterologous production of natural products.Natural products, in general, are known to feature complex chemical scaffolds with many stereocentres.As a result, chemical synthesis of natural products can be inefficient and costly.In many cases, the production of these biomolecules for use as drugs relies heavily on fermentation or on semi-synthetic approaches.The supply of the antimalarial drug artemisinin and the anticancer drug etoposide are examples of the semi-synthetic industrial production of clinically important drugs 122,123 .The commercial semi-synthetic production of etoposide starts from the natural product podophyllotoxin, which leads to the excessive harvesting of Podophyllum hexandrum and P. peltatum, both of which harbour podophyllotoxin in their roots 122 .As a result, P. hexandrum was included in the Convention on International Trade in Endangered Species.The global supply of artemisinin and derivatives thereof was improved significantly 124 through a largescale production of the key intermediate artemisinic acid in Saccharomyces cerevisiae and the efficient conversion of artemisinic acid to artemisinin via a photochemical process, creating a singlet oxygen 125 .
Heterologous expression can offer several advantages in comparison to the native host.In particular, a heterologous expression system may be useful when the native producer is a slow-growing microorganism not suitable for large-scale bioreactor fermentation and/or is genetically intractable, as these characteristics would impede efforts to increase the yield of the compound from the native producer.Ideally, a heterologous host should have characteristics such as amenability to genetic manipulations, a high success rate of heterologous expression of foreign BGCs, advantageous growth characteristics and provide decent yields of the secondary metabolite.
Essential steps in the workflow of heterologous expression of foreign biosynthetic genes include the mobilization of DNA encoding the biosynthetic pathway into a suitable vector, the transfer into the heterologous host and, finally, the successful expression leading to heterologous production of the natural product of interest (Fig. 4).The use of a phylogenetically closely related heterologous host often results in more efficient expression 126 .Similar codon usage and adaptation provides higher translational efficiency over genetically distant species and the higher likelihood for functionality of native transcriptional elements of the gene cluster, such as regulatory elements, promoters 127

Heterologous expression
The production of non-native biomolecules through the transfer of biosynthetic genes into a foreign host.The genetic information required for heterologous expression can be found in the DNA originating from environmental bacteria, plants and the human microbiome.Biosynthetic gene clusters (BGCs) are DNA sequences encoding biosynthetic machineries, which are often co-localized.These BGCs can be subcloned from the original source or acquired via gene synthesis to generate a heterologous expression construct.This construct has to be transferred into a bacterial host and maintained by integration into its genome or as a replicative plasmid, depending on the construct and heterologous host.Fermentation of the heterologous host and chemical extraction of the produced secondary metabolites can lead to the identification, isolation and structure elucidation of the heterologously produced compound.HPLC, high-performance liquid chromatography.
In the world of prokaryotes, some well-established chassis have been developed.E. coli has served for decades as a superb host for the production of various recombinant proteins 131 .A selection of success stories includes the clinically relevant insulin analogues 132 , the type XVIII collagen derivative endostatin 133 , the human granulocyte colony-stimulating factor filgrastim 134 and its PEGylated derivative pegfilgrastim 135 , and the fusion protein denileukin diftitox (interleukin-2 and diphtheria toxin) 136 .However, this species was found to be non-ideal for the production of many bacterial secondary metabolites.E. coli is not known as a gifted native producer of complex natural products and the expression of foreign biosynthetic genes in it as hosts is challenging.The major problems are the expression of large multimodular PKS and NRPS genes encoding proteins far larger than those normally found in E. coli, the supply of metabolic substrates required for complex natural-product formation, the higher guanine-cytosine content of the original biosynthetic genes and the different codon usage.Nevertheless, several BGCs from entomopathogenic bacteria were successfully expressed in E. coli 137 .The close phylogenetic relationship between entomopathogenic bacteria such as Xenorhabdus and Photorhabdus species to E. coli underpins the plausibility of E. coli as a heterologous host for BGCs originating from these bacteria.
The overexpression of biosynthetic pathways has proven to be a reliable production platform for intriguing natural products 126 .Members of the order Actinomycetales, in particular, the genus Streptomyces, the genus Bacillus and the order Myxococcales, are known to be prolific producers of bacterial natural products and have been widely investigated.Diverse engineered streptomycetes, such as S. coelicolor, S. avermitilis, S. lividans or S. albus, have been optimized for the expression of actinomycete-derived BGCs 138 , whereas BGCs from underexploited bacterial strains are preferably transferred in phylogenetically related hosts 126 .For myxobacterial BGCs, Myxococcus xanthus DK 1622 is the preferred heterologous host 116,139,140 , whereas Anabaena sp.strain PCC 7120 (reFs 141,142 ), Synechocystis sp.PCC 6803 (reF.143 ) and Synechococcus elongatus strain PCC 7942 (reFs 144,145 ) are the preferred chassis for cyanobacterial BGCs 146 .Other characterized heterologous hosts are Pseudomonas putida 140,147 , B. subtilis 148,149 , Lactococcus lactis NZ9000 (reFs 150,151 ) and Burkholderiales strain DSM 7029 (reFs 140,152 ), besides some less frequently used heterologous hosts.The importance of establishing different heterologous hosts and not relying on one 'ideal' broad host is shown by a study engineering Salinispora tropica CNB-440 for the heterologous expression of biosynthetic enzymes.This model organism of the marine actinomycete genus Salinispora featured a threefold higher production of thiolactomycins in comparison to the extensively engineered so-called superhost S. coelicolor M1152 (reF. 153 ).
The benefits of utilizing a heterologous host are highlighted by the production of the argyrins, a family of cyclic octapeptides exhibiting potent antimicrobial, antitumorigenic and immunosuppressant activities.The argyrins are biosynthesized by the myxobacterial strains Archangium and Cystobacter sp. 154,155, and the actinomycetes are also known as producers of argyrins, which were named antibiotics A21459 (reFs 156,157 ).The NRPS pathway employed in the production of these compounds results in at least ten derivatives (argyrins A-J), due to miscellaneous megasynthetase substrate specificities and inefficient tailoring chemistry.The native producers of argyrins are difficult to genetically manipulate and, thus, synthetic versions of the argyrin BGC were designed and constructed from synthetic DNA in order to produce the argyrin panel in a genetically modified derivative of M. xanthus DK 1622 (reF. 139 ).By circumventing the bottlenecks associated with the native producer -including low yield and a lack of methods for BGC engineering -this biotechnological production platform provides an efficient supply of this valuable biomolecule for medicinal applications.Reprogramming of the biosynthetic machinery enables the direct production of selected argyrins, as well as novel derivatives thereof.
In a recent study, several long-chain polyunsaturated fatty acid (LC-PUFA) BGCs based on synthetic DNA encoding polyketide synthase-like LC-PUFA synthases from myxobacteria were heterologously expressed in the oleaginous yeast Yarrowia lipolytica 158 .The myxobacterial iterative PKS-like PUFA synthase yields LC-PUFA with lower consumption of nicotinamide adenine dinucleotide phosphate and also enables direct incorporation of acyl-CoA precursors.In contrast, PUFA biosynthesis in eukaryotes occurs by less efficient aerobic pathways, in which several oxygen-dependent desaturases and elongases are involved for the conversion of saturated fatty acids into PUFAs.One of the constructed yeast strains produced the highest concentration of docosahexaenoic acid (350 mg l −1 ) among all published PUFA-producing Y. lipolytica strains, emphasizing the combination of a promising host strain and the myxo bacterial de novo synthesis of docosahexaenoic acid 158 .This platform is promising for the production of LC-PUFAs in high quantities and quality.The natural source of these compounds, fish oil, is in decline 159 .
The ability to activate foreign, downregulated biosynthetic gene clusters using heterologous expression has been demonstrated recently by Luzhetskyy's group.They constructed S. albus chassis strains, optimized for the discovery of natural products through the heterologous expression of foreign BGCs by chromosomal deletion of 15 gene clusters that usually encode the biosynthesis of secondary metabolites.Genomic integration of a foreign cryptic BGC originating from the distantly related Frankia alni strain ACN14a enabled the heterologous production of fralnimycin 160 .
Recent research on secondary metabolites does not focus exclusively on environmental microbes but also includes strains from the mammalian gut.In addition to investigations into their contributions to gut immunity and the provision of essential nutrients by enzymatic catalysis (see Beyond living factories), their ability to produce specific secondary metabolites that might account for important host-microorganism and microorganism-microorganism interactions has been studied 161 .To date, most of the natural products identified from the mammalian gut microbiota are from RiPPs pathways, such as the lantibiotics, for example, ruminococcin A 162 , nisin O 163 and nisin H 26 , as well as different bacteriocins, microcins and thiazole-modified/ oxazole-modified microcins 161 .A rare example of an NRPS natural product deriving from the mammalian gut microbiota is the pyrrolo [2,4]benzodiazepine derivative tilivalline (21) produced by Klebsiella oxytoca 164 , which has been identified as the pathogenicity factor in antibiotic-associated colitis 165 .The tilivalline gene cluster could be expressed heterologously and its biosynthesis was completely reconstituted in vitro with the recombinantly produced biosynthetic proteins 166 .
Bacterial heterologous production platforms can also be applied in the expression of biosynthetic genes found from uncultured environmental bacteria.Numerous sequencing projects have revealed that the majority of bacteria have yet to be cultured, indicating that the mobilization of environmental DNA (eDNA) into a cultured bacterium and subsequent heterologous expression in a surrogate host can lead to the production of novel secondary metabolites.One example of a successful sequence-guided metagenomic pipeline to probe for new congeners of the natural-product family is exhibited by the discovery of the calcium-dependent antibiotics malacidins A and B 167 .An arrayed collection of eDNA isolated from 2,000 unique soil samples was probed by using degenerate PCR primers, which targeted the NRPS A domains.The malacidins are structurally distinct from known calcium-dependent antibiotics, such as daptomycin (19) and friulimicin, which, nevertheless, facilitated a biosynthesis proposal for malacidin, since the underlying NRPS biosynthesis of all three natural products is similar.Another example of metagenomics-based exploration is given by the discovery of the potent anti-HIV lanthipeptide, divamide A, from uncultured symbiotic bacteria living in tunicates 168 .The scarcity of isolated material of this endophytic bacterium was circumvented by applying NMR structure elucidation combined with metagenomics and synthetic biology to fully characterize the newly discovered natural products and their underlying biosynthesis by heterologous expression in E. coli.However, identifying novel and complex scaffolds from eDNA remains challenging, as shown by a recent functional metagenomic screening study in which three new derivatives of the well-known antibiotic chloramphenicol were identified 169 .Besides the tedious generation of metagenomic DNA libraries harbouring eventually novel BGCs, the heterogeneous origin of metagenomic libraries impedes the taxonomical correlation of a BGC to the corresponding microorganism and, with it, the rational choice of an appropriate heterologous host.In addition, it cannot be guaranteed that the heterologously expressed gene cluster originating from eDNA will yield the same natural products, as it would have done in the native host.Nevertheless, these approaches enabled access to natural products from bacteria and eDNA samples, which would not have been possible by classical biotechnological approaches.
Obtaining plant secondary metabolites by extraction from natural host plants 170 is challenging, as a result of the low abundance of the secondary metabolites of interest and the complex metabolite background from plants 171 .In addition, unlike bacterial natural products, plant secondary metabolite genes are typically not found clustered in their chromosomes, making identification of the full set of biosynthetic genes difficult 172 .Thus, due to the advent of recombinant DNA techniques and further insights from microbial biosynthetic pathways, heterologous production platforms became an attractive alternative to the traditional production of phytochemicals by plant cell culture [173][174][175] .Numerous phytochemicals have been produced, mainly using genetically tractable plants such as Nicotiana benthamiana, and microbial hosts, such as S. cerevisiae and E. coli 172,176 .Reconstitution of the cannabinoid biosynthesis in S. cerevisiae underlines the applicability of yeast to produce complex phytochemicals 177 .Interestingly, it was possible to establish a stepwise in vivo fermentation process of four engineered E. coli strains to reconstitute hydrocodone and thebaine 178 , while the yield of the latter is increased 300-fold from the developed system in yeast 179 .
The anticancer plant natural product taxol (paclitaxel) exemplifies the obstacles to the clinical use of various plant phytochemicals.Originally isolated from the bark of the Pacific yew tree 180 , two to four fully grown trees were required to provide sufficient dosage for one patient 181 .In contrast, several total synthesis protocols have been reported, but overall yields do not exceed 0.4% [182][183][184][185] .A semi-synthetic route using the biosynthetic intermediate baccatin III (which can be isolated efficiently from the needles of the European yew or, alternatively, produced by plant cell culture 186 ) improved the supply of clinically required taxol; however, the generation of derivatives remains challenging 187 .The formation of the crucial paclitaxel precursor taxadiene in E. coli initially generated yields of 1 g l −1 (reF. 188 ).Further yield improvement could be achieved by optimizing the cytochrome P450 expression, reductase partner interactions and N-terminal modifications to increase the titre of oxygenated taxanes to ~570 ± 45 mg l −1 (with taxadiene-5α-ol as the major metabolite) in E. coli 189 .
Another plant biosynthetic pathway that has been thoroughly investigated for low-cost production in a microbial host is the terpenoid artemisinin, a major drug in the treatment of malaria.The production of 8-hydroxycadinen, the key intermediate in the production of gossypol and artemisinic acid (en route to artemisinin), could be achieved in E. coli 190 , and a novel semi-biosynthetic route to artemisinin itself using an engineered substrate-promiscuous cytochrome P450 was reported 191 .
These two bacterial-production platforms indicate the potential of fermentation strategies to yield valuable phytochemicals 192 .While the heterologous production of the paclitaxel precursor taxadiene in E. coli was mainly achieved due to the shared mevalonate pathway in plant and bacteria, other approaches to achieve successful biosynthesis of phytochemicals aim at co-culture fermentation 176 .The lack of compartmentalization in bacterial cells can be compensated by breaking the complete pathway into separate modules and expressing each module in a dedicated microbial strain to address challenges such as membrane-bound proteins, misfolded proteins or other non-physiological stress in a single host.Co-cultivation of multiple E. coli strains has been used to achieve the heterologous production of phytochemicals, including resveratrol 193 and other flavonoids 194 .In order to express the cytochrome P450 genes, co-culture of E. coli and S. cerevisiae are more commonly used, such as for the production of benzylisoquinoline alkaloids magnoflorine, corytuberine and scoulerine 195 .
In conclusion, a robust, heterologous expression platform often paves the way for targeted engineered approaches to the biosynthetic production of novel derivatives.
Engineering biosynthetic pathways.Two different approaches have been generally pursued to engineer biosynthetic machineries, with the aim of affording new or altered biomolecules 196 : directed evolution 197 and targeted/rational genetic engineering.Targeted/rational genetic engineering of biosynthetic machineries is also termed combinatorial biosynthesis and includes methods based on: (1) gene deletion, insertion or alteration to afford mutasynthesis and site-specific mutagenesis, (2) the reorganization of biosynthetic pathways, such as domain, module or subunit swapping and intersubunit docking-domain exchange and (3) engineering de novo biosynthetic pathways.
In contrast to the aforementioned combinatorial biosynthesis methods, precursor-directed biosynthesis simply involves the combination of chemical and biological approaches to generate novel derivatives of complex natural products, without genetic manipulation of the producing organism.Simple precursor analogues are fed to the producing organism, with the intention that they are incorporated directly into a biosynthetic pathway, resulting in structurally new products.However, precursor-directed biosynthesis has several drawbacks, such as the usually low incorporation rates of unnatural precursors.As a result, these procedures have been further developed by adding precursors to the culture of a mutagenized bacterium, in which natural-product biosynthesis has been blocked at a particular stage.This combinatorial biosynthesis approach is called mutasynthesis and was invented by Rinehart and Gottlieb 198 .They described feeding a mutant of S. fradiae 3535 (incapable of synthesizing deoxystreptamine and, consequently, deficient of the antibiotic neomycin) with the related aminocyclitols streptamine and 2-epistreptamine.The process resulted in the isolation of four new antibiotics, termed hybrimycins A and B (for each of the two different isomers).Mutasynthesis approaches are used mainly to incorporate simple biosynthetic intermediates, such as aromatic and heteroaromatic amino acids, to a 3-amino-5-hydroxybenzoic-acid-blocked mutant of S. hygroscopicus, yielding new analogues of geldanamycin 199 (Fig. 5a) or ansamitocin in Actinosynnema pretiosum 200 .Current case studies aim to incorporate more advanced biosynthetic intermediates or even complex parts of the final molecule, for example, during the mutasynthesis of the α-pyrone antibiotic myxopyronin in M. fulvus 201 .However, incorporation yields often suffer from stability and uptake problems.46) is more challenging than single-gene alteration, since the interplay of the modular enzymes is strictly regulated.Reorganization of these biosynthetic pathways enabled the production of 'JBIR-06' with a changed alkyl chain (47), a derivative of neoantimycin with a contracted core cyclic structure to a tri-lactone (48) and an expanded JBIR-06 derivative with a tetra-lactone core cyclic structure (49).
An example of rational site-specific mutation is given by the heterologous production of the antifungal vioprolides.The biosynthesis genes were engineered such that the adenylation (A) domain bound to serine rather than alanine (Fig. 5b).Site-directed mutagenesis of the N-methyltransferase domain provided further diversified vioprolide structures 140 .Previous NRPS studies focused mainly on changes to domains, modules and binding-pocket mutagenesis.In contrast, a simplified procedure for combinatorial biosynthesis of non-ribosomal peptides is based on subdomain swaps, which facilitated the alteration of the A domain specificity, for example, from phenylalanine to valine 202 (Fig. 5b).Since subdomains are shorter than domains or modules, subdomain swapping might fuel NRPS-engineering approaches based on bioinformatics-based comparisons.
The applicability of classical combinatorial biosynthesis 203 by the reorganization of modular machineries has been shown in studies conducted on the aromatic polyketide actinorhodin from a type II PKS 204 , the macrolide antibiotic erythromycin (6) from a type I PKS 205 and the lipopeptide antibiotic daptomycin (19)  from an NRPS 206,207 .The NRPS in the daptomycin biosynthetic pathway was genetically engineered for the biosynthesis of a library of novel lipopeptide antibiotics.Single or multiple modules and subunit exchanges, inactivation of the tailoring enzyme glutamic acid 3-methyltransferase and natural variations of the lipid tail of the NRPS have been conducted to successfully modify the daptomycin cyclic peptide core with high product titres 206 (Fig. 6a).
A recent study aimed at reprogramming the antimycin-type NRPS-PKS hybrid assembly lines by providing a heterologous expression platform for the tri-lactone JBIR-06 (45) and the tetra-lactone neoantimycin (46) (Fig. 6b).The goal was to shrink the core cyclic structure of neoantimycin A (46) to a tri-lactone (48), expand the core of JBIR-06 (45) to a tetra-lactone (49)  and diversify the alkyl chain of JBIR-06 (47) through the incorporation of various alkylmalonyl-CoA extender units 208 .Although module exchanges of evolutionarily homologous biosynthetic gene clusters have been successfully conducted in the past following sequence comparisons, the significance of this study lies not only in the modification of the cyclic structure or the alkyl-chain decoration.The interplay and rational engineering of the NRPS-PKS assembly machinery sheds light on how the bioengineering of productive biosynthetic assembly lines might be leveraged to produce unnatural polyketide-non-ribosomal-peptide hybrids.
Detailed in vitro experiments using NRPS model systems or investigations of individual domains followed by correlations of the conclusions to in vivo production of NRPSs 209 might lead to the establishment of a 'non-ribosomal peptide synthesizer' that can create artificial NRPSs 210 .In contrast to established heterologous expression platforms and the generation of so-called unnatural natural products, this approach aims to generate de novo engineered biosynthetic architectures for specific biomolecules.A recent strategy sets out to establish a platform for the de novo design of novel NRPS pathways by using exchange units (XUs) rather than modules as functional units to engineer NRPSs 211 (Fig. 7).These XUs are fused at specific positions to connect the condensation (C) and A domains of NRPSs in Xenorhabdus and Photorhabdus.The original specificity of the downstream module was taken into account to enable the production of the designed peptide.Internal C domains were employed as an alternative to other peptide-chain-releasing domains for the rational production of cyclic peptides.Using these principles, it was possible to generate modified and functional artificial NRPSs producing derivatives of known and novel peptides.Since the thioesterase (TE) domain also confers substrate specificity, the authors employed the internal C or C/E domains to release the peptide chains, as had been shown for fumiquinazoline F 212,213 .Based on the results of the XU concept, a new fusion point inside the upstream C domain was used to develop the exchange unit condensation domain (XUC) concept 214 .In contrast to XUs, the XUC concept enables the efficient assembly The recent characterization of four giant aminopolyol PKS gene clusters (the largest biosynthetic gene cluster that has been expressed to date) using heterologous expression successfully afforded a novel aminopolyol.Intriguingly, these enzymes contain modules that use either malonyl-CoA or methylmalonyl-CoA extender units and feature all known types of β-processing products with different stereochemical outputs.The phylogenetic analysis of highly homologous but functionally diverse domains from the giant PKSs demonstrate the evolutionary mechanisms for structural diversification of polyketides and indicate the set of (AT-ACP-KS) domains to hold promise for the successful engineering of multimodular PKSs 215 .These gene clusters might provide genetic building blocks for de novo production of unnatural polyketides in the future.
Directed evolution applications aim to simulate processes such as gene (here, for example, multimodular biosynthetic module) duplications, speciation including mutations, homologous recombination and gene-transfer events to access nature's biosynthetic potential 216 .The inherent advantage of directed-evolution-based screening is that efficient engineering of biosynthetic enzymes can be achieved without an in-depth understanding of structure-function relationships 217 .The capability of directed evolution was shown by Arnold and colleagues, who demonstrated that the catalytic function of cytochrome c from Rhodothermus marinus could be enhanced by directed evolution approaches to yield greater than 15-fold increase in the turnover of carbonsilicon-bond formation than industrial synthetic catalysts 218 .
The general experimental strategy for directed evolution comprises iterative cycles of mutagenesis of the gene of interest, gene expression and screening/ selection for beneficial variants and subsequent gene amplification of the evolved genes 217 (Fig. 8a).A fundamental limitation of directed evolution approaches is the requirement of a powerful high-throughput-screening platform, which is capable of identifying the rare variant with beneficial mutations of the biosynthetic enzymes or their products.The detection can be conducted either by selection systems that couple the protein function to survival of the host or by screening systems that individually assay the encoded function of the enzyme activity, most often using a colorimetric or fluorimetric assay.Combined semi-rational approaches are being investigated to address this limitation, by relying on focused libraries, which require a less powerful high-throughput-screening platform 219 .
A novel approach to high-throughput screening was shown during the reprogramming of an archetypal NRPS module from the tyrocidine biosynthesis (TycA F ).The reprogramming involved modification of the l-Phe-specific A domain to accept and process the backbone-modified (S)-β-Phe with near-native specificity and efficiency 220 (Fig. 8b).The successful production of cyclic βα-dipeptide ( 56) and (S)-β-phenylalanine-containing pentapeptide (57)  proves that downstream domains tolerate the new building block.Successful reprogramming relied on fluorescence-activated cell sorting-based screening of a large number of mutational variants using the adenylation/thioesterification reaction of the TycA A-PCP di-domain, displayed on yeast cell lines.The established system was capable of screening a larger library than conventional plate assays 221 , and simultaneous screening of adenylation and thioesterification using a 'clickable' amino acid proved to be superior to indirect engineering approaches based on ligand binding 222 or monitoring adenylation by pyrophosphate exchange 223 .
A different approach to directed evolution was initiated by a rational engineering strategy on the rapamycin PKS.The aim was to replace the inactive domains from module 3 with active domains from modules 11 and 13 to yield strains capable of producing 32-desketorapamycin 224 .Instead of the expected mixture of domain-replacement mutants and wild-type revertants, mutants were isolated producing different rapamycin analogues, called rapalogues, featuring truncated and expanded macrolactone rings, alternative cyclization patterns and linear polyketide scaffolds 225 .Reduced or increased numbers of extension modules logically explain the structure of these altered polyketides (Fig. 8c).A recombination-based process had led to a range of deletion or duplication events within the BGC encoding the rapamycin PKS, mimicking a plausible mechanism of natural evolution for modular polyketide synthases.The genetic basis, generality and mechanism of this so-called accelerated-evolution process was further investigated by whole-genome sequencing of 17 generated mutants and by replicating this process in the tylosin PKS of S. fradiae NRRL 2702.Interestingly, the utilized pSG5-based replicon of pKC1139 is critical for the accelerated-evolution process, since no new tylosin derivatives could be produced when a comparable pKC1132-based control plasmid was used.
This result pinpoints the disadvantage of the accelerated-evolution-based/directed-evolution-based approach, lacking the knowledge of the precise molecular mechanism.In addition, the inherent bottleneck of constructing truly efficient high-throughput-screening systems further impedes the full potential of directed-evolution approaches.
Fig. 8 | Directed evolution of biosynthetic pathways.a | Directed evolution comprises iterative cycles of mutagenesis of the gene of interest to create diversified genes, followed by gene expression and screening/selection for beneficial variants, and subsequent amplification of the evolved gene(s).b | A fluorescence-activated cell sorting-based high-throughput-screening platform led to the reprogramming of an A domain in the tyrocidine biosynthesis (an archetypal non-ribosomal peptide synthetase module) to accept and process (S)β-Phe instead of the natural precursor l-Phe.The mutated domain processes this unnatural precursor with near-native specificity and efficiency , yielding cyclic βα-dipeptide ( 56) and (S)β-phenylalanine-containing pentapeptide (57).c | Different analogues of 39-desmethoxyrapamycin (58) featuring expanded (59) and truncated (60) macrolactone rings, alternative cyclization patterns and linear polyketide scaffolds were generated by recombination-based processes within the biosynthetic gene cluster encoding the rapamycin polyketide synthase.This so-called accelerated-evolution process relies on the use of the pSG5-based replicon of pKC1139.

Conclusion
The production of biomolecules by bacteria and the application of bacteria as biocatalysts have seen tremendous progress, in large measure due to genomics-driven and pathway-modification approaches enabling unprecedented possibilities for metabolic engineering.The hitherto acquired knowledge on biosynthetic pathways paves the way for the programmed production of biomolecules inspired by biosynthetic chemical-assembly machineries.Since engineered and reprogrammed bacteria facilitate the production of biomolecules, which are otherwise difficult to access, these microorganisms are invaluable to deal with various future challenges, such as the increasing need of chemicals for energy supply, nutritional demand and pharmaceutical requirements, while limiting our reliance on the use of toxic catalysts or organic solvents.As a result, fermentations of microorganisms might, for example, provide a sustainable energy resource at comparatively low working temperatures without producing toxic waste.In addition, bacteria may even become helpful in addressing the ecological problems caused by environmental contamination.Numerous developments are currently using genetically modified bacteria as biological workhorses to expand the capacity of synthetic processes by creating engineered living biofactories.
This Review aims to provide an inspirational perspective of current and future applications of microorganisms as living biofactories by highlighting the indispensable value of bacteria to humans, rather than considering microorganisms mainly as pathogens.The famous quote of Louis Pasteur "…les microbes auront le dernier mot" (…the microbes will have the last word) is, thus, illuminated from an alternative perspective.

Published online 23 March 2020
might pave the way for a variety of modern functional devices in microscale machines based on bacteria.In addition, the functional transfer of the bacterial magnetosome gene cluster originating from Magnetospirillum gryphiswaldense MSR-1 into a foreign non-MTB, the photosynthetic bacterium initiates the transcription of a particular gene by binding rNA polymerase and transcription factors.Constitutive promoters are always active, while inducible promoters are regulated through molecules, temperature and light.

R e v i e w s volume 4 |
April 2020 | 179

Fig. 3 |
Fig. 3 | Reprogramming bacteria and activating biosynthetic dark matter.Top: A generic representation of a biosynthetic gene cluster (BGC) is shown, consisting of core biosynthetic genes (green), positive regulators (red), repressors (grey) and self-resistance-conferring genes (yellow).Bottom: Examples of natural products identified by activating biosynthetic dark matter.Promoter insertion in front of a positive regulator in the genome led to the formation of the macrolide stambomycin A (25) 118 (part A), whereas insertion in front of two biosynthetic genetic operons (part B) led to the discovery of pyxidicycline A (26) 116 .Inactivation of the repressor gene gbnR in Streptomyces venezuelae ATCC 10712 induced the production of gaburedin A (27) 119 (part C).Sequestration of repressors (black vehicles) by

Fig. 4 |
Fig.4| Heterologous expression workflow.The genetic information required for heterologous expression can be found in the DNA originating from environmental bacteria, plants and the human microbiome.Biosynthetic gene clusters (BGCs) are DNA sequences encoding biosynthetic machineries, which are often co-localized.These BGCs can be subcloned from the original source or acquired via gene synthesis to generate a heterologous expression construct.This construct has to be transferred into a bacterial host and maintained by integration into its genome or as a replicative plasmid, depending on the construct and heterologous host.Fermentation of the heterologous host and chemical extraction of the produced secondary metabolites can lead to the identification, isolation and structure elucidation of the heterologously produced compound.HPLC, high-performance liquid chromatography.

NATure revieWS | CHEMiSTRy R e v i e w s volume 4 |
April 2020 | 183

Fig. 5 |
Fig.5| Mutasynthesis and mutagenesis.a | Mutasynthesis involves -following gene deletion, insertion or alterationthe feeding of unnatural precursors to the culture of a mutagenized bacterium, in which the biosynthesis has been blocked at a particular stage.The unnatural precursors(32-35) are highlighted, as are the resulting modifications to the geldanamycin (36) products(37-41).b | Site-specific mutagenesis, meanwhile, aims to change the substrate specificity of biosynthetic enzymes by engineering of the binding pocket or by subdomain swapping.An example for the generation of an altered natural product upon binding-pocket engineering is vioprolide D ser(42), while diketopiperazine (43) was generated through subdomain swapping.

Fig. 6 |
Fig. 6 | Biosynthetic pathway reorganization.Reorganization of biosynthetic pathways such as domain, module or subunit swapping leads to the production of novel structures.a | The daptomycin (19) derivative CB182296 (44) was generated through the exchange of two modules.b | Intersubunit docking-domain exchange conducted on the antimycintype cyclic depsipeptide pathways of JBIR-06 (45) and neoantimycin A (46) is more challenging than single-gene alteration, since the interplay of the modular enzymes is strictly regulated.Reorganization of these biosynthetic pathways enabled the production of 'JBIR-06' with a changed alkyl chain(47), a derivative of neoantimycin with a contracted core cyclic structure to a tri-lactone(48) and an expanded JBIR-06 derivative with a tetra-lactone core cyclic structure(49).

Fig. 7 |R e v i e w s volume 4 |
Fig.7| Engineering de novo biosynthetic pathways.Exchange units (XUs) or exchange unit condensation domains (XUCs) afford novel or altered natural products.The method holds promise for the generation of a future non-ribosomal peptide synthesizer.XUs or XUCs from different biosynthetic pathways (green, grey and red) can be fused together to generate peptides with variable size, different functionalization and cyclized scaffolds(50-55).

◀
NATure revieWS | CHEMiSTRy R e v i e w s volume 4 | April 2020 | 189