Streptomyces and Saccharopolyspora hosts for heterologous expression of secondary metabolite gene clusters

Abstract

Natural products discovery from actinomycetes has been on the decline in recent years, and has suffered from a lack of innovative ways to discover new secondary metabolites within a background of the thousands of known compounds. Recent advances in whole genome sequencing have revealed that actinomycetes with large genomes encode multiple secondary metabolite pathways, most of which remain cryptic. One approach to address the expression of cryptic pathways is to first identify novel pathways by bioinformatics, then clone and express them in well-characterized hosts with known secondary metabolomes. This process should eliminate the tedious dereplication process that has hampered natural products discovery. Several laboratory and industrial production strains have been used for heterologous production of secondary metabolite pathways. This review discusses the results of these studies, and the pros and cons of using various Streptomyces and one Saccharopolyspora strain for heterologous expression. This information should provide an experimental basis to help researchers choose hosts for current application and future development to express heterologous secondary metabolite pathways in yields sufficient for rapid scale-up, biological testing, and commercial production.

Introduction

Secondary metabolites produced by actinomycetes continue to be excellent starting scaffolds for the development of antibiotics, anticancer agents, immunomodulators, anthelminthic agents, and insect control agents. Although the rate of discovery of novel secondary metabolite chemical structures has declined in recent years, there is good reason to believe that many more important new secondary metabolites await discovery [7, 10, 12]. Many secondary metabolites have eluded discovery because their biosynthetic pathways are not widely dispersed among actinomycetes commonly isolated in soils or marine sediments, or they may not be produced in sufficient yields for detection under standard fermentation conditions. With the advent of genome sequencing, it is now apparent that actinomycetes with large genomes have the potential to produce multiple secondary metabolites, and only a fraction of these are produced under standard fermentation conditions [12, 21, 51, 57, 95, 100, 102, 123]. Bioinformatic approaches can predict certain features of the products of cryptic biosynthetic pathways [4, 35, 59, 79, 117, 126], and there are physiological and genetic approaches to stimulate the expression of otherwise cryptic genes [30, 51, 52, 120, 121, 124]. The latter include fermentations in multiple media, genomo-isotopic enrichment labeling of predicted structures, and engineering positive regulation. Another important complementary approach reviewed here is to express cryptic pathways in heterologous hosts suitable for the expression of otherwise silent secondary metabolite gene clusters.

There are a number of potential actinomycete hosts identified from molecular engineering, combinatorial biosynthesis, and gene cluster expression studies. Some of these are widely used laboratory strains that are easy to manipulate genetically, such as Streptomyces coelicolor, Streptomyces lividans, and Streptomyces albus J1074 [50], some are industrial strains that produce products that have not been commercialized, and some are producers of commercial products. The last two groups include relatively high producing wild-type strains and genetically advanced strains capable of producing elevated and often very large quantities of specific compounds. With the current genomics efforts to identify cryptic secondary metabolite biosynthetic gene clusters, it will be important to develop a suite of streptomycete hosts for high-level heterologous expression of specific types of secondary metabolites, and to facilitate rapid identification and testing of novel products. In this review, I discuss the history, rationale, and examples of Streptomyces (and one Saccharopolyspora) hosts used for the heterologous expression of secondary metabolites. This information should provide a rationale and experimental basis to help choose hosts for the expression of cryptic secondary metabolite pathways. Some of these hosts have already proven to be valuable for combinatorial biosynthesis to produce novel derivatives of pharmacologically active secondary metabolites. In many cases, genomic sequences of the expression hosts are available.

General cloning hosts

Streptomyces albus J1074

The establishment of S. albus G as a readily transformable cloning host began years before cloning was possible in streptomycetes. Chater and Wilde [31] demonstrated that S. albus G expresses restriction endonuclease SalI that restricts the growth of bacteriophage Pal6. They isolated a mutant (S. albus J1074) defective in SalI restriction and modification. Subsequent studies showed that S. albus G restricts the growth of 8 of 10 broad host range streptomycete bacteriophages, but all 10 formed plaques on J1074 [36]. Bacteriophage FP22, which has no SalI sites in its double-stranded DNA, formed plaques on S. albus G, whereas FP4 and FP43, which have more than 30 SalI sites, did not form plaques on S. albus G, but formed plaques on J1074. Protoplasts of S. albus J1074 regenerated viable colonies efficiently on modified R2 medium [15], a prerequisite for protoplast transformation. These early studies established the potential suitability of S. albus J1074 as a host for cloning and expression of streptomycete genes.

S. albus J1074 was first used as a cloning host by the Salas group [48, 109]. They used S. albus for the heterologous expression of the steffimycin biosynthetic genes, and obtained a yield of purified compounds of about 10 mg/l [53]. In this case, they needed to clone a set of rhamnose biosynthetic genes with the steffimycin gene cluster to obtain steffimycin production, because S. albus lacks rhamnose biosynthetic genes. This system was used for combinatorial biosynthesis of molecules related to steffimycin, with yields of novel compounds with antitumor activities of up to 35 % of control [101].

S. albus J1074 has been used for the expression of several other secondary metabolite gene clusters (see Table 1). The fredericamycin biosynthetic gene cluster was expressed from both high copy number [127] and single copy number ϕC31-integrating vectors [32], yielding 120 and 132 mg/l product, respectively. In contrast, when the fredericamycin pathway was inserted into the S. lividans ϕC31 attB site, no product was detected [32]. In S. lividans, a small amount of product (0.5 mg/l) was detected when the fdmR1 gene encoding a positive regulatory protein belonging to the Streptomyces antibiotic regulatory protein (SARP) family [22, 23, 124] was added to the strain on a multicopy plasmid. The yield was further enhanced to 1.4 mg/l when the fdmR1 gene was expressed from the strong constitutive ermEp* promoter on a high copy number vector. Transcription studies identified the fdmC gene, encoding a putative ketoreductase, as a possible bottleneck. When they expressed both fdmR1 and fdmC genes driven by the ermEp* promoter, the yield of fredericamycin was increased to 17 mg/l, still only 13% of the yield obtained by expressing the fredericamycin gene cluster from its native promoters in S. albus J1074.

Examples of heterologous expression of complete secondary metabolic biosynthetic pathways in Streptomyces hosts

Native strain	Product	Insert (kb)	Expression plasmid	Expression host	Yield (mg/l)	Reference
S. griseus	Fredericamycin	33	pWHM3	S. albus J1074	120	[127]
			pSET152	S. albus J1074	132	[32]
				S. lividans K4-114	0	[32]
S. platensis	Iso-migrastatin	65	pStreptoBAC V	S. albus J1074	46	[47]
				S. lividans K4-114	25	[47]
				S. coelicolor M512	23	[47]
				S. avermitilis SUKA5	4.2	[47]
Streptomyces sp.	Napyradiomycin	36	pOJ446	S. albus J1074	NR	[129]
Micromonospora sp.	Thiocoraline	53	ϕC31 attP/int-based	S. albus J1074	NR	[80]
S. fradiae A54145	A54145	>100	pDA2002	S. ambofaciens BES2074	386	[3]
				S. roseosporus UA431	100	[3]
S. griseus	Streptomycin	41.2	pKU465cos	S. avermitilis SUKA5	~180	[66]
S. clavuligerus	Cephamycin C	~40	pKU465cos	S. avermitilis SUKA17	~80	[66]
			pKU465cos + rpsJp-ccaR	S. avermitilis SUKA17	~130	[66]
S. spheroides	Novobiocin	34.5	ϕC31 attP/int-based	S. coelicolor M512	31	[44, 45]
				S. lividans TK24	<1	[45]
S. erythraea	6dEB	~30	pSET152	S. coelicolor CH999	40, 47a	[110]
				S. lividans K4-114	33, 28a	[110]
S. roseochromogenes	Chlorobiocin	45	ϕC31 attP/int-based	S. coelicolor M512	26	[45]
S. alboniger	Puromycin	15	pKC505	S. griseofuscus	2–4	[72]
				S. lividans 66	8–15	[72]
S. rimosus	Oxytetracycline	34	pIJ916	S. lividans 1326	20	[25]
Streptomyces sp.	Staurosporine	34	ϕC31 attP/int-based	S. lividans TK23	2.6	[103]
S. griseus	Macrotetrolide	25	pSET152	S. lividans 1326	10	[71]
S. roseosporus	Daptomycin	128	pStreptoBac V	S. lividans TK23 Δact	20, 55b	[104]
S. mutabilis	Capreomycin	~35	pOJ436-based	S. lividans 1326	50	[46]
S. tendae	Nikkomycin	~43	pKC505	S. lividans TK23	400–500	[29]

NR not reported

^aYields for genes cloned from wild-type and high producing strains of S. erythraea, respectively

^bYields in standard and phosphate-supplemented media, respectively

Another interesting use of S. albus J1074 was the cloning and expression of the complete biosynthetic gene cluster for isomigrastatin from Streptomyces platensis NRRL18993 into pStreptoBac V [47]. In this study, the isomigrastatin gene cluster was inserted into the ϕC31 attB site in S. albus, S. lividans, S. coelicolor, and S. avermitilis, and fermentations were carried out in two media. The best yields were obtained in medium B2. The parental strain produced 58 mg/l, whereas the S. albus J1074 recombinant produced 46 mg/l. S. lividans K4-114 and S. coelicolor M512 produced about twofold lower yields, and S. avermitilis SUKA5 produced a tenfold lower yield (Table 1).

A particularly relevant application of S. albus J1074 was the expression of the thiocoraline pathway cloned from a marine Micromonospora strain [80]. In this study, the native thiocoraline pathway was not expressed in S. albus or S. lividans unless the tioA positive regulatory gene encoding an OmpR-family regulator was transcribed from the ermEp* promoter. This demonstrated that a non-streptomycete actinomycete secondary metabolite gene cluster can be expressed in S. albus and S. lividans, and points out the utility of ensuring the expression of all necessary genes by intervening in the expression of a single positive regulatory gene. This approach may be applicable to the heterologous expression of otherwise cryptic secondary metabolite pathways, particularly from non-streptomycete actinomycetes not amenable to standard molecular genetic and fermentation manipulations that are routinely executed in streptomycetes.

The genome of S. albus J1074 has been sequenced at the Broad Institute, and is available at the National Center for Biotechnology Information (NCBI) website (http://www.ncbi.nlm.nih.gov/).

S. coelicolor and S. lividans

S. coelicolor and S. lividans are two very closely related laboratory strains that have been used extensively in the development of streptomycete genetic tools [62]. The genomic sequences of both strains are available at NCBI, and they differ from each other primarily by their composition of horizontally acquired genes [58]. S. coelicolor has five large (>25 kb) and 18 smaller genomic islands that are not present in S. lividans. Much is known about the biosynthesis and regulation of secondary metabolites in these strains [22, 23, 30, 124]. In spite of the very close phylogenetic relationship, S. coelicolor and S. lividans differ dramatically in expression of their common secondary metabolite biosynthetic pathways. For instance, wild-type S. lividans contains essentially silent or cryptic biosynthetic gene clusters for actinorhodin (Act), undecyleprodigiosin (Red), and calcium-dependent antibiotic (CDA), whereas all three of these are normally produced by S. coelicolor. There are a number of ways to trigger the expression of one or more of these gene clusters in S. lividans. Notably, certain rpsL and rsmG mutations strongly activate the expression of Act, and certain rpoB mutations associated with resistance to rifampin (RifR) activate the expression of Act, Red, and CDA [99]. In addition, it has been shown that some rare, naturally RifR actinomycetes have a second paralogous rpoB(R) gene that has mutations associated with RifR [120]. Introduction of this gene into the chromosome of S. lividans triggered abundant expression of Act, Red, and CDA. It is not yet known if the mechanism(s) of activating these three pathways will be generally translatable to the activation of cryptic genes in other actinomycetes, but it is an area that needs further investigation. An advantage of the S. lividans system is that Act (blue) and Red pigments are easily observed, so this system can be used to screen genes, fermentation conditions, and chemical compounds for the activation of expression of these antibiotics. These in turn might be further screened to identify potential general activators of cryptic pathways in other actinomycetes.

S. coelicolor and S. lividans have been used as hosts for the expression of several secondary metabolite biosynthetic gene clusters (Table 1). S. coelicolor was used in early studies on the expression the erythronolide PKS genes, with production yields of about 50 mg/l [61]. The heterologous product yields of other compounds have varied from 23 to 47 mg/l in S. coelicolor and from 0 to ~450 mg/l in S. lividans (Table 1). In some cases, product yields were substantially better in S. coelicolor than in S. lividans. For instance, when the novobiocin biosynthetic gene cluster was inserted into the ϕC31 attB sites of S. coelicolor M512 and S. lividans TK24, the product yields were 31 and less than 1 mg/l, respectively [44]. The fermentation yield in S. coelicolor was increased to 54 mg/l by miniaturizing the fermentation and adding an oxygen carrier, siloxylated ethylene oxide/propylene oxide copolymer. The yield was further increased to 163 mg/l by overexpressing the novG gene encoding a StrR-like positive regulatory protein [113].

The complete daptomycin gene cluster present in pStreptoBAC V [92] was inserted into the ϕC31 attB site of a S. lividans strain deleted for the act gene cluster, and the recombinant produced the natural A21978C lipopeptides at ~20 mg/l [104]. The recombinant also produced the otherwise cryptic CDA lipopeptides, confounding the analysis. CDA production was suppressed, and A21978C production stimulated to 55 mg/l by increasing the phosphate level in the medium. This yield was about 50% of the yield normally obtained with wild-type S. roseosporus in an optimized medium, but only a fraction of that produced in mutationally advanced strains.

S. lividans 1326 has also been used for the heterologous expression of tuberactinomycin family nonribosomal peptide antibiotics capreomycin [46] and viomycin [19]. Capreomycin is normally produced by Saccharothrix mutabilis subsp. capreolus, a strain apparently intractable to molecular genetic manipulations. The capreomycin gene cluster was inserted into the ϕC31 attB site in S. lividans, and the recombinant produced 50 mg/l of capreomycin. The viomycin gene cluster was cloned from Streptomyces sp. strain ATCC 11861 and inserted in a similar manner into S. lividans. The gene cluster was manipulated by λ-Red-mediated recombination in E. coli to generate specific deletions, then inserted into S. lividans by conjugation and site-specific integration in the ϕC31 attB site to study gene functions [19]. These two examples point out the value of S. lividans as a host for molecular genetic and biosynthetic studies that otherwise cannot be carried out in the native hosts.

Recently, a strain of S. coelicolor deleted for the actinorhodin (act), prodiginines (red), PKSI (cpk), and calcium dependent antibiotic (cda) clusters was constructed, and combined with mutations in rpoB and rpsL that enhance secondary metabolite production (J. P. Gomez-Escribano and M. J. Bibb, Abstracts of the 15th International Symposium on the Biology of Actinomycetes, Shanghai, August 20–25, 2009). This strain has been used to produce enhanced levels of cloned secondary metabolite gene clusters, but product yields were not reported.

Streptomyces griseofuscus

S. griseofuscus was identified many years ago as an excellent host for the isolation and propagation of broad host range Streptomyces bacteriophages [36]. Of the 30 Streptomyces species tested, S. griseofuscus was the best strain to determine bacteriophage titers because it displayed no discernable DNA restriction. S. griseofuscus was also shown to form protoplasts that regenerated viable colonies at high efficiency [5, 15]. As anticipated from the protoplast regeneration and bacteriophage restriction studies, S. griseofuscus was shown to be readily transformable by unmodified plasmid DNA, and was used to generate bifunctional E. coli–Streptomyces shuttle plasmids and cosmids [74, 75, 107]. S. griseofuscus was also used for the development of bacteriophage FP43-mediated plasmid transduction [88], and transposon technology [54, 88, 89, 114, 115], and served as a host for cloning of tylosin, carbomycin, and spiramycin resistance genes [26, 42, 107], which provided entry points for the cloning and engineering macrolide antibiotic biosynthetic genes at Lilly [37, 43, 70].

S. griseofuscus was used as a host to clone and express the complete puromycin biosynthetic gene cluster from Streptomyces alboniger ATTC12461 from a low copy number bifunctional cosmid pKC505 [72, 107]. The puromycin yields were 2–4 mg/l from different clones. The same plasmids gave yields of 8–15 mg/l in S. lividans 66. S. griseofuscus represents a laboratory strain with potential for heterologous whole pathway expression that has been relatively unexplored. Its rapid growth (doubling time ~1 h in TS broth at 34°C [5]) and ease of genetic manipulation make it an attractive strain for further studies.

Hosts derived from industrial polyketide producers

Streptomyces ambofaciens

S. ambofaciens produces the clinically important macrolide antibiotic spiramycin [111]. Early work at Lilly demonstrated that, like S. griseofuscus, S. ambofaciens was a suitable host for gene cloning because it did not restrict double-stranded DNA from a variety of streptomycete bacteriophages [36], and protoplasts regenerated viable cells efficiently [15] and were readily transformable with plasmid DNA prepared from other hosts [81]. This enabled the use of S. ambofaciens as a reliable host for the development of site-specific integration vectors [67–69]. These studies identified the highly efficient chromosomal insertion by non-replicating plasmids containing ϕC31 attP and int functions [69], thus enabling the further development of a suite of very useful plasmid and cosmid vectors with site-specific integration and conjugal transfer functions [24]. Recent studies have reconfirmed that plasmid pSET152 can be introduced into S. ambofaciens at high efficiency (1.4 × 10⁻² per recipient) by conjugation from E. coli and site-specific insertion into the ϕC31 attB site when selections were done on AS-1 agar [63].

The initial problem with using S. ambofaciens as an expression host for the genetic engineering of macrolide antibiotic biosynthetic pathways was its genetic instability. In early protoplast studies, it was observed that spiramycin production was lost at very high frequencies among colonies derived from regenerated protoplasts [Matsushima and Baltz, unpublished]. Instability in S. ambofaciens has been the subject of extensive research (e.g., see [1, 2, 33, 39, 40, 76–78, 128]), and the chromosome ends which harbor regions prone to DNA deletion and amplification have recently been sequenced [33]. In the pre-genomic era of the early 1980s, the pragmatic solution to genetic instability in S. ambofaciens was the use of chemical mutagenesis and selection for stability. S. ambofaciens was mutagenized with N-methyl-N-nitro-N-nitrosoguanidine (MNNG), protoplasts were prepared from mutagenized cells and regenerated, colonies were screened for production of elevated levels of spiramycin, and the process was repeated [49]. The starting strain (ATCC 15154) produced 128 mg/l of spiramycin, and the stable mutant strain (S. ambofaciens 111–59) produced 1.7 g/l. A derivative of strain 111–59 was used as a cloning host to produce hybrid 16-member macrolides [70].

S. ambofaciens BES2074, a derivative of strain 111–59 blocked in spiramycin and netropsin production, has been used as a host for the heterologous production of the cyclic lipopeptide A54145 [3, 18]. A BAC vector containing a larger than 100-kb insert, which included the complete A54145 biosynthetic gene cluster [93], was conjugated into S. ambofaciens from E. coli, and the genes inserted into the chromosomal ϕC31 attB site. The recombinant produced an average of 385 mg/l of the A54145 lipopeptides (Table 1). This represented 91% of the yield of a moderately improved strain of S. fradiae, and an ~285 % increase over the wild-type S. fradiae strain [3, 28]. Now that S. ambofaciens BES2074 has shown clear potential to produce products derived from PKS and NRPS pathways, it would be interesting to sequence its genome to see how empirical mutation and selection for stability has influenced the genome size and structure.

Streptomyces avermitilis

S. avermitilis is the commercial producer of avermectins [57]. The S. avermitilis genome has been sequenced and annotated, and many cryptic secondary metabolite pathways have been identified [57, 60, 65, 95, 122]. Besides avermectins, wild-type S. avermitilis normally produces oligomycins and filipins as major secondary metabolites [66]. Tanaka et al. [121] have shown that the oligomycin biosynthetic gene cluster, which is normally very poorly expressed, can be highly expressed (~1 g/l) in an rpsL mutant (K43M) deleted for the aveA1 (PKS) gene. Several strains containing deletions of large segments of the S. avermitilis genome have been generated, including a series of strains with genomes reduced from 9.03 to 7.35–7.50 Mb [60, 66]. These include strains deleted for avermectins, oligomycin, filipin, the terpines geosmin and neopentalenolactone, and carotenoid. The large deletion mutation in the left arm of the sub-teleomeric region also removed about 78% of the 111 transposase genes, perhaps improving the overall stability of the genome-minimized strains.

The genome-minimized strains were used to express heterologous secondary metabolite pathways, including the streptomycin gene cluster from Streptomyces griseus, and the cephamycin C gene cluster from Streptomyces clavuligerus, and the pladienolide pathway from Streptomyces platensis. The streptomycin gene cluster was inserted into the ϕC31 attB site in the wild-type S. avermitilis strain and in a deletion mutant, SUKA5. The highest levels of production were obtained in an avermectin production medium. The wild-type S. avermitilis recombinant produced ~30 mg/l streptomycin, whereas the SUKA5 recombinant produced ~180 mg/l. The latter compared favorably with wild-type S. griseus which produced ~50 mg/l of streptomycin in a streptomycin production medium. They also showed that the positive regulatory gene strR could be driven by the rpsJ and the aveR promoters, and streptomycin yields of ~170 and ~200 mg/l, respectively, were obtained. The use of alternative promoters to drive expression of strR decoupled the pathway regulation from a more complex mechanism that involved induction by A-factor.

Cephamycin production was assessed in another deletion mutant, S. avermitilis SUKA17, by inserting the pathway into the ϕC31 attB site. In this case, no cephamycin production was observed in the cephamycin production medium, but cephamycin was produced in the avermectin production medium at ~80 mg/l. A second copy of the positive regulatory gene ccaR was introduced into the bacteriophage ϕK38-1 attB site under the transcriptional control of the rpsJ promoter, and the yield was increased to ~130 mg/l. In this case, the use of the rpsJ promoter decoupled expression of the cephamycin gene cluster from transcriptional control involving an anti-sigma factor antagonist encoded by bldG in S. clavuligerus, and which is fortuitously provided by the rsvV gene in S. avermitilis. Deletion of rsvV in S. avermitilis SUKA17 abolished cephamycin production, but this was overcome by driving expression of ccrR by the rpsJ promoter.

A third example was the expression of the S. platensis pladienolide biosynthetic gene cluster from a conjugal BAC vector inserted in the ϕC31 attB site. In this case, no expression was observed in wild-type or genome-minimized S. avermitilis strains, and transcription analysis indicated that the pathway transcriptional activator gene pldR was not transcribed. However, when a second copy of pldR under the control of the ermEp promoter was inserted into ϕK38-1 attB site in recombinants containing the complete pathway, pladienolides were produced in both strains, but at substantially higher yield in the genome-minimized strain. The authors speculated that the genome-minimized strain was able to channel acyl-CoA intermediates common to the avermectin and pladienolide polyketide pathways into pladienolides exclusively, thus accounting for the higher productivity.

Streptomyces fradiae (tylosin producer)

Tylosin and a semisynthetic derivative, tilmicosin, are important antibiotics used in animal health [64]. Tylosin is produced by S. fradiae T59235, C4 and related strains [112], and the genetics and biosynthesis of tylosin production have been under investigation for many years [6, 16, 17, 38, 112, 119]. S. fradiae C4 is a relatively high tylosin producer derived by multiple rounds of mutagenesis starting from the wild-type S. fradiae T59235, and many derivatives of C4 have been described [16, 17, 112]. The advantages of developing an advanced tylosin production strain of S. fradiae as a cloning host to produce hybrid macrolides were articulated many years ago [6, 17]. To achieve this, early genetic studies on S. fradiae focused on the development of protoplast fusion [5, 15], protoplast transformation [81, 84, 85], and tylosin biosynthesis [16, 17]. S. fradiae protoplasts are transformable by plasmids modified by passage through S. fradiae M1, a non-sporulating mutant that is less restricting than typical tylosin production strains, but poorly transformable by plasmids prepared from other streptomycetes [81]. S. fradiae was shown to express multiple restriction/modification systems, and the genes encoding restriction enzymes were systematically inactivated by multiple rounds of MNNG mutagenesis followed by selection for improved transformation by different plasmids, starting with strain JS85, a tylosin non-producing derivative of T59235 [84]. This process yielded S. fradiae PM76 that was highly transformable by plasmids prepared from other streptomycetes, and loss of restriction was supported by bacteriophage host range studies [118]. DNA from JS85 was insensitive to cleavage by ScaI, PstI, XhoI, and MstI, but PM76 was cleaved by all four enzymes, suggesting that the mutations associated with improved transformation disrupted restriction and modification for these systems. Like its parent strain JS85, PM76 lacked a functional tylosin biosynthetic pathway [17, 84], but was an efficient recipient for the conjugal transfer of tylosin genes from other strains [17, 118]. Since trans-conjugants retained the non-restricting phenotype of the recipient, this provided a means to reintroduce specific mutations in tylosin biosynthetic genes into a non-restricting background for cloning studies. A recombinant containing the tyl gene cluster with a tylB mutation was instrumental in identifying cosmids containing spiramycin biosynthetic genes by heterospecific complementation [108].

Since S. fradiae PM76 was derived from a wild-type tylosin producer, the problem of DNA restriction remained a barrier to the use of the best tylosin production strains for genetic engineering in the 1980s. This dilemma was solved by introducing DNA into S. fradiae by plasmid RP4-mediated conjugation of streptomycete plasmid DNA from E. coli. Mazodier et al. [87] demonstrated that plasmids containing oriT from RP4 could be transferred from E. coli to streptomycetes by conjugation, and Bierman et al. [24] extended this observation by showing that plasmids could be conjugated from E. coli into highly restricting S. fradiae strains at high frequencies. This indicated that the plasmid DNA introduced by RP4 transfer functions, which was likely transferred as single-stranded concatemers generated by rolling circle replication [73], was able to bypass the potent restriction barriers that normally cut double-stranded DNA. This system has also been used to bypass potent restriction barriers in Saccharopolyspora spinosa [82, 86], the producer of the commercial insect control agents spinosad and spinetoram [56], further emphasizing the utility of conjugation to engineer important industrial strains that otherwise might not be amenable to genetic manipulation.

Rodriquez et al. [110] used the approach of engineering an S. fradiae strain that had undergone mutagenesis and selection for elevated tylosin production. The strain produced 2 g/l tylosin in shake flask cultures. They deleted the tylGI-V PKS genes, then conjugated and inserted pSET152 containing the PKS genes from the wild-type S. fradiae strain into the ϕC31 attB site under the control of its natural tylGIp promoter, and the recombinant produced 1.3 g/l of tylosin. They concluded that mutations associated with high productivity resided outside of the main PKS cluster. This is consistent with the finding of the Stratigopoulos and Cundliffe [119] that strains derived from S. fradiae C4, which had undergone at least seven cycles of mutation and selection at Lilly [112], and produced more than 3 g/l of tylosin in shake flasks [Seno and Baltz, unpublished], had only one mutation in the tylosin gene cluster, mapping to a negative regulatory gene, so the other mutations must reside outside of the tylosin gene cluster. This observation is important in that the Kosan strain and the Lilly C4 strain may be genetically well suited to produce high levels of other heterologous type I polyketides.

The Kosan group used the S. fradiae expression system, driven by the tylGIp promoter, to replace the tylGI-V PKS genes with the comparable genes (>40 kb) from the chalcomycin pathway to produce hybrid macrolide antibiotics [125]. In a separate study, Reeves et al. [106] used this system to generate a hybrid 16-member macrolide derived from the first two chalcomycin PKS genes (chmGI–II) and the last three spiramycin PKS genes (srmIII–V), and the recombinant produced novel macrolide antibiotics at nearly 2 g/l.

The success of the Kosan group suggests that even higher yields of heterologous polyketides might be obtained by using more advanced tylosin production strains. For instance, a number of mutants derived from strain C4 are generally available [17]. Some of these (e.g., GS3 and GS22) are deleted for segments of the PKS genes [16, 17, 20], and might be useful hosts for the production of type I polyketides. Preferably, strains further developed along the production strain lineage could be used as cloning hosts after deleting the tylGI–V PKS genes, as demonstrated by the Kosan group, or after deleting the complete tyl gene cluster to eliminate the sugar biosynthetic and other tailoring functions as well.

Saccharopolyspora erythraea

S. erythraea is the commercial producer of erythromycin, and its genome has been sequenced [102]. Although S. erythraea is not a member of the genus Streptomyces, it is included in this review because it is as an example of how molecular genetic tools developed mainly in Streptomyces species can be adapted for use in other actinomycetes. S. erythraea does not have a high efficiency ϕC31 attB site for insertional cloning. In order to exploit the streptomycete vector pSET152, which has oriT for conjugation and ϕC31 att/int functions [24], Rodriquez et al. [110] generated a S. erythraea cloning host by deleting the PKS genes eryAI–III, and replacing them with the ϕC31 attB site from S. lividans. They also inserted the ermEp* promoter to drive expression of downstream ery genes. This expanded the utility of the ϕC31-integrating vectors to a genus lacking an efficient attB for integration, and provided a means for precise insertion of cloned PKS genes into the vacant eryAI–III PKS locus. They reintroduced the eryAI–III genes under the control of the natural eryAIp promoter from the high producer, which normally produces 1–2 g/l erythromycin in shake flasks, and from the wild-type strain, which normally produces about 200 mg/l erythromycin. Both recombinants produced about 1.3 g/l erythromycins, indicating that mutations in the high producer accounting for elevated erythromycin production must reside outside of the eryAI–III genes and the eryAIp promoter. Therefore, as demonstrated with the S. fradiae system, the S. erythraea system may also be suitable for the expression of cryptic PKS gene clusters from other actinomycetes. As with S. fradiae tylosin producers, there are S. erythraea industrial strains capable of producing much higher levels of erythromycin that might be manipulated as discussed above to produce heterologous macrolide biosynthetic pathways.

Hosts derived from nonribosomal peptide producers

Streptomyces roseosporus

S. roseosporus is used for the commercial production of the therapeutically important lipopeptide antibiotic daptomycin [41], and its genome has been sequenced by the Broad Institute. S. roseosporus is a facile host for genetic manipulation by conjugation, plasmid transduction, transposition, and gene replacement [55, 90, 91], and it has been developed as a cloning host for genetic engineering and combinatorial biosynthesis of novel lipopeptide antibiotics at Cubist Pharmaceuticals [13, 14, 18, 34, 94, 96, 97]. A variety of deletion mutants have been generated in a strain that normally produces about 300 mg/l of daptomycin in shake flasks, including strain UA431 which is deleted for the complete daptomycin biosynthetic gene cluster [34, 96, 97]. Many novel lipopeptides have been generated by combinatorial biosynthesis with yields ranging from 1 to 100 mg/l. The complete A54145 lipopeptide biosynthetic gene cluster cloned in a BAC vector was conjugated from E. coli into an S. roseosporus UA431, inserted into the ϕC31 attB site, and the recombinant produced 100 mg/l of A54145 factors [3]. S. roseosporus UA431 may be suitable host for cloning and expression of cryptic NRPS gene clusters.

Streptomyces fradiae A54145

S. fradiae A54145 is one of several streptomycete strains isolated at the Lilly Research Laboratories that produce a family of the cyclic lipopeptide antibiotics called A54145 factors [28]. S. fradiae A54145 was chosen for preclinical structure–activity relationship studies because it was the most robust strain, producing about 100 mg/l of A54145 factors without any strain improvement. A strain that has been through minimal mutagenesis and selection and produces more than 400 mg/l of A54145 factors has been developed at Cubist Pharmaceuticals as a host for the genetic engineering of lipopeptide biosynthesis [3, 98]. S. fradiae DA1187 is deleted for the complete A54145 biosynthetic gene cluster [3], and readily accepts BAC vectors containing more than 100-kb inserts via conjugation from E. coli, and site-specific insertion into the ϕC31 or ϕBT1 attB sites. Exconjugants containing the complete A54145 gene cluster inserted into either of these sites produced more than 400 mg/l of A54145 factors [3]. S. fradiae DA1187 may be another useful host for the expression of cryptic NRPS pathways.

Streptomyces toyocaensis A47934

S. toyocaensis A47934 produces the non-glycosylated “glycopeptide” A47934 which comprises the sulfated core heptapeptide of teicoplanin [27, 130]. Because A47934 contains no sugar residues, it is a convenient starting scaffold for the addition of sugar residues to generate novel glycopeptides. S. toyocaensis protoplasts are transformable by some plasmids prepared from S. lividans, and it is a good recipient for bacteriophage FP43-mediated plasmid transduction [83]. Although it appears to express some DNA restriction, non-replicating plasmids, including one with a 37-kb insert of heterologous DNA, could be introduced into S. toyocaensis by conjugation from E. coli and inserted into the ϕC31 attB site at high efficiencies. Recombinants produced A47934 at yields ranging from about 300 to 600 mg/l [83], which represented only minor reductions from control A47934 yields. The conjugation/insertion system was used to generate the first recombinant to produce a hybrid glycopeptide antibiotic, glucosyl-A47934, by introducing the gtfE gene from the vancomycin-producing Amycolatopsis orientalis [116]. Importantly, the product yield was about 500 mg/l. This system may be suitable for the heterologous expression of cryptic NRPS pathways. The genes for the biosynthesis of A47934 have been cloned and sequenced [105], so the nucleotide sequence information is available to delete A47934 biosynthetic genes to generate expression hosts.

Discussion

Model organism vs. a suite of production hosts

In academia, there is a tendency to focus on model systems to solve problems that may be relevant in the long term to commercial enterprise. In industry, there is a tendency to work on short-term problems that can be solved with existing technology. This dual approach worked reasonably well in the past, but it is probably not sufficiently robust for the discovery of novel secondary metabolites from cryptic pathways in actinomycetes. The concept of having a “model organism” for carrying out molecular engineering manipulations in bacteria has been well served by E. coli. The focused work on E. coli has paid off over the years, and many of its advantages have been exploited for the genetic engineering of actinomycetes, as witnessed by the facile recombineering of actinomycete secondary metabolite biosynthetic pathways in E. coli using λ-Red-mediated recombination, followed by RP4-mediated conjugation into actinomycetes and site-specific integration into different bacteriophage or IS element attB sites [14, 18]. The notion that a single model organism can be developed for antibiotic-producing actinomycetes for the heterologous expression of secondary metabolite pathways is not realistic, and does not reflect the current state-of-the-art of facile genetic manipulations or the current knowledge about the potential for high-level production of specific types of secondary metabolites in different Streptomyces species. Most streptomycetes, and many other actinomycetes, can now be manipulated genetically. The main driver for this is the ability to conjugate between E. coli and actinomycetes [24, 86, 87], and in so doing to bypass host restriction barriers, which are common in actinomycetes [36]. Another important driver is the ability to insert large blocks of genes cloned in cosmid or BAC vectors very efficiently into ϕC31, ϕBT1 or other bacteriophage attB sites for stable expression of the cloned genes [3, 14, 18, 66]. Having broadly applicable genetic tools has opened the possibility to have a “suite of production hosts” for whole pathway expression in Streptomyces, and very probably in other actinomycetes, as witnessed by the successes in S. erythraea [110].

S. coelicolor and S. lividans have served as model organisms for many years, and they have helped progress the field of Streptomyces genetics enormously. The sequencing of the S. coelicolor genome [21] has served as a catalyst for others to sequence more actinomycete genomes (e.g., see [57, 95, 102]), and as a harbinger of the observation that actinomycetes with large genomes encoded multiple secondary metabolite functions, most of which are silent under standard fermentation conditions. This observation has been expanded dramatically with the sequencing of multiple streptomycete and other actinomycete genomes by the Broad Institute and the US Department of Energy.

Not withstanding the important impact of S. coelicolor and S. lividans as model organisms in the past, the cumulative results of heterologous secondary metabolite expression levels in S. coelicolor and S. lividans do not support them as first choices for heterologous expression of cryptic pathways. In fact, another laboratory strain, S. albus J1074, which has been recently sequenced by the Broad Institute, has proven to be a more robust producer of a wide variety of heterologously expressed secondary metabolites (Table 1). Many industrial strains also have clear advantages for high-level production of heterologous secondary metabolites.

Industrial strains selected for high-level production of specific compounds

In principle, a suite of ideal expression hosts would include several species that are capable of producing very high levels of specific types of molecules (e.g., those derived from PKS or NRPS pathways). Since the speed of discovery and evaluation of new natural products is intimately linked to fermentation production yields, having high-level production strains will be important determinants of the success or failure of genomics efforts to identify novel secondary metabolites. The yields obtained for heterologous expression of complete secondary metabolite biosynthetic pathways reviewed here ranged from about 4 to 400 mg/l, and expression of hybrid pathways ranged from 1 to 2 g/l. It is not surprising that the highest yields for the production of hybrid PKS pathways were obtained in strains of S. fradiae and S. erythraea deleted for the production of the commercial antibiotics tylosin and erythromycin, respectively. These expression hosts were derived from mutationally improved strains that produce about 2 g/l of product in shake flask fermentations. It is noteworthy that most of the mutations that influence polyketide yields in S. fradiae and S. erythraea are located outside of the tylosin and erythromycin biosynthetic gene clusters. These strains have not been used for heterologous expression of complete secondary metabolite pathways, but the location of most mutations influencing product yields outside of the main polyketide biosynthetic clusters bodes well for the channeling of precursors into heterologously expressed PKS pathways in these hosts.

Silencing of major secondary metabolite pathways

Mutational silencing of major secondary metabolite biosynthetic pathways has two key advantages: (1) it simplifies the identification of the products of cloned heterologous gene clusters; and (2) it eliminates the channeling of key precursors into competing product pathways, thus improving the yields of the desired product. Two key examples are S. avermitilis and S. ambofaciens. In S. avermitilis, disruption of the avermectin pathway caused a large increase in the production of oligomycin, a product normally produced in low levels in S. avermitilis. Elimination of multiple secondary metabolite pathways in the genome-minimized strains provided hosts that produced several other heterologously expressed secondary metabolites at fairly good yields. There are undoubtedly strains of S. avermitilis in industry that produce much higher yields of avermectins than the starter strain used by Ōmura, Ikeda, and colleagues [66]. It would be interesting to see if minimizing the genome of such a strain would give even higher yields.

In S. ambofaciens BES2074 which was initially selected for stable, elevated production of the 16-membered macrolide spiramycin, then subsequently blocked in biosynthesis of spiramycin and netropsin, the heterologous expression of the lipopeptide antibiotic A54145 was surprisingly high (385 mg/l) without any modification of a medium optimized for A54145 production by S. fradiae. It would be instructive to sequence the genome of BES2074 to see how it differs from the unstable, low spiramycin-producing wild-type strain.

Engineering the regulation of secondary metabolite biosynthesis

The biosynthesis of secondary metabolites can have multiple levels of regulation [22, 23] that may not be predictable a priori. However, many secondary metabolite biosynthetic pathways have pathway-specific positive regulation that can be identified by the inclusion of the regulatory gene(s) within the biosynthetic gene clusters. There are several examples discussed in this review where the simple replacement of a promoter for a key positive regulatory gene can uncouple the expression from a more complicated regulatory cascade. For example, the use of the rpsJ or aveR promoters decoupled heterologous streptomycin production in S. avermitilis from a more complicated regulatory cascade requiring A-factor. Likewise, cephamycin C production in S. avermitilis was improved by adding a second copy of the positive regulatory gene ccaR under the transcriptional control of ermEp*. In a third case, the pladienolide biosynthetic gene cluster was not expressed in the genome-minimized S. avermitilis unless a second copy of the positive regulatory gene pldR was expressed from the ermEp* promoter. In these three cases, the genetic manipulations were facilitated by having vectors for the insertion of the original gene clusters into the ϕC31 attB site, and the second engineered regulatory gene into the ϕK38-1 attB site. The ϕBT1 attB site could potentially provide a third site for insertion of additional genes [3].

The yield of novobiocin was improved ~3-fold in S. coelicolor by overexpression of the positive regulatory gene novG [113]. In S. lividans, the fredericamycin biosynthetic gene cluster was not expressed unless the positive regulatory gene fdmR1 was expressed from ermEp* [127]. However, in the same study, the fredericamycin pathway was expressed at much higher levels without additional engineering in S. albus J1074. This is a good example of the desirability of testing multiple expression hosts. All of these examples (and one below) lend credence to the notion that expression of heterologous secondary metabolite biosynthetic pathways can often be assured by simply expressing a single positive regulatory gene from a strong constitutive promoter, and that ermEp* is a highly successful promoter of choice.

Expression of secondary metabolite biosynthetic pathways from non-streptomycete actinomycetes

There are well over 200 genera of actinobacteria described, and many more yet to be discovered [7]. Many grow slowly relative to their Streptomyces relatives, and are not easily scaled in fermentation. However, genome sequencing can be used to explore the potential of rare and poorly studied actinomycetes as sources for novel secondary metabolites, followed by cloning and expression of the novel pathways in more robust streptomycete hosts. In this regard, the two examples reviewed here are particularly compelling. Capreomycin is produced by a species of Saccharothrix which is apparently not tractable for genetic studies. The complete pathway was expressed in S. lividans at 50 mg/l without modifying its promoters or expression of regulatory elements [46]. On the other hand, the gene cluster encoding thiocoraline biosynthesis from a marine Micromonospora species was not initially expressed in S. albus J1074, but was expressed after the ermEp* promoter was inserted to drive the expression of the tioA positive regulatory gene [80]. These two examples suggest that streptomycete expression hosts may be widely applicable for the expression of cryptic pathways from non-streptomycete actinomycetes.

Enhancing production by manipulating fermentation conditions

Although fermentation optimization was not a major focus of this review, it is worth mentioning several observations. First, product yields for daptomycin production in S. lividans were improved substantially (from 20 to 55 mg/l) by simply increasing the phosphate level in a production medium that had been optimized previously for daptomycin production in S. roseosporus [104]. Similarly, novobiocin yields in S. coelicolor were improved ~2-fold by miniaturizing the fermentation vessels and increasing the oxygen transfer [113]. Therefore, medium development can improve productivity in laboratory strains. Secondly, in S. avermitilis, three structurally unrelated antibiotics, streptomycin, cephamycin C, and pladienolide, were produced in highest yield in a medium optimized for avermectin production. This may be a second general advantage of industrial production strains which have undergone both strain improvement and medium optimization during the course of scale-up studies. Finally, as observed in many industrial fermentations, optimum yields are ultimately obtained in large-scale fermentation vessels with continuous or batch feeds of carbon and nitrogen sources (e.g., glucose and ammonium), pH control, etc. This is a third advantage of applying highly productive industrial strains, which have undergone extensive large-scale fermentation optimization, as hosts for the expression of cryptic secondary metabolites.

In search of the “dream team” of actinomycete expression hosts

Although two hosts for high-level production of polyketides have been reported, much higher producing strains are used in industrial fermentations. In many cases, industrial fermentations yield more than 10 g/l of product. Many of these highly productive strains are proprietary in large pharmaceutical companies, and held as trade secrets. These same companies have witnessed the demise of their own programs in natural products discovery in recent years, including the very successful Eli Lilly and Company. In earlier years, Lilly was very open to sharing strains, plasmids, and gene clones, and this facilitated academic research on glycopeptide and macrolide biosynthesis. The “Lilly plasmids”, including the commonly used pSET152 [24], have become part of the repertoire of molecular genetic tools exploiting conjugation and site-specific integration that have helped progress the molecular genetic manipulation of actinomycetes. For rapid progress on the exploitation of actinomycete genome sequences to identify and express the products of novel secondary metabolite biosynthetic pathways in sufficient yields for robust discovery and development, it would be very useful if the very high producing “dream team” industrial strains were made available as hosts for heterologous expression. This could help revitalize natural products discovery and development.

Finally, a suite of strains suitable for heterologous expression of secondary metabolite biosynthesis to discover novel secondary metabolites will also be useful for the further development of combinatorial biosynthesis [8, 9], another technology that can help revitalize natural products discovery and development [11, 12]. We currently have a number of relatively poor to average strains to work with, and it might be advisable to pick four or five strains, including some laboratory strains and some industrial strains as hosts for heterologous expression experiments. Perhaps over time, the “dream team” strains will become available for exploitation. In the meantime, the funding agencies should be encouraged to expand genome sequencing initiatives directed at a broad range of actinomycete genera. This will undoubtedly help drive the renaissance of natural products discovery.