Biosynthetic engineering of polyene macrolides towards generation of improved antifungal and antiparasitic agents

Polyene macrolides are potent antifungal agents that are also active against parasites, enveloped viruses and prion diseases. They are medically important as antifungal antibiotics but their therapeutic use is limited by serious side effects. In recent years there has been considerable progress in genetic analysis and manipulation of the streptomycetes that produce nystatin, amphotericin B, candicidin, pimaricin and rimocidin/CE-108-related polyenes. This has led to engineered biosynthesis of several new polyenes that are not easily obtained as semisynthetic derivatives. This review summarises recent advances made since the subject was last reviewed in 2003. Polyene biosynthesis generally involves assembly and cyclisation of a polyketide chain, followed by oxidative modifications and glycosylation of the macrolactone ring. New derivatives have been obtained by engineering both early and late stages of polyene biosynthetic pathways. These compounds have allowed more detailed investigations of structure-activity relationships and some are likely to show improvements in therapeutic index. The biosynthetic approach is already yielding sufficient material for testing the toxicity and activity of new compounds, thus opening possibilities for discovery of leads for development of effective and safe antifungal and antiparasitic agents.


INTRODUCTION
Polyene macrolides are antifungal agents that are abundantly produced by streptomycetes and some other soil bacteria [1].Polyenes generally consist of macrolactone cores with between three and eight conjugated double bonds.The macrolactone also incorporates a six-membered hemiketal ring with a methyl branch that is oxidized to a carboxyl group.Typically there is a single aminodeoxyhexose sugar residue, mycosamine, attached to the ring through a glycosidic bond [1].A long-standing model for the mode of action proposes that polyene and sterol molecules interact to form a transmembrane channel that pierces the wall of the target cell [2].Most polyenes are selectively toxic towards fungal cells because they have a high affinity for ergosterol, the predominant sterol in fungal membranes [3].Polyenes interact to a lesser extent with cholesterol in mammalian membranes.These interactions are thought to cause toxic side effects when polyenes are used as antibiotics.
Transmembrane channel formation may not account for all of the biological activities of polyenes, however.The conjugated double bonds in polyenes are susceptible to auto-oxidation and form radical intermediates that may damage cells [4].
In addition to being efficient antifungal antibiotics, polyenes are also active against enveloped viruses, parasites and prion diseases [5].
Amphotericin B (AmB) (2, Fig. 1) has been the leading antifungal drug for many years, and is still being used in clinical practice [6].Several advantages of this drug attracted the attention of many laboratories in attempts for developing new derivatives: very potent antifungal activity, wide spectrum, and extremely low incidence of resistance development.
On the other hand, there are several undesirable side effects associated with administration of AmB, such as nephrotoxicity.Due to these side effects, the clinical use of AmB is limited to those cases in which other therapies have failed.The toxicity of polyene antibiotics can be reduced by liposomal formulation or by chemical modification.New lipid formulations with much lower toxicity and better pharmacokinetic profile have been introduced for clinical use [6], nevertheless, these formulations remain very expensive for regular treatments.Some semi-synthetic derivatives (see below) have very low toxicity but their production is not economically favourable.
The continuous need for new antibiotics has intensified interest in genetic manipulation of natural product biosynthesis in streptomycetes.After the discovery of the modular nature of complex polyketide synthases (PKSs) [7], several groups initiated genetic studies on polyene producers with a view to generating improved polyene-based drugs by engineered biosynthesis.The initial work on the biosynthetic gene clusters for nystatin (1), amphotericin (2, 3), pimaricin (4) and candicidin (5) (Fig. 1) was reviewed in 2003 [8].Since then, Deng and co-workers have characterised the complete biosynthetic gene cluster for the aromatic heptaene FR008 and showed that it is identical to candicidin [9].Malpartida's group characterized the cluster for the related tetraenes rimocidin (6) and CE108 (7) (Fig. 1) [10].A biosynthetic gene cluster for the pentaene filipin was identified in the genome sequence of Streptomyces avermitilis [11].Specialized reviews on pimaricin and nystatin biosynthesis have been contributed by Aparicio et al. [12] and by Fjaervik and Zotchev [13].In general, polyene biosynthetic gene clusters encode modular PKS proteins that assemble the macrolactone cores, cytochrome P450 enzymes that catalyse exocyclic carboxyl group formation and hydroxylation/epoxidation, and enzymes for biosynthesis and attachment of mycosamine.Genes involved in export and regulation have also been identified.Since the initial characterization of these clusters, there have been significant advances in the analysis of the functions of these genes.Engineered biosynthesis has yielded several new polyene analogues that would not have been easily obtained by chemical synthesis.This review summarises recent progress with each of these polyenes.To assist the reader, the functions of the various polyene biosynthetic genes are summarized in Table 1. Figure 2 gives an overview of the biosynthesis of a typical polyene, pimaricin.

GENERATING NOVEL POLYENES BY ENGINEERED BIOSYNTHESIS
Over the past 40 years, chemists have undertaken chemical modification of polyene antibiotics with the aim of reducing toxicity, and most of such work has focused on AmB.
These studies have shown that a positive charge on the amino group of mycosamine is important for antifungal activity, whereas suppression of charge on the exocyclic carboxyl group reduces toxicity [14].The AmB analogues N-methyl-N-D-fructosyl amphotericin B methyl ester (MFAME) (8) [15] and MS8209 (9) [16] (Fig. 3) show dramatic improvements in terms of toxicity, but are extremely expensive to produce.Fig. 3. Semi-synthetic derivatives of Am B AmB-arabinogalactan conjugates have improved water-solubility and reduced toxicity, while retaining antifungal activity [17].
Carreira and co-workers have synthesized bis(aminopropylene) AmB derivatives (10) with reduced toxicity and increased antifungal activity, as well as conjugates that are yielding important insights into the precise mode of action of polyene antibiotics [18].These studies show that structural modification of polyene antibiotics can significantly improve their pharmacological properties.However, it has so far not been possible to develop a semi-synthetic polyene that is economically viable.
Although total synthesis of AmB has been achieved by two research groups [19,20] most chemical manipulations use the natural product isolated from cultures of the producer organism, Streptomyces nodosus, as starting material.Engineered biosynthesis can generate analogues that are impossible to make by chemical modification, and it allows these compounds to be obtained as primary fermentation products.Previous reviews have discussed in detail the genes for polyene PKSs, the genes involved in biosynthesis of mycosamine from GDP-mannose, and genes for formation of exocyclic carboxyl groups and polyene hydroxylation [9,12,13].Generally, engineered biosynthesis involves manipulation of cloned polyene biosynthetic genes, and introduction of the modified DNA into the producing organism to allow gene replacement by homologous recombination with the chromosome.
Phage transduction, conjugation and protoplast transformation are standard methods for introduction of DNA into streptomycetes.These methods have been applied successfully to polyene producers, although progress has been delayed because these methods are timeconsuming.However, once a genetically modified streptomycete has been obtained, it is a sustainable source of a novel compound.Heterologous expression and "recombineering" methods have been applied extensively to smaller biosynthetic gene clusters [21] and could be useful for exploiting polyene biosynthesis more fully.
Although a number of recombinant bacteria have been designed to synthesise potentially useful new analogues, in many cases the yields of the products were greatly diminished.Further work on regulation of polyene biosynthesis should lead to yield improvements.Studies on regulatory genes are currently most advanced with the pimaricin and nystatin clusters.
The remaining sections describe recent progress in engineered biosynthesis of derivatives of nystatin, amphotericin B, FR008/candicidin, pimaricin and rimocidin and CE-108.
This mutant also produced nystatin 1068 (17), the heptaene analogue of nystatin 1070 ( 14), the 10-deoxy analogue of 1068, and octaenes with molecular weights of 950 Da and 1094.6 Da [23].The synthesis of the two latter analogues indicates that the NysC PKS protein with eliminated ER domain is capable of stuttering, a process in which a single module carries out two cycles of chain extension rather than one [24].
The ERD44 mutant has a sizeable deletion in the ER5 coding sequence [23].The yields of the S44HP heptaene analogue were quite low, possibly because the large deletion had a negative effect on the PKS overall structure.Site-directed mutagenesis and gene replacement were used to introduce single amino acid residue changes into the ER5 domain [25].Gly-5073 and Gly5074 of the NADPH-binding site were mutated to Ser and Pro, respectively.Identical changes have earlier been shown to efficiently inactivate ER4 domain of the erythromycin PKS [26].The resulting GG→SP mutant of S. noursei produced the S44HP analogue at a relatively high level.
Chemical mutagenesis of S. nodosus, which normally produces a mixture of tetraene AmA (3) and heptaene AmB (2) (Fig. 1), resulted in a mutant that produced AmB exclusively [25].The amphotericin PKS ER5 coding region from this mutant was amplified by PCR, sequenced, and found to contain a mutation that resulted in substitution of Ser4967 with Asn.
A similar Ser→Asn substitution was introduced into the ER5 domain of the nystatin PKS, and the resulting mutant was shown to produce a mixture of heptaene (16) and tetraene (1) nystatins [25].
When S. nodosus is grown under conditions of glucose limitation, the proportion of heptaene increases from 37% to 63% of the total polyene produced.These conditions reduce the flow of glucose through the pentose phosphate pathway, thus limiting the availability of NADPH [25].This suggests that ER5 of the amphotericin PKS has a low affinity for NADPH and that a low intracellular level of NADPH effectively inactivates the domain.

Mutagenesis of the loading module of the nystatin PKS
The loading module for the nystatin PKS has the domain composition KS 0 -AT 0 -DH 0 -ACP 0 .The amino acid sequence of the AT 0 domain suggested that it is specific for malonyl CoA.KS 0 was suggested to have a decarboxylase activity because it has a Ser residue at position 170 instead of the active site Cys residue found in C-C bond-forming domains [27].
Site-directed mutagenesis and domain exchange were used to investigate the functions of the loading module in vivo [28].Changing Ser170 of KS 0 to Cys, Gln or Gly had no effect on nystatin biosynthesis whereas replacement of the malonyl specific AT 0 domain with the methylmalonyl-specific AT 1 resulted in no nystatin production.The malonyl-specific AT 3 domain could substitute for AT 0 , indicating that malonyl groups can be loaded and decarboxylated to provide acetyl starter units.A fortuitous Ser413→Asn mutation in the KS 0 domain was found to abolish nystatin biosynthesis in a strain containing a KS 0 -AT 3 -DH 0 -ACP 0 loading module.This indicates that a functional decarboxylase is essential when NysA loads malonyl groups.The Ser413→Asn mutation, however, did not abolish nystatin production in the wild-type NysA (KS 0 S413N-AT 0 -DH 0 -ACP 0 ).This indicates that the decarboxylase activity is not absolutely essential for nystatin biosynthesis in a wild type strain and that AT 0 is capable of loading acetyl groups.These results indicate that the AT 0 domains of polyene PKS loading modules can use either acetyl CoA or malonyl CoA as a source of starter acyl units.
Extensive modification of the AT 0 domain of NysA by means of both site-specific mutagenesis and domain swapping aimed at changing the specificity of NysA towards methylmalonyl CoA has been attempted (Zotchev and Malpartida, unpublished results).However, none of the mutants was able to prime the nystatin biosynthesis with the propionyl starter, suggesting that the first extender module of nystatin PKS has strict specificity for the acetyl starter.
Zotchev and Lancelin suggest that a prerequisite for antifungal activity is that the polyene and polyol chains are of approximately equivalent length.Some reduction in activity was also observed with a hexaene derivative of nystatin (20), obtained by deletion of a module from the NysC PKS protein [23].

Analysis of the late steps in nystatin biosynthesis
Inactivation of nysL gene encoding P450 monooxygenase resulted in formation of 10deoxynystatin A 1 (12) [30].Interestingly, the same compound was produced when the nysG and nysH ABC transporters were inactivated in S. noursei [31].This indicates that C-10 hydroxylation is coupled to export through NysG-NysH.The Zotchev group succeeded in overproducing the NysL P450 in E. coli.The recombinant protein was obtained in a soluble form, and its C-10 hydroxylating activity was demonstrated in an in vitro system with 10deoxynystatin as a substrate [30].

AMPHOTERICIN
The first amphotericin analogue produced by engineered biosynthesis was 8deoxyamphotericin B (21) (Fig. 5), generated through insertional inactivation of the amphL cytochrome P450 gene [33].Inactivation of amphDIII for GDP-mannose 4, 6 dehydratase resulted in formation of the aglycone 8-deoxyamphoteronolide B (22) (Fig. 5) [32].The identification of these products indicated that the exocyclic carboxyl group formation can precede glycosylation and C-8 hydroxylation.The production of 8-deoxy amphotericin B indicates that glycosylation can proceed in the absence of C-8 hydroxylation.
The amphDIII mutant produced several different polyene aglycones.Analysis of minor products by NMR spectroscopy indicated that the amphotericin PKS makes stereochemical mistakes during cycles one and two of polyketide chain assembly [32].
Most polyenes undergo a unique late modification that consists of oxidation of a methyl branch to a carboxyl group.The amphN gene was targeted because amphotericins lacking exocyclic carboxyl groups were predicted to be analogous to AmB methyl ester, a less toxic semi-synthetic analogue.Numerous initial attempts to inactivate the amphN gene were unsuccessful.The amphNM region was eventually deleted from the chromosome of S. nodosus, and the resulting mutant was shown to produce the heptaene 16-descarboxyl-16-methyl amphotericin B (23) and the tetraene 8-deoxy-16-descarboxyl-16-methyl amphotericin A (24) (Fig. 5) [33].This indicated that glycosylation with mycosamine proceeded normally even though the methyl branch attached to C-16 was not oxidized to a carboxyl group, implying that the AmphDI glycosyl transferase has some flexibility towards its aglycone substrate.The C-8 hydroxylation appears to proceed normally with the heptaene but does not occur with the tetraene.It was proposed that the AmphM ferredoxin normally co-operates with the C-8 hydroxylase AmphL.In the absence of AmphM, AmphL may interact with a different ferredoxin to form a complex that recognizes the heptaene but not the tetraene [33].25) [33].The sugar is expected to be rhamnose, and NMR analysis is currently in progress to confirm this (B.Rawlings, personal communication).An amphDII mutant would be expected to synthesise a ketosugar, but S. nodosus apparently has a ketoreductase that reduces the ketone group to an alcohol.These results suggest that AmphDI glycosyltransferase has some sugar-flexibility as well.This analogue was shown to have low antifungal activity.
The first successful manipulation of the amphotericin PKS was the deletion of two modules from the AmphC protein, resulting in a mutant producing a pentaene (26) in good yield (Fig. 5) [34].The latter pentaene was water-soluble, but had no antifungal activity.
Recent ES-MS data indicate that while the exocyclic carboxyl group is formed on the pentaene, it is neither glycosylated nor C-8 hydroxylated (P.Caffrey, unpublished results).
The toxicity of amphotericin analogues was assessed by means of haemolytic assays.
The 16-methyl-16-descarboxyl analogues 23, 24 and 25 (Fig. 5) had lower haemolytic activities than AmB, while 23 and 24 retained antifungal activity.Thus, the properties of these new compounds appear to be comparable to those of AmB methyl ester.

FR008 / CANDICIDIN
Deng and co-workers have investigated biosynthesis of the aromatic heptaene FR008, produced by Streptomyces sp.FR008 [9].The organism produces a complex of four polyenes, FR008-I ( 27), FR008-II (28), FR008-III (29) and FR008-IV (30) (Fig. 6).Of these, only FR008-III had a hemiketal ring.FR008-I has a hydroxyl group at C-3 whereas the remaining three FR008-II, FR008-III and FR008-IV have a ketone group, indicating that the ketoreduction step may be omitted in the last cycle of polyketide chain extension during biosynthesis of the latter polyene.FR008-IV differs from the other three in having a methylene group at C-9 rather than a hydroxyl group, suggesting that both dehydration and enoyl reduction occur in the extension cycle 18.The FR008 cluster has been completely sequenced.The four FR008 polyenes were shown to be identical to the candicidin complex produced by Streptomyces griseus IMRU3570.The candicidin biosynthetic gene cluster from S. griseus was partially characterised by Campelo and Gil [35,36].The characterised candicidin biosynthetic genes showed between 93 and 100 % sequence identity with their counterparts in the FR008 cluster, clearly indicating their common origin.Deng's group found that deletion of the glycosyl transferase gene from Streptomyces FR008 resulted in production of the aglcyones of FR008 polyenes I to IV [9]. Figure 6 shows one example -FR008 VII (31) is the aglycone of FR008-III (29).Disruption of the aminotransferase gene involved in mycosamine biosynthesis resulted in production of mostly aglycones and small amounts of FR008 analogues glycosylated with a ketosugar.FR008-XI (32) is the ketosugar analogue of FR008-III (29).The low levels of these compounds indicates that polyene glycosyl transferases have a low affinity for GDP-ketosugar intermediates.Some aglycones were also detected in which the C-18 methyl branch was not oxidized to a carboxyl group.

PIMARICIN
The pimaricin analogue 4, 5-de-epoxypimaricin (33) (Fig. 7), obtained by inactivation of the pimD gene in Streptomyces natalensis [37], was the first biosynthetically engineered polyene macrolide produced.Biosynthesis of this compound at high yield indicated that the formation of the epoxide group at C4-C5 is the last step in pimaricin biosynthesis.The pimD gene was overexpressed in Escherichia coli to obtain a functional recombinant protein with a hexahistidine affinity tag.This enabled its purification and kinetic analysis and the in vitro C4-C5 epoxidation of 4, 5-deepoxypimaricin into pimaricin [38].
The inactivation of the pimK glycosyltransferase gene in S. natalensis resulted in the formation of the aglycone 4, 5-de-epoxypimaricinolide (34) (Fig. 7) as a major product, indicating that epoxidation is impaired in the absence of glycosylation with mycosamine and suggesting that the PimD epoxidase has a strong preference for the glycosylated polyene substrate.This aglycone product turned out to be extremely unstable and had no antifungal activity (J.F. Aparicio, unpublished results).This shows that the mycosamine sugar of pimaricin is essential for antifungal activity.This could not have been predicted because some smaller polyenes like filipin and fungichromin lack sugar moieties but display strong antifungal properties.The lack of an efficient gene delivery system for the pimaricin producing strain S.
natalensis has hampered the development of novel pimaricin derivatives.The recent development of a gene transfer system based on intergeneric conjugation from E. coli [39] will enable easier manipulation of the genes responsible for pimaricin biosynthesis, and could prove valuable for the generation of new designer polyene macrolides with better antifungal activity and pharmacological properties.

RIMOCIDIN / CE108
108B (37) and rimodicin B (36) (Fig. 8).The producer of polyene amide derivatives was unexpectedly generated when the parental strain was transformed with an SCP2*-based vector carrying the erythromycin resistance gene (ermE) for selection.There is no obvious explanation for the activation of this tailoring activity.A naturally-occurring strain catalyzing a similar chemical modification has also been isolated [42,43].This strain produces the tetraene pimaricin and its amide derivative AB-400 (35) (Fig. 7).The rimocidin amides had several times the antifungal activity of their counterparts with exocyclic carboxyl groups, while their haemolytic activities were not increased [43].The amide derivative (37) of CE-108 ( 7) is particularly interesting.While CE-108 ( 7) is weakly antifungal, its amide derivative CE-108B (37) has antifungal activity comparable to that of rimocidin (6).CE-108B (37) has a much lower haemolytic activity than rimodicin (6) and its amide (36).Thus, the amidated polyenes might be promising compounds for development of safer antifungal drugs.Similarly, the amide form of pimaricin (AB-400) (35) (Fig. 7), obtained from Streptomyces sp.RGU5.3, was a more potent antifungal drug than the parent compound [43], while no significant differences in their haemolytic properties were detected.All these data suggest that modification of the exocyclic carboxy groups of polyene macrolides by amidation improves selective toxicity of these compounds towards fungal cells.This is also consistent with the data on semisynthetic polyene macrolide derivatives altered in exocyclic carboxyl groups [44,45,46].The possibility of generating polyene amide derivatives in vivo is attractive because less toxic and highly active new polyenes could be generated by fermentation procedures.
Neither rimocidin nor CE-108 aglycones were converted to carboxamides in glycosyl transferase-deficient rimE mutants of S. diastaticus [43].This shows that carboxamide group formation occurs after glycosylation; thus, the substrates for the amide formation are the fully decorated polyenes.
Interestingly, the tailoring amidotransferase from S. diastaticus var.108 can catalyse not only conversion of rimocidin (6) and CE-108 (7) to amide forms (36,37) but also conversion of pimaricin to its amide form AB-400 (35) [43].The putative amidotransferase from S. diastaticus var.108 has a broader substrate range than that of the pimaricin amide producer Streptomyces sp.RGU5.3.However, conversion of AmB to its amide form could not be demonstrated.
Work is in progress to identify the gene for the amidotransferase.The substrate range could then be increased by protein engineering.
The rimG cytochrome P450 monoxygenase gene is an additional post-PKS decorating activity in rimocidin biosynthesis.Inactivation of rimG yielded exclusive production of decarboxy-methyl polyenes, even in those conditions known to be adequate for formation of amide derivatives [43].These results confirmed that the carboxamide group originates from modification of an exocyclic carboxyl group, and not from incorporation of a malonamyl extender unit during macrolide ring assembly.The latter mechanism was suggested for the biosynthesis of oxytetracycline [47], which is also made by S. diastaticus.The noncarboxylated rimocidin C (38) and CE-108C (39) had lower haemolytic activities than the corresponding parental polyenes, with similar antifungal activity.These new compounds, as well as decarboxy-methyl polyenes obtained for amphotericin B, nystatin and pimaricin by genetic engineering may well be considered as promising new antifungal drug leads.

USE OF NEW POLYENES AS ANTI-PARASITIC DRUGS
Fungi and several parasites such as Leishmania and Trypanosoma have ergosterolderived compounds as components of their membranes, making antifungal drugs targeting such membranes also potentially useful as antiparasitic agents [6].Despite its toxicity, AmB and its semisynthetic derivatives have been used successfully against several parasitic infections such as leishmaniasis and trypanosomiasis [48,49,50].Particularly promising were the new lipid formulations of AmB for clinical use against severe parasitosis; but the high cost of such formulations limits their use.The alternative drugs currently in use are being compromised because of the appearance of resistant strains, substantiating the need for new antiparasitic agents.

REGULATION
The nystatin biosynthetic gene cluster was first shown to contain the nysRI, nysRII and nysRIII genes for LAL (large ATP-binding regulators of the LuxR family) proteins [28].LAL proteins have an N-terminal ATP/GTP-binding domain and a C-terminal DNA-binding domain [52].Another regulatory protein NysRIV has an N-terminal PAS domain and a C-terminal DNA-binding domain [28].PAS domains monitor changes in light, redox potential, oxygen, overall energy level of a cell, and small ligands [53].PAS domains are located in the cytosol and detect internal signals, but they can also sense environmental factors that cross the cell membrane.Homologues of NysRI, RII, RIII and RIV are present in the amphotericin and candicidin/FR-008 clusters [35,36].
The pimaricin cluster contains two major regulatory proteins.PimR is the archetype of a new class of regulators [54].The C-terminal half is homologous to LAL proteins and the Nterminal half has an OmpR-like DNA-binding domain found in many Streptomyces antibiotic regulatory proteins (SARPs) [55].PimM, the second regulator of pimaricin biosynthesis, has been recently characterized [56].It is a regulator of the LuxR type with a PAS sensory domain.
The majority of prokaryotic PAS domains function as sensor modules of sensor kinases of two-component systems [53] but this is not the case with PimM.

Functional analysis of nystatin regulatory genes
The functions of the nystatin regulatory genes have been investigated in detail [57,14].
The related nysRI, nysRII, nysRIII genes appear to form an operon whereas the nysRIV gene is transcribed separately.An in-frame deletion was introduced into each gene so that inactivation of an upstream gene would not have a polar effect on downstream genes.Inactivation of any one of the nysRI, nysRII, nysRIII or nysRIV genes almost completely abolished nystatin production.This showed that the NysR proteins function as transcriptional activators.
Complementation experiments were also carried out where each of the nysR genes was placed under the control of a strong ermE promoter and re-introduced into the various nysR mutants.
Overproduced NysRII or NysRIII complemented the nysRI mutation.Overexpression of nysRIV complemented deficiencies in any of the four nysR genes.This indicates that NysRIV acts at a late stage in a regulatory cascade, probably as a direct transcriptional activator of biosynthetic genes.
Promoters from the nystatin cluster were also investigated [57].The biosynthetic genes appear to be organized into at least seven transcription units.The seven likely promoter regions were fused to xylE reporter gene sequences and promoter activity was assessed in wildtype S. noursei and in the nysR mutants.The promoter region for the first three PKS genes was strong but was absolutely dependent on all four nysR genes.The promoter for the last three PKS genes was weaker but had slightly reduced activity in all four nysR mutants, indicating that expression of these genes is less dependent on the activators.Transcription from the promoter for the export genes required functional NysRII, NysRIII and NysRIV proteins whereas two promoters for glycosylation genes were constitutive.The nysRI, nysRII and nysRIII genes are transcribed from a single promoter that is autoregulated by NysRI and NysRII.The nysRIV gene has a separate promoter that is activated by all three NysRI, NysRII and NysRIII proteins.
The overall picture seems to be that NysRI responds to external signals and activates transcription of nysRI-RII-RIII.The protein products activate transcription of nysRIV.The NysRIV protein, to varying degrees, activates promoters for the different groups of nystatin biosynthetic and export genes.The PAS domain of NysRIV suggests that its activity is further influenced by energy levels within the cell.The high degree of homology between the nystatin, amphotericin and candicidin/FR008 regulatory genes suggests that regulation of amphotericin biosynthesis is similar [57,35,36].

Functional analysis of pimaricin regulatory genes
Gene inactivation studies showed that PimR is a positive regulator of pimaricin biosynthesis [54].Gene expression analysis by reverse transcriptase-polymerase chain reaction (RT-PCR) revealed that S. natalensis ∆pimR shows little or no transcription of the pim genes except for the mutated pimR gene.This demonstrated that this regulator activates the transcription of all the genes of the pimaricin cluster but not its own transcription [54].
PimM is an ortholog of NysRIV [57], and its inactivation leads to impairment of pimaricin biosynthesis, suggesting that it constitutes a second activator.Gene expression analyses of the pimaricin gene cluster revealed that the genes responsible for initiation and first elongation cycles of polyketide chain extension (pimS0 and pimS1) are among the main targets for regulation, and also that PimM plays its regulatory role independently of PimR [56].
Preliminary results indicate that both PimM and NysRIV regulators might follow a similar regulatory pattern for the expression of their respective polyenes [56], thus suggesting that this pattern could be shared by homologous regulatory genes found in the amphotericin [35], candicidin [36] or filipin [12] clusters.

Indirect transcriptional regulation
The biosynthesis of polyenes in Streptomyces is traditionally considered to be very sensitive to phosphate regulation.Concentrations of inorganic phosphate above 1mM drastically reduced pimaricin production.Recently it has been demonstrated that this negative regulation is exerted at the transcriptional level in S. natalensis [58].At 10 mM phosphate, expression of all the pimaricin biosynthesis (pim) genes including the pathway-specific positive regulators pimR and pimM is fully repressed.Transcriptional control by phosphate of genes involved in antibiotic biosynthesis has been reported in a few other cases [59], including the polyene candicidin [60].
In S. coelicolor and S. lividans, the cellular response to phosphate limitation is controlled by the phoU-phoR-phoP genes that encode a two-component system [61].PhoR is a membrane-bound sensor kinase and PhoP is a DNA-binding response regulator that controls transcription of phoA (for alkaline phosphatase) and other target genes.PhoU is a modulator of the phosphate response.The phoU-phoR-phoP region of S. natalensis has been characterised [58].The PhoP protein from S. coelicolor binds to PHO box consensus sequences in the S. natalensis phoU-phoRP intergenic region, indicating that the system is autoregulated.Targeted gene disruptions yielded phoP and phoR-phoP deletion mutants of S.
natalensis.In both cases, production of pimaricin increased up to 80 % in complex media and both showed reduced sensitivity to phosphate control.Four of the pim genes, including the pimS1 and pimS4 PKS genes, showed increased expression in the phoP-disrupted mutant [58].
However, no consensus PHO boxes were found in the promoter regions of any of the pim genes, suggesting that phosphate control of these genes is mediated indirectly by PhoR and PhoP via modifications of other regulators.Manipulation of this regulatory mechanism will improve yields of other polyene macrolides.

Signal networks
In bacterial cultures there is a mechanism called "quorum sensing" that allows the culture to detect a high density of population and react by different mechanisms of adaptation [62].9) which elicits pimaricin production at nanomolar concentrations [63].The presence of similar quorum-sensing inducers in other polyene-producing strains has not been reported to date, but it seems conceivable that this kind of signal could play a fundamental role on the production of these compounds.This group has also discovered that glycerol, ethylene glycol, 1,2-propanediol or 1,3propanediol can elicit the production of the tetraene pimaricin in its producer strain S.
natalensis, and also increase the production of other polyenes by their respective producer strains, including S. noursei (nystatin), S. rimosus (rimocidin) or S. griseus (candicidin), among others [64].Although the precise mechanism by which glycerol elicits polyene production remains to be established, the action of glycerol seems to be independent of the PI-factor inducing effect [64].

Cholesterol oxidases as new modulators of small-ring polyene macrolide biosynthesis
The pimaricin cluster includes the pimE gene for a cholesterol oxidase that had no immediately obvious role in biosynthesis of the antibiotic.Cholesterol oxidase (EC 1.1.3.6) is a flavoprotein that catalyzes both, the oxidation of cholesterol to 5-cholesten-3-one with the reduction of molecular oxygen to hydrogen peroxide, and the isomerization of the ∆ 5 -bond to yield 4-cholesten-3-one as the final product.This enzyme participates in the first step of a metabolic pathway for utilizing cholesterol (or other sterols with a 3-β-hydroxyl group) as a carbon and energy source for primary metabolism [65].Gene inactivation and complementation studies led to the surprising finding that PimE is essential for the biosynthesis of pimaricin.This extracellular enzyme was purified from S. natalensis culture broths to homogeneity, and shown to restore pimaricin production when added to the mutant culture broths or "resting cells".Other cholesterol oxidases also triggered pimaricin production suggesting that these enzymes could act as signaling proteins for polyene biosynthesis [66].
Cholesterol oxidase genes are present in two other known biosynthetic gene clusters of small-size polyenes.These are the pte gene cluster of S. avermitilis [12], and the gene cluster responsible for the biosynthesis of the 28-membered tetraenes rimocidin and CE-108 in S.
diastaticus var.108 [43].It is therefore plausible that the genes for cholesterol oxidases might have been acquired in the process of evolution by the polyene antibiotic gene clusters in order to provide a prompt response to the presence of fungi (whose cell envelopes contain ergosterol) in the environment via expression of the polyene antibiotic biosynthesis genes.This modulation of polyene macrolide biosynthesis by extracellular proteins is totally new in bacteria and will constitute a novel working area that will aid in understanding the molecular mechanisms of signal transduction.
So far, the presence of common regulatory circuits controlled by cholesterol oxidase enzymes cannot be extended to larger polyenes since no cholesterol oxidase encoding genes have been found in the gene clusters for the nystatin, amphotericin or candicidin [28,67,35,36,37,38].

CONCLUSIONS
Over the past four years, engineered biosynthesis has yielded over twenty new polyene derivatives.Some of these compounds appear to have reduced toxicity and to retain antifungal and antiparasitic activity.Further manipulation of PKS and polyene modification enzymes will allow production of libraries of new compounds and complete investigation of structureactivity relationships.Recent developments in genetic manipulation of streptomycetes should accelerate this work.The most promising new polyenes will be required in large quantities for testing as potential drug candidates.Considerable progress has been made in understanding the regulation of polyene biosynthesis.This knowledge will be invaluable for maximizing yields of new polyenes.
Some cells produce signals of intercellular communication that transmit the message of nutrient limitation or high density population to the remaining cells of the culture, which in turn respond to this stimulus by the transcription (or repression) of given genes.Although this mechanism is general, the chemical molecules that act as signals are different depending on the species.Streptomyces species use butyrolactones; Vibrio, Xanthomonas and other Gramnegative bacteria use homoserine-lactones, and Bacillus use oligopeptides.In S. natalensis