Introduction

Mupirocin is a polyketide antibiotic produced by the soil bacterium Pseudomonas fluorescens NCIMB 10586, isolated from Hampstead Heath in London (LGC Standards, 2009). Pseudomonas spp. are Gram-negative rods found in soil and water as well as on surfaces of plant roots and leaves. Pseudomonas fluorescens produces pyoverdine during iron starvation, a green pigment that chelates iron and fluoresces under ultraviolet light. Pseudomonas fluorescens can grow at 4 °C and degrades lipids and proteins and has been implicated in the spoilage of refrigerated milk, meat, eggs and seafood (Prescott et al. 2005). It uses oxygen as electron acceptor when available, but nitrate can be used under anaerobic conditions. Different strains produce a variety of chemicals that interfere with neighbouring organisms inhabiting the rhizosphere. Production of secondary metabolites is activated by signal molecules in a phenomenon known as quorum sensing (Lugtenberg and Bloemberg 2004).

Polyketides are secondary metabolites produced by bacteria, fungi and some plants that display a wide range of biological activities that are increasingly exploited as therapeutic tools. Type I and II polyketide synthases (PKSs) are closely related to fatty acid synthases (FASs) and are classified according to their protein architecture (Ridley et al. 2008). Type I PKSs consist of large multifunctional enzymes with domains joined covalently, for example the 6-deoxyerythronolide B synthase (DEBS) that makes the core of erythromycin (Cane 2010; Hill and Staunton 2010). Type II PKSs consist of a multifunctional PKS complex built from separate monofunctional proteins, for example the Streptomyces coelicolor actinorhodin synthase (Tang et al. 2004a) and type II FAS have a similar structure (Schujman and Mendoza 2008). For bacteria, Escherichia coli has become a model organism, providing valuable information about the processes that take place during fatty acid synthesis (Magnuson et al. 1993). The FAS II of E. coli consists of seven core polypeptides (encoded by the fab genes) that work together to catalyse fatty acid synthesis (Maier et al. 2008) and the manipulation of this pathway can lead to increased production of biodiesel (Liu et al. 2010b). However, there is great interest in the diversity of PKS systems since this provides a growing source of genetic building blocks for synthetic biology which aims to generate novel biologically active molecules.

Mupirocin biosynthesis

The first reported study of mupirocin was in 1971 (Fuller et al. 1971). Subsequent investigations revealed a mixture of four pseudomonic acids (A–D), collectively named mupirocin, with extensive antibacterial activity. The structure of mupirocin, shown in Fig. 1, comprises a monic acid (a heptaketide) containing a pyran ring, attached to 9-hydroxynonanoic acid (9-HN) via an ester linkage (Fuller et al. 1971; Chain and Mellows 1974, 1977; Alexander et al. 1978; Whatling et al. 1995). Mupirocin targets bacterial isoleucyl-tRNA synthase (IleS) competitively inhibiting the formation of Ile tRNA, ultimately blocking protein synthesis (Hughes and Mellows 1978).

Fig. 1
figure 1

Top the structure of mupirocin, showing the monic acid and 9-hydroxynonanoic acid moieties. Mupirocin is a mixture of four pseudomonic acids: PA-A (about 90%) R = H; PA-B (about 8%) R = OH; PA-C (<2%) R = H and C10/11 E-alkene; PA-D (<2%) R = H, C4′/5′ E-alkene. Bottom structure of thiomarinol A, showing the marinolic acid (equivalent to pseudomonic acid/mupirocin) and holomycin moieties

Full size image

The mup cluster

A >65-kb region of the chromosome involved in mupirocin biosynthesis was identified by transposon mutagenesis (Whatling et al. 1995) and subsequently analysed by DNA sequencing, gene knockouts and complementation studies (El-Sayed et al. 2003). The cluster can be conveniently split into two sections—the first mainly encoding three large multifunctional PKS proteins as well as the multifunctional protein with acyltransferase (AT) domains, and the other encoding two smaller multifunctional PKS genes as well as 29 individual “tailoring” genes. Table 1 shows all of the genes involved in mupirocin biosynthesis and their putative functions. The cluster is unusual in that the order of the genes does not match the order of biosynthetic steps. The multifunctional genes mmpD and mmpA together encode the first four and last two elongating modules respectively and one putative transfer/non-elongating/processing module. They comprise appropriate ketosynthase (KS), acyl carrier protein (ACP), ketoreductase (KR), dehydratase (DH) and methyltransferase (MT) functions for monic acid backbone synthesis while mmpC encodes two AT domains and a putative enoyl reductase (ER) domain. The mmpB gene encoding single KS, KR and DH domains and triple ACP domains is thought to be responsible for synthesis of the 9-HN moiety, but also encodes the only thioesterase (TE) suggesting that it controls the final steps of the pathway and release of products. Resistance to mupirocin is encoded within the cluster by MupM which shows significant similarity to other mupirocin-resistant IleS proteins (El-Sayed et al. 2003).

Table 1 Gene functions of the mupirocin biosynthetic cluster
Full size table

Monic acid biosynthesis

The heptaketide chain that comprises the backbone of monic acid is synthesised by six condensation reactions of acetate-derived units catalysed by MmpD (modules 1–4) and MmpA (modules 5–6) as shown in Fig. 2a (Martin and Simpson 1989; El-Sayed et al. 2003). Synthesis could begin with one of the trans-acting ATs transferring an activated starter unit (acetyl-coenzyme A intermediate) to the 4′-phosphopantetheine arm of ACP-D1 and then to the thiol group of the active cysteine of KS-D1. An activated extender unit (malonyl-CoA) could then be transferred by one of the ATs to the vacant ACP-D1, prior to decarboxylative (Claisen) condensation catalysed by KS-D1. Alternatively, an activated starter unit could be transferred directly to the KS-D1 (see below). The first module then carries out ketoreduction and α-methylation to create the structure that mimics the isoleucine side chain. Synthesis continues through three further rounds of condensation and modification on successive modules of MmpD (KS/ACP-D2, D3 and D4), including a second α-methylation by the MT in module 3, before the acyl intermediate is passed to MmpA. The first module of MmpA (module 5 in the pathway) is likely to act as a transfer or processing (non-elongating) module due to the combination of atypical KS (KS0) and ACP domains. Deletion (as well as mutation of active site residues) of these domains resulted in a loss of antibiotic production, indicating an important role in mupirocin production (El-Sayed et al. 2003). Two further elongation modules on MmpA extend the chain to give the C17 precursor to monic acid.

Fig. 2
figure 2

Proposed scheme for biosynthesis of mupirocin. a The monic acid (MA) backbone is synthesised by condensation of extender units and further modifications by the six modules of MmpD and MmpA. 9-HN is synthesised by the iterative condensations and further ketoreduction and dehydration catalysed by MmpB on a proposed 3-hydroxypropionate starter unit. There are two possible pathways for the addition of 9-HN to the MA backbone: continued extension onto MA that is already esterified to 3-HP (pathway 1, R = monic acid); or, separate synthesis followed by esterification (pathway 2, R = H). b The MA structure is modified by tailoring enzymes, such as MupW, MupT, MupO, MupU, MupV, MupC and MupF, which function to form the pyran ring and the epoxide group at C-10,-11. Whether this happens before or after joining to 9-HN is not known. The ACP holding the intermediate in panel b before it is transferred to mAcpE is shown as green/purple to indicate that we do not know whether this normally happens on ACP5/6/7 (green) or ACP3/4 (purple). Key: all functional domains are labeled except for “docking domains” (hatched black and white oval) that are remnants of in-cis AT domains that may be required for function. Introduction of new nomenclature for the mupirocin gene cluster means that the modules and domains of multifunctional proteins are now named according to their position within each protein. For example, module 3 of MmpA is termed “A3” and the ACPs that form a didomain on MmpA are now termed ACP-A3a and ACP-A3b (formerly ACPs 3 and 4)

Full size image

The mupirocin cluster contains many genes that are needed to modify (“tailor”) the PKS-bound intermediate before completion of the final pseudomonic acid structure. The modifications involve incorporation of a methyl group at C15, hydroxylation at C6, epoxidation at C10,11 and formation of the pyran ring (Fig. 2b) (Cooper et al. 2005a; Hothersall et al. 2007). Bioinformatics can often predict both biochemical function and partner genes which work together (based on being found together in other genomes). These include the hydroxymethylglutaryl-CoA synthase (HCS) cassette genes mAcpC, mupG, mupH, mupJ, mupK (for function see below) as well as two other blocks mAcpD, mupS, mupQ, mmpF (functional currently unknown) and mupD, mupE (an enoyl reductase). However, gene knockouts and product analysis are essential since other gene sets such as mupW, mupT (pyran ring formation) and mAcpE, mupO, mupU, mupV, mupC and mupF (further reduction around the pyran ring) seem to be functions currently unique to this cluster. While it is clear at what stage in the pathway some of them work, for example the HCS cassette that functions at module A3, for others it is either flexible or still unclear. Gene functions relating to C6-hydroxylation and C-10,-11 epoxidation are currently under investigation.

Gene knockouts have been particularly instrumental in indicating a possible pathway for pyran ring formation and pseudomonic acid A synthesis (Fig. 2b). In this scheme, MupW and MupT catalyse the epoxidation of the C-8,-16 double bond (essential for formation of the tetrahydropyran ring) which makes the C-16 more receptive to attack by the C-5 hydroxyl group. Esterification with 9-HN and C-10,-11 epoxidation, which may occur before or after this, result in pseudomonic acid B. Mutation of mupW and mupT resulted in accumulation of a novel metabolite, mupirocin W, lacking the tetrahydropyran ring but having the attached 9-HN confirming the role of MupW/MupT in mupirocin biosynthesis but not defining when it normally occurs (Cooper et al. 2005b). To produce pseudomonic acid A, MupU is proposed to mediate the transfer to mAcpE before MupO, a cytochrome P450 may catalyse oxidation of the C-7 hydroxyl to the ketone, while subsequent dehydration by MupV generates a C-8,-9 enoyl bond. MupC is proposed to reduce the C-8,-9 bond, before MupF catalyses ketoreduction at C-7 and the resultant product is released as pseudomonic acid A (Cooper et al. 2005a, b; Hothersall et al. 2007; Wu et al. 2007).

The HCS cassette, comprised of MupG, MupH, MupJ, MupK and mAcpC, is responsible for the incorporation of the β-methyl group at C-15: MupG catalyses the decarboxylation of acetate from a malonate bound to mAcpC; MupH catalyses the condensation to produce a gluconate intermediate; dehydration, catalysed by MupJ is finally followed by decarboxylation mediated by MupK to produce the β-methylthioester (Wu et al. 2007, 2008). The functions of related HCS cassettes have been proposed by a number of groups and experimental evidence supporting these hypotheses have been provided from studies on myxovirescin (Simunovic et al. 2006), jamaicamide (Edwards et al. 2004), leinamycin (Tang et al. 2004b), bacillaene (Butcher et al. 2007) and curacin A (Chang et al. 2004).

9-Hydroxynonanoic acid biosynthesis

Since there are strong indications for the role of MmpD and MmpA in monic acid (MA) synthesis it seems logical that MmpB, the third type I PKS, is responsible for 9-HN although as yet there is no direct evidence for this. However, 9-HN is proposed to be derived from a 3-hydroxypropionate (3-HP) starter unit with MmpB catalysing three rounds of condensation with malonate as the extender unit (Fig. 2a). MmpB does not contain an ER domain but an in-frame deletion of MupE results in a 6′–7′ enoyl bond suggesting that it, possibly in conjunction with MupD, is responsible for at least part of the required ER activity (Hothersall et al. 2007; Macioszek, PhD thesis, University of Birmingham, 2009; Hothersall and Wu, unpublished data). Since mutagenesis of mupE did not result in the formation of a completely unsaturated fatty acid chain, an additional enzyme must be responsible for reduction of the fatty acid chain to give 9-HN. There is a third domain in MmpC which has predicted ER activity, that could function during the formation of 9-HN. Recent work by Bumpus et al. (2008) demonstrated that PksE, an enzyme from the dihydrobacillaene PKS consisting of AT and ER domains, provided ER activity in-trans. The terminal TE domain of MmpB could then either catalyse the release of the saturated 9-HN or provide a means of esterification with monic acid (El-Sayed et al. 2003; Hothersall et al. 2007). The order in which events occur is still under investigation. It is possible either that 9-HN and monic acid are synthesised separately and then joined together to complete the mupirocin structure, or that 9-HN is elongated on a starter unit (3-HP) esterified with the product of MmpD/MmpA (J Hothersall, J Wu and A Murphy et al. unpublished data).

Mupirocin production is regulated by quorum sensing

A quorum sensing mechanism controls expression of the mupirocin biosynthetic genes. This involves the constitutive production of diffusible signal molecules (autoinducers, in this case N-acyl homoserine lactones) that accumulate in the environment of the bacteria and, when the population reaches a critical density (quorum), switch on target gene transcription via an activator protein. Based on sequence alignments with the Vibrio harveyi lux system, the genes mupR and mupI were predicted to mediate quorum-regulated expression of the mupirocin cluster (El-Sayed et al. 2001). MupI is required to produce the diffusible substance that activates the mup promoter, while MupR is the activator that responds to it (El-Sayed et al. 2001). Thus, the MupR–MupI system activates transcription of the mup operon upon binding to the lux box promoter regions of mupA, mAcpC and mupF, while surprisingly mupI does not appear to have a lux box region (Fuqua et al. 1994; El-Sayed et al. 2001). At present, it is unknown whether mupR requires transcription right through the mupirocin cluster for expression or whether mupI gene expression is boosted when the quorum switch is turned on.

Special features of the mupirocin cluster

The mupirocin cluster contains various features that distinguish it from typical PKSs such as the DEBS system and whose activities may be useful in the generation of novel pathways.

The mupirocin cluster depends on in-trans ATs

The mupirocin PKSs differ from those of the well studied erythromycin biosynthetic system from Saccharopolyspora erythraea (Lal et al. 2000) in that the AT domains are absent from each module and are encoded by a separate gene, mmpC (El-Sayed et al. 2003). They thus belong to the group of PKS systems termed in-trans AT PKSs, of which a growing number have been described and analysed including those that produce myxovirescin, virginiamycin, leinamycin, lankacidin, pederin, rhizoxin, bryostatin, kirromycin, mycosubtilin, bacillaene, difficidin, macrolactin, chivosazol, disorazol and thiomarinol (Piel 2002, 2010; Cheng et al. 2003; Chen et al. 2006; Kopp et al. 2005; Mochizuki et al. 2003; Perlova et al. 2006; Simunovic et al. 2006; Aron et al. 2007; Partida-Martinez and Hertweck 2007; Pulsawat et al. 2007; Schneider et al. 2007; Sudek et al. 2007; Weber et al. 2008). There are several key active site amino acids thought to be critical in determining AT specificity for its substrate, particularly the YASH and YAFH motifs, defining methylmalonyl-CoA and malonyl-CoA specificity respectively (DelVecchio et al. 2003). Both of the mupirocin ATs contain partial YAFH motifs, particularly the crucial FH residues, indicating malonyl-CoA specificity. Phylogenetic analysis of in-trans ATs from different systems as shown in Fig. 3 indicates two evolutionary pathways with the mupirocin ATs falling in separate clades. Sequence analysis of these clades based on methods described by Yadav et al. (2003) shows the AT2 clade to be specific for malonyl-CoA, while the AT1 clade displays far more active site residue diversity indicating a broader range of substrate specificity. It is possible that the AT2-like enzymes represent the “main” ATs within clusters, while the AT1-like enzymes provide increased turnover, or more specialized functions.

Fig. 3
figure 3

Phylogenetic tree showing a selection of acyltransferases that operate in-trans throughout their respective PKS systems. Escherichia coli FabD is included as the out-group for reference. Each AT is also accompanied by connected symbols that show the polypeptide context (architecture) in which that AT is found—alone or joined to a second AT or to an ER domain. Accession numbers: Bacillus subtilis ATCC 6633 (AAF08794), Streptomyces collinus Tu365 (AM746336), P. fluorescens NCIMB 10586 (AAM12912), Pseudoalteromonas rava SANK 73390 (FN689524), Myxococcus xanthus DK 1622 (YP_632122), Burkholderia rhizoxina (CAL69887), Candidatus endobugula sertula (ABK51299), symbiont of Paederus fuscipes (AAS47559 and AAS47563), Bacillus amyloliquefaciens FZB42 (YP_001421286, YP_001421027, and CAG23974), P. fluorescens BCCM_ID9359 (GU479979), Streptomyces rochei (AB088224), Streptomyces atroolivaceus (AAN85520), Streptomyces virginiae (BAF50719), B. subtilis 168 (NP_389591 and CAB13584), and Polyangium cellulosum (DQ065771 and AJ874112)

Full size image

AT “docking domains” have been identified in several in-trans AT PKSs, including mupirocin and thiomarinol, and are located immediately downstream of the KS domains on the multifunctional proteins. Tang et al. (2004b) proposed that these domains were remnants of once-functional AT domains, and may have a role to play in interactions between ATs and domains within multifunctional proteins. By contrast, Aron et al. (2007) proposed that it was more likely that they played a role in regulating acyl-group transfer. Further investigation is required to determine the functions of these domains and whether they do indeed play a role in metabolite biosynthesis.

The mupirocin cluster does not contain integrated ER domains

The trans-AT systems either completely lack ER domains or have them in unusual positions. For example, for the myxovirescin biosynthetic cluster, it has been proposed that the ER domain of TaO is shared between modules 7 and 8, encoded on Ta-1 and TaO, respectively (Simunovic et al. 2006). ER domains are also thought to be encoded by modules on discrete proteins such as LnmG and MmpC (Cheng et al. 2003; El-Sayed et al. 2003). Figure 3 includes the domain architecture of the AT-encoding genes. There are four distinct architectures: AT didomains with an ER domain, AT didomains without an ER domain, single AT domains with an ER domain, and finally stand alone ATs. These architectures are distributed among the two evolutionary pathways. Their functionality and relevance is yet to be determined.

MT domains and a HCS cassette are responsible for incorporation of methyl groups

In type I systems with in-cis AT domains each module can specify the nature of the extender unit, for example choosing methyl-malonate rather than malonate. When all modules share one or two in-trans ATs incorporation of α-methyl groups can be specified by MT domains in a module. These occur in the mup cluster and all other in-trans AT PKSs mentioned previously, with the exceptions of macrolatin and mycosubtilin. However, this is not obligatory because some systems employ a mixture of in-trans and in-cis ATs. In mupirocin, the MT domains in modules 1 and 3 are responsible for the methyl groups C-17 and C-16 in the final structure, which are incorporated from S-adenosyl methionine (El-Sayed et al. 2003; Feline et al. 1977; Wu et al. 2008). In addition, many of the in-trans AT group of PKS contain β-branches to the polyketide backbone that are incorporated under the actions of an HCS cassette (Chen et al. 2006; Cheng et al. 2003; Partida-Martinez and Hertweck 2007; Pulsawat et al. 2007; Simunovic et al. 2006; Sudek et al. 2007). As detailed in the previous section, the mupirocin HCS cassette catalyses the incorporation of the methyl group at C-15. Studies on mupirocin have also provided evidence of the functions of the HCS cassette—mutation of mupH produced a new metabolite, mupirocin H, which appeared to be a truncated version of monic acid incorporating a 3-hydroxy-γ-lactone ring (Wu et al. 2007).

The mupirocin cluster has tandemly duplicated acyl carrier proteins

The multifunctional proteins MmpA and MmpB contain tandem doublet and triplet ACPs respectively (El-Sayed et al. 2003). There are several other trans-AT PKSs with similar unusual domain architecture: leinamycin, lankacidin, bacillaene, difficidin, chivosazol, virginiamycin and macrolactin (Chen et al. 2006; Cheng et al. 2003; Mochizuki et al. 2003; Perlova et al. 2006; Pulsawat et al. 2007; Schneider et al. 2007). The ACPs of the doublet (ACPs-A3a and -A3b—formerly ACPs 3 and 4) and triplet (ACPs-Ba, -Bb and -Bc—formerly ACPs 5, 6 and 7) are more closely related to each other than any other ACPs within the cluster, indicating they may have arisen from gene duplication events. Another unusual feature of the tandem ACPs is the unusually short linker regions between the individual domains—the spacers between domains on the Mmps are usually approximately 100 amino acids in length, but ACPs-A3a and -A3b are separated by just 12 amino acids and ACPs-Ba, -Bb and -Bc by only 3. Rahman et al. (2005) produced various mutants to determine the roles of these tandem ACPs: mutants of ACPs-A3a and -A3b, and ACPs-Ba, -Bb and -Bc resulted in loss of mupirocin production when analysed by bioassay and HPLC. Pairwise mutants of ACPs-Ba/Bb, -Bb/Bc and -Ba/Bc reduced mupirocin production to less than 20% of wild type, while individual mutants of ACPs-A3a and -A3b, and ACPs-Ba, -Bb and -Bc resulted in approximately 60% and 25–36% of production when compared to wild type, respectively. The authors concluded that while any one ACP from the cognate doublet and triplet clusters is sufficient for mupirocin biosynthesis, production was significantly improved by an increase in numbers, indicating that the doublet and triplet set of ACPs are functionally redundant (Rahman et al. 2005). Reducing the tandem ACPs to a single ACP would be rate limiting. The results indicated that the doublet ACPs (A3a/A3b) work in parallel, while ACP-Ba physically blocks access of other ACPs to some part of the machinery if it is inactivated by a point mutation (Rahman et al. 2005).

The mupirocin cluster does not have a loading module for system initiation

The lack of an obvious loading module within the mupirocin cluster is unusual. Many PKS systems, of both cis- and trans-AT architecture, have loading modules specifically designed to accept the starter unit for initiation of metabolite production (Hertweck 2009). The model DEBS system contains a loading domain consisting of an AT and an ACP, thought to provide the propionyl-CoA starter unit to the first module (Hill and Staunton 2010). The loading modules of trans-AT PKSs vary in the domains that are present, from the non-ribosomal peptide synthase of leinamycin to the minimal ACP of chivosazol and virginiamycin and to the more intricate loading module of bryostatin which contains four domains (Cheng et al. 2003; Perlova et al. 2006; Pulsawat et al. 2007; Sudek et al. 2007). Some such as the disorazol and macrolactin systems are similar to the mupirocin system in not having obvious loading modules. In these systems it has been proposed that a malonate residue is loaded to the first KS domain and then decarboxylated to provide the required starter acetate unit (Kopp et al. 2005; Schneider et al. 2007). Such a mechanism might occur in the mupirocin system, with an as yet unidentified protein providing the decarboxylative function to generate the first acetate for module 1 of MmpD. Alternatively, KS-D1 may decarboxylate malonate to acetate before the first condensation. It is also possible that one of the uncharacterised type II ACPs accepts malonate and catalyses the decarboxylation before transfer to KS-D1. Determining the substrate specificities of the mupirocin ATs may shed some light on how the whole system is initiated. In-frame deletion of AT1 only reduced mupirocin production while a similar mutation in AT2 abolished it (El-Sayed et al. 2003; Hothersall, unpublished data). Current work has shown that AT2 exclusively prefers malonyl-CoA as a substrate over acetyl-, propionyl-, and methylmalonyl-CoA (Gurney et al. unpublished data). As yet it has not been possible to produce soluble AT1 and achieving this is essential to allow direct comparison of the biochemical properties of AT1 and AT2.

Clinical significance and applications

Mupirocin competitively inhibits IleS, blocking the formation of Ile tRNA and thus inhibiting protein synthesis (Hughes and Mellows 1978). IleS catalyses the transfer of isoleucine onto its tRNA via the formation of aminoacyl-adenylate (aa-AMP). The C-14 to C-11 terminus of monic acid resembles the side-chain structure of isoleucine and interacts with the isoleucine-specific binding pocket of IleS (Yanagisawa et al. 1994). The pyran ring interacts with the ATP binding site of IleS, and it is thought the 9-HN moiety may stabilise the binding by its affinity for a hydrophobic groove (Nakama et al. 2001). Mupirocin has a remarkably broad spectrum of activity; it is active against both Gram-positive and Gram-negative organisms, and particularly effective against those staphylococcal and streptococcal species most commonly responsible for infections of the skin (Sutherland et al. 1985). Mupirocin cannot be used systemically due to its high affinity for serum protein and rapid metabolism, but has been successfully used topically for many years (Basker et al. 1980; Sutherland et al. 1985).

Pseudomonic acid was tentatively used as a potential therapeutic agent for skin infections and nasal carriage of antibiotic-resistant strains of Staphylococcus aureus in 1983 (Wuite et al. 1983; Dacre et al. 1983). In 1985, it was introduced for the treatment of bacterial skin infections, and in 1988 for nasal carriage of Staphylococci, including methicillin-resistant S. aureus (MRSA) (Cookson et al. 1990). MRSA colonising the skin and nose can easily be transferred to other areas of the body or wounds, thus causing particular concern during surgical procedures and when cuts or burns are open (Neu 1992). Marketed globally, mupirocin is now used worldwide for topical treatment of impetigo, infected skin lesions and for decolonisation of patients with nasal carriage of Staphylococci, including MRSA (GlaxoSmithKlein 2010; Medimetriks 2008; TEVA 2003). Decolonisation can reduce the risk of MRSA infection in patients and decrease transmission to other patients (Gilpin et al. 2010).

Resistance

The initial widespread use of mupirocin to treat MRSA lead to resistance first being recorded in 1987 (Rahman et al. 1987). Resistance to mupirocin has two distinct levels: high-level resistance >500 μg/ml, and the more common low level between 8–256 μg/ml (Ramsey et al. 1996). Low level resistance normally arises from spontaneous mutations in the chromosomally encoded IleS which reduce affinity for mupirocin, but are non-transferable and generally of little clinical significance (Eltringham 1997; Slocombe and Perry 1991). High-level resistance has a more substantial effect on clinical treatments and involves a plasmid-encoded gene, mupA, responsible for the production of a novel IleS (Dyke et al. 1991; Eltringham 1997; Farmer et al. 1992; Gilbart et al. 1993; Hodgson et al. 1994; Rahman et al. 1987). The mupA gene is associated with transposable elements as part of different plasmids, often self-transmissible, that also confer resistance to other antibiotics, such as gentamicin, tetracycline and trimethoprim (Patel et al. 2009; Perez-Roth et al. 2010). A recent study to assess the situation in China, where mupirocin is used for treating skin infections, determined that 6.6% of isolates displayed high-level mupirocin resistance and carried the mupA gene. Of these isolates, over 80% were also resistant to gentamicin, erythromycin and clindamycin, in addition to observations of resistance to trimethoprim, rifampicin and levofloxacin (Liu et al. 2010a).

Comparison of sequences surrounding mupA genes from self-transmissible plasmids in S. aureus indicates apparently multiple gene capture events with varying amounts of the same flanking sequences as if the gene comes from the chromosome of an as yet unidentified organism (Perez-Roth et al. 2010). Phylogenetic analysis of IleS proteins put MupA as well as MupM and TmlM proteins (that confer high-level resistance to mupirocin and thiomarinol on the strains that produce these antibiotics) in a group of IleS proteins that have eukaryote-like sequence motifs in their active sites, presumably explaining their resistance (Yanagisawa and Kawakami 2003). Further analysis of the growing wealth of bacterial genome sequences also reveals that many bacteria carry two IleS proteins—the second (IleS2) belonging to this eukaryote-like mupirocin resistance type correlating with a mupirocin-resistant phenotype. The reasons for carriage of a second IleS are at present unclear, but it seems quite likely that it is one such IleS2 gene in an as yet unidentified bacterium that is the source of mupA in S. aureus.

Resistance has been reported globally at varying degrees of severity. A study in Kuwait which sampled 53 MRSA isolates (39 of which expressed high-level mupirocin resistance) reported a significant increase in the number of high-level mupirocin-resistant MRSA isolates between 1993 and 1995; over the total study period (1990–1995), 42% of the isolates demonstrated high-level resistance (Udo et al. 1999; Vasquez et al. 2000). Between 1994 and 1995 at two closely situated hospitals in Brazil, resistance to mupirocin was >50% in one and approximately 6% in the other—the difference being that mupirocin was used far more frequently in the first hospital (Orrett 2008). In areas where mupirocin is readily available, the occurrence of resistance is high: in New Zealand, mupirocin became available over the counter in 1991, and by 1999, up to 28% of S. aureus isolates were mupirocin resistant (Upton et al. 2003). After increased mupirocin use in Western Australia, high-level mupirocin resistance reached 15% but subsequent government-issued guidance on limiting use reduced these levels to 0.3% after 4 years (Torvaldsen et al. 1999). A recent study showed that high-level mupirocin resistance was detected in 17% of patients involved, and this lead to decolonisation failure (Gilpin et al. 2010). Among these studies, a recurring conclusion is evident—prudent use of mupirocin can reduce the levels of resistance.

Future perspectives

While much has been learned about the mupirocin biosynthetic cluster, particularly over the last ten years, there are still aspects of the pathway to be fully worked out. Understanding these aspects in conjunction with what we already know may allow these genes to be used in combinatorial genetics, as in the case of erythromycin (Reeves et al. 2001), rapamycin (Gregory et al. 2005), daptomycin (Nguyen et al. 2006), lovastatin (Xie et al. 2006) and pikromycin (Gupta et al. 2008). The thiomarinols have a chemical structure which is essentially a combination of two independent antibiotics, pseudomonic acid and holomycin highlight the potential of making hybrid antibiotics based on mupirocin (Fig. 1). With a broad range of activity against both Gram-positive and -negative bacteria, thiomarinol A displays far more potency than mupirocin with activity approximately 20-fold higher against organisms such as MRSA and E. coli (Shiozawa et al. 1993). In addition to increasing potency, genetic manipulation may also be used to overcome the high affinity of mupirocin for serum protein—its rapid metabolism means that it cannot be used systemically (Basker et al. 1980; Sutherland et al. 1985). While chemical modifications have been explored to try to resolve these issues, genetic modifications have not (Basker et al. 1980). It is possible that analogues of mupirocin and thiomarinols could generate compounds that are both more metabolically stable and more active (El-Sayed et al. 2001; Marion et al. 2009).

Since the discovery of penicillin by Alexander Fleming in 1928, the development of antibiotics has revolutionised the treatment of infectious diseases. Resistance to penicillin was observed a mere 3 years after clinical development by Pfizer in 1944, despite the warnings of Fleming himself (Fleming 1945). Between the 1940s and 1970s, many clinically important antibiotics were discovered and developed, but there have been few in recent years (Bush and Macielag 2000; Critchley and Ochsner 2008; Larkin 2003) so novel compounds and strategies are needed. Despite resistance to mupirocin continuing to be an issue (Caffrey et al. 2010; Fawley et al. 2006; Gilpin et al. 2010), the potential to produce novel derivatives may mean that it will continue to play a role in the future treatment of infectious diseases.