The complex extracellular biology of Streptomyces


  • Editor: Lee Kroos

Correspondence: Keith F. Chater, John Innes Centre, Norwich Research Park, Colney, Norwich NR4 7UH, UK. Tel.: +44 1603 450297; fax: +44 1603 450045; e-mail:


Streptomycetes, soil-dwelling mycelial bacteria that form sporulating aerial branches, have an exceptionally large number of predicted secreted proteins, including many exported via the twin-arginine transport system. Their use of noncatalytic substrate-binding proteins and hydrolytic enzymes to obtain soluble nutrients from carbohydrates such as chitin and cellulose enables them to interact with other organisms. Some of their numerous secreted proteases participate in developmentally significant extracellular cascades, regulated by inhibitors, which lead to cannibalization of the substrate mycelium biomass to support aerial growth and sporulation. They excrete many secondary metabolites, including important antibiotics. Some of these play roles in interactions with eukaryotes. Surprisingly, some antibiotic biosynthetic enzymes are extracellular. Antibiotic production is often regulated by extracellular signalling molecules, some of which also control morphological differentiation. Amphipathic proteins, assembled with the help of cellulose-like material, are required for both hyphal attachment to surfaces and aerial reproductive growth. Comparative genomic analysis suggests that the acquisition of genes for extracellular processes has played a huge part in speciation. The rare codon TTA, which is present in the key pleiotropic regulatory gene adpA and many pathway-specific regulatory genes for antibiotic production, has a particular influence on extracellular biology.


Bacteria of the genus Streptomyces are abundant in soil. They use an extraordinary variety of extracellular mechanisms for primary growth and to facilitate intimate associations with other organisms, as well as to achieve their well-known attributes of developmental complexity and secondary metabolism. In the 65 years since Waksman discovered streptomycin as the first therapeutically useful Streptomyces antibiotic, it has been found that streptomycetes synthesize an amazing variety of chemically distinct inhibitors of many different cellular processes. They include antibiotics, fungicides, cytostatics, modulators of the immune response, and effectors of plant growth (Ōmura, 1992; Hopwood, 2007). Genes for the synthesis of such secondary metabolites probably emerged up to hundreds of million years ago (reviewed by Baltz, 2006).

The most recent common ancestor of modern streptomycetes seems to have lived about 440 million years ago, when it was probably intimately involved in the colonization of the land by green plants (see Chater & Chandra, 2006). This view allocates these bacteria a formative role in the evolution of the soil environment. In this habitat, dessication-resistant Streptomyces spores are usually abundant, giving rise to branching hyphal filaments under suitable conditions. This growth adaptation assists in adhering to and penetrating the insoluble organic remains of fungi, plants and other organisms, which are then broken down by hydrolytic exoenzymes to provide nutrients. When grown for a few days in the laboratory, whether on soluble nutrients or on insoluble materials more closely resembling the natural condition, such colonies form a densely packed substrate mycelium, from which aerial branches emerge, eventually forming chains of spores (Flärdh & Buttner, 2009). Aerial growth is achieved at the expense of the substrate mycelial biomass, which undergoes extensive lysis and reuse to fuel the reproductive phase (Wildermuth, 1970; Miguélez et al., 1999). The few milligrams (wet weight) of the colony's substrate mycelium are thereby converted into tens of millions of spores. The production of antibiotics and other secondary metabolites is largely associated with this developmental pathway.

Two streptomycetes have been particularly well studied: Streptomyces griseus, the first streptomycete to be used for the commercial production of an antibiotic (streptomycin) (Horinouchi, 2007) and Streptomyces coelicolor A3(2), the most widely used laboratory strain (Hopwood, 2007), which we refer to throughout as S. coelicolor for convenience, although it should be remembered that this strain is properly Streptomyces violaceoruber (Kutzner, 1981). Both genomes have been sequenced, as has that of the important industrial organism Streptomyces avermitilis (Bentley et al., 2002; Ikeda et al., 2003; Ohnishi et al., 2008) (others have been sequenced, but not yet fully analysed or published). The 8–10-Mb linear chromosomes comprise >7000 genes, about 45% of which are common to the three genomes: these are mostly confined to a 6.4-Mb central segment (Ventura et al., 2007). The S. coelicolor genome encodes an extraordinary 819 predicted secreted proteins, including 60 proteases, 13 chitinases/chitosanases, eight cellulases/endoglucanases, three amylases and two pectate lyases (Bentley et al., 2002), and extracellular proteins from diverse streptomycetes hydrolyse and modify many different high- and low-molecular-weight compounds, or have novel functions (Schrempf, 2007). Proteomic analysis (Kim et al., 2005; Widdick et al., 2006; Akanuma et al., 2009) confirms that streptomycetes secrete large numbers of proteins. The extent of intracellular–extracellular transactions is further emphasized by the plethora of surface-located signal transduction systems, including about 50 ‘extracytoplasmic function’σ factors (compared with one in Escherichia coli K-12), 50–100 two-component systems, and several tens of serine–threonine protein kinases.

Extending the comparative analysis to include two further complete genome sequences of streptomycetes to which we had access, nearly 13 000 genes are specific to one of the five (K.F. Chater & G. Chandra, unpublished data). Most of these are located in the 1–2-Mb terminal regions of chromosomes, confirming and extending previous findings (Choulet et al., 2006; Ohnishi et al., 2008). Genes likely to encode secreted proteins make up 7.6 % of species-specific genes, compared with 5% for the whole genomes. Genes for secondary metabolism are also over-represented among species-specific genes, which are considered to be highly enriched for genes acquired by horizontal gene transfer over the hundreds of millions of years since the last common ancestor of the five organisms (Chater & Chandra, 2006; Ventura et al., 2007).

In this article, we present a broad survey of an elaborate extracellular biology, dealing not only with secreted proteins and small molecules but also with larger extracellular structures. After a description of the range of protein secretion systems used by streptomycetes, we focus on aspects of the extracellular biology that are intimately bound up with the evolution and developmental adaptations of streptomycetes.

The Streptomyces genome encodes at least three different types of general protein secretion systems

The Sec pathway

The Sec pathway is, by analogy with most other prokaryotes, the main route by which proteins are exported across the cytoplasmic membrane of streptomycetes (Driessen & Nouwen, 2008). The S. coelicolor genome encodes all the Sec components, including the essential proteins SecY and SecE, along with SecD and SecF, which are encoded by adjacent genes. Both co- and post-translational routes of targeting secretory proteins to the Sec translocon are present. The SecA protein from Streptomyces lividans has been shown in vitro to be required for the secretion of the Sec-dependent α-amylase (Blanco et al., 1998). Streptomyces species also encode the RNA and protein components of the signal recognition particle (SRP; Palacín et al., 2003). In E. coli, SRP is essential for the targeting of almost all inner membrane proteins (Luirink et al., 2005), and as such, ffh, encoding the protein component of SRP, and ftsY, encoding the SRP receptor, are essential genes in E. coli (Seluanov & Bibi, 1997). Interestingly, however, ftsY can be deleted from S. coelicolor, and although the mutation affects sporulation and antibiotic production, the strain remains viable (Shen et al., 2008). This might suggest differences in the action of SRP in Streptomyces compared with other bacteria.

The Esx secretion system

A second type of protein secretion system, the recently described Gram positive-specific Esx or type VII secretion system, has been found in streptomycetes, based on an Esx genomic signature comprising a gene encoding an FtsK/SpoIIIE-like ATPase, in the vicinity of one or more genes encoding small proteins with a conserved W-x-G amino acid motif (Pallen, 2002). The system was first described in Mycobacterium tuberculosis, where it is required for the secretion of two small extracellular proteins, ESAT-6 and CFP10 (Hsu et al., 2003; Pym et al., 2003; Stanley et al., 2003). A related system in Staphyloccus aureus secretes similar proteins (Burts et al., 2005). Although in both of these organisms the Esx pathway is intimately related with virulence, the delivery of virulence factors is unlikely to be the only function of this transport machinery because it is also encoded in the genomes of presumptively nonpathogenic streptomycetes. Streptomyces griseus, S. avermitilis and S. coelicolor have two apparently distinct systems (the ATPases are encoded by SCO4508 and SCO5734 in S. coelicolor), while the plant pathogen Streptomyces scabies has just one (encoded by SSC58621, which is most closely related to SCO5734). The function of any of these systems in Streptomyces is currently unknown, but a significant physiological role is suggested by analogy to similar systems in other organisms. It should be noted that in addition to the secretion of small proteins with a W-x-G motif, this machinery, at least in Mycobacterium, also secretes other proteins that lack any apparent signal sequence (Abdallah et al., 2006; Raghavan et al., 2008). This means that there may be many substrates of this export system waiting to be uncovered in Streptomyces.

The Tat pathway

The third general protein transport system found in Streptomyces, the Tat pathway, differs from Sec in that it transports prefolded proteins across the cytoplasmic membrane. It was originally assumed that the Tat pathway was dedicated primarily to the export of proteins that bind complex redox-active cofactors, which by necessity are inserted into the preprotein in the cytoplasmic compartment (Berks et al., 2000); however, Streptomyces species and some other prokaryotes, including certain halophilic archaea, appear to export large numbers of non-cofactor-containing proteins by this pathway (Rose et al., 2002; Dilks et al., 2003; Widdick et al., 2006). Because of this unexpected extension of the substrate range of the Tat system, we describe the literature on the subject in some detail here.

Proteins are targeted to the Tat pathway by N-terminal signal peptides with two adjacent and invariant arginine residues (Berks, 1996), and hydrophobic regions that are less hydrophobic than those of Sec signal peptides (Cristobal et al., 1999). When either of the two useful prediction programs for twin-arginine signal peptides, tatfind and tatp (Rose et al., 2002; Dilks et al., 2003; Bendtsen et al., 2005), are applied to S. coelicolor, substrate numbers of 145 and 189, respectively, are predicted, and 80 of these pass both prediction programs. However, the fact that many of the predicted substrates do not apparently contain cofactors, coupled with the observation that the high G+C content of the S. coelicolor genome means that arginine often replaces lysine in signal peptides, suggested that these predictions might be an overestimate. Most of the 25 Tat substrates experimentally verified by proteomic analysis of wild-type S. coelicolor and a tatC mutant strain had previously been identified by one or both of the prediction programs, while many proteins secreted independently of the Tat system had consecutive arginines in their signal peptides, but, significantly, were mostly not predicted to be Tat signal peptides by tatfind or tatp (Widdick et al., 2006). This study therefore helped to validate the utility of the in silico analysis and strongly indicated that the Tat pathway is indeed a general route of protein export in streptomycetes.

It is clear that the Tat system makes a major contribution to the physiology of Streptomyces because mutations in tat genes of S. coelicolor, S. lividans and S. scabies result in highly pleiotropic phenotypes (Schaerlaekens et al., 2004; Widdick et al., 2006; S. Mann & T. Palmer, unpublished data). Such mutants of S. lividans and S. coelicolor grow in a very dispersed manner in liquid culture and fail to sporulate on solid media containing sucrose. The tatC strain of S. coelicolor is fragile and prone to lysis, indicating a probable cell wall defect. A similar tatC mutant of S. scabies is defective for the production of melanin (S. Mann & T. Palmer, unpublished data), consistent with a failure to secrete the enzyme tyrosinase, a known Tat substrate (Schaerlaekens et al., 2001). It also shows defects in sporulation. The secretion of tyrosinase is of particular interest because the enzyme, encoded by the melC2 gene, lacks any N-terminal signal peptide. Instead, it is exported as a complex with its chaperone, MelC1, which has a typical Tat-targeting signal (Chen et al., 1992). After transport through the Tat system, MelC1 apparently dissociates from tyrosinase, as it is not found in the mature enzyme. The dissociation of MelC1 depends on the presence of copper in the medium, and it is likely that MelC1 also aids the insertion of copper into the tyrosinase active site (Chen et al., 1992; Tsai & Lee, 1998).

The known or predicted Tat substrates in S. coelicolor include a diverse array of hydrolytic enzymes; thus, the Tat system most likely plays a major role in nutrient acquisition from complex sources, permitting growth when more readily used soluble nutrients are not available. For example, it is important for phosphate acquisition, with several phosphatase enzymes, including alkaline phosphatases of the PhoD and PhoX families as well as phosphodiesterases and phospholipases, being secreted Tat-dependently (Widdick et al., 2006). In addition, at least one of the three characterized xylanases of S. lividans is a Tat substrate (Faury et al., 2004; Schaerlaekens et al., 2004), and many other enzymes involved in extracellular carbohydrate metabolism are known or predicted Tat substrates (Widdick et al., 2006) (see also the list of predicted Tat substrates of S. coelicolor at Other Tat substrates include a suite of substrate-binding protein components of ATP-binding cassette (ABC) transporters, which in Gram-positive bacteria are usually anchored to the outer face of the membrane by lipid modification. Indeed, inspection of many of their signal peptides reveals apparent ‘lipoboxes’ with invariant cysteine residues at the +1 position of the mature protein sequence, and it has recently been demonstrated that SCO1639, SCO2780 and SCO3484 encode Tat-dependent lipoproteins (M.G. Hicks, B.J. Thomson, D.A. Widdick, G. Chandra, K. Findlay, I.C. Sutcliffe, T. Palmer & M.I. Hutchings, unpublished data).

In the interesting case of the extracellular agarase of S. coelicolor (an enzyme absent from most other characterized streptomycetes), the signal peptide has a canonical Tat signal peptide and indeed is exported by the Tat pathway (Widdick et al., 2006, 2008). This signal peptide was used for heterologous protein production long before the Tat system had been discovered (Buttner et al., 1987; Buttner et al., 1988; Isiegas et al., 1999), and expression of dagA from a multicopy vector resulted in as much as a 500-fold increase in extracellular agarase activity (Kendall & Cullum, 1984). This indicates that the Tat pathway has enormous capacity for protein secretion. Agarase itself has been used experimentally as a very effective reporter enzyme to verify Tat signal peptides from bacteria, archaea and plants. Its application to S. scabies genes indicates that a number of candidate virulence factors, as well as a likely paralogue of the secreted regulatory protein factor C, are substrates of the Tat pathway (S. Mann, D. Widdick & T. Palmer, unpublished data), suggesting that many other interesting Tat substrates remain to be uncovered.

Feasting on chitin and its derivatives

Chitin, an insoluble nitrogen-containing polysaccharide, is a major nutrient source for streptomycetes, being acquired even from living sources. Because chitin is insoluble, complex extracellular systems have been developed for its utilization.

Chitin synthesis and its inhibition by streptomycetes

Annually, >1 × 1010 tonnes of chitin are estimated to be produced in terrestrial and marine habitats, making it the second most abundant polysaccharide in nature. It consists of β-1,4-linked N-acetylglucosamine (NAG) chains associated in antiparallel (α) or parallel (β) fashion to form high-molecular-weight (up to several MDa) crystalline networks stabilized by hydrogen bonding and van der Waals' interactions. Chitin types differ in their degree of acetylation, in their molecular size (up to several MDa), and in their association with other organic and inorganic compounds. The high tensile strength of crystalline α-chitin makes it a very important component of fungal cell walls, the exoskeletons of arthropods, and the peritrophic membrane within the gut of various organisms. Chitin also appears to be a component of the surface ornamentation of spores of some streptomycetes (Smucker & Pfister, 1978; Gomes et al., 2008). Chitosans, which are encountered in the cell walls of certain fungi, are derived by enzymatic deacetylation from chitin (for a review, see Muzzarelli, 1999).

The formation of chitin depends on diverse chitin synthases. Chitin synthases require UDP-NAG as a substrate and NAG as an activator, but they vary in pH optimum and ion requirement. Filamentous fungi can have up to 10 types of these membrane-spanning enzymes, which have a similar catalytic region but different accessory domains. Some streptomycetes produce competitive inhibitors of chitin synthase, presumably aiding competition with chitin-containing organisms sharing the soil habitat. Different chitin synthases vary in their susceptibility to these inhibitors, which include complex peptidyl nucleosides, such as nikkomycins, and polyoxins that are believed to bind to the catalytic site (see Fig. 1) (reviewed by Ruiz-Herrera & San-Blas, 2003).

Figure 1.

 Survey of the chitinolytic system. Some streptomycetes produce nikkomycin or polyoxin (right, top), which inhibit chitin synthases from fungi and other chitin-producing organisms. With a range of binding proteins, streptomycetes target chitin or chitosan, or organisms containing these polysaccharides. This process can lead to a close interaction between Streptomyces and fungal hyphae (left, top). Streptomycetes produce a range of chitinases, and those belonging to family 18 are inhibited by allosamidin. As a result of chitin degradation by chitinases (right), chito-oligomers (NAG)n, chitobiose (N,N′-diacetylchitobiose) or NAG arise. In Streptomyces olivaceoviridis NAG is taken up either via a specific PTS transport system or via the ABC transporter Ngc (in cooperation with the ATP-binding protein MsiK), which also transports chitobiose. In Streptomyces coelicolor, the ABC transporter Das is required only for the uptake of chitobiose.

Tools for chitin recognition

Chitin is unusual among polysaccharides in containing nitrogen, and, unlike most other bacteria, streptomycetes can use it as a carbon and a nitrogen source (Schrempf, 2001). Binding to chitin is important for its degradation. Streptomycetes used specialized chitin-binding proteins (CHBs), such as the 77% identical CHB1 (18.7 kDa) of Streptomyces olivaceoviridis and CHB2 (18.6 kDa) of Streptomyces reticuli, which do not display any catalytic or antifungal activity (Schnellmann et al., 1994; Kolbe et al., 1998) (Fig. 1). Many CHB homologues, both close and more distant, are present in streptomycetes.

CHBs interact very strongly with various types of α-chitin, including nascent α-chitin fibres generated in vitro (Siemieniewicz et al., 2007), with submicromolar dissociation constants. Binding is less strong to colloidal chitin, a predominantly amorphous form derived from native α-chitin after acid treatment. CHB1 also targets the chitin of living organisms, including crab shells, the peritrophic membrane of potworms (abundant inhabitants of soil), and the hyphae and spores of various chitin-containing fungi (Zeltins & Schrempf, 1995). The 14.9-kDa CHB3 protein from S. coelicolor, which shares only 37% amino acid identity with CBH1, targets both α-chitin and β-chitin, but relatively loosely; furthermore, it binds to chitosan at a low salt concentration. The CHBs and the chitin-binding domains from several chitinases show no common amino acid motifs (Schrempf, 2001).

CHB1 consists of an elongated main globule about 4 nm long and an c. 2-nm-wide foot-like domain. Its structure and function depend on disulfide bonds (Svergun et al., 2000). Its efficient specific binding to α-chitin involves conserved tryptophan residues (Zeltins & Schrempf, 1997).

Chitin-hydrolysing enzymes

Diverse organisms, including bacteria, fungi and plants, produce chitinases and chitosanases, which cleave the glycosidic linkages between the sugar units of chitin or chitosan, to generate low molecular weight chito-oligosaccharides. Based on the primary sequence, three-dimensional structure, and catalytic mechanism, different glycoside hydrolase families have been defined, and chitinases belong mostly to families 18 and 19 (Cantarel et al., 2009). Family 18 chitinases occur in many different organisms, and use a substrate-assisted double-displacement reaction with retention of the configuration of the anomeric carbon, whereas family 19 chitinases are found mainly in plants and actinomycetes, and mediate a single-displacement reaction with inversion of the configuration of the anomeric carbon (Cantarel et al., 2009).

Chitinases have been purified from various Streptomyces strains (reviewed by Schrempf, 2001, 2007). Most streptomycetes contain many chitinase genes, potentially enabling them to hydrolyze the natural diversity of chitin types. For example, the chromosome of S. coelicolor encodes 11 deduced family 18 chitinases and two family 19 chitinases, with various modular arrangements comprising catalytic, substrate-binding and linker domains. However, detailed studies of enzymes that degrade crystalline chitin from exoskeletons or fungal walls are scarce. Streptomyces olivaceoviridis can degrade chitin very efficiently (Blaak et al., 1993), as it secretes many chitinases including an exochitinase (exo-chiO1) of 59 kDa with a C-terminally located family-18 catalytic domain, a central region containing an FnIII module, and an N-terminal region. This enzyme efficiently adheres to and hydrolyzes chitin in native fungal hyphae and crab shells. During cultivation, the 59-kDa chitinase is specifically processed to a 47-kDa truncated form, which retains the catalytic and the FnIII domain, but lacks the N-terminal part. The 47-kDa form has lost specific strong binding to crystalline α- or β-chitins, but retains full activity on colloidal chitin and low-molecular-mass substrates. Thus, strong adhesion of the large form of the enzyme mediated by the N-terminal domain (12 kDa) is needed for effective hydrolysis of crystalline chitin (Blaak & Schrempf, 1995; Vionis et al., 1996). This chitin-binding domain does not resemble the CHBs described above. Nevertheless, as for CHB1, tryptophan residues play a key role in the interaction with insoluble chitin.

The pseudotrisaccharide allosamidin, a secondary metabolite produced by about 5% of streptomycetes, is a competitive inhibitor of family-18 chitinases including exo-chiO1 and eukaryotic chitinases (see Fig. 1). Some evidence suggests that, at subinhibitory levels, it induces the production of some Streptomyces chitinases via a two-component regulatory system (Suzuki et al., 2008). The significance of this intriguing observation remains unclear.

Uptake systems associated with chitin degradation

Chitin breakdown products can be transported by at least three transport systems. During cultivation in the presence of NAG or chitin, Streptomyces mycelium efficiently takes up NAG. The S. olivaceoviridis membrane contains a protein (NgcE) that binds to NAG (KD=8.3 nM) and chitobiose (N,N′-diacetylchitobiose, KD=29 nM), and also has micromolar affinity for chitotetraose, chitopentaose, chitohexaose, and chitotriose. The corresponding ngcE gene encodes a predicted lipid-anchored member of the CUT-1 family of ABC transporters for carbohydrates, and is next to genes for two membrane-spanning proteins (NgcF and NgcG) that have a consensus motif for integral membrane ABC transporters (Wang et al., 2002). The NgcEFG system shares an ATP-hydrolysing protein, MsiK (Schlösser et al., 1997), with several other ABC transporters for carbohydrates (also see the section on cellulose breakdown), and transports both NAG and chitobiose (see Fig. 1). However, an ngcE mutant can still transport NAG, because it retains a functional phosphoenolpyruvate (PEP)-dependent phosphotransferase system (PTS) (see Fig. 1; Saito & Schrempf, 2004). This is encoded by two adjacent genes, ptsC1 and ptsC2, encoding proteins that are 56% identical and contain eight transmembrane regions. PtsC1 and PtsC2 each correspond to a single EIIC domain otherwise known only in some bacterial multidomain permeases for glucose/mannose or NAG. The functional soluble PTS components HPr, EI and EIIA complete the system (Xiao et al., 2002). Analysis of double mutants showed that PtsC2, and not PtsC1, mediates PEP-dependent specific transport of NAG (Km=5 μM).

In S. coelicolor, a chitobiose-specific ABC transporter encoded by the dasABC operon also requires the cooperation of MsiK (see Fig. 1). Chitinase induction by chitobiose or chitin is lost in an msiK mutant, indicating that chitobiose uptake is necessary for induction. The gene dasR is required for regulation of the dasABC operon (Saito et al., 2007, 2008), but also plays a wider role: remarkably, the NAG derived from chitin breakdown interacts with the DasR regulatory protein to exert general carbon catabolite repression in streptomycetes, which also affects their morphological differentiation and secondary metabolism (Rigali et al., 2006, 2008).

Interaction of streptomycetes with chitin-containing organisms

During cocultivation of the highly chitinolytic S. olivaceoviridis and the ascomycete Aspergillus proliferans, the Streptomyces spores germinate first, and produce the exo-ChiO1 chitinase described above. Outgrowth of some fungal spores follows. Streptomyces mycelium then undergoes massive growth at the expense of fungal hyphae before a balanced proliferation of closely interacting fungal and Streptomyces hyphae is attained (Fig. 1). This process involves the aggregation of CHB1 molecules to maintain a close glue-like contact of the partners, followed by several concerted responses (Siemieniewicz & Schrempf, 2007). CHBs mediate interaction of various Streptomyces strains with a range of fungi under different conditions.

In a soil-microcosm system, S. lividans carrying the gene for exo-ChiO1 degrades ground chitin or added A. proliferans hyphae very efficiently (Vionis et al., 1996; Schrempf, 2007), and survives well in the gut of the potworm Enchytraeus crypticus, where it can degrade fungal hyphae on which E. crypticus has been fed (Kristufek et al., 1999; Schrempf, 2001). The ecological roles of chitin synthase inhibitors, which exhibit antifungal but not antibacterial activity (Ruiz-Herrera & San-Blas, 2003), have yet to be evaluated, but their action can lead to hyphal bulging and eventual destruction of some fungi. Chitinolytic streptomycetes have been investigated as possible biocontrol agents against damaging fungi, such as the agents of Fusarium wilt of cucumber (Singh et al., 1999).

How streptomycetes gain access to insoluble crystalline cellulose

As with chitin, access to other insoluble polysaccharides as sources of nutrients involves elaborate extracellular systems, including mechanisms for binding to cellulose and extracellular proteolytic processing to change the behaviour of enzymes during cellulose degradation.

Nonenzymatic proteins for targeting cellulose

Cellulose, the main polymeric component of the plant cell wall, is the most abundant polysaccharide on earth. The thick secondary cell wall comprises highly ordered, crystalline cellulose and xylans, and can also be lignified to provide further strength. Its targeting by streptomycetes depends on a 35.6-kDa carbohydrate-binding protein (CbpC) that is present in most streptomycetes. CbpC is secreted via a typical N-terminal signal sequence. It is linked to the peptidoglycan layer of the mycelium via a C-terminal Streptomyces sortase recognition signal (LAETG, instead of the generic LPXTG consensus, plus a hydrophobic domain and charged tail). Purified CpbC interacts strongly with crystalline cellulose and moderately with different crystalline chitin types, but not with chitosan. A glycine-aspartate/serine rich region, which separates the carbohydrate-binding module from the sorting signal, is important for protein stability (Walter & Schrempf, 2008). In addition to CbpC, streptomycetes produce a 35-kDa protein (AbpS) that very specifically targets a predominantly crystalline type of cellulose (Avicel), but not soluble types of cellulose or other polysaccharides (Walter et al., 1998). Cellulose is bound by the N-terminus of AbpS, which protrudes from the murein layer, to which AbpS is tightly and very likely covalently linked, while the 35-aa C-terminus containing hydrophobic amino acids anchors AbsS to the membrane. Thus, AbpS connects the hyphal interior with the extracellular space, via a large centrally located α-helical structure, which has weak homology to tropomyosins and streptococcal M-proteins (Walter & Schrempf, 2003). Dimerization of two AbpS molecules seems to be critical for targeting insoluble cellulose, a mechanism that differs from that of cellulose-binding domains within cellulases. Oligomerization on contacting cellulose could possibly transduce extracellular signal(s) into the cell.

Degradation of crystalline cellulose

In addition to cellulose, the plant cell wall also contains diverse xylans, 1,4-linked xylose polymers commonly containing side branches of arabinosyl, glucuronosyl, acetyl, uronyl and mannosyl residues. Streptomycetes have a battery of extracellular enzymes to allow concerted hydrolysis of xylans, some of which have been investigated in detail (Pagéet al., 1996; Ruiz-Arribas et al., 1997; Tsujibo et al., 1997). Many streptomycetes can degrade soluble carboxymethyl cellulose by secreted enzymes generally like those known for many other bacteria. However, only 25 of 180 Streptomyces strains tested could hydrolyse Avicel.

With Avicel as the sole carbon source, S. reticuli produces no specific cellulolytic enzymes during early growth, but instead it produces a heme-containing dimeric catalase/peroxidase enzyme (CpeB, 82-kDa subunits) with a broad substrate specificity, which probably generates reactive oxygen species to mount an initial attack on the insoluble cellulose (Zou & Schrempf, 2000). CpeB is ‘mycelium associated’, in that it can be released from the mycelium by mild detergent treatment. Later in growth, CpeB is processed to a form lacking the C-terminal region and the associated heme-independent Mn-peroxidase activity (Zou & Schrempf, 2000).

After CpeB action, S. reticuli releases an unusual mycelium-associated 82-kDa cellulase (Avicelase), with an N-terminal cellulose-binding domain and a C-terminal catalytic domain belonging to cellulase family E (Schlochtermeier et al., 1992b). The binding domain targets the enzyme to Avicel during early stages of growth. After this, the C-terminal 42-kDa region is released by a 36-kDa metalloprotease (Moormann et al., 1993). The 42 kDa form hydrolyses Avicel to oligomers that are further processed to cellobiose and cellotriose by a glucosidase (Schlochtermeier et al., 1992a). Avicel is the only known inducer of Avicelase: low-molecular-weight breakdown products of cellulose could be excluded (Walter & Schrempf, 1996).

Uptake of cellulose degradation products

In S. reticuli, the specific uptake of cellobiose and cellotriose is mediated by a novel ABC transporter comprising a cellobiose-binding lipoprotein (CebE), and two integral membrane proteins (CebF and CebG). The requisite ATP-binding protein, MsiK, which is encoded by a distantly located gene, also cooperates with different sugar ABC transport systems in S. reticuli and S. lividans (Schlösser et al., 1997, 2000). CebE is synthesized specifically during growth with cellobiose, under the control of a regulatory protein, CebR, which is specified by a gene next to the cebEFG gene cluster. On the other hand, MsiK is produced in the presence of several sugars, including NAG and chitobiose (Schlösser & Schrempf, 1996; Schlösser et al., 1999; Wang et al., 2002).

Ecological considerations

In terrestrial habitats, streptomycetes are important for the initial decomposition of organic material by their large repertoire of cellulases, xylanases, lignocellulase and other enzymes (Schrempf, 2007). When specific types and amounts of carbon sources were added alone and in combination with native soil mesocosms, the addition of cellulose or lignin led to the highest Streptomyces densities (Schlatter et al., 2009), showing that the cellulolytic systems described above are highly relevant to the ability of streptomycetes to cope with cellulose types in natural habitats. The ability of some streptomycetes to degrade xylans and cellulose-related materials has led to their apparent symbiotic involvement, along with many other microorganisms, in the gut activities of termites (Schäfer et al., 1996).

Although most streptomycetes are saprophytic, a few species (including S. scabies) provoke visible lesions on the surface of various root and tuber vegetables, including potatoes (potato scab). Virulence is mainly dependent on their inhibition of plant cellulose synthesis, by the production of thaxtomins, 4-nitroindol-3-yl-containing 2,5-dioxopiperazines (Bischoff et al., 2009; King & Calhoun, 2009). In Streptomyces turgidiscabies, the DNA encoding thaxtomin biosynthesis is in a large pathogenicity island that also includes genes for necrosis (nec1) and a tomatinase (tomA) (Joshi et al., 2007; Loria et al., 2008; Seipke & Loria, 2008), and for production of nitric oxide, a key regulatory molecule for several plant processes (Johnson et al., 2008). However, some pathogenic isolates lack either nec1 or tomA or both (Lerat et al, 2009). The fact that the genes responsible for pathogenicity can disseminate by lateral transfer in different combinations implies that nonpathogenic strains can evolve to pathogenic ones.

The roles of extracellular proteases and protease inhibitors in the growth and development of streptomycetes

Streptomycetes are prodigious producers of proteases (the commercially available protease cocktail Pronase, prepared from culture fluids of S. griseus, contains as many as 10 different proteases: Sweeney & Walker, 1993). In S. coelicolor, 56 genes are annotated to encode proteases: 27 for serine proteases, eight for metalloproteases, and 21 for aminopeptidases (Bentley et al., 2002). Clearly, extracellular proteases are involved in assimilating extracellular proteinaceous nitrogen sources, some of which, such as keratin (Böckle & Müller, 1997), may be quite recalcitrant. One strain produces at least four different keratinases (Xie et al., 2009). However, much evidence has now accumulated that some extracellular proteases are implicated in Streptomyces development, and that this role involves elaborate cascades mediated by protease inhibitors and specific proteolytic processing.

Early studies

In early studies, two strains were analysed in detail: Streptomyces albidoflavus SMF301, one of a minority of streptomycetes that can form spores in submerged culture, and Streptomyces exfoliatus SMF13, which produces leupeptin, a tripeptide inhibitor of proteases (Kim & Lee, 1990; Kim et al., 1991, 1992, 1993). These strains produce at least three types of extracellular protease in culture fluids: a chymotrypsin-like protease (CLP); a trypsin-like protease (TLP); and a metalloprotease (MTP) (Shin & Lee, 1986; Rho et al., 1990; Jeong et al., 1993; Kang et al., 1995a, b).

The main role of the CLP appears to be to provide nutrients for growth: CLP activity is higher during growth on sodium caseinate as the sole nitrogen source, and a CLP-specific inhibitor inhibits growth, but not sporulation, of S. albidoflavus (Shin & Lee, 1986; Kang et al., 1995a; Kim & Lee, 1995a). In contrast, several lines of evidence implicate TLP and MTP in sporulation. Their activities increase with the number of spores from the middle of the S. albidoflavus exponential growth phase. At the end of growth, TLP activity declines as the mycelium lyses, whereas metalloprotease activity is maintained. In agar-grown cultures, addition of a TLP inhibitor interferes with aerial growth, while addition of an metalloprotease inhibitor suppresses spore formation (Rho & Lee, 1994; Kang et al., 1995b; Kang & Lee, 1997). It appears that TLP and MTP help to break down substrate mycelium for nutrients to support spore formation (Kang et al., 1998; Lee, 1998).

As in some eukaryotic microorganisms (e.g. Coccidioides immitis, Yuan & Cole, 1989; Saccharomyces cerevisiae, Holzer, 1983), protease activity associated with S. exfoliatus sporulation is regulated by an autogenous protease inhibitor, in this case leupeptin, in line with the suggestions of Aoyagi (1989) and Gräfe (1989). Thus, leupeptin accumulated during growth of S. exfoliatus holds in check a TLP that has broad substrate specificity. Upon the exhaustion of glucose in submerged culture, leupeptin is completely degraded by late-accumulating leupeptin-inactivating enzyme (LIE, a 34.7-kDa metalloprotease), and increased TLP activity is correlated with the subsequent decline of mycelium (Kim & Lee, 1995a; Kim et al., 1998). Both are delayed by adding leupeptin. In surface-grown cultures, substrate mycelium degradation and aerial growth are closely associated with TLP activity (Kim et al., 1995; Kim & Lee, 1995b, 1996).

Genetic dissection of the roles of proteases and protease inhibitors in differentiation

A leupeptin-centred cascade is not universal among streptomycetes, most of which produce neither leupeptin nor LIE (Kim et al., 1998). In addition, no proteases with N-terminal amino acid sequence similarity to mature TLP of S. exfoliatus are encoded in the genome of S. coelicolor, although they are widespread in other species. Instead, it is likely that low-molecular-weight protease inhibitor proteins (protein-Streptomyces subtilisin inhibitors, SSIs) may fulfil a leupeptin-like role in many streptomycetes. SSIs have been identified from many Streptomyces spp. (Taguchi et al., 1993; Kuramoto et al., 1996; Van Mellaert et al., 1998), mainly in the context of possible therapeutic uses (Maeda et al., 1971; Umezawa, 1982). The SSIs are conserved dimers of identical subunits of about 100 amino acids (Mitsui et al., 1977; Takeuchi et al., 1991, 1992). Their interactions with some of the various types of target serine protease such as subtilisin, chymotrypsin, trypsin, and zinc-containing metalloproteases from S. griseus (Kumazaki et al., 1993) have been analysed (Laskowski et al., 1983; Kojima et al., 1990, 1998; Taguchi et al., 1995). Significantly, for the biological roles of SSIs, one SSI (and presumably most of them, considering their end-to-end alignments) is ‘two-headed’, with the ability to interact simultaneously with two different types of proteases (Hiraga et al., 2000).

The evidence indicating that SSIs play a leupeptin-like role in the developmental control of autolysis in many streptomycetes comes from studies of developmental mutants of two phylogenetically distant species: the genetic model organism, S. coelicolor, and S. griseus, whose development has been analysed mainly as a consequence of the early discovery of bald colony mutants that could be phenotypically rescued by the addition of the signalling molecule A-factor.

Among the many S. coelicolor bld genes (named so because the mutant colonies lack aerial growth and thus are ‘bald’), bldA is unusual: it specifies the only tRNA for the leucine codon UUA, which (as TTA) is very rare in the high-GC genomes of streptomycetes (Lawlor et al., 1987; Leskiw et al., 1991). Mutations in bldA interfere with morphological differentiation and production of several antibiotics in a range of streptomycetes (Merrick, 1976; Chater & Chandra, 2008). One of a number of proteins absent from a bldA mutant of S. coelicolor is an abundant SSI-type protease inhibitor encoded by SCO0762 [termed Streptomyces trypsin inhibitor (STI) because of its predicted specificity for trypsin] (Kato et al., 2005b; Kim et al., 2005, 2008). SCO0762 does not contain a TTA codon, but instead is transcriptionally dependent on the developmentally important AdpA regulatory protein, which is encoded by a TTA-containing gene (adpA, also known as bldH: Nguyen et al., 2003; Takano et al., 2003). In S. griseus too, production of a protease inhibitor (SgiA) is dependent on AdpA (Hirano et al., 2006), which in this species is the route through which the A-factor mediates its effects (Vujaklija et al., 1993). There are likely AdpA-binding motifs upstream of SSI genes in Streptomyces albogriseolus S-3253 (Obata et al., 1989) and Streptomyces venezuelae (Van Mellaert et al., 1998). Moreover, all characterized adpA genes have a TTA codon. Thus, bldA and AdpA may influence the regulation of protease inhibitors in many streptomycetes.

The many targets for AdpA in S. griseus include 18 genes for proteases (Yamazaki et al., 2004; Kato et al., 2005a; Akanuma et al., 2009). One of these, sprT, encodes the target TLP for the SgiA protease inhibitor (Kato et al., 2005b; Oh et al, 2007). Another of these proteases, a metalloprotease, has been implicated in morphological differentiation (Kato et al., 2002, 2005a; Tomono et al., 2005). This suggests that the morphological defects of bldA and adpA mutants of S. griseus are at least partially due to changes in the abundance and activities of extracellular proteases. Consistent with this, the exogenously added SgiA protease inhibitor inhibits aerial mycelium formation in S. griseus (Hirano et al., 2006).

In S. coelicolor, changing the TTA codon in adpA largely, but not completely, suppresses the morphological deficiency of a bldA mutant (Nguyen et al., 2003; Takano et al., 2003), suggesting that while most of the deficiency is caused by the bldA dependence of adpA, at least one other TTA-containing gene also contributes. One of the many S. coelicolor protease determinants, SCO5913, contains a TTA codon, and an SCO5913 deletion mutant is delayed in differentiation, indicating that the TTA codon in SCO5913 probably does contribute to the bldA mutant phenotype (Li et al., 2007; Kim et al., 2008a). The SCO5913 mutant has higher and longer-lived STI activity than the wild type, indicating that SCO5913 protease is responsible for some inactivation of STI. In confirmation of this role of SCO5913, artificial induction of SCO5913 causes STI activity to disappear, and STI is degraded in vitro by SCO5913. However, STI inactivation also involves some other protease(s), because STI eventually disappears even from the SCO5913 null mutant, after which the mutant begins to differentiate. A role for STI in holding off the onset of development is also indicated by the accelerated differentiation of a constructed SCO0762 mutant (i.e. lacking STI) (Kim et al., 2008a).

Prokaryotic convertases interact with STI

A yeast two-hybrid cDNA analysis revealed two S. coelicolor proteases that interact with STI. Both have a C-terminal proprotein convertase P-domain that is characteristic of eukaryotic subtilisin-like proprotein convertase, and which is responsible for the two-hybrid interaction. One, encoded by SCO5447, is a putative neutral zinc metalloprotease homologous to an S. griseus protease, SgmA, that is AdpA-dependent (Ohnishi et al., 2005; Kim et al., 2008a). The other, encoded by SCO1355, is a putative secreted serine protease whose two catalytic domains (subtilisin N and peptidase S8) are not related to the SCO5447 catalytic domains, despite the 55% identical P-domains (Kim et al., 2008a). Independent evidence of STI–P-domain interactions comes from the finding that an STI-like protein (Q26 kexstatin I) of Streptomyces platensis inhibits the yeast proprotein convertase kexin Kex2 (Oda et al., 2001).

A SCO1355 mutant shows delayed development, consistent with a role for its product in development, but surprisingly, a SCO5447 mutant has normal development, possibly because of functional complementation by the homologous neighboring gene, SCO5446, under the conditions tested (Kim et al., 2008a).

In eukaryote cells, proprotein convertase is often involved in the processing of precursors of proteins (including a metalloprotease) that are important for the accurate regulation of cellular events (Rockwell et al., 2002; Ueda et al., 2003; Henrich et al., 2005; Szumska et al., 2008). By analogy, SCO5447 or SCO1355 proteases may process target proproteins required for differentiation in S. coelicolor.

Possible targets of the protease cascade

In Streptomyces (formally Streptoverticillium) mobaraensis, a protease called TAMEP whose experimentally determined N-terminal sequence corresponds to that of SCO5447 protease activates the proprotein form of an extracellular transglutaminase. Extracellular transglutaminases are industrially interesting enzymes that catalyse the cross-linking of proteins to generate high molecular weight aggregates. A few actinomycetes produce such enzymes, but their biological functions are not understood.

The transglutaminase from S. mobaraensis is synthesized initially as an inactive preproprotein. After cleavage of a secretion signal peptide (a possible Tat substrate according to our retrospective analysis), activation of the proprotein is by further proteolytic removal of a 45-aa N-terminal domain that, before cleavage, keeps the enzyme inactive and increases the protein's stability (Washizu et al., 1994; Pasternack et al., 1998). TAMEP, which is a metalloprotease, is secreted somewhat later than the protransglutaminase (Pasternack et al., 1998), and is regulated by an SSI-type protease inhibitor (Zotzel et al., 2003a, b). Similar findings have recently been reported for the activation of a transglutaminase from a Streptomyces hygroscopicus strain, with the interesting difference that the new protransglutaminase has a cluster of potential cleavage sites for different kinds of proteases, and activation can be carried out by either an endogenous serine protease or a metalloprotease (Ren et al., 2008; Zhang et al., 2008a).

Although homologues of the known actinobacterial transglutaminases are not encoded in most sequenced Streptomyces genomes, it has been suggested that this cell wall-associated enzyme may play a developmental role in cross-linking surface proteins of aerial hyphae. Intriguingly, the protease inhibitor of the S. mobaraensis cascade is itself a substrate for the transglutaminase (Schmidt et al., 2008), and the equivalent protease inhibitor from S. hygroscopicus has significant surfactant activity (Zhang et al., 2008b). This raises the possibility that the inhibitor may have a second function beyond its role in controlling the protease cascade: the surfactant property might induce it to coat the aerial mycelium (like the more strongly surface-active proteins and peptides described below), facilitating its cross-linking by transglutaminase to form a coherent surface layer.

It will be important to determine whether the protease cascade is involved in the activation of other kinds of proenzymes, which has been described in the context of Streptomyces growth and development. For example, in Streptomyces antibioticus, sporulation is regulated in part by nucleases, one of which is activated through processing by a serine protease (Nicieza et al., 1999; Fernández & Sánchez, 2002), while there are several examples of proteolytic separation of the substrate-binding and catalytic domains of extracellular cellulases and xylanases (e.g. Fernández-Abalos et al., 2003). In addition, processing of an N-terminal peptide sequence by an unidentified enzyme is needed for activation of a precursor of a secreted morphogenetic peptide, SapB, that is essential for aerial mycelium formation in streptomycetes under certain growth conditions (Kodani et al., 2004).

General model for the protease cascade

In Fig. 2, we show a model that links intracellular regulatory events with a complex multistep extracellular protease cascade. There is no information yet to indicate how transcription of the majority of the genes presented in the model is controlled, making it difficult at present to rationalize the timing of the events depicted. In an influential paper, Willey et al. (1993) deduced the existence of a cascade of still mostly uncharacterized extracellular signals needed for development, based on observations of extracellular cross-complementation of aerial mycelium deficiencies among a collection of S. coelicolor bld mutants, including bldA and adpA (=bldH). As summarized by Chater & Chandra (2006), the initiation of the cascade of Willey et al. (1993) requires that a bldJ-dependent signal is imported by the oligopeptide transporter specified by the bldK locus, leading to bldA- and bldH-dependent production of a further signal. This leads to the sequential activation of further genes encoding transcription factors, in each case followed by the production of a further signal, in the order bldG, bldC, and bldD plus bldN. If, as has been suggested, this cascade and the protease cascade might overlap or even be the same (Kim et al., 2008a), some of the protease genes, such as SCO1355 and SCO5447, could turn out to be regulated by bld genes downstream of bldA in the cascade of Willey et al. (1993). Interestingly, the only previously identified signal in the cascade is an oligopeptide, whose origin is unknown (Nodwell & Losick, 1998). Our model speculatively suggests that the cleaved signal sequence(s) of one or more of the secreted proteins might provide the oligopeptide (Fig. 2). Alternatively, the oligopeptide might be generated by the action of a protease. In their initial paper, Willey et al. (1993) noted that not all bld mutants fit neatly into the linear cascade. The nonlinearity of the events outlined in Fig. 2 might well provide an explanation for this.

Figure 2.

 General model of an extracellular protease cascade that contributes to Streptomyces development. (a) The cascade held in check late in the main growth phase. Inside the hyphae various genes such as ssi, aaa, kkk and zzz are activated, some (e.g. ssi and kkk) by the action of AdpA (bright green), which depends on the UUA-reading tRNA specified by bldA and is autoactivating. Secretion of the gene products results in the release of signal peptides (leaf-green). SSI (red) dimerizes to form a two-headed protease inhibitor, which binds the P-domain (ochre) of certain proteases as well as some other proteases (grey) to hold them inactive. As long as the P-domain-containing protease is inactive, the prosequence (pink) of zymogens is uncleaved, and the zymogens therefore remain inactive. (b) Release of the cascade in stationary phase, activating autolysis. An inhibitor-inactivating enzyme (IIE, bright blue) is secreted (in Streptomyces coelicolor, the production of IIE requires the translation of a UUA codon). We speculate that induction of IIE may depend on the uptake of some specific oligopeptide by the BldK oligopeptide transporter, and that this oligopeptide (leaf-green) may be derived from the secretion signal sequence of a component of the protease cascade. The IIE inactivates SSI, thereby releasing the bound proteases. The nonspecific protease participates in a general manner in autolysis, while the P-domain-containing protease activates zymogens to yield enzymes (black) that play more specific developmental or autolytic roles, such as transglutaminases of some species.

Despite all these intriguing observations, the complex roles and interactions of proteases and protease inhibitors during Streptomyces differentiation have not been characterized at the molecular level, and the connection between intracellular and extracellular events is not fully defined.

Other proteinaceous inhibitors of exoenzymes

Streptomyces clavuligerus secretes a protein [β-lactamase-inhibitory protein (BLIP)] that inhibits β-lactamases, thereby enhancing the effectiveness of the β-lactam antibiotic that it produces (Doran et al., 1990). BLIP-I and BLIP-II are also produced by a strain of S. exfoliatus that is not known to produce β-lactams (Kim & Lee, 1994). BLIP-I is similar to the S. clavuligerus BLIP, while BLIP-II is quite different. Disruption of either of the S. exfoliatus genes impairs colony differentiation, and at least in the case of the BLIP-I-defective mutant, the morphological defect can be ‘complemented’ by growth close to the wild type (Kang et al., 2000; Kim et al., 2008b). The BLIP-II mutant also has abnormalities in sporulation-associated septation and DNA condensation, raising the question of whether it also has some intracellular activity – a possibility also suggested by its close structural relatedness to the human regulator of chromosome condensation, RCC1 (Kim et al., 2008b).

Some streptomycetes produce proteins that inhibit other hydrolytic exoenzymes, such as the α-amylase inhibitor tendamistat produced by Streptomyces tendae (Koller & Riess, 1989). There is evidently a remarkably extensive level of entirely extracellular regulatory processes, but the extent to which they directly serve the biology of the producing organism, as opposed to interfering with the activities of neighbouring organisms, remains to be determined.

The diverse roles of secondary metabolites

In the preceding section, extracellular proteinaceous inhibitors of enzymes were discussed, but the extravagant secondary metabolism of streptomycetes includes the production of numerous small molecules (usually between 100 and 3000 Da) that are biologically active outside the producer cell, many being antibiotics that inhibit enzymes and cellular processes. At least 7000 different secondary metabolites have been discovered in Streptomyces isolates (Bérdy, 2005), and when natural product chemists interrogate the genomes of streptomycetes and closely related organisms, they usually find 20–30 or more gene sets for secondary metabolism, of which perhaps 30% are for antibiotic biosynthesis (Bentley et al., 2002; Ikeda et al., 2003; Challis, 2008).

A high proportion of gene clusters for secondary metabolism show strain specificity. Together with the sporadic distribution of some pathways across diverse streptomycetes (e.g. streptomycin biosynthesis: Egan et al., 1998), this implies an involvement of lateral gene transfer in establishing the secondary metabolic profile of streptomycetes. There are several examples to support the notion that lateral transfer of antibiotic biosynthetic genes, and even some aspects of pathway evolution, may be mediated by large linear plasmids, which are common among streptomycetes and that can sometimes integrate into the chromosome (Chater & Kinashi, 2007).

Here, we discuss the possible biological functions of secondary metabolites. In addition to those discussed, secondary metabolites produced by streptomycetes include volatile odour compounds such as geosmin and methylisoborneol, both of which are widespread among streptomycetes, although little is known about their functions (e.g. Komatsu et al., 2008). A thoughtful analysis of the possible reasons for the production of such large numbers of secondary metabolites was presented by Challis & Hopwood (2003).


Iron is essential for the growth of all living cells, but its usual Fe3+ form is insoluble and therefore unavailable. Most bacteria acquire iron through the agency of secreted siderophores, typically nonribosomally synthesized small peptides with an extremely high affinity for Fe3+. Iron–siderophore complexes are taken into the cells by highly efficient transport systems. Unusually, it has emerged from genome mining that it is common for streptomycetes to produce more than one kind of siderophore (Challis & Hopwood, 2003; Barona-Gómez et al., 2006). For example, S. coelicolor produces coelichelin and desferrioxamine, while S. tendae produces desferrioxamine and enterobactin. It has been argued that in the soil environment, competing bacteria may be able to exploit each other's siderophores by possessing siderophore uptake systems with appropriate specificity, and that the use of more than one siderophore reduces the advantage of such competitors (Challis & Hopwood, 2003). Evidence of an interspecies interaction was provided by the finding that growth and sporulation of Streptomyces tanashiensis colonies can be stimulated by growth near S. griseus colonies, and that this is entirely attributable to desferrioxamine produced by S. griseus (Yamanaka et al., 2005).

Adaptive benefits of antibiotics

Production of most antibiotics is more or less species specific. It seems reasonable to consider that the secondary metabolome of any one organism provides it with an effectively unique ability to compete against some of the other microorganisms that it may encounter, including closely related streptomycetes. However, because antibiotic production is usually delayed until most of the growth has been completed, some antibiotics may serve to defend colony biomass against overgrowth by other organisms during the autolysis that accompanies development, rather than to help in competition for the primary biomass accumulation (Chater & Merrick, 1979). In one interesting case, methylenomycin has been reported to inhibit aerial growth more potently than vegetative growth of sensitive streptomycetes (Vivian, 1971; Chater & Hopwood, 1989).

Roles of antibiotics in symbiotic interactions

The adaptive benefits of antibiotic production can sometimes be more subtle and involve interactions with more than one organism. For example, there are several examples of antibiotic production by Streptomyces symbionts being a key feature of the symbiotic interactions, providing protection against unwanted infection of a symbiotic partner by other microorganisms. A polyketide produced by the sponge symbiont Streptomyces dendra MSI051 preferentially inhibits biofilm-forming bacteria that might be detrimental to the host sponge (Selvin, 2009). Production of the well-known polyene antifungal agent candicidin by streptomycetes associated with leaf-cutting ants can inhibit fungi that might otherwise overgrow the ants' fungus farms (Currie et al., 2003; Mueller et al., 2008; Haeder et al., 2009). Another polyene antifungal agent, mycangimycin, is produced by a streptomycete symbiont of pine bark beetles. Mycangimycin antagonizes fungal attacks on symbiotic fungi growing in a special organ (mycangium) as food for the beetle larvae (Scott et al., 2008). Protection against fungal attack on the larvae of the hunting wasp Philanthus triangulum is afforded by the presence on the cocoon of a streptomycete that has been transmitted from a special organ in the adult's antennae (Kaltenpoth et al., 2005; Goettler et al., 2007).

There are also examples of antibiotics influencing symbioses with plants. Thus, antibiotics of some plant-associated streptomycetes appear to provide protection to the plant against pathogens, while in return the streptomycetes presumably benefit from the plant exudates (Castillo et al., 2002; Tokala et al., 2002), and undefined substances secreted by mycorhizal helper streptomycetes appear to promote the plant–mycorhizal symbiosis by affecting the branching pattern of the fungal hyphae (Schrey et al., 2007).

Secondary metabolites can also be involved in pathogenicity. This is particularly clear in the case of thaxtomins and potato scab, which was discussed earlier. It has also been reported that the symptoms of a disease of fish lungs induced by a strain of S. griseus can be reproduced by extracts from agar culture medium, although it is not known whether this effect is attributable to a secondary metabolite (Lewis et al., 2008).

Antibiotics as developmental regulators and signalling molecules

Early reports suggested that some antibiotics might play roles in controlling the development of the producer. For example, the macrolide pamamycin-607 is needed for the aerial growth and sporulation of Streptomyces alboniger (McCann & Pogell, 1979). This effect seems to involve the supply of calcium to aerial hyphae, with pamamycin perhaps acting as a calcium siderophore (Natsume et al., 1995; Hashimoto et al., 2003). Other antibiotics, notably germicidin and hypnosin, can act as inhibitors of spore germination, perhaps providing a means to prevent spore germination in situ in spore chains, and to synchronize germination of locally abundant spores (Grund & Ensign, 1985; Peterson et al., 1993; Aoki et al., 2007; Challis, 2008). It has further been suggested that antibiotics may even have originated as signalling molecules, because many antibiotics can induce striking changes in the expression of some genes that are not related to general stress responses, when added at levels well below those needed to inhibit growth (Goh et al., 2002; Davies et al., 2006; Yim et al., 2007). An example mentioned earlier is the induction of some chitinases by low concentrations of allosamidin via a two-component regulatory system (Suzuki et al., 2008), and a nonactinobacterial example is provided by the nonribosomal peptide antibiotic surfactin in Bacillus subtilis (López & Kolter, 2010). However, considering the noncongruence of host phylogeny with the distribution of most antibiotic pathways, which suggests extensive lateral transfer (Egan et al., 1998; Chandra & Chater, 2008), it is hard to to imagine that relatively recently acquired pathways can generally be sufficiently assimilated into the host's biology to take on immediate, adaptively significant signalling roles within the host. Many of the reported ‘signalling’ effects are mediated via the normal target of the antibiotic action. It may be sufficient to rationalize at least some such cases by assuming that the balance between the range of active states of the multiprotein complexes that are typical antibiotic targets (e.g. macromolecular biosynthesis machinery such as ribosomes) can be significantly shifted by small changes in the stability of particular states. Small changes in such centrally active complexes are likely to have perceptible effects on global gene expression.

Antibiotics with redox activity

Many antibiotics are redox-active pigments, including the red-blue actinorhodin and red prodiginines of S. coelicolor. Dietrich et al. (2008) have suggested, on the basis of mutant phenotypes, that at least the prodiginines may serve as sensors of redox stress, and then interact with SoxR to activate an appropriate response that is reflected both in levels of expression of the target genes and in a change in the growth pattern of colonies. However, in the absence of evidence from complementation to demonstrate that these effects are all attributable to the same mutation, these inferences should be considered provisional.

Spore pigments

Most streptomycetes have pigmented spores, whose colour has been used as a taxonomic characteristic. Two chemically quite different kinds of spore pigments have been described. One very widespread type, which is responsible for a range of spore colours, is synthesized by a type II polyketide synthetic route, resulting in polycyclic aromatic molecules (Davis & Chater, 1990; Metsä-Keteláet al., 2002). The other is synthesized by a type III polyketide synthetic route to generate a kind of melanin (Funa et al., 1999, 2005). At least in the latter case, the spore pigment seems to provide a small increment of UV protection. The difficulty of extracting pigments from spores indicates that the pigments are secreted from aerial hyphae and then covalently linked to the spore walls. Conceivably, this may contribute to the resistance of the walls to enzymatic digestion, either autolytically or by digestive enzymes of animals that might consume spores (e.g. earthworms, springtails, bacteriovorous protists, etc.) (Bloomfield & Alexander, 1967; Chater & Chandra, 2006). Because polyketide spore pigments are very widespread among streptomycetes, they were presumably present in the ancient progenitor of modern steptomycetes. This suggests that type II polyketide antibiotic synthesis may have originated from such a cluster (Chater & Chandra, 2006).

Some antibiotic biosynthetic enzymes are extracellular

Although it has often been assumed that antibiotic biosynthesis is an intracellular process, which ends in the export of the mature antibiotic, there is some surprising evidence for an extracellular location of a few true biosynthetic enzymes.

The production of antibiotics carries with it the potential to commit suicide, a danger averted by a variety of mechanisms (Cundliffe, 1989). In some pathways, the late intermediates are modified to keep them inactive, as in the cases of oleandomycin and methymycin (Quirós et al., 1998; Zhao et al., 2003). In these cases, the glycosylated (inactive) end-product is exported and extracellular enzymes remove the modification. Such extracellular enzymes are biochemically ‘acceptable’ because they are not dependent on activated substrates or nucleotide-based cofactors to provide energy or reducing power. However, in the case of chromomycin, the final acetylation step in biosynthesis to yield an active product appears to be carried out by a membrane-bound enzyme, after export of the deacetylated form via an ABC transporter (Menéndez et al., 2007). Also, an enzyme required for one of the late actinorhodin-tailoring steps (ActVI-ORF3, a dehydratase) is found specifically in the culture supernatant, in a form lacking the secretion signal peptide predicted from the gene sequence (Hesketh et al., 2002; Hesketh & Chater, 2003). Even more surprisingly, the NADPH-dependent ActIII ketoreductase protein, whose enzymatic activity determines the nature of the ring closure during formation of the actinorhodin backbone, is covalently attached to the cell wall, despite the absence of any obvious secretion signal sequence (Xu et al., 2008b). In addition, two FAD-dependent enzymes for biosynthesis of an uncharacterized polyketide, encoded by SCO6272 and SCO6281, are exported via the Tat system of S. coelicolor (Widdick et al., 2006). It would not normally be expected that reducing enzymes could function in an extracytoplasmic context; but there is evidence of an extracellular pool of reducing power in bacteria (Ju et al., 2005; Wos & Pollard, 2006).

In S. coelicolor strains manipulated to accumulate different amounts of extracellular STI protease inhibitor, actinorhodin and STI levels are inversely related (Kim et al., 2008a). It will be interesting to know whether this effect involves the extracellular steps in actinorhodin biosynthesis.

Extracellular signalling by small molecules and proteins

In pioneering early discoveries, the extracellular γ-butyrolactone signalling molecule A-factor was identified in culture fluids of streptomycin-producing S. griseus (Khokhlov et al., 1967), and investigators in Debrecen discovered a protein, factor C, that stimulated partial development in liquid cultures of a supposed S. griseus strain (Szabóet al., 1962). A-factor was subsequently comprehensively investigated by Beppu and colleagues at the University of Tokyo. The topic has been extensively reviewed recently (Ohnishi & Horinouchi, 2004; Takano, 2006; Nishida et al., 2007), and we summarize the major features here. The literature on factor C is reviewed here for the first time, and therefore in more detail. It appears that horizontal gene transfer has led to the acquisition of many of the signalling molecules.

Diffusible autoregulators

Biosynthesis, mode of action and physiological role of γ-butyrolactones. A-factor is produced by the condensation of 8-methyl-3-oxononanoyl-acyl carrier protein and the hydroxyl group of dihydroxyacetone phosphate, catalysed by the product of the gene afsA (Kato et al., 2007). This gene is readily lost at least from the most-studied strain, along with the chromosome end near which it is located (Lezhava et al., 1997), with the result that production of streptomycin and other secondary metabolites, as well as morphological differentiation, are lost, but can be restored by the external addition of A-factor. In many other streptomycetes, similar γ-butyrolactones are produced, but they have some degree of species specificity in stereochemistry and side-chain length and structure – thus, A-factor itself is not often encountered outside of S. griseus strains.

Under laboratory conditions, A-factor accumulates extracellularly during growth (it is thought to be more or less freely diffusible through membranes), and when its concentration reaches nanomolar levels, it binds to a cytoplasmic binding protein (ArpA), interfering with the ability of ArpA to bind to, and repress, the promoter of the key regulatory gene, adpA. AdpA activates a large regulon of genes, including the activator gene for streptomycin biosynthesis and several genes playing contributory roles in the early stages of morphological development, notably encoding some of the participants in the extracellular protease cascade described above.

Although adpA is universally present in streptomycetes, in most species it is not regulated by γ-butyrolactones. Instead, the γ-butyrolactone systems often seem to be specific to the pathway genes for particular antibiotics, and the factor biosynthetic gene(s) and the gene for the regulatory factor-binding protein are frequently located next to each other and to the pathway genes for the relevant antibiotic (Takano, 2006). (A-factor is a rare exception to this rule, arpA being located away from both afsA and the biosynthetic genes for streptomycin.)

Nevertheless, in several cases, deletion of the factor biosynthetic gene has pleiotropic effects, causing defects in colony differentiation and activation of other biosynthetic pathways (Folcher et al., 2001; Wang & Vining, 2003; Lee et al., 2008). It is not known whether these effects are mediated via AdpA.

In most cases, although apparently not in the case of afsA, the factor-binding protein also represses the factor biosynthetic gene(s), so that the factor accumulates only slowly, reaching a level high enough to lift autorepression only at a relatively high cell density. Then, the resulting increase in factor biosynthetic gene expression causes increased factor production and hence precipitates a population-wide activation of target genes (at least in that part of the population physiologically competent to respond – the regulation of secondary metabolism is very complex, involving diverse checkpoints mediated by a large number of genes: Chater & Bibb, 1997; Bibb, 2005). Strain specificity in the structures and recognition of autoregulatory factors presumably permits the factor-regulated functions of individual strains, such as production of specific antibiotics, to be regulated autonomously in the highly mixed communities in soil. It also implies acquisition by horizontal gene transfer, and indeed Nishida et al. (2007) provided phylogenetic evidence that the factor biosynthetic (afsA-like) and recognition (arpA-like) genes generally have not coevolved over long time periods, but rather were independently acquired.

Other diffusible autoregulatory factors. Other kinds of autoregulatory factors are emerging. Thus, an extracellular factor involved in the regulation of methylenomycin biosynthesis was found to be alkali resistant (unlike γ-butyrolactones), leading to the discovery of a new class of furan autoregulators (Corre et al., 2008). The furan biosynthetic and factor-binding proteins are closely related to those for γ-butyrolactones. A further kind of autoregulator, P1 factor [2,3-diamino-2,3-bis(hydroxymethyl)-1,4-butanediol], which is still uncharacterized biosynthetically and genetically, can activate pimaricin biosynthesis in a mutant of Streptomyces natalensis (remarkably, it is interchangeable with A-factor in this system) (Recio et al., 2004).

Factor C: an extracellular ‘signalling’ protein

Factor C was isolated from the culture fluid of ‘S. griseus 45H’, a strain able to sporulate in liquid culture. This strain was recently reclassified as S. albidoflavus (Kiss et al., 2008). The effect of factor C is detectable at very low concentrations (c. 1 ng mL−1) (Szeszak et al., 1990). Biochemical and DNA sequence information indicate that prefactor C has a 38-aa secretory Tat signal sequence (Biróet al., 1980, 1999; Szabóet al., 1999). It is therefore likely that factor C acquires its tertiary structure in a cytoplasmic context before export via the Tat system.

Mode of action. The mode of action of factor C remains a mystery. However, it is remarkable that A-factor production and associated attributes were restored to an uncharacterized nonsporulating, A-factor-negative (AFN) mutant of a strain of S. griseus (B2682, which naturally lacks a facC gene) by the introduction of facC (Birkóet al., 2007). [It is not obvious why the mutant lacks A-factor, because it has an intact copy of afsA that is expressed at an elevated level compared with the wild-type strain: this suggests that the mutation in the AFN strain lies in another yet unknown gene involved in the control of A-factor biosynthesis (Birkóet al., 2007).]

The bald AFN mutant overexpresses several ABC transporter solute-binding proteins and stress response proteins compared with the wild-type S. griseus B2682 strain or with the facC transformant of the AFN mutant, and it was suggested that these proteins helped the mutant to compensate for the loss of nutrients normally supplied by proteolysis (Birkóet al., 2009). A similar set of proteins was released by an adpA mutant of S. griseus, but this was associated with the higher level of cell death of the mutant compared with the wild type or a complemented mutant, leading to the extracellular accumulation of these proteins, which, it was speculated, might be particularly resistant to proteolysis (Akanuma et al., 2009).

Unusual distribution of facC-like genes. Southern blotting and genome sequence analysis show that facC is absent from most streptomycetes, although present in two Streptomyces albus strains (Birkóet al., 1999) and most S. albidoflavus isolates (Kiss et al., 2008). This indicates that the gene may have been acquired at the time when S. albidoflavus emerged about 140 million years ago (using 1% difference in 16S rRNA gene sequence as representing approximately 50 million years' divergence: Ochman et al. (1999) and A.C. Ward, pers. commun.). Other sequences related to facC can be found in three main phylogenetically distinct groups of microorganisms: frequently in low-GC Gram-positive bacteria, especially staphylococci, lactococci and their phages, where they are generally part of larger proteins, and occasionally in streptomycetes and related actinomycetes, and – quite surprisingly – in some mycelial fungi (Fig. 3). All the actinomycete and fungal proteins contain just a FacC-like domain, and are more closely related to each other than to the firmicute/phage proteins. Noticeable exceptions to this are two genes encoding lactobacterial bacteriocins (helveticins: Joerger & Klaenhammer, 1990), which fall into the actinmycete/fungal subfamily both phylogenetically and in size. Members of this subfamily all seem to be secreted, but the Tat-directed signal sequence of the Streptomyces prefactor C proteins is replaced in the fungal and helveticin proteins by sequences to signal export by the major secretion system (signalp analysis, Bendtsen et al., 2004). They may therefore be secreted in an unfolded form, to be folded in an extracytoplasmic environment.

Figure 3.

 Phylogenetic tree of proteins related to factor C. The sources of the proteins are colour-coded as follows: black, streptomycetes and other mycelial actinomycetes; dark red, mycelial fungi; blue, phages of Firmicutes; light red, firmicute bacteria. The factor C protein sequences from different species were selected to be representative of a larger collection. The Bacillus subtilis‘TagC’ protein was used as the outgroup in phylogenetic analysis. The protein sequences were aligned using the program t-coffee with default settings (Notredame et al., 2000). Tools from the phylip package version 3.67 (Felsenstein, 1989, 2005) were used for various phylogenetic tasks (all tools were used with their default settings): seqboot, for bootstrapping the alignment into 100 datasets; protdist, to calculate distances between sequences; neighbor, to calculate trees from the distance data; consense, to arrive at a consensus tree from the trees produced by neighbor. A dendroscope (Huson et al., 2007) was used to view the consensus tree in different ways. Key: Fungi: AJEDE, Ajellomyces dermatitidis; ASPFU, Aspergillus fumigatus; CHAGB, Chaetomium globosum; NEOFI, Neosartorya fischeri; PARBR, Paracoccidioides brasiliensis; PENCW, Penicillium chrysogenum; PODAN, Podospora anserina. Firmicutes: BACSU, B. subtilis; LACH4, Lactobacillus helveticus Helveticin; LACSK, Lactobacillus sakei ssp. carnosus Helveticin; LISMO, Listeria monocytogenes; Staphylococcus aureus strains: STAAE, strain Newman; STAA2, strain JH1; STAAR, strain MRSA252. Staphylococcus phages: Q4ZDU1, phage 69; Q4ZAV3, phage 52A; Q4ZBA3, phage 55; Q4ZCZ9, phage 42E; Q4ZCK5, phage 47; Q9BOC7, phage phiSLT; Q4ZE58, phage 66. Actinomycetes: Saery, Saccharopolyspora erythraea; Salbus, Streptomyces albus G; Sscab, Streptomyces scabies; STRGR, Streptomyces griseus; STRCL, Streptomyces clavuligerus; Strmg1, streptomyces_sp._mg1; THECU, Thermomonospora curvata DSM 43183.

The implied lateral gene transfer between streptomycetes and mycelial fungi might be facilitated by their being close coinhabitants of the soil with similar ecological adaptations. We have already described batteries of chitinases that help streptomycetes to digest fungal walls, and many fungi make β-lactam antibiotics active against bacterial cell wall biosynthesis. This potential for mutual protoplast production might facilitate occasional exchange of DNA. Indeed, the fungal genes for β-lactam biosynthesis are themselves likely to have been acquired by lateral transfer from a progenitor of one of the many kinds of soil bacteria (including a number of actinomycetes) that synthesize β-lactams (Aharonowitz et al., 1992).

Surface-active proteins involved in the attachment of hyphae to surfaces and aerial hyphal growth

The action of many bld genes of S. coelicolor leads eventually, via the previously mentioned extracellular signalling cascade of Willey et al. (1993), to the production of extracellular proteins that are directly involved in the physical process of aerial growth: SapB (Tillotson et al., 1998), and chaplin and rodlin proteins that make up the fibrous sheath material described in early ultrastructural studies of aerial mycelium (Glauert & Hopwood, 1961; Wildermuth et al., 1971; García, 1995; Claessen et al., 2002, 2003, 2004; Elliot et al., 2003; Elliot & Talbot, 2004). The very high surface activity of all three kinds of proteins has led to the widely quoted idea that their main role is to allow the aerial hyphae to breach the surface tension of the aqueous environment surrounding growing vegetative hyphae (reviewed in detail by Wösten & Willey, 2000). Nevertheless, because the proteins continue to cover the aerial hyphae after their emergence, they must enclose a hydrated compartment that includes the cell wall. It is hard to avoid the inference that this cylindrical compartment forms a channel through which nutrients and other metabolites can diffuse from the substrate to the growing tips of the aerial hyphae (Fig. 4; Chater & Chandra, 2006). A similar possibility also exists for the analogous hydrophobins of some fungi.

Figure 4.

 Roles of surface-active proteins in surface adherence of vegetative hyphae and growth of aerial hyphae. At points of contact between substrate hyphae and a hydrophobic surface, cellulose fibrils (orange) are extruded and act as nucleators for the assembly of amyloid complexes of chaplins (black dots) to form fimbriae, which serve as attachment organelles (De Jong et al., 2009b). Cellulose fibrils are also extruded at aerial hyphal tips, where a complex of cellulose synthase (CslA, pale blue) with the polarity determinant DivIVA (dark red) is located (Li et al., 2007). Because the tip-located cellulose appears to help chaplins to form the hydrophobic sheath of aerial hyphae, the model tentatively assumes that this is via the formation of fimbria-related structures. Once formed, the hydrophilic inner face of the sheath is predicted to hold a water column (presumably largely a water-saturated cell wall) in a kind of periplasmic compartment. On the further assumption that the sheath is open at the growing tip, we speculate that the evaporation of water from the tip will bring about transpiration, and hence carry nutrients from the base to the upper parts of the aerial hypha, while maintaining hydration of the aerial hypha.

SapB and its equivalents in other streptomycetes resemble lantibiotics

A detailed account of SapB was presented by Willey et al. (2006). In summary, SapB, which comprises just 21 aminoacyl residues, is detectable only on rich media (Capstick et al., 2007). It is generated by complex processing of the 42-aa product of the ramS gene, to generate lanthionine cross-links in steps similar to those involved in the biosynthesis of lantibiotics (a class of ribosomally synthesized peptide antibiotics containing modified cysteine residues cross-linked by lanthionine bridges). The adjacent ramC gene encodes a multifunctional enzyme that appears to carry out the modification steps, generating pro-SapB. A specific transporter (the ramAB product) is thought to be the main exporter of SapB, although it is not clear whether the 21 aa of the leader sequence are cleaved off before, during or after export. SapB forms a monolayer with hydrophobic and hydrophilic phases at air–water interfaces. This physical property is presumably central to facilitating hyphal growth into the air. Alignment of SapB-like peptides encoded by other Streptomyces genomes shows that the leader sequences are fairly highly conserved, while within the sequences present in the mature peptide there is greater conservation of the residues involved in modification than of the intervening residues, implying general structural similarity. The SapB equivalent of S. griseus is known as AmfS (Ueda et al., 2002, 2005). The role of SapB in aerial growth and sporulation can be substituted by a somewhat diverged lanthionine-containing amphipathic peptide (SapT) from S. tendae (Kodani et al., 2005); indeed, the completely unrelated (but functionally equivalent) hydrophobin protein SC3 of the fungus Schizophyllum commune can restore aerial mycelium formation to SapB-less bld mutants of various classes (Tillotson et al., 1998), and in the case of ram mutants, SC3 can even restore sporulation (Kodani et al., 2005).

Chaplins, amphipathic proteins that can substitute for SapB-like peptides under some growth conditions

During growth on minimal medium, SapB is not produced, but on rich medium, both SapB and the chaplins seem to be necessary for aerial growth (probably accounting for the medium dependence of the morphology of many bld mutants: Capstick et al., 2007). A mutant lacking SapB and all the chaplins is unable to form aerial mycelium on any medium tested, reinforcing the importance of these surfactants for aerial growth. The number of chaplins varies among species, but S. coelicolor has eight. Three of these (the ‘large’ chaplins A, B, C) have – along with an additional chaplin domain – C-terminal extensions that contain recognition sites for sortase enzymes. At least two, and perhaps as many as seven, sortases are encoded in the S. coelicolor genome (Pallen et al., 2001). Through their sortase recognition sites, the large chaplins are covalently attached to the cell wall, providing a scaffolding to which the ‘small’ chaplins (devoid of such cell wall-anchoring domains) can attach (although even in the absence of the large chaplins, the small chaplins are able to assemble and promote aerial growth and sporulation) (Claessen et al., 2003; Elliot et al., 2003). Extensive functional redundancy among chaplins means that multiple gene knockouts are necessary in order to detect their importance for aerial growth (Claessen et al., 2003, 2004; Elliot et al., 2003; DiBerardo et al., 2008). A strain carrying just the three chaplin genes chpC, E and H– the only ones conserved in all sequenced Streptomyces genomes, and the only ones expressed more or less constitutively, even during vegetative growth – showed almost wild-type aerial mycelium development. This was consistent with the following model: ChpC provides a sortase-linked anchor to the cell wall, on which ChpH assembles (in a manner shown to depend on its conserved cysteine residues), while ChpE, the only chaplin lacking the conserved cysteines, plays a supporting role in ChpH assembly (DiBerardo et al., 2008). It has been suggested that the assembly of a sheath layer contributes regulatory information necessary for the full development of sporulating aerial mycelium (the ‘sky pathway’; Claessen et al., 2006), and in this connection it is interesting that 17 of the genes downregulated in a chaplinless mutant are also downregulated in a bldA mutant (De Jong et al., 2009a).

Rodlins: chaplin-associated proteins involved in the surface architecture of spores

Extraction of spores with trifluoracetic acid removes not only the small chaplins but also rodlin proteins. Rodlins are encoded by a pair of adjacent genes that are also close to some of the chaplin genes. They interact with fibres of chaplin polymers to organize them into paired rodlets that give a basketwork-like appearance to the surface of S. coelicolor spores (Wildermuth et al., 1971; Claessen et al., 2002, 2004). Such paired rodlets are absent from the smooth spores of S. avermitilis, consistent with the absence of rodlin genes from its genome. This evidence of the dispensability of rodlins for sporulation is reinforced by the absence of overt phenotypic effects when the rodlin genes are deleted (Claessen et al., 2002). Other streptomycetes have elaborately decorated spores (e.g. spiny, warty or hairy in appearance) (Miyadoh, 1997), and it remains to be established whether these ornamentations are attributable to rodlins in those organisms, and whether they have a biological role such as attachment of spores to surfaces.

Interplay of surface-active proteins with cellulose

Calcofluor white is used for the cytological detection of β-glucans such as cellulose and chitin. It stains S. coelicolor hyphae preferentially at tips, which are the sites of active cell wall extension (Xu et al., 2008a). A cellulose synthase protein (CslA) is also located at hyphal tips. Structural predictions indicate that an extracytoplasmic C-terminus is connected via seven transmembrane domains to a cytoplasmic N-terminus, which appears to interact with the tip-located polarity determinant DivIVA as judged by two-hybrid analysis. The deletion of the CslA determinant (cslA, SCO2836) causes the loss of β-glucans from hyphal tips. It also generates a bald colony phenotype that can be partially corrected by the addition of an extract highly enriched for chaplins and rodlins. This led Xu et al. (2008a) to suggest that β-glucans synthesized by CslA function like a kind of bandage, to stabilize the nascent sheath material as the chaplins and rodlins are released during rapid aerial growth (if, as seems likely, they are secreted from the aerial hyphal tip). However, in view of recent studies of the roles of chaplins and cellulose in the fimbria-mediated attachment of hyphae to surfaces, a slightly different model can be proposed. The attachment of S. coelicolor hyphae to the hydrophobic surface of the culture vessel in static liquid culture (Van Keulen et al., 2003) is considerably reduced in a chaplinless mutant, attachment being mediated by extracellular fimbriae that are largely composed of chaplins (De Jong et al., 2009b) (Fig. 4). The fimbriae emerge from small spikes in the lateral walls of hyphae and are absent from the chaplinless mutant, although the spikes remain. Remarkably, the spike structures are also the site of attachment of much finer fibrils, seen most readily in a chaplinless mutant. Production of these fibrils depends on cslA, and they are eliminated by treatment with cellulase. Similar treatment of the wild type results in detachment of fimbriae from the hyphae and loss of the ability to attach to surfaces. In vitro, purified chaplins can assemble in bundles of amyloid fibres to form fimbria-like structures. It has been proposed that in vivo the chaplins aggregate into amyloids that may then assemble onto the cellulose fibrils to form fimbriae. The functionally similar fungal hydrophobin SC3 depends on glucan polymers for effective polymerization (Scholtmeijer et al., 2009). Mutants in cslA show only partial loss of attachment to surfaces, indicating that some cellulase-sensitive material that is made independently of CslA is also involved. In support of this, fimbriae of a cslA mutant showed moderate fluorescence when exposed to calcofluor white (De Jong et al, 2009b).

These results show that chaplins and cellulose fulfil at least two distinct functions in streptomycetes. The hydrophobins of mycelial fungi also play these two roles (Wösten & Willey, 2000). It seems reasonable to speculate that, in the early evolution of the genus, the use of such surface-active proteins in attachment (i.e. biofilm formation) may have preceded their role in facilitating aerial growth, and that the cellulose produced at growing aerial hyphal tips may provide a nucleation point for chaplin amyloids similar to that involved in attachment (Fig. 4 shows a naïve model incorporating this possibility). It will be interesting to investigate whether chaplins and/or rodlins could also play a further role by conferring resistance to grazing by bacteriovorous protists, as has been observed for LiCl-extractable surface material of other actinobacteria (Tarao et al., 2009).

As mentioned earlier, Smucker & Pfister (1978) suggested that the fibrous sheath had a chitinaceous component, a claim recently echoed by a study showing that chitin is present on the spiny spores of Streptomyces lunalinharesii (Gomes et al., 2008). Although it is difficult to square these suggestions with the absence of obvious chitin synthesis determinants from the S. coelicolor genome, the suggestions are reminiscent of the observations of De Jong et al. (2009b), implicating some other β-glucan as well as cellulose in fimbriation.

Concluding remarks

We believe that this is the first attempt to review the extracellular biology of streptomycetes. It is inevitably incomplete, but whatever the shortcomings of the survey, one must be impressed by the range and complexity of the phenomena described. It seems that the evolution and speciation of streptomycetes has involved continual interplay between the advantages and opportunities offered by controlling or influencing the extracellular environment, and the new evolutionary opportunities available within such autogenously created environments. Future investigations into this theme will probably find greater integration of the extracellular biology than we have been able to delineate.

Most of the aspects of extracellular biology described here appear to be associated with high population density/stationary phase. As reviewed by Chater & Chandra (2008) and Chandra & Chater (2008), many of these aspects are dependent on the UUA-reading tRNA encoded by the bldA gene. In some cases, this is because of a TTA codon universally present in the key pleiotropically acting regulatory gene adpA, while in other cases it is because of TTA codons in genes more specifically associated with particular processes (three-quarters of gene clusters for antibiotic production contain at least one TTA-containing gene, most often regulatory: Chater & Chandra, 2008). Remarkably, the abundance of this tRNA does not follow the pattern typical of other tRNAs, which are usually most abundant during rapid growth: instead, the bldA tRNA is maximally abundant at the onset of stationary phase (Leskiw et al., 1991; Trepanier et al., 1997). These considerations led to the proposal that the possession of TTA codons might result in a sensitivity to a checkpoint signalled by increased bldA tRNA levels. TTA-containing genes, which include many of those affecting the extracellular biology of streptomycetes, would thus depend on an ‘expression space’ guarded by bldA (Chater & Chandra, 2008).

Many extracellular proteins are not dependent on bldA for expression (as shown, for instance, by the existence of a substantial extracellular proteome in a bldA mutant). These bldA-independent proteins include many that are widespread or universal among streptomycetes, such as chaplins, rodlins and many chitinases. These are encoded by genes that may well have become important to the ancient progenitor of modern streptomycetes before the evolution of bldA-dependent expression space.

Presumably, the profile of species-specific (i.e. horizontally acquired) extracellular activities is an important determinant of the specificity and flexibility of the ecology of each organism. Unrelated coinhabitants of a common niche may be a source of adaptively useful DNA for streptomycetes (a source made more accessible by the ability of streptomycetes to kill and lyse bacteria). Further lateral transfer among streptomycetes would be readily achievable through the abundant transfer-proficient plasmids of streptomycetes, which frequently have chromosome-mobilizing ability.


We thank Govind Chandra for carrying out bioinformatic analysis.


During the preparation of this article, we learned of the sad premature passing of Professor Sueharu Horinouchi, who contributed profoundly to our knowledge of Streptomyces extracellular biology. We dedicate the article to his memory.