ORIGINAL PAPER

D. Schneider á C. J. Bruton á K. F. Chater

Duplicated gene clusters suggest an interplay of glycogenand trehalose metabolism during sequential stagesof aerial mycelium development in Streptomyces coelicolor A3(2)

Received: 12 November 1999 /Accepted: 31 January 2000

Abstract DNA sequencing and operon disruption ex-periments indicate that the genes glgBI and glgBII,which code for the two developmentally speci®c glyco-gen branching enzymes of Streptomyces coelicolor A3(2)each form part of larger duplicated operons consisting ofat least four genes in the order pep1-treS-pep2-glgB. Thesequences of the TreS proteins are 73% identical (93%similar) to that of an enzyme that converts maltose intotrehalose in Pimelobacter, a distantly related actino-mycete; and the Pep1 proteins show relatedness to thea-amylase superfamily. Disruptions of each operon havespatially speci®c e�ects on the nature of glycogen de-posits, as assessed by electron microscopy. Upstream ofthe glgBI operon, and diverging from it, is a gene (glgP)that encodes a protein resembling glycogen phosphory-lase from Thermatoga maritima and a homologue inMycobacterium tuberculosis. These three proteins form adistinctive subgroup compared with glycogen phos-phorylases from most other bacteria, which more closelyresemble the enzymes from eukaryotes. Diverging fromthe glgBII operon, and separated from the pep1 gene bytwo very small ORFs, is a gene (glgX) encoding aprobable glycogen debranching enzyme. It is suggestedthat most of these gene products participate in the de-velopmentally modulated interconversion of immobile,inert glycogen reservoirs, and di�usible forms of carbon,both metabolically active (e.g. glucose-1-phosphate

generated by glycogen phosphorylase) and metabolicallyinert but physiologically signi®cant (trehalose).

Key words Glycogen á Trehalose á Sporulation

Introduction

True mycelial growth in bacteria appears to be con®nedto high G + C gram-positive bacteria (actinomycetes).This growth habit is probably an adaptation that facil-itates the utilisation of insoluble substrates for growth.In streptomycetes, which have a particularly complexmorphology (Chater and Losick 1997), the resultingmycelial biomass (the substrate mycelium) itself even-tually supports a further round of growth (Me ndez et al.1985), to generate a more or less parasitic aerial myce-lium, in which many hyphae turn into long chains ofspores. There is no detailed information about the ge-netic and physiological mechanisms that must controlthe reuse within hyphae, during the sequential mor-phological phases of development, of nutrient resourcesaccumulated as biomass and storage materials in thesubstrate mycelium.

Two forms of carbohydrate likely to be implicated inthe mobilisation of accumulated carbon resources inStreptomyces are the polysaccharide glycogen (mainlya-1,4-linked glucosyl residues) and the non-reducingdisaccharide trehalose (glucose a-1,1-glucose). Glycogenaccumulates biphasically over time (e.g. Karandikaret al. 1997; MartõÂ n et al. 1997). This is correlated with itsobserved accumulation in two locations in colonies: inthe older parts of the substrate mycelium, from whichaerial hyphae emerge (phase I accumulation); and inimmature spore chains (phase II accumulation) (BranÄ aet al. 1986; Bruton et al. 1995; Plaskitt and Chater 1995;MartõÂ n et al. 1997). Glycogen appears to be virtuallyabsent from the stalks of aerial hyphae and the maturespores. On the other hand, trehalose appears to bepresent at all growth stages, though there are disparitiesbetween available data with respect to its relative

Mol Gen Genet (2000) 263: 543±553 Ó Springer-Verlag 2000

Communicated by A. Kondorosi

D. Schneider1 á C. J. Bruton á K. F. Chater (&)John Innes Centre, Norwich Research Park,Colney, Norwich NR4 7UH, UKE-mail: Chater@bbsrc.ac.ukTel.: +44-1603-450297; Fax: +44-1603-450045

Present address:1 Plasticite et Expression des Ge nomes Microbiens,CNRS EP2029, CEA LRC12, Universite Joseph Fourier,BP53, 38041 Grenoble Cedex 9, France

The ®rst two authors contributed equally to this paper

abundance at di�erent growth stages in di�erent strep-tomycetes: in S. antibioticus, trehalose levels are rela-tively high in aerial hyphae and, particularly, in spores(BranÄ a et al. 1986), whereas in S. coelicolor the highestlevels coincide with the most rapid growth phase (Ka-randikar et al. 1997). The periods of maximum abun-dance of the two carbohydrates given in publishedreports do not coincide. Is their metabolism indepen-dent, or are there metabolic links between the two? Forexample, glycogen could perhaps provide local carbonresource depots which could be moved from one loca-tion to another by interconversion with readily di�usiblebut metabolically rather inactive trehalose. In this paperwe describe evidence for close linkage and probablecotranscription of genes for glycogen and trehalosebiosynthesis in S. coelicolor A3(2), and argue that theremay be interconversion of these carbohydrates to sup-

port the needs of development. An analogy is drawnwith the roles of the polysaccharide starch and the non-reducing disaccharide sucrose in plants.

Materials and methods

Bacterial strains, plasmids, phages and culture conditions

Cosmids 6A11, 4G10 and 2C12, all of which contained approxi-mately 40-kb insertions of S. coelicolor A3(2) DNA in the vectorSupercos 1, were maintained in Escherichia coli SURE (see Re-denbach et al. 1996, for details). The S. coelicolor A3(2) derivativeJ1508 (hisA1 uraA1 strA1 pg1 SCP1NF SCP2); Ikeda et al. 1984)was the host for construction of disruptions of glgB gene clustersby insert-mediated single-crossover integration of the bacterio-phages FC31 KC515 (Rodicio et al. 1985) and KC862 (Brutonet al. 1991), which are deleted for the attP site. Details of thesedisruption mutants are shown in Fig. 1. S. lividans 1326 was thehost for transfections with ligation mixtures containing FC31DNA. Procedures for phage propagation, phage DNA isolation,transfection and transduction were as in Hopwood et al. (1985).For electron microscopic analysis of polysaccharide deposits(Thie ry 1967; BraÄ na et al. 1980; Bruton et al. 1995) colonies weregrown on agar minimal medium (MM; Hopwood et al. 1985)containing 0.5% mannitol instead of glucose as carbon source.

DNA manipulations

Standard procedures were used (Sambrook et al. 1989), unlessstated otherwise. For Southern analysis, probes were labelled non-radioactively using the DIG system (Boehringer Mannheim). Hy-bridization was at 65 °C and washing was at 65 °C in 0.1 ´ SSC,

Fig. 1 Pattern of homology between, and strategy for disruption of,genes upstream of glgBI and glgBII. The DNA extending some 5.5 kbupstream of glgBI and glgBII showed extensive homology whenSouthern blots of di�erent restriction digests were hybridized with thethree probe fragments indicated by the shaded bars. Sequence analysisrevealed the genetic organisation indicated by the dart-shaped boxes.The fragments indicated as black bars were inserted, in the indicatedorientations, into FC31 vectors (KC515 or KC862) and used to directinsertion of the vector at the homologous point in the genome. Thephenotypes of glycogen deposits in phases I and II are indicated in theboxes adjacent to the black bars. Examples of the correspondingelectron micrographs are shown in Figs. 2 and 3

544

0.1% SDS (these conditions are predicted to give strong signals forsequences with at least 90% identity to the probes, assuming aG + C content of 70%). DNA sequencing and subsequent com-puter analysis were done as in Bruton et al. (1995).

Results

Interruption of duplicated DNA upstreamof glgBI and glgBII has spatially speci®c e�ectson glycogen deposits

A previous study (Bruton et al. 1995) revealed thatS. coelicolor possesses two glycogen branching enzymes.One, encoded by glgBI, is needed for phase I but not forphase II glycogen synthesis. The other, encoded byglgBII, is needed for phase II, but not for phase Iglycogen synthesis. Limited analysis showed that thesequences immediately upstream of glgBI and glgBIImight also be related to each other. To investigate thisfurther, we used digoxigenin-labelled segments of theDNA upstream of glgBI to probe Southern blots of re-striction digests of cosmid 4G10, which contains theglgBII region, and similar probes for the region up-stream of glgBII were used on Southern blots of re-striction digests of cosmid 6A11 containing the glgBIregion. Sequences extending up to about 5.5 kb up-stream of the two unlinked glgB genes were found to berelated (Fig. 1).

The occurrence of cross-hybridizing DNA upstreamof each glgB gene might merely indicate the presence ofan ancient ``fossil'' duplication of a DNA segmentcontaining adjacent genes unrelated functionally toglgB, or might re¯ect functional and/or regulatory in-volvement of the whole duplicated region in glycogenmetabolism. To examine this, potentially mutagenic in-sertions of a FC31-derived prophage were made into theclusters, using single crossover recombination mediatedby several di�erent restriction fragments (Fig. 1). Arepresentative colony of each class of lysogen was thenanalysed by electron microscopy for the presence ofpolysaccharide deposits (Figs. 2 and 3). In each case,aberrant deposits resembling those previously seen in thecorresponding glgB disruption mutants (Bruton et al.1995) were found in one location in the colonies, andnormal wild-type deposits were found in the other.Disruptions in the glgBI cluster (e.g. J1881; Fig. 2) af-fected only phase I deposits, and disruptions in theglgBII cluster (e.g. J1890, J1891, J1883; Fig. 3) a�ectedonly phase II deposits. These experiments veri®ed thatextensive lengths of DNA upstream of each glgB genewere involved in spatially speci®c glycogen metabolism,either by virtue of their own function or because they arecotranscribed with glgB genes, such that the disruptionshad polar e�ects on glgB expression. The occurrence ofpolysaccharide deposits, albeit abnormal ones, ruled outsevere impairment in basic a-1,4 polyglucan biosynthe-sis. However, it was noticeable that sporulation-associ-ated polysaccharide deposits were small and di�cult to®nd in spore chains of strains J1891 and J1892, in which

the DNA fragment directing the integration of thevector derived from a location only about 2.5 kb fromglgBII. This suggests a quantitative as well as a quali-tative e�ect of this disruption on glycogen deposits,more marked than had previously been noted with dis-ruption of glgBII itself (Bruton et al. 1995).

The occurrence of mixed phenotypes in some mutantsin the glgBII cluster complicated this analysis: a mi-nority of spore chains showed apparently normal gly-cogen deposits. To ®nd out whether this was due tooccasional excision of the mutagenic prophage, sporepopulations were harvested after non-selective (thio-strepton-free) culture of the mutants, and tested forthiostrepton resistance. In no case were more than 3%of colonies thiostrepton sensitive, whereas somewhathigher proportions (albeit di�cult to quantify precisely)

Fig. 2A±E Electron microscopy reveals phase I-speci®c changes inglycogen deposits caused by gene disruption upstream of glgBI.A J1508 (parental strain), phase I deposits (thin arrow). B J1508, phaseII deposits in developing spore chain (thin arrow). C±E J1881, inwhich the DNA upstream of glgBI has been disrupted (see Fig. 1).Phase I deposits are abnormal (C and D, thick arrows). Themorphology of phase II deposits (E, thin arrows) in a developingspore chain is normal. Bars 0.5 lm

545

of spore chains contained normal glycogen deposits,making strain instability an unlikely explanation. Thephenotypic variation was noticed only with strains inwhich one end of the cloned fragment used for disrup-tion included part of glgBII. Possibly, the cloned seg-ment or the adjacent vector sequences contain apromoter that usually remains unexpressed, but can beactivated in a quasi-random, stochastic manner in somedeveloping spore chains. Such an interpretation wassupported by the fact that the same phase II variationwas observed with J1884, in which looping out of thedisrupting vector was prevented by a deletion extendingleftwards from within the vector (Fig. 1). Nevertheless,since all the insertions gave aberrant, presumptively

unbranched, glycogen deposits resembling those previ-ously obtained by speci®c disruption of glgBI or glgBII,we conclude that the major promoter for glgB in eachcluster is at least 3 kb upstream of glgB (i.e. upstream ofthe inserts in J1880/J1881 or J1891/J1892).

A putative trehalose synthase geneand a gene for a possible glycosyl hydrolaseor glycosyl transferase are locatedupstream of each glgB gene

In view of the evidence suggesting that the duplicatedDNA sequences were cotranscribed with the adjacentglgB genes, the regions were sequenced to investigatetheir possible function. As shown in Fig. 1, three genespreceding each glgB gene encode proteins that show verysigni®cant conservation between the clusters (Table 1).The genes were arranged in the same order in each case(pep1-treS-pep2-glgB). Moreover, non-coding intergenicregions were absent or very short, and possible transla-tional coupling was indicated by small overlaps betweenthe ®rst and second conserved genes of each cluster(Table 1).

We also carried out limited further sequencingdownstream of the two glgB genes (Fig. 1). The genedownstream of glgBI converges on it, while the gene(pep3) downstream of glgBII could potentially be co-transcribed with it (though a terminator-like invertedrepeat lies between the two genes). Downstream of pep3is a converging gene. None of the genes downstream ofglgB genes resemble genes of known function, and we donot consider them further in this paper.

In database searches, the deduced Pep1 proteins(which are about 87% identical to each other) showedevident relatedness to proteins of the a-amylase super-family from a wide range of prokaryotes and eukar-yotes. The greatest similarity (65% identity end-to-end)was to the uncharacterised product of a gene ofMycobacterium tuberculosis. The remaining 36 proteins

Fig. 3A±D Electron microscopy reveals phase II-speci®c changes inglycogen deposits caused by disruptions upstream of glgBII (forparental controls, see Fig. 2). A J1891 showing normal phase Iglycogen granules (thin arrow). B J1891 spore chain showing one ofthe few phase II polysaccharide deposits detected (thick arrow), whichis a clump typical of unbranched phase II material. C Typicalsporulating segment of J1883 showing blobs of unbranched glycogen(thick arrows). D Less frequent example of a sporulating segment ofJ1883 that contains normal phase II glycogen (thin arrow). Bars0.5 lm

Table 1 Comparison of genes and gene products of the glgBI and glgBII clusters

Gene Cluster I Cluster 2 Shared featuresa

pepA Absent 90-amino acid gene product NAIntergenic region 61-nt gap (includes strong

inverted repeat)NA

pep1 675-amino acid gene product 669-amino acid gene product 71.6% (156 residues); 98.0% (394 residues);72.7% (110 residues)(with one gap at N-ter and one at C-ter)

Intergenic region 4-nt overlap 4-nt overlaptreS 566-amino acid gene product 572-amino acid gene product 92.3% (no gaps)Intergenic region 20-nt gap 28-nt gap 12-nt identical segmentpep2 464-amino acid gene product 453-amino acid gene product 73.4% (6 gaps)Intergenic region 40-nt gap 3-nt gapglgB 788-amino acid gene product 741-amino acid gene product 73.5% (3 gaps)Intergenic region 16-nt to converging gene

(includes strong inverted repeat)72-nt gap (includes stronginverted repeat)

pep3 Absent 364-amino acid gene product NA

aNA, not applicable

546

that exceeded the default level of similarity to Pep1needed for reporting by BLASTP were all members ofthe a-amylase superfamily. The most signi®cantmatches (P < 0.001) were to a-glucosidases and mal-tases from organisms as diverse as a mosquito, the fruit¯y, mouse, a gram-negative bacterium (Sinorhizobium)and a thermophilic actinomycete (Thermomonospora).The aligned regions picked out by BLASTP for theseproteins allowed, overall, almost continuous alignmentof Pep1 across the segment from amino acid residues230±530. In the proteins aligned, this region makes upmost of the [ab]8 barrel structure characteristic of thea-amylase superfamily, and includes a series of residueswhose role in enzyme function has been characterised(Svensson 1994; Fig. 4). The conserved regions includ-ed the three active-site carboxylates involved in cleav-ing a-1,4-glucan bonds (D in region 2, E in region 3and D in region 4), and an aspartate in region 1 and anarginine in region 2 that form a structurally essentialsalt bridge. However, some normally conserved fea-tures are absent from Pep1. These include the histidineresidues in regions 1 and 4 which interact with theglycone residue at the bond to be cleaved, and the tworesidues involved in Ca2+ binding in some starch-metabolising enzymes (an asparagine in region 1 and ahistidine in region 2). Interestingly, the same diver-gences from conserved features were noted in themaltosyltransferase of Thermatoga maritima, an un-usual non-hydrolytic enzyme which speci®cally trans-fers maltosyl units, generating a set of multiples ofmaltose (i.e. maltose, maltotetraose, maltohexaose, etc)from starch or glycogen (Meissner and Liebl 1998).Among the enzymes listed in the BLASTP search,many of those with poorer alignments in the [ab]8barrel region showed similarity to di�erent segments ofthe C-terminal region of Pep1 proteins (residues �500±600), notably including several cyclomaltodextrin glu-canotransferases (residues ca. 550±570 and 580±600),two trehalose-6-phosphate hydrolases and twomaltodextrin glycosyltransferases (residues �500±550).Overall, these comparisons are strongly indicative of anenzyme capable of glycohydrolase or glycotransferaseactivity on glycogen (for details of references and ac-cession numbers, see Discussion).

The product of the second gene common to eachcluster showed extremely high end-to-end relatedness totrehalose synthases of Pimelobacter and Thermus spp.(ca 73% sequence identity and 95% similarity overnearly the whole length of the gene products, in the caseof the Pimelobacter enzyme), making it virtually certainthat their enzymatic function is the same (Fig. 5), hencethe gene designation treS. These enzymes have beendemonstrated biochemically to convert maltose intotrehalose (for details of references and accession num-bers, see Discussion). As shown in Fig. 4, the TreSsequence can also be aligned with those of a-amylase-related proteins. Possibly, this re¯ects the breaking ofthe a-1,4-linkage of maltose necessary for its conversionto trehalose.

No resemblance between the products of the thirdconserved gene in each cluster (pep2) and any func-tionally characterised proteins was discernible.

Sequences related to the genes of the glgB clustersare present in M. tuberculosis, but are not duplicatedand are di�erently organised

The four genes common to each glgB cluster each haveclosely related homologues (but only once each) in thegenome of M. tuberculosis (Cole et al. 1998), which is anactinomycete distantly related to streptomycetes. How-ever, in M. tuberculosis the genes are organised as twowell separated pairs: the homologues of treS and the(potentially translationally coupled) functionally un-characterised pep2 gene immediately downstream of itlie about 1,340 kb away from glgBI, which is itself lo-cated immediately downstream of the pep1 homologue(with an intergenic gap of 7 nucleotides).

Di�erent genes encoding probable glycogenmetabolising enzymes are adjacent to,and diverge from, the two glgB operons

In the M. tuberculosis sequence annotation, a putativeglgP (glycogen phosphorylase) gene lies upstream of,and diverges from, the pep1glgB gene pair (glycogenphosphorylase degrades a-1, 4-glucans to glucose-1-phosphate). This encouraged us to sequence DNAfurther upstream from the glgB operons. Indeed, adivergent glgP-like gene was found upstream of theglgBI cluster (Figs. 1 and 6). The product of this genewas particularly similar to that of the M. tuberculosisglgP and, more tellingly, to that of a gene fromThermatoga maritima that has been demonstratedbiochemically to encode an a-glucan phosphorylase(Bibel et al. 1998). These proteins are somewhat di-vergent from all other known glycogen phosphorylases,which are otherwise particularly well conserved acrossorganisms from E. coli and Bacillus subtilis throughyeasts to humans (Fig. 7). No glgP gene of a more``typical'' type is present in M. tuberculosis (Cole et al.1998).

When we sequenced further upstream from glgBII,no glgP-like gene was found. Instead we encountered aglgX-like gene. In other bacteria, such genes are believedto encode debranching enzymes able to cleave a-1,6-linked lateral branches from glycogen molecules.

The glpP and pep1I coding sequences are separatedby only 370 bp of non-coding DNA. However, thelikely translational starts of glgX and pep1II are sepa-rated by 1137 bp. A search of this segment for possibleprotein-coding DNA revealed two possible small genes(Fig. 1). One is potentially cotranscribed with glgX(and its stop codon overlaps the likely glgX startcodon). The other is potentially cotranscribed withpep1, from which it is separated by 61 bp including a

547

GC-rich inverted repeat. The predicted translationalstarts of the two ``minigenes'' are 291 bp apart. Nohomology of the minigene products to known proteinscould be detected.

Discussion

Duplicated operons for storage carbohydratemetabolism

The results described here show that S. coelicolor con-tains two highly conserved �8 kb gene clusters that in-clude a determinant of glycogen synthesis (glgB) and,

almost certainly, one of trehalose synthesis (treS). Theclusters also contain two other conserved genes; theorder of the genes is pep1-treS-pep2-glgB. Each ofthe clusters is likely to be transcribed as a single operon,because (1) single-crossover disruptions with fragmentsinternal to treSI and treSII gave rise to phenotypes likeglgBI and glgBII mutant phenotypes, respectively, asmight be expected if the disruptions had polar e�ects onglgB genes further downstream in the same transcriptionunit; (2) the mutant phenotype (unbranched glycogendeposits) generated by ``mutational cloning'' with afragment extending from inside treSII to inside glgBIIsuggests that the fragment is entirely internal to a tran-scription unit; and (3) the pep1 and treS genes overlap by4 nucleotides in each cluster, suggesting translationalcoupling of adjacent genes transcribed on a singlemRNA. Each operon is located near to, and divergesfrom, another gene likely to be involved in glycogendegradation (glgP, which encodes a probable glycogenphosphorylase, in the case of the glgBI cluster; and glgX,which speci®es a probable debranching enzyme, in thecase of the glgBII cluster).

Before discussing further the implications of theseresults, it is appropriate to review the strength of theevidence that treS encodes a trehalose synthase, partic-ularly because until recently the only known route fortrehalose synthesis in bacteria involved two enzymes:

Fig. 4 Conservation in Pep1, TreSSc and GlgXSc of critical regions ofa-amylase-related proteins. The alignments and details use informa-tion from Svensson (1994) andMeissner and Liebl (1998). The GlgXSc

sequences (Scx) are all highly conserved in relation to other bacterialGlgX proteins (residues marked with dots are identical in at least ®veout of seven GlgX proteins examined; where letters are shown, theresidues indicated are predominant in the set of seven). Pep1 (Scp;Pep1I and Pep1II are identical in the regions shown) shows low butclear-cut similarity to many enzymes in this family, and this Figureshows that suitably placed residues correspond to many functionallysigni®cant residues of these proteins. Other abbreviations: Aor,Aspergillus oryzae; Ppa, porcine pancreas; Bce, Bacillus cereus; Sct,S. coelicolor TreSI or TreSII (identical in these regions); Eco, E. coli;Scb, S. coelicolor GlgBI; Thm, Thermatoga maritima; Mtb, Myco-bacterium tuberculosis Pep1 homologue; Scp, S. coelicolor Pep1

548

trehalose-6-phosphate synthase, which condenses glu-cose-6-phosphate and UDP-1-glucose (or, in at least onecase, GDP-1-glucose; Elbein 1968) to give trehalose-6-phosphate, and trehalose-6-phosphate phosphatase.However, enzymes that are able to generate a-1,1-linkedglucosyl moieties by intramolecular glucosyl transfer,using a-1,4-glucans, malto-oligosaccharides or, in somecases, maltose itself as the substrate, have now beendescribed in the archaebacteria Sulfolobus spp. (Lama

et al. 1990, 1991; Kato et al. 1996; Maruta et al. 1996c;Nakada et al. 1996a, 1996b; Kobayashi et al. 1996); inThermus aquaticus (Nishimoto et al. 1996a) and theactinomycetes Arthrobacter (Maruta et al. 1995, 1996a;Nakada et al. 1995a, 1995b) and Pimelobacter (Nishi-moto et al. 1995, 1996b); and in gram-negative organ-isms (Rhizobium spp.) (Maruta et al. 1996b; Streeterand Bhagwat 1999). Tsusaki et al. (1996, 1997) puri®edtrehalose synthase (which converts maltose into treha-lose), from Pimelobacter sp. R48 and Thermus aquati-cus, and used experimentally determined amino acidsequence data to clone the relevant treS genes. Thedegree of relatedness between the Pimelobacter enzymeand the S. coelicolor TreS proteins is much greater thanthat between the Pimelobacter and Thermus trehalosesynthases, justifying the presumption that the S. coeli-color genes do indeed encode trehalose synthase iso-forms.

Fig. 5 Alignment of TreSI and TreSII with trehalose synthases ofPimelobacter (pim; Accession No. E10495) and Thermus aquaticus(thaq; Accession No. E11499). The T. aquaticus enzyme has a 390-residue C-terminal extension (not completely shown) relative to theenzyme from Pimelobacter. The sequence of TreSI is shown in full(scoel), while for TreSII only those residues that di�er from theequivalent residues in TreSI are shown above the sequence alignment.The M. tuberculosis homologue of TreS is also included (myct;Accession No. O07176)

549

The signi®cance of duplicated treS genes

Glycogen and trehalose have been reported to showsomewhat di�erent patterns of accumulation in strep-

tomycetes. Trehalose appears to be present at all stagesof colony development, and may reach very high levelsin spores (BranÄ a et al. 1986; McBride and Ensign 1987;Karandikar et al. 1997). Glycogen is found at the in-terface between the substrate and aerial mycelium and inimmature spores, and is largely absent from vegetativehyphae, the stalks of aerial hyphae, and spores (BranÄ aet al. 1980, 1986; Plaskitt and Chater 1995; Karandikaret al. 1997). At least some of the genes of glycogensynthesis (i.e. glgB and glgC) are duplicated, one copy ofeach being speci®c for one of the phases of glycogendeposition (Bruton et al. 1995; Homerova et al. 1996;Martõ n et al. 1997). It now appears that there are du-

Fig. 6 Alignment of the deduced glgP gene products from S. coelicolor(scoel) and M. tuberculosis (MycT; Accession No. Q10639) with ana-glucan pyrophosphorylase from T. maritima (thmar; Accession No.AJ001088). The colons above the alignment indicate amino acidscommon to all three sequences that are universally conserved in all theGlgP sequences compared in Fig. 7, which are derived from diverseeukaryotes and prokaryotes. The letters indicate residues universallyconserved in the other GlgP sequences, but not conserved in the threesequences aligned here. The dashed arrows indicate segments whoselength varies in di�erent sequences

550

plicated treS genes for trehalose synthesis, one beingcotranscribed with each glgB gene. It is therefore quitelikely, but unproven, that the glgBI-linked treS genemay be involved in making trehalose at the substrate/aerial mycelial interface, and that the glgBII-linked treSgene may contribute to the accumulation of trehalosein spores. This leaves open the question of how tre-halose is made at other stages such as during substratemycelium growth; but both enzymatic (Matula et al.1971) and genome studies (Cole et al. 1998) haveshown that conventional enzymes of trehalose synthesisare present in mycobacteria, and at least M. tubercu-losis also contains a treS gene that is extremely similarto those of Pimelobacter and S. coelicolor: thus, thesetwo systems for trehalose synthesis appear to coexist inmycobacteria. Moreover, trehalose phosphate synthase(albeit GDP-glucose-dependent; see above) and treha-lose phosphate phosphatase activities have been dem-onstrated in S. hygroscopicus (Elbein 1967, 1968),making it likely that actinomycetes often have both

systems. Possibly, the conventional enzymes of treha-lose synthesis are present in S. coelicolor, and the treSgenes of S. coelicolor serve to boost the localised supplyof trehalose to the aerial mycelium and ultimately tothe spores. This boost could permit increased aerialgrowth and spore yield in some conditions, and con-tribute to resistance to stresses such as dessication,against which trehalose may o�er protection (MartõÂ net al. 1986; McBride and Ensign 1987; Rod et al. 1988),and which are likely to be associated with the aerialformation of spore chains and the long-term survival ofspores.

The possible roles of pep1 and pep2

The pep1 homologue in M. tuberculosis is not closelylinked to a treS homologue; rather, it is located imme-diately upstream of glgB in this organism, emphasizingits likely connection with glycogen. Conceivably, Pep1 isan intracellular glucanohydrolase (or maltosyltransfer-ase) that converts glycogen into units ± perhaps maltose± that serve as the substrate for TreS; this putative in-timate connection between Pep1 and TreS would beconsistent with the overlap of the two relevant genes ineach cluster (we note that some related enzymes areknown to generate maltose from branched a-glucans;Ferrari et al. 1993; Meissner and Liebl 1998).

While this paper was being reviewed, Belanger andHatfull (1999) provided evidence that mycobacterialPep1 homologues recycle carbon from intracellularstored glycogen during growth on certain media. Thisis consistent with the idea that Pep1 is a glycanase, andled Belanger and Hatfull to name the mycobacterialgene glgE. In a nice phrase, they considered that,during mycobacterial growth, glycogen acts as a ``car-bon capacitor'', which can be mobilised for glycolysisby GlgE (ºPep1). Our model for S. coelicolor has bothsimilarities to, and di�erences from, that for myco-bacteria. In each case, the Pep1 homologue mobilisescarbon from glycogen. However in our model thebalance between glycogen accumulation and degrada-tion is largely developmentally determined in S. coeli-color, rather than responding to growth rate as inmycobacteria; and we suppose that the carbon unitsgenerated by Pep1 are converted to trehalose, ratherthan entering glycolysis as is proposed for mycobacte-ria. Since at least M. tuberculosis, and perhaps M.smegmatis, contains a treS gene, some of the carbonmobilised by GlgE may be converted into trehalose inmycobacteria. If so, some of the phenotypic e�ects ofthe M. smegmatis glgE mutation may be due to tre-halose limitation.

There are no characterised Pep2-like proteins, so nofunctional attribution can be attempted, except that theclose linkage and likely cotranscription of treS and pep2in both Streptomyces clusters and in M. tuberculosisindicates that Pep2 proteins are involved in trehalosemetabolism.

Fig. 7 Dendrogram based on degrees of similarity between variousa-glucan phosphorylases. The PILEUP analysis shows that theproteins from S. coelicolor (Scoel), M. tuberculosis (Myct: AccessionNo. Q10639) and T. maritima (Thm: Accession No. AJ001088) clusterseparately from the major cluster of such proteins, which includesthose from E. coli (Accession No. P13031) Haemophilus in¯uenza(Hinf: Accession No. P45180), man (Accession no. P11216), bean(Accession No. P53536), potato (Pot: Accession No. P04045), slimemoulds (Slime: Accession No. Q00766) and Bacillus subtilis (Bsu:Accession No. P39123)

551

The concept of a carbon relayduring the Streptomyces life cycle

The combined consideration of these genetic data withphysiological data obtained previously (see above forreferences) suggests an interplay of trehalose and gly-cogen metabolism. The following model will provide aframework for subsequent experiments.

1. When hyphal growth is limited in conditions ofcarbon source excess, available carbon is sequesteredinside non-growing hyphae as storage macromolecules,including glycogen. The formation of such phase Iglycogen deposits involves the activation of the glgBIoperon to give rise to the proper branched structure ofglycogen, as well as the activation of other unlinkedgenes needed for glycogen synthesis, such as glgC(MartõÂ n et al. 1997).

2. When the hyphae containing glycogen depositssend reproductive branches into the air, some of thecarbon that supports their growth comes from glycogendegradation. This would be achieved partly by ``con-ventional'' degradation by glycogen phosphorylase, togive readily metabolisable glucose-1-phosphate, but wealso suggest that some of the glycogen may be convertedto trehalose in a process involving TreSI (a suggestionprompted by the fact that treSI is linked to, and ap-parently cotranscribed with, glgBI). The conversion ofglycogen to maltose, the probable TreSI substrate, mayinvolve the action of PepI.

3. When aerial hyphae undergo sporulation septa-tion, glycogen is deposited in the prespore compart-ments, probably being synthesized by the conventionalroute from glucose-1-phosphate (MartõÂ n et al. 1997).Some of the glucose-1-phosphate may originate from theaction on phase I glycogen of glycogen phosphorylaseencoded by glgP. It is not clear whether or how trehalosemight contribute glucosyl residues to this phase II gly-cogen. The branches are put into phase II glycogen bythe glgBII gene product (Bruton et al. 1995). Tran-scription of glgBII implies cotranscription of pep1II andtreSII, perhaps sowing the seeds of the eventual con-version of the phase II glycogen into trehalose as sporematuration takes place.

4. This speculative model does not address the ques-tion of how the biosynthetic and degradative functionsattributed to the glgB operons could be orchestrated,though regulation of enzyme activity by allosteric ef-fectors or covalent modi®cation seems quite likely.

Streptomyces aerial hyphae are unusual among bac-terial cells in that their cellular growth takes place awayfrom external nutrient sources (Migue lez et al. 1994),necessitating the movement of nutrients from the sub-strate to the aerial hyphal growth zones. The ability tostockpile a local nutrient reserve such as glycogen and tomobilise it partially in the form of a substance (treha-lose) that is immune to the action of normal metabolicenzymes would make a colony less dependent on the(unreliable) continued supply of carbon sources from theenvironment during the developmental process. An in-

teresting analogy may be drawn with the generation ofstarch deposits in plants during daylight as a store ofphotosynthate, and the subsequent utilisation of thesedeposits in the dark, in part to generate sucrose, anothernon-reducing, metabolically inactive disaccharide wellsuited for movement to parts of the plant remote fromthe sites of primary assimilation (Stitt and Sonnewald1995).

Acknowledgments We thank Kim Findlay for help with electronmicroscopy, Stephen Bornemann for helpful discussion, and AlisonSmith and Mervyn Bibb for thoughtful comments on the manu-script. DS was supported by an EC post-doctoral fellowship. Ad-ditional support was provided by a grant from the Biotechnologyand Biological Sciences Research Council to the John Innes Centre.

References

Belanger A, Hatfull G (1999) Exponential-phase glycogen recyclingis essential for growth of Mycobacterium smegmatis. J Bacteriol181: 6670±6678

Bibel M, Brettl C, Gosslar U, KriegshaÈ user G, Liebl W (1998)Isolation and analysis of genes for amylolytic enzymes of thehyperthermophilic bacterium Thermotoga maritima. FEMSMicrobiol Lett 158: 9±15

BranÄ a AF, Manzanal MB, Hardisson C (1980) Occurrence ofpolysaccharide granules in sporulating hyphae of Streptomycesviridochromogenes. J Bacteriol 144: 1139±1142

BranÄ a AF, Me ndez C, Dõ az LA, Manzanal MB, Hardisson C(1986) Glycogen and trehalose accumulation during colonydevelopment in Streptomyces antibioticus. J Gen Microbiol 132:1319±1326

Bruton CJ, Guthrie EP, Chater KF (1991) Phage vectors that allowmonitoring of transcription of secondary metabolism genes inStreptomyces. Biotechnology 9: 652±656

Bruton CJ, Plaskitt KA, Chater KF (1995) Tissue-speci®c glycogenbranching isoenzymes in a multicellular proaryote, Streptomy-ces coelicolor A3(2). Mol Microbiol 18: 89±99

Chater KF, Losick R (1997) The mycelial life-style of Streptomycescoelicolor A3(2) and its relatives. In: Shapiro JH, Dworkin M(eds) Bacteria as multicellular organisms. Oxford UniversityPress, New York, pp 149±182

Cole ST, et al. (1998) Deciphering the biology of Mycobacteriumtuberculosis from the complete genome sequence. Nature 393:537±544

Elbein AD (1967) Carbohydrate metabolism in Streptomyces.J Bacteriol 94: 1520±1524

Elbein AD (1968) Trehalose phosphate synthesis in Streptomyceshygroscopicus: puri®cation of guanosine diphosphate D-glucose:D-glucose-6-phosphate 1-glucosyl-transferase. J Bacteriol 96:1623±1631

Ferrari E, Jarnagin AS, Schmidt BF (1993) Commercial produc-tion of extracellular enzymes. In: Sonenshein AL, Hoch JA,Losick R (eds) Bacillus subtilis and other gram-positive bacte-ria: biochemistry, physiology and molecular genetics. AmericanSociety for Microbiology, Washington DC, pp 917±938

Homerova D, Benada O, Kofronova O, Rezuchova B, Kormanec J(1996) Disruption of a glycogen branching enzyme gene, glgB,speci®cally a�ects the sporulation-associated phase of glycogenaccumulation in Streptomyces aureofaciens. Microbiology 142:1201±1208

Hopwood DA, Bibb MJ, Chater KF, Kieser T, Bruton CJ, KieserHM, Lydiate DJ, Smith CP, Ward JM, Schrempf H (1985)Genetic manipulation of Streptomyces. A laboratory manual.John Innes Foundation, Norwich

Ikeda H, Seno ET, Bruton CJ, Chater KF (1984) Genetic mapping,cloning and physiological aspects of the glucose kinase gene ofStreptomyces coelicolor. Mol Gen Genet 196: 501±507

552

Karandikar A, Sharples GP, Hobbs G (1997) Di�erentiation ofStreptomyces coelicolor A3(2) under nitrate-limited conditions.Microbiology 143: 3581±3590

Kato M, Miura Y, Kettoku M, Shindo K, Iwamatsu A, KobayashiK (1996) Puri®cation and characterisation of new trehalose-producing enzymes isolated from the hyperthermophilic archae,Sulfolobus solfataricus KM1. Biosci Biotechnol Biochem 60:546±550

Kobayashi K, Kato M, Miura Y, Kettoku M, Komeda T,Iwamatsu A (1996) Gene cloning and expression of newtrehalose-producing enzymes from the hyperthermophilicarchaeum Sulfolobus solfataricus KM1. Biosci BiotechnolBiochem 60: 18820±18825

Lama L, Nicolaus B, Trincone A, Morzillo P, De Rosa M, Gam-bacorta A (1990) Starch conversion with immobilised thermo-philic Archaebacterium Sulfolobus solfataricus. Biotechnol Lett12: 431±432

Lama L, Nicolaus B, Trincone A, Morzillo P, Calandrelli V,Gambacorta A (1991) Thermostable amylolytic activity fromSulfolobus solfataricus. Biotechnology ForumEurope 8: 201±203

MartõÂ n MC, DõÂ az LA, Manzanal MB, Hardisson C (1986) Role oftrehalose in the spores of Streptomyces. FEMS Microbiol Lett35: 49±54

MartõÂ n MC, Schneider D, Bruton CJ, Chater KF, Hardisson C(1997) A glgC gene essential only for the ®rst of two spatiallydistinct phases of glycogen synthesis in Streptomyces coelicolorA3(2). J Bacteriol 179: 7784±7789

Maruta K, Nakada T, Kubota M, Chaen H, Sugimoto T,Kurimoto M, Tsujisaka Y (1995) Formation of trehalose frommaltooligosaccharides by a novel enzymatic system. BiosciBiotechnol Biochem 59: 1829±1834

Maruta K, Hattori K, Nakada T, Kubota M, Sugimoto T,Kurimoto M (1996a) Cloning and sequencing of trehalosebiosynthesis genes from Arthrobacter sp. Q36. Biochim BiophysActa 1289: 10±13

Maruta K, Hattori K, Nakada T, Kubota M, Sugimoto T,Kurimoto M (1996b) Cloning and sequencing of trehalosebiosynthesis genes from Rhizobium sp. M-11. Biosci BiotechnolBiochem 60: 717±720

Maruta K, Mitsuzumi H, Nakada T, Kubota M, Chaen H, FukudaS, Sugimoto T, Kurimoto M (1996c) Cloning and sequencing ofa cluster of genes encoding novel enzymes of trehalose bio-synthesis from thermophilic archaebacterium Sulfolobus acido-caldarius. Biochim Biophys Acta 1291: 177±181

Matula M, Mitchell M, Elbein AD (1971) Partial puri®cation andproperties of a highly speci®c trehalose phosphate phosphatasefrom Mycobacterium smegmatis. J Bacteriol 107: 217±222

McBride MJ, Ensign JC (1987) E�ects of intracellular trehalosecontent on Streptomyces griseus spores. J Bacteriol 169: 4995±5001

Meissner H, Liebl W (1998) Thermotoga maritima maltosyltrans-ferase, a novel type of maltodextrin glycosyltransferase actingon starch and malto-oligosaccharides. Eur J Biochem 250:1050±1058

Me ndez C, BranÄ a AF, Manzanal MB, Hardisson C (1985) Role ofsubstrate mycelium in colony development in Streptomyces.Can J Microbiol 31: 446±450

Migue lez EM, Garcõ a M, Hardisson C, Manzanal MB (1994)Autoradiographic study of hyphal growth during aerial myce-lium development in Streptomyces antibioticus. J Bacteriol 176:2105±2107

Nakada T, Maruta K, Mitsuzumi H, Kubota M, Chaen H, Su-gimoto T, Kurimoto M, Tsujisaka Y (1995a) Puri®cation andcharacterisation of a novel enzyme, maltooligosyl trehalosetrehalohydrolase, from Arthrobacter sp. Q36. Biosci BiotechnolBiochem 59: 2215±2218

Nakada T, Maruta K, Tsusaki K, Kubota M, Chaen H, SugimotoT, Kurimoto M, Tsujisaka Y (1995b) Puri®cation and prop-erties of a novel enzyme, maltooligosyl trehalose synthase, fromArthrobacter sp. Q36. Biosci Biotechnol Biochem 59: 2210±2214

Nakada T, Ikegami S, Chaen H, Kubota M, Fukuda S, SugimotoT, Kurimoto M, Tsujisaka Y (1996a) Puri®cation and charac-terisation of thermostable maltooligosyl trehalose synthasefrom the thermoacidophilic archaebacterium Sulfolobus acido-caldarius. Biosci Biotechnol Biochem 60: 263±266

Nakada T, Ikegami S, Chaen H, Kubota M, Fukuda S, SugimotoT, Kurimoto M, Tsujisaka Y (1996b) Puri®cation and charac-terisation of thermostable maltooligosyl trehalose trehalohy-drolase from the thermoacidophilic archaebacterium Sulfolobusacidocaldarius. Biosci Biotechnol Biochem 60: 267±270

Nishimoto T, Nakano M, Ikegami S, Chaen H, Fukuda S, Su-gimoto T, Kurimoto M, Tsujisaka Y (1995) Existence of a novelenzyme converting maltose into trehalose. Biosci BiotechnolBiochem 59: 2189±2190

Nishimoto T, Nakada T, Chaen H, Fukuda S, Sugimoto T,Kurimoto M, Tsujisaka Y (1996a) Puri®cation and charac-terisation of a thermostable trehalose synthase fromThermus aquaticus. Biosci Biotechnol Biochem 60: 835±839

Nishimoto T, Nakano M, Nakada T, Chaen H, Fukuda S, Su-gimoto T, Kurimoto M, Tsujisaka Y (1996b) Puri®cation andproperties of a novel enzyme, trehalose synthase, from Pime-lobacter sp. R48. Biosci Biotechnol Biochem 60: 640±644

Plaskitt KA, Chater KF (1995) In¯uences of developmental geneson localised glycogen deposition in colonies of a mycelialprokaryote, Streptomyces coelicolor A(3)2: a possible interfacebetween metabolism and morphogenesis. Philos Trans R SocLond B Biol Sci 347: 105±121

Redenbach M, Kieser HM, Denapaite D, Eichner A, Cullum J,Kinashi H, Hopwood DA (1996) A set of ordered cosmids anda detailed genetic and physical map for the 8 Mb Streptomycescoelicolor A3(2) chromosome. Mol Microbiol 21: 77±96

Rod ML, Alam KY, Cunningham PR, Clark DP (1988) Accu-mulation of trehalose by Escherichia coli K-12 at high osmoticpressure depends on the presence of amber suppressors.J Bacteriol 170: 3601±3610

Rodicio MR, Bruton CJ, Chater KF (1985) New derivatives of theStreptomyces temperate phage /C31 useful for the cloning andfunctional analysis of Streptomyces DNA. Gene 34: 283±292

Sambrook J, Fritsch EF, Maniatis T (1989) Molecular cloning: alaboratory manual (2nd edn). Cold Spring Harbor LaboratoryPress, Cold Spring Harbor, NY

Stitt M, Sonnewald U (1995) Regulation of metabolism in trans-genic plants. Annu Rev Plant Physiol Plant Mol Biol 40:341±368

Streeter JG, Bhagwat A (1999) Biosynthesis of trehalose frommaltooligosaccharides in Rhizobia. Can J Microbiol 45:716±721

Svensson B (1994) Protein engineering in the a-amylase family:catalytic mechanism, substrate speci®city, and stability. PlantMol Biol 25: 141±157

Thie ry JP (1967) Mise en e vidence de polysaccharides sur coupe®nes en microscopie e lectronique. J Microscopie (Paris) 6: 987±1018

Tsusaki K, Nishimoto T, Nakada T, Kubota M, Chaen H, Su-gimoto T, Kurimoto M (1996) Cloning and sequencing of tre-halose synthase gene from Pimelobacter sp. R48. BiochimBiophys Acta 1290: 1±3

Tsusaki K, Nishimoto T, Nakada T, Kubota M, Chaen H, FukudaS, Sugimoto T, Kurimoto M (1997) Cloning and sequencing oftrehalose synthase gene from Thermus aquaticus ATCC33923.Biochim Biophys Acta 1334: 28±32

553