Identiﬁcation and Structure of the Anti-sigma Factor ... · Norwich NR4 7UH, UK 3Department of Molecular Microbiology, John Innes Centre, Norwich NR4 7UH UK 4School of Biological

Identification and Structure of the Anti-sigma Factor-binding Domain of the Disulphide-stress RegulatedSigma Factor sR from Streptomyces coelicolor

Wei Li1, Clare E. M. Stevenson2, Nicolas Burton2, Piotr Jakimowicz3

Mark S. B. Paget4, Mark J. Buttner3, David M. Lawson2 andColin Kleanthous1*

1School of Biological SciencesUniversity of East AngliaNorwich NR4 7TJ, UK

2Department of BiologicalChemistry, John Innes CentreNorwich NR4 7UH, UK

3Department of MolecularMicrobiology, John InnesCentre, Norwich NR4 7UHUK

4School of Biological SciencesUniversity of Sussex, FalmerBrighton BN1 9QG, UK

The extracytoplasmic function (ECF) sigma factor sR is a global regulatorof redox homeostasis in the antibiotic-producing bacterium Streptomycescoelicolor, with a similar role in other actinomycetes such as Mycobacteriumtuberculosis. Normally maintained in an inactive state by its bound anti-sigma factor RsrA, sR dissociates in response to intracellular disulphide-stress to direct core RNA polymerase to transcribe genes, such as trxBAand trxC that encode the enzymes of the thioredoxin disulphide reductasepathway, that re-establish redox homeostasis. Little is known about whereRsrA binds on sR or how it suppresses sR-dependent transcriptionalactivity. Using a combination of proteolysis, surface-enhanced laserdesorption ionisation mass spectrometry and pull-down assays weidentify an N-terminal, ,10 kDa domain (sRN) that encompasses region 2of sR that represents the major RsrA binding site. We show that sRN

inhibits transcription by an unrelated sigma factor and that this inhibitionis relieved by RsrA binding, reaffirming that region 2 is involved in bind-ing to core RNA polymerase but also demonstrating that the likely mech-anism by which RsrA inhibits sR activity is by blocking this association.We also report the 2.4 A resolution crystal structure of sRN that revealsextensive structural conservation with the equivalent region of s70 fromEscherichia coli as well as with the cyclin-box, a domain-fold found in theeukaryotic proteins TFIIB and cyclin A. sRN has a propensity to aggregate,due to steric complementarity of oppositely charged surfaces on thedomain, but this is inhibited by RsrA, an observation that suggests apossible mode of action for RsrA which we compare to other well-studiedsigma factor-anti-sigma factor systems.

q 2002 Elsevier Science Ltd. All rights reserved

Keywords: structure; RsrA; S. coelicolor; sigma factor; X-ray crystallography*Corresponding author

Introduction

Promoter recognition by bacterial RNA polymer-ase requires the core enzyme (a2bb

0v ¼ E) toassociate with a specificity subunit called s. Thefirst s factor to be characterised was s70, the princi-pal, essential s factor of Escherichia coli.1 Othermembers of the s70 family, known as alternative sfactors, each confer a different promoter specificityon the holoenzyme (Es), and bacteria use alterna-tive s subunits to control expression of specificsets of genes (regulons). Alignment of members ofthe s70 family has identified four major conservedregions (1–4), each further divided intosubregions.2,3 The binding motifs that recognise

0022-2836/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved

Present addresses: W. Li, Department of Haematology,Cambridge Institute for Medical Research, University ofCambridge, Wellcome Trust/MRC building, Hills RoadCB2 2XY, UK; C. Kleanthous, Department of Biology(Area 10), P.O. Box No 373, University of York,Heslington, York YO10 5YW, UK.

E-mail address of the corresponding author:[email protected]

Abbreviations used: ECF, extracytoplasmic function;ZAS, zinc-containing anti-sigma factor; SELDI-MS,surface-enhanced laser desorption ionisation massspectrometry.

doi:10.1016/S0022-2836(02)00948-8 available online at http://www.idealibrary.com onBw

J. Mol. Biol. (2002) 323, 225–236

the 210 and 235 promoter sequences are locatedin regions 2.4 and 4.2, respectively,2,4 – 6 whereasthe interface of s with core is more extensive, withregions 2–4 all involved.7 – 10

It is common for bacteria to express alternative sfactors before they are needed, but maintain themin an inactive state until their release is triggeredby appropriate conditions. This inhibition isachieved through the reversible action of inhibitorproteins known as anti-s factors.11,12 Whereas allmembers of the s70 family are homologous, anti-sfactors belong to a rapidly increasing number ofphylogenetically unrelated families. Although sfactor:anti-s factor interaction has been analysedin detail in only a few cases, most particularlysF:SpoIIAB in Bacillus subtilis,13 s28:FlgM inSalmonella typhimurium14,15 and s70:AsiA inE. coli,16,17 it is already clear that inhibition can beachieved through a variety of distinct mechanisms.

In Streptomyces coelicolor, sR activates expressionof a regulon of .30 genes that help the bacteriumsurvive “disulphide stress”, the unwanted for-mation of disulphide bonds in the cytoplasm.18– 20

sR target genes include trxBA and trxC, which

encode the enzymes of the thioredoxin disulphidereductase pathway.19,20 The activity of sR is con-trolled by the anti-s factor RsrA, a redox-sensitive,zinc metalloprotein that modulates sR activity inresponse to changes in the thiol-disulphide redoxbalance of the cytoplasm.21,22 Under reducing con-ditions, RsrA binds to sR to form a 1:1 complex,preventing it from activating transcription.Oxidative stress induces intra-molecular disul-phide bond formation in RsrA, causing it to dis-sociate and thereby release sR to activate its targetgenes. Increased trxBA and trxC expression inturn leads to the thioredoxin-dependent reductionof oxidised RsrA back to its sR-binding con-formation, shutting off sR-dependenttranscription.19 – 22 The sR–RsrA system also existsin other actinomycetes, including pathogens likeMycobacterium tuberculosis, where it is namedsH-RshA.19,20,23 RsrA is the first-described memberof a new and growing family of anti-s factors, thezinc-containing anti-s factor (ZAS) family, thatcontrol the activity of certain members of the extra-cytoplasmic function (ECF) subfamily of sigmafactors.3,22 – 25

Figure 1. Identification of sRN, theRsrA binding domain on sR.(a) Limited trypsin digestion profileof sR under native conditions.Three major fragments were gener-ated, two from the N terminus(sRN) and one from the C terminus(sRC) that were mapped by N-termi-nal sequencing of blotted fragments(indicated by the positions of thearrows) and SELDI-MS; the Ctermini of the fragments are pre-dicted on the basis of the cleavagespecificity of trypsin and correlationof observed and expected molecu-lar masses. The numbered sectionsof sR represent the characteristicregions of sigma factors (afterLonetto et al.3). (b) SELDI-MS of sR

and its proteolytic fragments. (i)The major sR tryptic fragments thatare detected using a normal phase(NP20) chip, where both sRN andsRC are evident. (ii)–(iv) Data col-lected from IMAC3 (immobilizedmetal affinity capture) chips. In (ii),the chip was loaded with a mixtureof His-RsrA and the sR limitedtryptic digest (,20 mM). In (iii), thesR tryptic digest was added withoutHis-RsrA while in (iv) only His-RsrA was loaded onto the chip. Inall cases the chips were washedextensively with buffer prior tolaser desorption (see Materials andMethods). Only sRN fragmentswere retained on the His-RsrA-IMAC3 chip, indicating an associ-ation between the domain and theanti-sigma factor.

226 Characterisation of the RsrA-binding Domain of sR

There is currently no information on how ZASproteins inhibit their target sigma factors, andsince the mechanisms of action of anti-sigma fac-tors are diverse, it is impossible to extrapolatefrom other well-studied systems. Hence, weembarked on a series of biochemical experimentsto delineate the binding site on sR for its ZASprotein, RsrA, and to demonstrate how this associ-ation inhibits sR-dependent transcriptional activity.Through this work we identify an RsrA bindingdomain on sR, determine its three-dimensionalstructure by X-ray crystallography, and show thatits structural conservation goes beyond the sigmafactor family to include eukaryotic transcriptionfactors and signalling proteins. Our data illustrateyet another variation on the rich mechanisticthemes by which sigma factors are inhibited bytheir anti-sigma factor proteins.

Results

Identification of the RsrA-binding domain on sR

sR (25 kDa) and RsrA (12 kDa) were over-produced in E. coli and purified as described19,21

and in Materials and Methods. Both heterologouslyproduced proteins are functional in vitro; demon-strated for sR through transcription run-off assaysdirected by sR-specific promoters, and for RsrA bysuppression of this activity when in the reducedstate.

Proteolysis has been used extensively todelineate anti-sigma factor binding sites on sigmafactors, either through the identification of sites ofprotease protection or fragments that bind theanti-sigma factor protein.26,27 Adopting the laterapproach, we used proteolysis to define regions ofsR that interact with RsrA, identifying initially themajor trypsin-generated fragments of sR in theabsence of its anti-sigma factor. Although there

are 28 theoretical trypsin cleavage sites in the sR

sequence we found that limited digestion withtrypsin under native conditions (0.4% (w/w) at25 8C, pH 7.5 for 30 minutes) generated only threemajor species, as deduced by SDS-PAGE (data notshown). These were readily mapped to thesequence of sR through a combination of N-termi-nal sequencing of the blotted fragments andanalysis of their molecular weights by surface-enhanced laser desorption ionisation mass spec-trometry (SELDI-MS) (Figure 1(a) and (b), (i). Twoof the fragments originate from the N terminus ofthe protein and correspond to a 9.5–10 kDadomain that encompasses region 2 and part ofregion 1, which we term sRN. The third fragmentcorresponds to a ,6 kDa domain that encom-passes regions 3.2, 4.1 with some of 4.2 and comesfrom the C-terminal half of sR, which we termsRC. Together, the two proteolytically defineddomains account for approximately two-thirds ofthe sR sequence (Figure 1(a)).

To determine if either of these domains interactswith RsrA we performed a pull-down assay usinga SELDI-IMAC chip (see Materials and Methods).This approach capitalises on the ability to attach abait protein to the metal-chelate surface of theIMAC chip through an engineered histidine tagand to then identify proteins that bind specificallyto such an activated surface by SELDI-MS. In thisinstance, His-tagged RsrA and a limited trypticdigest of sR were loaded onto an IMAC chip andthe chip washed repeatedly with binding bufferprior to the identification of bound fragments bySELDI-MS. Both N-terminal sRN fragments wereretained specifically on the His-RsrA-activatedchip, with no detectable retention of sRC (Figure1(b), (ii). Since both sRN fragments were identified,it is reasonable to assume that the deletion of theC-terminal six residues does not affect the abilityof this domain to bind RsrA. Control experimentsin which either His-RsrA or sR tryptic fragmentsalone were bound to IMAC chips, washedrepeatedly and then analysed by SELDI-MS, indi-cated that retention of the sRN fragments requiredbound RsrA (Figure 1(b), (iii) and (iv)). We con-clude that sRN is a stable N-terminal 10 kDadomain, encompassing regions 2.1–2.4 of sR,which also represents a major interaction site forRsrA.

Characterisation of the sRN domain and itscomplex with RsrA

In order to obtain quantities of sRN sufficient forbiophysical analysis, the corresponding fragmentof sigR encoding its 88 amino acid residues(residues 23–110) was subcloned into pET15b,thereby placing an N-terminal hexa-histidine tagonto the domain (His-sRN) (see Materials andMethods). Circular dichroism (CD) spectroscopyof purified His-sRN at pH 5.4 indicated that thedomain is folded and composed of ,68% a-helix,with little or no b-sheet structure (Figure 2(a)). We

Figure 2. Far-UV CD spectrum of overexpressed andpurified His-sRN. His-sRN (3.8 mM) was dialysed into50 mM potassium phosphate buffer (pH 5.4) and thespectrum collected in a 1 mm pathlength cell from 190–260 nm. The resulting spectrum was baseline-subtractedand the secondary structure composition estimated as,68% a-helix with little or no b-structure.

Characterisation of the RsrA-binding Domain of sR 227

found that preparations of His-sRN (with or with-out the histidine tag) were more stable at low pH(,6) but prone to aggregation at neutral pH unlessimidazole or RsrA was included in the buffer (datanot shown). We return to this issue below.

Overexpressed His-sRN was capable of bindingnon-His-tagged RsrA in vitro. Figure 3 shows theresults of a pull-down assay (at pH 7.5 in the pre-

sence of 100 mM imidazole) in which His-sRN andRsrA were added to Ni2þ-chelate beads, the beadswashed extensively with binding buffer, and thenfinally eluted with EDTA. RsrA and His-sRN onlyco-eluted during the final EDTA strip (Figure 3(a),lane 5). Omission of His-sRN from the Ni2þ-chargedresin (Figure 3(a), lane 4) demonstrated that reten-tion of RsrA on the beads required bound His-sRN. Complex formation between overexpressedHis-sRN and reduced RsrA was also stable enoughto be detected by gel-filtration chromatography(data not shown).

Interaction of sRN with core RNA polymerase isinhibited by RsrA

sRN is an autonomous, stably folded domain thatcontains sequences (region 2.4) required for bind-ing to the 210 box of a promoter but lacks keysequences in the C-terminal half of sR that areneeded for 235 recognition. Hence, it was to beexpected that sRN could not initiate transcriptionfrom sigR-dependent promoters, and preliminaryexperiments indicated that this was indeed thecase (data not shown). Binding to core RNA poly-merase by bacterial sigma factors involves con-served regions of the protein that have beendelineated for many different sigma factors.8 Thepresence of at least one of these (region 2.1)suggested that sRN might be able to associate withthe enzyme even though it could not initiatetranscription.

To test this hypothesis we investigated whetherHis-sRN could compete with the unrelated Strepto-myces sigma factor, sH, for the s binding site oncore RNA polymerase. sH plays an important rolein the general stress response in Streptomyces.28

Figure 3. His-sRN and RsrA form a complex in vitro.The Figure shows the results of a pull-down assay usingNi2þ-chelate resin. Lanes 1 and 2 indicate the migrationpositions of purified His-sRN and reduced RsrA, respect-ively, in SDS-15% PAGE. Lanes 3–5 show the proteinsthat are eluted from Ni-chelate resin by an EDTA stripafter two intervening wash steps with binding buffer(5 £ resin volume). Lanes 3 and 4 show purified His-sRN and reduced RsrA individually added to the resin,while in lane 5 equimolar His-sRN and reduced RsrAwere incubated at room temperature for ten minutesand then added to the resin prior to the wash and elutionsteps. RsrA was only retained on the resin in the pre-sence of His-sRN. Additional controls using an unrelated,histidine-tagged protein (His-Im9) showed that retentionof RsrA was specific to His-sRN and not a consequence ofthe tag (data not shown).

Figure 4. RsrA competes withcore RNA polymerase for sRN bind-ing. In vitro transcription of the ctcpromoter by EsH. Lane 1, no sH

control; lanes 2–4 contain 1.9 pmolsH plus 0, 9.5 and 19-fold of His-sRN to sH; lanes 5–7, contain thesame amount of sH and His-sRN asin lane 4 but include two-, four-and eightfold excess (relative toHis-sRN) of reduced RsrA. Theamount of transcript in lanes 2–7was quantified by densitometryand plotted as the percentageof sH-dependent transcriptionobserved in lane 2. The decrease insH-dependent transcription whenHis-sRN was added (lanes 3 and 4)indicates that this small domaincompetes effectively for core RNApolymerase binding (although itselfcannot activate transcription) butthat this inhibition is relieved byRsrA.


Transcription from a sH-dependent promoter(Figure 4, lane 2) was largely inhibited when amolar excess of His-sRN was included in the assay(Figure 4, lane 4), consistent with sRN competingwith sH for the sigma factor binding site on coreRNA polymerase. Importantly, the majority of sH-dependent transcription could be recovered bytitrating-in an excess of reduced RsrA relative toHis-sRN (Figure 4, lanes 5–7). Taken together,these results suggest that sRN retains the capacityto bind core RNA polymerase but that this associ-ation is likely inhibited by RsrA.

Crystallization and structure of sRN

Crystallographic information is currently avail-able for s70 from E. coli and sA from Thermusaquaticus both bound to core RNA polymerase andin the unbound form,5,6,9,10 and for sF fromB. stearothermophilus in complex with its anti-sigmafactor SpoIIAB.29 No structural information is yetavailable for any ECF sigma factor. We attemptedto grow crystals of intact His-sR that had beenpurified through Ni-affinity chromatography andthe N-terminal histidine tag removed by thrombindigestion (see Materials and Methods). Hexagonal-shaped crystals were obtained after only a fewdays using polyethylene glycol (PEG) 8000 as pre-cipitant at pH 8 and at 4 8C. However, subsequentanalysis of the crystals by SDS-PAGE indicatedthat they did not consist of intact sR but rather ofa proteolytic digestion product identical with the10 kDa N-terminal fragment generated by trypsin,sRN. It seems that thrombin digestion of the intactsR continued during crystallisation experimentsand that crystals were only formed after the for-mation of sRN. Similar proteolytic susceptibilityduring crystallisation has been noted for sA fromT. aquaticus.6

We solved the structure of sRN by multipleanomalous displacement (MAD) phasing to aresolution of 2.7 A using intact selenomethionine-labelled sR as the starting material, which alsobecame proteolysed during crystallisation. Thestructure was subsequently refined against nativedata collected to 2.4 A resolution. Crystals werealso obtained for overexpressed sRN but these didnot improve the data quality (data not shown). Ofthe 88 amino acid residues that comprise sRN only78 could be unambiguously placed in the electrondensity; the residues that are missing from thefinal model originate primarily from the C termi-nus, suggestive either of static or dynamic disorderin the crystal. sRN is an antiparallel three-helixbundle with three residues of 310 helix in the loopjoining helices 2 and 3 (Figure 5). a-Helix accountsfor 68% of the observed sRN domain structure,identical with the predicted helical content fromCD. Helix 1 is the longest and displays a pro-nounced kink approximately halfway along itslength.

Structural homologues of sRN in the Protein DataBank (PDB) were identified using the DALI

server.30 This yielded several significant matcheswith fragments of other structures (Figure 6), thebest match being that with the structure of s70

(PDB accession code 1SIG5). A least-squares super-position of the s70 structure onto sRN, on the basisof the main-chain atoms from 75 equivalent resi-dues, gave an overall root-mean-square deviation(rmsd) of 2.9 A. It is notable that the 310 helix andthe kinked helix, which encompasses region 2.1,are also preserved in the s70 structure. Other

Figure 5. Stereo diagram showing the a-carbon back-bone of sRN, the major RsrA binding domain of sR

from S. coelicolor. The N and C termini correspond toresidues 25 and 102 of the full-length protein,respectively.

Figure 6. Structural similarities between sRN and otherproteins. (a) The three a-helices of sRN colour-coded,with the equivalent helices similarly coloured ((b)–(d))in s70,5 cyclin A31 and TFIIB,32 respectively. Note howthe kinked helix of sRN, involved in binding to coreRNA polymerase, is only conserved in s70. The struc-tures are depicted in ribbon representation and weregenerated using MOLSCRIPT57 and Raster3D.58


significant matches were found with cyclin A (PDBaccession code 1VIN31), rmsd of 2.4 A over 68 resi-dues, and transcription factor TFIIB (PDB accessioncode 1AIS32), rmsd of 1.7 A over 59 residues,although in both of these the kink in the first helixis not present (Figure 6).

Protein–protein interactions in the crystal andimplications for the mode of action of RsrA

In the sRN crystal, the molecules pack as a net-work of non-intersecting fibres that run along the2-fold screw axes of the unit cell (Figure 7). Theasymmetric unit contains two copies of sRN that

are related to one another by a 908 rotation aboutthe 2-fold screw axis and a quarter cell edge trans-lation along the axis. Further 908 rotations andquarter cell edge translations in both directionsrelate adjacent molecules in neighbouringasymmetric units. Consequently, there is a 41 non-crystallographic screw axis coincident with thecrystallographic 2-fold screw axis. This is apparentin a self-rotation function (calculated using data inthe range 10.0–3.5 A resolution) as a peak at 15%of the origin height that coincides with the crystal-lographic 2-fold screw axis. Thus the crystallo-graphic and non-crystallographic interfaces alongthe fibre are virtually indistinguishable.

The total buried surface area at the interfacesbetween adjacent molecules amounts to ,810 A2.The surfaces are remarkably complementary interms of shape (a shape complementarity index of0.64 makes them comparable to antibody/antigencomplexes33) and charge (Figure 7). Thesecharacteristics likely explain the propensity of sRN

to precipitate at neutral pH and the higher orderoligomers that are readily detected in chemicalcross-linking experiments (data not shown).

The protein–protein interface observed in thesRN fibre is unlikely to be biologically significantgiven that the domain represents less than 40% ofintact sR (and that sR itself is not prone to pre-cipitate at physiological pH). However, the self-association provides clues as to how RsrA mightbind and sequester sR in an inactive, heterodimericcomplex. This stems from the fact that wefound RsrA to be very effective at preventing theaggregation of sRN. In our handling of purifiedsRN, we found that the domain (with or without

Figure 7. Protein–protein interactions in the sRN

crystal. (a) Packing of molecules in the fibre. Crystallo-graphically equivalent molecules are shown in the samecolour and are related by the 2-fold screw axis. Adjacentmolecules are related by a non-crystallographic 4-foldscrew axis. (a)–(g) The central pair of molecules fromthe fibre in (a) rotated by 908 about the vertical axis inopposite directions so that their interacting surfaces facethe viewer. (b), (d) and (f) show the left-hand molecule,(c), (e) and (g) show the right-hand molecule. (b) and (c)are in ribbon representation and use the same colours asFigure 6. (d) and (e) show the electrostatic surface poten-tials as red, white and blue for values less than 210 kT,neutral and greater than 10 kT, respectively. (f) and (g)show the interaction surfaces. Those regions of the sur-face that are closer than 2 A away from the neighbouringsurface are coloured yellow, and the remainder iscoloured grey. (a)–(c) were generated as for Figure 6.(d)–(g) were produced using GRASP.59

Figure 8. RsrA prevents the self-association of sRN. Amixture of His-sRN and sRN in 20 mM Tris–HCl (pH7.5), 0.1 M NaCl, 400 mM imidazole was dialysed over-night at 4 8C against the same buffer, but lackingimidazole, in the presence (lanes 3 and 4) or absence(lanes 1 and 2) of a twofold molar excess of RsrA. Theaddition of BSA to dialysed His-sRN/sRN served as anegative control (lanes 5 and 6). Soluble (S) and pellet(P) fractions from each of the three samples were thenanalysed by SDS-20% PAGE. The experiment demon-strates that His-sRN/sRN readily precipitates at neutralpH, even in the presence of BSA, but that this is pre-vented by the inclusion of RsrA. The sRN preparationused in the experiment was a mixture of histidine andnon-histidine-tagged protein (resulting from prolongedstorage of His-sRN), demonstrating that both are proneto aggregation. The asterix indicates an impurity in thepreparation.


the histidine tag) was only soluble at neutral pH inthe presence of high concentrations (.100 mM) ofimidazole. Removal of the imidazole by dialysisagainst Tris–HCl at pH 7.5 results in the precipi-tation of both sRN and His-sRN (Figure 8, lane 2).Aggregation was largely abolished by the presenceof RsrA during dialysis but not by the inclusion ofBSA (Figure 8, lanes 3–6). Since RsrA binds specifi-cally to sRN (see Figures 1 and 3), this effect is mostreadily interpreted in terms of RsrA inhibiting theself-association of sRN. Since RsrA is an acidicprotein (pI, 4.8) we speculate that it prevents sRN

self-association by binding to its electropositivesurface (Figure 7(e)).

Discussion

We have defined the anti-sigma factor bindingdomain on the ECF sigma factor sR fromS. coelicolor. sR regulates the transcription of manygenes involved in the maintenance of thiol-disulphide redox homeostasis, while its anti-sigmafactor RsrA is the redox sensor. Trypsin treatmentof sR under native conditions generated two majorfragments, from the N and C terminus, respect-ively, with only sRN able to bind RsrA (Figure 1).While this demonstrates that sRN is most likely themajor binding site for RsrA we cannot discountthe possibility that additional contacts are made toother regions but are lost during proteolysis. Over-expressed sRN inhibits transcription by anunrelated sigma factor (Figure 4), suggesting thatsRN is an autonomous domain able to bind at thesigma factor binding site on core RNA polymerase.Reduced RsrA abolished this effect, implying thatRsrA may inhibit sR-dependent transcriptionalactivity by inhibiting its binding to polymerase.The self-association of sRN, evident in solution as apropensity to aggregate and observed crystallo-graphically as sRN fibres (Figure 7), is due toelectrostatically complementary surfaces in thedomain coming together, a phenomenon that islargely prevented by the binding of RsrA (Figure 8).

The crystal structure of sRN revealed a similaramount of a-helix to that suggested by circulardichroism, and that this domain is structurallysimilar to the equivalent region in s70 from E. coli(Figure 6), consistent with the presence of well-defined and conserved regions involved in bindingto core RNA polymerase (region 2.1), DNA melting(region 2.3) and 210 recognition (region 2.4). sRN

also shows strong structural similarity to a foldthat is shared by the eukaryotic signalling proteincyclin A and transcription factor TFIIB, termed thecyclin box fold.34 The cyclin box fold is character-ised by five helices, three of which are found insRN. The cyclin box fold seems to be a generaladapter motif employed in transcriptional regu-lation, as well as cell cycle control, able to bind adiverse set of proteins and DNA.34 Although a pro-karyotic protein, the presence of the cyclin box foldwithin sR is consistent with these properties, since

it binds other proteins (core RNA polymerase andRsrA) and DNA.

Genome sequencing has revealed that the ECFsubfamily of sigma factors is far larger than firstimagined; S. coelicolor alone has 49 members.35

Overlapping promoter specificity, where morethan one ECF sigma factor recognizes a single pro-moter, is an emerging theme. For example, almost50% of promoters recognized by EsR are alsorecognized by at least one other form ofholoenzyme,20 and sX and sW of B. subtilis haveoverlapping specificity at several promoters.36 Inboth cases, the 210 region appears to play animportant role in promoter selectivity. The sRN

structure, and the identification of residues alongthe likely DNA-binding face of helix a3, will aidthe design of experiments directed towards under-standing the mechanism of promoter selection bysR and other ECF sigma factors.

Comparison with other s factor:anti-sfactor complexes

One of the best understood sigma factor:anti-sigma factor interactions is that of sF and its com-plex with SpoIIAB.13 sF initiates forespore-specificgene expression during sporulation in B. subtilisand forms a 1:2 complex with its anti-sigma factorSpoIIAB. In the crystal structure of the low-affinity(SpoIIAB(ADP)) form of the complex, region 3 ofsF makes specific contacts with each of the SpoIIABmonomers, while the remaining 80% of sF isdisordered.29 Genetic and biochemical data indi-cate that SpoIIAB also contacts regions 2 and 4 ofsF in the high-affinity (SpoIIAB(ATP)) form of thecomplex.26,27,37 Direct occlusion of a core RNA poly-merase binding surface appears to be the basis ofSpoIIAB’s anti-s activity.29

In S. typhimurium, s28, which transcribes flagel-lum and chemotaxis genes, is regulated by a cog-nate anti-sigma factor, FlgM.14,15 In contrast to themode of inhibition of sF by SpoIIAB, FlgM activelydissociates the Es28 holoenzyme to yield a 1:1s28:FlgM complex.38 Genetic and biochemical datasuggest that FlgM contacts multiple regions of frees28, including 2.1, 3.1 and 4,39,40 whereas anti-holoenzyme activity appears to be mediated solelythrough its interaction with region 4.40

During infection of E. coli by T4 bacteriophage,the phage-encoded protein AsiA inhibits host tran-scription by binding to s70 to form a 1:1 complex,interacting specifically with region 4.16,17,41 –44 FreeAsiA is a symmetric dimer with the residues thatform the hydrophobic dimer interface alsoinvolved in binding s70, suggesting that an AsiAprotomer is displaced from the dimer through itsinteraction with s70.45,46 Unlike the mode of actionof SpoIIAB and FlgM, AsiA can form a stablecomplex with the Es70 holoenzyme. Availableevidence suggests that AsiA modifies Es70

function, masking the 235 helix-turn-helix DNAbinding motif in region 4 while bound to theholoenzyme.41


The present work, along with that reported pre-viously, suggests that the anti-ECF action of theZAS protein RsrA is distinct from that of otheranti-sigma factors. Unlike the 1:2 complex of thesF:SpoIIAB complex, sR forms a 1:1 complex withRsrA21 and so, since RsrA is a monomer (data notshown), dimer dissociation as seen with AsiA isunlikely. Region 2 plays a prominent role in theinteraction of both sF and s28 with their anti-sigmafactors, which is also the case for sR. Region 4,which plays a central in the regulation of sF, s28

and s70 by anti-sigma factors, may not be involvedin binding to RsrA, since the C-terminal domain,sRC, which includes region 4, did not interact withRsrA (Figure 1).

sRN can inhibit sigma-dependent transcriptionand RsrA can inhibit this effect, suggesting thatRsrA prevents RNA polymerase binding by sR

(Figure 4), the binding site on the sigma factorlikely comprising a positively charged surface(Figure 7(e)). This site is composed of a3 and theC-terminal end of a1 of sRN, the former incorporat-ing residues involved in DNA melting and 210recognition (regions 2.3 and 2.4), the latter, part ofthe RNA polymerase binding site (region 2.1).Hence, RsrA likely obscures sR sequences involvedin binding both promoter DNA and RNA polymer-ase. This form of sigma factor inhibition wouldrepresent a novel mechanistic variation in sigmafactor:anti-sigma factor interactions and mayunderpin the regulation of other ZAS protein–ECF sigma factor complexes.

Materials and Methods

Protein production

RsrA was purified as described by Kang et al.21 T7-based overexpression of sR was accomplished essentiallyas described by Paget et al.19 but where sR carried anN-terminal hexa-histidine tag and an intervening throm-bin site, transformed into E. coli BL21lDE3/plysS forunlabelled cells or BL21lDE3 met2/plysS for seleno-methionyl-labelled cells. For the expression of unlabelledprotein, cells were grown at 37 8C in LB medium contain-ing 200 mg/ml ampicillin and 30 mg/ml chloramphenicolto mid-log and then induced for three hours with IPTG(1 mM). For selenomethionyl-labelled protein, an over-night LB culture was harvested, washed with 10 ml M9medium and then resuspended in 5 ml M9 medium,3 ml of which was used to inoculate a 1 l culture of M9medium containing 200 mg/ml ampicillin and 30 mg/mlchloramphenicol and supplemented with 2 mM MgSO4,10 mg thiamine, 0.1 mM CaCl2, 10 g glucose, 40 mg of19 amino acid residues (except methionine) and 60 mgselenomethionine (SeMet). This culture was grown at37 8C and induced at mid-log with 1 mM IPTG for 20hours. Harvested cells were disrupted using either aSoniprobe sonicator (Dawes) at 4 8C or by French press-ing, and clarified supernatant passed down a nickel-affinity column (POROS 20 MC, PerSeptive Biosystems)in binding buffer (20 mM Tris–HCl (pH 7.5), 0.5 MNaCl, 5 mM imidazole) at room temperature. Bound sR

was eluted with 50–600 mM imidazole gradient on aBioCad SPRINT (PerSeptive Biosystems) using a flow

rate of 10 ml/minute. Following thrombin digestion toremove the hexa-histidine tag (0.5 units thrombin/mgsR for two hours at 27 8C), the protein was further puri-fied by gel-filtered chromatography using a SuperdexS75 26/60 column (Pharmacia Biotech) equilibrated in50 mM Tris–HCl (pH 7.5), 1 mM EDTA. Pure sR

fractions were pooled, concentrated and stored in 50%(v/v) glycerol at 220 8C.

The N-terminal domain of sR, corresponding to resi-dues 23–110 (sRN), was generated with a hexa-histidinetag (His-sRN) by cloning the corresponding DNA intopET15b (pET15b-sigRN ), and purified as describedabove for sR often without the inclusion of the thrombincleavage or gel-filtration steps, the protein being essen-tially pure after Ni2þ-affinity chromatography. His-sRN

was prone to aggregation at pH 7.5 unless imidazolewas included in the buffer and so was generally storedat 220 8C as pooled fractions from the Ni2þ-column(final imidazole ,300 mM) and included 50% glycerol.The protein was soluble at pH 7.5 but only in the pre-sence of RsrA (see the text for details), alternatively theprotein remained soluble in low pH buffers such assodium acetate or sodium phosphate (pH 5.4).

All protein molecular masses were verified either byelectrospray ionisation mass spectrometry (VG Platform;Micromass, UK) or by SELDI-MS (CiphergenBiosystems, USA) and were within 1–2 Da of theexpected mass. Protein concentrations were determinedfrom molar absorption coefficients derived from aminoacid analysis (sR, 1280 ¼ 27,682 M21 cm21; sRN,1280 ¼ 16,298 M21 cm21).

Trypsin digestion and mass spectroscopy

sR was digested with trypsin (0.4%, Sigma) for 30 min-utes at 25 8C in 50 mM Tris–HCl (pH 7.5) and the diges-tion mixture subjected to SDS-16% (w/v) PAGE. Majorproteolytic fragments were blotted onto PVDF mem-brane and subjected to N-terminal sequencing (AltaBioscience, Birmingham, UK). SELDI-MS was generallyused to determine fragment masses using either normalphase silicon oxide surface (NP1 and NP20) chips, forsimple verification of peptide masses, or immobilisedmetal affinity capture (IMAC3) chips for the identifi-cation of peptide fragments binding to RsrA. In the latterexperiments, 10 mM DTT reduced His-RsrA and a sR

trypsin digestion mixture (each ,20 mM) were incubatedat 30 8C for one hour before spotting onto an IMAC3 chippre-loaded with NiSO4 and washed repeatedly withbinding buffer ( £ 5) and distilled water ( £ 3) beforedetection of captured peptides by SELDI-MS. In allexperiments, 3,5-dimethoxy-4-hydroxycinnamic acid ora-cyano-4-hydroxycinnamic acid were used as thematrix and the endonuclease domain of colicin E9(15,088 Da) and the immunity protein Im9 (9882.6 Da)used as calibrants of molecular mass.

Circular dichroism spectroscopy

CD data were collected on an Applied Photophysicspp-180 CD spectrophotometer (Leatherhead, UK) in a1 mm pathlength cell at 20 8C using 3.8 mM His-sRN in50 mM potassium phosphate buffer (pH 5.4). A total of50,000 data points were collected between 190 nm and260 nm and 20 scans averaged. Secondary structurecontent was estimated using the deconvolution programCDNN.


In vitro binding assays

Pull-down assays were performed using Ni-NTA resin(Novagen) and in 100 mM Tris–HCl (pH 7.5), 150 mMNaCl containing 100 mM imidazole. His-sRN, whichremains soluble under these conditions, and reducedRsrA (non-His-tagged) were each incubated with Ni-NTA resin (pre-charged with NiSO4) either individuallyor in complex. After washing the beads with bindingbuffer (see above) containing 1 mM of the reducingagent Tris (2-carboxyethyl) phosphine hydrochloride,bound proteins were eluted with strip buffer (100 mMTris–HCl (pH 7.5), 150 mM NaCl, 100 mM EDTA) andfractions analysed by SDS-16% PAGE.

In vitro transcription run-off assays

In vitro transcription assays were performed asdescribed21 using transcription buffer containing1.9 pmol sH and 0.5 pmol of the sH-specific promoterctc.47 The effects of His-sRN and RsrA on this activitywere monitored by pre-incubating these proteins, in thepresence of 10 mM DTT, at 30 8C for 30 minutes, andthen with 1 ml of E. coli core RNA polymerase (1.5 pmol,Cambio, Cambridge, UK) for a further five minutesbefore adding transcription buffer and initiating tran-scription by the addition of an NTP mix (0.4 mM) con-taining [a-32P]CTP (400 Ci/mmol). Transcripts wereanalysed on a 6% polyacrylamide gel containing 7 Murea.

sRN aggregation assay

A mixture of His-sRN and sRN (,1 mg/ml total),dissolved in 20 mM Tris–HCl (pH 7.5), 0.1 M NaCl con-taining 400 mM imidazole, was dialysed overnight at4 8C against 20 mM Tris–HCl (pH 7.5), 0.1 M NaCl inthe presence or absence of a twofold molar excess ofdithiothreitol-reduced RsrA. In a third dialysis experi-ment, BSA was added to His-sRN/sRN. Soluble and pelletfractions from each of the three samples were then ana-lysed by gel-electrophoresis.

Crystallization

Thrombin-cleaved sR was concentrated to approxi-mately 10 mg/ml, filtered through a 0.1 ml Ultrafree filterand crystals grown by vapour diffusion in hangingdrops using VDX plates (Hampton Research) at 4 8C.The precipitant consisted of 10–15% (w/v) PEG 8000, in100 mM Tris–HCl at pH 8.5. These conditions were sup-plemented with 1 mM DTT for the SeMet protein.

Drops consisted of 1 ml of protein solution mixed with1 ml of precipitant. The crystals appeared as hexagonalplates after two days, being approximately 100 mmacross and 20 mm thick. In order to ascertain the preciseprotein sequence of the crystallized material, severalcrystals were dissolved and N-terminally sequencedusing a Procise Model 491 protein sequencer (AppliedBiosystems, Warrington, Cheshire), while SELDI-MS(see above) was used to ascertain the molecular weightand hence define the C terminus. The crystallizedprotein corresponded to the 10.3 kDa fragment of sRN,consisting of residues 23–110.

Data collection

All crystal manipulations were performed usingHampton Research tools. Initially, crystals were cryo-protected by soaking for up to five minutes in motherliquor containing ethylene glycol in place of 25% of thebuffer volume, after which they were mounted in cryo-loops and flash-cooled by plunging into liquid nitrogen.They were then transferred to a dewar using the CrystalCap system for storage and subsequent transportationto the synchrotron. For data collection, crystals weremounted on the goniometer using cryotongs and main-tained at 100 K with an Oxford Cryosystems Cryostreamcooler. The observed diffraction was consistent with tri-gonal symmetry giving approximate cell parameters ofa ¼ b ¼ 71.6 A, c ¼ 102.8 A. Systematic absences along c p

indicated that the space group was either P3121 orP3221. An estimation of solvent content gave values of66% for two copies of sRN per asymmetric unit, and 50%for three copies.48 X-ray data for the native protein werecollected on the microfocus beamline ID13 (l ¼ 0.782 A)at the European Synchrotron Radiation Facility (ESRF)in Grenoble, France, using a 165 mm MAR-ResearchCCD detector, yielding virtually complete data to 2.4 Aresolution. The SeMet X-ray data were collected using a165 mm MAR-Research CCD detector on beamlinePX9.5 at the Synchrotron Radiation Source (SRS) atDaresbury, UK. Prior to data collection, a fluorescencescan was used to determine the best wavelengths atwhich to collect data. Subsequently, X-ray data werecollected at wavelengths of 0.9794 A (peak), 0.9801 A(inflection point) and 0.8500 A (high-energy remote). Alldata sets were collected in a single 1808 sweep of 1 deg.images to 2.7 A resolution.

Data processing and structure solution

The native X-ray data were processed with DENZOand scaled and merged using SCALEPACK49 and all

Table 1. X-ray data collection statistics

Beamline PX9.5 (SRS) PX9.5 (SRS) PX9.5 (SRS) ID13 (ESRF)Data set SeMet-peak SeMet-inflection SeMet-remote NativeWavelength (A) 0.9794 0.9801 0.8500 0.782Resolution (A) 2.7 2.7 2.7 2.4Unique reflectionsa 15,984 15,984 15,980 12,713Completeness (%)b 100.0 (100.0) 100.0 (100.0) 100.0 (100.0) 97.9 (83.4)Redundancya 5.9 5.9 5.9 3.8Rmerge

b,c 0.045 (0.282) 0.053 (0.324) 0.055 (0.343) 0.047 (0.142)kIl/ksIlb 24.5 (3.8) 25.1 (4.2) 22.5 (3.4) 25.2 (5.3)

a Anomalous pairs treated as separate reflections for MAD data.b Figures in parentheses refer to data in the highest resolution bin.c Rmerge ¼

PðlIj 2 kIjllÞ=

PkIjl; where Ij is the intensity of an observation of reflection j and kIjl is the average intensity for reflection j.


subsequent downstream processing and statistical anal-ysis was effected using programs from the CCP4 suite.50

A subset of the data comprising a random 5% of thereflections was set aside for the calculation of “free”(Rfree) crystallographic R-factors during modelrefinement.51 For the MAD data, after running SCALE-PACK (using the “no merge” option) the resultant fileswere processed using the program SOLVE†.52 SOLVEwas run with each possible enantiomorph, and onlygave a satisfactory result in space group P3221. sRN con-tains two Met residues. SOLVE automatically deter-mined the positions of four Se atoms, suggesting twocopies of the fragment per asymmetric unit. Phasing cal-culations on the basis of these Se coordinates gave anoverall figure of merit of 0.62 at 2.7 A resolution, whichincreased to 0.91 after solvent flattening and histogrammatching with DM.53 An experimentally phased Fobs

electron density map calculated at this stage gave clearcontrast between protein and bulk solvent regions andseveral helices were apparent. Data collection and pro-cessing statistics of all data used in the structure solutionand refinement are summarised in Table 1.

Model building and refinement

Model building was performed using the program Oand alternated with refinement using REFMAC.54,55 Thefour Se positions were arranged in two pairs and, uponcloser inspection, each pair was seen to be associatedwith an a-helix, the two atoms lying approximately twoturns of helix apart. The latter observation was consistentwith the likely separation of Met47 and Met54 in thestructure. These positions were subsequently used todefine the 2-fold non-crystallographic symmetry axisrelating one protein molecule to the other. This enabledthe initial electron density maps to be improved by2-fold averaging. After model building and refinement,the final model consisted of a total of 154 residues intwo chains. Molecule A comprised residues 28–102 of

the full length sR, whilst molecule B contained residues25–102. In other words, some eight residues at the Cterminus of sRN were not modelled for both chains,whereas five and two residues were absent at the Ntermini of the A and B chains, respectively. After refine-ment, the final Rcryst and Rfree values were 25.8% and27.8%, respectively. The model quality was evaluatedusing PROCHECK,56 approximately 88% of the residueslay in the most favoured regions of the Ramachandranplot, otherwise the overall criteria required for a struc-ture at a resolution of 2.4 A were either satisfied orexceeded. The parameters of the final structure aresummarised in Table 2.

Accession numbers

The coordinates and structure factor data for sRN havebeen deposited in the Protein Data Bank with accessioncodes 1H3L and R1H3LSF, respectively.

Acknowledgements

This work was funded by The Wellcome Trustand the BBSRC. N.B. was supported by theBBSRC-funded Cambridge and East Anglia Centrefor Structural Biology. We are grateful for supportand access to the SRS in Daresbury and the ESRFin Grenoble and particularly thank beamlinescientists J. Nicholson (SRS) and A. Perrakis(ESRF). L. Mitchenall, S. Mayer and T. Clarke areacknowledged for assistance during X-ray datacollection. We thank the EMBL Grenoble Out-station for supporting data collection at the ESRFunder the European Union TMR/LSF Programme.Finally we acknowledge M. Naldrett forN-terminal sequencing and A. Leech for assistancewith SELDI-MS data acquisition.

References

1. Burgess, R. R., Travers, A. A., Dunn, J. J. & Bautz,E. K. F. (1969). Factor stimulating transcription byRNA polymerase. Nature, 221, 43–46.

2. Lonetto, M., Gribskov, M. & Gross, C. A. (1992). Thesigma 70 family: sequence conservation andevolutionary relationships. J. Bacteriol. 174,3843–3849.

3. Lonetto, M. A., Brown, K. L., Rudd, K. E. & Buttner,M. J. (1994). Analysis of the Streptomyces coelicolorsigE gene reveals the existence of a subfamily ofeubacterial RNA polymerase s factors involved inthe regulation of extracytoplasmic functions. Proc.Natl Acad. Sci. USA, 91, 7573–7577.

4. Severinova, E., Severinov, K., Fenyo, D., Marr, M.,Brody, E. N., Roberts, J. W. et al. (1996). Domainorganization of the Escherichia coli RNA polymerases70 subunit. J. Mol. Biol. 263, 637–647.

5. Malhotra, A., Severinova, E. & Darst, S. A. (1996).Crystal structure of a s70 subunit fragment fromE. coli RNA polymerase. Cell, 87, 127–136.

6. Campbell, E. A., Muzzin, O., Chlenov, M., Sun, J. L.,Olson, A., Weinman, O. et al. (2002). Stucture of the

Table 2. Model parameters and quality

RefinementResolution range (A) 40–2.4Rcryst

a (on the basis of 95% of data; %) 25.8Rfree

a (on the basis of 5% of data; %) 27.8Residues with most favoured f/cb (%) 87.8

Root-mean-square deviations from idealityBond distances (A) 0.029Bond angles (8) 2.44Chiral volume (A3) 0.204Planar torsion angles (8) 4.2Staggered torsion angles (8) 20.8

Average B factors (A2)Main-chain atoms 34Side-chain atoms 36Water molecules 52Overall 34

a The R-factors Rcryst and Rfree are calculated as follows: R ¼PðlFobs 2 FcalclÞ=

PlFobsl £ 100; where Fobs and Fcalc are the

observed and calculated structure factor amplitudes,respectively.

b As calculated using PROCHECK.56

† http://www.solve.lanl.gov


http://www.solve.lanl.gov

bacterial RNA polymerase promoter specificity s

subunit. Mol. Cell, 9, 527–539.7. Sharp, M. M., Chan, C. L., Lu, C. Z., Marr, M. T.,

Nechaev, S., Merritt, E. W. et al. (1999). The interfaceof s with core RNA polymerase is extensive, con-served, and functionally specialized. Genes Dev. 13,3015–3026.

8. Burgess, R. R. & Anthony, L. (2001). How sigmadocks to RNA polymerase and what sigma does.Curr. Opin. Microbiol. 4, 126–131.

9. Murakami, K. S., Masuda, S. & Darst, S. A. (2002).Structural basis of transcription initiation: RNA poly-merase holoenzyme at 4 A resolution. Science, 296,1280–1284.

10. Vassylyev, D. G., Sekine, S.-I., Laptenko, O., Lee, J.,Vassylyeva, M. N., Borukhov, S. & Yokoyama, S.(2002). Crystal structure of a bacterial RNA polymer-ase holoenzyme at 2.6 A resolution. Nature, 417,712–719.

11. Hughes, K. T. & Mathee, K. (1998). The anti-sigmafactors. Annu. Rev. Microbiol. 52, 231–286.

12. Helmann, J. D. (1999). Anti-sigma factors. Curr. Opin.Microbiol. 2, 135–141.

13. Rudner, D. Z. & Losick, R. (2001). Morphologicalcoupling in development: lessons from prokaryotes.Dev. Cell, 1, 733–742.

14. Ohnishi, K., Kutsukake, K., Suzuki, H. & Iino, T.(1992). A novel transcriptional regulation mechanismin the flagellar regulon of Salmonella typhimurium: ananti-sigma factor inhibits the activity of the flagel-lum-specific sigma factor sF. Mol. Microbiol. 6,3149–3157.

15. Hughes, K. T., Gillen, K. L., Semon, M. J. & Karlinsey,J. E. (1993). Sensing structural intermediates inbacterial flagellar assembly by export of a negativeregulator. Science, 262, 1277–1280.

16. Orsini, G., Ouhammouch, M., Le Caer, J. P. & Brody,E. N. (1993). The asiA gene of bacteriophage T4codes for the anti-s70 protein. J. Bacteriol. 175, 85–93.

17. Colland, F., Orsini, G., Brody, E. N., Buc, H. & Kolb,A. (1998). The bacteriophage T4 AsiA protein: amolecular switch for s70-dependent promoters. Mol.Microbiol. 27, 819–829.

18. Aslund, F. & Beckwith, J. (1999). Bridge overtroubled waters: sensing stress by disulfide bondformation. Cell, 96, 751–753.

19. Paget, M. S. B., Kang, J.-G., Roe, J.-H. & Buttner, M. J.(1998). sR, an RNA polymerase sigma factor thatmodulates expression of the thioredoxin system inresponse to oxidative stress in Streptomyces coelicolorA3(2). EMBO J. 17, 5776–5782.

20. Paget, M. S. B., Molle, V., Cohen, G., Aharonowitz, Y.& Buttner, M. J. (2001). Defining the disulphide stressresponse in Streptomyces coelicolor A3(2): identifi-cation of the sR regulon. Mol. Microbiol. 42,1007–1020.

21. Kang, J.-G., Paget, M. S. B., Seok, Y.-J., Hahn, M.-Y.,Bae, J.-B., Kleanthous, C. et al. (1999). RsrA, an anti-sigma factor regulated by redox change. EMBO J.18, 4292–4298.

22. Paget, M. S. B., Bae, J.-B., Hahn, M.-Y., Li, W.,Kleanthous, C., Roe, J.-H. & Buttner, M. J. (2001).Mutational analysis of RsrA, a zinc-binding anti-sigma factor with a thiol-disulphide redox switch.Mol. Microbiol. 39, 1036–1047.

23. Fernandes, N. D., Wu, Q. L., Kong, D., Puyang, X.,Garg, S. & Husson, R. N. (1999). A mycobacterialextracytoplasmic sigma factor involved in survival

following heat shock and oxidative stress. J. Bacteriol.181, 4266–4274.

24. Newman, J. D., Anthony, J. R. & Donohue, T. J.(2001). The importance of zinc-binding to the func-tion of Rhodobacter sphaeroides ChrR as an anti-sigmafactor. J. Mol. Biol. 313, 485–499.

25. Helmann, J. D. (2002). The extracytoplasmic function(ECF) sigma factors. Advan. Microbiol. Physiol. 46,47–110.

26. Decatur, A. L. & Losick, R. (1996). Three sites of con-tact between the Bacillus subtilis transcription factorsF and its antisigma factor SpoIIAB. Genes Dev. 10,2348–2358.

27. Campbell, E. A. & Darst, S. A. (2000). The anti-sigmafactor SpoIIAB forms a 2:1 complex with sF, contact-ing multiple conserved regions of the sigma factor.J. Mol. Biol. 300, 17–28.

28. Kelemen, G. H., Viollier, P. H., Tenor, J. L., Marri, L.,Buttner, M. J. & Thompson, C. J. (2001). A connectionbetween stress and development in the multicellularprokaryote Streptomyces coelicolor A3(2). Mol.Microbiol. 40, 804–814.

29. Campbell, E. A., Masuda, S., Sun, J. L., Muzzin, O.,Olson, C. A., Wang, S. & Darst, S. A. (2002). Crystalstructure of the Bacillus stearothermophilus anti-sfactor SpoIIAB with the sporulation s factor sF. Cell,108, 795–807.

30. Holm, L. & Sander, C. (1995). DALI: a network toolfor protein structure comparison. Trends Biochem.Sci. 20, 478–480.

31. Brown, N. R., Noble, M. E. M., Ednicott, J. A.,Garman, E. F., Wakatsuki, S., Mitchell, E. et al.(1995). The crystal structure of cyclin A. Structure, 3,1235–1247.

32. Kosa, P. F., Ghosh, G., DeDecker, B. S. & Sigler, P. B.(1997). The 2.1 A crystal structure of an archaeal pre-initiation complex: TATA-box-binding protein/tran-scription factor (II)B core/TATA-box. Proc. NatlAcad. Sci. USA, 94, 6042–6047.

33. Lawrence, M. C. & Colman, P. M. (1993). Shape com-plementarity at protein–protein interfaces. J. Mol.Biol. 234, 946–950.

34. Noble, M. E. M., Endicott, J. A., Brown, N. R. &Johnson, L. N. (1997). The cyclin box fold: proteinrecognition in cell cycle and transcription control.Trends Biochem. Sci. 22, 482–487.

35. Bentley, S. D., Chater, K. F., Cerdeno-Tarraga, A. M.,Challis, G. L., Thomson, N. R., James, K. D. et al.(2002). Complete genome sequence of the modelactinomycete Streptomyces coelicolor A3(2). Nature,417, 141–147.

36. Qiu, J. & Helmann, J. D. (2001). The 210 region is akey promoter specificity determinant for the Bacillussubtilis ECF sigma factor sX and sW. J. Bacteriol. 183,1921–1927.

37. Garsin, D. A., Paskowitz, D. M., Duncan, L. & Losick,R. (1998). Evidence for common sites of contactbetween the antisigma factor SpoIIAB and its part-ners SpoIIAA and the developmental transcriptionfactor s28 in Bacillus subtilis. J. Mol. Biol. 284, 557–568.

38. Chadsey, M. S., Karlinsey, J. E. & Hughes, K. T.(1998). The flagellar anti-sigma factor FlgM activelydissociates Salmonella typhimurium s28 RNA polymer-ase holoenzyme. Genes Dev. 12, 3123–3136.

39. Kutsukake, K., Iyoda, S., Ohnishi, K. & Iino, T. (1994).Genetic and molecular analyses of the interactionbetween the flagellum-specific sigma and anti-sigmafactors in Salmonella typhimurium. EMBO J. 13,4568–4576.


40. Chadsey, M. S. & Hughes, K. T. (2001). A multipartiteinteraction between Salmonella transcription factors28 and its anti-sigma factor FlgM: implications fors28 holoenzyme destabilization through stepwisebinding. J. Mol. Biol. 306, 915–929.

41. Adelman, K., Orsini, G., Kolb, A., Graziani, L. &Brody, E. N. (1997). The interaction between theAsiA protein of bacteriophage T4 and the s70 subunitof Escherichia coli RNA polymerase. J. Biol. Chem. 272,27435–27443.

42. Severinova, E., Severinov, K. & Darst, S. A. (1998).Inhibition of Escherichia coli RNA polymerase bybacteriophage T4 AsiA. J. Mol. Biol. 279, 9–18.

43. Minakhin, L., Camarero, J. A., Holford, M., Parker,C., Muir, T. W. & Severinov, K. (2001). Mapping themolecular interface between the s70 subunit of E. coliRNA polymerase and T4 AsiA. J. Mol. Biol. 306,631–642.

44. Urbauer, J. L., Adelman, K., Urbauer, R. J., Simeonov,M. F., Gilmore, J. M., Zolkiewski, M. & Brody, E. N.(2001). Conserved regions 4.1 and 4.2 of s70 consti-tute the recognition sites for the anti-sigma factorAsiA, and AsiA is a dimer free in solution. J. Biol.Chem. 276, 41128–41132.

45. Lambert, L. J., Schirf, V., Demeler, B., Cadene, M. &Werner, M. H. (2001). Flipping a genetic switch bysubunit exchange. EMBO J. 20, 7149–7159.

46. Urbauer, J. L., Simeonov, M. F., Urbauer, R. J.,Adelman, K., Gilmore, J. M. & Brody, E. N. (2002).Solution structure and stability of the anti-sigmafactor AsiA: implications for novel functions. Proc.Natl Acad. Sci. USA, 99, 1831–1835.

47. Igo, M. M. & Losick, R. (1986). Regulation of a pro-moter that is utilised by minor forms of RNA poly-merase in Bacillus subtilis. J. Mol. Biol. 191, 615–624.

48. Matthews, B. W. (1968). Solvent content of proteincrystals. J. Mol. Biol. 33, 491–497.

49. Otwinowski, Z. & Minor, W. (1997). Processing ofX-ray diffraction data collected in oscillation mode.Methods Enzymol. 276, 307–326.

50. Collaborative Computational Project, No. 4 (1994).The CCP4 suite: programs for protein crystallo-graphy. Acta Crystallog. sect. D, 50, 760–763.

51. Brunger, A. T. (1993). Assessment of phase accuracyby cross validation—the free R-value—methods andapplications. Acta Crystallog. sect. D, 49, 24–36.

52. Terwilliger, T. C. & Berendzen, J. (1999). AutomatedMAD and MIR structure solution. Acta Crystallog.sect. D, 55, 849–861.

53. Cowtan, K. (1994). DM: an automated procedure forphase improvement by density modification. JointCCP4 ESF-EACBM Newsletter Protein Crystallog. 31,34–38.

54. Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard,M. (1991). Improved methods for building proteinmodels in electron density maps and the location oferrors in these models. Acta Crystallog. sect. D, 47,110–119.

55. Murshudov, G. N., Vagin, A. A. & Dodson, E. J.(1997). Refinement of macromolecular structures bythe maximum-likelihood method. Acta Crystallog.sect. D, 53, 240–255.

56. Laskowski, R. A., Macarthur, M. W., Moss, D. S. &Thornton, J. M. (1993). PROCHECK: a program tocheck the stereochemical quality of protein struc-tures. J. Appl. Crystallog. 26, 283–291.

57. Kraulis, P. J. (1991). MOLSCRIPT: a program toproduce detailed and schematic plots of proteinstructures. J. Appl. Crystallog. 24, 946–950.

58. Merritt, E. A. & Bacon, D. J. (1997). Raster3D: photo-realistic molecular graphics. Methods Enzymol. 277,505–524.

59. Nicholls, A., Sharp, K. A. & Honig, B. (1991). Proteinfolding and association—insights from the interfacialand thermodynamic properties of hydrocarbons.Proteins: Struct. Funct. Genet. 11, 281–296.

Edited by J. Karn

(Received 19 June 2002; received in revised form 30 August 2002; accepted 5 September 2002)


Documents

Identiﬁcation and Structure of the Anti-sigma Factor ... · Norwich NR4 7UH, UK 3Department of Molecular Microbiology, John Innes Centre, Norwich NR4 7UH UK 4School of Biological