The structure of TolB, an essential component of the tol-dependent translocation system, and its...

The structure of TolB, an essential component of the tol-dependent translocation system, and its protein–proteininteraction with the translocation domain of colicin E9Stephen Carr1,2†, Christopher N Penfold1†, Vicki Bamford1, Richard James1*and Andrew M Hemmings1,2*

Background: E colicin proteins have three functional domains, each of which isimplicated in one of the stages of killing Escherichia coli cells: receptor binding,translocation and cytotoxicity. The central (R) domain is responsible forreceptor-binding activity whereas the N-terminal (T) domain mediatestranslocation, the process by which the C-terminal cytotoxic domain istransported from the receptor to the site of its cytotoxicity. The translocation ofenzymatic E colicins like colicin E9 is dependent upon TolB but the details ofthe process are not known.

Results: We have demonstrated a protein–protein interaction between the Tdomain of colicin E9 and TolB, an essential component of the tol-dependenttranslocation system in E. coli, using the yeast two-hybrid system. The crystalstructure of TolB, a procaryotic tryptophan–aspartate (WD) repeat protein,reveals an N-terminal α+β domain based on a five-stranded mixed β sheet anda C-terminal six-bladed β-propeller domain.

Conclusions: The results suggest that the TolB-box residues of the T domain ofcolicin E9 interact with the β-propeller domain of TolB. The protein–proteininteractions of other β-propeller-containing proteins, the yeast yPrp4 proteinand G proteins, are mediated by the loops or outer sheets of the propellerblades. The determination of the three-dimensional structure of the Tdomain–TolB complex and the isolation of mutations in TolB that abolish theinteraction with the T domain will reveal fine details of the protein–proteininteraction of TolB and the T domain of E colicins.

IntroductionColicins are protein antibiotics produced by strains ofEscherichia coli and closely related bacteria. They are clas-sified into groups corresponding to the cell-surface recep-tor to which they bind in E. coli cells; for example theE colicins bind to the BtuB receptor, which is an essentialcomponent of the high-affinity vitamin B12 transportsystem [1]. Killing of E. coli cells by E colicins occurs inthree stages: receptor-binding, translocation and cytotoxic-ity. Commensurate with this, E colicin proteins have threefunctional domains, each of which is implicated in one ofthe stages. The central (R) domain is responsible forreceptor-binding activity whereas the N-terminal (T)domain mediates translocation, the process by which theC-terminal cytotoxic domain is transported from thereceptor to the site of its cytotoxicity. The mechanism ofcytotoxicity of E colicins results from either the formationof ion channels in the cytoplasmic membrane (colicin E1),an endonuclease activity that degrades DNA (colicins E2,E7, E8 and E9), or a ribonuclease activity that specificallycleaves 16S RNA (colicins E3, E4 and E6) or specific

transfer RNAs (tRNAs) (colicin E5) [2]. The DNase andRNase type E colicins have been termed the enzymaticE colicins [1]. The crystal structures of the DNasedomains of colicins E7 [3] and E9 [4] with their respectiveimmunity proteins have recently been determined. Theseare the first structures of members of the HNH family [5]of homing endonucleases that are found in both procary-otes and eucaryotes. The HNH endonucleases constituteone of four classes of homing-endonuclease enzymes thatare encoded by both introns and inteins and promote thehoming of genetic elements into allelic intronless orinteinless sites [6].

The Group A colicins (colicins A, E1–E9, K, L, N and S4)use the tol-dependent translocation system [7] whereasthe Group B colicins (colicins B, D, G, H, Ia, Ib, M, Q andV) use the ton-dependent translocation system [8]. Thetol-dependent translocation pathway is also used for theuptake of the DNA of filamentous phages. Four geneproducts are involved, tolQ, tolR, tolA and tolB, which areclustered at 17 minutes on the E. coli chromosomal map.

Addresses: Colicin Research Group, 1School ofBiological Sciences and 2School of ChemicalSciences, University of East Anglia, Norwich NR47TJ, UK.

*Corresponding authors.E-mail: [email protected]

[email protected]

†These authors contributed equally to this work.

Key words: β propeller, E colicins, protein–proteininteraction, TolB, translocation

Received: 7 September 1999Revisions requested: 11 October 1999Revisions received: 18 October 1999Accepted: 21 October 1999

Published: 22 December 1999

Structure 2000, 8:57–66

0969-2126/00/$ – see front matter © 2000 Elsevier Science Ltd. All rights reserved.

Research Article 57

st8106.qxd 02/09/2000 02:03 Page 57

The TolA, TolB, TolQ and TolR proteins are required totranslocate colicins A, E2–E9 and K whereas only theTolA and TolQ proteins, along with TolC, are required totranslocate colicin E1. The nucleotide sequences of thetol genes have been determined and the Tol proteinshave been found to be localized in the E. coli cell enve-lope (see [1] for a review). The TolQ, TolR and TolA pro-teins are associated with the cytoplasmic membrane.TolQ contains three membrane-spanning segments [9].TolA and TolR are anchored to the cytoplasmic mem-brane by their N-terminal domains whereas TolB is prin-cipally located in the periplasm, but is associated with theouter membrane via an interaction with the peptidogly-can-associated lipoprotein (Pal) [10]. The roles of the Tolproteins in the translocation of E colicins are the subjectof intensive experimental study.

The tol-dependent colicins contain a pentapeptidesequence (DGSGW; single-letter amino acid code) close tothe N terminus, which was originally designated the TolAbox [11], even though there was no experimental evidenceof any interaction with the TolA protein. The involvementof this pentapeptide sequence in tol-dependent transloca-tion was investigated by site-directed mutagenesis [12].Mutation of the two glycine residues to alanine had noaffect on the biological activity of colicin E9, whereasmutation of any of the other three residues to alanine com-pletely abolished biological activity. Using in vivo and invitro cross-linking experiments an interaction between theN-terminal 325 residues of colicin E3 and TolB wasdemonstrated [13]. This interaction was abolished if theN-terminal polypeptide of colicin E3 contained the muta-tion S37F, which is located in the putative TolA box. Thismutation also results in loss of biological activity of the full-size colicin E3 [14]. This suggests that the TolA boxshould rather be designated as a TolB box. Recently,residues located in the T domain of colicin N that arerequired for its interaction with TolA have been identified[15], but, because of the limited sequence homology, it hasproven difficult to identify a specific TolA-box sequencein other Group A colicins. The region of the T domain ofcolicin A that interacts with TolA has been identified asbeing localized between residues 32–107 by deletion sub-cloning and is thus located distal to the TolB box(DGTGW) of this colicin, which is located betweenresidues 11 and 15 [16]. Colicin A is the only colicin that isknown to contain both a TolA and a TolB box.

A detailed understanding of the translocation systemmight be of value in the design of novel antibiotics, asthere is evidence that some of the tol–pal genes areinvolved in maintaining cell-envelope structure andintegrity [17]. The tol–pal genes are also conserved inGram-negative bacteria, suggesting that their physiologi-cal role is important [18]. The TolB protein is also ofinterest as it is essential for mouse-lethal infection by

Salmonella typhimurium [19]. A particularly intriguingquestion is the difference between the translocation ofthe pore-forming colicins such as A and B, which act onthe surface of the cytoplasmic membrane, and the enzy-matic colicins like colicin E3 and E9. The ability todeliver the DNase or RNase domain of an E colicin,across both the outer and cytoplasmic membranes, intothe cytoplasm of E. coli cells is a unique event in procary-otic biology and might have potential in the design ofnew polypeptide antibiotics. Here we investigate theprotein–protein interaction between the T domain ofcolicin E9 with TolB using the yeast two-hybrid system,and we report the crystal structure of TolB.

ResultsInteractions between T(E9) and TolB Previous in vitro cross-linking experiments had indicatedan interaction between the N-terminal polypeptides ofcolicin A and E1 (but not of E3) with the TolA protein[20]. In contrast, both in vivo and in vitro cross-linkingexperiments showed an interaction between the N-termi-nal 325 residues of colicin E3 with TolB [13]. The yeasttwo-hybrid system has proven to be a valuable techniquein the study of protein–protein interactions [21]. In thissystem the two test proteins are fused to either the Gal4activation domain (AD) or to the Gal4 binding domain(BD). Interaction between the two test proteins reconsti-tutes the functional Gal4 transcriptional activator in theyeast nucleus, leading to expression of a β-galactosidasereporter gene, which can be assayed qualitatively by re-streaking colonies onto X-gal-containing plates. In theinitial series of experiments TolB, lacking the N-terminal23 residues that constitute the signal sequence, was fusedto BD and the N-terminal 164 residues of colicin E9[T(E9)] were fused to AD. Yeast cultures co-transformedwith both plasmids gave rise to blue colonies whenstreaked on X-gal plates (Table 1). The same result wasobserved with a TolB–AD and a T(E9)–BD construct, butnot when a TolB–BD construct was co-transformed withan empty AD vector. A positive yeast two-hybrid interac-tion was also observed between TolB and the N-terminal450 residues of colicin E9, which therefore contain boththe T and the R domains. This indicates that the interac-tion of the T domain with TolB is not affected by thepresence of the R domain in the fusion protein. Kineticassays with purified proteins will be needed to accuratelydetermine the equilibrium dissociation constant (Kd) forthe interaction between these proteins, but it has beenestimated that a Kd of < 1 µM is required for a detectableinteraction in the yeast two-hybrid system [22].

Mutations in T(E9) that affect its interaction with TolBIn order to confirm that the TolB box located in T(E9)was responsible for the protein–protein interaction withTolB we tested the effect of the alanine mutations of eachresidue in the TolB box (DGSGW), which we had made

58 Structure 2000, Vol 8 No 1

st8106.qxd 02/09/2000 02:03 Page 58

previously [12], on the protein–protein interactionbetween T(E9) and TolB in the yeast two-hybrid system.The results clearly show that the three alanine mutationsthat abolished the biological activity of colicin E9 alsoabolished the interaction with TolB, whereas mutation ofthe two glycine residues had no effect on biological activ-ity or on the interaction with TolB (Table 1). Theseresults confirm that the pentapeptide DGSGW constitutesa TolB box, and not a TolA box. Deletion of residues54–164 or 100–164 of the T+R(E9) construct fused to BDhad no affect on the protein–protein interaction withTolB. This result strongly suggests that the sole region ofT(E9) that interacts with TolB is located in the N-termi-nal 53 amino acids, which include the ‘TolB box’ pen-tapeptide (Figure 1). Recently it was shown from analysisof nuclear magnetic resonance (NMR) spectra that the Tdomain of colicin E9 is flexible and lacks a significantamount of regular secondary structure (GR Moore, per-sonal communication). It will be of interest to determinehow an unstructured T domain containing a TolB boxpentapeptide is specifically recognised by TolB.

Deletions in TolB that affect its interaction with T(E9)We attempted to localize the region of TolB that interactswith T(E9) by constructing N-terminal and C-terminaldeletions of the tolB gene in the BD vector. The preciselocation of the deletions was chosen, in the absence of thestructural information subsequently available, in order todivide the protein into four similarly sized regions. Theresults of yeast two-hybrid experiments using these con-structs showed that deletion of the first 119 residues ofTolB had no affect on its interaction with T(E9); however,deletion of the N-terminal 203 residues abolished theprotein–protein interaction (Table 2). Similarly, deletionof the C-terminal residues 205–431 or 304–431 abolishedthe protein–protein interaction. This clearly shows thatthe region of TolB that interacts with T(E9) is locatedbetween residues 119 and 431. The tolB2 (H147D)mutant of E. coli results in a significant reduction in sensi-tivity to the enzymatic E colicins (tolerance) [23], in con-trast to mutations in the BtuB receptor that lead to highlevel resistance to all E colicins. We used this clonedmutant gene to construct a fusion between the TolB2protein lacking the N-terminal signal sequence and BD ofGal4. When yeast cells were co-transformed with theresulting plasmid and the T(E9)–AD construct, there wasno effect of the H147D tolB2 mutation on theprotein–protein interaction between these two proteins

(Table 2). This result indicates that the colicin toleranceof the tolB2 mutation is not the result of a reduced interac-tion with T(E9).

TolB is a WD-repeat proteinBy sequence homology, TolB is a member of a large familyof proteins (> 2000) that contain between four and tenrepeats that often contain a conserved tryptophan–aspar-tate (WD) dyad, which are known as WD repeats [24].Such WD repeats are present in eucaryotic proteinsinvolved in a wide range of cellular functions, includingsignal transduction, transcription, splicing, cytoskeletalorganisation and vesicular fusion [25]. The crystal struc-ture of a small number of WD-repeat proteins have beendetermined and have revealed the presence of a β-pro-peller domain in which each repeat represents a singleblade and is folded into a four-stranded antiparallel twistedβ sheet. Few procaryotic proteins that contain WD repeatshave been identified. They include a serine/threonineprotein kinase PkwA from Thermomonospora curvata, theHatA and HatR proteins from Synechocystis sp. PCC6803,and the prolyl oligopeptidase homologue from Bacillus sub-tilis [24]. In a PSI-BLAST [26] search of the nonredundantGenBank database using the TolB protein sequence weidentified these sequences and also identified homology tothe procaryotic proteins X-Pro dipeptidyl-peptidase fromXanthomonas maltophilia, oligogalacturonide lyase fromErwinia caratovera, f416 from E. coli and orf c06024 (openreading frame c06024) of the archaebacteria Sulfolobus sol-fataricus. A reciprocal PSI-BLAST search using each of the

Research Article Structure of TolB Carr et al. 59

Table 1

Protein–protein interactions between T(E9) and TolB.

BD clone AD clone Colony colour

TolB T(E9) BlueT(E9) TolB BlueTolB – WhiteTolB T+R(E9) BlueT(E9–D35A) TolB WhiteT(E9–G36A) TolB BlueT(E9–S37A) TolB WhiteT(E9–G38A) TolB BlueT(E9–W39A) TolB WhiteT+R(E9) TolB BlueT+R(E9) ∆54–164* TolB BlueT+R(E9) ∆100–164 TolB Blue

*T+R(E9) ∆54–164 indicates that residues 54–164 have beendeleted from the 450-residue T+R–BD construct.

Figure 1

Sequence alignment of the translocationdomains of the tol-dependent enzymatic Ecolicins. The N-terminal 53 residues of theprotein sequences of the colicins are aligned,with the residues of the TolB box shown in red.

st8106.qxd 02/09/2000 02:03 Page 59

above protein sequences identified a significant homologyto TolB proteins in all cases. Until now no structural infor-mation was available for any of the procaryotic WD repeatproteins, although it was recently suggested, fromsequence-homology data, that the C terminus of TolB(residues 165–431) might contain a β-propeller domain

[24]. Methylamine dehydrogenase from Paracoccus denitrif-icans and cytochrome cd1 from Thiosphaera pantotrophawere not found in a PSI-BLAST search using the TolBsequence, but they have also been shown to contain β-pro-peller domains [27,28].

Crystal structure of TolBIn order to develop a rapid purification system for TolBwe introduced an NcoI site at base pair 70 (bp 70) and anXhoI site to replace the stop codon of the tolB gene usingthe polymerase chain reaction (PCR). The resultingNcoI–XhoI fragment was subcloned into the expressionvector pET21d, resulting in pRJ379 which encodes a 416-residue TolB protein lacking the N-terminal 23 signalsequence residues but having two additional C-terminalresidues (leucine, glutamate) introduced by the XhoI siteand a hexahistidine tag to facilitate purification of TolB(residues 24–439) by metal-chelate chromatography fromthe cytoplasm of induced E. coli BL21 (DE3) cells. Wehad shown, by complementation of an E. coli tolB2mutant, that the presence of the additional eight C-termi-nal residues did not effect the biological activity of the


Table 2

Deletions/mutations in TolB that affect the protein–proteininteractions with T(E9).

BD clone AD clone Colony colour

TolB T(E9) BlueTolB ∆1–119* T(E9) BlueTolB ∆1–203 T(E9) WhiteTolB ∆205–431 T(E9) WhiteTolB ∆304–431 T(E9) WhiteTolB2† T(E9) BlueTolB2 T+R(E9) Blue

*TolB ∆1–119 indicates that residues 1–119 of TolB have beendeleted from the TolB–BD fusion construct. †The TolB2 constructcontains the H147D mutation found in the tolB2 mutation of E. coli.

Table 3

Crystallographic statistics of the TolB structure.

λ1 = 1.387 Å λ2 = 1.07 Å λ3 = 1.386 Å λ4 = 0.91 Å

Resolution (Å) 2.6 2.6 2.6 2.0Completeness (%) 98.6 (98.0) 98.9 (98.5) 94.0 (93.0) 98.4 (97.4)Anomalous completeness (%) 94.6 (93.0) 95.7 (94.8) 85.2 (83.0) 98.2 (97.5)Rsym* (%) 6.6 (18.2) 5.4 (16.1) 6.3 (17.0) 5.2 (17.6)<I/σ(I) > 21.9 (7.9) 24.3 (9.5) 18.2 (7.1) 23.5 (7.0)Overall B factor (Å2) 23.1 23.3 23.3 18.7Anomalous multiplicity 2.0 (1.9) 1.8 (1.6) 1.9 (1.7) 1.9 (1.8)Phasing power

Acentrics† 2.05/3.00 1.14/2.33 2.08/2.18 –/2.95Centrics 1.13 0.69 1.25 –

RCullis‡ 0.71/0.63 0.81/0.78 0.70/0.80 –0.65

Refinement statisticsResolution range (Å) 20–2.0Rcryst

§# (%) 23.4 (23.5)Rfree

§# (%) 27.1 (27.3)Rmsds

Bonds (Å) 0.012Angles (°) 1.72Dihedrals (Å) 17.1Planarity of peptide groups (Å) 0.004Planarity of aromatic groups (Å) 0.02B values (mainchain atoms) (Å) 27.9B values (sidechain atoms) (Å) 29.1

Figures in parentheses refer to data in the highest resolution bin (butsee#). *Rsym = Σ Σ I i – ⟨I⟩/Σ I i, where ⟨I⟩ is the average ofsymmetry-equivalent reflections and the summation extends over allobservations for all unique reflections. †Phasing power for acentricreflections at each wavelength is presented for isomorphous and thenanomalous differences. The reference wavelength for phasingpurposes was λ4. ‡The Cullis R factor is defined asΣ [ FH –( FPH – FP )/Σ (FPH – FP )], where FH representsthe calculated heavy-atom structure factor. Figures for acentric

reflections only. §Rcryst = Σ Fo–Fc/ΣFo, where Fo and Fc arethe observed and calculated structure factors, respectively, and thesummation extends over all unique reflections in the quoted resolutionrange for which F > 2.0 σ(F) (i.e. the working set comprised 92.3% ofunique reflections). For Rfree the summation extends over a subset(5%) of reflections excluded from all stages of refinement. #Figures inparentheses for Rcryst and Rfree refer to these quantities calculatedusing all available data (resolution range 20–2.0 Å).

st8106.qxd 02/09/2000 02:03 Page 60

full-size TolB protein in vivo (data not shown). Crystalsthat diffracted to 2.0 Å resolution were prepared of thefusion protein. A ytterbium derivative was prepared andthe structure was solved by Yb-MAD (ytterbium multi-wavelength anomalous dispersion) methods (Table 3).The refined structure contains residues 35–431 of TolBand 230 water molecules (Figure 2). The N-terminal 11

residues of TolB and the C-terminal histidine-tag residuesare disordered and are not visible in the maps. The overallfold of the protein comprises a 131-residue N-terminaldomain (residues 35–165) consisting of a five-strandedmixed β sheet that sandwiches two major α helices againsta C-terminal six-bladed β-propeller domain (residues166–431). The ytterbium-binding site lies in the central


Figure 2

Crystal structure of TolB at 2.0 Å resolution.(a) Stereoview of the electron density for ablade from the β-propeller domain (residues202–242) contoured at 1.1σ. (b) Stereodiagram showing a Cα trace of TolB, withevery tenth residue labelled. (c) Overview ofthe TolB structure in which the α helices arecoloured red and the β sheets are colouredgreen. A number of short helical segments inthe loops connecting β strands of theC-terminal domain are also coloured red. Thepositions of residue His147 (which is mutatedto aspartate in the tolB2 mutant of E. coli andleads to significantly reduced sensitivity to theenzymatic E colicins) and Asp189 areindicated. Asp62, the other residue which ion-pairs with His147, is obscured in this view.

st8106.qxd 02/09/2000 02:03 Page 61

channel of the β-propeller domain, with ligands to the lan-thanide ion provided by the sidechain of Asp337 and fourclosely bound water molecules.

Searches of the Protein Data Bank (PDB) using DALI[29] for three-dimensional (3D) structural homology withthe refined TolB structure revealed alignments betweenthe N-terminal domain of TolB and a large number ofsimilar motifs in proteins of known structure. The most

significant alignment was with porphobilinogen deami-nase [30] (entry 1PDA) with a Z score of 6.1. The rootmean square deviation (rmsd) for this alignment was 2.8 Åand encompassed 60% of the residues of the TolBdomain. As expected, significant structural homology wasdetected between the blades of the C-terminal β-pro-peller domain of TolB and blades from the β-propellerdomains of a number of proteins in the databank. Forexample, in a pairwise structural alignment with the


Figure 3

Topology and internal structural homology ofthe β-propeller domain. (a) Closing the circleof β-propeller blades. A view down the centralbarrel of the β-propeller domain of TolB whereeach of the six repeated sequence motifs ofthe amino acid sequence that fold to give theβ-propeller structure is shown in a differentcolour (1, purple; 2, yellow; 3, red; 4, royalblue; 5, gold; and 6, green). This viewdemonstrates the mechanism of propellerclosure in TolB where the C-terminal β strandfrom the final sequence motif forms the innerβ strand of the four-stranded sheet that formsblade 1. (b) An alignment of the six sequencerepeats (motifs 1–6) that form the β-propellerdomain of the TolB protein. The startingresidue number of each motif is indicated.Residues that form β strands are enclosed inboxes. (c) Orthogonal views of asuperposition of the six blades of the TolBpropeller domain. The colouring of residues ineach of the six sequence motifs is the sameas in (a) and the nomenclature for the strandsfollows that in (b).

st8106.qxd 02/09/2000 02:03 2

six-bladed β-propeller domain of neuraminidase [31](entry 1EUT) the rmsd was 2.9 Å for an alignmentencompassing 85% of residues 166–431 of TolB. The Zscore for this alignment was 18.3.

The ββ-propeller domainEach blade of the propeller consists of twisted β sheetsthat are radially arranged around a central tunnel. Theinner β strand of blade 1 is made up of the C-terminalresidues 422–428, whereas β strand 2 of this blade consistsof the N-terminal residues of the β-propeller domain(residues 167–174; Figure 3). The mechanism by whichβ-propeller proteins close the circle between the first andlast blades (velcro) varies in the β-propeller structures thatare known. In a group of seven-bladed β-propeller pro-teins that includes the G-protein β subunit, galactoseoxidase and the yPrp4 protein, the N-terminal residuesconstitute the outer β sheet of blade 7 [27]. In contrast,the seven-bladed methylamine dehydrogenase [28] andthe five-bladed tachylectin-2 protein [32] use the samecircle-closing arrangement as TolB. It is of interest to

compare the individual blades of the propeller domain.Pairwise superimposition of blades 2–6 of the propellerdomain with first blade reveal that low sequence identitywithin the 44 amino acids of the repeat motif does notresult in significant structural divergence and the bladescan be superimposed with an average rmsd of 1.09 Å cal-culated over all aligned Cα positions (Table 4).

Our yeast two-hybrid results clearly show that theprotein–protein interaction between the N-terminaltranslocation domain of colicin E9 and TolB is essentialfor the cytotoxicity of the enzymatic E colicin. Three ofthe five residues of the TolB box of colicin E9 are criticalin the protein–protein interaction. Much less is known ofthe essential region(s) of TolB, except that the N-termi-nal 119 residues are not required and that the H147Dmutation does not abolish the interaction, even thoughthis mutation confers tolerance to E colicins on host E. colicells. His147 lies on the C-terminal helix of the N-termi-nal domain of TolB and forms both intradomain and inter-domain ion-pairing interactions (with residues Asp62 andAsp189, respectively; Figure 2).

As we had shown that residues 1–203 of TolB, whichcontain all of the N-terminal domain and only a small part ofthe β propeller, did not interact with T(E9) in the yeast two-hybrid experiments (Table 1), it is reasonable to assumethat T(E9) interacts with the β-propeller domain of TolB.As single mutations in the TolB box of T(E9) abolish theinteraction with TolB, it is likely that similar mutations willbe located in the β-propeller domain of TolB which affectits interaction with T(E9). It is our aim to isolate furthermutants in the tolB gene that affect the interaction betweenTolB and T(E9) and then to map them on the structure ofthe TolB–T(E9) complex when it is determined.

All β-propeller proteins use their central tunnel or theentrance to the tunnel to coordinate a ligand or to carry


Table 4

Comparison of TolB propeller blades.

Blade Residues Matched Identity (%) Match rate (%) Rmsd (Å)

2 204–242 30 33.3 68 0.763 246–286 32 28.1 73 1.054 290–330 29 13.8 66 1.095 334–374 33 15.2 75 1.216 377–418 33 15.2 75 1.35

The coordinates of residues of TolB (less those in connecting loops)comprising each of the propeller blades 2–6 were compared withthose of blade 1 (residues 166–200 and 421–429) using TOP [51].The number of residues matched in the structural alignment is reportedwith the percentage of these having the same identity in both blades.The match rate reflects the percentage of residues that are structurallyaligned. The rmsd is calculated over matched Cα atoms.

Figure 4

Van der Waals molecular surface of TolBcolour coded according to electrostaticpotential, where blue and red areas representpositively and negatively charged regions,respectively. The two views are from oppositeends of the β propeller and show the acidicnature of the channel.

st8106.qxd 02/09/2000 02:03 Page 63

out some catalytic function that has to be preserved bythe structural rigidity of the propellers [27]. For example,the central tunnel of prolyl oligopeptidase allows accessof oligopeptides of up to 30 amino acids to the active site,which is hidden in a large cavity at the interface of thecatalytic and β-propeller domains. In contrast, analysis ofmutations in the β-propeller protein yPrp4 that abolish itsprotein–protein interaction with yPrp3 show that they arelocalized in loops of the wide surface of the torus [33].Similarly, activation of phospholipase Cβ2 or adenylcyclase by G proteins involves different residues in theouter sheets of the blades of the β propeller of the Gprotein [34]. It will be of interest to see how the Tdomain of colicin E9 interacts with residues of TolBwhen the structure of the complex is determined. Inter-estingly, when an electrostatic-potential-coded van derWaals surface is calculated for TolB it clearly reveals thecentral channel of the β-propeller domain to be highlyacidic (Figure 4). A similar analysis of the β-propellerdomains in protein structures available in the PDB revealthis observation to be qualitatively unique to TolB. Inretrospect then, the tight binding of a lanthanide ion inthis channel is not so surprising. However, the relevanceof this charge distribution (if any) to the function of theprotein is not so clear. Certainly, without significant per-turbation to what seems to be a rather rigid propellerstructure, the channel is of insufficient diameter toaccommodate the passage of any species larger than a sol-vated ion; the maximum diameter of the channel shrinksto only 1.5 Å when a solvent-accessible molecular surfaceis calculated (data not shown).

Biological implicationsIn contrast to the ton-dependent translocation system,which has a role in the energy-dependent uptake ofligands like vitamin B12 and iron siderophores [35], thenormal role of the tol-dependent translocation system inEscherichia coli is not clear. The diversity of phenotypes oftol–pal mutants of E. coli suggests a role for the Tol–Palproteins in maintaining cell-envelope integrity [36]. Therecent finding that TolA and TolB interact with porins invitro has suggested a possible role in the integration ofporin proteins in the E. coli outer membrane [37], which isconsistent with the localization of Tol proteins in contactsites between the cytoplasmic and outer membranes of E.coli cells [38]. The precise function of the Tol proteins inthese processes remains to be determined, as does themechanism by which E colicins have parasitized thissystem to facilitate cell killing. Our results indicate thatthe protein–protein interaction between the TolB box inthe T domain of colicin E9 and the β-propeller domain ofTolB is important in the translocation of E colicins. Thephenotype of the H147D mutation also perhaps indicatesa direct role for the N-terminal domain of TolB in thetranslocation of E colicins, possibly via its interaction withthe outer-membrane proteins Pal, Lpp and OmpA [10].

It is intriguing that the Sec13 protein is amongst theWD-repeat proteins to which TolB shows homology;this protein is involved in the vesicular trafficking of pro-teins in yeast and nuclear-pore biogenesis in humans[39]. In addition, the G-protein β subunit is the only β-propeller structure known not to be part of a secretedprotein [27]. It is tempting to speculate that the β-pro-peller structure has properties that make it particularlysuitable for a role in protein trafficking, or, that inexported proteins like TolB, perhaps the β-propellerstructure can accurately refold after crossing a mem-brane to produce a rigid structure capable of an enzy-matic activity or a specific protein–protein interaction.

Materials and methodsYeast two-hybrid experimentsThe yeast two-hybrid vectors pAS2-1 and pACT2 were purchased fromClontech, UK. The construction of Gal4 fusion clones using thesevectors and transformation of the resulting clones into S. cerevisae strainY187 were as described in the manufacturer’s instruction manual.Fusions of polypeptides with the Gal4 BD domain were usually con-structed by introducing an NdeI site at the 5′ end and an XhoI site at the3′ end of the fusion fragment and subsequent ligation of the fragmentinto the pAS2-1 plasmid. Fusions of polypeptides with the Gal4 ADdomain were usually constructed by introducing an NcoI site at the5′ end and an XhoI site at the 3′ end of the fusion fragment and subse-quent ligation of the fragment into the pACT2 plasmid. Qualitativeβ-galactosidase assays were performed by streaking test yeast coloniesonto SD-Leu-Trp (SD, synthetic dropout media; Clontech) plates contain-ing X-gal (80 µg ml–1) and one tenth of the volume of 10 × BU salts (70 gNa2HPO4⋅7H20, 30 g NaH2P04 per litre of H2O, adjusted to pH 7.0).

Structure determinationC-terminally histidine-tagged TolB lacking the 23 amino acid signalsequence (residues 24–439) was produced by overexpression inE. coli BL21(DE3). The protein was purified by conventional metal-chelate chromatography and concentrated to 5 mg/ml in 10 mMTris.HCl at pH 7.5. Large, single crystals of typical dimensions250 × 150 × 10 µm3 grew after 24–72 h from solutions containing18% (w/v) polyethylene glycol (PEG) of average molecular weight 8Kand 100 mM MES pH 6.5 at 18°C. These crystals could be cryopro-tected by transferring them to crystallization solution containing 25%(v/v) ethyene glycol. The space group was P21 with apparent cell para-meters a = 63.6 Å, b = 40.2 Å, c = 77.7 Å, β = 110.2° containing asingle molecule of TolB in the asymmetric unit. The crystallization con-ditions and cell parameters are similar to those previously reported[40]. Preliminary attempts to solve the structure of TolB using the multi-ple isomorphous replacement method revealed only a single usefulderivative prepared by soaking crystals in 22 mM ytterbium trichloridefor at least 1 h at 20°C prior to cryoprotection. It was therefore decidedto attempt to solve the structure by Yb-MAD. X-ray diffraction datawere collected from a single crystal derivitized with ytterbium at threewavelengths (λ1–λ3) around the ytterbium LIII edge on station X31 atDESY, Hamburg, and processed using DENZO [41] to 2.6 Å resolu-tion. All data collection was performed at 100K. The crystal wasretrieved and transferred to station X11 at DESY where a further high-energy remote dataset was collected at an X-ray wavelength of 0.91 Å,which was processed to 2.0 Å resolution (λ4). The position of thesingle ytterbium ion in the asymmetric unit was straightforwardly deter-mined by inspection of isomorphous and anomalous difference Patter-son maps of exceptional clarity. Initial phase estimates were obtainedby MAD phasing with the programme SHARP [42] and improved bysolvent flattening using SOLOMON [43] at a nominal solvent contentof 44% using the λ4 data as the reference set. The mean figure ofmerit for acentric reflections to 2.6 Å resolution was 0.697 before


st8106.qxd 02/09/2000 02:03 Page 64

solvent flattening. The initial Fourier map calculated with solvent-flat-tened MAD phases showed clear and contiguous electron density forlarge parts of the molecule and it was possible to completely trace a C-terminal β-propeller domain and the two major helices of an N-termi-nal domain of the protein using O [44]. Following limited positionalrefinement using REFMAC [45] the R factor for this model (residues55–72 and 141–431) for all amplitudes to 2.6 Å resolution had con-verged to 35.3% (Rfree = 40.2%). After three cycles of phase combina-tion the model consisted of residues 34–86 and 96–431. Inspection ofdifference Fourier maps at this stage showed additional interpretableelectron density in the region of residues 87–95 in the N-terminaldomain and this was duly added to the structural model. For refine-ment, 5% of the reflections were set aside for calculation of the Rfreevalue [46] and maximum likelihood refinement of this partial structureusing REFMAC and with a single overall temperature factor gave amodel with Rcryst = 27.3% (Rfree = 34.7%).

From this point, refinement of the model consisting of residues 35–431proceeded with simulated annealing with CNS [47] using all data in therange 20–2.0 Å resolution. Refinement was interspersed with cycles ofmanual rebuilding with reference to σA-weighted difference maps [48].This process converged to give Rcryst = 26.4% (Rfree = 30.3%) for amodel with individual isotropic temperature factors. At this point watermolecules were added into significant residual electron density usingARP [49]. A total of 230 water molecules were added to the modeland refinement and adjustment of these was performed until no furtherdifference density could reasonably be ascribed to water. At this stagethe refinement had converged to give Rcryst = 23.4% (Rfree = 27.1%)for data in the range 20–2.0 Å resolution with F > 2.0 σ(F) and withrefinement of individual isotropic temperature factors and use of a bulk-solvent correction. The corresponding values calculated using all datain the same resolution range are Rcryst = 23.5% and Rfree = 27.3%. Weascribe the relatively high residual R-factor values of the final model toour difficulty in accurately modelling diffraction from a ytterbium ion andthe general mobility of the N-terminal domain. This mobility is reflectedin the mean temperature factor for atoms of the N-terminal domain(34.0 Å2) relative to that of the propeller domain (25.8 Å2). Despite this,the stereochemistry of the final model is good; when analyzed usingPROCHECK [50] the final structure has 99.2% of residues in themost-favoured regions of the Ramachandran plot and only threeresidues (Ala50, Thr122 and Arg245) are in generously allowedregions. No residues fall in disallowed regions. In addition, the rmsds ofbond lengths and angles fall within acceptable limits. The averageisotropic temperature factors for mainchain and sidechain atoms of themodel are 27.9 Å2 and 30.3 Å2, respectively. For the water moleculesthe corresponding average is 38.0 Å2. Searches for structural homol-ogy were performed using DALI [29] and structural alignments of theblades of the β-propeller domain were performed using TOP [51].Figures were produced using a combination of the programsMOLSCRIPT [52], Raster3D [53] and SPOCK [54].

Accession numbersThe atomic coordinates for TolB have been deposited in the ProteinData Bank with accession code 1C5K.

Note added in proofThe structure of TolB has also recently been reported in Abergel, C., Bou-veret, E., Claverie, J-.M., Brown, K., Rigal, A., Lazdunski, C. and Bénédetti,H. (1999). Structure of the Escherichia coli TolB protein determined byMAD methods at 1.95 Å resolution. Structure 7, 1291-1300.

AcknowledgementsWe thank Jean Claude Lazzaroni for providing the tolB2 mutant. We alsothank all members of our labs, past and present, for their hard work andenthusiastic support of the colicin project at UEA. We acknowledge accessto the EMBL X31 and X11 beamlines at the DORIS storage ring, DESY,and to the SRS station 9.6, Daresbury Laboratory. The Colicin ResearchGroup is generously supported by The Wellcome Trust and the Biotechnol-ogy and Biological Sciences Research Council of the UK.

References1. James, R., Kleanthous, C. & Moore, G.R. (1996). The biology of E

colicins: paradigms and paradoxes. Microbiology 142, 1569-1580.2. Ogawa, T., Tomita, K., Ueda, T., Watanabe, K., Uozumi, T. & Masaki,

H. (1999). A cytotoxic ribonuclease targeting specific transfer RNAanticodons. Science 283, 2097-2100.

3. Ko, T.P., Liao, C.C., Ku, W.Y. Chak, K.F. & Yuan, H.S. (1999). Thecrystal structure of the DNase domain of colicin E7 in complex with itsinhibitor Im7 protein. Structure 7, 91-102.

4. Kleanthous, C., et al., & Hemmings, A.M. (1999). Structural andmechanistic basis of immunity towards endonuclease colicins. Nat.Struct. Biol. 6, 243-252.

5. Shub, D.A., Goodrich-Blair, H. & Eddy, S.R. (1994). Amino acidsequence motif of group I intron endonucleases is conserved in openreading frames of group II introns. Trends Biochem. Sci. 19, 402-404.

6. Belfort, M. & Roberts, R.J. (1997). Homing endonucleases: keepingthe house in order. Nucleic Acids Res. 25, 3379-3388.

7. Lazdunski, C. (1995). Colicin import and pore formation: a system forstudying protein transport across membranes? Mol. Microbiol.16, 1059-1066.

8. Braun, V. (1995). Energy-coupled transport and signal transductionthrough the Gram-negative outer membrane via TonB-ExbB-ExbD-dependent receptor proteins. FEMS Microbiol. Rev. 16, 295-307.

9. Kampfenkel, K. & Braun, V. (1993). Membrane topologies of the TolQand TolR proteins of Escherichia coli: inactivation of TolQ by amissense mutation in the proposed first transmembrane segment.J. Bacteriol. 175, 4485-4491.

10. Bouveret, E., Derouiche, R., Rigal, A., Lloubès, R., Lazdunski, C. &Bénédetti, H. (1995). Peptidoglycan-associated lipoprotein-TolBinteraction. A possible key to explaining the formation of contact sitesbetween the inner and outer membranes of Escherichia coli. J. Biol.Chem. 270, 11071-11077.

11. Pilsl, H. & Braun, V. (1995). Novel colicin 10: assignment of fourdomains to TonB- and TolC-dependent uptake via the Tsx receptorand to pore formation. Mol. Microbiol. 16, 57-67.

12. Garinot-Schneider, C., Penfold, C.N., Moore, G.R., Kleanthous, C. &James, R. (1997). Identification of residues in the putative TolA boxwhich are essential for the toxicity of the endonuclease toxin colicinE9. Microbiology 143, 2931-2938.

13. Bouveret, E., Rigal, A., Lazdunski, C. & Bénédetti, H. (1997). TheN-terminal domain of colicin E3 interacts with TolB which is involvedin the colicin translocation step. Mol. Microbiol. 23, 909-920.

14. Escuyer, V. & Mock, M. (1987). DNA sequence analysis of threemissense mutations affecting colicin E3 bactericidal activity. Mol.Microbiol. 1, 82-85.

15. Raggett, E.M., Bainbridge, G., Evans, L.J., Cooper, A. & Lakey, J.H.(1998). Discovery of critical Tol A-binding residues in the bactericidaltoxin colicin N: a biophysical approach. Mol. Microbiol.28, 1335-1343.

16. Bouveret, E., Rigal, A., Lazdunski, C. & Bénédetti, H. (1998). Distinctregions of the colicin A translocation domain are involved in theinteraction with TolA and TolB proteins upon import into Escherichiacoli. Mol. Microbiol. 27, 143-157.

17. Lazzaroni, J.C. & Portalier, R. (1992). The excC gene of Escherichiacoli K-12 required for cell envelope integrity encodes thepeptidoglycan-associated lipoprotein (PAL). Mol. Microbiol. 6, 735-42.

18. Journet, L., Rigal, A., Lazdunski, C. & Bénédetti, H. (1999). Role ofTolR N-terminal, central, and C-terminal domains in dimerization andinteraction with TolA and TolQ. J. Bacteriol. 181, 4476-4484.

19. Bowe, F., Lipps, C.J., Tsolis, R.M., Groisman, E., Heffron, F. & Kusters,J.G. (1998). At least four percent of the Salmonella typhimuriumgenome is required for fatal infection of mice. Infect. Immunol.66, 3372-3377.

20. Bénédetti, H., Frenette, M., Baty, D., Knibiehler, M., Pattus, F. &Lazdunski, C. (1991). Individual domains of colicins confer specificityin colicin uptake, in pore-properties and in immunity requirement.J. Mol. Biol. 217, 429-439.

21. Chien, C.T., Bartel, P.L., Sternglanz, R. & Fields, S. (1991). The two-hybrid system: a method to identify and clone genes for proteins thatinteract with a protein of interest. Proc. Natl Acad. Sci. USA88, 9578-9582.

22. Estojak, J., Brent, R. & Golemis, E.A. (1995). Correlation of two-hybridaffinity data with in vitro measurements. Mol. Cell. Biol.15, 5820-5829.

23. Nagel de Zwaig, R. & Luria, S.E. (1967). Genetics and physiology ofcolicin-tolerant mutants of Escherichia coli. J. Bacteriol.94, 1112-1123.


st8106.qxd 02/09/2000 02:03 Page 65

24. Ponting, C.P. & Pallen, M.J. (1999). A β-propeller domain within TolB.Mol. Microbiol. 31, 739-740.

25. Neer, E.J., Schmidt, C.J., Nambudripad, R. & Smith, T.F. (1994). Theancient regulatory-protein family of WD-repeat proteins. Nature371, 297-300.

26. Altschul, S.F., et al., & Lipman, D.J. (1997). Gapped BLAST and PSI-BLAST: a new generation of protein database search programs.Nucleic Acids Res. 25, 3389-3402.

27. Baker, S.C., Saunders, N.F., Willis, A.C., Ferguson, S.J., Hajdu, J. &Fulop, V. (1997). Cytochrome cd1 structure: unusual haemenvironments in a nitrite reductase and analysis of factors contributingto β-propeller folds. J. Mol. Biol. 269, 440-455.

28. Chen, L., et al., & Mathews, F.S. (1998). Refined crystal structure ofmethylamine dehydrogenase from Paracoccus denitrificans at 1.75 Åresolution. J. Mol. Biol. 276, 131-149.

29. Holm, L. & Sander, C. (1993). Protein structure comparison byalignment of distance matrices. J. Mol. Biol. 233, 123-138.

30. Louie, G.V., et al., & Jordan, P.M. (1992). Structure of porphobilinogendeaminase reveals a flexible multidomain polymerase with a singlecatalytic site. Nature 359, 33-39.

31. Gaskell, A., Crennell, S. & Taylor, G. (1995). The three domains of abacterial sialidase: a β-propeller, an immunoglobulin module and agalactose-binding jelly-roll. Structure 3, 1197-1205.

32. Beisel, H.G., Kawabata, S., Iwanaga, S., Huber, R. & Bode, W.(1999). Tachylectin-2: crystal structure of a specific GlcNAc/GalNAc-binding lectin involved in the innate immunity host defense of theJapanese horseshoe crab Tachypleus tridentatus. EMBO J.18, 2313-2322.

33. Ayadi, L., Callebaut, I., Saguez, C., Villa, T., Mornon, J.P. & Banroques,J. (1998). Functional and structural characterization of the prp3 bindingdomain of the yeast prp4 splicing factor. J. Mol. Biol. 284, 673-687.

34. Panchenko, M.P., et al., & Neer, E.J. (1998). Sites important forPLCβ2 activation by the G protein βγ subunit map to the sides of theβ propeller structure. J. Biol. Chem. 273, 28298-28304.

35. Buchanan, S.K., et al., & Deisenhofer, J. (1999). Crystal structure ofthe outer membrane active transporter FepA from Escherichia coli.Nat. Struct. Biol. 6, 56-63.

36. Bernadac, A., Gavioli, M., Lazzaroni, J.C., Raina, S. & Lloubès, R.(1998). Escherichia coli tol–pal mutants form outer membranevesicles. J. Bacteriol. 180, 4872-4878.

37. Rigal, A., Bouveret, E., Lloubès, R., Lazdunski, C. & Bénédetti, H.(1997). The TolB protein interacts with the porins of Escherichia coli.J. Bacteriol. 179, 7274-7279.

38. Guihard, G., Boulanger, P., Bénédetti, H., Lloubès, R., Besnard, M. &Letellier, L. (1994). Colicin A and the Tol proteins involved in itstranslocation are preferentially located in the contact sites betweenthe inner and outer membranes of Escherichia coli cells. J. Biol.Chem. 269, 5874-5880.

39. Siniossoglou, S., et al., & Hurt, E.C. (1996). A novel complex ofnucleoporins, which includes Sec13p and a Sec13p homolog, isessential for normal nuclear pores. Cell 84, 265-275.

40. Abergel, C., et al., & Bénédetti, H. (1998). Crystallization andpreliminary crystallographic study of a component of the Escherichiacoli tol system: TolB. Acta Crystallogr. D 54, 102-104.

41. Otwinowski, Z. & Minor, W. (1997). Processing of X-ray diffractiondata collected in oscillation mode. Methods Enzymol. 276, 307-326.

42. De la Fortelle, E. & Bricogne, G. (1997). Maximum-likelihood heavy-atom parameter refinement for multiple isomorphous replacement andmultiwavelength anomalous diffraction methods. Methods Enzymol.276, 472-494.

43. Abrahams, J.P. & Leslie, A.G.W. (1996). Methods used in thestructure determination of bovine mitochondrial F-1 ATPase. ActaCrystallogr. D 52, 30-42.

44. Jones, T.A., Zou, J.Y., Cowan, S.W. & Kjeldgaard, M. (1991). Improvedmethods for building protein models in electron-density maps and thelocation of errors in these models. Acta Crystallogr. A 47, 110-119.

45. Murshudov, G.N., Vagin, A.A. & Dodson, E.J. (1997). Refinement ofmacromolecular structures by the maximum-likelihood method. ActaCrystallogr. D 53, 240-255.

46. Brünger, A.T. (1992). Free R value – a novel statistical quantity forassessing the accuracy of crystal structures. Nature 355, 472-475.

47. Brünger, A.T., et al., & Warren, G.L. (1998). Crystallography and NMRsystem: a new software suite for macromolecular structuredetermination. Acta Crystallogr D 54, 905-921.

48. Read, R.J. (1986). Improved fourier coefficients for maps usingphases from partial structures with errors. Acta Crystallogr. A42, 140-149.

49. Lamzin, V.S. & Wilson, K.S. (1997). Automated refinement for proteincrystallography. Methods Enzymol. 277, 269-305.

50. Laskowski, R.A., Macarthur, M.W., Moss, D.S. & Thornton, J.M.(1993). PROCHECK — a program to check the stereochemical qualityof protein structures. J. Appl. Crystallogr. 26, 283-291.

51. Lu, G. (1996). A WWW service system for automatic comparison ofprotein structures. Protein Databank Quarterly Newsletter 78, 10-11.

52. Kraulis, P.J. (1991). MOLSCRIPT — a program to produce bothdetailed and schematic plots of protein structures. J. Appl. Crystallogr.24, 946-950.

53. Merritt, E.A. & Bacon, D.J. (1997). Raster3D: photorealistic moleculargraphics. Methods Enzymol. 277, 505-524.

54. Christopher, J.A. (1997). SPOCK: The Structure Properties andObservation Toolkit. Center for Macromolecular Design, Texas A&MUniversity, College Station Tx 77843-2128.


Because Structure with Folding & Design operates a‘Continuous Publication System’ for Research Papers, thispaper has been published on the internet before being printed(accessed from http://biomednet.com/cbiology/str). Forfurther information, see the explanation on the contents page.

st8106.qxd 02/09/2000 02:03 Page 66

The structure of TolB, an essential component of the tol-dependent translocation system, and its...

Documents

Transcript of The structure of TolB, an essential component of the tol-dependent translocation system, and its...