The NMR Structure of FliK, the Trigger for the Switch of Substrate Specificity in the Flagellar Type III Secretion Apparatus

doi:10.1016/j.jmb.2011.04.008

J. Mol. Biol. (2011) 409, 558–573 Contents lists available at www.sciencedirect.com

Journal of Molecular Biology j o u r n a l h o m e p a g e : h t t p : / / e e s . e l s e v i e r. c o m . j m b

The NMR Structure of FliK, the Trigger for the Switch of Substrate Specificity in the Flagellar Type III Secretion Apparatus Shino Mizuno 1 , Hirokazu Amida 2 , Naohiro Kobayashi 3 , Shin-Ichi Aizawa 1 ⁎ and Shin-ichi Tate 2 ⁎ 1

Department of Life Sciences, Prefectural University of Hiroshima, 526 Nanatsuka, Shobara, Hiroshima 727-0023, Japan 2 Department of Mathematical and Life Sciences, Graduate School of Science, Hiroshima University, 1-3-1 Kagamiyama, Higashi-Hiroshima 739-8526, Japan 3 Institute for Protein Research, Osaka University, 3-2 Yamadaoka, Suita, Osaka 565-0871, Japan Received 20 January 2011; received in revised form 1 April 2011; accepted 4 April 2011 Available online 12 April 2011 Edited by M. F. Summers Keywords: flagellum; length control; type III secretion system; FlhB; NMR

The flagellar cytoplasmic protein FliK controls hook elongation by two successive events: by determining hook length and by stopping the supply of hook protein. These two distinct roles are assigned to different parts of FliK: the N-terminal half (FliKN) determines length and the C-terminal half (FliKC) switches secretion from the hook protein to the filament protein. The interaction of FliKC with FlhB, the switchable secretion gate, triggers the switch. By NMR spectroscopy, we demonstrated that FliK is largely unstructured and determined the structure of a compact domain in FliKC. The compact domain, denoted the FliKC core domain, consists of two α-helices, a β-sheet with two parallel and two antiparallel strands, and several exposed loops. Based on the functional data obtained by a series of deletion mutants of the FliKC core domain, we constructed a model of the complex between the FliKC core domain and FlhBC. The model suggested that one of the FliKC loops has a high probability of interacting with the C-terminal domain of FlhB (FlhBC) as the FliK molecule enters the secretion gate. We suggest that the autocleaved NPTH sequence in FlhB contacts loop 2 of FliKC to trigger the switching event. This contact is sterically prevented when NPTH is not cleaved. Thus, the structure of FliK provides insight into the mechanism by which this bifunctional protein triggers a switch in the export of substrates. © 2011 Elsevier Ltd. All rights reserved.

Introduction

*Corresponding authors. E-mail addresses: [email protected]; [email protected]. Abbreviations used: NOE, nuclear Overhauser enhancement; HSQC, heteronuclear single quantum coherence; RDC, residual dipolar coupling; GST, glutathione S-transferase; hetNOE, heteronuclear nuclear Overhauser enhancement; EDTA, ethylenediaminetetraacetic acid; PDB, Protein Data Bank.

Many bacterial species can swim in liquid, using one flagellum or many flagella.1 The flagellum selfassembles from a large number of different proteins; it is structurally divided into three substructures: the filament, the hook, and the basal body.2,3 The assembly of the flagellum proceeds from the cellproximal to the cell-distal substructures: the basal body formation is followed by the biogenesis of the hook and finally by growth of the filament.4 The filament and the hook are external to the cytoplasmic

0022-2836/$ - see front matter © 2011 Elsevier Ltd. All rights reserved.

559

NMR Structure of the Flagellar Protein Flik

membrane; their components, the hook protein (FlgE) and the filament protein (FliC), are exported by a flagellar-specific secretion apparatus that belongs to the type III secretion system (T3SS).1,4 In the Gram-negative bacterium Salmonella enterica serovar Typhimurium (S. Typhimurium), the filament has a variable length that depends upon the growth phase or the age of the filament. In contrast, the hook, a short tubular structure connecting the basal body and the filament, has a controlled length; the average length is 55 nm with a standard deviation of 6 nm.5 Although its length is controlled, this rather large deviation, 10% of the average, suggests a loose mechanism of length control. Two flagellar-specific proteins are known to be directly involved in hook length control: FliK and FlhB.5,6 FlhB is an integral membrane protein that switches the substrate specificity of the T3SS.7,8 FlhB has putative transmembrane domains in its N-terminal part and a hydrophilic globular domain in its C-terminal part,9,10 which is cytoplasmic. The soluble C-terminal domain of FlhB (FlhBC), undergoes autocleavage at the NPTH loop, where the N–P bond is cleaved; FlhBC is thereby divided into two subdomains: FlhBCN (residues 211–269) and FlhBCC (residues 270–383).11,12 A site-directed mutant having APTH instead of NPTH prevents cleavage of the loop.12,13 Mutants that slow down autocleavage of this loop cause delayed switching.13 Autocleavage of the loop was suggested to cause a structural change in FlhBC and thereby to switch its substrate specificity.12 However, recent X-ray studies that compared the structures of the non-cleaved and cleaved forms of the C-terminal domains of YscU and EscU (FlhB homologs of Yersinia and Escherichia coli injectisomes, respectively) have demonstrated that the cleavage of the NPTH loop does not alter the structure; the only structural changes are limited to the loop.14–16 Thus, the substrate specificity switch does not seem to be mediated by the changes in the overall fold of FlhBC. FliK, a soluble cytoplasmic protein, is secreted during flagellar formation.17 In fliK deletion mutants, flagellar filament formation is not initiated. Instead, the hook continues to grow, giving rise to the “polyhook” phenotype.6,18 FliK appears to have two domains, the N-terminal domain (FliKN) and the C-terminal domain (FliKC), which are easily separated by limited proteolysis.8,17,19,20 FliKN is less stable and is more easily degraded by prolonged proteolysis than FliKC, which remains intact after prolonged proteolysis, suggesting that FliKC has a stable fold.11 Conservation of the hydrophobic residues suggests that FliKC should have a structure in common with its homologs and orthologs in the T3S specificity switching system (T3S4) domain.21 Mutant FliKs that have deletions in the T3S4 domain lose their ability to switch substrate specificity, suggesting that the T3S4 domain plays an essential role in the switch.9,19 In summary, FliK is structurally and functionally divid-

ed into two parts: FliKN (residues 1–203) and FliKC (residues 204–405).6,9 FliKN is directly involved in length control, while FliKC controls the hook length by switching the substrate specificity of the secretion apparatus. It should be noted that these two functions are not independent of one another. If FliKC is deleted, the hooks keep elongating into polyhooks even in the presence of intact FliKN. In contrast, as long as FliKC is intact and even if FliKN is deleted, the hooks elongate to produce polyhooks with filaments attached, the socalled polyhook-filament phenotype.5 Although there are several hypotheses for the role of FliK in determining hook length, none of them can fully explain all of the observatons.19,22–24 Biochemical experiments on FliK gave limited insight into the FliK structure, leading to inadequate hypotheses about its mechanism of action. We were motivated therefore to determine the FliK structure using NMR spectroscopy. In this study, we analyzed both the N- and C-terminal domains and compared their structural features with those of the full-length FliK. We confirmed earlier observations that FliKN is unstable and that FliKC has both an unstructured and a structured parts and a folded domain. We then solved the solution structure of the folded domain (core) of FliKC. Based on results from a series of experiments using fliK mutants, we modeled the structure of a complex between the FliKC core domain and FlhBC. We modeled the FlhBC structure based on the sequence homology with its homolog EscU.16 In our model, the cleaved loop in FlhBC exposes the site of interaction with the FliKC core domain; this site of interaction is sterically blocked in the non-cleaved form of FlhBC. Overall, the FliK structure obtained from the present NMR analyses sheds new light on the regulatory roles of FliK as a bifunctional player in the flagellar formation.

Results Our strategy In the alignment of amino acid sequences of FliK orthologs from various species, there are no conserved sequences in the N-terminal region (FliKN), whereas conserved sequences are found in the C-terminal region (FliKC) (Fig. 1a). Not surprisingly, genetic, biochemical, and molecular biological assays showed that FliKN is structurally disordered while FliKC has a compact globular domain.6,9,24 FliKN contains an unusual sequence of Gly and Pro, which are structure-breaking residues (Fig. 1a). We employed PONDR† to predict disordered regions, that is, those that have a score greater than 0.5 (Fig. 1b).26,27 Of the FliK structure, 61% is predicted to † http://www.pondr.com/2daypage.html

560

NMR Structure of the Flagellar Protein Flik

Fig. 1 (legend on next page)

NMR Structure of the Flagellar Protein Flik

561

Fig. 1. (a) Clustal W25 alignment of the sequences of FliK orthologs from eight bacteria and from YscP, an equivalent protein in the injectisome assembly. Gene ID and the name of the organism are indicated for each sequence. The locations of the fragments studied in this work, FliKN and FliKC, are marked with shaded bars on the alignment; the numbers marked correspond to the positions for the residues in FliK from S. Typhimurium. The location for the T3S4 domain is indicated by the dotted bar, whose location was slightly modified (from that original proposed) according to the determined structure and sequence conservation. The secondary elements determined by NMR are depicted on the alignment by the filled bars: two a-helices and four β-strands. (b) PONDR scores plotted against the residue number of FliK. PONDR scores indicate the propensities for folding; the regions having residues with lower PONDR score tend to be part of a folded structure, while the regions having a score over 0.5 are predicted to be disordered (indicated with bold black bars). (c) SDS gels for indicating the purity of each fragment used in this study.

be disordered. Failure of attempts to grow crystals of FliK is probably due to the disordered parts. Proteins with some disordered regions, however, can be studied by NMR spectroscopy,28 and PONDR predicts folded domains around residue 51 in FliKN and around residue 275 in FliKC. These regions are consistent with the two fragments obtained by limited proteolysis of FliK, namely, FliKN (1–147) and FliKC (204–320).9 Since FliK can be functionally and structurally divided into two, we separately determined the structures of the N-terminal and C-terminal domains by NMR.

Purification of FliK fragments of various sizes In preliminary experiments, we attempted to purify overproduced FliK, using various types of chromatography columns, but often observed that FliK is degraded probably by trace amounts of nonspecific proteases. Addition of protease inhibitors failed to completely inhibit cleavage; in a concentrated FliK solution used for NMR measurements, we always saw cleavage, which gradually increased over the 1 week needed for the measurements. Therefore, we tested various methods to

562 suppress protease activity during the purification of FliK and its derivatives. For purification of the full-length FliK, heat treatment was effective in stopping degradation during the 1 week of NMR data collection (Fig. 1c, left). The heat treatment did not cause any significant changes in the 1H–15N heteronuclear single quantum coherence (HSQC) spectrum compared to that of the untreated samples (see below), which suggests that FliK can readily recover from any denaturation caused by the heat treatment. For several reasons, heat treatment was not effective for the purification of FliK partial fragments. When expressed in E. coli, the His-tagged FliKN(1–147) fragment, which interacts with FlgE,9 was not detected by SDS-PAGE, suggesting that FliKN(1–147) is unstable and is degraded. However, the glutathione S-transferase (GST)-tagged FliKN(1–147) fragment was successfully expressed, purified, and used for NMR data collection without heat treatment (Fig. 1c, middle). The molecular size of GST itself is comparable to that of the FliKN(1–147) fragment. It was possible nevertheless to make assignments for the NMR peaks from FliKN(1–147) alone because FliKN(1–147) has a high mobility in solution and gives stronger NMR peaks than the less mobile GST. The His-tagged FliKC fragment (residues 200–405) was successfully overexpressed in E. coli. However, when left overnight in the presence of Factor Xa to remove His-tags, FliK fragments were degraded to a stable peptide. Amino acid sequencing and matrixassisted laser desorption/ionization–time of flight mass spectrometry revealed that the stable fragment was a peptide from residues 204 to 3709 (data not shown). Thus, we constructed and expressed a Histagged FliKC(204–370) fragment, purified it, and used it in further experiments (Fig. 1c, right). The purity of each FliK fragment was checked by SDS-PAGE prior to each experiment. The structure of FliK revealed by NMR To reveal the structure of FliK, we first examined the full-length FliK by 1H–15N HSQC. The spectra showed that approximately 70% peaks are in the limited region of the amide proton chemical shifts at 8.4 ppm, indicating that the corresponding residues are unstructured (Fig. 2a, middle). Then, we examined the GST-tagged FliKN(1–147) fragment and the FliKC(204–370) fragment in the same way. The 1H–15N HSQC spectrum for the GST-tagged FliKN(1–147) fragment also shows a limited chemical shift dispersion along the 1H dimension, suggesting that the corresponding residues are unstructured (Fig. 2a, left). Backbone resonance assignments of the fragment were not completed due to either peak overlap or peak broadening in the spectra. The broadened peaks suggest that parts of FliKN(1–147) are conformationally unstable with a time constant

NMR Structure of the Flagellar Protein Flik

in the microsecond–millisecond range. This is consistent with the observation that there is proteolytic cleavage but with a protease-resistant part in the N-terminal portion of FliK.9 In addition, the PONDR score of 0.2 around the residue 51 suggests a sequence that may transiently fold but is not stable enough to be protected from prolonged proteolysis. In contrast to the overlapping peaks in the spectrum for FliKN(1–147), the spectrum for FliKC(204–370) shows well-dispersed peaks, indicating that the fragment is stably folded (Fig. 2a, right). NMR signals from the FliKC(204–370) fragment were mostly assigned, although some of the side-chain signals coming from the C-terminal Q-rich regions remained unassigned due to the overlap of peaks. Structural characterization of FliK terminal fragments based on NMR spectra Since we employed different methods to purify FliK fragments of varying sizes, we wanted to know whether these fragments have the native fold. We compared spectra of the FliKN(1–147) fragment and of the FliKC(204–370) fragment with that from the fulllength FliK (Fig. 2b). Most of the peaks in the FliKN (1–147) spectrum acceptably overlap those in the spectrum for the full-length FliK, but some showed significant displacements (Fig. 2b, left). These displacements are due to small backbone differences between the fold of the fragment and that of the FliKN portion of the full-length FliK. We suggest that the changes are introduced in the refolding process after heat denaturation. Thus, FliKN(1–147) is not completely unstructured (see Discussion). In contrast to the N-terminal spectra, the peaks in the spectrum for FliKC(204–370) can be almost perfectly overlaid onto the corresponding peaks in the spectrum for the full-length FliK (Fig. 2b, left), indicating that FliKC(204–370) is more structurally stable than FliKN(1–147). In summary, FliK consists of two halves: the disordered FliKN and the folded FliKC. We examined the two terminal regions separately, focusing first on the FliKC(204–370) fragment because it is small enough to solve with ease and because the dispersed NMR peaks are technically easier to assign than the more heavily overlapped signals from the FliKN(1–147) fragment. Structural flexibility and secondary structure of FliKC The sequence-specific backbone assignment of the FliKC(204–370) fragment was obtained from tripleresonance experiments (see Materials and Methods). Each amino acid residue thus assigned is indicated in the 1H–15N HSQC spectrum (Fig. 2c). Some of the peaks showing significant deviations from the spectrum of the full-length FliK are Ala204,

NMR Structure of the Flagellar Protein Flik

Ala220, Asn311, Val324, and Arg370 residues. Ala204 is the first residue in the sequence and follows the two-residue sequence HM tacked onto our construct; thus, its deviation from the full-length FliK peak position is not unreasonable. For Arg370, the last residue in FliKC(204–370), the large deviation is also reasonable due to the adjacent residues missing in the fragment but not to the full-length FliK. The chemical shift deviations observed for Ala220, Asn311, and Val324 are not as easily explained. To gain insight into the structural flexibility of FliKC(204–370), we measured the steady-state 15N{1H} heteronuclear nuclear Overhauser enhancement (NOE) (hetNOE). The signals within the limited range of the 1H chemical shift were negative peaks when hetNOE was active (Fig. 3a). For each residue, we plotted the hetNOE value, that is, the peak intensity with 1H irradiation divided by that without it (Fig. 3b). The residues ranging from Pro253 to Ser355 showed positive values, suggesting that the corresponding segment has a stable fold. This segment overlaps the T3S4 domain that was pre-

563 dicted to be folded by comparison to its orthologs (Fig. 1a). Those sequences showing negative values of hetNOE are apparently unstructured. Two residues (Ala204 and Gln360) at both ends in this profile showed the values lower than − 3.9, the theoretical limit for 15N nuclear spin, which occurs in the absence of 1H radio frequency irradiation. Characteristic features of the FliKC core domain All of 162 amide proton signals in the 1H–15N HSQC spectra of FliKC(204–370) were assigned. We determined the tertiary structures of FliKC(204–370) on the basis of distances, dihedral angles, and residual dipolar coupling (RDC) restraints (Table 1). We selected and superposed 20 lowest-energy structures by minimizing the RMSD of Cα, C, O, and N backbone atoms of the α-helices. The superposition clearly shows a structured domain and the unstructured terminal tails (Fig. 4a). The compact domain is formed from residue Pro253 to Ser355, a stretch that corresponds to positive hetNOE factors.

Fig. 2. (a) 1H–15N HSQC spectra for the full-length FliK (FliKfull) (middle), GST-fused form of FliKN (left), and FliKC (right). (b) The spectral overlays for the full-length FliK and GST-fused form of FliKN (left) and also the full-length FliK and FliKC (right). (c) The backbone and side-chain resonance assignments are indicted in the spectrum for FliKC.

564

NMR Structure of the Flagellar Protein Flik

Fig. 2 (legend on previous page)

This stretch of the sequence is a little smaller than that of the T3S4 domain but larger than that predicted by PONDR. We call this newly visualized domain the FliKC core domain. As anticipated from the negative hetNOE factors, the other parts are unstructured; no long-range NOE factors that indicate a defined fold were found. The FliKC core domain consists of two α-helices and a β-sheet structure composed of one pair of antiparallel β-strand and one pair of parallel β-strand (Fig. 4b). Each structural element is connected by a loop: a short loop between α1 and β1 (loop 1), a long loop between β1 and β2 (loop 2), a long loop between β2 and β3 (loop 3), a short loop between β3 and α2 (loop 4), and a long loop between α2 and β4 (loop 5). Loop 2 contains the conserved sequence LHPEELG (see below). The close contact between α1 and α2 (see the next section) may be due to Pro333, which

causes a slight kink in α2. Loop 3 is also remarkable with its run of negatively charged residues, a feature found among most of its orthologs; one exception (found in Bacillus licheniformis) does not have any acidic residues in the loop (Fig. 1a). Loop 2 appears partially buried in the protein, while loop 3 is located in a rather open space. The remaining loops are exposed but may only be necessary for maintaining the whole structure. Both terminal regions (Pro253–Gly267 and Ser353–Ser355) are unstructured but remain close to the core structure (Fig. 4b). Roles of loop 2 in the substrate specificity switching By careful sequence analysis of the proteins of FliK orthologs and the injectisome equivalent proteins

565

NMR Structure of the Flagellar Protein Flik

Fig. 3. (a) The steady-state 15N{1H} NOE spectra collected with (right) and without (left) 1H irradiation for FliKC at 303 K on a 600-MHz spectrometer. The signals in red have negative intensities. (b) The NOE efficiency for each residue is plotted against the residue number.

Table 1. Summary of conformational restraints of the final 15 structures of the FliKC NOE upper distance restraints Total Intraresidue Sequential (|i − j| = 1) Medium range (1 b |i − j| b 5) Long range (|i − j| N = 5) Dihedral angle restraints (φ and ψ) Vicinal coupling restraints (3JNH, Ha) RDC restraints (1H–15N) PROCHECK Ramachandran plot statistics (%) Residues in most favored regions Residues in additional allowed regions Residues in generously allowed regions Residues in disallowed regions RMSD among the structures (Å) Average backbone RMSD from the mean (residues 60–150) Average backbone RMSD from the mean (residues 60–150)

1870 496 645 252 478 165 109 64 75.9 20.5 1.8 1.8 0.55 ± 0.12 1.00 ± 0.11

including YscP, we found that the most highly conserved sequence is LxPx1x2LG. In this motif, acidic residues are found at the x1 and/or x2 positions (Fig. 1a). In loop 2 of the FliKC core domain, the corresponding sequence is LHPEELG placing two negative charges in the site (as marked in red stick representation in Fig. 4b). Notably, Pro296 in this conserved sequence has a cis-peptide bond, which was suggested from its 13Cβ and 13Cγ chemical shifts29 and also confirmed by the sequential NOEs between the α-protons of His295 and Pro296.28 This unusual Pro residue may contribute to the structural stability of the core domain or may have a special role in triggering the switch (see Discussion). Genetic analysis of fliK in-frame deletion mutants gave a hint about the role of loop 2 in the substrate specificity switching19 (Fig. 5). The mutant lacking residues Thr248 to Gln278 produces short hooks with attached filaments, indicating that this peptide

566

NMR Structure of the Flagellar Protein Flik

Fig. 4. Solution structure of FliKC at pH 6.4, 303 K. (a) A bundle of 20 energy-minimized conformers were superimposed for minimal RMSD of the backbone N, Cα, and C′ atoms of the FliKC core domain, residues 255–355. An example of a conformer is shown in red. (b) Ribbon representation of the FliKC core domain having the minimal energy structure among the 20 conformers. Two views are depicted from the opposing angles. Two Glu residues on the exposed loop 2, or the LHPEELG loop, are depicted in red in the stick representation. The drawings were prepared by the programs MOLMOL52 and PyMOL for (a) and (b), respectively.

is dispensable for switching. The mutant lacking residues Thr248 to Gln288 produces a mixture of polyhooks (64%) and polyhook filaments (36%), indicating that the switching ability of the resultant domain is impaired and stochastic. This mutant was described as producing polyhooks in

the previous paper, but since we occasionally observed polyhook filaments as shown +/− in a table, we reexamined the strain. The mutant lacking residues Thr248 to Glu298 produces polyhooks with no filament attached, indicating that this peptide is indispensable for switching. In the

567

NMR Structure of the Flagellar Protein Flik

α1 β1 250 260 270 280 300 β2 310 AAMPTLSSAT AQPLPVASAP VLSAPLGSHE WQQTFSQQVM LFTRQGQQSA QLRLHPEELG QVHISLKLDD loop2

Short hook+filament Polyhook+filament Polyhook

TH10338 TH10337 TH10336

Polyhook (no filament)

Polyhookfilament

TH10338(Δ248-278)

0

TH10337(Δ248-288)

64

0 36

TH10336(Δ248-298)

99

1 (N=100)

Fig. 5. Deletion map in the FliKC core domain and consequent flagellar structures. Three fliK mutant strains (TH10336, TH10337, and TH10338) previously examined19 were reexamined in this study. Secondary structures are indicated by arrows above the FliK amino acid sequence, and the position of loop 2 is indicated under the sequence (top). Deleted segments in the FliK sequence of mutant strains are indicated by broken lines, and the flagellar structures observed by electron microscopy are shown above the lines (middle). The numbers of partial flagellar structures isolated from each mutant strain are shown in the table; 100 particles were counted for each strain (bottom). The TH10338 strain produced ordinary filaments, but the hooks were shorter than the wild type.19

tertiary structure of the protein, the peptide from Val279 to Gln288 corresponds to the N-terminal unstructured fragment followed by the first half of α1, while the peptide from Ser289 to Glu298 corresponds to the C-terminal half of α1 followed by loop 2 (Fig. 4b). As the deletion is extended toward loop 2, the switching ability becomes

Table 2. Residues possibly involved in the interaction between FliKC and FlhBC Flexible regions Residues Active D309, D310, R336, S342–S349 E341, G343 Passive L339, I344, L346 Active T268, E314, R302, P270–Y273 FlhBC R320, R324 Passive L313, P316 Status for the complex structure model of FliKC and FlhBC in the lowest energy van der Waals energy - 29.8 (kcal/mol) Electrostatic energy - 273.1 (kcal/mol) 2 766.6 Buried surface area (Å ) Protein FliKC

HADDOCK restrain type

The residues are picked up in consideration of the result identifying that the segment comprising the residues 301–350 is responsible for the export substrate specificity switch, which result comes from a systematic deletions to FliK,19 and the shape and electrostatic complementarities between the FliKC core domain and the FlhBC structures.

weaker and is eventually lost. Thus, we conclude that loop 2 is essential for triggering the switch of substrate specificity. Implication of the switching of FlhB by FliK FlhB forms the secretion gate and determines the substrate specificity.1 The structure of the C-terminal domain of FlhB (FlhBC) was obtained by structural homology modeling. The crystal structure of the cytoplasmic domain of YscU, a T3SS homolog of FlhB, was used as a template.14 Two conformations of the FlhBC structure, one having a cleaved NPTH loop and the other having an uncleaved NPTH loop, were modeled based on the YscU crystal structures. The surface charge distribution on the modeled FlhBC is shown in Fig. 6a. One notable structural feature is the basic patch formed by the αhelix and β-sheets adjacent to the NPTH loop. Autocleavage of the loop opens this region, providing access to the basic residue hidden behind the uncleaved loop; the result is an expansion of the basic patch. Using the modeled FlhBC and the NMR structure of FliKC, we manually docked these two proteins (Fig. 6b). In the docked structure, the FliKC loop 2 functions as a “wedge” inserted into the “basic cleft” that is formed by the α-helix and the β-sheets. Note that, in this docked structure, the FliK loop 2 would have interfered with the NPTH loop were the latter not cleaved. Thus, autocleavage of the NPTH loop

568

NMR Structure of the Flagellar Protein Flik

Fig. 6. (a) The homology modeled structure of FlhBC and the surface charge distribution in the intact NTPH loop form and its cleaved form. The homology modeling was done with the program MODELLER using UscU structures in the cleaved (3BZL) and non-cleaved (3BZP) forms as templates. The surface charge distributions of the cleaved form of FlhBC (right) and its non-cleaved form (left) are shown. (b) Our docked structure made by choosing plausible interacting residues on the FliKC core domain and on the FlhBC as listed in Table 2. The docked structure was calculated with the HADDOCK algorithm. The side view of the complex structure is shown. Structural parts colored pink and green represent FlhBCN and FlhBCC subdomains in the cleaved form of FlhBC. FliKC structure is depicted in yellow ribbon with van der Waals surface. The drawing was prepared using PyMOL.

seems essential for access of the FliK loop 2 to FlhB. The docked structure reasonably explains the requirement for the NPTH autocleavage in switching specificity, that is, it permits interaction between FliK and FlhB.

Discussion The switching of substrate specificity in the type III flagellar protein export is the second step in the regulation of flagellar assembly. In this step, the

569

NMR Structure of the Flagellar Protein Flik

export apparatus changes from exporting rod/hook proteins to exporting filament proteins. Hook assembly would have continued were the switch not made. FliK plays a major role in the switching by its interaction with autocleaved FlhB, the secretion gate for substrates. FliK has two distinct functions each associated with one of its two structural domains: FliKN and FliKC. In this study, we focused on FliKC only and reported the structure of its FliKC core domain as determined by solution NMR spectroscopy. Before a discussion of our findings on the FliKC core domain, we need to discuss the FliKN domain. FliKN is mostly unstructured but may have transiently folded parts We have shown that FliKN(1–147) is unstructured as judged by the limited signal dispersion in the 1 H–15N HSQC spectra, but the fragment may not be completely unstructured. Heat treatment of the full-length FliK, which was essential for purification of the protein, caused subtle but significant changes in the NMR spectra; these changes arise from the FliKN(1–147) fragment. Thus, the FliKN portion in the full-length FliK may be more completely folded. We did not pursue the complete structure of FliKN but plan to do so in the future. We conclude that FliKN is largely unfolded, except for a small proportion that might be folded albeit transiently. This NMRderived structural insight is consistent with the protease-resistant fragment we observed. Flexible or unstable structural parts in FliKC In striking contrast to FliKN(1–147), residues Ser258 to Gln359 of FliKC(204–370) form a compact domain consisting of two α-helices and four β-strands. The remaining residues at either end are unstructured as judged from negative hetNOE factors (Fig. 3c). Although unstructured, both terminal segments of the FliKC(204–370) fragment are functionally important. The N-terminal segment (Ala204 to Ala257) connects the compact domain to the unstructured portion of FliKN. Since protease digestion stops at Ala204, however, the segment may bind to the compact domain that becomes part of the core domain. The C-terminal tail segment starting at Gln360 is necessary for function as well; deletion of this tail abolishes switching.24 Although the segment is also unstructured, it is too resistant to protease digestion and may interact with the compact domain to form a larger domain (e.g., the T3S4 domain). We do not know how these segments work to effect switching. Which loop is important for switching? FliK function has been explored using various deletion mutants: spontaneous deletions, in-frame

deletions at random sites, and a series of successive 10-amino-acid deletions.6,9,19 Accumulated data from those mutants provided insight into the FliK structure. Now that we have solved the structure of the FliKC domain, let us review those observations. Minamino et al. constructed a series of 10-aminoacid deletion, several of which cover the FliKC core domain, making a large contribution in identifying the functional region.9 A 10-amino-acid deletion surely destroys the core domain structure and abolishes switching. Unfortunately, however, since we do not know the structure of the mutant protein, we cannot predict how function will be altered in detail. Shibata et al. constructed a series of deletions, with each one being longer by 10 residues.19 As described in Results, eventually, the ability to switch was lost. In the continuous deletion mutants exhibiting the gradual loss of function, we expect that some functioning structural elements would remain in some of the deletions. Interaction between FliK and FlhB Because the surface of FlhBC is rich in positive charges, it is reasonable to focus on the acidic surface patch in the FliKC core domain as a possible binding site for FlhBC. Loop 2 in the core domain has the negatively charged sequence of LHPEEL; therefore, it will lead off the interaction with FlhB as the N-terminal peptide of FliK enters the secretion channel. Clever experiments using YscP indicated that only one molecule is required to affect switching.30 The interaction between the core domain and FlhB in a limited space will insure switching, with a resulting conformational change in the secretion gate. The unusual Pro residue in loop 2 may play a role in stabilizing the loop structure and promoting the conformational change. As the FliK molecule enters the secretion channel, the core domain will unfold, making the remaining loops accessible to FlhB. We do not have any information about these loops at the moment. We also do not know if the cleaved FlhB segment might become detached after interaction with FliK. Recently, Morris et al . measured interaction between FliK and FlhB by optical biosensing methods.31 They found that the affinity between the two components is strong (order of micromolar) and that FliK binds to both wild-type (autocleaved) and mutant (non-cleaved) FlhBs with similar strength. We have not measured the affinity constant between FliKC and FlhB either cleaved or uncleaved by solution NMR. However, their results do not necessarily disagree with ours because, in our model, FliK is half extended as it is secreted, which is a different conformation from the isolated molecules used by Morris et al.31

570

Materials and Methods Strains, growth conditions, and chemicals used We used the fliK gene and its derivatives from S. Typhi SJW1103 throughout this study. Three fliK inframe deletion mutants (TH10336, TH10337, and TH10338) were previously described.19 E. coli DH5α was used for cDNA insert cloning host and E. coli BL21 (DE3) for protein expression host. E. coli cells on LB agar plates were inoculated and grown at 37 °C in an M9 minimal medium containing 15NH4Cl or 15NH4Cl and [13C]glucose for protein isotope labeling. Plasmids were maintained in E. coli with appropriate antibiotics for selection (100 mg/l of ampicillin). Plasmids and genomic extractions, restriction enzyme digestions, DNA ligations, and transformations into E. coli were carried out, using enzymes provided by Promega or New England Biolabs. Purification of FliK Construction of overexpression systems for FliK fragments of various sizes We constructed three types of FliK fragments: the fulllength FliK (405 residues), the FliK N-terminal domain (FliKN) (residues 1–147), and the FliK C-terminal domain (FliKC) (residues 204–370) (Fig. 1). We cloned the fulllength fliK gene into the pET22b vector (Novagen), between NdeI and BamHI restriction sites. The DNA fragment encoding FliKN was cloned into pGEX-4T-2 (GE Healthcare) at the BamHI and EcoRI, and those encoding FliKC (residues 200–405 or residues 204–370) were cloned into pCold I (Takara Bio) at the NdeI and BamHI sites. All plasmids were introduced into E. coli strain BL21(DE3) (Stratagene). Protein overexpression E. coli cells were grown in M9 minimal medium containing 15NH4Cl or 15NH4Cl and [13C]glucose for protein isotope labeling and ampicillin (50 μg/ml) at 37 °C to an OD600 of 0.6. IPTG was added to a final concentration of 0.5 mM, and the cells were incubated at 37 °C for 5 h to activate expression vectors. For overexpression of each of the FliK terminal halves, the culture temperature was lowered to 15 °C after addition of IPTG, and cells were incubated for 15 h. At the late log phase, the cells were harvested by centrifugation (4000g for 15 min). The cell pellet was suspended in 50 ml buffer solution [50 mM Tris–HCl (pH 8.0)] and was lysed by sonication on ice (amplitude, 17%, 45 min) (TAITEC), and the cell debris were removed by low-speed centrifugation (26,000g for 30 min) or high-speed centrifugation (52,000g for 30 min). The supernatants, denoted Sup A, were collected and used for further experiments.

NMR Structure of the Flagellar Protein Flik

added to the supernatant at 4 °C; the solution was gently stirred overnight and then centrifuged to recover the pellet by low-speed centrifugation. The pellet was suspended in the dialysis buffer [20 mM Tris–HCl (pH 8.0) and 1 mM ethylenediaminetetraacetic acid (EDTA)] and was dialyzed overnight to remove ammonium sulfate. After removing nonspecific aggregations by centrifugation (13,000g for 5 min), the supernatant was applied to a HiTrap HP anion-exchange column (GE Healthcare). The column was equilibrated with 10 column volumes of the cleaning buffer [20 mM Tris–HCl (pH 8.0) and 1 mM EDTA]. A linear gradient of the elution buffer (cleaning buffer containing 1 M NaCl) was used to elute proteins. Each fraction was analyzed by SDS gel electrophoresis, and fractions containing FliK were pooled and dialyzed against NMR buffer [50 mM sodium phosphate (pH 6.4) and 1 mM EDTA]. Purification of the FliK N-terminal domain The Sup A containing GST-fused FliKN was directly applied to a Glutathione Sepharose column (GE Healthcare). After washing the column with 20 column volumes of the cleaning buffer [50 mM Tris–HCl (pH 8.0)], proteins were eluted with 10 ml of the elution buffer (cleaning buffer containing 10 mM glutathione reductase). Fractions containing FliKN were pooled (10 ml) and dialyzed against the NMR buffer. The FliKN fragment was linked to the C-terminal end of GST protein by a six-residue sequence, LVPRGS. The FliKN–GST fusion was used in the NMR experiments because removal of the GST by cleavage caused unwanted degradation of the N-terminal fragment of FliKN. The GST protein was purified in the same way and was used as a control. Purification of the FliK C-terminal domain The Sup A was applied to a HisTrap HP column (5 ml volume) (GE Healthcare). After washing the column with 10 column volumes of the cleaning buffer [50 mM Tris–HCl (pH 8.0) and 20 mM imidazole), proteins were eluted with 10 ml of the elution buffer [50 mM Tris–HCl and 500 mM imidazole (pH 8.0)]. Fractions containing FliKC were pooled (10 ml) and dialyzed against buffer solution [20 mM Tris–HCl (pH 8.0) and 1 mM EDTA]. The N-terminal His-tag was removed by Factor Xa treatment at 4 °C for overnight. Then, the sample was again applied to an anion exchanger, HiTrap HP column (GE healthcare) for further purification. After washing the column with 10 column volumes of the cleaning buffer [20 mM Tris–HCl (pH 8.0) and 1 mM EDTA], the elution buffer (cleaning buffer containing 1 M NaCl) was used to elute proteins. Fractions containing FliKC were pooled and dialyzed against the NMR buffer. All samples for NMR measurements were concentrated using Amicon Centriprep (10 kDa cutoff; Millipore). SDS-PAGE

Purification of the full-length FliK The Sup A was heated at 100 °C for 20 min, and the precipitate was removed by low-speed centrifugation. Next, ammonium sulfate [50% (w/v)] was gradually

Laemmli SDS-PAGE was used to analyze sample and to estimate the purity of the product. The gel used is Tris–HCl and 12.0% or 14.7% precast polyacrylamide gel.

571

NMR Structure of the Flagellar Protein Flik

NMR spectroscopy All NMR samples were dissolved in 50 mM sodium phosphate buffer (pH 6.4), containing 1 mM EDTA. NMR experiments were performed at 303 K (30 °C) inner sample temperature calibrated using ethylene glycol. Data collection was done on a Bruker DMX600 or an Avance 600 equipped with a cryogenic TXI probe, both of which are operating at 600 MHz for 1H resonance frequency. For the resonance assignment, a conventional set of triple-resonance experiments was used,32 although all of the data were collected in a nonuniform sampling manner.33 The resonance assignments have been deposited to the Biological Magnetic Resonance Data Bank [Protein Data Bank (PDB) code: 2RRL]. A set of 1H–1H distance retrains was derived from 15N-edited NOE spectroscopy (mixing time, 100 ms) and 13C-edited NOE spectroscopy (mixing time, 100 ms) spectra. All NMR data were processed by the program NMRPipe (a multidimensional spectral processing system based on UNIX pipes)34 in combination with NMR toolkit (Hoch), and spectral analyses were carried out with KUJIRA utilities35 on the platform of NMR View.36 Steady-state 15N{1H} hetNOE measurements37 were carried out on a DMX600 at 303 K. 1H saturation was done with a train of 270° flip pulses for the duration of 2 s with additional relaxation delay of 3 s. For the reference spectra collected without 1H saturation, a 5-s relaxation delay was applied. The vicinal 3JHNHα couplings were determined by a series of two-dimensional 1H–15N correlation spectra with signals showing 3J-dependent amplitude modulation.38 The RDCs were measured using anisotropically stretched acrylamide gel39; 5% acrylamide gel concentration with 75:1 molar ratio for acrylamide and bis-acrylamide with 3% doping with DADMA was used to make the gel positively charged. For collecting the reference data, the acrylamide gel made with the same composition with a smaller diameter was used; the reference gel was not stretched in an NMR sample tube so that the positions of the protein bands were not changed. Protein concentration in the gel was about 0.5 mM. The IPAP (in-phase/anti-phase) HSQC spectra collected in an interleaved manner were used to measure the RDCs.40,41 Structure determination Automatic NOE assignments and structure calculations in a torsion angle dynamics mode were performed with the program CYANA.42,43 TALOS restraints44 to backbone dihedral angles were included in the calculation. The CYANA calculation finally gave 20 conformers with lowest values for the target function, which were further minimized using explicit solvent with NOE-derived restraints, experimentally derived and TALOS backbone dihedral restrains with RDC restraints using the program Xplor-NIH.45 The resultant structures were validated using PROCHECK-NMR.46 Homology modeling of FlhB structure The C-terminal domain structure of FlhB, FlhBC, was modeled from the X-ray structures of the corresponding part of EscU in cleaved (PDB code: 3BZL) and non-cleaved forms (PDB code: 3BZP); the non-cleaved EscU includes

the mutated APTH loop instead of the intact NPTH in the autocleavage site.16 The program MODELLER 9v8 was used in the modeling.47 In the case of the cleaved form of FlhB, the short fragment from residue 253 to residue 269 was not properly modeled because the fragment contains only a small number of residues. The corresponding part was incorporated from the model built as a non-cleaved form. The corresponding parts in the crystal structures of the cleaved and non-cleaved forms of EscU showed few structural differences; therefore, the modeling of the cleaved form of FlhB seems reasonable. To release the short contacts among the atoms in the modeled structures, we applied energy minimization using the suite in the program Chimera.48 Surface charge calculation The surface electrostatic potentials were calculated with the program APBS (Adaptive Poisson–Boltzmann Solver) Tools249 and were displayed in the range −5.0 kT/e to + 5.0 kT/e using the program PyMOL (DeLano Scientific, San Carlos, CA). The structure presentation with solventaccessible surface colored according to the surface charge in 30% transparency was also generated by PyMOL. Modeling the complex structure between FliKC and FlhBC In building the docked structure of FliKC into FlhBC, the plausible restraints were collected from the manually docked structure with considering the functionally responsible regions identified by a series of FliK mutants, the surface charge distributions, and the shape complementarities between the structures; for FlhBC, the energyminimized model structure in the cleaved form was used. The manually collected restraints were used as inputs for the HADDOCK calculation50 to build the model complex structure; the active and passive restraints are listed in Table 2. The docking modeling calculation was done on the HADDOCK web server.51 Accession numbers Coordinates and structure factors have been deposited in the PDB with accession number 2RRL. NMR assignment data have been deposited in the Biological Magnetic Resonance Data Bank with accession number RCSB11423.

Acknowledgements This work was supported by a Grant-in-Aid for Scientific Research on Innovative Areas “Molecular Science of Fluctuations toward Biological Functions” (area number 2006) from the Japan Society for the Promotion of Science. Part of the work was also supported by a Grant-in-Aid for Scientific Research on Priority Areas “Molecular Science for Supra

572 Functional Systems–Development of Advanced Methods for Exploring Elementary Process” (area number 477) from the Japan Society for the Promotion of Science. S.T. also acknowledges the financial support from a Grant-in-Aid for Creative Scientific Research Program (18GS0316). The authors appreciate RIKEN NMR facility (Yokohama, Japan) for providing the instrument times for collecting a part of NMR data used in this research. We thank K. Uchida for technical help and M. Hashimoto, J. Uewaki, and S. Nakano for their useful discussion and help in preparing the manuscript. Finally, the authors are grateful to David DeRosier for his kind help in revising the manuscript.

References 1. Macnab, R. M. (2003). How bacteria assemble flagella. Annu. Rev. Microbiol. 57, 77–100. 2. Chevance, F. F. V. & Hughes, K. T. (2008). Coordinating assembly of a bacterial macromolecular machine. Nat. Rev. Microbiol. 6, 455–465. 3. Macnab, R. M. (1999). The bacterial flagellum: reversible rotary propellor and type III export apparatus. J. Bacteriol. 181, 7149–7153. 4. Aizawa, S. I. (2009). Flagella, P.393-403. in Encyclopedia of Microbiology. (Moselio Schaechter). Elsevier, Oxford, United Kingdom. 5. Hirano, T., Yamaguchi, S., Oosawa, K. & Aizawa, S. I. (1994). Roles of FliK and FlhB in determination of flagellar hook length in Salmonella typhimurium. J. Bacteriol. 176, 5439–5449. 6. Williams, A. W., Yamaguchi, S., Togashi, F., Aizawa, S. I., Kawagishi, I. & Macnab, R. M. (1996). Mutations in fliK and flhB affecting flagellar hook and filament assembly in Salmonella typhimurium. J. Bacteriol, 178, 2960–2970. 7. Macnab, R. M. (2004). Type III flagellar protein export and flagellar assembly. Biochim. Biophys. Acta, 1694, 207–217. 8. Minamino, T., Imada, K. & Namba, K. (2008). Mechanisms of type III protein export for bacterial flagellar assembly. Mol. Biosyst. 4, 1105–1115. 9. Minamino, T., Ferris, H. U., Moriya, N., Kihara, M. & Namba, K. (2006). Two parts of the T3S4 domain of the hook-length control protein FliK are essential for the substrate specificity switching of the flagellar type III export apparatus. J. Mol. Biol, 362, 1148–1158. 10. Ferris, H. U. & Minamino, T. (2006). Flipping the switch: bringing order to flagellar assembly. Trends Microbiol. 14, 519–526. 11. Minamino, T., Saijo-Hamano, Y., Furukawa, Y., Gonzalez-Pedrajo, B., Macnab, R. M. & Namba, K. (2004). Domain organization and function of Salmonella FliK, a flagellar hook-length control protein. J. Mol. Biol, 341, 491–502. 12. Ferris, H. U., Furukawa, Y., Minamino, T., Kroetz, M. B., Kihara, M., Namba, K. & Macnab, R. M. (2005). FlhB regulates ordered export of flagellar components via autocleavage mechanism. J. Biol. Chem. 280, 41236–41242.

NMR Structure of the Flagellar Protein Flik

13. Fraser, G. M., Hirano, T., Ferris, H. U., Devgan, L. L., Kihara, M. & Macnab, R. M. (2003). Substrate specificity of type III flagellar protein export in Salmonella is controlled by subdomain interactions in FlhB. Mol. Microbiol. 48, 1043–1057. 14. Lountos, G. T., Austin, B. P., Nallamsetty, S. & Waugh, D. S. (2009). Atomic resolution structure of the cytoplasmic domain of Yersinia pestis YscU, a regulatory switch involved in type III secretion. Protein Sci. 18, 467–474. 15. Deane, J. E., Abrusci, P., Johnson, S. & Lea, S. M. (2010). Timing is everything: the regulation of type III secretion. Cell. Mol. Life Sci. 67, 1065–1075. 16. Zarivach, R., Deng, W., Vuckovic, M., Felise, H. B., Nguyen, H. V., Miller, S. I. et al. (2008). Structural analysis of the essential self-cleaving type III secretion proteins EscU and SpaS. Nature, 453, 124–127. 17. Minamino, T., Gonzalez-Pedrajo, B., Yamaguchi, K., Aizawa, S. I. & Macnab, R. M. (1999). FliK, the protein responsible for flagellar hook length control in Salmonella, is exported during hook assembly. Mol. Microbiol. 34, 295–304. 18. Patterson-Delafield, J., Martinez, R. J., Stocker, B. A. D. & Yamaguchi, S. (1973). A new fla gene in Salmonella typhimurium -flaR- and its mutant phenotype superhooks. Arch. Microbiol, 90, 107–120. 19. Shibata, S., Takahashi, N., Chevance, F. F., Karlinsey, J. E., Hughes, K. T. & Aizawa, S. (2007). FliK regulates flagellar hook length as an internal ruler. Mol. Microbiol. 64, 1404–1415. 20. Moriya, N., Minamino, T., Hughes, K. T., Macnab, R. M. & Namba, K. (2006). The type III flagellar export specificity switch is dependent on FliK ruler and a molecular clock. J. Mol. Biol. 359, 466–477. 21. Agrain, C., Callebaut, I., Journet, L., Sorg, I., Paroz, C., Mota, L. J. & Cornelis, G. R. (2005). Characterization of a type III secretion substrate specificity switch (T3S4) domain in YscP from Yersinia enterocolitica. Mol. Microbiol. 56, 54–67. 22. Journet, L., Agrain, C., Broz, P. & Cornelis, G. R. (2003). The needle length of bacterial injectisomes is determined by a molecular ruler. Science, 302, 1757–1760. 23. Waters, R. C., O'Toole, P. W. & Ryan, K. A. (2007). The FliK protein and flagellar hook-length control. Protein Sci. 16, 769–780. 24. Hirano, T., Shibata, S., Ohnishi, K., Tani, T. & Aizawa, S. I. (2005). N-terminal signal region of FliK is dispensable for length control of the flagellar hook. Mol. Microbiol. 56, 346–360. 25. Larkin, M. A., Blackshields, G., Brown, N. P., Chenna, R., McGettigan, P. A., McWilliam, H. et al. (2007). ClustalWand Clustal X version 2.0. Bioinformatics, 23, 2947–2948. 26. Li, X., Romero, P., Rani, M., Dunker, A. K. & Obradovic, Z. (1999). Predicting protein disorder for N-, C-, and internal regions. Genome Inform. Ser. Workshop Genome Inform. 10, 30–40. 27. Romero, P., Obradovic, Z., Li, X., Garner, E. C., Brown, C. J. & Dunker, A. K. (2001). Sequence complexity of disordered protein. Proteins, 42, 38–48. 28. Wüthrich, K. (1986). pp. 117–129, John Wiley & Sons, Ltd, New York, NY. 29. Schubert, M., Labudde, D., Oschkinat, H. & Schmieder, P. (2002). A software tool for the prediction of Xaa–Pro

NMR Structure of the Flagellar Protein Flik

30.

31.

32.

33.

34.

35.

36. 37.

38.

39.

peptide bond conformations in proteins based on 13C chemical shift statistics. J. Biomol. NMR, 24, 149–154. Wagnera, S., Stentab, M., Metzgera, L. C., Perarob, M. D. & Cornelis, G. R. (2010). Length control of the injectisome needle requires only one molecule of Yop secretion protein P (YscP). Proc. Natl Acad. Sci. USA, 107, 13860–13865. Morris, D. P., Roush, E. D., Thompson, J. W., Moseley, M. A., Murphy, J. W. & McMurry, J. L. (2010). Kinetic characterization of Salmonella FliK–FlhB interactions demonstrates complexity of the type III secretion substrate-specificity switch. Biochemistry, 49, 6386–6393. Cavanagh, J., Fairbrother, W. J., Palmer, A. G., III & Skelton, N. J. (1996). Heteronuclear NMR Experiments. Protein NMR Spectroscopy. Academic Press, Inc., New York, NY. Rovnyak, D., Frueh, D. P., Sastry, M., Sun, Z. Y. J., Stern, A. S., Hoch, J. C. & Wagner, G. (2004). Accelerated acquisition of high resolution tripleresonance spectra using non-uniform sampling and maximum entropy reconstruction. J. Magn. Reson. 170, 15–21. Delaglio, F., Grzesiek, S., Vuister, G. W., Zhu, G., Pfeifer, J. & Bax, A. (1995). NMRPipe: a multidimensional spectral processing system based on UNIX pipes. J. Biomol. NMR, 6, 277–293. Kobayashi, N., Iwahara, J., Koshiba, S., Tomizawa, T., Tochio, N., Güntert, P. et al. (2007). KUJIRA, a package of integrated modules for systematic and interactive analysis of NMR data directed to high-throughput NMR structure studies. J. Biomol. NMR, 39, 31–52. Johnson, B. A. & Blevins, R. A. (1994). NMR View: a computer program for the visualization and analysis of NMR data. J. Biomol. NMR, 4, 603–614. Farrow, N. A., Muhandiram, R., Singer, A. U., Pascal, S. M., Kay, C. M., Gish, G. et al. (1994). Backbone dynamics of a free and phosphopeptide-complexed Src homology 2 domain studied by 15 N NMR relaxation. Biochemistry, 33, 5984–6003. Kuboniwa, H., Grzesiek, S., Delaglio, F. & Bax, A. (1994). Measurement of HN-Hα J couplings in calciumfree calmodulin using new 2D and 3D water-flip-back methods. J. Biomol. NMR, 4, 871–878. Sass, H. J., Musco, G., Stahl, S. J., Wingfield, P. T. & Grzesiek, S. (2000). Solution NMR of proteins within polyacrylamide gels: diffusional properties and residual alignment by mechanical stress or embedding of oriented purple membranes. J. Biomol. NMR, 18, 303–309.

573 40. Ishii, Y., Markus, M. A. & Tycko, R. (2001). Controlling residual dipolar couplings in high-resolution NMR of proteins by strain induced alignment in a gel. J. Biomol. NMR, 21, 141–151. 41. Ottiger, M., Delaglio, F. & Bax, A. (1998). Measurement of J and dipolar couplings from simplified twodimensional NMR spectra. J. Magn. Reson. 131, 373–378. 42. Herrmann, T., Güntert, P. & Wüthrich, K. (2002). Protein NMR structure determination with automated NOE assignment using the new software CANDID and the torsion angle dynamics algorithm DYANA. J. Mol. Biol. 319, 209–227. 43. Güntert, P. (2009). Automated structure determination from NMR spectra. Eur. Biophys. J. 38, 129–143. 44. Cornilescu, G., Delaglio, F. & Bax, A. (1999). Protein backbone angle restraints from searching a database for chemical shift and sequence homology. J. Biomol. NMR, 13, 289–302. 45. Schwieters, C. D., Kuszewski, J. J., Tjandra, N. & Marius Clore, G. (2003). The Xplor-NIH NMR molecular structure determination package. J. Magn. Reson. 160, 65–73. 46. Laskowski, R. A., Rullmannn, J. A., MacArthur, M. W., Kaptein, R. & Thornton, J. M. (1996). AQUA and PROCHECK-NMR: programs for checking the quality of protein structures solved by NMR. J. Biomol. NMR, 8, 477–486. 47. Sanchez, R. & Sali, A. (2000). Comparative protein structure modeling. Introduction and practical examples with MODELLER. Methods Mol. Biol. 143, 97–129. 48. Pettersen, E. F., Goddard, T. D., Huang, C. C., Couch, G. S., Greenblatt, D. M., Meng, E. C. & Ferrin, T. E. (2004). UCSF Chimera—a visualization system for exploratory research and analysis.. J. Comput. Chem. 25, 1605–1612. 49. Baker, N. A., Sept, D., Joseph, S., Holst, M. J. & McCammon, J. A. (2001). Electrostatics of nanosystems: application to microtubules and the ribosome. Proc. Natl Acad. Sci. USA, 98, 10037–10041. 50. Dominguez, C., Boelens, R. & Bonvin, A. M. (2003). HADDOCK: a protein–protein docking approach based on biochemical or biophysical information. J. Am. Chem. Soc. 125, 1731–1737. 51. de Vries, S. J., van Dijk, M. & Bonvin, A. M. (2010). The HADDOCK web server for data-driven biomolecular docking. Nat. Protoc. 5, 883–897. 52. Koradi, R., Billeter, M. & Wüthrich, K. (1996). MOLMOL: a program for display and analysis of macromolecular structures. J. Mol. Graphics, 14, 51–55.