X-ray Analysis of Biological Macromolecules. Dissertation

Transcription

1 X-ray Analysis of Biological Macromolecules Dissertation zur Erlangung des akademischen Grades des Doktors der Naturwissenschaften (Dr. rer. nat.) an der Universität Konstanz Lehrstuhl für Biophysik Mathematisch-Naturwissenschaftliche Sektion vorgelegt von Dipl. Biol. Patrick Polzer Tag der mündlichen Prüfung: Referent: Prof. Dr. Wolfram Welte Referent: Prof. Dr.Hans-Jürgen Apell Konstanz, Juni 2007

2 Contents List of Tables 5 List of Figures 6 1 Zusammenfassung 9 2 Abstract 12 3 TonB from Escherichia coli 15 4 Dimerization of TonB is not essential for its binding to FhuA Abstract Introduction Experimental procedures Construction of plasmids encoding TonB proteins Bacterial strains, plasmids, and growth conditions Purification of FhuA Purification of the C-terminal TonB fragments Purification of the FhuA405.H 6 /TonB complexes Crystallization, data collection, and structure solution Analytical ultracentrifugation Tryptophan fluorescence of the C-terminal TonB fragments Assay of bacteriophage susceptibility Assay of siderophore-dependent growth and iron transport Results Analysis of FhuA/TonB interaction in vitro

3 Contents Analytical ultracentrifugation of the C-terminal TonB fragments Tryptophan fluorescence of the C-terminal TonB fragments TonB fragments shorter than 96 amino acids inhibit TonB function in vivo very weakly Crystal structure of the TonB77 fragment Discussion Crystallization and preliminary x-ray analysis of TonB Abstract Introduction Materials and methods Expression and purification Crystallization and data collection Results and discussion Crystal structure of TonB Abstract Introduction Experimental procedures Protein expression and purification Crystallization and data collection Structure determination Dynamic light scattering Analytical ultracentrifugation Results Description of experimental structure Comparison with crystal structures of TonB77 and TonB Oligomerization of TonB92 in solution, and complex formation with FhuA Discussion

4 Contents Why does the presence of the additional N-terminal residues in TonB92 as compared with TonB85 change a dimer to a monomer in solution and cause such a significant conformational difference in the crystal structure? Is there a role for the two conformations in the transport process? Acknowledgments C-ring from Ilyobacter tartaricus 73 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus Abstract Introduction Structure of the c-ring Structure of the sodium ion binding site Ion translocation in F 0 complexes Supporting Online Material Materials and Methods GalU from Escherichia coli Concluding remarks and perspectives 95 Bibliography 98 Acknowledgements 108 List of Publications 110 Selbstständigkeitserklärung 111 4

5 List of Tables 4.1 Oligodeoxynucleotides used in creation of pbadtonb and ptb recombinant clones Strains of E. coli K-12 and plasmids used Data from the sedimentation velocity and sedimentation equilibrium experiments done with the C-terminal TonB fragments 76, 86, 96, and Summary of results for the TonB fragments Growth of E. coli AB2847 ara transformants on NB medium containing ampicillin (100 µg/ml) and dipyridyl (250 µm) Susceptibility of E. coli AB2847 ara transformants to phage Φ80λi Data collection and refinement statistics for the TonB77 homodimer Crystal data and x-ray data-collection statistics for a native TonB92 crystal Results of molecular replacement Data collection and refinement statistics of TonB Summary of data collection and refinement statistics

6 List of Figures 3.1 Select siderophore-mediated iron acquisition systems of E. coli Amino acid sequence of the C-terminal TonB fragments used in our studies Purification of FhuA405.H 6 and of the C-terminal TonB fragments Size exclusion chromatography of FhuA405.H 6 FC TonB protein complexes Three single TonB77 crystals grown in 2 M sodium formate and 0.1 M sodium citrate, ph Stereo ribbon diagram of the C-terminal fragment TonB77, showing the intertwined dimer Topological diagrams derived from the structures of C-terminal TonB and TolA fragments A native TonB92 crystal of dimensions mm grown in space group P Putative topology model of TonB Three-dimensional structure of the dimeric TonB92 in ribbon representation Topological diagram of TonB92 showing secondary structure elements derived from the crystal structure Ball-and-stick representation of the amino acid residues of TonB- 92, which are involved in stabilizing the extended N-terminal ES1 segment by formation of hydrogen bonds

7 List of Figures 6.4 Electron density map (2mF o - DF c ) at 3σ (dark blue) and at 2σ (light blue) around amino acid residue Gln 160 of ES Superposition of the three-dimensional structures of one molecule of TonB92 (in red) with one molecule of the tight dimer (i.e. the structure of TonB77 or TonB85) (in blue) C-terminal amino acid sequence of TonB from E. coli Superposition of one cluster of aromatic residues, Phe 180, Trp 213, and Tyr 215, of the TonB structures Superposition of the three-dimensional structures of one molecule of TonB92 (red) and the C-terminal domain of TolA from E. coli (blue) (PDB accession code: 1Tol) Model of the F 0 F 1 -ATP-synthase Structure of the I. tartaricus c 11 ring in ribbon representation Section of the c-ring showing the interface between the N-terminal and two C-terminal helices with those side chains discussed in the text Electron density map (red, Na + omit map at 3.0σ; blue, 2F obs - F calc map at 1.4σ) and residues of the Na + binding site formed by two c subunits, A and B Alignment of selected c subunit sequences Motional flexibility within the c-ring Van der Waals surface of the c-ring Ribbon model of the E. coli c 10 oligomer obtained by homology modeling according to the I. tartaricus c 11 structure Schematic model for the interconversion of the binding site in the subunit a/c interface from an alternately locked conformation to an open one Ribbon diagram of the search model G1P-TT Ribbon diagram of the GalU tetramer Superposition of GalU and G1P-TT Superposition of the active sites of GalU and G1P-TT

8 List of Figures 9.5 Superposition of the B-site of G1P-TT with one monomer of GalU

9 1 Zusammenfassung Thermodynamisch gesehen basiert Leben auf der energieaufwendigen Erzeugung von Ordnung entgegen dem allgemeinen entropischen Streben nach Gleichverteilung. Die Auswirkungen des zweiten Hauptsatzes der Thermodynamik auf einen Organismus haben gleich mehrere Konsequenzen für dessen Aufbau. Zum einen benötigen sie freie Energie zur Erzeugung ihrer höheren Ordnung und zum anderen kann diese Ordnung nur durch Kompartimentierung aufrecht erhalten werden, das heißt, Organismen brauchen eine Grenzschicht zwischen sich und ihrer Umgebung. Ein weiterer wichtiger Punkt ist Selektivität. Es muß ein gerichteter, kontrollierbarer Transport bestimmter Moleküle durch die oben erwähnte Grenzschicht stattfinden. Darüber hinaus müssen auch Signale über die Grenzschicht weitergeleitet werden, was eine Grundvoraussetzung für die Interaktion des Organismus mit seiner Umgebung ist. Die Zellmembran stellt diese Grenzschicht dar und die eben geforderten Aufgaben werden von speziellen Proteinen erfüllt. Dieser Sachverhalt macht die Zellmembran und die darin vorhandenen, beziehungsweise an sie assoziierten, Proteine zu einem der wichtigsten Untersuchungsgebiete der Biologie. Zum besseren Verständnis der Funktionsweise von Proteinen ist insbesondere deren Struktur von enormer Bedeutung. Ist sie bekannt, dann lassen sich Substratbindungsstellen finden und es ergeben sich Hinweise auf die der Funktion zugrundeliegenden katalytischen Mechanismen. Die Pharmaindustrie nutzt dieses Wissen zum Beispiel, um Medikamente zu entwickeln, die gezielt an bestimmte Stellen am Zielprotein binden können. Während die Primärstruktur der Proteine, also die Reihenfolge der Aminosäuren, relativ leicht zu bestimmen ist und auch die Sekundärstruktur, also die Abfolge von α- Helices und β-faltblättern, recht zuverlässig vorhergesagt werden kann, stellt die Auflösung der dreidimensionalen Anordnung der Atome eines Proteins für die Wissenschaft immer noch eine Herausforderung dar. 9

10 1 Zusammenfassung Der Großteil der heute bekannten Proteinstrukturen wurde mittels Röntgenstrukturanalyse gelöst (80.2% bis Ende ). Bei dieser Methode, die erstmals erfolgreich von Perutz [1] und Kendrew [2] zur Bestimmung der Hämoglobin-, beziehungsweise Myoglobinstruktur eingesetzt wurde, macht man sich die Kristallisierbarkeit von Proteinen zunutze. Die Proteinkristalle werden mit einem hochenergetischen, monochromatischen Röntgenstrahl bestrahlt und erzeugen ein Beugungsbild mit einer Vielzahl von Reflexen. Zur Strukturlösung benötigt man zum einen die so gemessenen Reflex-Intensitäten (und -Positionen) und zum anderen die Phasen der Reflexe, welche jedoch experimentell nicht bestimmt werden können. Dieser Umstand ist als das Phasenproblem der Röntgenstrukturanalyse bekannt. Es gewinnt zusätzlich an Gewicht, wenn man bedenkt, daß die Phasen mehr Information über die Struktur in sich tragen als die Intensitäten, was man sehr eindrucksvoll an Kevin Cowtans Katzen und Enten in seinem Bilderbuch der Fourier-Transformationen (Picture Book of Fourier transforms) sehen kann 2. Die Lösung dieses Problems stellt, nach dem Kristallisieren der Proteine, die letzte zu überwindende Hürde dar. Daher ist es nicht verwunderlich, daß es dafür eine Vielzahl von Lösungsansätzen gibt und es ist vorab schwer zu sagen, welcher Ansatz die größten Erfolgsaussichten liefert. Im Grunde kann man die Methoden der Phasenbestimmung in drei Bereiche einteilen. Zunächst einmal existieren sogenannte direkte Methoden, bei denen versucht wird, die Phasen unmittelbar aus den Reflexintensitäten zu berechnen. Dafür ist jedoch eine sehr hohe Auflösung (ca. 1.0 Å) erforderlich, die mit Proteinkristallen nur sehr selten erreicht werden kann. Die zweite Methode basiert auf dem Schweratom-Ersatz. Hierbei wird versucht, Schweratome an bestimmte Stellen in jeder Elementarzelle eines Proteinkristalls zu binden (oder bereits vorhandene Schweratome zu nutzen, wie zum Beispiel bei Metallo-Proteinen) und dann mittels verschiedener solcher Derivate, oder durch Messung an der Absorptionskante solcher Schweratome, Unterschiede in den Reflexintensitäten zu bestimmen. Aus diesen kann dann die Substruktur der Schweratome, also deren Positionen in der Elementarzelle, berechnet werden [3, 4] und daraus dann, mittels Harkerkonstruktion [5], die 1 Protein Data Bank: http : // 2 http : // cowtan/fourier/fourier.html 10

11 1 Zusammenfassung Phasen der Reflexe. Die letzte Methode schließlich, der sogenannte molekulare Ersatz, wird in Zukunft immer größere Bedeutung erlangen. Dabei wird versucht, die gemessenen Reflexintensitäten mit Phasen einer ähnlichen, bereits gelösten Struktur zu kombinieren (Sequenzidentität > 25%; RMSD 3 der Positionen der C α -Atome < 1.8 Å) und mit den so erhaltenen Daten die noch unbekannte Struktur zu lösen. Je mehr Proteinstrukturen bekannt sind, desto höher wird auch die Wahrscheinlichkeit, daß es bereits eine ähnliche Struktur gibt, die für diese Methode verwendet werden kann. So konnten zum Beispiel im Jahre 2005 etwa 66% aller Proteinstrukturen mittels molekularem Ersatz gelöst werden. Im Rahmen dieser Doktorarbeit wurden die Strukturen von drei Proteinen untersucht. Zunächst einmal wurden die Strukturen von zwei C-terminalen Fragmenten des Proteins TonB gelöst (ab Kapitel 3). Desweiteren ist es gelungen, die Struktur des C-Rings einer natriumabhängigen F 0 F 1 -ATP-Synthase aus Ilyobacter tartaricus (I. tartaricus) zu bestimmen (ab Kapitel 7). Das letzte Projekt umfaßt die Strukturlösung des Enzyms GalU, einer Uridylyltransferase aus Escherichia coli (E. coli) (siehe Kapitel 9). 3 RMSD: root-mean-square deviation 11

12 2 Abstract Thermodynamically, life is based on the energy dependent creation of order against the natural entropic aim of achieving unity distribution. The effects of the second principle of thermodynamics have a number of consequences for the building plan of a living organism. First of all, it needs free energy to maintain its higher internal order. Furthermore, order is only possible, if there exists a barrier between the organism and its surrounding. Another important point is selectivity. There must exist a selective, controlled transport of certain molecules and in addition, also signals must be transmitted across the barrier, so that an organism can react on its surrounding. To achieve all this, living cells possess at least one cell membrane and certain proteins in or near it. This fact makes the cell membrane with its proteins an interesting target for investigation in modern biology. For a better understanding of the function of a protein, its three dimensional structure is of crucial importance. If it is known, one can identify substrate binding sites and catalytical mechanisms can be fathomed. This knowledge is used for example by pharmaceutical companies to design new drugs, which bind selectively at certain sites of a target protein. While it is relatively easy to determine the primary structure of a protein, which is the sequence of its amino acids, and even the succession of α-helices and β-sheets, the secondary structure, can be predicted with growing dependability, the determination of the three dimensional structure (tertiary structure) of proteins is still a challenging area in science. The major part of today s known protein structures was solved by x-ray crystallography (80.2% until ). Since the first protein structures were solved by Perutz (Hemoglobine [1]) and Kendrew (Myoglobine [2]) the principle of this technique remained the same. One takes advantage of the crystalliza- 1 Protein Data Base (PDB), http : // 12

13 2 Abstract bility of biological macromolecules and irradiates those crystals with a highly energetic, monochromatic x-ray beam to collect diffraction patterns with a multitude of reflexions. The intensities (and positions) of these reflexions are relatively easy to measure, but they are not sufficient to solve a structure. It is also necessary to determine the phase of each reflexion, which is a property that cannot be measured experimentally. This constitutes the so called phase problem in crystallography. Unfortunately, the phases carry even more information than the intensities, which was nicely demonstrated by Kevin Cowtans cats and ducks in his Picture Book of Fourier Transforms 2. The last step in structure solvation is therefore to overcome the problem of phase determination. There exists a multitude of solutions to this problem, which can be classified in three major groups. First of all, there are the so called direct methods, where one tries to calculate the phases directly from the intensities. However, very high resolution is necessary to utilize this method and protein crystals tend to refuse to diffract to such high resolution (approximately 1.0 Å). The second method makes use of heavy atoms. The aim is to bind heavy atoms (or compounds) at certain positions in each elementary cell of the crystal (or to use already existing heavy atoms as for example in the case of metallo-proteins). If this succeeds, one can either measure different heavy atom derivatives, or it is possible to measure only one crystal at different wavelengths around the absorption edge of the heavy atom. Evaluating the differences in intensities of related reflexes, one can solve the substructure of the heavy atoms [3, 4]. Subsequently, the experimental phases can be calculated using the Harker construction [5]. The last method of interest is called molecular replacement. Here, measured intensities are combined with phases from a properly oriented, similar structure, which is already known (sequence identity > 25%; root-mean-square deviation (RMSD) of the C α atom positions < 1.8 Å). If the right orientation can be found and the structures are similar enough, it is possible to solve the new structure using the combined informations of phases and intensities. In the near future, this method will steadily increase in importance, since more 2 http : // cowtan/fourier/fourier.html 13

14 2 Abstract and more structures are available and so the probability of finding a similar structure rises. For example, in the year 2005, 66% of all structures submitted to the Protein Data Base (PDB) were solved by molecular replacement. During this PhD thesis, the structures of three proteins were investigated. In the first place, the structures of two C-terminal fragments of TonB were solved (see chapter 3). Secondly, the structure of the C-ring of a Na + -dependent F 0 F 1 - ATP-synthase from Ilyobacter tartaricus (I. tartaricus) was determined (see chapter 7). Finally, the structure of the enzyme GalU, a Uridylyl transferase from Escherichia coli (E. coli) was solved (see Chapter 9). 14

15 3 TonB from Escherichia coli TonB is anchored in the inner membrane of E. coli cells and forms a complex with two further membrane proteins, ExbB and ExbD. These two make use of the chemiosmotic potential over the inner membrane and deliver this energy via TonB, which spans the periplasmic space, to various outer membrane receptors like FhuA, FecA or BtuB (TonB-dependent receptors). In this manner, the energy dependent transport of ferrichrome (FhuA), ferric enterobactin (FepA), ferric citrate (FecA) or vitamin B 12 (BtuB) [7] across the outer membrane is fueled. Thus, the TonB system is crucial for the transport of iron compounds into the periplasmic space of E. coli cells. The transport through the inner Figure 3.1: Selected siderophore-mediated iron acquisition systems of E. coli from Ferguson et al. [6]. 15

16 3 TonB from Escherichia coli membrane is utilized by separate protein complexes as shown in figure 3.1. During this PhD-thesis, the structures of two TonB fragments of different length were solved (77 and 92 residues). These C-terminal parts of TonB interact with the outer membrane receptors. Before the structures were solved, one of the questions was, whether TonB interacts with its receptor as a monomer or as a dimer. Interestingly, the shorter fragment was crystallized as a dimer and the longer fragment was monomeric in solution and showed a weakly connected dimer in the crystal. The comparison between the two structures showed significant structural differences. Recent structures of complexes of TonB with BtuB and FhuA have been reported [8, 9]. In these structures TonB resembles more the structure of the longer fragment and interacts as a monomer. 16

17 4 Dimerization of TonB is not essential for its binding to the outer membrane siderophore receptor FhuA of Escherichia coli 1 JOURNAL OF BIOLOGICAL CHEMISTRY, Vol. 279, No. 11, Issue of March 12, pp , 2004 Jiri Ködding 2, Peter Howard 3, Lindsay Kaufmann 3, Patrick Polzer 2, Ariel Lustig 4 and Wolfram Welte 2,5 4.1 Abstract FhuA belongs to a family of specific siderophore transport systems located in the outer membrane of Escherichia coli (E. coli). The energy required for the transport process is provided by the proton motive force of the cytoplasmic membrane and is transmitted to FhuA by the protein TonB. Although the structure of full-length 1 The atomic coordinates and structure factors (code 1QXX) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (http : // 2 Fakultät fuer Biologie, Universität Konstanz, Universitätsstrasse 10, Konstanz, Germany 3 Department of Microbiology and Immunology, University of Saskatchewan, Saskatoon, Saskatchewan S7N 5E5, Canada 4 Biozentrum Basel, 4056 Basel, Switzerland 5 To whom correspondence should be addressed. Tel.: ; Fax: ; wolfram.welte@uni-konstanz.de. 17

18 4 Dimerization of TonB is not essential for its binding to FhuA TonB is not known, the structure of the last 77 residues of a fragment composed of the 86 C-terminal amino acids was recently solved and shows an intertwined dimer (Chang, C., Mooser, A., Pluckthun, A., and Wlodawer, A. (2001) J. Biol. Chem. 276, ). We analyzed the ability of truncated C-terminal TonB fragments of different lengths (77, 86, 96, 106, 116, and 126 amino acid residues, respectively) to bind to the receptor FhuA. Only the shortest TonB fragment, TonB77, could not effectively interact with FhuA. We have also observed that the fragments TonB77 and TonB86 form homodimers in solution, whereas the longer fragments remain monomeric. TonB fragments that bind to FhuA in vitro also inhibit ferrichrome uptake via FhuA in vivo and protect cells against attack by bacteriophage Φ Introduction The cell wall of Gram-negative bacteria consists of two lipid bilayers, the outer membrane and the cytoplasmic membrane enclosing the peptidoglycan layer. A number of different transport pathways regulate the uptake of essential compounds into the cell. One class of outer membrane transporters is connected to the cytoplasmic membrane by the TonB protein; therefore, they are called TonB-dependent receptors. The three-dimensional structure of a short C-terminal fragment of TonB is available in the literature [10]. One of these receptors in E. coli is the ferric hydroxamate uptake system containing the integral outer membrane protein FhuA [11], which serves as a receptor for the iron siderophore ferrichrome (FC) 6, the antibiotics albomycin and rifamycin CGP 4832, colicin M, and microcin J25, and the phages T1, T5, and Φ80. Other TonB-dependent iron transporters of the outer membrane include FecA for ferric dicitrate (Cit) uptake [12], FepA for enterobactin uptake [13], and BtuB for vitamin B 12 uptake [14]. The transport of all of these ligands requires energy, which is provided by the electrochemical potential of the proton gradient across the cytoplasmic membrane (proton motive force) and is mediated by 6 The abbreviations used are: FC, ferrichrome; Cit, ferric dicitrate; NB, nutrient broth; LB, Luria-Bertani media; r.m.s.d., root-meansquare deviation. 18

19 4 Dimerization of TonB is not essential for its binding to FhuA the protein complexes ExbB, ExbD, and TonB [15, 16, 17]. ExbB/D is located in the cytoplasmic membrane, whereas TonB is attached to the membrane by an N-terminal hydrophobic anchor [18]. The major part of TonB spans the periplasmic space to reach the outer membrane receptors. The crystal structure of FhuA reveals a two domain architecture [19, 20]: a β-barrel consisting of 22 antiparallel strands and a globular domain at the N-terminus (residues 1-160), called the cork or plug domain filling most of the interior of the barrel. Stability studies using differential scanning calorimetry experiments have shown the autonomous behavior of the cork and the β-barrel that unfold at different temperatures [21]. The interactions between the cork domain and the β-barrel consist of nine salt bridges and more than 60 hydrogen bonds [20]. Located at the periplasmic side of FhuA there is a short α-helix, the so-called switch helix (residues 24-29). This α-helix has been found to unwind during or following ligand binding, indicating that this structural change might be a signal for TonB to bind FhuA [19, 20]. This unwinding was observed in the crystal structures of FhuA with bound ferrichrome [19] or albomycin [22]. On the other hand, the crystal structure of FhuA with the rifamycin derivative CGP-4832 demonstrates that ligand binding causes destabilization rather than unwinding of the switch helix [23]. These structures present a specific ligand binding site that is exposed to the external medium and determined by specific hydrogen bonds between the substrate and residues of both the cork and the β-barrel domain. The crystal structures of FepA [24], FecA [25], and BtuB [26] show similar molecular architectures. The presence of a switch helix has only been observed in the structures of FhuA and FecA but not in FepA and BtuB, implying that this structure element is not essential for TonB recognition in general. The pathway of the ligand from the binding site to the periplasm and the mechanism of its transport have not yet been elucidated. Two possibilities are discussed in the literature: 1) conformational change of the cork domain opens up a channel large enough for the siderophore to slide through [7, 27] or 2) the cork domain leaves the barrel together with the bound siderophore [28]. A highly conserved motif among all TonB-dependent siderophore receptors is the TonB-box (residues 7-11: DTITV in FhuA), which plays an important role in the receptor-tonb interaction [29, 30]. The TonB-box is located at the periplasmic side of the cork domain close to the switch helix. Furthermore, the 19

20 4 Dimerization of TonB is not essential for its binding to FhuA globular domains of FhuA and FepA are exchangeable without loss of substrate specificity. For example, a mixed mutant consisting of an FhuA-barrel and an FepA-cork retains the specificity for ferrichrome, the natural substrate for FhuA [31]. Different cork-barrel combinations from several bacterial strains led to the same results [32]. Complexes between wt FhuA or wt FepA with the periplasmic domain of TonB were characterized in vitro [33]. However, up to now there has been no in vitro evidence for interactions between the receptor lacking the cork domain and the TonB protein, and new investigations of FepA indicated that the barrel domain alone could not behave as an active transporter [34]. The TonB protein of E. coli is composed of 239 amino acids of which 17% are proline residues. Most of these are located between residues 75 and 107, spanning the periplasmic space to link the outer membrane receptor with the cytoplasmic membrane [35]. The elongated conformation of this proline-rich region has been demonstrated by NMR studies [36]. This region is not essential for the process of energy transduction [37]. Two other significant regions can be distinguished: 1) a hydrophobic region at the N-terminus (residues 1-32) anchoring TonB to the cytoplasmic membrane. The amino acids between Ser-16 and His-20 were found to be essential for the interaction with the membrane-embedded proteins ExbB and ExbD [38] and 2) a C-terminal domain that forms the contact to the outer membrane receptor. The threedimensional structure of a C-terminal fragment (residues ) reveals a cylindershaped dimer [10]. Each monomer contains three β-strands and a short α-helix arranged in a dimer so that the six β-strands build a large antiparallel β-sheet. The first 10 N-terminal amino acids of this fragment are not visible in the electron density map because of their flexibility. The structure of another energy transducing protein, TolA from Pseudomonas aeruginosa (P. aeruginosa), has been solved recently [39]. Despite a sequence identity of only 24% (Lalign server 7 ) the crystal structure of the periplasmic domain of TolA shows a similar structure and topology, however without dimer formation. The importance of the dimer formation for the mechanism of energy transduction is thus not yet understood. However, it has been shown that a region of TonB 7 form.html) 20

21 4 Dimerization of TonB is not essential for its binding to FhuA Figure 4.1: Amino acid sequence of the C-terminal TonB fragments used in our studies. The location of the site around residues known to be involved in binding to FhuA is shown in boldface [41]. Structural elements derived from the crystal structure of TonB77 are indicated [10]. The amino acid sequence region predicted to form a β-sheet is shown underlined. contributing the critical interaction with the receptor is located around amino acid 160 [40]. This finding was supported by the observation that synthetic nonapeptides with sequence identity to the amino acid region between residues 150 and 166 of TonB are able to inhibit the capacity of wt FhuA to transport siderophores [41]. To understand the role of the C-terminal domain of TonB in the interaction with FhuA, we have investigated FhuA-TonB interactions using purified C-terminal TonB fragments of different lengths shown in Fig. 4.1 (consisting of 77, 86, 96, 106, 116, or 126 amino acid residues, respectively). All TonB fragments except TonB77 were able to form a complex with FhuA. Analytical ultracentrifugation experiments and tryptophan fluorescence measurements also showed that the fragments with 86 or more amino acid residues behave differently than TonB77. In parallel, we analyzed the ability of these TonB fragments to inhibit ferrichrome (FC) and ferric citrate (Cit) uptake in vivo and to protect cells against attack by bacteriophage Φ80. 21

22 4 Dimerization of TonB is not essential for its binding to FhuA 4.3 Experimental procedures Construction of plasmids encoding TonB proteins All constructions, with the exception of pbadtonb118, were created using PCR, and the products were first cloned into an intermediate vector (pskii + or pksii + ). The oligodeoxynucleotides used are listed in Table 4.1. The plasmid pcstonb30 [42], which encodes residues of the periplasmic domain of TonB cloned into pet30a (Novagen), was used as a template to generate the four smaller tonb fragments. Standard PCR conditions were used, with US10-US12 and US26 being the forward primers unique for each fragment as indicated, each one giving a PstI cut site on the 5 -end of the fragment, and US5 as the return primer, creating a HindIII restriction site on the 3 -end of the fragment. In combination with US5, oligonucleotide US10 was used to create pbadtb86, oligonucleotide US11 for pbadtb77, oligonucleotide US12 for pbadtb96, and oligonucleotide US26 for pbadtb106. Each fragment encodes the final number of amino acids of the periplasmic domain of TonB as specified by the TonB fragment number, i.e. pbadtb77 encodes the final 76 amino acids of the periplasmic TonB domain plus a methionine as the first amino acid. The PstI-HindIII digested product was then electrophoresed, and the TonB fragment isolated and cloned into PstI-HindIIIdigested pbad/giii. The construct pbadbtb118 was obtained by digesting Table 4.1: Oligodeoxynucleotides used in creation of pbadtonb and ptb recombinant clones. Oligodeoxy Sequence (5 3 ) nucleotide US5 GAA TTC AAG CTT TTA CCT GTT GAG TAA TAG TCA US10 CTG CAG CAT TAA GCC GTA ATC AGC C US11 CTG CAG CAC CGG CAC GAG CAC AGG CA US12 CTG CAG CAC CGG TTA CCA GTG TGG CTT CA US26 CTG CAG TCA AGT ACA GCA ACG GCT GCA ACC A UR134 CAT ATG GCA TTA AGC CGT AAT CAG CC UR135 CAT ATG CCG GCA CGA GCA CAG GCA UR136 GCT AGT TAT TGC TCA GCG G UR141 CAT ATG CCG GTT ACC AGT GTG GCT TCA UR142 CAT ATG TCA AGT ACA GCA ACG GCT GCA UR143 CAT ATG TTT GAA AAT ACG GCA CCG GCA C UR144 CAT ATG AAA CCC GTA GAG TCG CGT C 22

23 4 Dimerization of TonB is not essential for its binding to FhuA pmftlp [42] with PstI and HindIII and cloning the fragment into PstI- HindIII-digested pbad/giii. Each of these recombinant clones codes for an 18-amino acid (54-bp) signal sequence provided by the vector. Cloning the TonB fragment into the PstI site of pbad/giii downstream of this sequence also adds an 8-amino acid linker at the N-terminal side. For ptb77 to ptb126, UR134, UR135, and UR141 through UR144 were the forward primers for each fragment as indicated, each one creating an NdeI site on the 5 -end of the fragment, and UR136 was the return primer, which hybridizes to the pet30a vector just downstream of the multiple cloning site and contains a Bpu1102I site. Cloning of the resulting PCR fragment back into pcstonb30 created the plasmids ptb77-ptb126, which in each case expresses the indicated TonB fragment without a signal sequence Bacterial strains, plasmids, and growth conditions The strains and plasmids used in this study are shown in Table 4.2. The media used were Tryptone yeast extract (2xYT), nutrient broth (NB) (Difco) and Luria-Bertani media (LB). The growth temperature was 37 C for all experiments. Ampicillin was used at a concentration of 100 µg/ml (Ap100). Strain AB2847 ara was created by P1 transduction of leu::tn10 and ara714 from LMG194 (Invitrogen) into AB2847 [43] Purification of FhuA FhuA405.H 6 was expressed in E. coli strain AW 740 [ ompf zcb::t n10 ompc fhua31] [44] on plasmid phx 405 with a his 6 tag inserted between residues 405 and 406 [45]. The protein was purified as described in the literature [46] with the following changes: for binding experiments the purification was stopped before the detergent exchange from LDAO to DDAO. Fractions containing FhuA were concentrated to 10 mg/ml and dialyzed overnight against 50 mm ammonium acetate ph 8.0 with 0.05% LDAO (N,N-dimethyldodecylamine-Noxide/FLUKA). 23

24 4 Dimerization of TonB is not essential for its binding to FhuA Table 4.2: Strains of E. coli K-12 and plasmids used. Strains and plasmids Genotype Source Strains AB2847 arob tsx malt thi Hantke [43] AB2847 ara arob tsx malt thi ara714 leu::tn10 This study XL Blue reca1 enda1 gyra96 thi-1 hsdr17 supe44 Strata gene rela1 lac [F proab laclqzdm15::tn10] Str W3110 IN(rmD-rmE)1 rph-1 Jensen [47] LMG194 F- lacx74 gal E thi rpsl phoa (PvuII) Invitrogen ara714 leu::tn10 DH5α Delta (argf lac)u169 enda1 reca1 hadr17 Hanahan[48] supe44 thil gyra1 rela1 (F Φ80 lacz M15) Plasmids pcstonb30 tonb fusion in pet30a Howard et al. [42] psk + T/pKS + T ColE1 ori, lacz, Apr Stratagene pbad/giii pbr322 ori, arabad promoter, arac, Apr Invitrogen pmftlp feca tonb fusion in pmalc2g Howard et al. [42] ptb77 tonb fragment in pet30a This study ptb86 tonb fragment in pet30a This study ptb96 tonb fragment in pet30a This study ptb106 tonb fragment in pet30a This study ptb116 tonb fragment in pet30a This study ptb126 tonb fragment in pet30a This study pbadtonb77 tonb fragment in pbad/giii This study pbadtonb86 tonb fragment in pbad/giii This study pbadtonb96 tonb fragment in pbad/giii This study pbadtonb106 tonb fragment in pbad/giii This study pbadtonb119 tonb fragment in pbad/giii This study Purification of the C-terminal TonB fragments C-terminal fragments of TonB (77, 86, 96, 106, 116 and 126 residues, respectively) were over-expressed in E. coli BL21(DE3) cells containing the plasmids ptb77 to ptb126 (shown in Table 4.2) and induced at OD600 = 0.7 by addition of 0.4 mm IPTG (isopropyl-β-d-thiogalactopyranoside, Bio-Vetra). Protein expression was maintained at 37 C for 2 h. The pellets from ml cell culture (2xYT/Kan50) were resuspended in buffer A (20 mm Tris-HCl ph 8.0, 100 mm NaCl, 1 mm EDTA) and the cells were broken by french press (4000 PSIG, 3 passes). After centrifugation at 15, 000 g for 30 min the 24

25 4 Dimerization of TonB is not essential for its binding to FhuA supernatant was loaded on an SP Sepharose cation-exchange column (Amersham Biosciences) and was then washed with buffer A. TonB was eluted from the column with a NaCl gradient at a salt concentration of about 300 mm NaCl. The eluate was then desalted on a Sephadex G25 column (Amersham Biosciences) before loading onto another strong cation-exchange column (Source 15s/Amersham Biosciences). The eluted TonB protein containing about 250 mm NaCl was again desalted on a Sephadex G25 column with buffer A (no EDTA) and yielded protein at a concentration of ca. 4 mg/ml. The mobility of the fragments on 15% SDS-PAGE corresponded to their theoretical molecular masses (Fig. 4.2). The purification was carried out within 1 day to avoid protein degradation. For analytical ultracentrifugation and crystallization experiments an additional gel filtration step was added. The protein was concentrated up to 10 mg/ml (Amicon spin-column with YMCO 5,000) and glycerol was added to a final concentration of 10%. The TonB sample was then loaded onto a gel filtration column (Superose 12 HR 60/10, Amersham Biosciences). Binding experiments with FhuA were done with TonB fragments that were purified without this gel filtration step but mixed with 0.05% LDAO immediately before the incubation with FhuA Purification of the FhuA405.H 6 /TonB complexes Protein solutions containing FhuA (10 mg/ml) and TonB fragment (4 mg/ml), respectively, were mixed in a weight ratio of 1:2 resulting in a large molar excess of TonB in the samples. The protein mixture was then incubated overnight in the presence of 60 µm ferrichrome (Mr= 740, Biophore Research). Glycerol was subsequently added to the protein solution to a final concentration of 10%. The sample was then applied to a Superose 12 HR 60/10 column (Amersham Biosciences), equilibrated, and eluted with the following buffer: 20 mm Tris, ph 8.0, 50 mm NaCl, 0.05% LDAO. The flow rate was kept at 0.1 ml/min. The protein-containing fractions were analyzed by 15% SDS-PAGE and stained with Coomassie blue (Fig. 4.3). For Western blots to detect TonB we used anti-tonb antiserum from rabbit as previously described [42]. 25

26 4 Dimerization of TonB is not essential for its binding to FhuA Figure 4.2: Purification of FhuA405.H 6 and of the C-terminal TonB fragments. The proteins were purified as described under Experimental Procedures 4.3. Their purity was tested by 15% SDS-PAGE followed by Coomassie staining. Apparent molecular masses (in kda) are given on the left. A, TonB77; B, TonB86; C, FhuA405.H 6 (lane 1), TonB96 (lane 2); D, TonB106 (lane 1), TonB116 (lane 2), and TonB126 (lane 3) Crystallization, data collection, and structure solution The C-terminal TonB77 fragment was crystallized under the following conditions: TonB77 was purified as described above and concentrated to 20 mg/ml (Centricon YM 5,000). Hanging drop crystallization plates were used with 1-ml reservoir solution containing 2.0 M sodium formate and 0.1 M sodium citrate, ph 5.6, mixing 2 µl of reservoir solution with 2 µl of protein solution in the drop. Crystals of the size µm 3 grew at 18 C within 2 weeks (Fig. 4.4). For diffraction data collection single TonB77 crystals were soaked in cryobuffer: reservoir solution with 20% glycerol for 1 min and were then flash-frozen in liquid nitrogen. X-ray diffraction data were collected at beamline ID14-4 at the Electron Synchrotron Radiation Facility in Grenoble, France. The crystals diffracted to a resolution of 2.5 Å. Raw data were processed with the program package XDS [49] to a final resolution of 2.7 Å. Higher resolution shells were omitted from the refinement process because of very high R values (> 50%). The space group was determined to be P6422 with the following unit cell parameters: a = Å, b = Å, c = Å, α = 90, β = 90 and δ = 120. The structure of TonB77 was solved using molecular 26

27 4 Dimerization of TonB is not essential for its binding to FhuA Figure 4.3: Size exclusion chromatography of FhuA405.H 6 FC TonB protein complexes. Complex formation of C-terminal TonB fragments of different length and FhuA405.H 6 was tested by 15% SDS-PAGE following the gel filtration step. Panels A-E show the elution peaks of the gel filtration experiments. Lane 1 corresponds to peak 1, and lane 2 corresponds to peak 2. A.) FhuA405.H 6 (lane 1), TonB77 (lane 2). B.) FhuA405.H 6 and TonB86 (lane 1), TonB86 (lane 2). C.) FhuA405.H 6 and TonB96 (lane 1), TonB96 (lane 2). The elution peak 1 containing the FhuA FC TonB96 complex shown in lane 1 was incubated with 1% betaine and purified by gel filtration again. Peak 1 from this purification step contains FhuA405.H 6 (lane 3), peak 2 contains TonB96 (lane 4). D.) FhuA405.H 6 and TonB116 (lane 1), TonB116 (lane 2). E.) FhuA405.H 6 and TonB126 (lane 1), TonB126 (lane 2), FhuA405.H 6 and TonB106 (lane 3), TonB106 (lane 4). 27

28 4 Dimerization of TonB is not essential for its binding to FhuA Figure 4.4: Three single TonB77 crystals grown in 2 M sodium formate and 0.1 M sodium citrate, ph 5.6. replacement with the program MOLREP [50] and REFMAC5 [51] from the program package CCP4 [52]. The search model consisted of all protein atoms of the published model of TonB86 (PDB entry 1IHR) 8. Chain tracing and model building was done with the graphical interface O [53]. The program LSQCAB from CCP4 [52] was used to calculate the r.m.s.d. for the Cα atoms between TonB77 and the existing structure 1IHR of TonB Analytical ultracentrifugation The purified C-terminal fragments of TonB (77, 86, 96, and 116, respectively) were analyzed by sedimentation velocity and sedimentation equilibrium experiments using an AN 60-Ti rotor 316 in a Beckman XL-A Optima equipped with an optical absorbance system (Ariel Lustig, Biozentrum Basel, Switzerland). All protein solutions were freshly purified and gel-filtrated. The buffer was 20 mm Tris, ph 8.0, and 100 mm NaCl in all experiments. Velocity sedimentation data were obtained from 0.5 mg/ml protein solutions and a rotor speed of 54, 000 rpm at room temperature obtaining the sedimentation coefficient (s 20,ω ). Sedimentation equilibrium experiments were done at different 8 The worldwide repository for the processing and distribution of 3-D biological macromolecular structure data. Available at = &page = 0&pdbId = 1IHR. 28

29 4 Dimerization of TonB is not essential for its binding to FhuA concentrations between 0.5 and 2 mg/ml and a rotor speed of 24, 000 and 28, 000 rpm at room temperature. The partial specific volume (ν) of the proteins was calculated on the basis of the amino acid distribution [54] and was near the mean value of globular proteins 0.73 cm 3 /g. These experiments were used to determine the molecular mass (M r ), hydrodynamic radius (R H ), and the frictional ratio (f/f 0 ) [55] of the purified TonB fragments. The calculations were done with the computer program SEGAL 9 based on the numerical fitting of the sedimentation equilibrium pattern to one or two exponential functions Tryptophan fluorescence of the C-terminal TonB fragments Fluorescence spectra were measured from TonB77, TonB86, TonB96, and TonB116, respectively, at an excitation wavelength of 295 nm over the range from 320 to 400 nm (PerkinElmer Life Sciences, L550B). The fragments were purified as described above and used at a final concentration of 0.1 mg/ml Assay of bacteriophage susceptibility Susceptibility to bacteriophage Φ80λi 21 was measured by dropping 5 µl aliquots of 10-fold dilutions of the phage onto freshly poured overlays (100 µl containing 10 8 cells of the various strains added to 3 ml of LB soft agar and poured onto LB plates). The LB soft overlay, LB plates, and bacterial cultures each contained the indicated concentration of arabinose when measuring the effect of the arabinose induction level on the susceptibility of the cells to bacteriophage. The susceptibility was recorded as the log of the highest dilution of phage that gave a confluent lysis zone of the bacterial lawn Assay of siderophore-dependent growth and iron transport The ability of the strains to gain iron from either ferrichrome (FC) or ferric citrate (Cit) was assayed on NB agar plates [42]. To limit the free iron available to the cells, dipyridyl was added to both the agar plates and NB soft overlay 9 SEGAL program description. Analytical Ultracentrifugation at Biozentrum Basel. Available at C/sof tware00.html. 29

30 4 Dimerization of TonB is not essential for its binding to FhuA at a final concentration of 250 µm. When measuring the effect of the arabinose induction level on the ability of the cells to transport iron, the various indicated levels of arabinose were added. Sterile paper discs (6-mm diameter) were saturated with 10 µl of either 1 mm FC, 10 mm sodium citrate, or 100 mm sodium citrate and left to dry. The discs were then placed onto the overlays, which consisted of 10 8 bacteria added to 3 ml of NB soft agar. The plates were incubated overnight at 37 C, and the diameter of rings of growth around each siderophore disc was measured (in millimeters), including the diameter of the disc. No growth was recorded as 6 mm. 4.4 Results Analysis of FhuA/TonB interaction in vitro Several uptake processes across the outer membrane of Gram-negative bacteria are energized by the proton motive force of the cytoplasmic membrane. TonB from E. coli is the protein that transduces the energy from the inner membrane to the outer membrane transporter. One of the transporters is the ferric hydroxamate receptor FhuA of E. coli. The amino acid region around residue 160 of TonB is known to interact with the periplasmic side of this outer membrane [40, 41]. For a detailed analysis of the interaction of FhuA with TonB we prepared protein complexes of FhuA with C-terminal fragments of TonB. Gel filtration (Superose 12 column) of protein mixtures containing FhuA and the C-terminal fragment of TonB led to an elution profile with two well separated peaks, which were monitored on a 15% SDS-PAGE gel stained with Coomassie blue (Fig. 4.3). If a protein complex was present in the sample, the first elution peak contained both FhuA and the TonB fragment (Fig. 4.3, lanes B1, C1, D1, E1, and E3). The second elution peak consisted only of TonB (Fig. 4.3, lanes A2, B2, C2, C4, D2, E2, and E4). In the case of the shortest fragment TonB77, the first elution peak contained only FhuA (Fig. 4.3, lane A1), whereas all longer fragments of TonB co-eluted with FhuA from the gel filtration column. The formation of the FhuA FC TonB complex was only observed in the presence of ferrichrome, which also colors the solution of the first yellowish peak (data not shown). Obviously FhuA is able to form a 30

31 4 Dimerization of TonB is not essential for its binding to FhuA complex with the longer C-terminal fragments of TonB (86, 96, 106, 116, and 126). We also noticed that NaCl had to be present in the buffer at a concentration of at least 50 mm for the FhuA FC TonB106 complex to be stable over a ph range from 4.6 to 8.0. To stabilize the complex we tried 10% glycerol, 1% glucose, 100 mm glycine, or 1% betaine hydrochloride, respectively. Except for betaine none of these additives led to a significant alteration in the behavior of the proteins in these binding experiments. The presence of 1% betaine in the buffer abolished complex formation between FhuA and the TonB fragments. Moreover we have shown that this purified FhuA FC TonB96 complex could be disrupted by the addition of 1% betaine. In the subsequent gel filtration step with 1% betaine in the elution buffer the first peak contained FhuA alone (Fig. 4.3, lane C3), whereas the second peak contained TonB96 (Fig. 4.3, lane C4). This FhuA fraction was colorless, whereas fractions containing FhuA with bound ferrichrome had a yellowish color, suggesting that ferrichrome had dissociated from the receptor after the addition of betaine. It is possible that betaine replaces ferrichrome in the FhuA binding site and thereby induces the release of the TonB-fragment Analytical ultracentrifugation of the C-terminal TonB fragments The crystal structures of TonB88 [10] and of TonB77 (this work) show identical dimers. This led us to investigate the aggregation state of the C-terminal TonB fragments (TonB77, TonB86, TonB96, and TonB116) in solution by analytical ultracentrifugation. Sedimentation coefficients were determined by sedimentation velocity analyses and yielded increasing values from 1.54 to 1.98 S for the fragments TonB77 and TonB86, respectively. The TonB fragments 96 and 116 showed significantly lower sedimentation values of 1.39 and 1.37 S, respectively. The radii calculated from the sedimentation coefficients were thus much smaller for the two longer fragments than for the two shorter ones. The frictional ratio was larger than 1.2, the typical value for spherical globular proteins, suggesting that the conformation is more elongated for these longer TonB fragments. The molecular mass of the fragments was assessed from sedimentation equilibrium (see Experimental Procedures 4.3). Molecular masses 31

32 4 Dimerization of TonB is not essential for its binding to FhuA of 8.5 and 12 kda for the fragments TonB96 and TonB116, respectively, correspond very well with the theoretically calculated masses based on the amino acid sequences of these proteins being 10.8 kda for TonB96 and 12.8 kda for TonB116. The masses of the shorter fragments, TonB77 and TonB86, of 15 and 18.3 kda, respectively, correspond well with the theoretical masses of the homodimers of 17.4 kda for TonB77 and 19.5 kda for TonB86. In both cases no monomeric protein were detectable in the sedimentation equilibrium profiles at 0.25, 0.5, and 1.0 mg/ml protein concentration. The results from analytical ultracentrifugation are summarized in Table 4.3. Table 4.3: Data from the sedimentation velocity and sedimentation equilibrium experiments done with the C-terminal TonB fragments 76, 86, 96, and 116. All fragments were freshly purified (see Experimental Procedures). The partial specific volume necessary for radius and weight determination was calculated based on the amino acid composition of the fragments. The actual molecular weight (given in kilo- Daltons) of each TonB-fragment was measured by analytical ultracentrifugation (see Experimental Procedures) and is compared here to the theoretical molecular weight calculated by the SwissProt server ( TonB fragment TonB77 TonB86 TonB96 TonB116 Sedimentation coefficient (S) Partial specific volume (ml/g) Frictional ratio Radius (nm) Molecular mass (kda)(actual) Molecular mass (kda) (theoretical) Tryptophan fluorescence of the C-terminal TonB fragments Another way to test the aggregation state of the TonB fragments is to measure their tryptophan fluorescence. This method is based on the fact that the molecular neighborhood of tryptophan influences its fluorescence charac- 32

33 4 Dimerization of TonB is not essential for its binding to FhuA teristics. Each of the C-terminal TonB fragments used in this study contains only one tryptophan (Trp-213, see Fig. 4.1) that projects its indole group into the hydrophobic core of the TonB77 dimer (Ref. [10] and this study). The intensity maximum of the fluorescence spectra of the two dimers TonB77 and TonB86 is similar at λ max = 343 nm and at λ max = 340 nm, respectively. The maximum for the TonB fragments 96, 106, 116, and 126 is also similar, but shifted to a shorter wavelength of λ max = 333 nm (Table 4.4). We are not able to explain this blue shift, because we have no information about the environment of tryptophan 213 in the monomeric form of TonB. The fact that the fluorescence spectra of the shorter and the longer fragments is in agreement with the results of the analytical ultracentrifugation analyses indicating that the C-terminal fragments of TonB with a length of 77 and 86 amino acids, respectively, form homodimers in solution, whereas the longer fragments TonB96 and TonB116 remain monomeric (see Table 4.4) TonB fragments shorter than 96 amino acids inhibit TonB function in vivo very weakly It has previously been demonstrated that the entire periplasmic C-terminal domain of TonB can inhibit the function of native TonB in vivo [42]. Various C-terminal fragments of TonB were produced as periplasmic proteins by expression as a fusion protein with the signal sequence of FecA [42]. The periplasmic C-terminal domain was shown to inhibit both ferrichrome and ferric citrate transport as well as growth on iron-limited media when iron was provided as ferrichrome or ferric citrate. In addition, induction of the periplasmic TonB fragment was shown to rescue the producing cells from the lethal effects of colicin M and bacteriophage Φ80, both of which depend on TonB for uptake. In those studies, the smallest fragment of TonB to be assayed and shown to be inhibitory was that produced by pmftlp, which contained the last 118 amino acids of TonB. Here, we similarly assayed fragments containing the last 77, 86, 96, and 106 amino acids of TonB, again when produced as periplasmic proteins. In this case due to the very slight inhibitions observed with some of the fragments (see Tables 4.5 and 4.6), we expressed the fragments as fusion proteins with the signal sequence of the GeneIII protein of the 33

34 4 Dimerization of TonB is not essential for its binding to FhuA Table 4.4: Summary of results for the TonB fragments. Correlation between the ability of the C-terminal TonB fragments to bind to the FhuA ferrichrome complex in vitro, their state of aggregation and in vivo inhibition of siderophore uptake. The aggregation state was determined by analytical ultracentrifugation (see Experimental Procedures). The structure of of the dimeric TonB77 and TonB86 fragments has been solved. TonB fragment TonB77 TonB86 TonB96 TonB106 TonB116 TonB126 In vitro binding to FhuA/Fc Aggregation state Dimeric Dimeric Monomeric * a Monomeric * λ max of Trp fluorescence (nm) Inhibition of si- - +/ + + * * derophore uptake in vivo a Asterisks, samples not measured filamentous phage fd from the vector pbad/giii. This vector allowed higher expression than was obtained for the FecA signal sequence/tonb fusions produced by the pmalc2g vector used earlier (data not shown). As a control, we also created a GeneIII signal sequence fusion to the LP fragment encoding the last 118 amino acids of TonB and assayed it as well. Cells containing the plasmids were grown in varying concentrations of arabinose to induce the fusion proteins and plated on iron-deficient media containing discs soaked in ferrichrome or sodium citrate and were challenged with serial dilutions of bacteriophage Φ80. As before, the LP fragment containing the last 118 amino acids of TonB was capable of inhibiting siderophore-dependent growth, such 34

35 4 Dimerization of TonB is not essential for its binding to FhuA Table 4.5: Growth of E. coli AB2847 ara transformants on NB medium containing ampicillin (100 µg/ml) and dipyridyl (250 µm). 6-mm filter discs were saturated with 10 µl of 1 mm ferrichrome (Fc), 10 mm or 100 mm sodium citrate (Cit) as indicated, left to dry, and placed onto a lawn of the bacteria on media containing the indicated concentration of arabinose to induce the TonB fragments. After overnight incubation the growth zone was measured (in millimeteres) and includes the diameter of the filter disc (6 mm). Therefore a measurement of 6 indicates no visible growth around the disc. Typical results from one of three experiments are shown. Strain and solution (mm) Growth on arabinose at: % AB2847 ara (pbad) FC (1) Cit (10) Cit (100) AB2847 ara (pbadtonb119) FC (1) Cit (10) Cit (100) AB2847 ara (pbadtonb77) FC (1) Cit (10) Cit (100) AB2847 ara (pbadtonb86) FC (1) Cit (10) Cit (100) AB2847 ara (pbadtonb96) FC (1) Cit (10) Cit (100) AB2847 ara (pbadtonb106) FC (1) 33 6 a 6 6 Cit (10) Cit (100) a In one of the three experiments, a very faint full-sized growth ring was observed around the discs for these assay conditions. that at inducer concentrations of 0.02% or higher, growth on ferrichrome or ferric citrate was completely inhibited (Table 4.5). In addition the synthesis 35

36 4 Dimerization of TonB is not essential for its binding to FhuA of this fragment in the presence of 0.002% arabinose or greater was capable of rescuing the cells from the Φ80 challenge (Table 4.6). As can also be seen in Tables 4.5 and 4.6, very similar inhibition and rescue results were observed for the C-terminal 96- and 106-amino acid TonB fragments. In contrast, there was no inhibition of siderophore-dependent growth by the C-terminal 77 amino acid TonB fragment, even when induced with 0.2% arabinose. The C-terminal 86 amino acid fragment inhibited growth only on ferrichrome and only at concentrations of 0.02% arabinose or higher (Table 4.5). Cells expressing the TonB77 fragment were only poorly rescued from the Φ80 challenge, even at the highest inducing concentration of 0.2% arabinose (Table 4.6). Again the cells expressing the TonB86 fragment showed an intermediate phenotype, being rescued more than the cells expressing the TonB77 fragment but substantially less than those containing the TonB96, TonB106, and TonB118 fragments. Table 4.6: Susceptibility of E. coli AB2847 ara transformants to phage Φ80λi [30]. 5 µl of serial 10-fold dilutions of a phage lysate were dropped onto a lawn of the bacteria shown, on media that contained the indicated concentrations of arabinose to induce the expression of the TonB fragments. Results are given as the log of the highest dilution of the phage lysate that gave a confluent lysis zone of the bacterial lawn. Results given in parentheses indicate that there was confluent lysis, but the zones were cloudy. Typical results from one of three experiments are shown. Strain and treatment Strain susceptibility at arabinose: % AB2847 ara (pbad) AB2847 ara (pbadtonb118) 6 1(2) 1 1 AB2847 ara (pbadtonb77) AB2847 ara (pbadtonb86) (3) AB2847 ara (pbadtonb96) AB2847 ara (pbadtonb106)

37 4 Dimerization of TonB is not essential for its binding to FhuA Crystal structure of the TonB77 fragment Model building and refinement ended with a final R-factor of 26.7% and an R free of 27.1%. The data collection and refinement statistics are summarized in Table 4.7. The three-dimensional structure of TonB77 presents a dimer shown in Fig. 4.5, and is very similar to the structure of TonB86 [10]. Effectively, the TonB86 model comprises only 76 amino acids in the electron density map because of the high flexibility of the first 10 N-terminal amino acids. These additional amino acids apparently do not influence dimer formation and the crystal structure of the protein. The Cα atom positions of the two TonB models can be superimposed with an r.m.s.d. of Å. These results are in agreement with our experiments showing that TonB77 and TonB86 both behave as dimers in solution. Table 4.7: Data collection and refinement statistics for the TonB77 homodimer. Values in parentheses refer to the highest resolution shell ( Å). Resolution range (Å) Space group P Unit cell parameters a = b = Å, c = Å, α = β = 90, γ = 120 No. of molecules per ASU 1 No. of observations 39,238 No. of unique reflections 4,053 (409) Completeness (%) 97.5 (97.6) Solvent content (%) 68.2 R merge for all reflections (%) 4.9 (46.3) Average I/σ(I) (3.56) R, R free values (%) 26.7, Discussion The mechanism of energy transduction between the cytoplasmic membrane and the outer membrane via the TonB-ExbB-ExbD system is still unclear. We know, however, that the outer membrane siderophore receptors FhuA and FecA 37

38 4 Dimerization of TonB is not essential for its binding to FhuA Figure 4.5: Stereo ribbon diagram of the C-terminal fragment TonB77, showing the intertwined dimer. One TonB77 molecule is shown in black; the other one is grey. The atomic coordinates have been deposited in the Protein Data Bank (accession code 1QXX). from E. coli form a complex with TonB mediating the transport of siderophores through the membrane. With respect to the crystal structure of FhuA, alone and in complex with the siderophore [19], it is assumed that the conformational changes of the receptor caused by the siderophore create a TonB binding site at the periplasmic side of the receptor. Our in vitro results correlate with this model. In the absence of the siderophore ferrichrome we observed only a very weak complex formation between FhuA and TonB, correlating with results of earlier studies [33]. We also found that ferrichrome can be displaced by betaine in a purified FhuA ferrichrome TonB complex. This exchange of the ligand is followed by a dissociation of TonB underlining the necessity of the specific FhuA-ferrichrome interaction for effective binding of FhuA to TonB. Recently it has been shown that a C-terminal fragment of TonB (TonB86) 38

39 4 Dimerization of TonB is not essential for its binding to FhuA crystallizes as a dimer [10]. We were able to confirm this dimer structure by solving the structure of TonB77 in a different crystal form. Based on these crystallographic data, it was proposed that the dimer formation of TonB is of critical importance for the mechanism of the energy transduction. Our analytical ultracentrifugation experiments support the existence of a TonB77 dimer in solution indicating that this observation is not a crystallization artifact (Tables 4.3 and 4.4). On the other hand, we found no complex formation between TonB77 and FhuA in vitro (Fig. 4.3, lane A1). This observation correlates with the failure of TonB77 to inhibit both ferrichrome uptake via FhuA and ferric citrate uptake via FecA in vivo (Table 4.5). In addition infection of the cells by bacteriophage Φ80 is only very weakly influenced by TonB77 (Table 4.6). TonB86 was found to be dimeric in solution as well (Table 4.4). This fragment, however, is able to bind to FhuA in vitro (Fig. 4.3, lane B1) and to interfere with ferrichrome uptake in vivo (Table 4.5). These findings indicate the special role of the additional amino acid residues in TonB86 compared with TonB77 for binding to FhuA. On the other hand these residues do not influence the dimeric structure of TonB, as shown in both the crystal structure of TonB86 and our analysis of in vitro complex formation. The addition of 10 further amino acid residues or more at the N terminus led to stable monomers in solution as we observed in case of the longer C-terminal fragments: TonB96, TonB106, and TonB116 (Table 4.4). It was also possible to correlate these results with the in vivo inhibition studies, because each of the fragments that were monomeric in vitro strongly inhibited siderophore and bacteriophage Φ80 uptake (Tables 4.5 and 4.6). Our observations agree with the results obtained for the whole periplasmic domain of TonB (residues ). Sedimentation analyses showed this fragment to be monomeric in solution [33], whereas in vivo studies showed this fragment to be inhibitory for all TonB-dependent functions assayed [42]. It was also shown that this fragment binds to FhuA as a monomer [33]. Sauter et al. [56] came to similar conclusions in vivo using a bacterial two-hybrid system. TonB76 formed dimers or multimers in these experiments, whereas TonB207 did not. Full-length TonB, containing the transmembrane part, also showed dimer formation or multimer formation. Based on the data presented here we propose that the dimer formation of the short C-terminal TonB fragments (TonB77 and TonB86) is an exception 39

40 4 Dimerization of TonB is not essential for its binding to FhuA Figure 4.6: Topological diagrams derived from the structures of C-terminal TonB and TolA fragments. The arrows represent β-strands, and cylinders represent α-helices. Panel A corresponds to the structure of the C-terminal fragment TonB86 and TonB77 from E. coli (Ref. [10] and this study). Panel B shows the C-terminal domain of TolA from P. aeruginosa [39], which is very similar to the structure of TolA from E. coli [57]. Panel C shows a putative topology for the C-terminal domain of TonB96 derived from the structure of the TonB77 monomer with an additional N-terminal β-strand consisting of 20 amino acid residues. and not the energetically favored oligomer of native TonB. The stability of the TonB77 dimer is supported by the formation of a 6-stranded β-sheet with three β-strands from each monomer (Fig. 4.6 A). We suppose that a C-terminal TonB fragment longer than 96 amino acid residues can form a 4-stranded β- sheet by itself so that it is monomeric in solution (Fig. 4.6 C). This hypothesis is supported by a secondary structure prediction for the longer TonB fragments, which indicates an additional zone of β-strand around amino acid 148 (Fig. 4.1). The shortest TonB fragment harboring this region is TonB96. The additional β-strand might fold between β-strand numbers 1 and 3 (Fig. 4.6 C). In the TonB77 dimer this strand of the β-sheet is filled by the β-strand 40

41 4 Dimerization of TonB is not essential for its binding to FhuA number 3 of a second TonB molecule [10]. A monomeric form of this proposed topology was also found in the crystal structure of TolA, a protein functionally related to TonB from the TolA TolQ TolR protein complex [58]. The three-dimensional structure of the C-terminal domain of TolA from E. coli in complex with G3p [59] shows a very similar fold to the three-dimensional structure of the same domain of TolA from P. aeruginosa [39] despite an amino acid sequence identity of only 20%. Both TolA structures are composed of three β-strands and two α-helices in the order β-β-α-α-β forming a three-stranded antiparallel β-sheet (Fig. 4.6 B). A similar structure is achieved by the dimeric TonB77 through β-strand swapping [39, 57]. The domain-swapped TonB77 can be generated by connecting the helix from one monomer with the β-strand number 3 of the other monomer (Fig. 4.6, A and B). We hypothesize that the monomeric fragment TonB96 folds into a three-dimensional structure similar to the domain-swapped TonB77 dimer. Alternatively, it seems more likely that the additional 20 N-terminal amino acid residues of a TonB96 monomer might form a β-strand that slides between β-strands 1 and 3 (Fig. 4.6 C), building a four-stranded β-sheet. A comparison of the dimerization state of the C-terminal TonB fragments and their ability to bind to FhuA (Table 4.4) demonstrates that the C-terminal amino acid sequence of the TonB fragments and not their aggregation state determines their binding behavior. The C-terminal fragment TonB77, a stable dimer in solution (Tables 4.3 and 4.4), cannot effectively interact with the membrane receptor protein FhuA (Fig. 4.3, lane A1). This finding correlates with the observation that the TonB sequence around amino acid residue 160 might contribute critical interactions to the binding of the cork domain of FhuA [30, 41]. This amino acid region is missing in TonB77 (Fig. 4.1). On the other hand TonB86, which forms a dimer as does TonB77, contains the major part of the sequence around residue 160 for binding to FhuA (Fig. 4.1) and is able to effectively interact with the receptor molecule (Fig. 4.3, lane B1). In the crystal structure, these residues could not be resolved, probably due to disorder. Because the longer TonB fragments are monomeric in solution and inhibit TonB-dependent transport far more effectively than do shorter ones, we cannot find any evidence that TonB functions as a dimer in the energy transduction process. 41

42 4 Dimerization of TonB is not essential for its binding to FhuA Acknowledgments - We are grateful to the following persons from the University of Konstanz: Dietmar Schreiner for helping us with the protein purifications, Andre Schiefner for x-ray diffraction data collection, Milena Roudna for helping us with the tryptophan measurements, and Kinga Gerber for helpful discussions. 42

43 5 Crystallization and preliminary x-ray analysis of a C-terminal TonB fragment from Escherichia coli Acta Crystallographica Section D Acta Cryst. (2004). D60, Jiri Koedding 1, Patrick Polzer 1, Frank Killig 1, S. Peter Howard 2, Kinga Gerber 1, Peter Seige 1, Kay Diederichs 1 and Wolfram Welte Abstract The TonB protein located in the cell wall of Gram-negative bacteria mediates the proton motive force from the cytoplasmic membrane to specific outer membrane transporters. A C-terminal fragment of TonB from Escherichia coli (E. coli) consisting of amino acid residues (TonB92) has been purifed and crystallized. Crystals grew in space group P2 1 to dimensions of about mm. A native data set has been obtained to 1.09 Å resolution. 5.2 Introduction The cell wall of Gram-negative bacteria consists of two lipid bilayers, the outer membrane and the cytoplasmic membrane, enclosing the peptidoglycan layer. 1 Department of Biology, University of Konstanz, Konstanz, Germany 2 Department of Microbiology and Immunology, University of Saskatchewan, Saskatoon, Saskatchewan S7N 5E5, Canada 43

44 5 Crystallization and preliminary x-ray analysis of TonB92 All essential compounds have to be transported across the outer membrane by diffusion or specific transport pathways. Specifc transporters such as the iron siderophore receptors FhuA, FepA and FecA or the vitamin B 12 transporter BtuB are known to be TonB-dependent as they are connected to the cytoplasmic membrane by the TonB protein. TonB mediates the chemical potential of the proton gradient across the cytoplasmic membrane (proton motive force) to the specific outer membrane receptors. TonB belongs to a protein complex together with ExbB and ExbD [15, 16, 17], which are both located in the cytoplasmic membrane. TonB consists of 239 amino acid residues, with the first 33 residues forming a hydrophobic anchor [18] that attaches TonB to the cytoplasmic membrane. The major part of TonB spans the periplasmic space to reach the outer membrane receptors. This function is achieved by a flexible proline-rich region between residues 75 and 107 [35] that is not essential for the process of energy transduction [37]. The C-terminal domain of TonB forms the contact to the outer membrane receptor, but almost nothing is known about this interaction. It has been shown that a region of critical importance for this protein-protein interaction is located around amino acid residue 160 [40]. The three-dimensional structures of two C-terminal fragments of TonB, TonB86 (residues [10]) and TonB77 (residues [60]), have already been determined. These two TonB fragments crystallized under different conditions and in different space groups. Despite these differences, both structures are similar and reveal a cylinder-shaped dimer. Each monomer contains three β-strands and a short α-helix arranged in a dimer so that the six β-strands can build up a large antiparallel β-sheet. The structure of another energy-transducing protein, TolA from Pseudomonas aeruginosa, has also been solved recently [39]. TolA belongs to the TolA/Q/R system, a system analogous to the TonB/ExbB/ExbD complex, involved in nutrient import [58]. Despite having a sequence identity of only 24% (LALIGN server 3 ), the crystal structure of the periplasmic domain of TolA shows a similar topology, but without dimer formation. The importance of dimer formation for the mechanism of energy transduction is thus not yet understood. It has been shown that C-terminal fragments of TonB with 3 form.html 44

45 5 Crystallization and preliminary x-ray analysis of TonB92 more than 90 amino acid residues behave as monomers in solution [60]. However, Sauter et al. (2003) showed that the periplasmic part of TonB is able to dimerize in vivo [56]. Complex formation between monomeric C-terminal fragments of TonB and FhuA has been observed in vitro [33]. On the other hand, a stoichiometry of 2:1 was recently found for TonB-FhuA complexes in vitro [61]. Here, we present the expression, purification and crystallization of TonB92, a C-terminal fragment of TonB from E. coli consisting of amino acid residues We are currently trying to crystallize TonB92 using the selenomethionine-substitution method [62]. 5.3 Materials and methods Expression and purification The C-terminal fragment of TonB (TonB92) containing the last 92 amino acid residues of the TonB protein was overexpressed in E. coli BL21(DE3) cells containing the plasmid ptb92. For ptb92 the forward primer was US20 (5 -CAT ATG GTG GCT TCA GGA CCA CGC GCA-3 ), creating an NdeI restriction site at the 5 -end of the fragment. The return primer was UR136 (5 -GCTAGT TAT TGC TCA GCG G-3 ), which hybridizes to the pet30a vector (Novagen) just downstream of the multiple cloning site and contains a Bpu1102I restriction site. Cloning of the resulting PCR fragment into pc- STonB30 [42] created the plasmid ptb92. Cells were grown in 2 Y T tryptone yeast extract supplemented with the antibiotic kanamycin (50 mg/l) at 310 K and were induced at OD600 = 0.7 by the addition of 0.4 mm IPTG (isopropyl-β-d-thiogalactopyranoside, BioVetra). Expression of the 10.2 kda protein TonB92 was maintained at 310 K for 2 h. The pellets from ml cell cultures were resuspended in buffer A (20 mm Tris-HCl ph 8.0, 100 mm NaCl, 1 mm EDTA) and the cells were broken using a French press (28 MPa; three passes). After centrifugation at 15, 000 g for 30 min, the supernatant was loaded onto an SP Sepharose cation-exchange column (Amersham Biosciences) and washed with buffer A. TonB was eluted from the column with a NaCl gradient at a salt concentration of about 300 mm NaCl. The eluate 45

46 5 Crystallization and preliminary x-ray analysis of TonB92 Figure 5.1: A native TonB92 crystal of dimensions mm grown in space group P2 1. was then desalted on a Sephadex G25 column (Amersham Biosciences) before loading onto another strong cation-exchange column (Source 15S, Amersham Biosciences). The eluted TonB protein containing about 250 mm NaCl was again desalted on a Sephadex G25 column with buffer A (without EDTA) and yielded protein at a concentration of 4 mg/ml. The mobility of the fragments on 15% SDS-PAGE corresponded to their theoretical molecular weights. The purification was carried out within 1 d in order to avoid protein degradation. An additional gel-filtration step was added. The protein was concentrated to 10 mg/ml (Amicon spincolumn with YMCO 5000) and glycerol was added to a final concentration of 10%. The TonB sample was then loaded onto a gelfiltration column (Superose 12 HR 60/10, Amersham Biosciences) and eluted with buffer A Crystallization and data collection For crystallization, the protein sample was concentrated to 20 mg/ml. Initial screening was performed using Crystal Screen I [63], Crystal Screen II (Hampton Research) and Wizard Screens I and II (Emerald BioStructures Inc.) at 291 K in 96-well sitting-drop plates (Hampton Research). Crystals of dimensions mm were obtained with Wizard Screen I condition No. 46

47 5 Crystallization and preliminary x-ray analysis of TonB92 Figure 5.2: Putative topology model of TonB Further refinement yielded crystals in 24-well hanging-drop plates (Hampton Research) with 1 ml reservoir solution (100 mm imidazole ph 8.0, 1.1 M sodium citrate, 100 mm NaCl) within 5 d (Fig. 1). The crystallization drop contained 3 µl protein solution (20 mg/ml) and 3 µl reservoir solution. Prior to data collection, single crystals were soaked in three different cryoprotectant solutions for 1 min each and were then transferred into liquid nitrogen. The cryoprotectant solutions contained reservoir solution supplemented with 5, 15 and 25% glycerol. Data collection from the native TonB92 crystals was carried out to a resolution of 1.09 Å at beamline X06SA at the SLS, Villigen, Switzerland. Raw data were processed using XDS [49]. 5.4 Results and discussion We have expressed a C-terminal fragment of E. coli TonB (TonB92) containing the last 92 amino acid residues of the protein. TonB92 was purified to near-homogeneity as determined by SDS-PAGE analysis (data not shown) and crystallized at 20 mg/ml with the hanging-drop method (Fig. 5.1). A native data set was collected to 1.09 Å resolution and the raw data were processed 47

48 5 Crystallization and preliminary x-ray analysis of TonB92 Table 5.1: Crystal data and x-ray data-collection statistics for a native TonB92 crystal. Values in parentheses refer to the highest resolution shell. Protein concentration (mg/ml) 20 Crystallization conditions 100 mm imidazole ph 8.0, 1.1 M sodium citrate, 100 mm NaCl Unit-cell parameters (Å, ) a = 22.58, b = 49.32, c = 72.22, α = 90, β = , γ = 90 Space group P2 1 Resolution (Å) ( ) Wavelength (Å) (13.0 kev) Total measured reflections Unique reflections (1727) Completeness (%) 99.8 (99.0) I/σ(I) (1.71) Rmeas (%) 8.3 (59.2) Rmeas was calculated according to Diederichs & Karplus [64]. with the program XDS [49]. The space group was determined to be P2 1, with two molecules per asymmetric unit. Further data-collection statistics are given in Table 5.1. Molecular replacement with the search model 1qxx was carried out with MOLREP [50] from the CCP4 program package [52]. This structure represents the TonB77 dimer [60], which is very similar to the TonB86 dimer [10]. Unfortunately, no useful phase information was obtained using this search model either with or without side-chain atoms. Molecular replacement with a Table 5.2: Results of molecular replacement. C-terminal TonB Results of rotation Results of translation search fragment search (Rf/σ) Corr Corr-2 TonB77 dimer TonB77 dimer, polyalanine model TonB77 monomer TonB77 monomer, polyalanine model

49 5 Crystallization and preliminary x-ray analysis of TonB92 model consisting of an isolated monomer of the TonB77 dimer also failed to give sufficient phase information (Table 5.2). Additionally, a search model was created based on the putative topology model of TonB92 (Fig. 5.2) that we proposed previously for the TonB96 fragment [60]. This model consists of a TonB77 monomer with an additional β-strand at the N-terminus (β1 in Fig. 5.2) that might fold between β-strands 1 and 3. The results of the molecular replacement showed bad packing interactions. Refinement of the data with the program REFMAC5 [51] failed. The fact that none of these models gave us useful phase information indicates that the structure of TonB92 may differ significantly from the published TonB77 and TonB86 dimers. We are currently working on phase determination by direct phasing methods and we are also trying to crystallize TonB92 with incorporated selenomethionine. We thank the staff at the SLS synchrotron beamline for their support. 49

50 6 Crystal structure of a 92-residue C-terminal fragment of TonB from Escherichia coli reveals significant conformational changes compared to structures of smaller TonB fragments JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 280, No. 4, Issue of January 28, pp , 2005 Jiri Ködding 1, Frank Killig 1, Patrick Polzer 1, S. Peter Howard 2, Kay 6.1 Abstract Diederichs 1 and Wolfram Welte 1,3 Uptake of siderophores and vitamin B 12 through the outer membrane of Escherichia coli (E. coli) is effected by an active transport system consisting of several outer membrane receptors and a protein complex of the inner membrane. The link between these is TonB, a protein associated with the cytoplasmic membrane, which forms a large periplasmic domain capable of interacting with several outer membrane receptors, e.g. FhuA, FecA, and FepA for siderophores and BtuB for vitamin B 12. The active transport across the outer membrane 1 Department of Biology, University of Konstanz, Universitätsstraße 10, Konstanz, Germany 2 Department of Microbiology and Immunology, University of Saskatchewan, Saskatoon, Saskatchewan S7N 5E5, Canada 3 To whom correspondence should be addressed. Tel.: ; Fax: ; wolfram.welte@uni-konstanz.de. 50

51 6 Crystal structure of TonB92 is driven by the chemiosmotic gradient of the inner membrane and is mediated by the TonB protein. The receptor-binding domain of TonB appears to be formed by a highly conserved C-terminal amino acid sequence of 100 residues. Crystal structures of two C-terminal TonB fragments composed of 85 (TonB85) and 77 (TonB77) amino acid residues, respectively, have been previously determined (Chang, C., Mooser, A., Pluckthun, A., and Wlodawer, A. (2001) J. Biol. Chem. 276, and Koedding, J., Howard, S. P., Kaufmann, L., Polzer, P., Lustig, A., and Welte, W. (2004) J. Biol. Chem. 279, ). In both cases the TonB fragments form dimers in solution and crystallize as dimers consisting of monomers tightly engaged with one another by the exchange of a β-hairpin and a C-terminal β-strand. Here we present the crystal structure of a 92-residue fragment of TonB (TonB92), which is monomeric in solution. The structure, determined at 1.13 Å resolution, shows a dimer with considerably reduced intermolecular interaction compared with the other known TonB structures, in particular lacking the β-hairpin exchange Introduction The cell wall of Gram-negative bacteria consists of two lipid bilayers, the outer membrane and the cytoplasmic membrane with the peptidoglycan layer in between. A number of different transport pathways regulate the uptake of essential compounds into the cell. Most substances are translocated through the outer membrane by diffusion porins using a concentration gradient. However, substances occurring at very low concentrations like iron siderophores and vitamin B 12 use specific, active, high affinity uptake systems that are driven by chemiosmotic energy transduced to the outer membrane by the TonB protein [65]. Three-dimensional structures of the following TonB-dependent receptors have been determined by x-ray crystallography: FhuA [19, 20], FepA [24], FecA [25], and BtuB [26]. All of them share the same basic architecture, a 22-strand antiparallel β-barrel, which is partially filled with an N-terminal 4 The atomic coordinates and structure factors (code 1U07) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (http : // 51

52 6 Crystal structure of TonB92 globular domain (also called plug or cork domain). The ligand binding site is exposed to the external medium, whereas the TonB binding site is located at the periplasmic side of the receptor. A peptide motif near the N-terminus is conserved among all TonB-dependent receptors: D7TITV in FhuA, D12TIVV in FepA, D81ALTV in FecA, and D6TLVV in BtuB. This conserved region is called the TonB box [14, 66, 30]. Infection of E. coli by bacteriophages T1 and Φ80 and uptake of bacterial toxins (Colicins M and Ia and Microcin 25) also occurs in a TonB-dependent manner, and TonB box motifs are found in these colicins. Binding of the ligand to TonB-dependent receptors induces conformational changes of the cork domain. In the case of the receptor FhuA the unwinding of a short α-helix, which directly follows the TonB box and is exposed to the periplasm (the so-called switch-helix ), was observed [19, 20, 26]. The relocation of the switch-helix likely changes the position and accessibility of the TonB box on the periplasmic side of the receptor [26, 67]. These allosteric conformational transitions, propagated from the ligand binding site to the periplasmic side, may serve to signal the ligand-loaded state of the receptor. The TonB-dependent transporters receive their energy from the chemiosmotic gradient of the cytoplasmic membrane mediated by an inner membrane protein complex composed of ExbB, ExbD, and TonB [15, 16, 17]. The TonB protein of E. coli is composed of 239 amino acid residues and can be divided into three domains. A hydrophobic region at the N-terminus (residues 1-32) anchors the TonB protein to the cytoplasmic membrane [35]. Residues are predicted to assume an α-helical conformation, which contains four highly conserved residues, the so-called SHLS-motif, which was found to be essential for the interaction with the integral membrane protein ExbB [38]. The transmembrane domain is followed by a periplasmic part with high proline content and a conserved C-terminal domain, each composed of 100 amino acid residues. 17% of the TonB sequence are proline residues; most of them are located between residues 75 and 107 [36, 33]. Several observations indicate that the C-terminal domain of TonB (approximately residues ) interacts directly with the TonB-dependent receptors, particularly with their TonB box [68, 40]. Synthetic nonapeptides corresponding to the amino acid sequence of TonB between residues 155 and 166 were 52

53 6 Crystal structure of TonB92 found to be able to inhibit FhuA-dependent transport of ferrichrome in vivo [41]. Complex formation between the C-terminal domain of TonB and the outer membrane receptors FhuA or FepA, respectively, has also been demonstrated by co-purification [33, 60], and disulfide cross-linking was demonstrated between cysteine substitutions in the Q160 region of TonB and the TonB box of BtuB [30] and FecA [69]. Recently the three-dimensional crystal structures of two C-terminal TonB fragments were reported. One of the fragments is composed of the C-terminal 85 amino acid residues of TonB from E. coli ( TonB85, residues [10]). The other fragment is composed of the C-terminal 77 amino acid residues ( TonB77, residues [60]). Both atomic models contain only residues , because the eight additional N-terminal residues of TonB85 are disordered and cannot be identified in the electron density map. Apart from that the two structures are virtually identical and show two molecules tightly engaged with one another as an intertwined dimer. Each molecule forms three β-strands and a short α-helix in the order β-β-α-β. The dimer is stabilized by the exchange of the first two β-strands, which form a β-hairpin. The arrangement of the β-strands in the dimer thus leads to a large 6-strand antiparallel β-sheet. The finding that the C-terminal 77 residues of TonB crystallized as a dimer has led to models in which a dimer of native TonB may respond to occupied high affinity receptors, possibly by rotating [10]. Other recent results have indicated that the situation may be more complicated, however. In a study of C-terminal TonB fragments of increasing length (from 77 to 126 residues), it was found that fragments of 85 residues and shorter formed homodimers in solution, whereas the longer fragments were monomeric [60]. In addition, the shortest fragment with 77 residues (which lacked glutamines 160 and 162) was unable to bind to FhuA, whereas the longer ones (which contained the glutamines) formed a complex with it as determined by gel filtration studies and could inhibit the function of native TonB when produced in vivo. Furthermore, in a recent analytical ultracentrifugation and surface plasmon resonance study [70], it was found that His-tagged C-terminal fragments of 85 and 208 residues both bind to FhuA, with the shorter fragment binding as a dimer. In the absence of FhuA the longer fragment that comprises the entire periplas- 53

54 6 Crystal structure of TonB92 mic part of TonB was monomeric, whereas in the presence of FhuA there were heterotrimers composed of one receptor and two fragments. The three-dimensional structure of the C-terminal domain of TolA that belongs to the related TolQ/R/A system was recently reported alone for the Pseudomonas aeruginosa TolA [39] and in complex with the bacteriophage coat protein G3p for the E. coli TolA [59]. Despite an insignificant sequence identity (20%) between these two TolA proteins, their structures are remarkably similar. They both crystallize as monomers and consist of a three-stranded antiparallel β-sheet flanked by four α-helices positioned on one side of the β- sheet. A structure-based alignment of the C-terminal domains of TolA from P. aeruginosa with TonB from E. coli results in an amino acid sequence identity of only 18% [39]. Although TolA shares the secondary structure pattern β1-β2-α-β3 with TonB77 and TonB85, it lacks the β1-β2 hairpin exchange observed in the structure of TonB77 and TonB85 that enables the formation of a stable dimer. Herein we report the crystal structure of a new C-terminal fragment of TonB from E. coli at 1.13 Å resolution. TonB92 contains the C-terminal 92 amino acid residues of TonB, being only 7 residues longer than TonB85. Its threedimensional structure, however, differs significantly from those of TonB85 and TonB77, and more closely resembles that of TolA, because of the absence of the β1-β2 hairpin exchange. In combination with recent results demonstrating that the longer TonB fragments are able to interact with FhuA more effectively and inhibit native TonB function in vivo, while the shorter fragments do not, these results cast new light on the question whether TonB in vivo functions as a monomer or a dimer. 54

55 6 Crystal structure of TonB Experimental procedures Protein expression and purification TonB92, a C-terminal fragment containing the last 92 amino acid residues of the TonB protein of E. coli, was overexpressed in BL21(DE3) cells containing the plasmid ptb92 and was subsequently purified to near homogeneity [60]. Purification of the SeMet 5 -TonB92 was performed according to the protocol for the native TonB92 with the following exceptions: cells were grown in M63 minimal medium [71] to an A 600 of 0.6, and a selected set of amino acids was then added to a medium containing lysine, phenylalanine, and threonine at a concentration of 100 mg/l; leucine, isoleucine, and valine at 50 mg/l; and selenomethionine at 60 mg/l. 15 min later, 0.3 mm isopropyl 1-thio-β- D-galactopyranoside was added to start the overexpression of SeMet-TonB92. Purified SeMet-TonB92 was concentrated in a 5-kDa filter (Vivaspin), and the flow-through was discarded. The final yield was 1 mg of SeMet-TonB92 per liter of culture. For crystallization, SeMet-TonB92 was used at a concentration of 20 mg/ml Crystallization and data collection Crystallization and data collection of native TonB92 at 1.08 Å resolution has been described elsewhere [72]. Crystals of selenomethionine-substituted TonB92 were grown in 100 mm imidazole, ph 8.0, 1.1 M sodium citrate, and 100 mm sodium chloride using the hanging-drop method. The crystallization drop contained 3 µl of protein solution and 3 µl of reservoir solution. For x-ray data collection the crystal was flash frozen at 100 K using reservoir solution supplemented with 20% ethylene glycol as a cryoprotectant. SeMet data were collected at the Swiss Light Source beamline X06SA to 2.0 Å resolution and measured at three wavelengths corresponding to the peak, inflection, and remote high wavelength of selenium. The data were processed using XDS [49]. Data collection statistics are listed in Table 6.1. The space group was determined to be P2 1 ; the Matthews coefficient is consistent with two molecules per 5 The abbreviations used are: SeMet, selenomethionine; NCS, noncrystallographic symmetry. 55

56 6 Crystal structure of TonB92 Table 6.1: Data collection and refinement statistics of TonB92. R-factors were calculated according to [64]. TonB92 SeMet Data processing TonB92 native Peak Inflection Remote High Wavelength (Å) Unit Cell Parameters a (Å) b (Å) c (Å) β ( ) Unit cell volume (Å 3 ) 79,600 77,900 Solvent content (%) Resolution (Å) ( ) a ( ) ( ) ( ) No. of observed reflections 426,417 (4768) 42,065 (5668) 37,950 (5121) 38,020 (5138) No. of unique reflections 66,106 (2763) 20,072 (2878) 19,945 (2827) 20,027 (2853) Completeness (%) 98.7 (78.5) 96.9 (86.1) 96.5 (85.1) 96.2 (85.1) R meas (%) 7.9 (68.1) 5.6 (17.0) 5.6 (16.9) 5.5 (18.3) R mrgd F (%) 6.5 (77.5) 6.3 (18.7) 6.5 (20.2) 6.8 (21.9) I/σI b 12.7 (1.5) 13.3 (4.9) 12.7 (4.7) 12.9 (4.4) S norm/s ano 1.25 (1.10) 1.12 (1.07) 1.24 (1.07) Refinement statistics Resolution range (Å) ( ) R (%) 13.9 (19.9) Rfree (%) 18.3 No. of nonhydrogen atoms 1,464 No. of residues (atoms) with two conformations 14 (110) Residues with two conformations in molecule A Ser157; Asn159; Gln160; Gln175; Asn190; Ser195; Glu216; Ser222; Val226; Ile228 Residues with two conformations in molecule B Asn159; Arg171; Met201; Asn227 No. of solvent waters 201 Root mean square deviation of bond lengths (Å) Root mean square deviation of bond angles ( ) a Numbers in parentheses correspond to the highest resolution shell. b Size of anomalous signal as calculated in XDS/XSCALE. asymmetric unit and a solvent content of 35% Structure determination Experimental phases were derived by the MAD method using the SeMet data. Four heavy atom sites were identified by using SOLVE [73], corresponding to two methionines in the TonB92 amino acid sequence and two molecules in the asymmetric unit. Initial protein phases were calculated by using SOLVE. These phases were further improved by solvent flattening with RESOLVE [74, 75] leading to a first polyalanine model of TonB92 with 123 amino acid residues out of 184. The program also modeled the side chains of 50 amino acid residues. Because of the small number of heavy atoms, non-crystallographic symmetry (NCS) could not be determined from the heavy atom sites. For finding the 56

57 6 Crystal structure of TonB92 NCS, a program 6 was used that identified regions with 2-fold NCS from the available main-chain fragments. Three corresponding Cα atoms from each NCS region were chosen to generate a new pseudo-heavy atom coordinate file for input to RESOLVE. In the next run, RESOLVE found the 2-fold NCS in the pseudo-heavy atom sites and used it for phase improvement, leading to a new polyalanine model with 140 amino acid residues (76% of total residues). At this time 69 side chains were modeled correctly. The resulting electron density maps were taken for further manual model building using the programs O [53] and COOT [76]. This model was refined against the high resolution data of the native TonB92 using REFMAC5 [51], which is part of the CCP4 program package [52]. The structure was subsequently refined with SHELXL [77]. After modeling 198 water molecules and adding all hydrogen atoms, the last anisotropic refinement resulted in a final R-factor of 13.4% (R free, 18.5%). Refinement statistics are presented in Table 6.1. Atomic coordinates have been deposited in the Protein Data Bank with the accession code 1U Dynamic light scattering Dynamic light scattering experiments with purified TonB92 were carried out using a DynaPro MS instrument from Proterion Corp., High Wycombe, UK. The sample volume was 12 µl of protein solution at 2 mg/ml in 20 mm Tris at ph 8.5 and 100 mm NaCl. Each sample was filtered through a µm pore filter (Whatman) before measurement. The time-dependent intensity signal of the scattered light was evaluated with the program Dynamics Version Analytical ultracentrifugation Experiments were performed on a Beckman XL-A Optima analytical ultracentrifuge equipped with an AN 60-Ti rotor 316 and an optical absorbance system. All experiments were done at 20 C with freshly prepared solution of TonB92 at 2 mg/ml in 20 mm Tris ph 8.5 containing 100mM NaCl. The protein was finally purified by gel-permeation chromatography. In sedimentation 6 K. Diederichs, unpublished work. 57

58 6 Crystal structure of TonB92 velocity experiments absorption was scanned 86 min after the rotor reached the top speed of 52, 000 rpm. Results of sedimentation equilibrium were obtained at protein concentrations of 0.5, 1.0, and 2.0 mg/ml, respectively. At a rotor speed of 34, 000 rpm, equilibrium was achieved after 20 h. Data were analyzed using the software SEGAL In vitro binding experiments of TonB92 to ferrichrome-loaded FhuA were done as previously described for other C-terminal TonB-fragments [60]. 6.4 Results The crystal structure of a C-terminal fragment of TonB from E. coli containing 92 amino acid residues ( TonB92, residues ) was solved at 1.13 Å resolution. Because this fragment contains two methionine residues, we were able to gain phase information by the selenomethionine substitution method (see Experimental Procedures, chapter 6.3). The structure refinement with SHELXL resulted in a well defined model of both chains (called a and b in the following) of TonB92 in the asymmetric unit with an R-factor of 13.4% and an R free of 18.5% (Fig. 6.1). All amino acid residues except the first two N-terminal residues of the a-chain and the first four N-terminal residues of the b-chain were clearly visible in the electron density map Description of experimental structure The overall size of the molecule is 50 Å by 20 Å by 20 Å. The structure of TonB92 presents secondary structure elements in the order of α -β1-β2-αβ3 (Fig. 6.2). Strands β1, β2 and β3 associate to a three-stranded β-sheet. The C-terminal strand β3 is longer than β1 or β2 and interacts with the corresponding part of a second TonB92 molecule by forming an intermolecular antiparallel β-sheet. Dimerization leads to a non-crystallographic 2-fold symmetry (Fig. 6.1). Near the N-terminus a segment of eleven residues ( ES1, Arg154-Pro164) containing the Gln160 region possesses backbone conformation close to β- 7 See 58

59 6 Crystal structure of TonB92 Figure 6.1: Three-dimensional structure of the dimeric TonB92 in ribbon representation. One molecule is shown in red, and the other one is blue. The aromatic residues forming four aromatic clusters are shown in ball-and-stick representation. The C-terminal β-strand forms an antiparallel β-sheet with the other TonB92 molecule. sheet except for one residue, Leu156. A short helix ( α ) formed by six residues (Ala 165 -Leu 170 ) follows which is part of a loop reversing the direction of the main chain. It is followed by a type I β-hairpin composed of β1 (residues Gly 174 -Val 182 ) and β2 (residues Asp 189 -Lys 197 ). The three residues PDG (Pro 184 -Asp 185 -Gly 186 ) at the tip of the hairpin are a conserved motif for the Ton-B family and are a frequent sequence in three-residue β-hairpins [78] which confer high turn stability even in peptides [79]. After a further turn, helix α (residues Glu 203 -Met 210 ) and another segment with conformation close to β-sheet ( ES2, Trp 213 -Glu 216 ), which is apposed to part of ES1 (Arg 154 -Arg 158 ) follow. Surprisingly only four hydrogen bonds of this interaction stabilize the conformation of ES1: the main chains of Arg 154, Leu 156, and Ser 157 form hydrogen bonds with the main chain of Arg 214 and Glu 216 (Fig. 6.3). Nevertheless, the electron density of residues in ES1 is well defined (Fig. 6.4), and their B-factors do not deviate from the average value. After β2, the main-chain again changes direction and forms the long C-terminal strand β3 59

60 6 Crystal structure of TonB92 Figure 6.2: Topological diagram of TonB92 showing secondary structure elements derived from the crystal structure. β-strands are indicated by arrows, and α-helices are indicated by cylinders. Structural elements that were not observed in the structures of TonB77 and TonB85 are marked with an asterisk. (Pro Thr 236 ). β3 protrudes out of the domain by about 8 residues and associates with an antiparallel β3 of another molecule resulting in an intermolecular β-sheet β3-β Comparison with crystal structures of TonB77 and TonB85 In the following, the crystal structure of TonB92 will be compared with that shared by TonB77 and TonB85. The common structure of the latter two fragments contains the residues and will be referred to as the tight dimer. A superposition of one molecule from the crystal structure of TonB92 and one from the tight dimer (Fig. 6.5) and their secondary structure assignments (Fig. 6.6) show that the secondary structure elements β1, β2, α, and β3 are formed and arranged similarly in both structures. In the tight dimer, the two molecules are engaged by exchanging their β1-β2 hairpins with one another resulting in the formation of a six-stranded intermolecular antiparallel β-sheet 60

61 6 Crystal structure of TonB92 Figure 6.3: Ball-and-stick representation of the amino acid residues of TonB92, which are involved in stabilizing the extended N-terminal ES1 segment by formation of hydrogen bonds. The bonds are shown as dotted green lines with the distances given in Ångstroms (Å). The representation of the side-chain atoms is incomplete. (see Fig. 5 of Ref. [60]). In contrast, in the TonB92 structure the β1-β2 hairpin does not exchange with another molecule but rather takes up the same place that is filled in the tight dimer with the β1 -β2 hairpin from the other molecule. As seen in Fig. 6.5, the orientation of the β-hairpin of TonB92 and of the tight dimer can be superimposed onto each other very well. The comparison shows that the TonB chain has additional flexibility that is not obvious from the tight dimer structure and that results in two hinges before and after the β1-β2 hairpin. The backfolding of the β-hairpin weakens the monomer-monomer interaction and leads to both a reduced length of β1 in TonB92 and the formation of a new helix α (Fig. 6.6). Strand β1 of the tight dimer starts with the residues Ala 169, Leu 170, and Arg 171, whereas in case of 61

62 6 Crystal structure of TonB92 Figure 6.4: Electron density map (2mF o - DF c ) at 3σ (dark blue) and at 2σ (light blue) around amino acid residue Gln 160 of ES1. The sidechains of Asn 159 and Gln 162 show multiple conformations. TonB92 these residues are part of the helix α. The presence of seven additional N-terminal residues of TonB92 as compared with TonB85 thus seems to abolish the β1-β2 exchange leading to a remarkably different crystal structure but retaining the basic arrangement of the secondary structure elements β1, β2, α, and β3 (see Figs. 6.5 and 6.6). In the tight dimer of TonB85, 10 N-terminal residues (Ala 155 -Pro 164 ), previously shown to be important in interactions between TonB and the TonB box of the receptors [30], could not be modeled due to a high flexibility of this region. In the corresponding ES1 segment of the TonB92 structure, residues possess dihedral angles close to β-sheet but lack any hydrogen bonds of the backbone and the side chains to other parts of the molecule. Interactions via four hydrogen bonds occur only further N-terminally between residues Arg 154 -Ser 157 and ES2. Conversely, in the tight dimer the residues of helix α show β-like backbone conformation and are positioned close to the ES2 residues to which they form 62

63 6 Crystal structure of TonB92 Figure 6.5: Superposition of the three-dimensional structures of one molecule of TonB92 (in red) with one molecule of the tight dimer i.e. the structure of TonB77 or TonB85 (in blue). The β1-β2 hairpin of the second molecule from the tight dimer is shown in yellow. The PDG-loop between β1 and β2 containing residues Pro 184, Asp 185, and Gly 186 is indicated. two hydrogen bonds. The interaction partners of ES2 thus shift by approximately ten residues from the N-terminal end of ES1 in the TonB92 structure to its C-terminal end. These differences are surprising, because TonB77 and TonB85 both contain all residues which in the TonB92 structure form α and ES2 and in fact TonB85 even contains most of the residues that form ES1, except for one residue, Arg154. Five aromatic amino acid residues were found to be conserved in TonB of several Gram-negative bacteria, forming four clusters in the tight dimer. Point mutations of these residues lead to a reduced activity of TonB [80]. In the two different crystal structures, these aromatic residues cluster in the same way, and the residues even have similar orientations (Fig. 6.1). In particular the cluster composed of Phe 180, Trp 213, and Tyr 215, respectively, can be superimposed upon the corresponding residues of TonB85 with low deviation (Fig. 6.7). Phe 180 resides on the exchangeable β-hairpin. For this reason the aromatic cluster (Phe 180, Trp 213, and Tyr 215 ) is formed by residues 63

64 6 Crystal structure of TonB92 Figure 6.6: C-terminal amino acid sequence of TonB from E. coli. The position of the N-termini of the fragments TonB92, TonB85, and TonB77 are indicated by arrows. The secondary structure elements found in the structure of TonB92 and the tight dimer (i.e. TonB77 and TonB85) are indicated by arrows (β- strands) and boxes (helices). Conserved residues are shown with a black background. The segments ES1 and ES2 are indicated by lines. of both molecules of the tight dimer, whereas in the structure of TonB92 all three residues belong to one molecule. In the other cluster of the tight dimer, residues Phe 202 and Phe 230 interact by stacking of their aromatic side chains. In the TonB92 structure a similar arrangement of these two residues is found but enlarged by Tyr 163 from ES1. The interaction of Phe 202 with Tyr 163 meets the criterion of an edge on interaction [81] and thus may contribute to stabilizing the folding back of the β1-β2 hairpin in the two hinges before β1 and after β2. Tyr 163 has been shown to be critical for FecA function [69]. The structure of TonB92 and the tight dimer share the formation of an intermolecular antiparallel β-sheet β3-β3. The residue pairing is, however, 64

65 6 Crystal structure of TonB92 Figure 6.7: Superposition of one cluster of aromatic residues, Phe 180, Trp 213, and Tyr 215, of the TonB structures. The residues from the structure of TonB92 are colored in red. Residues of the tight dimer are given in blue and yellow depending on which of the two TonB molecules they belong to. slightly different. In TonB92 the center is shifted by one residue toward the C-terminus compared with the tight dimer Oligomerization of TonB92 in solution, and complex formation with FhuA Experiments were carried out to compare oligomerization of TonB92 and complex formation with FhuA in solution with other fragments, TonB77, TonB86, TonB96, and TonB116, which were studied by us recently [60]. In contrast to the fact that TonB92 forms a dimer in the crystal, we find that it behaves monomeric in solution as determined by dynamic light scattering and analytical ultracentrifugation. Dynamic light scattering experiments show monodispersity, and the autocorrelation function of scattered light intensity can be fitted by assuming a globular protein in solution with a calculated molecular mass of 13 kda (data not shown). This finding is consistent with the results of the analytical ultracentrifugation experiments presenting an average molecular 65

66 6 Crystal structure of TonB92 mass of 12 kda and a sedimentation coefficient of s 20 = The experimentally determined molecular weights thus correspond to the calculated molecular mass of 10.2 kda for monomeric TonB92. In addition, a complex of purified TonB92 and FhuA loaded with ferrichrome could be isolated by size exclusion chromatography (data not shown), providing evidence for the ability of TonB92 to bind to its specific siderophore receptor FhuA in vitro. 6.5 Discussion TonB from E. coli is known to be essential for the transduction of chemiosmotic energy to TonB-dependent outer membrane receptors like FhuA and FepA. The atomic structure of the complete TonB (239 residues) is not yet known, but a rather elongated shape has been inferred from centrifugation and gel-filtration data [33, 70]. Recently, crystal structures of two C-terminal fragments of TonB, TonB77 [60] and TonB85 [10], have been determined. Both fragments crystallized as an intertwined homodimer. Here we present the structure of TonB92, a C-terminal fragment of TonB from E. coli containing 92 amino acid residues. Current knowledge of the molecular mechanism of TonB-dependent transport through the outer membrane is still very rudimentary. TonB transduces the energy that is needed for active transport of siderophores and vitamin B 12 through its cognate outer membrane receptors. The low copy number of TonB molecules compared with the number of TonB-dependent receptors [65] suggests that TonB probes many receptors and transduces energy only to ligand-loaded ones. Transport is initiated by binding of the ligand to the receptor binding site with submicromolar affinity. In the paradigmatic case of FhuA, the binding site is formed by residues in the external loops of the β- barrel and residues at the apices of three loops of the cork domain [19, 20]. The binding is accompanied by a conformational transition propagated through the cork domain to the periplasmic surface and results in the allosteric unwinding of a short switch helix and a relocation of the periplasmic N-terminus, likely accompanied by a relocation of the TonB box as found in the structures of unliganded BtuB and BtuB with bound vitamin B 12 [26, 82]. Several lines of evidence support the existence of a specific interaction be- 66

67 6 Crystal structure of TonB92 tween the C-terminal domain of TonB and the receptor, which is critical for TonB-dependent transport. A close apposition of the TonB box of the receptor with the C-terminal TonB region around Gln 160 has been inferred from in vivo experiments using tonb suppressor mutants [68], crosslinking data [66], disulfide formation in cysteine substitution mutants [30], and competitive binding of peptides [41]. In vitro, complexes between receptors and TonB have been described that are more stable when the receptor contains bound ligand [33, 67, 60, 69, 83, 84]. Because of the crystallographic evidence for an allosteric conformational change at the periplasmic surface of the receptor, TonB could in principle bind there without significantly altering its conformation, but an induced fit of TonB upon binding cannot be excluded. Conflicting views exist about the oligomeric form at which TonB persists in the free form and when bound to its receptor. Under in vivo conditions, Sauter et al. [56], using ToxR fusions, found that the periplasmic part of TonB existed as a monomer, whereas the entire TonB and the C-terminal TonB76 fragment formed dimers. Under in vitro conditions, Koedding et al. [60] with analytical ultracentrifugation experiments found that TonB77 and TonB85 formed dimers, whereas longer TonB fragments up to TonB116 formed monomers. Under similar conditions, Khursigara et al. [70] found monomeric behavior of the periplasmic part of TonB and dimers for TonB85. Also under similar conditions, Moeck and Letellier [33] report preliminary results indicating a 1:1 stoichiometry for complexes of the periplasmic part of TonB with FhuA, whereas Khursigara et al. [70] found that TonB85 as well as the periplasmic part of TonB bound to FhuA at a 2:1 stoichiometry. A conclusion of Khursigara et al. was that the periplasmic part of TonB, while forming monomers in absence of FhuA, formed dimers upon binding to FhuA that remained stable after dissociation from the receptor. Moreover, Khursigara et al. observed with surface plasmon resonance experiments that the periplasmic TonB fragment displayed binding kinetics indicative of two binding sites, whereas TonB85 displayed binding to FhuA only by a single binding site. The binding site of the periplasmic part of TonB, which was of higher affinity, apparently was not shared with TonB85 and was responsive (increasing its affinity) to the presence of the FhuA ligand ferrichrome. This site, which must reside N-terminal of the C-terminal TonB domain, might be identical with the high affinity site in 67

68 6 Crystal structure of TonB92 the proline-rich region described by Brewer et al. [85]. Because the transport process is coupled to the use of chemiosmotic energy, the transduction of that energy to the receptor must be triggered at a distinct moment that likely coincides with the recognition of a liganded receptor by TonB. Furthermore, because the ligand has to be dissociated from its engagement with the high affinity binding site, one would expect the complex between TonB and the receptor to be rather tight. Because no high affinity binding site for FhuA has so far been found in the C-terminal domain [60], a tight complex may be formed only transiently during energy transduction, or its formation requires another high affinity site that is further N-terminal as, e.g., the site inferred from the results of Khursigara et al. [70] and Brewer et al. [85]. Finally, because the different TonB-dependent receptors show no overall conservation of residues on the periplasmic surface, the interaction of TonB with the receptors probably relies rather on backbone interactions than on sequence-specific side-chain interactions. In the following, we will attempt to correlate all of the structural results concerning TonB with these various biochemical and functional data Why does the presence of the additional N-terminal residues in TonB92 as compared with TonB85 change a dimer to a monomer in solution and cause such a significant conformational difference in the crystal structure? In solution, TonB85 and shorter C-terminal fragments of TonB behave as dimers, whereas TonB92 and larger C-terminal fragments form monomers (data from this work and Ref. [60]). Indeed, the loss of the β1-β2 hairpin exchange significantly weakens the intermolecular interactions in TonB92 as compared with the tight dimer. Nevertheless TonB92 still forms a dimer in the crystal via its long β3 strand indicating that under special circumstances it might dimerize in solution as well. One of the keys to understanding the structural transition from the tight dimer of TonB85 to the TonB92 structure must be the seven additional residues at the N-terminus of TonB92. The critical residue in the TonB92 structure 68

69 6 Crystal structure of TonB92 appears to be Arg 154, which is the first residue missing in the TonB85 fragment and which contributes one hydrogen bond to the ES1-ES2 interaction in the structure of TonB92. Moreover, the other additional six N-terminal residues (Ser 148 -Pro 153 ) do not interact with the rest of the molecular structure Is there a role for the two conformations in the transport process? Experimentally two crystal structures are observed that also show a simple way to derive one from the other. It is possible that these obviously stable structures of the tight dimer and of TonB92 reflect two conformational states of native full-length TonB independent of whether the dimer actually forms in vivo or not. Given that, it could be that each state will play a role in two different, distinct moments of the transport process. Because the basic steps of that process are virtually unknown, only rather tentative hypotheses can be put forward. Cross-linking data and experiments sensitive to molecular mass indicate that TonB binds the receptors at two different affinities, at a low affinity via a site in the C-terminal domain and at a higher affinity via a site further to the N- terminus. Among the subsequent associations, some likely are of higher affinity to dissociate the ligand from its micromolar binding site. The structure of TonB92 seems to be a good candidate for a conformation that binds to the unliganded receptor at low affinity, because the Gln 160 region of ES1 is surfaceexposed. The TonB box may be suitably placed for the interaction only in the liganded receptor, which would explain the cross-linking results of BtuB and FecA with TonB [30, 69]. The association of ES1 with the TonB box could result in the dissociation of the weak interaction between ES1 and ES2 (which requires only four hydrogen bonds to be broken) and thus trigger a change to a conformation that resembles that of one molecule of the tight dimer. The β1-β2 hairpin with its rather stable PDG turn could fold out (see Fig. 6.5) and would render the C-terminal TonB domain a reactive protein toward the receptor due to the many unsaturated hydrogen bonds. This could be the conformation of the C-terminal TonB domain in the complexes with its receptor. The dimer structure itself 69

70 6 Crystal structure of TonB92 might not form because of the smaller copy number of TonB compared with its receptors. Interestingly, a mutation of the Glycine in the PDG motif of the hairpin, G186D, abolishes growth on ferrichrome and confers strong resistance to colicins B and M (Table 3 in Traub et al. [86]), which supports the idea of an important role for the hairpin. Other scenarios can be envisaged. Some of the proposed transport mechanisms involve a pulling force generated by conversion of chemiosmotic energy. That force may apply to the N-terminus of TonB92 after binding to the receptor and disrupt the ES1-ES2 bond. Subsequently ES1 may be pulled along ES2. Helix α may unwind, and part of its residues eventually become apposed and hydrogen-bonded to ES2. The unwinding of α would abolish the hinge, which bends the main chain backwards between ES1 and β1, and the β1-β2 hairpin would fold out. The two conformations of TonB92 and of the tight dimer obviously are both rather stable, and their interconversion in solution may be inhibited by a rather high activation barrier. Khursigara et al. [70] describe dimer formation of the periplasmic part of TonB by FhuA. We speculate that liganded FhuA may act as a catalyst for TonB dimerization due to its interaction with the ES1 segment. Moreover, Khursigara et al. [70] and Brewer et al. [85] describe a high affinity interaction site of TonB with FhuA further N-terminal of the C-terminal domain. After high affinity binding of one TonB molecule to FhuA the receptor may catalyze dimer formation of TonB if the latter is present at sufficiently high concentration. The result would be a heterotrimer formed by one liganded FhuA with a TonB tight dimer. After dissociation from FhuA, the TonB tight dimer would persist as observed by Khursigara et al. [70]. This could serve as an inactivation mechanism for TonB, preventing TonB-dependent receptors from becoming blocked by an excess of inactive periplasmic TonB domains. In the above scenarios, the folding out of the β1-β2 hairpin is presumed to lead either to a transient but tight FhuA-TonB complex of 1:1 stoichiometry or to the formation of the tight dimer with an adjacent TonB molecule. An idea of the mechanism by which the former complex may arise may be provided by two structures of the related domain from TolA. This energy-transducing protein which belongs to the TolQ/R/A system is known to connect the cy- 70

71 6 Crystal structure of TonB92 Figure 6.8: Superposition of the three-dimensional structures of one molecule of TonB92 (red) and the C-terminal domain of TolA from E. coli (blue) (pdb accession code: 1Tol, Lubkowski et al. [59]). The N- terminal α-helix of TolA, which is not present in TonB, is not shown. toplasmic membrane with the outer membrane [87]. The C-terminal domain of TolA has been crystallized as a monomer [39]. A superposition of TonB92 with the C-terminal fragment of TolA is given in Fig. 6.8 and shows a close superposition of the three β-strands and the long α-helix. A second structure shows this TolA domain complexed as a monomer with a bacteriophage coat protein, with which it interacts during infection [59]. The association shows an intermolecular β-sheet formed and stabilized by intermolecular hydrogen bonds between TolA and G3p. At present, the details of the interactions and conformational changes that take place during energy-dependent transport involving TonB and its receptors remain far from clear. However, the two structures of the tight dimer and TonB92 represent constraints, which a final model of transport will have to take account of, and perhaps clues to the dynamic mechanism by which TonB interacts with the receptors and transduces energy to the transport process. 71

72 6 Crystal structure of TonB Acknowledgments We thank the staff at the X06SA (Swiss Light Source/Switzerland) synchrotron beamline for their support. We are grateful to Ariel Lustig from the Biozentrum Basel/Switzerland for performing the analytical ultracentrifugation of TonB92. We also thank Ramon Kanaster for assisting us in purifying TonB92 and Kinga Gerber for helpful discussions. 72

73 7 C-ring from Ilyobacter tartaricus All living cells require energy, which is predominantly provided in the form of ATP (Adenosine-Tri-Phosphate). The hydrolysis of ATP to ADP (Adenosine- Di-Phosphate) and phosphate delivers about 30 kj/mol, which can be used by the cell [88, 89]. To generate ATP, cells utilize so called ATP synthases. These are multisubunit protein complexes [90] (see figure 7.1). The ATP synthase can be separated into two subcomplexes. One is the soluble F 1 part, responsible for ATP generation and the other one is the membrane spanning F 0 part. The latter makes use of the energetically favoured transport of protons (or sodium ions) across the cell membrane. It consists of one a- and two b-subunits and a ring of c-subunits, the c-ring. The protein structure of the F 1 part was solved in 1994 by Abrahams et al. [92] and gave impressing insights into the mechanism of ATP synthesis. However, the structure solution of the F 0 part proved to be challenging due to the fact that it comprises the membrane spanning part of the ATP synthase. Pending this PhD-thesis, we successfully solved the structure of the c-ring of a sodium dependent F-type ATP synthase from I. tar- Figure 7.1: Model of the F 0 F 1 -ATPsynthase (Image by R. L. Cross [91]). 73

74 7 C-ring from Ilyobacter tartaricus taricus, consisting of 11 monomers. Since the phase problem could neither be overcome by heavy atom derivatization nor by Se-Met substitution, we finally solved the structure by molecular replacement using a new software package called PHASER [93]. A low resolution (6 Å) electron microscopy model which was provided by Vonck et al. was used as search model. 74

75 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus SCIENCE VOL 308 (2005), pp Thomas Meier 1, Patrick Polzer 2, Kay Diederichs 2,3, Wolfram Welte 2 and Peter Dimroth 1,3 8.1 Abstract In the crystal structure of the membrane-embedded rotor ring of the sodium ion-translocating adenosine 5 -triphosphate (ATP) synthase of Ilyobacter tartaricus at 2.4 Ångstrom resolution, 11 c-subunits are assembled into an hourglassshaped cylinder with 11-fold symmetry. Sodium ions are bound in a locked conformation close to the outer surface of the cylinder near the middle of the membrane. The structure supports an ion-translocation mechanism in the intact ATP synthase in which the binding site converts from the locked conformation into one that opens toward subunit a as the rotor ring moves through the subunit a/c interface. 8.2 Introduction In the F 1 F 0 ATP synthase, the cytoplasmic F 1 catalytic domain (subunits α 3 β 3 γδɛ) is linked by means of a central and a peripheral stalk (subunits γ/ɛ 1 Institut für Mikrobiologie, Eidgenössische Technische Hochschule (ETH), Zürich Hönggerberg, Wolfgang-Pauli- Str. 10, CH-8093 Zürich, Switzerland. 2 Fachbereich Biologie, Universität Konstanz M656, D Konstanz, Germany. 3 To whom correspondence should be addressed. dimroth@micro.biol.ethz.ch (P.D.); kay.diederichs@uni-konstanz.de (K.D.) 75

76 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus and b 2 /δ, respectively) to the intrinsic membrane domain called F 0 (subunits ab 2 c ). Each of these domains functions as a reversible rotary motor and exchanges energy with the opposite motor by mechanical rotation of the central stalk. During ATP synthesis, energy stored in an electrochemical gradient of protons or Na + ions fuels the F 0 motor, which causes the stalk to rotate with the inherently asymmetric γ subunit acting as a camshaft to continuously change the conformation of each catalytic β subunit. These sequential interconversions, which result in ATP synthesis, endow the binding sites with different nucleotide affinities (for reviews see [94, 95, 96]). The rotational model, which explains a wealth of biochemical and kinetic data, is impressively supported by the crystal structure of F 1 [92] and was experimentally verified by biochemical, spectroscopic, and microscopic techniques [97]. The F 0 motor consists of an oligomeric ring of c subunits that is abutted laterally by the a and b 2 subunits [98]. The c-ring, together with γ/ɛ subunits, forms the rotor assembly, which spins against the stator components ab 2 δα 3 β 3. Ion translocation at the interface between subunit a and the c-ring, driven by the ion motive force, is thought to generate torque [99, 100, 101, 102] applied to the γ subunit, which is then used to promote the conformational changes required for ATP synthesis at the F 1 catalytic sites. Despite intense efforts, little is known about the structural details of F 0. This lack of information hinders our understanding of how this molecular motor functions. The nuclear magnetic resonance (NMR) structures of the c monomer of Escherichia coli [99, 103] showed that the protein is folded into two α-helices linked by a loop. Other structural studies indicated that the c subunits of the oligomer are tightly packed into two concentric rings of helices [90, 104]. The number of c monomers per ring varies between n = 10, 11, and 14 units in the ATP synthases from yeast [90], I. tartaricus [104], and spinach chloroplasts [105], respectively. 76

77 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus Figure 8.1: Structure of the I. tartaricus c 11 -ring in ribbon representation. Subunits are shown in different colors. (A) View perpendicular to the membrane from the cytoplasmic side. Two subunits are labeled. (B) Side view. The blue spheres represent the bound Na + ions. Detergent molecules inside the ring are shown with red and grey spheres for clarity. The membrane is indicated as a grey shaded bar (width, 35 Å). The images were created with PyMOL [106]. 8.3 Structure of the c-ring We chose to determine the crystal structure of the I. tartaricus c-ring because of its inherent stability and relative ease of isolation [107]. After purification and crystallization of wild-type c-ring (see chapter 8.6), the structure was solved by molecular replacement [93] using a medium-resolution (6 Å) c-ring backbone model derived from electron crystallography [104]. The asymmetric unit of the crystal contains 4 c-rings. These rings are arranged in two parallel, but laterally translated c-ring dimers each formed by a coaxial association of two rings that interact with their termini in a tail-to-tail fashion. A noncrystallographic symmetry restraint was imposed during refinement at 2.4 Å over the 44 monomers in the asymmetric unit (3916 residues) with the exclusion of the loop regions that form crystal contacts, which substantially improved the electron density and resulted in an atomic model without Ramachandran plot outliers (see table 8.1 for a summary of data collection and refinement 77

78 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus Table 8.1: Summary of data collection and refinement statistics. Data Processing Wavelength [Å] Spacegroup P2 1 Unit cell parameters [Å] a= 147.7, b= 140.0, c= α = γ = 90, β = Solvent content [%] 65.1 Resolution [Å] [ ] No. of observed reflections [165501] No. of unique reflections [23376] Completeness [%] 97.9 [95.2] R meas [%][64] 18.3 [63.5] R mrgd F [%][64] 12.8 [48.7] I/σI 10.0 [3.3] Refinement statistics Resolution [Å] R-factor 20.1% R free 24.6% No. of residues 3916 No. of solvent waters 513 No. of ions 44 No. of fatty acid chains 44 r.m.s. deviation of bond length [Å] r.m.s. deviation of bond angles [ ] Numbers in brackets correspond to the highest resolution shell. statistics). The electron density map of a single c-ring shows a cylindrical, hourglassshaped protein complex of 70 Å in height and with an outer diameter of 40 Å in the middle and 50 Å at the top and bottom. Eleven c subunits, each composed of two membrane-spanning α helices forming a hairpin, are arranged around an 11-fold axis, creating a tightly packed inner ring with their N-terminal helices (Fig. 8.1). The C-terminal helices pack into the grooves formed between N-terminal helices, producing an outer ring, in agreement with previous medium resolution structures [90, 104]. In the electron density map, the backbone and side chains of all amino acids are clearly defined, except for the C-terminal glycine. The N- and C-terminal helices are connected by a loop formed by the highly conserved peptide Arg45, Gln46, and Pro47, which is exposed to the cytoplasmic surface (Fig. 8.2) [108, 109]. The chain termini are exposed to the periplasm. 78

79 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus Figure 8.2: Section of the c-ring showing the interface between the N-terminal and two C-terminal helices with those side chains discussed in the text. This three-helix bundle represents a functional unit responsible for Na + binding and allowing access of the ion to the binding site. The view is normal to the external surface of the c-ring with the ring axis approximately vertical. The color coding of the subunits is the same as in Fig. 8.1 A. The C-terminal helices are shorter than the N-terminal helices, owing to a break at Tyr 80 followed by another short helix of one turn (Figs. 8.1 and 8.2). For each helix, an individual plane can be found that roughly contains the axis of the helix and the c-ring symmetry axis (Fig. 8.1 A). All helices show a bend of about 20 in the middle of the membrane (at Pro 28 and Glu 65 in the 79

80 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus N-terminal and the C-terminal helices, respectively), causing the narrow part of the hourglass shape. Moreover, the bend tilts the helices in the cytoplasmic half out of the plane by 10, yielding a right-handed twisted packing (Fig. 8.1 A). When the c-ring is viewed from the cytoplasm, it rotates counterclockwise during ATP synthesis [110] against the drag imposed by the F 1 motor components. Thus, the resulting torque might decrease the bend and increase the interhelical distance in the cytoplasmic part of the c-ring, depending on the energies involved. Such a conformational change under load might serve to store elastic energy in the c-ring, adding to that described for the central and peripheral stalk subunits [111]. A change in the twist of the helices is supported by calculations [112] that show in the lowest-order mode a torsional movement of the cytoplasmic against the periplasmic surface of the c-ring (Fig. 8.5). Vonck et al. [104] determined the position of the periplasmic and the cytoplasmic interfaces of the membrane surrounding the c-ring, which are approximately at Tyr 80 and Ser 55, respectively (Fig. 8.2). On the inner surface of the c-ring, near the N- and C-termini, the map shows an extended electron density that can be modeled by an alkyl chain of nine C atoms per c subunit. We propose that detergents are bound here at a position corresponding to that of the external leaflet of the membrane. The Phe 5 residues at the periplasmic end of the inner surface form a ring that possibly marks the position of the glycerol backbone of phospholipids. Such a positioning of lipids would correlate to the central plug feature seen in two-dimensional crystals of c-rings [113]. Despite the lack of direct evidence for a bilayer inside the c-ring, its existence can be inferred from the hydrophobic surface that extends beyond the electron density of the alkyl chains until Tyr 34 (Fig. 8.6 B). However, the internal bilayer appears to be wider and shifted toward the periplasmic surface with respect to its external counterpart. 8.4 Structure of the sodium ion binding site The c-ring was crystallized in buffer containing 100 mm sodium acetate, which promotes Na + binding as shown by established techniques [107]. The Na + ions are seen in the map as 11 distinct densities at the bend of the helices, 80

81 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus Figure 8.3: Electron density map (red, Na + omit map at 3.0σ; blue, 2F obs - F calc map at 1.4σ) and residues of the Na + binding site formed by two c subunits, A and B. Na + coordination and selected hydrogen bonds are indicated with dashed lines. Distances are given in Å. The blue sphere indicates the center of the bound Na + ion. The view is the same as in Fig close to the outer surface of the c-ring. Figure 8.2 shows a section of the c-ring with two C-terminal and one N-terminal helix forming a Na + -binding unit. The coordination sphere is formed by side-chain oxygens of Gln 32 (Oɛ1) 81

82 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus and Glu 65 (Oɛ2) of one subunit and the hydroxyl oxygen of Ser 66 and the backbone carbonyl oxygen of Val 63 of the neighboring subunit (Figs. 8.2 and 8.3). The distance between the liganding atoms and the ion is 2.37 ± 0.14 Å. The liganding residues are in good accordance with mutational studies, which recognized Gln 32, Glu 65, and Ser 66 as being essential for Na + binding [114], and their arrangement was confirmed as a Na + binding site using the algorithm of Nayal and Di Cera [115]. The crystal structure shows the bound ion surrounded by a network of hydrogen bonds: The Oɛ1 of Glu 65 accepts hydrogen bonds from the hydroxyl groups of Ser 66 and Tyr 70, and the Nɛ2 of Gln 32 donates a hydrogen bond to Oɛ2 of Glu 65 (Fig. 8.3). These hydrogen bonds serve to keep Glu 65 deprotonated at physiological ph [116] in order to allow Na + binding and to lock it into its ion-binding conformation. This arrangement of residues and their hydrogen bonds obviously serves to optimize the solvation energy [117] of the Na + ion and results in a locked conformation of the ion binding site. We propose that the present structure with the side chain of Tyr 70 facing outward and stabilizing the side-chain conformation of Glu 65 represents the conformation outside of the subunit a/c interface. Toward the periplasmic surface from the binding site, the structure forms a cavity (Fig. 8.6 B) to which the side chain of Tyr 70 could relocate to allow unloading and loading of the binding site to and from subunit a. This relocation may be part of an unlocking mechanism. The Na + binding signature is conserved in all known Na + -translocating ATP synthases and appears in the c subunits of other anaerobic bacteria, suggesting that they may indeed harbor a Na + -ATP synthase (Fig. 8.4). Notably, none of these amino acids are conserved in proton-translocating ATP synthases, except for the acidic residue at position 65 (I. tartaricus numbering, Fig. 8.4), which is implicated in proton binding. Despite these differences in sequence, protonand Na + -driven F 0 motors must share the same basic structure because of the demonstration that chimeric ATP synthases are functional [118]. 82

83 8 Structure of the Rotor Ring of F-Type Na+ -ATPase from Ilyobacter tartaricus Figure 8.4: Alignment of selected c subunit sequences. Amino acids are shaded with a threshold of 70% for identical and similar amino acids among the sequences shown. Orange: G, C; yellow: P; blue: A, V, L, I, M, F, W; green: S, T, N, Q; red: D, E, R, K; cyan: H and Y. Also shown are the total number of amino acids (T), the percentages of identical (I) and similar (S) amino acids (compared with IT), and the coupling ion (C) bound to the acidic residue marked by an asterisk. (Na+ ) marks putative Na+ -binding c subunits. I. tartaricus (IT), Propionigenium modestum (PM), Acetobacterium woodii c2/3 (AW), Fusobacterium nucleatum (FN), Ruminococcus albus (RA), Thermotoga maritima (TM), Mycoplasma genitalium (MG), Synechococcus elongatus (SE), Spinachia oleracea (SO), Mycobacterium tuberculosis (MT), Escherichia coli (EC). Secondary-structure elements and numbering according to IT are indicated above the sequences. 8.5 Ion translocation in F0 complexes. Previous models of the E. coli c-ring were based on monomeric structures of the c subunit determined by NMR at ph 5 [103] and ph 8 [99] in solutions using organic solvents and were assumed to represent the protonated and unproto- 83

84 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus nated structures outside and inside the subunit a/c interface, respectively. In these models, the proton binding Asp 61 (position 65 in I. tartaricus) of those c subunits, which are in contact with the lipids, faces toward the N-terminal helices. Within the subunit a/c interface, the C-terminal helix is proposed to swivel by 140 clockwise (as viewed from the cytoplasm), which relocates Asp 61 to become accessible to subunit a [119]. The latter conformation is supported by cysteine cross-linking in F 0 complexes. These experiments indicate that residues at the same helical face as the proton binding site (Ile 55, Ala 62, Met 65, Gly 69, Leu 72, and Tyr 73 ) are accessible from subunit a and thus must be located on the surface of the c-ring [120]. In the structure presented here, the placement of the Na + binding site near the surface of the c-ring between two C-terminal helices allows for ion transfer to and from subunit a after side-chain movements and does not require large rearrangements of the protein backbone. If the structure of the decameric E. coli c-ring is modeled according to that of I. tartaricus, all residue positions that formed cross-links with subunit a are exposed to the surface (Fig. 8.7). We therefore presume that the tertiary fold of the E. coli c monomer in the NMR experiments does not match that in the oligomeric c-ring, and we note that the compact structure of the c-ring would impose severe sterical restraints against the proposed swiveling. In addition, the known specific reactivity of the ion-binding carboxyl group with the bulky dicyclohexylcarbodiimide (DCCD) could not be explained if the former is occluded between N- and C-terminal helices. With this specificity in mind, we inspected the c-ring surface near Glu 65, looking for a DCCDaccessible binding site. A pocket sufficiently large to accommodate binding of bulky organic molecules extends from Glu 65 to Val 58 at the ring surface within the interface between an N-terminal and two C-terminal helices (Fig. 8.6 A). The pocket is lined by small hydrophobic residues and is open to the membrane from where the hydrophobic DCCD molecule is thought to bind. The pocket might also contribute a binding site for other hydrophobic organic molecules such as the new class of diarylquinoline antibiotics against tuberculosis that target subunit c of Mycobacterium tuberculosis [121]. To allow for unloading and loading of the ion binding site, the locked conformation must be converted to an open one during passage of the site through 84

85 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus the subunit a/c interface. This interconversion may be enabled by modulating the electrostatic interactions of the binding site with the universally conserved Arg 227 of subunit a [122] as the site passes this positively charged residue during rotation (Fig. 8.8). The path of the ion leading from the binding site into the cytoplasm is presently unknown. Two possibilities are discussed: a channel through subunit a [102] or through the c-ring itself [96]. In accord with the latter possibility, biochemical data suggest that the Na + exit path is built intrinsically into the c-ring [107]. In the present c-ring structure, a channel leading from the binding site to the cytoplasmic surface could not be identified. Similarly, a channel-like pore is not visible in the structure of the Ca 2+ -ATPase (adenosine triphosphatase), which operates at a rate comparable to that of the ATP synthase [123], and in myoglobin [124]. In contrast to these proteins with a low rate of ion flux, proteins with a structurally evident channel (e.g., potassium channel) have diffusion-limited transport rates up to 10 8 ions per second [125], six orders of magnitude faster than the ATP synthase. However, dynamic fluctuations of the protein may open transient pathways for Na + conduction so that a wider pore is not evident in the structure. Taken together, the structural features of ion binding in the membraneembedded c-ring have profound implications for loading and unloading of the binding site because they represent stringent restraints for possible explanations of the F 0 motor function. 8.6 Supporting Online Material Materials and Methods ATP synthase was isolated from wild type Ilyobacter tartaricus cells. After disrupting the complex with N-lauroylsarcosine, pure c-ring was obtained by precipication of all other subunits of the enzyme with ammonium sulfate [107]. The purified c-ring was dialyzed against 10 mm Tris/HCl buffer (ph 8.0) to precipitate the protein by removal of the detergent. The material was collected by centrifugation and dissolved at 10 mg protein/ml in dialysis buffer containing 0.02% (w/v) NaN 3 and 0.78% (w/v) Zwittergent 3-12 (Calbiochem, 85

86 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus Figure 8.5: Motional flexibility within the c-ring. The low frequency normal modes of the c-ring were analyzed according to [112]. The data presented show two different views of the c-ring from I. tartaricus ATP synthase in the lowest-order nontrivial normal mode. (This image was modified due to the fact, that motion is not visible on paper. The arrows indicate the direction of movement; black stands for the cytoplasmic and light grey for the periplasmic endings, the dashed line specifies the hinge region.) La Jolla, CA, USA). Crystals were grown within 2-3 days at 17 C to a size of approximately µm 3 using the vapor diffusion method in hanging drops. A drop consisted of 1 µl protein solution (passed through a 0.22 µm sterile filter) and 1 µl crystallization buffer containing 15% (v/v) polyethylene glycol 400 in 0.1 M sodium acetate buffer, ph 4.5. The crystals were flash-frozen in liquid nitrogen after an equilibration for 24 h with 30% (v/v) polyethylene glycol 400 as cryo-protectant in the reservoir buffer at 17 C. Data used for structure solution were collected at the beamline X06SA of the Swiss Light Source (SLS, Switzerland) and the structure was refined against 2.4-Å data obtained at beamline ID29 of ESRF (France). 86

87 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus Figure 8.6: Van der Waals surface of the c-ring. The surface is colored lightgrey for apolar, yellow for polar, red for acidic and blue for basic surface exposed residues. A: The outer ring surface is shown. A putative DCCD-binding pocket is indicated by an arrow. B: The proximal half of the ring is removed to view the inner surface and some of the helices are shown in ribbon representation for clarity. Dashed lines near Phe 5 and Tyr 34 mark the putative positions of polar/apolar interfaces of the inner lipid bilayer. The cavity below the binding site is marked by an arrow. 87

88 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus Figure 8.7: Ribbon model of the E. coli c 10 oligomer obtained by homology modeling [126] according to the I. tartaricus c 11 structure. The residues which form cross-links with subunit a after cysteine substitution and the proton binding Asp 61 are also shown. 88

89 8 Structure of the Rotor Ring of F-Type Na + -ATPase from Ilyobacter tartaricus Figure 8.8: Schematic model for the interconversion of the binding site in the subunit a/c interface from an alternately locked conformation to an open one. The view is from the cytoplasmic side on the section of the interface at the level of the binding sites. The side chain of Glu 65 locks the Na + ion (yellow circle) in the binding site by its orientation approximately horizontal to the membrane plane (Fig. 8.7). We propose that a site entering the interface as indicated by the solid arrow encounters electrostatic interactions upon approaching Arg 227 in two ways: the Na + is repelled into the cytoplasmic outlet path (i.e. in a plane perpendicular to the membrane) and the Glu 65 is attracted, maintaining approximately its original conformation. Please note that the Glu 65 is on the distal helix with respect to the Arg 227 of the functional unit if rotation proceeds in the ATP synthesis direction. After rotation of the binding site to the opposite side of Arg 227, Glu 65 is pulled backwards towards the arginine thus opening the gate to the periplasmic inlet channel. Na + ions can now pass from subunit a to subunit c (dashed arrow) and the locked conformation of the binding site reforms when the site moves on and the electrostatic attraction between Arg 227 and Glu 65 is attenuated. The reverse order of events takes place in the ATP hydrolysis direction. 89

90 9 GalU from Escherichia coli The nucleotide-sugar UTP-α-D-glucose is a key metabolite in procaryotes. UDP-glucose plays a central role in the synthesis of the cell envelope of E. coli and is the glycosyl donor during the biosynthesis of lipopolysaccharides [127, 128], membrane-derived oligosaccharides (MDO) [129], and capsular polysaccharides [130]. It is also necessary in galactose and trehalose metabolism. UDP-α-D-glucose is an essential intermediate for growth on trehalose [131], the synthesis of trehalose [132] and growth on galactose [133]. In addition, it interferes with the expression of σ s during steady state growth [134]. UDP-α-D-glucose is produced enzymatically by the glucose-1-phosphate uri- Figure 9.1: Ribbon diagram of the search model G1P-TT in blue. The red part was cut out before molecular replacement attempts. 90

91 9 GalU from Escherichia coli Figure 9.2: Ribbon diagram of the GalU tetramer. dylyltransferase (GalU). This enzyme consists of 301 amino acid residues and catalyses the reaction of α-d-glucose-1-phosphate and UTP to UDP-glucose and pyrophosphate, according to the following scheme: UTP + glucose-1-phosphate UDP-glucose + pyrophosphate. The overexpression system was constructed by Reinhold Horlacher (AG Boos, University of Konstanz). Crystallization and preliminary x-ray structure analysis were performed by Joachim Diez (AG Welte/Diederichs, University of Konstanz) [135]. He collected data sets of native crystals and various derivatives down to a resolution of 1.7 Å but structure solution with experimental phases failed. 91

92 9 GalU from Escherichia coli Figure 9.3: Superposition of GalU (in red) and G1P-TT (in blue). The bound sugar molecule (in ball-and-stick representation) belongs to the structure of G1P-TT. A BLAST search 1 through the PDB-database allowed us to recognize enzymes with known structures and similar sequences. They are all nucleotidyltransferases, but none of them uses UTP as nucleotidyldonor: glucose-1- phosphate thymidylyltransferase (G1P-TT; PDB-code 1H5R) [136] from E. coli, glucose-1-phosphate thymidylyltransferase (RmlA) [137] from Pseudomonas aeruginosa, glucose-1-phosphate thymidylyltransferase (Ep) [138] from Salmonella eneterica L2 the phosphocholine cytidylyltransferase (LicC) [139] from Streptococcus Pneumoniae. Like GalU, all these are active as homotetramers. During this PhD-thesis, the structure of GalU was determined in cooperation with Alexei Vagin using his program MOLREP [50]. The search model for molecular replacement was a truncated model of the glucose-1-phosphate thymidylyltransferase from E. coli [136] missing the 44 C-terminal residues following residue 247. This structural fragment was omitted because it consists of three helices which protrude out of the globular domain of G1P-TT as can be seen in figure 9.1. After an initial rotation search with MOLREP

93 9 GalU from Escherichia coli Figure 9.4: Superposition of the active sites of GalU (in red) and G1P-TT (in blue). The substrate (in ball-and-stick representation) belongs to the structure of G1P-TT. [50], 50 peaks were used for a subsequent dyad search [140] which also incorporated informations from a previously performed calculation of the self-rotation function. The dyad search lead to a dimer in which two properly rotated monomers were oriented correctly with respect to each other. Subsequently this dimer was used as a search model in a conventional MOLREP [50] search yealding one tetramer, as expected. Model building was done with several automated, semi-automated and manual methods. Automated model building was performed with the programs ARP/wARP [141] and RESOLVE [142, 74]. Semiautomated and manual model building was done with O [53] and COOT [76]. Refinement was performed with REFMAC5 [51] (Collaborative Computational Project Number 4, 1994 [52]). The structure of GalU was solved at 1.7 Å resolution despite a very high 93

94 9 GalU from Escherichia coli Figure 9.5: Superposition of the B-site of G1P-TT (in blue and light blue) with bound substrate (in ball-and-stick representation) and the corresponding area in one monomer of GalU (in red). The putative B-site in GalU is blocked completely by the residues root mean square deviation of 1.8 Å (for 222 Cα-atoms in a cutoff range of 4.5 Å (calculated with LSQMAN [143])) and a low sequence identity of only 24% compared with the search model G1P-TT. It was refined to a final R- factor of 19% (R free 24%). Figure 9.2 gives an overview of the GalU tetramer and figure 9.3 shows the superposition of one monomer of GalU with one of G1P-TT. A comparison of the active sites of both structures (figure 9.4) suggests the same catalytic mechanism for GalU as predicted for G1P-TT [136], despite the absence of substrate in the structure of GalU. Many nucleotidylyltransferases possess a so called secondary binding site (B-site) whose function is not yet understood. This secondary site is missing in GalU due to a different mode of tetramerisation (figure 9.5). The B-site in G1P-TT is formed in the interface of two monomers. Due to the different tetramerization, this interface is missing in GalU. Furthermore, the loop from residue 140 to 146 blocks the putative B-site completely. The results of this work are yet unpublished. In this short overview my contribution to the project is outlined. 94