Functional and Structural Studies on Trimeric Autotransporters

Jack C. Leo

Research Program in Structural Biology and Biophysics Institute of Biotechnology, University of Helsinki

Genetics

Department of Biological and Environmental Sciences Faculty of Biosciences, University of Helsinki

and

National Graduate School in Informational and Structural Biology

Åbo Akademi University

ACADEMIC DISSERTATION

To be presented, with the permission of the Faculty of Biosciences of the University of Helsinki, for public examination in Lecture Hall 5, University main building

(Fabianinkatu 33), on 20 November, at 12 noon.

Helsinki 2009

2

Supervised by Research Director, Professor Adrian Goldman Structural Biology and Biophysics Institute of Biotechnology University of Helsinki Finland Reviewed by Susan Buchanan, Ph.D. Structural Biology & Cell Signalling Section National Institute of Diabetes and Digestive and Kidney Diseases Bethesda, Maryland United States of America and Professor Sarah Butcher Structural Biology and Biophysics Institute of Biotechnology University of Helsinki Finland

Opponent Assistant Professor Peter van Ulsen Department of Molecular Microbiology Faculty of Earth and Life Sciences VU University Amsterdam The Netherlands

Custos Professor Tapio Palva Genetics Department of Biological and Environmental Sciences Faculty of Biosciences University of Helsinki Finland ISBN 978-952-10-5791-5 (pbk.) ISBN 978-952-10-5792-2 (PDF) ISSN 1795-7079 Helsinki University Press Helsinki 2009

3

Science means looking – a special kind of looking. Looking especially hard at the

things you didn’t understand. Looking at the stars, say, and not fearing them, not worshipping them, just asking questions, finding the questions that would unlock the door to the next question and the question beyond that.

-Robert Charles Wilson

4

Contents

Abbreviations 6

Amino acids 7

List of original publications 8

Abstract 9

1. Introduction 10

2. Review of the literature 12 2.1 Secretion of proteins in bacteria: an overview 12

2.1.1 Secretion in Gram-negative bacteria 12 2.1.1 Type I secretion 13 2.1.2 Type II secretion 13 2.1.3 Type III secretion 15 2.1.4 Type IV secretion 15 2.1.5 Type VI secretion 16 2.1.6 Other secretion pathways in Gram-negative bacteria 16

2.1.6 Secretion in Gram-positive bacteria 17 2.2 Type V secretion: autotransportation and two-partner secretion 18

2.2.1 Classical autotransporters 18 2.2.1.1 Structure of autotransporters 22 2.2.1.2 Mechanism of autotransportation 25

2.2.2 Two-partner secretion 28 2.2.3 Trimeric autotransporters 31

2.3.2.1 Structure of trimeric autotransporters 32 2.2.3.3 Mechanism of trimeric autotransportation 37

2.3 Biology of individual trimeric autotransporters 39 2.3.1 The Yersinia adhesin YadA 39

2.3.1.1 Enteropathogenic yersiniae 39 2.3.2.2 YadA and ECM binding 41 2.3.2.3 Other virulence-associated functions of YadA 42 2.3.2.4 Mapping of YadA activities 44

2.3.2 Other ECM-binding trimeric autotransporters 45 2.3.2.1 Trimeric autotransporters from the genus Haemophilus 46 2.3.2.2 The ubiquitous surface proteins of Moraxella catarrhalis 47 2.3.2.3 Trimeric autotransporters from the genus Bartonella 47 2.3.2.4 ECM-binding trimeric autotransporters from other pathogens 48

2.3.3 Immunoglobulin-binding trimeric autotransporters 49 2.3.3.1 The Eib proteins from E. coli 49 2.3.3.2 The Moraxella IgD-binding protein 50

2.3.4 Other trimeric autotransporters 51 2.3.4.1 Trimeric autotransporters in plant pathogens 51 2.3.4.2 BimA: a trimeric autotransporter for intracellular motility 51

5

3. Aims of the study 53

4. Materials and methods 54 4.1 Bacterial strains 54 4.2 DNA techniques 54

4.2.1 Cloning of expression constructs 54 4.2.2 Site-directed mutagenesis 55

4.3 Recombinant protein expression and purification 55 4.3.1 Protein expression 55 4.3.2 Protein purification 55 4.3.3 Cell lysate preparation 56

4.4 Collagens, collagen-like peptides and immunoglobulins 56 4.4.1 Collagens and collagen-mimicking peptides 56 4.4.2 Antibodies 57

4.5 Binding studies 57 4.5.1 ELISA 57 4.5.3 Immunoblots 58 4.5.4 SPR measurements 58 4.5.4 Isothermal titration calorimetry 59 4.5.5 Gel shift assays 59

4.6 Crystallography 59 4.6.1 Protein crystallisation 59 4.6.2 Diffraction data collection, data processing and structure solution 60

5. Results and discussion 61 5.1. The YadA-collagen interaction 61

5.1.1 A triple-helical conformation is necessary for YadA binding 61 5.1.2 YadA binding determinants in fibrillar collagens 63

5.1.2.1 Collagens do not contain a single binding site for YadA 63 5.1.2.2 YadA binds to collagen regions with many hydroxyprolines and few charges 65

5.1.3 Crystallisation of YadA in complex with a collagen-mimicking peptide 66 5.2 The Eib proteins are receptors for IgG Fc 67 5.3 Crystallisation of trimeric autotransporter fragments 68

5.3.1 Crystallisation of Eib passenger domains 68 5.3.2 Crystallisation of the translocation unit of SadA 69

5.4 The crystal structure of the EibD passenger domain fragment 71 5.4.1 Overview of the structure 71 5.4.2 Mapping of the Ig-binding regions 71

6. Conclusions and future perspectives 74

Acknowledgements 77

References 79

6

Abbreviations

ABC Adenosine triphosphate-binding cassette ATP Adenosine triphosphate BSA Bovine serum albumin CEACAM-1 Carcinoembryonic antigen-related cell adhesion molecule 1 DNA Deoxyribonucleic acid ECL Enhanced chemiluminescence ECM Extracellular matrix ELISA Enzyme-linked immunosorbent assay ExPEC Extraintestinal pathogenic Escherichia coli Fab Fragment antigen-binding (of an immunoglobulin) Fc Fragment crystallisable (of an immunoglobulin) Ig Immunoglobulin IM Inner membrane IL-8 Interleukin-8 IPTG Isopropyl thiogalactoside ITC Isothermal titration calorimetry HRP Horse radish peroxidase kDa Kilodalton LAIR Leukocyte-associated Ig-like receptor MFP Membrane fusion protein OM Outer membrane OMP Outer membran protein PCR Polymerase chain reaction PBS Phosphate-buffered saline PDB Protein Data Bank (www.rcsb.org/pdb) PMN Polymorphonuclear leukocyte POTRA Polypeptide transport-associated Prn Pertactin RT Room temperature SAAT Self-associating autotransporter SDS-PAGE Sodium dodecyl sulfate polyacrylamide gel electrophoresis SEM Standard error of mean SPATE Serine protease autotransporter of Enterobacteriacaea SP Signal peptide SPR Surface plasmon resonance SRP Signal recognition particle STEC Shiga-toxin producing Escherichia coli Tat Twin arginine translocation TB Terrific broth TBS Tris-buffered saline TPS Two-partner secretion YeYadA Yersinia enterocolitica YadA Yop Yersinia outer protein YpYadA Y. pseudotuberculosis YadA Å Ångström

7

Amino acids

A Ala Alanine C Cys Cysteine D Asp Aspartic acid E Glu Glutamic acid F Phe Phenylalanine G Gly Glycine H His Histidine I Ile Isoleucine K Lys Lysine L Leu Leucine M Met Methionine N Asn Asparagine O Hyp 4-Hydroxyproline P Pro Proline Q Gln Glutamine R Arg Arginine S Ser Serine T Thr Threonine W Trp Tryptophan V Val Valine Y Tyr Tyrosine

8

List of original publications

This thesis is based on the following publications: I Leo JC, Elovaara H, Brodsky B, Skurnik M, Goldman A (2008): The

Yersinia adhesin YadA binds to a collagenous triple-helical conformation but without sequence specificity. Protein Engineering, Design and Selection 21: 475-484.

II Leo JC, Elovaara H, Bihan D, Kilpinen SK, Raynal N, Farndale RW, Goldman A (2009): The Yersinia adhesin YadA binds to collagenous sequences with a high hydroxyproline-to-charged residue ratio. Submitted.

III Leo JC & Goldman A (2009): The immunoglobulin-binding Eib proteins from Escherichia coli are receptors for IgG Fc. Molecular Immunology 46: 1860-1866

IV Leo JC*, Lyskowski* A, Goldman A (2009): Crystal structure of an IgA- and IgG-binding fragment of the trimeric autotransporter EibD. Manuscript.

*equal contribution These publications are referred to in the text by their Roman numerals (I-IV). Additional publications: Nummelin H, Merckel MC, Leo JC, Lankinen H, Skurnik M, Goldman A

(2004): The Yersinia adhesin YadA collagen-binding domain structure is a novel left-handed parallel β-roll. The EMBO Journal 23: 701-711.

Leo JC & Goldman A (inventors): Use of Escherichia coli Immunoglobulin-

Binding Proteins in Research and Biotechnology Applications. Patent application no. 20095041.

9

Abstract

Trimeric autotransporters are a family of secreted outer membrane proteins in Gram-negative bacteria. These obligate homotrimeric proteins share a conserved C-terminal region, termed the translocation unit. This domain consists of an integral membrane β-barrel anchor and associated α-helices which pass through the pore of the barrel. The α-helices link to the extracellular portion of the protein, the passenger domain. Autotransportation refers to the way in which the passenger domain is secreted into the extracellular space. It appears that the translocation unit mediates the transport of the passenger domain across the outer membrane, and no external factors, such as ATP, ion gradients nor other proteins, are required.

The passenger domain of autotransporters contains the specific activities of each protein. These are usually related to virulence. In trimeric autotransporters, the main function of the proteins is to act as adhesins. One such protein is the Yersinia adhesin YadA, found in enteropathogenic species of Yersinia. The main activity of YadA from Y. enterocolitica is to bind collagen, and it also mediates adhesion to other molecules of the extracellular matrix. In addition, YadA is involved in serum resistance, phagocytosis resistance, binding to epithelial cells and autoagglutination. YadA is an essential virulence factor of Y. enterocolitica, and removal of this protein from the bacteria leads to avirulence.

In this study, I investigated the YadA-collagen interaction by studying the binding of YadA to collagen-mimicking peptides by several biochemical and biophysical methods. YadA bound as tightly to the triple-helical model peptide (Pro-Hyp-Gly)10 as to native collagen type I. However, YadA failed to bind a similar peptide that does not form a collagenous triple helix. As (Pro-Hyp-Gly)10 does not contain a specific sequence, we concluded that a triple-helical conformation is necessary for YadA binding, but no specific sequence is required.

To further investigate binding determinants for YadA in collagens, I examined the binding of YadA to a library of collagen-mimicking peptides that span the entire triple-helical sequences of human collagens type II and type III. YadA bound promiscuously to many but not all peptides, indicating that a triple-helical conformation alone is not sufficient for binding. The high-binding peptides did not share a clear binding motif, but these peptides were rich in hydroxyproline residues and contained a low number of charged residues. YadA thus binds collagens without sequence specificity. This strategy of promiscuous binding may be advantageous for pathogenic bacteria.

The Eib proteins from Escherichia coli are immunoglobulin (Ig)-binding homologues of YadA. I showed conclusively that recombinant EibA, EibC, EibD and EibF bind to IgG Fc. I crystallised a fragment of the passenger domain of EibD, which binds IgA in addition to IgG. The structure has a YadA-like head domain and an extended coiled-coil stalk. The top half of the coiled-coil is right-handed with hendecad periodicity, whereas the lower half is a canonical left-handed coiled-coil. At the transition from right- to left-handedness, a small β-sheet protrudes from each monomer. I was able to map the binding regions for IgG and IgA using truncations and site-directed mutagenesis to the coiled-coil stalk and identified residues critical for Ig binding.

10

1. Introduction

One of the most evident features of bacteria in everyday life is their ability to cause disease. Though the vast majority of our microbiota is harmless and in fact many species are beneficial and even necessary for human health, most people think of bacteria first and foremost as agents of disease. Indeed, infectious diseases have been a scourge of humanity and have shaped the course of human history. Classic examples are the bubonic plague pandemics, particularly the Black Death of 14th century Europe, where the disease wiped out approximately 30% of the population (Perry & Fetherston, 1997). In developed countries, the introduction of antibiotics in the early 20th century has meant the relegation of bacterial infections from the leading cause of mortality to a subsidiary role. However, in Western countries it is easy to forget that infectious diseases are still a major cause of death globally. Recently, the emergence of antibiotic-resistant strains and the increasing threat of bioterrorism have called attention to the resurgence of what were considered largely defeated diseases. Research efforts are thus being directed at finding new antimicrobial compounds to combat these pathogenic bacteria.

During infection, disease-causing bacteria secrete a variety of proteins that interact with the host organism to establish or maintain the pathogen in host tissues. These extracellular proteins are collectively called virulence factors, due to their function in promoting infection. Virulence factors may be enzymes, such as proteases that break down host proteins, toxins that cause damage to the host by interfering with some vital functions, or adhesins that mediate the attachment of the pathogen to host cells and tissues. Virulence factors are often essential for the pathogenesis of disease-causing bacteria and so make attractive targets for specific antimicrobial drugs. For drug development, understanding how virulence factors are secreted and how they interact with the host are of central importance. Elucidation of the atomic structures of virulence factors, the proteins of the secretion pathways and the target molecules of the host gives valuable information about the mechanisms of secretion and the function of virulence factors. This would permit the rational design of inhibitor molecules that could either obstruct the interaction of a virulence factor with its target in the host or prevent its secretion from the bacterial cell.

Trimeric autotransporters comprise a group of virulence-related proteins in Gram-negative bacteria. These obligate homotrimeric proteins are found in the outer membrane (OM), where their main function is to act as adhesins. In addition to their primary adhesive activity, many trimeric autotransporters have secondary functions such as autoagglutination and protecting the bacterium from host immune responses (reviewed in Linke et al., 2006). Trimeric autotransporters share a highly conserved C-terminal region, the translocation unit. In addition to acting as an integral membrane anchor, the translocation unit translocates the N-terminal portion, the “passenger” domain, of the protein into the extracellular space. This translocation event appears to be independent of any auxiliary factors such as adenosine triphosphate (ATP), ion gradients or even other proteins; hence the name autotransporter (Henderson et al., 2004). As these proteins are involved in virulence, and as they all seem to be translocated in a highly similar manner,

11

the C-terminal translocation unit is a potential target for drugs that would inhibit the translocation across the outer membrane.

The prototypical trimeric autotransporter is the Yersinia adhesin YadA, found in two human enteropathogenic species, Y. enterocolitica and Y. pseudotuberculosis. I have investigated the binding of the YadA from Y. enterocolitica to its major ligand, collagen (Studies I & II). Collagens are ubiquitous molecules of the extracellular matrix (ECM) and give tissues form, strength and rigidity (for a review, see Gelse et al., 2003). Collagens have a distinctive triple-helical structure, which arises from the repeated three-amino acid pattern of G-X-X'. The glycine at position one is absolutely required for the formation of the collagenous triple-helical conformation, and the positions X and X' are usually occupied by the imino acids proline and 4-hydroxyproline, respectively. The large size and heterogeneity of collagens makes working with native collagens difficult. To investigate the YadA-collagen interaction, I used collagen-mimicking peptides, which also adopt a triple-helical structure, but which are more amenable than native collagen to biochemical and biophysical techniques (Study I). To further determine the binding determinants for YadA in collagen, I studied the adherence of YadA to a library of collagenous peptides spanning the entire sequences of human collagens type II and type III, the collagen ‘Toolkits’ from Richard Farndale at the University of Cambridge (Study II).

The Eib proteins from Escherichia coli are another group of trimeric autotransporters that are structurally related to YadA. However, rather than binding to ECM molecules, these proteins bind immunoglobulins (Igs), antibody molecules which are of central importance in immune responses. Coating the bacterial cell with host-derived Igs presumably helps to protect the bacterium from immune attack. I investigated the binding of four recombinantly produced Eibs to two classes of Igs, IgG and IgA (Study III). In addition, I crystallised an Ig-binding fragment of the passenger domain of one member of the family, EibD, and, together with Andrzej Lyskowski, solved its structure (Study IV). I further investigated the mechanism of Ig binding by site-directed mutagenesis (Study IV).

12

2. Review of the literature

2.1 Secretion of proteins in bacteria: an overview

To interact with other cells and the environment, cells must secrete proteins into the extracellular space, either as soluble molecules or membrane-bound proteins. Bacteria are no exception, and pathogenic bacteria especially secrete a wide variety of proteins that are required for virulence-related functions such as adhesion, invasion, toxicity or protection from immune responses.

In this section, I will give a brief overview of bacterial protein secretion. The emphasis will be on Gram-negative bacteria, as this thesis deals with a class of secreted proteins found in these organisms. However, I will also give a short summary of secretion in Gram-positive bacteria for comparative purposes.

2.1.1 Secretion in Gram-negative bacteria

In contrast to eukaryotic cells and Gram-positive bacteria, secreted molecules have to cross a second membrane in Gram-negative bacteria. The OM of Gram-negative bacteria acts as an additional protective barrier, increasing resistance to e.g. antibiotics and detergents  (Bos et al., 2007). In addition to mechanistic difficulties involved in translocating proteins across two membranes, the space between the inner and outer membranes, the periplasm, contains no immediate sources of energy. The periplasm does not contain the energy-donating molecule ATP, which is usually used to drive energetically unfavourable biological reactions  (Thanassi et al., 2005). Furthermore, the permeability of the OM to small molecules below approximately 700 Da means that ion gradients cannot be maintained across it, making secondary active transport impossible. In spite of, or perhaps because of, these challenges, Gram-negative bacteria have evolved no less than six different protein secretion mechanisms, all of which are implicated in some virulence-related functions (Tseng et al., 2009).

The six secretion mechanisms can be classed into two groups, depending on whether they use the Sec translocation machinery to cross the inner membrane. The Sec translocase consists of the SecA ATPase and the SecYEG translocation channel (Natale et al., 2008). As the majority of periplasmic and secreted proteins traverse the IM via the Sec pathway, Sec-mediated secretion has also been called the general secretion pathway. Sec-dependent proteins can be easily identified by the presence of an N-terminal signal peptide, which is usually 20-25 amino acids in length (Fekkes & Driessen, 1999). Proteins containing this signal are delivered to the translocase by either the chaperone SecB or the signal recognition particle (SRP)  (Natale et al., 2008). Sec-independent secretion systems have other, specific signals.

In the following text, I give a short introduction to five of the six Gram-negative secretion systems, and additionally I describe shortly some alternative secretion pathways.

13

Type V secretion, which consists of autotransportation and the two-partner secretion system, will be covered in detail in section 2.2.

2.1.1 Type I secretion

Type I secretion is a Sec-independent system for protein secretion. It is a one-step system; there is no intermediate periplasmic step for the secreted proteins. The type I secretion machinery is fairly simple, consisting of only three protein components (Figure 2.1): an ATP-binding cassette (ABC) protein at the inner membrane, a membrane fusion protein (MFP) and an outer membrane protein (OMP) (Holland et al., 2005). The ABC protein acts as the energiser of the system, using the free energy released by the hydrolysis of cytosolic ATP to transport proteins across the cell envelope. ABC transporters are an evolutionarily conserved family of proteins and can be found in all three domains of life (Moussatova et al., 2008). The MFP and the OMP form a channel through the periplasm and outer membrane to allow secretion. The secretion signal of substrate proteins is located in the 50 C-terminal residues of the protein (Holland et al., 2005). However, the exact nature of the signal remains elusive. The prototype for type I secretion is the α-haemolysin of extraintestinal pathogenic Escherichia coli (ExPEC) (Gentschev et al., 2002). However, many other toxins, proteases, lipases and surface proteins are exported by the type I machinery. Similar ABC protein-containing systems are used by cells in transporting a number of different molecules, including peptides and carbohydrates, and for excreting ions and drugs (Moussatova et al., 2008). In the α-haemolysin system, the membrane-integral ABC transporter HlyB interacts with HlyD, the MFP. The OMP component is TolC, an integral OM β-barrel protein with a large periplasmic domain formed by an α-helical barrel, which is involved in several export pathways (Buchanan, 2001). During secretion, TolC and HlyD form the lumen of the channel through which the substrate protein passes, probably in an unfolded conformation (Holland et al., 2005).

2.1.2 Type II secretion

The type II secretion system is a prevalent but not ubiquitous secretion pathway that, in contrast to type I secretion, is a two-step process (Figure 2.1). It has also been called the main terminal branch of the general secretion pathway (Sandkvist, 2001). The prototype for type II secretion is pullulanase, a starch-hydrolysing lipoprotein, from Klebsiella oxytoca, but a number of hydrolytic enzymes and toxins are also exported by this secretion apparatus (Johnson et al., 2006). Type II secretion is a Sec-dependent system, though some proteins using the type II pathway can transit the IM by an alternative system (discussed in 2.1.6). The proteins thus form distinct periplasmic intermediates, which have had their N-terminal signal peptides removed and which are partially or wholly folded before export across the OM (Sandkvist, 2001). Transport across the OM requires 12-15 accessory proteins, collectively referred to as the secreton (Sandkvist, 2001). The OM components of the secreton belong to the secretin family of proteins that form a 12-14 subunit ring-like structure, which is the OM secretion channel (Johnson et al., 2006). The

14

other proteins in the secreton anchor the pore to the IM and energise secretion, and include OM proteins, integral IM proteins and a cytoplasmic ATPase (Johnson et al., 2006). The periplasm is spanned by a pseudopilus structure, which is required for OM transport. The pseudopilus possibly acts as a piston that drives the secretion through the secretin pore (Sandkvist, 2001). The signal that sorts proteins to be secreted from those resident in the periplasm is still not understood, as the various proteins have little primary sequence similarity (Sandkvist, 2001). Type IV pilus assembly bears many similarities to type II secretion, and this suggests they have a common evolutionary origin (Hansen & Forest, 2006).

Figure 2.1 Schematic of major secretion systems (types I-IV and type VI) in Gram-negative bacteria. In type I secretion, substrate proteins are transported in a single step across the cell envelope. In type II secretion, substrates are first translocated across the OM by the Sec machinery, followed by translocation across the OM by the secreton, a complex of 12-15 proteins. In type III secretion, effector proteins are delivered directly into the cytoplasm of target cells by the injectisome. Effector proteins and in some cases DNA can also be secreted into target cells by type IV secretion, also known as adaptive conjugation. Type VI secretion is a third mechanism for injecting proteins into target cells. For details, see text. Figure adapted from Henderson et al. (2004) and Tseng et al. (2009).

15

2.1.3 Type III secretion

The type III secretion system is a Sec-independent mechanism for introducing effector proteins directly into the cytoplasm of host cells. Type III secretion systems are widespread among pathogenic bacteria that infect both animals and plants (He et al., 2004). As this secretion system is heavily involved in virulence, it has been the subject of intense study in recent years. However, recent studies have shown that some mutualists also have type III secretion systems (Coombes, 2009). The prototype for this system is the Yersinia type III secretion system, which is responsible for injecting anti-phagocytic and anti-inflammatory effector proteins, the Yersinia outer proteins (Yops), into immune cells (Cornelis, 2002). The type III system spans the cell envelope and transports effectors across three biomembranes: the two bacterial membranes and the host plasma membrane. The machinery resembles a syringe (Figure 2.1), hence its name, the injectisome. The basal body contains two ring structures, one in both the IM and OM, which interact across the periplasm via a rod-like structure to form a closed channel (Cornelis, 2006). As in type II secretion, proteins of the secretin family form the OM rings. An ATPase is associated with the basal ring and supplies the energy for translocation (Cornelis, 2006). It is attractive to assume that effectors are transported through the prominent extracellular needle and injected from its tip into the target cell cytoplasm, but this is still under debate. Accessory translocator proteins are needed to form a pore in the target cell membrane (Galan & Wolf-Watz, 2006). The signal for type III secretion is not entirely clear, but it seems to be located in the N-terminal 20-30 amino acids. Alternatively, in some cases the secretion signal may be in the mRNA molecule (Anderson & Schneewind, 1997). In addition, certain committed chaperones may be responsible for delivering effector proteins to the type III secretion system (Galan & Wolf-Watz, 2006). The type III secretion system is highly similar to the biogenesis of bacterial flagella, suggesting the systems have a common ancestor (Journet et al., 2005).

2.1.4 Type IV secretion

Like type III secretion, type IV secretion (also called ‘adaptive conjugation’) delivers secreted proteins directly in the cytoplasm of target cells. In addition to transporting proteins, type IV secretion is capable of mobilising DNA. Indeed, bacterial conjugation machineries are now considered a subtype of type IV secretion (Juhas et al., 2008). The prototype of this secretion pathway is the VirB/D4 T-DNA transport system of Agrobacterium tumefasciens, which transports oncogenic nucleoprotein complexes into plant cells to promote the formation of crown gall disease (Escobar & Dandekar, 2003). Other pathogens with type IV secretion systems include Bordetella pertussis, Legionella pneumophila, Bartonella henselae and Helicobacter pylori (Backert & Meyer, 2006). Like the injectisome, the type IV machinery spans the cell envelope. Transfer of unfolded protein substrates (or of single-stranded DNA) is energised by nucleoside triphosphatases at the IM. A number of IM, periplasmic and OM-associated proteins form the type IV

16

secretion machinery. A channel structure spans the periplasm and the OM. The contact to target cells is mediated by a pilus structure (Christie et al., 2005). Target proteins are recognised by a C-terminal signal determined by an arginine cluster, but other internal motifs may also be required for translocation (Christie et al., 2005). In most cases of type IV secretion, IM translocation is Sec-independent. However, the B. pertussis system is an exception where the secretion of pertussis toxin is a two-step process: the toxin first crosses the IM via the Sec machinery, followed by OM translocation by the type IV system (Burns, 2003).

2.1.5 Type VI secretion

Recently, a new secretion system was described in Gram-negative bacteria. This novel secretion apparatus was termed type VI secretion, and it is now known to be widely distributed among Gram-negative bacteria, in both pathogenic and mutualistic species (Bingle et al., 2008). Like type III and type IV secretion, secretion by the type VI system is Sec-independent and probably delivers substrate proteins into the cytoplasm of target cells (Bingle et al., 2008). The prototypes of proteins secreted by the type VI system are the Hcp and VrgG proteins of Vibrio cholerae (Pukatzki et al., 2006). VrgG resembles bacteriophage tail proteins, which assemble into a spike that inserts into the OM to initiate phage infection (Cascales, 2008). Hcp forms oligomeric rings which might form a channel or pilus (Cascales, 2008). These proteins may thus constitute the extracellular structure for puncturing target cell membranes and transporting other substrate proteins of type VI secretion. This would mean that these proteins are both substrates for and part of the type VI secretion system. Gene clusters encoding type VI secretion systems contain between 12 and 25 genes, so the structure and assembly of the machinery appears to be rather complex (Cascales, 2008). A cytoplasmic ATPase called ClpV is necessary for secretion and presumably energises the system (Cascales, 2008). It is unclear whether the secretion is a one- or two-step process, but substrate proteins are secreted uncleaved and do not contain N-terminal signal sequences, suggesting that type VI secretion is Sec-independent (Cascales, 2008).

2.1.6 Other secretion pathways in Gram-negative bacteria

In addition to the Sec machinery, which transports unfolded proteins across the IM, Gram-negative bacteria contain another system that allows folded proteins to traverse the IM. This is the twin-arginine translocation (Tat) pathway, as the signal peptides for this transport contain an S/T-R-R-X-F-L-K motif, where X is any polar amino acid (Natale et al., 2008). The Tat translocase of Escherichia coli consists of three proteins, TatA, TatB and TatC. The translocase is powered not by the hydrolysis of ATP, but the proton motive force across the IM (Natale et al., 2008). After translocation, the Tat signal peptide is usually removed by signal peptidase (Natale et al., 2008). The Tat system is typically used to transport proteins that acquire cofactors in the cytoplasm, such as proteins involved in redox reactions. However, some integral membrane proteins, periplasmic enzymes and

17

binding proteins, and protein complexes are also transported into the periplasm by the Tat pathway. The Tat pathway also delivers some proteins for transport across the OM by the type II secretion system (Johnson et al., 2006).

There is yet another secretion pathway in Gram-negative bacteria, used for the biogenesis of fimbriae and pili, termed the chaperone-usher pathway (Sauer et al., 2004). The structural subunits of fimbriae are transported into the cytoplasm by the Sec machinery, where a chaperone protein binds to the emerging subunit and maintains it in a semi-folded state. The chaperone then delivers the fimbrial subunits to the usher, which is a pore-forming protein in the OM (Thanassi et al., 2005). The usher also acts as the base of the fimbria. The first subunits added form the tip structures of the nascent fimbria, and these are followed by sequential addition of pilin subunits which form the often lengthy (up to several micrometres) fimbrial rod (Sauer et al., 2004). The fimbria thus grows from the top down. Remarkably, the energy requirement of the system seems to come from the energy of folding of the subunits as they are added to the fimbria. The chaperone keeps its subunit in a high-energy primed state and the release of this energy on folding drives secretion through the usher and formation of the fimbria (Sauer et al., 2004). In this respect, the chaperone-usher pathway is analogous to autotransport, where the energy for translocation is also provided by protein folding (Thanassi et al., 2005).

A recent study has proposed a mechanism for pilus assembly, where the usher plays a central role (Remaut et al., 2008). Polymerisation of fimbrial subunits takes place through a process called donor strand exchange, where an N-terminal extension from a preceding subunit completes a β-sheet in the following subunit by displacing a strand donated by the chaperone. The usher, which functions as a dimer, recruits chaperone-subunit complexes through its N-terminal domain. The N-terminal domain of the usher remains associated with the chaperone-subunit complex until donor strand exchange is complete. Only one pore of the dimeric usher is used for pilus assembly; the N-terminal domain of the other monomer is used to recruit new chaperone-subunit complexes to the growing pilus. In this way, the N-terminal domains of the two usher monomers act consecutively in a see-saw fashion: while the N-terminal domain of one is attached to the growing pilus during donor strand exchange, the other binds to a periplasmic chaperone-subunit complex and is ready to deliver this complex to the pilus as soon as the already present chaperone is released after donor strand exchange (Remaut et al., 2008).

2.1.6 Secretion in Gram-positive bacteria

Having only a single membrane to cross, protein secretion in Gram-positive bacteria is much easier than in Gram-negative bacteria. The main bulk of secreted proteins pass through the Sec pathway, which transports proteins directly into the extracellular space. In contrast to Gram-negative bacteria, Gram-positive bacteria do not contain SecB, so substrate proteins are delivered to the translocase through the SRP pathway (van Wely et al., 2001). Some Gram-positive bacteria also contain a second, non-essential Sec system, dependent on a separate SecA homologue, SecA2 (Rigel & Braunstein, 2008) Gram-positive bacteria can also export proteins via the Tat pathway (Pallen et al., 2003). Recently, a novel secretion system has been described for the secretion of the virulence-

18

related protein ESAT-6 in Mycobacterium tuberculosis. This protein lacks a classical signal sequence, and it is secreted by a cluster of genes referred to as ESX-1. This secretion system has been called Type VII secretion (Abdallah et al., 2007). Mycobacteria contain several such secretion systems, and type VII secretion clusters are present in several other Gram-positive taxa (Abdallah et al., 2007).

2.2 Type V secretion: autotransportation and two-partner secretion

Type V secretion is a Sec-dependent protein secretion system consisting of three distinct pathways, which share a number of common features. Type Va secretion comprises the classical autotransporter pathway, type Vb refers to two-partner secretion and the trimeric autotransporters form the type Vc secretion pathway. All three systems share a similar two-step mechanism, where the proteins are translocated across the IM in a Sec-dependent manner, while secretion across the OM depends on an integral β-barrel membrane component. In the autotransporter pathways, the membrane-embedded β-barrel, also referred to as the translocation unit or β-domain, and the secreted portion (the passenger or α-domain) are both contained in a single polypeptide, or three identical polypeptide chains in the case of trimeric autotransporters (Dautin & Bernstein, 2007). In contrast, the β-barrel and secreted passenger domain in two-partner secretion are coded for by different genes (Jacob-Dubuisson et al., 2001). It is unclear whether the structural and mechanistic similarities of the different subtypes of the type V secretion system are due to divergence from a common ancestor or whether this is a case of convergent evolution.

Compared to most other secretion systems in Gram-negative bacteria, type V secretion is conceptually very simple (Figure 2.2). The term autotransporter was coined by Klauser et al. (1993), as it appeared that all the information required for translocation was contained in the polypeptide chain and no extraneous factors were required. This apparent mechanistic simplicity may explain why type V secretion is the most prevalent mechanism for protein export in Gram-negative bacteria: over 500 classical autotransporters have been reported (Junker et al., 2006).

2.2.1 Classical autotransporters

Pohlner et al. (1987) were the first to describe the phenomenon of autotransportation. They studied the secretion of IgA protease from Neisseria gonorrheae and proposed the following model for the secretion of IgA protease: the N-terminal signal peptide is recognised by the Sec machinery, which leads to the translocation of the protein across the inner membrane. After this, the C-terminal domain forms a pore in the OM through which the rest of the protein is then translocated into the extracellular space, where the protease domain folds into its native structure and is autoproteolytically cleaved from the pore domain (Figure 2.2) (Pohlner et al., 1987). This model of transportation has been accepted

19

as the paradigm for type Va secretion, though some aspects are still under debate (discussed more fully in section 2.2.1.3).

Figure 2.2 Schematic model of type V secretion. In classical autotransportation (type Va), the protein is synthesised in the cytoplasm and then translocated across the IM by the Sec machinery (1), after which the N-terminal signal peptide is cleaved off (not shown). The translocation domain (light grey) inserts into the OM (2), and the passenger domain (dark grey) is then transported, presumably through the pore of the translocation domain, into the extracellular space where it folds into its native structure (3). In two-partner secretion (type Vb), the passenger (TpsA; dark grey) and translocation domain (TpsB; light grey) are separate polypeptide chains which both contain a signal sequence. After transportation by the Sec translocase (1), TpsB inserts into the membrane and recognises TpsA by its periplasmic POTRA domain (2). TpsA is then transported through the pore of TpsB into the extracellular space (3). Trimeric autotransportation (type Vc) follows the same sequence of events as in classical autotransporters, with the obvious difference that the translocation domain trimerises and three chains are transported into the extracellular space. After translocation by the Sec machinery (1), the translocation unit (light grey) inserts into the OM and trimerises (2), followed by the transport of three passenger domains (dark grey) through the pore. These then fold into the trimeric structure on the outside of the cell (3). Adapted from Henderson et al. (2004).

Since this discovery, a large number of proteins that resemble IgA protease both in structure, function and secretion have been described (Henderson et al., 2004). These monomeric, classical autotransporters share the conserved C-terminal translocation

20

domain of approximately 300 residues (Klauser et al., 1993), but also have structural features in common in the passenger domain.

Autotransporters display a wide variety of functions, listed in Table 2.1. Although as a class autotransporters have a number of different activities, a unifying aspect is that these are commonly involved in virulence. Below, I will give short description of some usual autotransporter functions, but for more extensive reviews see Henderson et al. (2001 & 2004) .

A significant proportion of autotransporters are proteases, including the prototypical IgA protease. One subfamily of autotransporters is the serine protease autotransporters of Enterobacteriaceae (SPATEs). These include well-characterised proteins such as EspC, EspP, Tsh (also referred to as Hbp) and Pet from pathogenic strains of Escherichia coli (Provence & Curtiss, 1994; Stein et al., 1996; Brunder et al., 1997; Eslava et al., 1998) and SigA and SepA from Shigella flexneri (Benjelloun-Touimi et al., 1995; Al-Hasani et al., 2000). Other proteases include Prts (formerly Ssp) from Serratia marcescens and Pta from Proteus mirabilis (Ohnishi et al., 1994; Alamuri & Mobley, 2008). The adhesion and penetration protein Hap from Haemophilus influenzae has serine protease activity that mediates the release of the Hap passenger domain from the cell surface (Fink et al., 2002). However, Hap also functions as an adhesin and autoagglutination factor (Hendrixson & St. Geme, 1998). These activities are possible because not all Hap molecules are proteolysed; some remain attached to the OM (Hendrixson & St. Geme, 1998).

A second significant function for autotransporters is to act as adhesins. The classical example is pertactin (Prn) from Bordetella pertussis and related species, an immunodominant protein that is part of acellular pertussis vaccines (Leininger et al., 1991; Greco et al., 1996). Prn mediates binding of B. pertussis to several types of epithelial cells, though its receptor remains elusive. A related B. pertussis protein, BrkA, also functions as an adhesin but in addition promotes serum resistance (Fernandez & Weiss, 1994). E. coli expresses a group of three autotransporter adhesins, the adhesin involved in diffuse adherence (AIDA-I), the common surface protein antigen 43 (Ag43) and TibA (Benz & Schmidt, 1989; Henderson et al., 1997; Lindenthal & Elsinghorst, 2001). These three proteins promote aggregation and biofilm formation, and can cross-react with each other. Therefore, the term self-associating autotransporter (SAAT) has been proposed for this group (Klemm et al., 2006). Helicobacter pylori produces an autotransporter, BabA, which mediates attachment to human gastric epithelial cells by binding to the blood group antigens fucosylated Lewis b and H1 (Ilver et al., 1998).Several other functions have been attributed to autotransporters. Pseudomonas aeruginosa and Salmonella enterica serovar typhimurium both produce esterases. EstA from P. aeruginosa is required for the production of rhamnolipids, which act as a type of biosurfactant, and is also involved in cell motility and biofilm formation (Wilhelm et al., 2007). S. flexneri is an intracellular pathogen, and an autotransporter, IcsA (also known as VirG), is responsible for its intracellular motility (Bernardini et al., 1989). IcsA is an actin-polymerising protein that localises to the bacterial pole (Goldberg et al., 1993). An interesting and important autotransporter toxin is the vacuolating cytotoxin VacA of H. pylori (Reyrat et al., 1999). This toxin is largely responsible for the induction of gastric

21

Table 2.1 Selected characterised classical autotranporters.

Autotransporter Organism Functions Reference

Aae Aggregatibacter

actinomycetemcomitans

Adhesin Rose et al., 2003

Ag43 Escherichia coli SAAT, adhesin Henderson et al., 1997

AIDA-I E. coli SAAT, adhesin Benz & Schmidt, 1989

AoaA Azorhizobium caulinodans Exopolysaccharide production,

symbiosis promotion, N2 fixation

Suzuki et al., 2008

ApeE Salmonella typhimurium Esterase Carinato et al., 1998

BabA Helicobacter pylori Adhesin Ilver et al., 1998

BrkA Bordetella pertussis Serum resistance, adhesin Fernandez & Weiss, 1994

EspC E. coli SPATE Stein et al., 1996

EspP E. coli SPATE Brunder et al., 1997

EstA Pseudomonas aeruginosa Esterase, rhamnolipid production,

cell motility, biofilm formation

Wilhelm et al., 2007

Hap Haemophilus influenzae Protease, adhesin Provence & Curtiss, 1994

Hbp/Tsh E. coli SPATE; haemoglobin binding,

adhesin

Otto et al., 1998

IcsA Shigella flexneri Actin-polymerisation Bernardini et al., 1989

IgA1 protease Neisseria gonorrheae Protease Pohlner et al., 1987

McaP Moraxella catarrhalis Adhesin Lipski et al., 2007

Nalp N. meningitidis Protease van Ulsen et al., 2003

Pet E. coli SPATE, enterotoxin Eslava et al., 1998

Prn Bordetella spp. Adhesin Leininger et al., 1991

PrtS (Ssp) Serratia marcescens SPATE Ohnishi et al., 1994

Pta Proteus mirabilis Protease, toxin, autoagglutination Alamuri & Mobley, 2008

rOmpA Rickttsia spp. Adhesin Li & Walker, 1998

rOmpB Ricketssia spp. Epithelial cell invasion Chan et al., 2009

Sat E. coli SPATE, vacuolating toxin Guyer et al., 2002

SepA S. flexneri SPATE Benjelloun-Touimi et al.,

1995

SigA S. flexneri SPATE, toxin Al-Hasani et al., 2000

SphB1 B. pertussis Protease, FHA maturation Coutte et al., 2001

TibA E. coli SAAT, adhesin Sherlock et al., 2005

VacA H. pylori Vacuolating cytotoxin Reyrat et al., 1999

YapE Yersinia pestis Adhesin Lawrenz et al., 2009

22

epithelial lesions in animal models (Cover, 1996). After translocation across the OM, VacA is proteolysed into two subunits, p35 and p55. However, the two subunits remain in contact and form flower-shaped oligomers. p55 functions as an adhesin, whereas p35 is responsible for the uptake of the protein into cells and the vacuolating activity (Reyrat et al., 1999).

Although most characterised autotransporters have been described from human pathogens, they are widely distributed in environmental and plant-associated species. An example from a plant-associated bacterium is AoaA, an autotransporter of Azorhizobium caulinodans (Suzuki et al., 2008). This species is involved in nitrogen fixation. Intriguingly, an AoaA-deficient mutant induced expression of plant defence genes, suggesting that this protein is an immunosuppressant. AoaA controls exopolysaccharide production, which may be the mechanism by which it mediates immunosuppresion. By repressing plant defences against A. caulinodans this way, AoaA may promote the symbiotic lifestyle needed for nitrogen fixation (Suzuki et al., 2008)

2.2.1.1 Structure of autotransporters

Classical autotransporters are composed of a single polypeptide chain. Within this chain, three distinct regions can be discerned. The N-terminal residues form a signal sequence, which directs the nascent polypeptide chain to the Sec machinery for translocation into the periplasm. Usually this signal peptide is approximately 25 amino acids in length, but significantly longer signal sequences (> 50 residues) are not uncommon (Henderson et al., 2004). These extended signal sequences may play a role in autotransporter biogenesis, as discussed in more detail in section 2.2.1.3 (Szabady et al., 2005).

The central, largest portion of the protein is called the passenger domain, the extracellular domain of the autotransporter which harbours its specific activity or activities. In many cases, the passenger domain is cleaved from the translocation unit (Henderson et al., 2000). Release of the passenger domain can result in a fully soluble protein, as for IgA protease or Tsh, but often the cleaved passenger domain remains associated with membrane. Examples of this are the adhesins Prn and AIDA-I, where the passenger domains stay attached to the cell surface despite cleavage (Li et al., 1991; Benz & Schmidt, 1992). Some passenger domains are cleaved from their translocation units by an autoproteolytic mechanism. In SPATEs, this autoproteolysis is independent of the serine protease activities of these proteins (Stein et al., 1996). The mechanism involves an aspartic acid-asparagine dyad and proceeds through a succinamide intermediate. The cleavage reaction takes place within the pore of the translocation unit, where the peptide backbone is cut after the asparagine residue of the catalytic dyad located in the linking α-helix (Dautin et al., 2007). The same autoproteolytic cleavage mechanism occurs in the B. pertussis autotransporters BrkA and Prn, though these are only distantly related to SPATEs (Dautin et al., 2007). In Neisseria meningitidis, many autotransporters including the IgA protease are processed by NalP (van Ulsen et al., 2003), and IcsA is cleaved by a dedicated protease, IcsP (Shere et al., 1997). Ag43 may undergo autoproteolytic cleavage through an endogenous aspartyl protease activity (Henderson & Owen, 1999). The

23

cleavage of AIDA-I is also suggested to be autocatalytic, although it does not contain any protease motifs (Suhr et al., 1996). Interestingly, cleavage of the passenger domain appears not to be necessary for its function (Charbonneau et al., 2006).

There are currently three crystal structures of autotransporter passenger domains in the protein data bank (PDB). The first to be crystallised was Prn (Emsley et al., 1996). The fold is an extended right-handed parallel β-helix with three-strands per rung of the helix (Figure 2.3A). Prn contains sixteen turns of the β-helix, and the C-terminal region contains two rungs of a β-roll. Several loops protrude from the structure, including a loop containing an RGD motif that is possibly involved in adhesion.

Figure 2.3 Structures of classical autotansporters. A) Pertactin passenger domain (PDB 1DAB). The β-helix is in green, the RGD motif-containing loop is in blue and the β-roll in yellow. B) Hbp passenger domain (PDB 1WXR). The β-helix is in green, the trypsin-like protease domain in blue, and other extrahelical regions (from N- to C-terminus) in orange, yellow, cyan and magenta. C) VacA passenger domain fragment (PDB 2QV3) D) NalP translocation unit (PDB 1UYN). The β-barrel is in blue and the linking α-helix in yellow. E) The EspP translocation unit (PDB 2QOM). The β-barrel is in blue and the linking α-helix in yellow. The figures were prepared using PyMOL (Delano Scientific).

The crystal structure of the passenger domain of Hbp (Figure 2.3B) also has a three-stranded β-helical fold, but with 24 turns of the helix (Otto et al., 2005). Similar to Prn, Hbp has several loops extending from the helical stem, but in addition the structure contains two globular domains. The N-terminal domain contains the serine protease activity of this SPATE. The structure of the domain is a mixed α/β-fold which is very similar to bovine trypsin (mainchain root mean square deviation 2.5 Å) (Otto et al., 2005).

24

The second extrahelical domain is much smaller than the protease domain, with a length of only 76 residues. Its core is an antiparallel β-sheet with three strands connected by long loops, and it has distant structural similarity to a chitin-binding domain of chitinase B. The function of this domain is unclear, but it is not the haem-binding region as deletion of the domain does not alter the affinity of Hbp to haem. It may instead be required for docking the protein back onto the cell surface after haem acquisition (Otto et al., 2005). Alternatively, it might be involved in one of the other various activities ascribed to Hbp, such as haemagglutination or binding to fibronectin or collagen type IV (Kostakioti & Stathopoulos, 2004).

The third and most recent structure is that of a fragment of the vacuolating toxin VacA from H. pylori (Figure 2.3C). The crystal structure of the p55 fragment of VacA, which is important in binding to cellular receptors, is again a three-stranded β-helix (Gangwer et al., 2007). In VacA, the β-sheets in the helix are not continuous, but are interrupted by several kinks which change the orientation of the sheets. The C-terminus of the structure contains a small globular α/β domain. Interestingly, this domain contains a disulphide between two cysteines located close to each other in the sequence. The disulphide is necessary for efficient VacA secretion (Letley et al., 2006). Disulphides are absent from Hbp and Prn, and autotransporters in general have very few cysteines (Jose et al., 1995).

It is interesting to note the structural similarity between Prn, Hbp and VacA. All three have a central β-helical core. In fact, the β-helix is a widespread structural motif in autotransporters (Junker et al., 2006). β-helices are very stable structures, and the β-helical fold of autotransporters may have evolved as a response to the proposed method of secretion, where the folding of the passenger domain on the outside of the cell could provide the free energy needed for transport across the membrane (Junker et al., 2006). Several autotransporters contain a “folding domain”, a region that facilitates or nucleates the assembly of the passenger domain into its native tertiary structure, at the C-terminus of the passenger domain (Henderson et al., 2004). However, this region is not required for translocation. A folding domain was first described in the serine protease autotransporter PrtS from Serratia marcescens, where it was termed the junction region (Ohnishi et al., 1994). The junction region acts as an intramolecular chaperone that aids in folding of the passenger domain, but it can also promote passenger domain folding in trans. In BrkA, the C-terminal of the passenger domain is also a folding region, without which the secreted passenger domain is susceptible to proteolysis, indicative of an unfolded state (Oliver et al., 2003). Prn contains a very stable structure at the C-terminus of its passenger domain, which corresponds to the six C-terminal turns of Prn β-helix (Junker et al., 2006). Prn contains a junction region homologous to those found in PrtS and BrkA, but the stable folding domain of Prn region extends further into the β-helix than the predicted junction region. Similar folding domains have been identified in numerous other autotransporters (Oliver et al., 2003).

The C-terminal 250-300 residues of autotransporters are responsible for insertion into and translocation across the OM (Pohlner et al., 1987). This translocation unit or domain (formerly also termed the β-domain or helper domain) is formed of a membrane integral β-barrel. The β-barrel was predicted to contain 14 strands (Jose et al., 1995); however, structural studies have shown that the number of strands is in fact 12 (Oomen et al., 2004; Barnard et al., 2007). The N-terminus of the translocation unit preceding the β-barrel is

25

referred to as the linking region or linker (Klauser et al., 1993), and it forms an α-helix (Suzuki et al., 1995). The linker is necessary for autotransportation, as constructs with a deleted linker no longer exported a heterologous passenger domain (Maurer et al., 1999).

There are two structures of autotransporter translocation domains in the PDB. The first was NalP (Figure 2.3D) (Oomen et al., 2004). Strikingly, the linking α-helix is inserted into the central pore of the barrel. This arrangement has implications for the mechanism of autotransportation, as discussed in section 2.2.1.2. The pore of the barrel is lined with predominantly hydrophilic side chains, and the pore is 10 × 12.5 Å in size. Sequence alignment suggests that the linking α-helix is conserved in translocation units of other autotransporters (Oomen et al., 2004).

The second translocation unit to be solved was that of EspP (Barnard et al., 2007). The overall structure is highly similar to NalP; however, there is a major difference in the linker helix (Figure 2.3E). In EspP, the autoproteolytic cleavage site is located in the α-helix, and consequently only one and a half turns of the linker helix remain within the barrel. The orientation of this helix is perpendicular to the axis of the β-barrel. A further dissimilarity is the positioning of one of the extracellular loops (loop 5) of the EspP barrel, which folds over to cover the extracellular side of the barrel pore. Presumably, these changes take place after the cleavage of the passenger domain, and the structure before cleavage is similar to that of the NalP translocation unit. However, the pores of the two proteins have different electrostatic properties: the pore of EspP is mostly acidic, whereas the NalP pore is basic on the cytoplasmic side and neutral on the extracellular side. This suggests that the interactions between the linker and the pore are specific for each protein, although the overall mechanism of autotransportation is probably the same (Barnard et al., 2007).

2.2.1.2 Mechanism of autotransportation

Three models have been proposed for the phenomenon of autotransportation, (1) the single-chain model, (2) the oligomeric model and (3) the Omp85/YaeT model (Figure 2.4). All three models are supported by some experimental evidence, as discussed below.

The single-chain model corresponds more or less to the original hypothesis for autotransportation (Pohlner et al., 1987). Translocation of the passenger domain is an activity contained entirely in the single autotransporter polypeptide. The model assumes that the C-terminal β-barrel folds into the outer membrane first, after which the passenger domain, which has been kept in an unfolded state in the periplasm, passes through the pore of the barrel, and the folding of the passenger domain makes the process essentially irreversible (Klauser et al., 1990).

The second model, referred to as the multimeric model, is based on electron micrographs of the translocation domain of the IgA protease (Veiga et al., 2002). In this study, IgA protease was found to form oligomeric ring structures on the cell surface. Additionally, biochemical data corroborated the view that the IgA protease translocation unit formed a multimeric structure, and other experiments suggested that the pore formed by the translocation domain was ~2 nm in width, consistent with the size of the pore of the multimeric ring in the electron micrographs (Veiga et al., 2002). A similar pore size was

26

reported for the β-barrel of PalA based on liposome swelling assays (Lee & Byun, 2003). Furthermore, IgA protease is capable of translocating a large, disulphide-bridged single-chain antibody, albeit at low frequency (Veiga et al., 1999). It seems unlikely that the pore of the β-barrel itself could accommodate such a large, folded molecule. The multimeric model offers an explanation: the passenger domains are secreted through the central pore of a ring structure formed by several translocation domains, and this pore is wide enough to allow the passage of folded proteins.

Figure 2.4 Models for the mechanism of autotransportation. In the single-chain model, all the activities required for OM translocation are contained within the autotransporter polypeptide chain. The passenger inserts through the pore of the translocation unit β-barrel (shown in cross-section) in an unfolded state and folds on the outside of the cell. The passenger domain can transit the pore in one of two ways: N-terminus first (threading model) or C-terminus first (hairpin model). In the Omp85 model, the passenger domain is not secreted by the translocation unit (full cylinder). Rather, the passenger domain crosses the OM with the assistance of Omp85 (shown in cross-section) in a folded or semi-folded state. In the multimeric model, several translocation units multimerise to form a ring structure, and the passenger domains are secreted through the central pore of this ring. Figure adapted from Jacob-Dubuisson et al. (2004).

However, most of the evidence supports the single-chain model. The strongest individual piece of evidence is the crystal structure of the NalP translocation domain (Oomen et al., 2004). In the structure, the α-helical linker is positioned inside the β-barrel, which is completely inconsistent with the multimeric model. Furthermore, the NalP translocation domain does not form multimers the way the IgA protease is reported to (Oomen et al., 2004). In addition, the outside of the NalP barrel is hydrophobic, and so the formation of a hydrophilic, protein-translocating channel by multimerisation does not seem possible for this protein. The crystal structure of the EspP translocation unit has further confirmed these findings (Barnard et al., 2007). This suggests that either the results obtained for the IgA protease translocation domain are artifactual, or that it is a special case among autotransporters.

If the single-chain model is correct, the passenger domain can be translocated in two ways: either by threading the N-terminus through the pore of the β-barrel first (threading model), or alternatively by inserting the C-terminus of the passenger domain into the

27

barrel to create a hairpin structure into the pore first, after which the rest of it is pulled through in a hairpin-like structure (hairpin model) (Oomen et al., 2004). In both submodels, the folding of the passenger domain on the outside of the cell contributes the force to pull the rest of the protein through the pore. Of these two, the former model seems unlikely, as heterologous passenger domains are exported efficiently by the translocation domain of AIDA-I (Maurer et al., 1997).

There is considerable evidence to suggest that the passenger domain is extruded through the pore of the β-barrel by the hairpin model. The position of the folding promoting regions in BrkA, Prn and Pet suggest that this region should be secreted first to aid in the folding of the rest of the passenger domain (Oliver et al., 2003; Junker et al., 2006; Renn & Clark, 2008). The linking α-helix of SPATEs contains a conserved 14-residue motif (Kostakioti & Stathopoulos, 2006). Mutagenesis experiments replacing residues in this motif of Tsh interfered with passenger domain secretion but not the membrane insertion of the β-barrel (Kostakioti & Stathopoulos, 2006). A recent study provided further evidence. Paired cysteines were introduced that form disulphide bridges at different positions in the Prn passenger domain and OM translocation was followed (Junker et al., 2009). A disulphide located in the C-terminus of the Prn passenger domain prevented the surface exposure of the passenger domain, whereas more N-terminally located disulphides allowed part of the passenger domain to be exported. Consistent with this, adding a reducing agent allowed full passenger domain translocation for all double-cysteine mutants (Junker et al., 2009).

Electrochemical and antibiotic diffusion experiments have shown that the NalP translocation unit can function as a pore, and that the linking α-helix can insert into and plug the pore both in vitro and in vivo (Oomen et al., 2004). The pore of NalP (10.5 × 12.5 Å) is wide enough to accommodate two extended polypeptide chains, but the fit would be very tight (Oomen et al., 2004). The NalP pore could not accommodate a folded helix and another chain, so if the hairpin model is true, the helix presumably only forms after translocation. The residues on the barrel wall forming hydrogen bonds with the linking helix are flexible and possibly fold backwards during transport (Oomen et al., 2004).

The unusually long signal peptides of some autotransporters may have a function in OM translocation and thus provide indirect evidence in favour of the hairpin model. Studies of the role of the extended signal peptide of EspP suggest that the signal peptide remains in contact with the IM Sec machinery for longer than standard-length signal peptides, thus acting as a kind of membrane anchor (Szabady et al., 2005). This may aid in keeping the passenger domain in an unfolded state in the periplasm while the translocation domain inserts and folds into the OM. This is supported by the findings that replacing the long signal peptide of EspP with a shorter one from maltose-binding protein or removing the N-terminal extension reduce the OM translocation efficiency of EspP without having a significant effect on IM transport (Szabady et al., 2005). Retarding the release of the passenger domain from the Sec machinery and thus its folding domain is consistent with the single-chain model, where the passenger domain is required to remain in a largely unfolded state prior to translocation across the OM.

In the fourth model for autotransportation, an auxiliary OM protein complex is required for translocation. The Omp85 complex is absolutely required for the biogenesis of all outer membrane proteins in N. meningitidis, including the autotransporter IgA1

28

protease (Voulhoux et al., 2003). The E. coli homologue of this protein, YaeT, was later shown to be necessary also for the OM insertion of AIDA-I and heterologously expressed BrkA (Jain & Goldberg, 2007). YaeT is also required for the surface display of the autotranporters IcsA and SepA in S. flexneri (Jain & Goldberg, 2007).

A possible role for Omp85 in autotransporter biogenesis is to provide a channel in the outer membrane through which the passenger domain is translocated across the OM (Jacob-Dubuisson et al., 2004). Alternatively, the Omp85 complex might stabilise an open conformation of the β-barrel to allow the passage of bulky, partially folded regions of passenger domains through an extended pore (Oomen et al., 2004). Indeed, such partially folded intermediates have been reported for IcsA and EspP (Brandon & Goldberg, 2001; Skillman et al., 2005), and these could presumably not fit through the NalP or EspP pores seen in the crystal structures. The Omp85 model thus provides a possible explanation for the autotransport of folded or partially folded structures. However, there is also evidence showing that introducing disulphides or domains with strong intrinsic folding propensity into autotransporter passenger domains can significantly impede OM translocation (Rutherford et al., 2006; Jong et al., 2007; Junker et al., 2009). Thus the exact role of Omp85 still remains to be elucidated. Although bona fide autotransportation and the Omp85 model appear to be inconsistent, they are not necessarily mutually exclusive. Omp85 is apparently required for the biogenesis of all β-barrel OM proteins. It is conceivable that Omp85 is needed for the insertion of autotransporter β-barrels into the OM, but after this Omp85 is no longer required for the transport of the passenger domain across the OM. In this scenario, once the β-barrel has assumed its native fold in the OM, translocation of the passenger domain is genuinely autonomous, and so the term ‘autotransporter’ would still be applicable.

2.2.2 Two-partner secretion

The two-partner secretion system (TPS) is the second pathway of type V secretion. It resembles autotransportation in that the secretion apparatus also consists of two factors, a secreted passenger domain and a translocation unit. However, in contrast to autotransporters where both functionalities are found in the same polypeptide chain, in TPS the two domains are separate proteins. The passenger polypeptide is referred to as the TpsA protein, whereas the membrane-embedded translocation unit is called TpsB (Jacob-Dubuisson et al., 2001). Because they are produced as discrete polypeptides, one TpsB could in principle translocate several different TpsAs. This, however, does not appear to be the case; each TpsB seems to be devoted to secreting only its cognate TpsA partner (Jacob-Dubuisson et al., 2001). The genes coding for a single TpsA and TpsB partnership are usually organised as an operon at the same locus in the genome (Jacob-Dubuisson et al., 2001), an arrangement which allows for the expression of both partners to be tightly coordinated.

Characterised two-partner secretion proteins are listed in Table 2.2. Though TPS-like proteins are widely dispersed in different taxa, occurring even in higher eukaryotes (Yen et al., 2002), Gram-negative bacteria contain many more classical autotransporters or even trimeric autotransporters than TPSs. The best-known system is the filamentous

29

haemagglutinin (FHA) of B. pertussis and the high-molecular weight (HMW) proteins 1 and 2 from Haemophilus influenzae, which are adhesins (Relman et al., 1989; St. Geme et al., 1993). Another well-characterised system is the Ca2+-independet haemolysin/cytolysin ShlA from Serratia marcescens (Poole et al., 1988).

Table 2.2 Selected characterised two-partner secretion proteins.

TpsA protein

TpsB protein Organism TpsA functions Reference

CdiA CdiB Escherichia coli Contact-dependent growth inhibition

Aoki et al., 2005

EthA EthB Ewardsiella tarda Cytolysin Hirono et al., 1997 FHA FhaC Bordetella pertussis Adhesin, immunomodulator Inatsuka et al., 2005 HecA HecB Erwinia crysanthemi Attachment to epidermal

cells, aggregation Rojas et al., 2002

HhdA HhdB H. ducreyi Cytolysin Palmer & Munson, 1995

HlyA HlyB E. coli Cytolysin Goebel & Hedgpeth, 1982

HMW1 HMW1B Haemophilus influenzae

Adhesin St. Geme et al., 1993

HMW2 HMW2B H. influenzae Adhesin St. Geme et al., 1993 HpmA HpmB Proteus spp. Cytolysin Uphoff & Welch,

1990 HxuA HxuB H. influenzae Haem:haemopexin binding Cope et al., 1994 LspA1 LspA1B H. ducreyi Cytolysin Ward et al., 1998 ShlA ShlB Serratia marcescens Cytolysin Poole et al., 1988

TpsAs are fairly large (over 100 kDa and up to 500 kDa) (Jacob-Dubuisson et al.,

2001), and many of them are proteolytically cleaved and released into the medium. However, some TpsAs, like FHA, remain attached to the OM, in accordance with their function as adhesins (Relman et al., 1989). TpsAs contain a conserved N-terminal domain, the TPS domain (Jacob-Dubuisson et al., 2001). Like the passenger domains of autotransporters, many TpsAs are predicted to form β-helical structures (Kajava et al., 2001). The crystal structure of a 30-kDa fragment of FHA TPS domain (Fha30) confirmed this (Clantin et al., 2004). Fha30 forms a three-stranded right-handed parallel β-helix. In addition, three extrahelical elements are present: a β-hairpin, a four-stranded β-sheet and three β-strands capping the β-helix (Figure 2.5A). Alignment of this structure with sequences from other TpsAs showed two regions with high conservation in the β-helix (C1 and C2) and two zones with lower conservation (LC1 and LC2). The LC regions are in the N-terminal cap and the extrahelical β-sheet. The C regions are crucial for the interaction of FHA with FhaC and its consequent secretion (Hodak et al., 2006).

30

Figure 2.5 Structures of TPS proteins. A) FHA TPS domain (PDB 1RWR). The β-helical core is green, the capping β-hairpin yellow, the extrahelical β-hairpin blue and the extrahelical β-sheet magenta. B) FhaC (PDB 2QDZ). The β-barrel is blue, the inserted α-helix yellow, POTRA 1 is green, POTRA 2 magenta and L6 cyan. The figure was prepared in PyMOL (DeLano Scientific).

TpsBs are proteins of ~60 kDa that form β-barrels in the OM. These proteins are responsible for the secretion of their substrate TpsA across the OM. In addition, some TpsBs (e.g. ShlA) are also needed for the activation of their TpsA partner (Schönherr et al., 1993). The TspBs are homologous to Omp85 (Yen et al., 2002). The pore-forming β-barrel is located in the C-terminus of the protein (Meli et al., 2006). The N-terminus of TpsBs was predicted to contain periplasmic POTRA (polypeptide transport-associated) domains that would interact with the TpsA partner (Sanchez-Pulido et al., 2003), which was later confirmed by the crystal structure of FhaC (Clantin et al., 2007). The transmembrane domain is a 16-stranded β-barrel, and the periplasmic region consists of two POTRA domains (Figure 2.5B). The N-terminus is an α-helix, which is inserted into the pore of the β-barrel. In addition, a large extracellular loop (L6) also folds into the pore. These two insertions effectively obscure the hydrophilic pore, and it seems unlikely that polypeptide translocation could occur with them in place (Clantin et al., 2007). However, electrophysiological measurements suggest that the blocking α-helix is outside the pore during secretion, thus giving the pore a diameter of 8 Å (Meli et al., 2006; Clantin et al., 2007). Deletion of L6 abolished secretion entirely, suggesting it may be involved in the export process (Clantin et al., 2007).

The mechanism of TPS has many similarities to autotransportation. In fact, the only major difference is the genetic separation of the transporter and passenger proteins. TpsAs often contain extended signal peptides reminiscent of the long signals seen in many autotransporters (Jacob-Dubuisson et al., 2001). The signal for OM secretion in the TpsAs appears to reside in the conserved regions of the N-terminal TPS domain of the proteins (Jacob-Dubuisson et al., 2004). These regions are postulated to interact with TpsB. The earliest models for TPS secretion postulated that TpsA transits the pore N-terminus first (Guédin et al., 1998). However, recent work has suggested that at least FHA might be

31

secreted C-terminus first, i.e. as a hairpin. The N-terminal TPS domain of FHA would remain in contact with FhaC while the C-terminus is exported first to fold outside the OM (Mazar & Cotter, 2006; Clantin et al., 2007). This model is strikingly similar to the hairpin model suggested for classical autotransporters. A further similarity is that secretion of FHA requires that the polypeptide be in an unfolded conformation in the periplasm, and the folding of the passenger protein on the cell surface provides the free energy for translocation (Guédin et al., 1998; Hodak et al., 2006). For a more complete discussion of the current understanding of the mechanism of the TPS secretion mechanism, the reader is referred to (Mazar & Cotter, 2007).

2.2.3 Trimeric autotransporters

In addition to the classical, monomeric autotransporters, there is a second class of autotransported proteins − the trimeric autotransporters. Unlike many classical autotransporters, the passenger domains of trimeric autotransporters are not cleaved from the translocation unit following secretion. This is in accordance with their major activity, which is to act as adhesins (Linke et al., 2006). The prototypical members of the family are the Yersinia adhesin YadA (previously known as Yop1 or P1) and the Haemophilus influenzae adhesion protein Hia.

As their common name implies, proteins of the type Vc secretion pathway are obligate homotrimers, and indeed trimerisation is necessary for the proteins to be active (Cotter et al., 2006). YadA was known to be a stable multimer, as the protein migrated as a much larger molecule than expected in SDS-PAGE, with a molecular weight of approximately 200 kDa, but it dissociated to a monomer of ~50 kDa after prolonged boiling or incubation in concentrated urea (Skurnik et al., 1984; Zaleska et al., 1985). The nucleotide sequence of YadA showed that the expected size of the polypeptide chain was ~45 kDa, and that the stability of YadA could not be explained by the presence of disulfide bridges, as there were no cysteines (Skurnik & Wolf-Watz, 1989). The first indication that the number of monomers in YadA was three was the coexpression of two YadA polypeptides, a full-length chain and a truncation, which allowed the deduction of the number of chains per molecule based on the banding pattern (Gripenberg-Lerche et al., 1995). More detailed analysis of the migration of different YadA species in SDS-PAGE and gradient density centrifugation gave further support for a quaternary structure containing three chains (Mack et al., 1994), as did a later study repeating the use of hetero-oligomeric YadA (Roggenkamp et al., 2003). Also the Hia membrane anchor was predicted to form trimers (Surana et al., 2004). The trimeric structure of these proteins was finally confirmed by the crystal structures of the Hia and YadA head domains (Nummelin et al., 2004; Yeo et al., 2004).

Various names have been proposed for the molecules transported in a type Vc manner. The first term used to describe YadA was non-pilus associated adhesin, used to differentiate it from the fimbriae and pili that are common adhesion molecules in Gram-negative bacteria (Hoiczyk et al., 2000). To distinguish the type Vc-secreted molecules from other non-fimbrial adhesins, a second name, the oligomeric coiled-coil adhesin or Oca family, was suggested due to the presence of coiled-coil structures in the

32

characterised members of the family and the fact that these proteins formed oligomeric structures on the cell surface (Roggenkamp et al., 2003). Once the number of monomers in these proteins was confirmed to be three, the name trimeric autotransporters was introduced (Cotter et al., 2005a). A slightly longer version, trimeric autotransporter adhesins, was also put forward, as all characterised proteins had adhesive activity (Linke et al., 2006). In this text, I will use the term trimeric autotransporters when referring to the proteins of the type Vc secretion system, as this name is both descriptive and succinct.

2.3.2.1 Structure of trimeric autotransporters

Electron microscopy revealed that YadA of Yersinia enterocolitica and UspA1/2 of Moraxella catarrhalis form lollipop-like projections from the outer membrane (Hoiczyk et al., 2000). Based on this observation, the proteins were divided into structurally distinct regions: the globular head domain (i.e. the bulge of the lollipop), a coiled-coil stalk and the membrane-embedded anchor (Figure 2.6). In addition, a fourth structural region was discovered through sequence comparison. This was

called the neck, as it connected the globular head with the stalk (Hoiczyk et al., 2000).

As in classical autotransporters, the C-terminal membrane anchor, i.e. translocation unit, is also responsible for transporting the passenger domain across the OM in trimeric autotransporters. The membrane anchor of trimeric autotransporters is

much shorter than the ~300 amino acids of monomeric translocation units. The length of the Hia translocation unit is 76 residues, and just 70 residues in YadA (Roggenkamp et al., 2003; Surana et al., 2004). This can of course be explained by the fact that these proteins are trimeric. The C-termini of Hia and YadA were predicted to contain four β-strands, and so the intact trimeric barrel would have 12 strands (Hoiczyk et al., 2000; Surana et al., 2004). This was confirmed by the crystal structure of the Hia translocation unit and Fourier transform infrared spectroscopy studies on the YadA barrel (Meng et al., 2006; Wollmann et al., 2006). The Hia crystal structure (Figure 2.7A) shows a barrel remarkably similar in size and general shape to the NalP barrel, with the central pore 1.8 nm in diameter (Meng et al., 2006). Three α-helices traverse the pore, where they do not coil, but the helices form a distinct supercoil upon exiting the pore.

The passenger domain can be structurally divided into a head, a neck and a stalk. Most trimeric autotransporters have an N-terminal head followed by a neck (Hoiczyk et al., 2000). However, some larger trimeric autotransporters may contain several head-like

Figure 2.6 Model of a typical trimeric autotransporter showing the domain architecture of the lollipop-shaped protein. Adapted from Hoiczyk et al. (2000).

33

regions and necks. For example, BadA contains a total of 24 neck sequences throughout the passenger domain, interspersed with a repeated “BadA” sequence in the stalk (Riess et al., 2004). Such modular architecture is a common feature of trimeric autotransporters and appears to be a factor promoting recombination and rapid evolution in these proteins (Linke et al., 2006). Bioinformatic studies have shown that certain domains are found in a number of trimeric autotransporters, and a database annotating the domain architecture of trimeric autotransporters (the daTAA database; http://toolkit.tuebingen.mpg.de/dataa) has been established (Szczesny & Lupas, 2008).

No experimental structures of full-length trimeric autotransporter passenger domains are currently available. However, there are presently structures from three head domains of trimeric autotransporters in the PDB (Figure 2.7). The first structure to be solved was that of binding region 1 (BD1), a fragment of the head domain of Hia (Yeo et al., 2004). This structure shows a mushroom-shaped architecture with three-fold symmetry containing a broad stem and an elongated cap (Figure 2.7B). The structure contains three well-defined domains. The N-terminal domain consists of an antiparallel β-sheet which forms a β-prism in the trimer. Sequence alignment revealed a prominent motif with the consensus sequence GIN, and consequently this type of β-prism domain is referred to as a GIN domain (Szczesny & Lupas, 2008). One side of the β-sheets in the Hia GIN domain is made up predominantly of hydrophobic residues, thus forming a hydrophobic core in the trimer. Domain 2 (which forms the edge of the mushroom “cap”) is in fact an interrupted neck-like structure (discussed below), called an IsNeck (Meng et al., 2008), with the insertion adopting a globular mixed α/β-fold. This is connected to domain 3 by a very short stretch of coiled-coil. Domain 3, the stalk of the mushroom, is a β-meander composed of 5 strands. Strands from adjacent monomers interdigitate to form the trimer. A striking feature of domain 3 is a tryptophan residue positioned at the N-terminus of the domain, thus forming a ring of three tryptophans at the narrow top of the trimeric domain. The motif GW is conserved in homologous domains, and the domain is consequently referred to as a Trp-ring domain (Szczesny & Lupas, 2008).

Later, structures of BD2 and a non-binding fragment of Hia with similar architecture were crystallised (Meng et al., 2008). The overall structure and topology of BD2 are very similar to BD1 (Figure 2.7C), though it does not contain the N-terminal GIN domain of BD1 (Meng et al., 2008). The folds of the IsNeck domains of BD2 and BD1 are essentially identical, as are the Trp-ring domains from BD1, BD2 and the Trp-ring domain from the third Hia fragment without any characterised binding activity (residues 307-422). This non-binding fragment also contains an IsNeck-like element, with a similar positioning of secondary structural elements as the BD1 and BD2 IsNeck domains but a different topology (Figure 2.7D). This is referred to as a KG domain (Meng et al., 2008). The KG domain of Hia307-422 lacks one α-helix present in the IsNeck domains of BD1 and BD2, and a second helix in the KG domain is aligned differently from the IsNeck domains. Both these elements are involved in forming the binding pocket in BD1 and BD2, and this could explain why Hia307-422 lacks detectable binding activity (Meng et al., 2008).

34

Figure 2.7 Crystal structures of trimeric autotransporters. A) Hia translocation unit (PDB 2GR7) B) Hia BD1 (PDB 1S7M) C) Hia BD2 (PDB 3EMF) D) Hia non-adhesive domain (PDB 3EMI) E) YadA collagen-binding domain (PDB 1P9H) F) BadA head domain fragment (PDB 3D9X) G) UspA1 stalk fragment (PDB 2QIH). Trp-ring domains are in magenta, GIN domains in cyan, β-rolls in green, necks in red, IsNecks in grey, KG domains in black, coiled-coils in orange, linker helices in yellow and β-barrels in blue. Coordinating chlorides (green) and phosphates (red) in coiled-coils are shown in space-filling representation. The figures were prepared with PyMOL (Delano Scientific).

The structure of YadA revealed a novel left-handed β-roll fold (Nummelin et al., 2004). The short β-strands, which form the layers of the β-helical roll, have the consensus sequence NSVAIGXXS. These repeats, of which there are 8 in YadA, were originally supposed to lie on the outside of the protein and form a collagen-binding interface (El Tahir et al., 2000). However, as shown by the crystal structure (Figure 2.7E), these repeats play a structural role. Residues in positions 3-5 of the motifs form the β strand, whereas residues in positions 1-2 are part of a β-turn which is variable in length. The residues in positions 6-8 form a second turn, which is always the same length. The glycine in position 6 is completely conserved in these motifs (Nummelin et al., 2004). This glycine in the second turn is in the αL conformation and forms a hydrogen bond to the hydroxyl of the

35

serine at position 9. This interaction stabilises the turn. The glycine is required in position 6, as larger sidechains could not be accommodated with this conformation. A similar glycine-serine pair is often found stabilising the first turn (glycine at position -2 outside the motif and serine at position 2), but the conservation is not as strong. The residues at positions 2 and 4 are usually small (serine, alanine), and form the hydrophobic core of the monomer. The larger hydrophobic residues (valine, isoleucine) found in positions 3 and 5 contribute towards the shared hydrophobic core of the trimer.

At the top, i.e. the N-terminus of the β-roll, a loop from one monomer crosses over to contact another monomer. This arrangement traps the N-terminus of one monomer under the loop crossing from its anticlockwise neighbour. A similar arrangement is seen at the bottom of the β-roll, where the neck forms a safety pin-like structure, with the chain from one monomer forming contacts with the anticlockwise adjacent monomer. The N-terminal loops and neck together constitute a “lock nut”, giving the YadA head domain its remarkable stability. An ionic network within the neck makes a further contribution to this stability.

The third available structure of a head domain fragment is from the giant trimeric autotransporter BadA (Figure 2.7F). The crystal structure of this fragment contains a Trp-ring domain, a GIN domain, a neck and the beginning of a coiled-coil stalk. In BadA, the Trp-ring and GIN domains are in the reverse order compared to Hia. In addition, BadA also contains a YadA-like β-roll domain with 11 layers lying N-terminal to the Trp ring domain, connected by a short segment of coiled-coil. Though the β-roll and connecting coiled-coil are not part of the crystal structure, homology modelling gives a good picture of the structure of the full head domain of BadA (Szczesny et al., 2008). BadA also contains a neck region, connecting the GIN domain to the start of the coiled-coil stalk, which is readily recognisable from the sequence. Intriguingly, the IsNeck domains of Hia also contain a neck structure, though this is not immediately apparent when examining the sequence, largely due to the insertion of a 44-residue domain into the neck sequence (Szczesny et al., 2008). However, structurally the neck of IsNeck domains is clearly identifiable, and has a nearly identical backbone structure to both the YadA and BadA necks. The conserved fold is stabilised by mainchain hydrogen bonds and sidechain interactions are apparently not needed apart from forming a small hydrophobic core (Szczesny et al., 2008). The network of charged residues seen in the YadA neck is absent from both BadA and Hia, and appears to be a feature specific to YadA.

The stalk of trimeric autotransporters is commonly rich in trimeric coiled-coil segments (Hoiczyk et al., 2000). The main function of the stalk seems to be in projecting the adhesive head domain away from the bacterial cell surface, but stalks are also often involved in protection against host defences (Linke et al., 2006). In YadA, the entire stalk is formed by a coiled-coil. In the model of the YadA stalk, the N-terminal part of the stalk shows a pentadecad periodicity (Hoiczyk et al., 2000). This corresponds to a right-handed coiled-coil with 3.75 residues per turn. At the C-terminus, the stalk adopts a more canonical, left-handed coiled-coil structure with heptad periodicity (Koretke et al., 2006). The transition from the right-handed coiled-coil into the left-handed form is mediated by an insertion of four residues (called a “stutter”) into the final pentadecad repeat of the right-handed coiled-coil (Koretke et al., 2006). The length of the YadA stalk varies between different strains, which may be a reflection of different lengths of O-antigen

36

chains, as YadA normally masks these chains (Hoiczyk et al., 2000). Interestingly, the length of YadA correlates with the length of the type III secretion system injectisome needle (Mota et al., 2005). The injectisome needle must reach beyond occluding structures at the bacterial cell surface in order to make contact with the host cell, and lengthening YadA presumably prevents the needle from interacting with the host cell membrane.

In contrast to YadA, the stalks of UspA1 and UspA2 appear to be canonical left-handed coiled-coils throughout (Hoiczyk et al., 2000). A stretch of the stalk of UspA1 (residues 527-665) has been crystallised. The structure of this region (Figure 2.7G) shows a left-handed coiled-coil with a periodicity of 3.5 residues per turn and a pitch of 150 Å (Conners et al., 2008). This coiled-coil, the longest continuous coiled-coil yet described, has eight instances where chloride ions are located in the core of the coiled-coil. These are in positions where the residues pointing into the core are asparagines. In addition, three histidine residues are part of the hydrophobic core. In two cases, the potential positive charge of the histidines is compensated for by a coordinating phosphate ion situated in the centre of the core. Furthermore, the C-terminus of the coiled-coil contains a stutter that relaxes the degree of supercoiling in this region. These features led to the hypothesis that the stalk may be deformable. The binding site in UspA1 for its cellular receptor, carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM-1), lies in an extended coiled-coil region of the stalk. Small-angle X-ray scattering, electron microscopy and molecular modelling suggest that the UspA1 coiled-coil bends significantly (30°-60°) upon binding CEACAM-1 (Conners et al., 2008). This bending is physiologically relevant as the binding site for CEACAM-1, located in the stalk, would otherwise not be accessible to its ligand due to steric constraints (Conners et al., 2008). Bends and kinks have also been observed for the trimeric autotransporter EmaA (Ruiz et al., 2006), and bending has been postulated for YadA in order to avoid steric clashes upon binding to its ligand collagen (Nummelin et al., 2004).

In large trimeric autotransporters especially, the modular structure may confer a degree of flexibility on the passenger domain. For example, the binding model suggested for Hia and its close homologue Hsf is a two-step mechanism, where the more distally situated, lower-affinity BD2 interacts with the host cell first, after which the membrane-proximal high-affinity BD1 binds to its receptor (Meng et al., 2008). For this to happen, the protein must undergo significant bending, so the passenger domain requires intrinsic flexibility.

MID, the IgD-binding protein from M. catarrhalis, displays an interesting and unusual structural feature. Electron microscopy experiments suggest that the passenger domain doubles back on itself, as gold-labelled antibodies directed against the N- and C-terminal regions of the passenger domain bound close to the membrane. However, internal regions of the passenger domain harbouring the IgD-binding and adhesive activities were found distal from the membrane (Hallström et al., 2008). This implies that the stalk would consist of 6 chains, as each monomer would contribute two antiparallel helices to the stalk. This remarkable arrangement, though following the anchor-stalk-neck-head topology displayed by other trimeric autotransporters, is structurally very different from other trimeric autotransporters for which structural information is available. In these other molecules, the head consists of the N-terminus of the sequence, whereas in MID the protruding head seems to be located roughly in the middle of the sequence.

37

MID also has some other unusual features. The binding site for IgD is apparently tetrameric, rather than trimeric, based on sedimentation equilibrium and native electrophoresis analyses (Nordström et al., 2002). In addition, in contrast to YadA and Hia, the translocation unit is suggested to contain 10 β-strands per monomer, giving a barrel with a total of 30 transmembrane β-strands (Hallström et al., 2008). However, these findings are highly incongruous with the results obtained for other trimeric autotransporters, and so should perhaps be re-evaluated.

2.2.3.3 Mechanism of trimeric autotransportation

The mechanism of trimeric autotransportation has been less studied than classical autotransportation, but is widely assumed to be similar. The models suggested for classical autotransporters could apply also to trimeric autotransporters, though the oligomeric model (here referring to oligomers of trimeric β-barrels) can quite safely be ruled out, as there are no reports of oligomers of trimeric β-barrels. In addition, the structure of the Hia translocation unit, with its obvious parallels to the NalP translocator, suggests that the passenger domain is transported through the pore of the β-barrel. Whether the N- or the C-terminus is extruded first is not entirely clear, but the fact that the entire passenger domain is not required for translocation favours the hairpin model (Meng et al., 2006). The involvement of Omp85 in trimeric autotransportation has not been proven, but as this complex is required for the biogenesis of all studied β-barrel OM proteins, it seems a safe bet to assume that Omp85 also plays a role in the insertion of trimeric autotransporters into the OM. However, as for classical autotransporters, whether Omp85 takes part in translocating the passenger domain across the OM remains an open question.

The first member of the type Vc secretion pathway to be shown to be an autotransporter was Hia (St. Geme & Cutter, 2000). In contrast to many classical autotransporters, the Hia passenger domain is not cleaved from the translocation unit during biogenesis (St. Geme & Cutter, 2000). In fact, no cleavage of the passenger from the translocation domain has been reported for any trimeric autotransporter. However, Hia contains a long signal peptide of 49 residues, similar to the extended signal peptides of some classical autotransporters (St. Geme & Cutter, 2000). This suggests that the N-terminus of Hia is also held by the Sec machinery for an extended period. Several other trimeric autotransporters, including MID, BadA and XadA also contain long signal peptides (Forsgren et al., 2001; Ray et al., 2002; Riess et al., 2004). In contrast to Hia, the signal peptides of YadA and some related proteins, such as Eibs and DsrA, have the more usual number of approximately 25 amino acids (Skurnik & Wolf-Watz, 1989; Elkins et al., 2000; Sandt & Hill, 2000).

As for classical autotransporters, the C-terminal translocation domain contains the autotransporting activity of trimeric autotransporters. The β-barrel of trimeric autotransporters contains a glycine residue (Gly389 in YadA from Y. enterocolitica serotype O:8) that is highly conserved throughout the family, being present in over 95% of trimeric autotransporters – only in a few cases is it replaced by alanine, serine, threonine or asparagine (Hoiczyk et al., 2000; Grosskinsky et al., 2007) (Figure 2.8). Mutating Gly389 results in reduced YadA expression, destabilised trimers and increased

38

periplasmic degradation (Grosskinsky et al., 2007). The effect of the mutation increased with the size of the substituting side chain. Although the translocation domains of trimeric autotransporters have high similarity, it seems that the C-termini of trimeric autotransporter do not cross-trimerise. Coexpression of YadA and Hia did not affect the

adhesive activity of either protein, though cross-trimerisation would most probably have completely wiped out any binding activity (Cotter et al., 2006). The same study suggests that trimerisation of the passenger domain takes place after trimerisation of the translocation domain rather than at the same time, as non-trimerising hetero-oligomers of full-length and truncated passenger domains could be detected on the cell surface in the absence of proteases (Cotter et al., 2006). Surface localisation of the passenger domain naturally requires a functional translocation domain. However, although the translocation domains do not cross-trimerise, it appears that any translocation domain is capable of transporting a heterologous trimeric autotransporter passenger domain, as chimeric proteins consisting of the translocation units of EibA, Hia or UspA1 and the YadA passenger domain displayed functional, trimeric YadA passenger domain on the cell surface (Ackermann et al., 2008).

The only major difference between trimeric and classical autotransportation appears to be the number of chains translocated. In the case of the trimeric autotransporters, three polypeptide chains must cross the OM instead of the single chain of classical autotransporters. If the single-chain model is correct, this presents some mechanistic problems for trimeric autotransporters, especially if the passenger domain is transported as a hairpin as in the C-terminus first version. This would mean that the pore would have to accommodate three hairpins if the three monomers

forming the passenger domain are translocated simultaneously, or one or two hairpins and one or two α-helices if the translocation of the monomers is sequential. An extended polypeptide hairpin has an approximate diameter similar to that of an α-helix, approximately 10 Å, and the dimensions of Hia barrel could accommodate three hairpins (Meng et al., 2006). However, the fit would be a very tight one, so the plausibility of this model is still questionable.

Figure 2.8 Monomer of the Hia translocation unit (PDB 2GR7) showing the position of the conserved glycine (Gly1064 in Hia, Gly389 in YadA) in magenta. In addition, residues in the linker close to the glycine are shown. β-sheets are in blue, the linker helix is in yellow. The figure was prepared with PyMOL (Delano Scientific).

39

2.3 Biology of individual trimeric autotransporters

As mentioned earlier, all characterised trimeric autotransporters have adhesive activity, but in addition they usually have a number of secondary activities (Table 2.3). In the following section I will discuss the biological functions of some of these molecules in detail.

2.3.1 The Yersinia adhesin YadA

At over 100 publications, YadA is by far the best characterised of the trimeric autotransporters and one of the most intensely studied bacterial adhesins. However, several aspects of the biology and biochemistry of YadA still remain to be clarified, and its exact role during infection in vivo is still uncertain. In this section, I will give a concise overview of the current understanding of the biological functions of YadA.

2.3.1.1 Enteropathogenic yersiniae

Three species of yersiniae are pathogenic to humans: Y. enterocolitica, Y. pseudotuberculosis and Y. pestis. The two former species are enteropathogens, causing gastrointestinal infection by ingestion of contaminated food or water. The infamous plague agent Y. pestis, on the other hand, is a zoonotic pathogen and one of the most invasive disease-causing bacteria. This species is transmitted to humans by flea bites, or more rarely by aerosols from individuals suffering from pneumonic plague. It appears that Y. pestis evolved from Y. pseudotuberculosis as recently as 20,000 to 1,500 years ago (Achtman et al., 1999); the difference in virulence and infection route are thus remarkable, but much progress has been made in elucidating the genetic changes that enabled Y. pestis to adopt a different life style and become so devastatingly pathogenic (Chain et al., 2004; Gu et al., 2007).

Compared to Y. pestis, Y. enterocolitica and Y. pseudotuberculosis are much milder, though of course infection by either of these organisms, with associated gastroenteritis and pseudoappendicitis, is highly unpleasant (Bottone, 1997). Both species are abundant in the environment, with Y. enterocolitica being more common in the cooler climate of Scandinavia and North America (Cover & Aber, 1989). Most cases of yersiniosis in humans are caused by Y. enterocolitica. Y. pseudotuberculosis more commonly infects animals, even though swine are a major reservoir for human pathogenic strains of Y. enterocolitica (Bottone, 1997). Both have a similar progression of infection: after ingestion, the bacteria reach the small intestine where they cross the ileal mucosa through M cells, which are transcytotic cells associated with the lymphoid tissue of the gut, Peyer’s patches (Grützkau et al., 1990). Proliferation takes place extracellularly within Peyer’s patches and mesenteric lymph nodes, and the infection is usually contained within these organs, but in severe cases the bacteria may disseminate to the spleen and the liver, and even cause systemic infection (Bottone, 1997). On occasion, the primary infection is

40

Table 2.3 Characterised trimeric autotranporters.

Trimeric autotransporter

Organism(s) Functions Reference

BadA Bartonella henselae Collagen binding, binding to epithelial cells, phagocytosis resistance, proangiogenic factor

Riess et al., 2004

BimA Burkholderia pseudomallei

Actin binding, actin polymerisation, actin-based motility

Stevens et al., 2005

DsrA Haemophilus ducreyi Serum resistance, keratinocyte binding, ECM binding

Cole et al., 2002

EibA, EibE Escherichia coli IgG binding, serum resistance Sandt & Hill, 2000 EibC, EibD, EibF E. coli IgA and IgG binding, serum resistance Sandt & Hill, 2001 EibG E. coli Binding IgA and IgG, chain-like

adhesion to mammalian cells Lu et al., 2006

EmaA Aggregatibacter actinomycetemcomitans

Collagen binding Mintz, 2004

HadA H. influenzae Autoagglutination, ECM binding, epithelial cell binding and invasion

Serruto et al., 2009

Hia, Hsf H. influenzae Epithelial cell binding Barenkamp & St. Geme, 1996

MID Moraxella catarrhalis IgD binding, adhesion to lung and middle ear epithelial cells, biofilm formation

Riesbeck et al., 2006

NadA Neisseria meningitidis Adhesion to and invasion of epithelial cells

Capecchi et al., 2005

NhhA N. meningitidis ECM binding, binding to epithelial cells, serum and phagocytosis resistance

Sjölinder et al., 2008

UpaG E. coli ECM binding, biofilm formation, autoagglutination

Valle et al., 2008

UspA1, UspA2 M. catarrhalis ECM and epithelial cell binding, serum resistance, biofilm formation

Riesbeck et al., 2006

VompA, VompC B. quintata Collagen binding, autoagglutination Zhang et al., 2004 XadA Xanthomonas oryzae

pv. oryzae Leaf attachment and entry Ray et al., 2002

YadA Yersinia enterocolitica, Y. pseudotuberculosis

ECM binding, serum and phagocytosis resistance, autoagglutination, binding to epithelial cells and neutrophils

El Tahir et al., 2000

YadB, YadC Y. pestis Epithelial cell invasion Forman et al., 2008

41

followed by sequelae, of which reactive arthritis is a typical example (Cover & Aber, 1989; Bottone, 1997). Of the two species, Y. pseudotuberculosis is the more invasive, but in contrast to Y. enterocolitica, diaorrhea is not associated with Y. pseudotuberculosis infections (Jani & Chen, 2008).

Both enteropathogenic species of Yersinia express a number of virulence factors, including YadA. YadA is encoded by the virulence plasmid pYV, and its expression is temperature-regulated so that the protein is produced only at +37 °C, the body temperature of the host (Bölin et al., 1982; Skurnik & Toivanen, 1992). YadA is an essential virulence factor of Y. enterocolitica, as evidenced by the fact that knocking out the yadA gene makes the bacteria avirulent (Roggenkamp et al., 1995). In contrast, YadA does not appear to be necessary for the pathogenicity of Y. pseudotuberculosis (Han & Miller, 1997). The yadA gene is also present in Y. pestis, but as a pseudogene due to frame shift within the coding sequence resulting from a single base pair deletion (Rosqvist et al., 1988; Skurnik & Wolf-Watz, 1989). However, two related, chromosomally encoded genes, yadB and yadC, have recently been discovered and shown to produce proteins active in adhesion and invasion (Forman et al., 2008).

2.3.2.2 YadA and ECM binding

After crossing the intestinal mucosa, YadA is the major adhesin expressed by the enteropathogenic yersiniae. In Y. enterocolitica, YadA (YeYadA) is a collagen-binding protein (Emödy et al., 1989; Schulze-Koops et al., 1992). YeYadA binds a number of collagen types, including the fibrillar collagens type I, II, III, V and XI (Emödy et al., 1989; Schulze-Koops et al., 1992), and the network-forming collagen type IV (Flügel et al., 1994), but not the microfibrillar collagen type VI (Schulze-Koops et al., 1992). Immunoblotting and blocking experiments suggest that the binding site for YeYadA resides in the α1 chains of collagens type I and type II, the α3 chain of collagen type XI and both chains of the heterodimeric collagen type V (Schulze-Koops et al., 1992). A more detailed study of the binding sites in the α1(I) and α1(II) chains indicated that the binding site lies in a conserved 134-amino acid region (Schulze-Koops et al., 1995). This binding was not dependent on the triple-helical conformation of collagen. Similarly, the binding site for YeYadA in collagen type IV was reported to reside in the amino-terminal region of the α1(IV) chain, which is part of the tetrameric 7sL fragment of collagen type IV, and not in the triple-helical domain (Flügel et al., 1994).

The YeYadA-collagen interaction is very stable, tolerating extremes of pH and temperature, and is resistant to proteases (Emödy et al., 1989). However, the affinity of recombinant YeYadA for collagen type I is only 0.28 µM as measured by surface plasmon resonance (Nummelin et al., 2004). This apparent discrepancy between the very tight binding observed for whole cells expressing YeYadA and the modest affinity of recombinant YeYadA can be explained by the cooperative effect of multiple YadA molecules coating the cell surface bestowing a high avidity to Y. enterocolitica cells for collagen. Indeed, when expressed, YeYadA is present at very high levels, creating a brush border on the cell surface visible in electron micrographs (Zaleska et al., 1985; Hoiczyk et al., 2000).

42

In addition to collagen, YadA binds other molecules of the ECM, such as fibronectin and laminins (Tertti et al., 1992; Tamm et al., 1993; Flügel et al., 1994). YadA from both Y. enterocolitica and Y. pseudotuberculosis (YpYadA) bound plasma fibronectin immobilised on glass slides (Tertti et al., 1992), though a later study showed that YeYadA binds more tightly to cellular fibronectin (Schulze-Koops et al., 1993). The binding region for YadA in fibronectin has not been characterised, but the YadA-fibronectin interaction is not dependent on the RGD repeats of fibronectins (Schulze-Koops et al., 1993). A recent study suggests that fibronectin, rather than collagen, is the major ligand for YpYadA, whereas YeYadA binds preferentially to collagen (Heise & Dersch, 2006). Similarly, YpYadA does not appear to bind laminin (Heise & Dersch, 2006), but YeYadA binds laminin-1 and laminin-2, though at a lower level than collagen (Tertti et al., 1992; Flügel et al., 1994). YeYadA bound the elastase fragment E1, but not other elastase fragments of laminin-1 (Flügel et al., 1994).

The YadA-mediated ability to bind ECM molecules is necessary for attachment of Y. enterocolitica to the intestinal submucosa (Skurnik et al., 1994). Y. enterocolitica expressing YadA bound to the submucosa of frozen sections, but YadA-negative bacteria did not, nor did Y. enterocolitica cultured at 22 °C, a temperature which represses YadA expression. Furthermore, when the sections were incubated with a bacterial extract and probed with an anti-YadA monoclonal antibody, the observed pattern of binding was the same as for intact YadA-positive cells. The binding could be partially blocked by collagen or laminin, and a combination of these two ECM molecules decreased the binding of the bacteria significantly. Whether this YadA-dependent binding to the submucosa is relevant in vivo is an open question, but one possibility is that binding to ECM of the lamina propria and submucosa might reduce the risk of phagocytosis (discussed further below) and allow the bacteria to multiply before spreading to other tissues (Skurnik et al., 1994).

The ECM-binding activity of YadA is associated with reactive arthritis. Reactive arthritis is a sterile inflammation of the joints that can develop after a triggering infection by a number of bacterial species, including Y. enterocolitica and Y. pseudotuberculosis. YadA is required for the arthritogenicity of Y. enterocolitica in a rat model (Gripenberg-Lerche et al., 1994). Collagen has an important structural role and is particularly exposed in joints, and the collagen-binding ability of YeYadA is implicated in eliciting reactive arthritis. Bacteria expressing a truncation of YadA that did not bind collagen were still arthritogenic, but at a significantly lower level than wild type bacteria and with much milder arthritic symptoms (Gripenberg-Lerche et al., 1995). Skurnik (1995) has hypothesised that polymorphonuclear leukocytes (PMNs) or macrophages presenting either a collagen-binding fragment of YadA or the intact molecule on the cell surface after uptake of bacteria could bind to collagen or other ECM molecules and induce inflammation. There is, however, currently no experimental evidence in support of this.

2.3.2.3 Other virulence-associated functions of YadA

YadA has several other functions associated with pathogenicity apart from binding to ECM molecules. Among other effects, YadA promotes adherence to and invasion of epithelial cells (Heesemann & Grüter, 1987; Bliska et al., 1993). YpYadA in particular

43

can act as an invasin and assist the bacteria in entering epithelial cells (Bliska et al., 1993; Eitel & Dersch, 2002). In contrast, YeYadA does not promote invasion, but favours adherence and mediates a related phenomenon, resistance to phagocytosis by epithelial cells (Heesemann & Grüter, 1987). The receptors for YadA on epithelial cells appear to be β1 integrins, and binding probably takes place through bridging ECM molecules (Eitel & Dersch, 2002). YadA-mediated cell entry and the induction of the pro-inflammatory cytokine interleukin-8 (IL-8) require intracellular signalling events in epithelial cells (Schmid et al., 2004; Eitel et al., 2005). Production of chemoattractant cytokines such as IL-8 may in fact be beneficial for pathogenic yersiniae. Yersiniae are resistant to phagocytosis and killing by professional phagocytes (discussed below), and thus recruitment of leukocytes may lead to tissue disruption and facilitate dissemination of the bacteria, rather than hinder disease (Grassl et al., 2003). A further function of YadA is to inhibit the action of interferon, which normally protects cells from bacterial invasion (Bukholm et al., 1990).

Resistance to phagocytosis by professional phagocytes such PMNs is important for Y. enterocolitica virulence (Lian et al., 1987). YadA is implicated in PMN phagocytosis resistance, and requires binding of the bacteria to the surface of PMNs (China et al., 1994). This presumably docks the bacteria onto the PMN so that the intracellular antiphagocytic effector proteins YopH and YopE can be delivered to the cytoplasm of the PMN by the type III secretion machinery (Ruckdeschel et al., 1996). YadA, in concert with YopH and YopE, thus prevents phagocytosis and killing of yersiniae by professional phagocytes and allows extracellular residence in host tissues. YadA also plays a role in protecting the bacteria from antimicrobial polypeptides produced by granulocytes, as these polypeptides caused a significantly larger decrease in the number of viable cells when exposed to YadA-negative as compared to YadA-expressing bacteria (Visser et al., 1996). The binding of Y. enterocolitica to the submucosa mentioned above may be a further mechanism to avoid phagocytosis, as adherent bacteria are more difficult to ingest by PMNs (Skurnik et al., 1994).

In addition to acting as an adhesion molecule and invasin and resisting phagocytosis, YadA has several other virulence-related functions. One of the first activities assigned to YadA was autoagglutination (Skurnik et al., 1984), and expression of YadA renders the bacteria hydrophobic (Martinez, 1983). Furthermore, YadA is involved in the formation of densely packed microcolonies of Y. enterocolitica when grown in three-dimensional collagen gels, though this phenotype is not dependent on the ability to bind collagen (Freund et al., 2008). The exact role of these aggregation phenomena in yersiniosis is not clear, but it seems reasonable to assume they help protect the bacteria from host attack.

A major virulence-related activity exhibited by enteropathogenic yersiniae is serum resistance (Pai & DeStephano, 1982). Under normal circumstances, incubation in serum leads to the lysis of Gram-negative bacteria through the action of complement, an innate immune defence mechanism. The alternative complement activation pathway is particularly important in defending against Gram-negative pathogens, and resisting the activity of complement is a requirement for invasive pathogenic bacteria (reviewed by Lambris et al., 2008). YadA is the major factor conferring serum resistance to Y. enterocolitica (Balligand et al., 1985), and expressing recombinant YeYadA in E. coli imparts serum resistance to this normally serum-sensitive bacterium (Martinez, 1989). In

44

contrast, Y. pseudotuberculosis is serum resistant at +37 °C even without YadA (Perry & Brubaker, 1983). The complement component C3b leads to opsonisation of the bacteria and the formation of the membrane attack complex, and less C3b is deposited on the cell surface of YeYadA-expressing bacteria than on YeYadA-negative bacteria (Pilz et al., 1992; China et al., 1993). YeYadA-expressing bacteria also promote the conversion of C3b to the inactive iC3b faster than YeYadA-negative bacteria (Pilz et al., 1992). YeYadA seems to increase serum resistance by binding the complement regulatory serum factors H and C4b-binding protein (China et al., 1993; Kirjavainen et al., 2008). Factor H, which inhibits the alternative pathway, binds C3b and promotes its cleavage to iC3b by factor I, and several pathogenic bacteria exploit the protective activity of factor H by binding it to the cell surface (Lambris et al., 2008). YadA binds directly to factor H, apparently along the whole of its length (Biedzka-Sarek et al., 2008a). However, yersiniae also produce the chromosomally encoded adhesin Ail which binds factor H, and serum resistance in Y. enterocolitica appears to be a multifactorial phenomenon (Biedzka-Sarek et al., 2005; Biedzka-Sarek et al., 2008b)

A final activity of YadA is binding to intestinal mucus (Mantle et al., 1989; Paerregaard et al., 1991). Y. enterocolitica binds to intestinal mucus, mucin and brush border vesicles from rabbits (Mantle et al., 1989). The ability to bind mucus correlated strongly with YadA expression (Paerregaard et al., 1991). Compared to rabbit mucin, Y. enterocolitica bound much tighter to purified mucin from human but weaker to rat mucin (Mantle et al., 1989). Binding to purified mucin was reduced by addition of monosaccharides, and the mucin binding was not dependent on the hydrophobic surface of the virulent bacteria (Mantle & Husar, 1993). However, mucin binding only accounted for approximately 40% of the observed binding of Y. enterocolitica to mucus (Mantle et al., 1989). Binding to intestinal mucus could favour colonisation of the gut as this may prevent clearance of the bacteria by abrasive forces, if the rate of replication and mucus penetration is greater than the turnover of the mucus (Mantle & Husar, 1993). Alternatively, this binding may not be a virulence-related function at all but may rather be a host defence mechanism, as preincubation of Y. enterocolitica with mucus or purified mucin decreased the ability of the bacteria to adhere to the intestinal epithelium or brush border vesicles (Mantle et al., 1989; Paerregaard et al., 1991).

2.3.2.4 Mapping of YadA activities

The various activities of YadA have been mapped to different regions in the molecule (Figure 2.9). The collagen-binding activity of Y. enterocolitica YadA is known to reside in the head domain (Tamm et al., 1993). However, the YadA head domain structure does not contain any obvious groove or pocket that would serve as a binding site. Site-directed mutagenesis pinpointed several residues critical for collagen binding, suggesting that the triple-helix of collagen binds diagonally across the binding face (Nummelin et al., 2004). Nevertheless, the detailed mechanism of collagen binding remains unclear. In YpYadA, a 30-amino acid extension at the N-terminus, not present in YeYadA, is responsible for the change of specificity from collagen to fibronectin (Heise & Dersch, 2006). In addition to collagen binding, the head domain of YadA is responsible for autoagglutination (Tamm et

45

al., 1993), and the very N-terminus of the protein (residues 29-81) binds neutrophils (Roggenkamp et al., 1996). An intact head domain is also required for epithelial cell adhesion and induction of IL-8 production (Schmid et al., 2004).

The site for laminin binding has not been elucidated in detail. 22-amino acid deletions in the head domain of YeYadA that abrogate collagen binding do not affect binding of YeYadA to laminin, nor does removing one of the NSVAIGXXS repeats required for collagen binding (Tamm et al., 1993; El Tahir et al., 2000). However, the 30-amino acid extension of YpYadA seems to reduce laminin binding (Heise & Dersch, 2006). Removal of this extension increased the binding to the level of YeYadA. Also the double mutation V133D,N134A, which eliminates collagen binding (Nummelin et al., 2004), abolished binding to laminin (Heise & Dersch, 2006). This latter finding suggests that the head domain does play a role in laminin binding, but the exact region remains to be determined.

The serum resistance properties of YeYadA map to the coiled-coil stalk (Roggenkamp et al., 2003). Short deletions

in the stalk that did not interrupt the periodicity of the coiled-coil did not affect factor H binding, suggesting that factor H binds to several conformationally similar motifs along the stalk (Biedzka-Sarek et al.,

2008b). In fact, deletions in the N-terminal half of the stalk seemed to increase the binding of factor H, which may be because this region of YeYadA contains binding sites for other serum factors that reduce the binding of factor H (Biedzka-Sarek et al., 2008b).

2.3.2 Other ECM-binding trimeric autotransporters

A number of other trimeric autotransporters that bind to molecules of the ECM have been described from a variety of Gram-negative bacteria. Many of these bind to collagen, and a large proportion of these adhesins contain YadA-like repeats in their head domains.

Figure 2.9 Model of YeYadA showing binding regions for different ligands (model from Koretke et al., 2006)

46

2.3.2.1 Trimeric autotransporters from the genus Haemophilus

After YadA, the best-characterised trimeric autotransporter is Hia from non-typeable Haemophilus influenzae. This organism, which is a common resident of the upper respiratory tract in humans, can cause various mucosal infections, including sinusitis, otitis media and chronic bronchitis. Hia was discovered through its ability to mediate the adherence of H. influenzae strains lacking other adhesin molecules to epithelial cells (Barenkamp & St. Geme, 1996). The passenger domain of Hia contains two homologous binding pockets (BD1 & BD2), the higher affinity BD1 at the C-terminal end and the lower-affinity BD2 at the N-terminus of the passenger domain, both of which bind to the same cellular receptor and are required for full adhesive activity (Laarmann et al., 2002). The binding site consists of an acidic pocket contained within a single monomer of trimeric Hia (Yeo et al., 2004). However, the receptor for Hia is still unknown.

In addition to Hia, members of the genus Haemophilus express several other trimeric autotransporters. H. influenzae type b contains a protein highly similar to Hia named Hsf (St. Geme et al., 1996). Indeed, Hsf and Hia are allelic variants, with 81% similarity (St. Geme et al., 1996). Hsf contains three regions homologous to the binding sites of Hia; two of these harboured a similar acidic pocket to the Hia binding sites and displayed adhesive activity, whereas no binding could be detected for the third site situated between the other two (Cotter et al., 2005b). Interestingly, in contrast to Hia, the two binding pockets of Hsf appear to bind independently of one another, as just one site was sufficient to promote wild-type adherence (Cotter et al., 2005b).

A very recently described trimeric autotransporter is HadA of H. influenzae biogroup aegyptius, the causative agent of Brazilian purpuric fever, a form of fulminant septicaemia in children (Serruto et al., 2009). Interestingly, this protein appears to lack a head domain completely, the passenger domain being predicted to consist entirely of an extended coiled-coil. However, the trimeric (interchain) coiled-coil is possibly interrupted by hairpin-like protrusions of dimeric, antiparallel coiled-coils formed by a single chain (Serruto et al., 2009). HadA promotes autoagglutination, ECM binding and adherence and invasion of epithelial cells. A KGD motif, located in the putative dimeric coiled-coil protrusions, appears to be involved in the adhesive activity of HadA (Serruto et al., 2009).

H. ducreyi, which causes the sexually transmitted genital ulcer disease chancroid, produces the multifunctional trimeric autotransporter DsrA (Elkins et al., 2000). This protein confers serum resistance by inhibiting the classical activation pathway of complement, possibly by covering structures on the bacterial surface that would otherwise be epitopes for bactericidal IgM (Abdullah et al., 2005). In addition, DsrA recruits C4-binding protein and vitronectin (Abdullah et al., 2005), as well as promoting binding to keratinocytes (Cole et al., 2002). It also binds fibronectin, and DsrA is the major fibronectin-binding adhesin of H. ducreyi (Leduc et al., 2008). Both the ECM-binding and serum resistance activities of the protein are contained in the 140 C-terminal residues of DsrA (Leduc et al., 2009). This fragment also contains the translocation unit, which is unlikely to contain the adhesive or serum resistance activities, thus suggesting that these activities map to the C-terminal ~70 residues of the DsrA passenger domain.

47

2.3.2.2 The ubiquitous surface proteins of Moraxella catarrhalis

Moraxella catarrhalis, an opportunistically pathogenic bacterium commonly found in the respiratory tract, causes middle ear infections and sinusitis in children and can exacerbate chronic obstructive pulmonary disease in adults (Verduin et al., 2002). Among the most extensively characterised M. catarrhalis OM proteins are the ubiquitous surface proteins UspA1 and UspA2. They share a common epitope of 140 residues, though their sequences are otherwise divergent, with UspA1 reminiscent of Hia and UspA2 more similar to YadA (Helminen et al., 1994; Aebi et al., 1997). A hybrid of these two proteins, designated UspA2H, also exists (Lafontaine et al., 2000). Both UspA1 and UspA2 bind to the ECM components fibronectin and laminin (Tan et al., 2005; Tan et al., 2006), and in addition UspA2 binds vitronectin (McMichael et al., 1998). UspA1 binds to epithelial cells which express CEACAM-1, the receptor for UspA1, and this binding induces apoptosis in the epithelial cells (N'Guessan et al., 2007). In addition, both proteins mediate the adherence of M. catarrhalis to epithelial cells through binding to bridging fibronectin molecules (Tan et al., 2005). UspA2 increases the serum resistance of M. catarrhalis (Aebi et al., 1998). The vitronectin-binding activity of UspA2 is also implicated in serum resistance, as soluble vitronectin inhibits the terminal pathway of the complement system (McMichael et al., 1998). In addition, UspA2 and to a lesser extent UspA1 bind C4-binding protein (Nordström et al., 2004). Interestingly, M. catarrhalis binds the complement factor C3 directly through a non-covalent, ionic interaction, which may prevent C3 being processed to C3b and thus inhibit the activation of the alternative complement pathway (Nordström et al., 2005). This binding is mediated mostly by UspA2, with UspA1 making a minor contribution. UspA1 is also involved in the formation of biofilms (Pearson et al., 2006).

2.3.2.3 Trimeric autotransporters from the genus Bartonella

The genus Bartonella contains zoonotic pathogens that cause several diseases, usually associated with vasculoproliferative disorders. B. henselae and B. quintata, express trimeric autotransporters that bind to ECM molecules. B. henselae, which causes cat scratch disease and tumorous proliferation of endothelial cells in immunocompromised patients, produces the giant trimeric autotransporter BadA, over 3,000 residues in length (Riess et al., 2004). BadA binds ECM molecules and epithelial cells, promotes autoagglutination and has antiphagocytic properties (Riess et al., 2004). In addition, it mediates a proangiogenic host reponse, i.e. tumorous proliferation of endothelial cells. BadA accomplishes this through the activation of hypoxia-inducible factor 1 in host cells and promoting secretion of the proangiogenic cytokines vascular endothelial growth factor and IL-8 (Riess et al., 2004; Kaiser et al., 2008). The head domain (residues 48-469) of BadA is required for binding to collagen, autoagglutination, binding to endothelial cells and eliciting the proangiogenic host response (Kaiser et al., 2008). However, the head domain alone was not sufficient for binding to fibronectin.

The related B. quintata, which causes trench fever, produces the variably expressed OM proteins VompA and VompC (Zhang et al., 2004). In addition, the locus encoding

48

these genes contains up to two other vomp genes, vompb and vompd, which undergo phase variation (switching of surface antigens) during infection. Both VompA and VompC, which contain YadA-like NSVAIGXXS motifs, mediate binding to collagen type IV, with VompC binding tighter. VompA also promotes autoagglutination of B. quintata (Zhang et al., 2004). In contrast to BadA and the Vomps, the role of the homologous protein BrpA from B. vinsonii subsp. arupensis in infection has not yet been elucidated (Gilmore et al., 2005).

2.3.2.4 ECM-binding trimeric autotransporters from other pathogens

A YadA-like collagen adhesin has also been described from Aggregatibacter actinomycetemcomitans (formerly Actinobacillus actinomycetemcomitans), a bacterium commonly associated with periodontitis (Mintz, 2004). This protein, the extracellular matrix adhesin EmaA, mediates the binding of A. actinomycetemcomitans to collagen and forms antenna-like protrusions ~23 nm in length on the cell surface (Mintz, 2004; Ruiz et al., 2006). The collagen-binding activity is located in the ellipsoidal head domain visible in transmission electron micrographs (residues 70-386) (Yu et al., 2008). Mutagenesis studies implicated that Gly162 in one of the ten YadA-like repeat of EmaA is critical for collagen binding (Yu et al., 2008). Gly162 is part of a GIAIG motif, similar to the NSVAIGXXS motifs of YadA, where Gly162 is the first glycine in the GIAIG motif. Mutating Gly162 to serine, the amino acid usually found in this position in both YadA and EmaA, abolished collagen binding. This was probably due to a structural change, as the ellipsoidal head could no longer be detected by electron microscopy (Yu et al., 2008). However, mutating G162 to alanine did not affect collagen binding or the structure of EmaA. The GIAIG motif containing G162 is separated from its neighbouring motifs by 23 residues on either side, whereas normal spacings are between 9 and 13 residues. This unusual arrangement suggests that G162 is situated between two loop structures, and Yu et al. (2008) offer this as an explain as to why substituting the glycine at this position with the larger and more polar serine leads to disruption of the head domain.

Certain uropathogenic E. coli and ExPEC strains contain a gene encoding the trimeric autotransporter UpaG, which has YadA-like repeats in the N-terminal region of the passenger domain (Durant et al., 2007; Valle et al., 2008). Like YadA, it is an ECM-binding adhesin, binding strongly to fibronectin and laminin, and in addition promotes autoagglutination and biofilm formation. UpaG also mediates specific adherence to bladder epithelial cells, possibly because these cells possess high levels of attached fibronectin (Valle et al., 2008). However, the role of UpaG in infection is uncertain, as in vitro the wild type upaG+ E. coli strains do not express the protein. In addition, the constitutive expression of UpaG or knocking out the corresponding gene had no effect on mouse urinary tract colonisation (Valle et al., 2008). However, immunisation with UpaG conferred significant protection against ExPEC challenge in mice, suggesting that UpaG is expressed on the bacterial surface in vivo (Durant et al., 2007).

Neisseria meningitidis, an encapsulated Gram-negative bacterium, is one of the leading causes of meningitis and septicaemia worldwide. N. meningitidis produces several autotransporters, including the trimeric autotransporters NhhA and NadA (Peak et al.,

49

2000; Comanducci et al., 2002; Scarselli et al., 2006). The Neisseria Hia homologue protein NhhA was first identified through its similarity to the classical autotransporter AIDA-I, but is more similar to the H. influenzae trimeric proteins Hia and Hsf, though NhhA is only approximately half the size of Hia (Peak et al., 2000). NhhA binds to epithelial cells and the ECM components laminin and heparan sulphate (Scarselli et al., 2006). It also protects the bacteria from phagocytosis and increases serum resistance (Sjölinder et al., 2008). Adherence to epithelial cells upregulates nnha expression, and NhhA is required for efficient colonisation of the murine nasopharynx (Sjölinder et al., 2008). NadA also promotes binding to and invasion of epithelial cells, and in addition stimulates macrophages ( Capecchi et al., 2005; Franzoso et al., 2008). The head domain of NadA (residues 24-87) is required for bacterial internalisation. However, NadA does not bind ECM components.

2.3.3 Immunoglobulin-binding trimeric autotransporters

2.3.3.1 The Eib proteins from E. coli

Not all ligands of trimeric autotransporters are large ECM proteins. The Escherichia coli immunoglobulin-binding (Eib) proteins were discovered through their ability to bind soluble antibody molecules (Sandt et al., 1997). These proteins are found in certain E. coli reference strains, such as ECOR-2, ECOR-9 and ECOR12, where they were expressed on the cell surface and expression was induced in stationary phase due to nutrient limitation (Sandt et al., 1997). Furthermore, expression was maximal at +37 °C and undetectable at +27 °C. Screening a genomic library of ECOR9 led to the identification of the gene eibA, which coded for a protein with IgG-binding activity (Sandt & Hill, 2000). Southern blotting using eibA as a probe uncovered four other paralogous loci that also produced Ig-binding proteins, designated eibB, eibC, eibD and eibE. In the following text, I will use the notation eibABCDE to refer to multiple members of the family. Sequencing of the genes showed that eibCDE were unique sequences, and that eibB was the product of a recombination event between eibA and eibC (Sandt & Hill, 2000). The gene products of eibACDE bound IgG Fc (the constant portion or “stalk” of the Y-shaped IgG; short for ‘fragment crystallisable’). Interestingly, genetic analysis suggested that the eib genes were part of prophage sequences, and indeed the eib sequences could be transferred to an eib-negative strain by irradiating the ECOR-9 stain with UV light and then exposing the non-binding strain to supernatant from UV-induced cells (Sandt & Hill, 2000). However, expression of Eib proteins was very weak in the resulting lysogenic strains. This is probably because the non-binding strain lacked the activating factor IbrAB, found in the Ig-binding ECOR strains and which possibly regulates eib expression at the level of transcription (Sandt et al., 2002).

Further analysis of ECOR-2 uncovered an IgA-binding Eib, designated EibF, which also bound IgG Fc (Sandt & Hill, 2001). EibCD also bound IgA, but EibAE failed to do so, and all Eibs were negative for IgE and IgM binding (Sandt & Hill, 2001). The IgG-

50

binding region of EibA and EibF is located in the 141 C-terminal residues of the sequence, but the IgA-binding activity of EibF is located separately in a 100-residue stretch in the middle of the protein (residues 181-280) (Sandt & Hill, 2001).

Though the ECOR strains from which the Eibs were cloned are not reported to be pathogenic, the properties of the Eib proteins suggest that they could be involved in virulence functions. The ability to bind Igs is widespread among pathogenic bacteria, and although the reason for this remains somewhat unclear, there is some evidence that covering the surface of the cell with host-derived proteins aids pathogens in evading host immune responses (Lambris et al., 2008). In addition, the Eib proteins increase the serum resistance of the bacteria (Sandt & Hill, 2000; Sandt & Hill, 2001). These attributes suggest that Eibs could, in principle, be virulence factors.

In accordance with this view, a sixth member of the Eib family was discovered from a shiga-toxin producing E. coli (STEC) strain of serogroup O91, designated EibG. This protein, in addition to binding IgG Fc and IgA, is responsible for a distinctive “chain-like adhesion” phenotype when adhering to mammalian cells (Lu et al., 2006). ECOR-2 and ECOR-9 also bind to mammalian cells, though they do not display the chain-like adhesion seen for STEC O91 (Lu et al., 2006).

A recent study has questioned the function of Eibs as receptors for IgG Fc (Ghumra & Pleass, 2007). Despite extensive efforts, the authors could not replicate the non-immune binding of IgG Fc in cell lysates, and suggested instead that the observed binding responses in the earlier studies were due to the polyclonal antibody preparations used. Their explanation was that these preparations contained antibodies specific to epitopes on the Eib proteins, and thus any observed binding was the result of an immunological antigen-antibody reaction. The conclusion was that the Eibs are not receptors for IgG Fc, and that E. coli, for which the Eibs are the only Fc-binding proteins reported, does not contain such receptors.

2.3.3.2 The Moraxella IgD-binding protein

M. catarrhalis also expresses an immunoglobulin-binding trimeric autotransporter, but one that binds IgD rather than the more abundant IgA or IgG classes (Forsgren et al., 2001; Hallström et al., 2008). This is the Moraxella IgD-binding protein MID, also referred to as Hag for haemagglutinin (Pearson et al., 2002). The IgD-binding region is located in the central part of this 200 kDa protein, between residues 962-1200 (Nordström et al., 2002). MID is a potent activator of IgD+ B cells, and the IgD-binding region alone, when conjugated to sepharose beads, was enough to activate B cells (Gjörloff Wingren et al., 2002).

In addition to binding IgD, MID can also agglutinate erythrocytes and bind to lung and middle ear epithelial cells (Forsgren et al., 2003; Holm et al., 2003). In addition, MID mediates binding to type IV collagen. The region responsible for these adhesive phenomena is adjacent to but distinct from the IgD-binding region, but the YadA-like β-roll region at the N-terminus of the protein is not needed for adhesion (Bullard et al., 2007). MID-deficient bacteria were cleared more effectively from mouse lungs than MID-expressing bacteria, suggesting that the adhesive activity of MID is relevant in vivo

51

(Forsgren et al., 2004). However, it appears that expression of MID hinders biofilm formation in M. catarrhalis (Pearson et al., 2006).

2.3.4 Other trimeric autotransporters

2.3.4.1 Trimeric autotransporters in plant pathogens

Though most of the characterised trimeric autotransporters are from animal pathogens and bind to the ECM, a few have been described from bacteria that infect plants. Based on genomic sequencing, trimeric autotransporters occur in at least in the genera Ralstonia, Xanthomonas and Xylella (Cotter et al., 2005a). The prototype of trimeric autotransporters from phytopathogens is XadA, an adhesin from Xanthomonas oryzae pv. Oryzae; this is the only one currently characterised. X. oryzae pv. oryzae causes leaf blight, a serious disease of rice. This species expresses XadA in minimal medium, as this environment may mimic the xylem sap in which the organism usually resides (Ray et al., 2002). XadA is a large (1,265 residues) protein with regions of homology to YadA and Hia (Ray et al., 2002). XadA is required for full virulence and normal colony morphology in X. oryzae pv. oryzae, and probably functions as an adhesin, but its ligand is as yet unidentified (Ray et al., 2002). XadA is also present in the related pathogens X. anoxopodis pv. citri and X. campestris pv. campestris (da Silva et al., 2002).

X. oryzae pv. oryzae also contains genes for two other trimeric autotransporters, a paralogue of XadA designated XadB and a YadA-like protein called YapH (Das et al., 2009). XadA, XadB and YapH have a role in establishing disease, as shown by knock-out studies. XadA appears to be required in the early phases, i.e. leaf attachment and entry, whereas YapH seems to be important in later stages once the bacteria are growing inside the xylem vessels (Das et al., 2009). Similarly to XadA, XadB is needed during early stages of infection, but it plays a subsidiary role to XadA (Das et al., 2009).

Xylella fastidiosa, a xylem-limited phytopathogen which causes several economically important plant diseases such as citrus variegated chlorosis and Pierce’s disease in grapevine, contains a homologue of XadA which is involved in virulence and biofilm formation by mediating the attachment of bacteria to abiotic surfaces (Feil et al., 2007). This protein and other afimbrial adhesins present in the organism might also be required for binding to tissues of its vector, the sharpshooter leafhopper (Lambais et al., 2000).

2.3.4.2 BimA: a trimeric autotransporter for intracellular motility

Burkholderia pseudomallei is a facultative intracellular pathogen that causes the severe invasive disease melioidosis in humans and animals. B. pseudomallei has evolved the ability to exploit the cytoskeleton of host cells for intracellular motility by polymerising and consequently moving along actin filaments (Stevens & Galyov, 2004). Remarkably, the protein responsible for this is the trimeric autotransporter Burkholderia intracellular

52

motility protein A (BimA) (Stevens et al., 2005). BimA contains proline-rich segments that are typically found in activators of the Arp2/3 complex, a cellular factor that nucleates actin polymerisation and regulates the formation of actin networks (Welch & Mullins, 2002). However, BimA does not appear to interact with Arp2/3 but displays a weak actin-polymerising activity independent of this complex in vitro (Stevens et al., 2005). BimA also has regions similar to Wiskott-Aldritch syndrome protein homology domain-2, which is known to bind actin monomers (Paunola et al., 2002). Indeed, BimA binds directly to actin monomers in vitro and mutations of bimA abolish the actin-based motility of B. pseudomallei. Furthermore, BimA localises to the bacterial pole where actin polymerisation and tail formation take place (Stevens et al., 2005). BimA orthologues are present in the related species B. mallei and B. thailandesis and can restore actin-based motility in a bimA mutant of B. pseudomallei (Stevens et al., 2005).

53

3. Aims of the study

The main aim of this study was to characterise the interaction of YadA with collagen using structural, biochemical and biophysical techniques. A further aim was to crystallise and solve the structure of one of the immunoglobulin-binding Eib proteins. A third aim was to crystallise the translocation unit of a trimeric autotransporter to gain insight into the mechanism of outer membrane translocation. Specific aims were:

• To investigate the binding of YadA to model collagen-mimicking peptides.

• To determine the binding site(s) for YadA in collagen.

• To obtain a crystal structure of YadA in complex with a collagenous

peptide.

• To crystallise and solve the structure of a translocation unit of SadA.

• To show that the Eibs are receptors for IgG Fc.

• To map the binding regions in the Eibs for IgG and IgA.

• To solve the crystal structure of the entire passenger domain or an Ig-

binding fragment of an Eib.

54

4. Materials and methods

In this section, I will summarise the methods used in the experiments. For a more detailed description of the methods, I refer the reader to the individual studies.

4.1 Bacterial strains

Yersinia enterocolitica serotype O:3 strain 6471/76 expresses wild type YadA. The strain YeO3-c is cured of the virulence plasmid pYV, and is thus a YadA-negative strain (Skurnik, 1984). Escherichia coli K-12 strains DH5α and JM109 are standard laboratory cloning strains. The expression host E. coli M15(pREP) is from QIAGEN. The expression strain E. coli BL21(DE3) (Novagen) expresses T7 polymerase under the inducible lacUV5 promoter, and is thus suitable for use with vectors of the pET series (Novagen).

4.2 DNA techniques

4.2.1 Cloning of expression constructs

The cloning of Y. enterocolitica O:3 YadA26-241 and YadA24-378 has been described before (Nummelin et al., 2002). I amplified the coding sequences of the passenger domains (or fragments thereof) of the Eib proteins by polymerase chain reaction (PCR) according to standard protocols using the proofreading enzyme Pfu (Fermentas). The primer sequences used for amplification are given in the original studies (Studies III & IV). I digested the resulting PCR products with appropriate restriction enzymes and ligated them into the expression vector pET28b+ (Novagen), which had also been digested with the same enzymes and dephosphorylated with calf intestinal alkaline phosphatase (Fermentas). I transformed the ligation mix into chemically competent DH5α cells and selected for transformants on Luria-Bertani (LB) agar supplemented with kanamycin (25 µg/ml). To verify cloning success, I extracted plasmid DNA from 5 ml overnight cultures grown in LB + kanamycin (25 µg/ml) and sequenced the insert (performed as a service either at the DNA sequencing laboratory at the Institute of Biotechnology, University of Helsinki or by the company Eurofins MWG Operon, Ebersberg, Germany).

The passenger domain of EibA was ligated into the BamHI and SalI sites of the expression vector pQE30 (QIAGEN), transformed into JM109 and transformants were selected on LB + ampicillin (100 µg/ml). The cloning was verified as above.

55

4.2.2 Site-directed mutagenesis

The production of YadA surface mutants has been described before (Nummelin et al., 2004). Site-directed mutagenesis to produce point, deletion and truncation mutants of EibD was carried out according to the Quickchange® protocol (Stratagene). The primer sequences used for this are given in the relevant study (Study IV). To verify that the proteins retained a native tertiary structure, far-UV circular dichroism (CD) spectra were recorded on a Jasco J-715 instrument. The final spectra were the average of five separate measurements and the values obtained were converted to molar ellipticities (θ). For details, see Studies I & IV.

4.3 Recombinant protein expression and purification

4.3.1 Protein expression

YadA26-241, YadA24-378, and EibA28-315 were produced in M15(pREP4). Cells were grown to mid-log phase at +37 °C in terrific broth (TB) medium containing ampicillin (150 µl/ml) and kanamycin (25 µg/ml), after which I induced expression by adding isopropyl thiogalactoside (IPTG) to a final concentration of 1 mM. The cells were allowed to grow a further 3 hours, after which I harvested them by centrifuging for 10 minutes at 4000 x g and resuspended in phosphate-buffered saline (PBS; 20 mM sodium phosphate, 150 mM NaCl, pH 7.4) or buffer 1 (20 mM sodium phosphate, 300 mM NaCl, pH 8.0).

For Eib protein (other than EibA) expression, I transformed the verified expression constructs into the production strain BL21(DE3). For expression, I grew the cells either to mid-log phase at +37 °C in TB medium supplemented with kanamycin (25 µg/ml) and 0.5% (w/v) glucose and induced with IPTG as above, or overnight at +37 °C in autoinducing ZYP-5052 medium (Studier, 2005) containing kanamycin at 100 µg/ml.

4.3.2 Protein purification

YadA and Eib passenger domains or passenger domain fragments were all initially purified by nickel-affinity chromatography. I broke the cells by sonication and clarified the lysate by centrifuging 30 minutes at 18 000 x g. The clarified lysate was then applied to a Ni column (either Ni-nitrilotriacetic acid (NTA) from QIAGEN or Ni-tris-carboxymethyl ethylene diamine (TED) from Macherey-Nagel) and the bound hexahistidine-tagged protein eluted with imidazole.

For YadA24-378 and YadA surface mutants, the nickel affinity chromatography was the only purification step. For the Eib proteins, I digested the eluate of the first Ni step with 20 units of thrombin (Sigma) overnight at +37 °C to remove the hexahistidine tag. I then submitted the digested protein to a second round of Ni purification and collected the flow-through containing the digested protein.

56

YadA26-241 and the Eib proteins were further purified by size exclusion chromatography using a Superdex 200™ 26/60 column on the Äkta Purifier system (both from GE Healthcare). The column was equilibrated with Tris-buffered saline (TBS; 10 mM Tris-HCl, 150 mM NaCl, pH 7.4). I collected the fractions corresponding to the trimeric protein, concentrated the protein using a centrifugal filter device (Amicon concentrator from Millipore) and estimated the concentration by measuring the absorbance at 280 nm.

For small-scale preparation of EibD mutants, I used Ni-iminodiacetic acid columns (Protino Ni-IDA 2000; Macherey-Nagel) as the only purification step.

The translocation unit of SadA (for Salmonella adhesin A (Hernandez Alvarez et al., 2008)) was produced and purified by our collaborator Dr. Dirk Linke (Max Planck Institute for Developmental Biology, Tübingen) using a phase separation technique, as described in Wollmann et al. (2006). Upon receiving it, I concentrated the protein using a centrifugal filter device (Amicon from Millipore). To remove excess detergent or to exchange detergent, I dialysed the protein in a microdialysis button against buffer (20 mM Tris-HCl pH 8.0, 1 mM EDTA, 300 mM NaCl) containing 1% octyl polyoxyethylene or 1% octyl tetraoxyethylene (Bachem Chemicals, Germany). After overnight dialysis, the protein was recovered and the concentration estimated using a bicinchoninic acid kit (Sigma).

4.3.3 Cell lysate preparation

To prepare cell lysates containing EibD fragments, I inoculated the appropriate strains in 5 ml autoinducing ZYP-5052 medium (Studier, 2005) with kanamycin (100 µg/ml) and grew overnight at +37 °C. The following morning, I collected the cells (centrifugation at 3500 x g for 10 minutes) and resuspended in 2 ml of PBS and broke the cells by sonication. After centrifuging as above, I transferred the supernatant (the cell lysate) to a new tube.

4.4 Collagens, collagen-like peptides and immunoglobulins

4.4.1 Collagens and collagen-mimicking peptides

Collagens type I, type II, type III and type IV were purchased from Sigma (Study I). Monomeric collagen type I used in Study III was from Devro, Stirling, UK. Fibrous equine (Ethicon) collagen was from Ethicon Corp., Somerville, NJ, USA.

To produce cyanogen bromide cleavage fragments of bovine collagen type II, 5 mg of collagen was incubated with a single crystal of cyanogen bromide (CNBr; Sigma) in 70% formic acid overnight at room temperature (RT). The resulting CNBr fragments were separated by SDS-PAGE and extracted by sonication using a modification of the protocol

57

by Retamal et al. (1999) . The correct sizes of the fragments were ascertained by mass spectrometry.

The collagen-like peptides (POG)10, (PPG)10 and (POG)5 (where ‘O’ stands for hydroxyproline) were purchased from Peptides International Inc., Lousiville, KY, USA. The peptides T3-785 (sequence (POG)3-ITG-ARG-LAG-(POG)4) and Gly- (sequence (POG)4-PO-(POG)5) were either produced by our collaborators as described in Long et al. (1993) and Kramer et al. (1999)(Long et al., 1993; Kramer et al., 1999) or purchased from Innovagen AB, Lund, Sweden. The synthesis of the Toolkit peptides has been described before (Raynal et al., 2006; Konitsiotis et al., 2008), as has the synthesis of collagen-related peptide (CRP) (Morton et al., 1995), the peptides GPP10 (Raynal et al., 2006), GFOGER (Knight et al., 2000), and the GPO1–GPO6 series (Smethurst et al., 2007). Our collaborators in Cambridge synthesised the GPO6a–GPO6j and GKO–GPO peptides using essentially the same protocol as for the GPO1–GPO6 (Smethurst et al., 2007). The cyclic GPO peptides were first synthesised as linear precursors in the same way as the GPO peptides above. For the full method for cyclisation, see Study II. The sequences of the peptides are given in the corresponding studies (Studies I & II).

4.4.2 Antibodies

Human IgG, IgG Fc and IgG Fab (κ-chain) were from Bethyl Laboratories Inc., Montgomery, TX, USA. Human IgA (from serum) was either from Bethyl Laboratories or Sigma. We purchased horseradish peroxidase (HRP)-conjugated human IgG Fc and IgA, and rabbit γ-globulin, from Jackson Immunoresearch Laboratories Inc., Newmarket, UK. Mouse IgG isotype control antibodies were from Exbio Praha a.s., Prague, Czech Republic. Goat anti-mouse-HRP was from GE Healthcare or Santa Cruz Biotechnology Inc., CA, USA. For YadA detection in ELISA experiments, we used the monoclonal anti-YadA antibody 3G12 (Skurnik et al., 1994).

To produce deglycosylated IgG, IgG Fc and IgA, I digested 100 µg of each with both peptide N-glycosidase F (PNGase F) and endoglycosidase H (both from New England Biolabs) overnight at +37 °C. The results were assessed by SDS-PAGE. PNGase F gave the better result, with complete removal of the sugar moiety, so I used the Igs digested with PNGase F for enzyme-linked immunosorbent assays (ELISAs) and surface plasmon resonance (SPR) experiments.

4.5 Binding studies

4.5.1 ELISA

I used ELISA to study the binding of YadA or YadA-expressing bacteria to collagens and collagen-like peptides and the binding of immunoglobulins to Eibs. In the YadA experiments, wells of a microtitre plate were coated with collagen or peptide. I blocked the

58

wells with bovine serum albumin (BSA) and added either YadA or bacteria. The wells were washed with PBS, and the 3G12 antibody was added (diluted in PBS). After incubation, I washed the wells as above and added the secondary antibody (anti-mouse-HRP). The wells were washed thoroughly. For detection, I added substrate solution (either O-phenylene diamine (OPD) or tetramethylbenzidine (TMB)), allowed colour to develop, and then stopped the reaction by adding 2.5 M H2SO4. I measured absorbance at 450 nm (TMB) or 490 nm (OPD). All steps were carried out at RT. For full details, see the relevant studies (Studies I & II).

For detecting Eibs, I coated plates directly with purified Eib proteins or Eib cell lysate. The wells were blocked with BSA, and HRP-conjugated immunoglobulin was added. The wells were washed thoroughly with PBS, and detection was performed as above. To determine binding of Eibs to different IgG isotypes and deglycosylated IgG fragments, I coated wells with Ig, blocked and then added Eibs. I then washed the wells, added HRP-conjugated antibody, washed again and detected as above.

For blocking experiments, I coated wells with Eibs and blocked with BSA. I then added IgG fragments at different concentrations. The wells were washed, after which HRP-conjugated antibody was added. Detection was performed as above. For full details, see the relevant studies (Studies III & IV).

4.5.3 Immunoblots

Binding of Igs by Eibs was assessed by immunoblotting. I separated 1 µg of Eibs in SDS-PAGE and transferred the proteins to a Hybond C nitrocellulose membrane (GE Healthcare). After transfer, I blocked the membrane with 1% fat-free milk powder (in PBS) and probed with anti-mouse-HRP, IgG Fc-HRP or IgA-HRP. After washing, the binding was visualised using enhanced chemiluminescence (ECL) detection (ECL plus kit from GE Healthcare) and exposing to photographic film (ECL Hyperfilm, GE Healthcare).

For blotting of native Eibs, I used dot blotting. 1 µg of purified protein or 10 µl of cell lysate was applied to a Hybond C nitrocellulose membrane, either by applying the protein by hand or using a vacuum dot blot device (from VWR). Once the protein had adsorbed to the surface of the membrane, the membrane was blocked and probed as above. I detected the binding using the ECL plus kit and exposing either to film or a cooled CCD camera (ChemiDoc XRS from Bio-Rad). For more detailed protocols, see the relevant studies (Studies III & IV).

4.5.4 SPR measurements

SPR experiments were performed either on a Biacore2000™ or a Biacore T100™ (GE Healthcare) instrument. The ligands were coupled to the surface of a CM5 sensor chip by amine coupling according to the manufacturer’s instructions. To determine dissociation constants, the analytes were passed over the surface of the chip at a range of concentrations. The contact and dissociation times were both 5 minutes. For evaluation,

59

we subtracted the response from an empty reference channel from the sensorgrams. As I was unable to obtain a good fit for a kinetic 1:1 binding model in the Biacore evaluation software (GE Healthcare), I calculated the steady-state affinity values to obtain an apparent dissociation constant. For more details, see the relevant studies (Studies I, III & IV).

4.5.4 Isothermal titration calorimetry (ITC)

ITC measurements were performed on a VP-ITC machine from GE Healthcare. Degassed YadA26-241 was put into the measurement cell at 4.5 µM per trimer in PBS. The collagen model peptide (POG)10 was added to the syringe at 125 µM of trimer and 35 injections of 3 µl were added to the cell at +25 °C with intervals of 240 seconds. For the peptides (PPG)10 and T3-785, the concentration was 350 µM for the trimeric peptide and 50 µM for YadA26-241. Gly- was added at a concentration of 600 µM (per monomer) and YadA26-241 at 50 µM. The runs other than (POG)10 were at +21 °C. I analysed the results in the programme Origin7. For full details, see Study I.

4.5.5 Gel shift assays

I used gel shift assays to assess the binding of YadA26-241 to the model peptides (POG)10, (PPG)10, T3-785 and Gly-. 5 µg YadA26-241 was mixed with an equal amount of peptide and run in native PAGE. To control for temperature effects, I ran gels at both RT and +5 °C. To determine the effect of peptide concentration on the migration of YadA in native PAGE, we mixed a constant amount of YadA with an increasing concentration of peptide. For details of the protocol, see Study I.

4.6 Crystallography

4.6.1 Protein crystallisation

Crystallisation was done by vapour diffusion using the sitting drop technique. To find preliminary crystallisation conditions, I used a standard sparse matrix approach using the Helsinki Random Screens I and II (http://www.biocenter.helsinki.fi/bi/xray/automation/services.htm). Preliminary screening was done using a Cartesian Microsys nanopipettor to set up plates with 100 nl + 100 nl drops. To optimise hit conditions, we constructed grid screens around the original condition, varying precipitant concentrations, pH and protein concentration. Once promising crystallisation conditions were found, larger drops were set up either by hand (1 µl +1 µl or 2 µl + 2 µl drops) or using the nanopipettor (0.5 µl + 0.5 µl drops).

60

To crystallise YadA26-241 in complex with a collagen-mimicking peptide, I used two approaches. In the first, I mixed 0.5 mg of YadA26-241 with a large molar excess (1 mg) of (POG)10 or (PPG)10. This solution was then run through a size exclusion column (Superdex™ 10/300; GE Healthcare). The fractions corresponding to non-aggregated complex where pooled and concentrated using an Amicon centrifugal filter device (Millipore). Crystallisation plates were then set up.

The other approach was to mix YadA26-241 with peptide at an approximately 5-fold molar excess of the peptide, and use the mix directly for crystallisation. As controls, I included drops with only either peptide or YadA26-241.

4.6.2 Diffraction data collection, data processing and structure solution

For data collection, crystals were frozen by flash freezing in liquid nitrogen. Where required, glycerol was mixed with the well solution to a final concentration of 10-15% and this solution was used as a cryoprotectant.

Diffraction data were collected at the European Synchrotron Radiation Facility beamline BM14 using a MARCCD 225 detector. The images were processed with Mosflm (Leslie, 1992) and analysed with programme the Pointless in CCP4 version 6.1.1 (1994). Afterwards the data were scaled in the programme Scala (Evans, 2006) of the CCP4 suite.

Initial molecular replacement was done using the CaspR web server (Claude et al., 2004). Coot was used for manual model building (Emsley & Cowtan, 2004) with refinement rounds performed in Refmac (Vagin et al., 2004). The final refinement round and automated water picking was performed in Phenix (Adams et al., 2002). The structure quality was assessed with Molprobity (Davis et al., 2004).

61

5. Results and discussion

5.1. The YadA-collagen interaction (Studies I & II)

5.1.1 A triple-helical conformation is necessary for YadA binding (Study I)

I investigated the binding of YadA to four collagen model peptides: (POG)10, (PPG)10, T3-785 and Gly-. (POG)10 acts as a model for a triple-helical conformation, forming a very stable collagenous triple helix (Sakakibara et al., 1973). However, the superhelix formed by (POG)10 is somewhat tighter than suggested for native collagen based on fibre diffraction experiments (Berisio et al., 2000). (PPG)10 forms a similar triple-helix to (POG)10, but lacks the hydroxyl groups of (POG)10 and is consequently less stable (Kramer et al., 1998). T3-785 contains a stretch of collagen type III sequence, which is closer to the level of supercoiling postulated for native collagen (Kramer et al., 1999). Gly- is essentially a control peptide: it lacks one of the absolutely required glycines and therefore does not form a triple-helix (Long et al., 1993).

I utilised a variety of techniques to study the interaction of YadA with these peptides, including SPR, ITC, ELISA and gel shift assays. (POG)10 bound strongly to YadA in all these methods. However, the binding of YadA to Gly- was barely detectable. Gly- is non-triple-helical; YadA therefore requires a triple-helical conformation for binding, as the only difference in sequence between (POG)10 and Gly- is the single glycine missing in the latter peptide.

(PPG)10 and T3-785 both bind YadA with intermediate affinities. Both of these are triple-helical peptides at RT. (PPG)10 bound YadA more strongly than T3-785 as measured by SPR, though both exhibited clear binding compared to Gly- (Figure 5.1). In the gel shift assays, both (POG)10 and T3-785 aggregate YadA, whereas (PPG)10 does not. This aggregation effect is concentration-dependent, and disappears at a large excess of peptide. As the only difference between (POG)10 and (PPG)10 are the hydroxyl groups of hydroxyproline, we conclude that hydroxyprolines promote aggregation of YadA. T3-785 contains POG repeats, which account for its aggregating YadA. Although (PPG)10 did not aggregate YadA, it clearly bound YadA, causing a small but detectable shift in the migration pattern of YadA. In addition, the staining of the YadA band intensified as the concentration of (PPG)10 increased, indicating an increase in the amount of bound peptide.

The affinity of YadA for (POG)10 is 0.17 µM by SPR and 0.28 µM by ITC (Table 5.1). These values are very close to the affinity of YadA for native collagen measured by SPR. The affinity of (PPG)10 by ITC was 10 µM, i.e. two orders of magnitude lower than that of (POG)10. An association constant could not be calculated for T3-785 or Gly-; however, the binding isotherms for the former indicated an endothermic reaction, as for (POG)10 and (PPG)10 (Figure 5.2). In contrast, the reaction for Gly- was exothermic and close to the heat of dilution, indicating no binding.

62

Figure 5.1 Binding of collagen model peptides to immobilised YadA by SPR.

To show that (POG)10 binds YadA in the same way as collagen type I, the binding of surface residue mutants of YadA to immobilised (POG)10 was investigated by SPR. These mutants had previously been used to study the binding of YadA to collagen type I (Nummelin et al., 2004). The results were similar in that mutants which had a large effect on collagen binding also bound (POG)10 poorly. This suggests that YadA binds to collagen and (POG)10 by the same mechanism.

Table 5.1 Affinities of YadA to collagen and collagen model peptides. In the SPR experiments, collagen type I or (POG)10 were coupled to the sensor chip and soluble YadA26-241 was run over the surface at different concentrations. The steady-state affinity values were calculated assuming a one-to-one binding model. In ITC, YadA26-241 was placed in the measurement cell and collagen model peptides were added by injection. The association constants were calculated from the integrated enthalpy values in the program Origin7 (GE Healthcare) assuming a one-site binding model.

Collagen type I (POG)10 (PPG)10

SPR (µM ± SEM) 0.28 ± 0.06* 0.17 ± 0.02 - ITC (µM ± SEM) - 0.28 ± 0.05 10 ± 0.08

*from (Nummelin et al., 2004)

63

Figure 5.2 Isotherms and integrated enthalpy values of collagen model peptides binding to YadA measured by ITC. A) (POG)10 B) (PPG)10 C) T3-785 D) Gly-.

5.1.2 YadA binding determinants in fibrillar collagens (Studies I & II)

5.1.2.1 Collagens do not contain a single binding site for YadA (Studies I & II)

Previous studies suggested that there would be a single, specific binding site for YadA in collagens type I and type II (Schulze-Koops et al., 1992; Schulze-Koops et al., 1995). However, my results with the collagen model peptides indicated that YadA binds to triple-helical collagenous conformation but without sequence specificity. I therefore investigated whether YadA could bind to several sites in collagen type II. I fragmented collagen type II with CNBr and examined the binding of YadA to the two largest fragments (CB10 and

64

CB11). CB10 contains the putative YadA-binding site (Schulze-Koops et al., 1995). However, I found that YadA bound to both fragments in an ELISA experiment, showing that YadA has more than one binding site in collagen type II.

To further study the binding of YadA to collagens, I utilised the collagen Toolkits. This library of triple-helical peptides span the sequences of homotrimeric human collagens type II and type III (Raynal et al., 2006; Konitsiotis et al., 2008). YadA bound promiscuously to the toolkits with widely varying affinities, showing conclusively that there is no single high-affinity binding site (Figure 5.3). However, YadA did not bind all the peptides, indicating that a triple-helical conformation alone is not sufficient for binding.

Figure 5.3 Binding of YadA to collagen Toolkits by ELISA. Wells of microtitre plates were coated with Toolkit peptides and probed with YadA. Collagens of the appropriate type, GPP10 and GFOGER (Knight et al., 2000) were included as controls. The columns show the mean absorbance at 450 nm from three replicate wells; error bars denote SEM. In Toolkit II, the peptides corresponding to the major CNBr fragments (CB10 and CB11) are shown.

65

5.1.2.2 YadA binds to collagen regions with many hydroxyprolines and few charges (Study II)

Unlike human collagen-binding proteins, which bind to a single or just a few peptides (reviewed by Farndale et al., 2008), YadA bound with a continuum of affinites to the Toolkits and unambiguous classification of peptides into high-binders and non-binders was not possible. The sequences of high-binding peptides (i.e. peptides with a binding response at the level of collagen or above) did not contain a clear binding sequence or motif. Therefore, several statistical analyses were performed on the binding data. Hydroxyproline-containing X-X' dipeptides, particularly PO, were enriched in the high-binding set. In contrast, charged dipeptides occurred less frequently than expected. Conversely, low-binders contained a higher proportion of the charged dipeptides. Binding correlated strongly with the hydroxyproline content of the peptides (Pearson correlation; p < 0.0001). Proline and phenylalanine also correlated positively with binding (p < 0.01). The charged residues lysine, aspartic acid and glutamic acid all had a strong negative correlation (p < 0.0001 for lysine, p < 0.05 for aspartic and glutamic acid).

In addition, the effect of other peptide properties on binding was examined. Hydrophobicity values for the Toolkit peptides were calculated using the scale of Black & Mould (Black & Mould, 1991) as this included a value for hydroxyproline. Binding correlated positively with peptide hydrophobicity (p = 0.0012), and also with the number of imino acids and GPO repeats in the peptides (p < 0.0001). A high content of charged residues had a negative impact on YadA binding (p < 0.0005), but the polarity of the net charge did not have an effect (p = 0.4). However, peptides having large number of charged residues but a net charge close to zero bound YadA better than peptides with a high net charge. This suggests that neutralising charges can to some degree compensate for a large proportion of charged residues.

I wished to determine whether the number and position of GPO repeats had an effect on YadA binding. To do this, I investigated the binding of YadA to a series of peptides with a varying number and spacing of GPO repeats in a GPP background. A single GPO repeat is sufficient to promote tight YadA binding, and the spacing between GPO repeats did not have an effect. A single GPO triplet was also enough to promote YadA aggregation in a gel shift assay. Placing GKO triplets around the GPO repeats had a strong inhibitory effect on YadA binding. Even a spacing of six GPO repeats between GKO triplets was not enough to promote tight (i.e. GPO6-level) binding. However, increasing the spacing between the GKO triplets increased the binding in a graded fashion.

Based on these results, we concluded that YadA binds promiscuously to stretches of collagen that have a high hydroxyproline-to-charged residue ratio. However, no specific sequence is needed. A triple-helical conformation is necessary but not sufficient for YadA binding. Based on the experiments with the GKO-GPO peptides, a stretch of at least six G-X-X' triplets is needed to form a strong YadA binding site. This is consistent with a previously proposed model for YadA binding to a collagen-like peptide based on mutagenesis experiments (Nummelin et al., 2004). In this model, YadA makes contacts to a stretch of seven POG repeats (Figure 5.4). A binding site of this length is significantly longer than those proposed for human collagen-binding proteins, where contacts are made to only two or three G-X-X' triplets (Emsley et al., 2000; Smethurst et al., 2007).

66

Figure 5.4 Model of YadA binding to a (POG)10 peptide. The model was created by docking the peptide onto the surface of YadA in the programme BIGGER, as described (Nummelin et al., 2004), with critical residues identified through site-directed mutagenesis. Seven of the 10 POG repeats make contacts with the surface of YadA. The figure was prepared in PyMOL (DeLano Scientific).

5.1.3 Crystallisation of YadA in complex with a collagen-mimicking peptide

I attempted to crystallise the head domain of YadA in complex with (PPG)10 and (POG)10. Unfortunately, no crystals of a complex were obtained. There are several explanations for this. YadA has three equivalent binding sites, all of which should be filled to ensure a monodisperse solution. However, the gel shift assays suggest that saturation of YadA binding sites requires a large molar excess of peptide. At less than saturating concentrations, (POG)10 aggregates YadA, a condition under which crystals will not form. Attempts to saturate the binding sites of YadA with (POG)10 and isolate the complex by size exclusion chromatography did not yield good results. One reason for this is that (POG)10 has limited solubility, and thus reaching the high concentrations required proved difficult. (PPG)10 does not exhibit this aggregation behaviour, but it binds ~300-fold less tightly to YadA than (POG)10, and so isolating a complex by size exclusion chromatography did not work. In addition, adding an excess of either peptide to the crystallisation solutions often resulted in the peptide itself crystallising. Furthermore, as

67

the peptides do not contain specific sequence information, they can contact the surface of YadA in a number of different registers. Thus attempts to crystallise a YadA-peptide complex were confounded by a number of factors.

5.2 The Eib proteins are receptors for IgG Fc (Study III)

The Eib proteins are trimeric autotransporters homologous to YadA (Sandt & Hill, 2000; Sandt & Hill, 2001). They were first identified by their ability to bind to the Fc region of IgG. However, recently Ghumra & Pleass (2007) challenged this view and claimed that the Fc-binding activity of Eibs was in fact artifactual. I produced the recombinant passenger domains of EibACDF and investigated their Ig-binding properties by immunoblot, ELISA and SPR. By each of these methods, I could show that Eibs bind to IgG Fc. The affinities of Eibs for IgG Fc were in the range of 50-200 nM. Furthermore, I was able to block the binding of Eibs to HRP-conjugated IgG by adding IgG Fc, but not when I added Fab. Eibs also failed to bind IgG Fab by SPR. These results show conclusively that Eibs are receptors for human IgG Fc.

Differential glycosylation of Fc could be a possible explanation for the failure of Ghumra & Pleass (2007) to detect the Fc-binding activity of Eibs. I therefore deglycosylated IgG and IgG Fc in order to investigate whether this would have an effect on binding to Eibs. In both ELISA and SPR experiments, removing the glycan moiety of Fc had a negligible effect on binding. Glycosylation of Fc thus does not play a role in Eib binding. The experimental set up of the ELISA experiment with deglycosylated antibodies shows the Eibs can bind more than one Ig, as would be expected for these homotrimeric molecules.

The experiments above show that Eibs bind to at least human and goat IgG. By ELISA, I showed that EibACDF bound to all isotypes of murine IgG, with a slight apparent preference for the IgG1 and IgG3 isotypes (Figure 5.5). In addition, Eibs bound to rabbit γ-globulin (a crude preparation of IgG), with EibF exhibiting the tightest binding (Figure 5.5).

I also investigated the binding of Eibs to IgA by the same methods. EibCDF all bound IgA, but EibA failed to do so, as reported before (Sandt & Hill, 2001). The affinities of EibCDF for IgA were between 70 and 500 nM, with EibF binding the tightest and EibC the weakest. Blocking the binding of IgA to EibCD with IgG Fc caused a significant (p < 0.01) reduction in binding levels, suggesting that Fc binding can to some extent interfere with IgA binding. For EibF, no significant reduction was seen.

68

Figure 5.5 Binding of Eibs to murine IgG isotypes (mIgG) and rabbit γ-globulin (rγ-globulin) by ELISA. Human IgG (hIgG) was a positive control. Columns represent the mean value from three replicate wells; error bars denote SEM.

5.3 Crystallisation of trimeric autotransporter fragments

5.3.1 Crystallisation of Eib passenger domains

In order to elucidate the structural basis of Ig-binding, I set up crystallisation trials for the passenger domains of EibACDF. From these, I was able to obtain crystals EibADF (Figure 5.6A-C). EibA crystallised in a variety conditions where polyethylene glycol was the major precipitant, producing hexagonal crystals. EibF crystallised in a single condition with a high concentration of ammonium formate. Unfortunately, these crystals only diffracted to ~10 Å resolution at best.

EibD crystallised in a variety of conditions. The best crystals were grown from jeffamine at near neutral pH (Figure 5.6C). These crystals diffracted well, to 2.7 Å. However, one dimension of the unit cell was almost 440 Å in length, thus giving rise to very closely spaced spots in the diffraction pattern. The data proved very difficult to process, and we could not assign a space group unambiguously. In addition, we suspected possible pseudosymmetry or twinning of the crystal.

To circumvent this problem, I produced a number of fragments of the EibD passenger domain and tested which ones still bound to both IgA and IgG. Based on these experiments, EibD161-418 was selected for large-scale production and purification. This

69

protein also crystallised readily, producing numerous, small (~100 µM) rod-like crystals in several different conditions (Figure 5.6D). The best diffraction was observed in crystals grown from ammonium phosphate. We collected a data set to 2.0 Å at beamline BM14 at the European Synchrotron Radiation Facility. We were able to solve the structure of EibD161-418 by molecular replacement using the YadA head domain structure (PDB 1P9H) as the search model. The crystallographic statistics are summarised in Table 5.2.

5.3.2 Crystallisation of the translocation unit of SadA

Together with our collaborators, we attempted to solve the structure of a translocation unit from a trimeric autotransporter. The membrane anchor of SadA was selected as the candidate for structure solution. We crystallised the translocation unit using standard sitting drop vapour diffusion methodology and were able to grow crystals in several conditions. In many cases, the crystals were clusters of needles, but some more three-dimensional crystals could be grown (Figure 5.6E & 5.6F). The crystals remained quite small, however, and even the largest of these did not diffract to sufficiently high resolution to allow the collection of a reasonable data set.

Figure 5.6 Crystals of trimeric autotransporter fragments. A) EibA passenger domain (grown from 8% PEG 10,000 at pH 7.5) B) EibF passenger domain (grown from 2.5 M ammonium phosphate) C) EibD passenger domain (grown from 30% jeffamine at pH 7.0) D) EibD161-418 (grown from 1 M ammonium phosphate) E) SadA translocation unit (solubilised in 1% octyl polyoxyethylene; grown from 12% PEG 4,000 at pH 4.6) F) SadA translocation unit (solubilised in 1% polyoxoethylene; grown from 14% PEG 3,350 at pH 5.6). The scale bar shows approximately 100 µm.

70

Table 5.2 Crystallographic statistics of EibD161-418 structure solution

Wavelength 1.03892 Å

Space group (number) H 3 (146)

Unit cell a = b = 48.972 Å c = 409.512

Data Overall Inner Shell Outer Shell

Resolution range 45.50 - 1.99 Å 45.50 – 6.29 Å 2.10 -1.99 Å

Rmerge 0.049 0.020 0.279

Rmeas (all I+ & I-) 0.069 0.028 0.392

Rpim(all I+ & I-) 0.048 0.020 0.275

Nr. of observations 38396 1206 5537

Nr. unique 24845 698 3623

Mean ((I)/σ(I)) 7.6 17.1 2.4

Completeness 98.7 86.8 98.8

Multiplicity 1.5 1.7 1.5

Refinement

R value 0.1784

Rfree 0.2153

Rfree set size 5.12 %

Rfree set count 1271

Final structure

Deviation from ideal

RMSD 0.006 Å

Angle 0.901

Chirality 0.062

Planarity 0.003

Dihedral 16.287

Min. nonbonded dist. 2.274 Å

Ramachandran stat.

Favoured 97.7%

Outliers 1.2%

Bad rotamers 5.3%

71

5.4 The crystal structure of the EibD passenger domain fragment (Study IV)

5.4.1 Overview of the structure

EibD161-418 is an elongated lollipop-like structure with a globular head domain, a neck and an extended stalk (Figure 5.7A). The N-terminal, globular domain is a YadA-like β-roll with eight NSVAIGXX motifs. The final loop of the β-roll domain before the neck is significantly larger than the others.

The first half of the stalk is a right-handed coiled-coil with hendecad periodicity, i.e. 11 residues per three turns (Lupas & Gruber, 2005). This is followed by a structure we refer to as a collar, a small minidomain formed by a twisted three-stranded antiparallel β-sheet protruding from the stalk. This region is stabilised by a network of ion bridges and hydrogen bonds (Figure 5.7 B). The collar shifts the last part of the stalk by 120 °, so that the lower half of the α-helix of one monomer essentially continues from the top half of the α-helix of an adjacent monomer (Figure 5.7A). The structure of the collar is stabilised by ionic and hydrogen-bonding networks at both the points of exit and re-entry into the coiled-coil (Figure 5.7B). The C-terminal half of the stalk is a canonical left-handed coiled-coil with heptad periodicity (3.5 residues per turn).

There are four anions positioned in the core of the left-handed coiled coil. In three cases, asparagine residues are in position d of the coiled-coil heptad (the seven positions of the heptad are designated a-g). Normally, hydrophobic residues are expected in positions a and d of coiled-coils, as these point into the hydrophobic core of the supercoil (Lupas & Gruber, 2005). The anion seems to be required to stabilise the structure when the amino acid in the core is polar (Figure 5.7C). Similar cases are seen in the UspA1 stalk, where chlorides and phosphates are found in the core of the coiled-coil in conjunction with polar residues (Conners et al., 2008). In addition, the fourth anion in the EibD stalk is coordinated by threonine residues in an RTD motif, where the threonine is in position a (Figure 5.7D). In figure 5.7, all anions are modelled as chlorides. In addition to the positions occupied by anions, both the right and left-handed regions of the stalk contain cavities in the hydrophobic core of the coiled-coil, devoid of electron density (Figure 5.7 E).

5.4.2 Mapping of the Ig-binding regions

To pinpoint the binding sites for IgA and IgG in the EibD passenger domain, I introduced stop codons at four positions along the EibD stalk (after residues 308, 329, 344 and 384). Truncating at any of these positions abolished binding to IgG, whereas only the first two truncations affected IgA binding, suggesting that the IgA binding site was in the right-handed region, whereas the IgG binding site was located in the left-handed region at the bottom of the stalk.

72

Figure 5.7 Crystal structure of EibD161-418. A) Overview of the structure. The β-roll domain is in green, the neck is red, the right-handed hendecad coiled-coil is coloured magenta, the collar is cyan and the left-handed heptad coiled-coil is orange. B) The EibD collar showing the stabilising ionic network. The monomers are coloured blue, red and yellow. C) Coordination of threonines by a chloride in an RTD motif (residues 382-384). The monomers are coloured blue, red and yellow. D) Coordination of a chloride ion by Asp residues in the d position of the coiled-coil. The figure shows Asp407. The chains of the monomers are coloured blue, red and yellow. E) The right-handed region of the stalk showing the cavities (circled) in the hydrophobic core of the right-handed coiled-coil. Two monomers are in space-filling representation; the third monomer is shown as a backbone ribbon. F) Residues important in Ig-binding. The figure shows only the stalk. Critical residues are coloured red, whereas residues with some effect on binding are orange. The figure was produced in PyMOL (DeLano Scientific).

To pinpoint residues critical for Ig binding, I produced a number of double or triple point mutations in the putative Ig-binding regions where surface-exposed residues where changed into alanines. By ELISA, IgA binding was abolished by the triple mutant K332A-K333A-Y334A. This mutant also had a large effect in SPR, and no apparent dissociation constant could be reliably measured for this mutant. It also failed to bind IgA in an immunoblot experiment. In addition, the mutations E327A-E329A-K330A and D336A-K338A had an effect on IgA binding by both ELISA and SPR. This suggests that at least one of the residues K332, K333 and Y334 is critical for binding IgA, and other residues in the vicinity (327-330 and 336-338) increase the affinity (Figure 5.7F). Interestingly, the mutation T415A-R416A-T417A at the bottom of the left-handed stalk also had an effect on IgA binding, although this mutation is distant from the putative binding site. The CD spectrum of this mutant is somewhat altered compared to the wild type, so some subtle structural change in this mutant may be propagated along the stalk and so affect IgA binding.

73

The IgG binding site is in the lower half of the stalk. Binding studies on the mutants suggested that D387A-Y388A had the strongest effect on IgG binding in both ELISA and immunoblot. Furthermore, a reliable estimate for an apparent dissociation constant could not be calculated for it by SPR. However, the CD spectrum of this mutant showed some differences compared to the wild type, which could be interpreted as a change in structure. Interestingly, none of the other mutants in the vicinity had a large effect on IgG binding. In contrast, S408A-K409A and T415A-R416A-T417A caused a slight reduction in the level of IgG (and IgA) binding. This may be due to minor structural effects as mentioned above.

To examine whether the collar had an effect on Ig binding, I deleted this region from the protein. Two deletions were made, Δ350-372 and Δ352-372. The first deletion keeps the two segments of coiled-coil in register, whereas the second should disrupt the periodicity of the coiled-coil. Surprisingly, neither deletion appeared to affect Ig binding. In fact, deleting the collar seemed to improve IgA binding, possibly by reducing the steric hindrance of the interaction. However, by SPR the deletions reduced binding to both IgA and IgG, although binding was still detectable. Thus, the collar does not have a critical role in Ig binding.

74

6. Conclusions and future perspectives

Trimeric autotransporters are important virulence factors. Because they usually have multiple virulence-related functions, their role in the biology and pathogenicity of the host bacterium is complex. However, these proteins offer fascinating insights into how bacteria adhere to substrates, protect themselves from environmental insults and secrete proteins. Consequently, the structure-function relationship of trimeric autotransporters has elicited a great deal of interest in recent years.

YadA is the prototype for trimeric autotransporters. Although the adhesion of YeYadA to its major ligand, collagen, has been studied for almost two decades, the details of the interaction are still vague. This prompted me to examine the binding of YadA to collagen model peptides and the Toolkits.

YadA binds promiscuously to fibrillar collagens, and contrary to previous reports (Schulze-Koops et al., 1992; Schulze-Koops et al., 1995), there is no single binding site. A collagenous triple-helix is required, and YadA binds tightly to collagen regions with a high hydroxyproline-to-charged residue ratio without particular sequence specificity.

This strategy could be advantageous for a pathogenic organism such as Y. enterocolitica. This organism is unlikely to encounter extensive collagenous surfaces, so binding without strict sequence specificity may aid the bacterium in establishing a foothold in host tissues. Other pathogenic organisms may share a similar strategy of binding unspecifically to host structures. YadA would thus be a paradigm for bacterial adhesins that mediate promiscuous adherence.

Intriguingly, YadA binds to Toolkit peptides that are also targets for human collagen-binding proteins. This suggests that YadA may interfere with some regulatory processes in the host. A particularly interesting example is leukocyte-associated Ig-like receptor (LAIR)-1, a collagen-binding inhibitor of immune responses (Meyaard, 2008). LAIR-1 modulates inflammation, and Y. enterocolitica infection can be followed by a number of inflammatory sequelae such as reactive arthritis. YadA could promote inflammation by competing with LAIR-1 for binding to collagen. Future research should thus be directed to uncovering whether there is a link between LAIR-1 and the inflammatory sequelae of yersiniosis.

The major unanswered question about the YadA-collagen interaction is how it can bind to diverse collagen sequences. This requires structural information about the interaction. Although a crude model of the binding of YadA to a collagen model peptide exists (Nummelin et al., 2004), it cannot be considered very accurate. Unfortunately, I was unable to crystallise the YadA headgroup in complex with a collagen model peptide despite considerable efforts. Other strategies should therefore be used to obtain a structure. Towards this end, I have initiated experiments to produce a YadA-PPG peptide fusion. Using a fusion protein would make controlling the stoichiometry of the interaction easier, and also prevent crystallisation of the peptide alone. In addition, genetically modifying the sequence of the peptide ligand in the fusion protein would be significantly easier than chemically synthesising a new peptide for each experiment. One or more complex structures obtained in this way would significantly help to illuminate the mechanism of collagen binding.

75

The Eib proteins are homologous to YadA. I have shown that these proteins are bona fide IgG Fc receptors. However, the role of Ig-binding by Eibs has not been determined. One possibility is that Ig-binding is involved in serum resistance. The stalks of several trimeric autotransporters, including YadA, protect the bacteria from complement-mediated lysis (Linke et al., 2006). Analogously, the location of the Ig-binding sites in the stalk of EibD suggests that Ig binding does have a protective function for the bacteria. To confirm this, I have planned experiments to study the role of Ig binding in serum resistance. Comparing the resistance of EibD-expressing bacteria to normal serum and serum depleted of antibodies should give an indication of whether the Ig-binding activity plays a major role.

There are several other unanswered questions about Eib function. The role of the collar is not clear. Furthermore, the function of the YadA-like β-roll domain is currently unknown. For EibD at least, the β-roll domain does not bind to collagen. It may be involved in adhesion to epithelial cells, but not all Eibs have a β-roll domain. The role of the N-terminal Eib domain, which is missing in the EibD161-418 structure and for which no structural predictions are available, is also undetermined. As all Eibs mediate adherence to epithelial cells, so adhesion may be an activity of the Eib domain, which is present in all six members of the family.

The in vivo functions of Eibs are still to be elucidated, as is their potential role as virulence factors. Eibs are encoded by prophage sequences, which strongly suggests that the presence of the Ig-binding proteins in ECOR strains is a result of horizontal gene transfer from other, possibly pathogenic strains or species. Additionally, many different eib genes are present in a single strain, which raises the question why multiple proteins with essentially identical functions would be needed and maintained in the same strain.

Another intriguing question is whether the Eib proteins can cross-trimerise. YadA and Hia do not cross-trimerise (Cotter et al., 2006), but YadA polypeptides of different lengths can be exported and form trimers (Roggenkamp et al., 2003). The translocation unit must therefore be responsible for recognition of compatible monomers and trimerisation. However, the translocation units of the Eibs have virtually identical sequences, so cross-trimerisation should occur in strains harbouring more than one eib gene if recognition is a function of the translocation unit. Whether cross-trimerisation takes place or not has implications for the mechanism of autotransportation: cross trimers would support the single-chain model whereas exclusive formation of homotrimers would mean that the passenger domains are responsible for recognising their partners. The latter model suggests that trimerisation of the passenger domain happens before translocation, which would be inconsistent with the hairpin model.

The mechanism of autotransportation remains the central undetermined aspect of the biochemistry of these proteins. I have been able to characterise the binding of Eibs and YadA to their ligands, but how the ligand-binding domains are transported outside the cell remains unclear. There are no clear-cut approaches to investigating this; possibly pulse-chase experiments or designing mutations to trap folding intermediates may provide a handle on revealing the mechanism of autotransportation. YadA and the Eibs would be the proteins of choice for investigating this phenomenon: they relatively small compared to most other trimeric autotransporters and are easily detected. Furthermore, the structure of these proteins is now well understood, and unlike some larger trimeric autotransporters

76

such as Hia and BadA, they do not contain multiple head domains. Because of their small size, well-known biochemistry and relatively simple structure, YadA and the Eibs would be the natural model systems with which to study autotransportation. Understanding the details of the mechanism of passenger domain translocation through the OM would of great biochemical interest. In addition, it would also enable rational design of antimicrobials targeting diverse pathogens that express trimeric autotransporter virulence factors.

77

Acknowledgements

This work was carried out during the years 2004-2009 in the Structural Biology and Biophysics research program at the Institute of Biotechnology, and the Division of Genetics in the Department of Biological and Environmental Sciences, both in the University of Helsinki. I would like to thank the heads of both the Institute, Professor Mart Saarma (and Professor Tomi Mäkelä, head of the Institute from summer 2009), and the Department, Professor Kielo Haahtela, for providing the excellent facilities in which to conduct this research. I also thank Professor Mark Johnson, the head of the National Graduate School in Information and Structural Biology, for accepting me into the school and arranging such invigorating meetings.

Among the many people deserving my gratitude, I would first and foremost like to thank my supervisor (and sometime director), Professor Adrian Goldman. Adrian’s enthusiasm, keen intelligence and breadth of knowledge have always inspired me. In addition to science, we both share an interest in theatre – in fact, it was through this interest that we first met and I ended up working in his lab. It has been an honour and a privilege to have been both supervised and directed by him.

I am deeply grateful to both members of my thesis follow-up committee, Professors Sarah Butcher and Mikael Skurnik. Sarah’s no-nonsense approach helped focus my work, and Miikki’s support and advice were always beneficial and welcome. I would also like to thank Sarah and Dr Susan Buchanan for reviewing my thesis.

Of the many great colleagues that I have had over the years, I would particularly like thank my predecessor in the YadA project, Heli Elovaara. Heli was a delight to work with and she taught me not to take science too seriously. I’m also grateful to her for continuing to to help out with the YadA project even after she left Adrian’s lab. Warm thanks also go to “my” post doc, Andrzej Lyskowski. His positive attitude, easygoing manner and friendly nature made working with him a pleasure. I’d also like to thank Andrzej for always being ready to help with any computer problems I had.

I would also like to thank all other members of the Goldman group, past and present, for creating such a friendly working environment in the lab. This has been quite a lot of people over the years: Ansko, Arnab, Igor, Jaana, Jussi, Katja, Lari, Maria, Michael, Mika, Praseet, Sanjay, Saurabh, Sirpa, Tommi, Veli-Matti, V-P and Vimal. And of course the PI of our “competitor group”, Pirkko, and the members of her group: Chiara, Elina and Marika. Special thanks go to our regular lunch crowd: Esko, Juho, Seija, Wesley and Danielle. And thanks to Heidi for putting up with me and my piles of paper in our shared office.

My Ph.D. work was essentially funded by the Division of Genetics at the Department of Biological and Environmental sciences, where I held a teaching post during this time. I would like to thank my boss there, Professor Tapio Palva, for choosing me for the post and providing a cordial working environment. And of course thanks are due to all my friends and colleagues at Genetics, particularly my regular co-teachers: Hannu, Pekka, Liisa, Elina, Krisse, Saara, Reetta, Päivi and Sami. I’m also very grateful to the invaluable technical staff, especially the two Arjas, Ikävalko and Välimäki, for always being ready to help out with any administrative problems I might have had.

78

I would like to thank my collaborators, Barbara Brodsky and particularly Richard Farndale, whose laboratory I visited to work on the Collagen Toolkits. Thanks go to the other members of the Farndale lab in Cambridge, especially Samir and Gavin, for making my stay there so enjoyable, and of course Dominique for synthesising collagen-like peptides for my experiments. I would also like to take this opportunity to thank Dirk Linke and Andrei Lupas for some very stimulating discussions and good advice. Unfortunately, we didn’t get a structure of a translocation domain, but I learned a lot trying.

Last, but certainly not least, I would like to thank all my dear friends and family who have supported me in my efforts, showed an interest in my work, or helped take my mind off science for a while. I am especially grateful to my parents, Riitta and John, for their constant encouragement and unwavering support.

Helsinki, October 2009

79

References

Abdallah A. M., Gey van Pittius N. C., Champion P. A., Cox J., Luirink J., Vandenbroucke-Grauls C. M., Appelmelk B. J., and Bitter W. (2007). Type VII secretion--mycobacteria show the way. Nat Rev Microbiol. 5(11):883-891.

Abdullah M., Nepluev I., Afonina G., Ram S., Rice P., Cade W., and Elkins C. (2005). Killing of dsrA mutants of Haemophilus ducreyi by normal human serum occurs via the classical complement pathway and is initiated by immunoglobulin M binding. Infect Immun. 73(6):3431-3439.

Achtman M., Zurth K., Morelli G., Torrea G., Guiyoule A., and Carniel E. (1999). Yersinia pestis, the cause of plague, is a recently emerged clone of Yersinia pseudotuberculosis. Proc Natl Acad Sci U S A. 96(24):14043-14048.

Ackermann N., Tiller M., Anding G., Roggenkamp A., and Heesemann J. (2008). Contribution of trimeric autotransporter C-terminal domains of oligomeric coiled-coil adhesin (Oca) family members YadA, UspA1, EibA, and Hia to translocation of the YadA passenger domain and virulence of Yersinia enterocolitica. J Bacteriol. 190(14):5031-5043.

Adams P. D., Grosse-Kunstleve R. W., Hung L. W., Ioerger T. R., McCoy A. J., Moriarty N. W., Read R. J., Sacchettini J. C., Sauter N. K., and Terwilliger T. C. (2002). PHENIX: building new software for automated crystallographic structure determination. Acta Crystallogr D Biol Crystallogr. 58(Pt 11):1948-1954.

Aebi C., Lafontaine E. R., Cope L. D., Latimer J. L., Lumbley S. L., McCracken G. H. J., and Hansen E. J. (1998). Phenotypic effect of isogenic uspA1 and uspA2 mutations on Moraxella catarrhalis 035E. Infect Immun. 66(7):3113-3119.

Aebi C., Maciver I., Latimer J. L., Cope L. D., Stevens M. K., Thomas S. E., McCracken G. H. J., and Hansen E. J. (1997). A protective epitope of Moraxella catarrhalis is encoded by two different genes. Infect Immun. 65(11):4367-4377.

Al-Hasani K., Henderson I. R., Sakellaris H., Rajakumar K., Grant T., Nataro J. P., Robins-Browne R., and Adler B. (2000). The sigA gene which is borne on the she pathogenicity island of Shigella flexneri 2a encodes an exported cytopathic protease involved in intestinal fluid accumulation. Infect Immun. 68(5):2457-2463.

Alamuri P., and Mobley H. L. (2008). A novel autotransporter of uropathogenic Proteus mirabilis is both a cytotoxin and an agglutinin. Mol Microbiol. 68(4):997-1017.

Anderson D. M., and Schneewind O. (1997). A mRNA signal for the type III secretion of Yop proteins by Yersinia enterocolitica. Science. 278(5340):1140-1143.

Aoki S. K., Pamma R., Hernday A. D., Bickham J. E., Braaten B. A., and Low D. A. (2005). Contact-dependent inhibition of growth in Escherichia coli. Science. 309(5738):1245-1248.

Backert S., and Meyer T. F. (2006). Type IV secretion systems and their effectors in bacterial pathogenesis. Curr Opin Microbiol. 9(2):207-217.

Balligand G., Laroche Y., and Cornelis G. (1985). Genetic analysis of virulence plasmid from a serogroup 9 Yersinia enterocolitica strain: role of outer membrane protein P1 in resistance to human serum and autoagglutination. Infect Immun. 48(3):782-786.

Barenkamp S. J., and St. Geme J. W. r. (1996). Identification of a second family of high-molecular-weight adhesion proteins expressed by non-typeable Haemophilus influenzae. Mol Microbiol. 19(6):1215-1223.

80

Barnard T. J., Dautin N., Lukacik P., Bernstein H. D., and Buchanan S. K. (2007). Autotransporter structure reveals intra-barrel cleavage followed by conformational changes. Nat Struct Mol Biol. 14(12):1214-1220.

Benjelloun-Touimi Z., Sansonetti P. J., and Parsot C. (1995). SepA, the major extracellular protein of Shigella flexneri: autonomous secretion and involvement in tissue invasion. Mol Microbiol. 17(1):123-135.

Benz I., and Schmidt M. A. (1989). Cloning and expression of an adhesin (AIDA-I) involved in diffuse adherence of enteropathogenic Escherichia coli. Infect Immun. 57(5):1506-1511.

Benz I., and Schmidt M. A. (1992). AIDA-I, the adhesin involved in diffuse adherence of the diarrhoeagenic Escherichia coli strain 2787 (O126:H27), is synthesized via a precursor molecule. Mol Microbiol. 6(11):1539-1546.

Berisio R., Vitagliano L., Mazzarella L., and Zagari A. (2000). Crystal structure of a collagen-like polypeptide with repeating sequence Pro-Hyp-Gly at 1.4 Å resolution: implications for collagen hydration. Biopolymers. 56(1):8-13.

Bernardini M. L., Mounier J., d'Hauteville H., Coquis-Rondon M., and Sansonetti P. J. (1989). Identification of icsA, a plasmid locus of Shigella flexneri that governs bacterial intra- and intercellular spread through interaction with F-actin. Proc Natl Acad Sci U S A. 86(10):3867-3871.

Biedzka-Sarek M., Jarva H., Hyytiainen H., Meri S., and Skurnik M. (2008a). Characterization of complement factor H binding to Yersinia enterocolitica serotype O:3. Infect Immun. 76(9):4100-4109.

Biedzka-Sarek M., Salmenlinna S., Gruber M., Lupas A. N., Meri S., and Skurnik M. (2008b). Functional mapping of YadA- and Ail-mediated binding of human factor H to Yersinia enterocolitica serotype O:3. Infect Immun. 76(11):5016-5027.

Biedzka-Sarek M., Venho R., and Skurnik M. (2005). Role of YadA, Ail, and Lipopolysaccharide in Serum Resistance of Yersinia enterocolitica Serotype O:3. Infect Immun. 73(4):2232-2244.

Bingle L. E., Bailey C. M., and Pallen M. J. (2008). Type VI secretion: a beginner's guide. Curr Opin Microbiol. 11(1):3-8.

Black S. D., and Mould D. R. (1991). Development of hydrophobicity parameters to analyze proteins which bear post- or cotranslational modifications. Anal Biochem. 193(1):72-82.

Bliska J. B., Copass M. C., and Falkow S. (1993). The Yersinia pseudotuberculosis adhesin YadA mediates intimate bacterial attachment to and entry into HEp-2 cells. Infect Immun. 61(9):3914-3921.

Bölin I., Norlander L., and Wolf-Watz H. (1982). Temperature-inducible outer membrane protein of Yersinia pseudotuberculosis and Yersinia enterocolitica is associated with the virulence plasmid. Infect Immun. 37(2):506-512.

Bos M. P., Robert V., and Tommassen J. (2007). Biogenesis of the gram-negative bacterial outer membrane. Annu Rev Microbiol. 61:191-214.

Bottone E. J. (1997). Yersinia enterocolitica: the charisma continues. Clin Microbiol Rev. 10(2):257-276. Brandon L. D., and Goldberg M. B. (2001). Periplasmic transit and disulfide bond formation of the

autotransported Shigella protein IcsA. J Bacteriol. 183(3):951-958. Brunder W., Schmidt H., and Karch H. (1997). EspP, a novel extracellular serine protease of

enterohaemorrhagic Escherichia coli O157:H7 cleaves human coagulation factor V. Mol Microbiol. 24(4):767-778.

Buchanan S. K. (2001). Type I secretion and multidrug efflux: transport through the TolC channel-tunnel. Trends Biochem Sci. 26(1):3-6.

81

Bukholm G., Kapperud G., and Skurnik M. (1990). Genetic evidence that the yopA gene-encoded Yersinia outer membrane protein Yop1 mediates inhibition of the anti-invasive effect of interferon. Infect Immun. 58(7):2245-2251.

Bullard B., Lipski S., and Lafontaine E. R. (2007). Regions important for the adhesin activity of Moraxella catarrhalis Hag. BMC Microbiol. 7:65.

Burns D. L. (2003). Type IV transporters of pathogenic bacteria. Curr Opin Microbiol. 6(1):29-34. Capecchi B., Adu-Bobie J., Di Marcello F., Ciucchi L., Masignani V., Taddei A., Rappuoli R., Pizza M., and

Arico B. (2005). Neisseria meningitidis NadA is a new invasin which promotes bacterial adhesion to and penetration into human epithelial cells. Mol Microbiol. 55(3):687-698.

Carinato M. E., Collin-Osdoby P., Yang X., Knox T. M., Conlin C. A., and Miller C. G. (1998). The apeE gene of Salmonella typhimurium encodes an outer membrane esterase not present in Escherichia coli. J Bacteriol. 180(14):3517-3521.

Cascales E. (2008). The type VI secretion toolkit. EMBO Rep. 9(8):735-741. Chain P. S., Carniel E., Larimer F. W., Lamerdin J., Stoutland P. O., Regala W. M., Georgescu A. M.,

Vergez L. M., Land M. L., Motin V. L., Brubaker R. R., Fowler J., Hinnebusch J., Marceau M., Medigue C., Simonet M., Chenal-Francisque V., Souza B., Dacheux D., Elliott J. M., Derbise A., Hauser L. J., and Garcia E. (2004). Insights into the evolution of Yersinia pestis through whole-genome comparison with Yersinia pseudotuberculosis. Proc Natl Acad Sci U S A. 101(38):13826-13831.

Chan Y. G., Cardwell M. M., Hermanas T. M., Uchiyama T., and Martinez J. J. (2009). Rickettsial outer-membrane protein B (rOmpB) mediates bacterial invasion through Ku70 in an actin, c-Cbl, clathrin and caveolin 2-dependent manner. Cell Microbiol.

Charbonneau M. E., Berthiaume F., and Mourez M. (2006). Proteolytic processing is not essential for multiple functions of the Escherichia coli autotransporter adhesin involved in diffuse adherence (AIDA-I). J Bacteriol. 188(24):8504-8512.

China B., N'Guyen B. T., de Bruyere M., and Cornelis G. R. (1994). Role of YadA in resistance of Yersinia enterocolitica to phagocytosis by human polymorphonuclear leukocytes. Infect Immun. 62(4):1275-1281.

China B., Sory M. P., N'Guyen B. T., De Bruyere M., and Cornelis G. R. (1993). Role of the YadA protein in prevention of opsonization of Yersinia enterocolitica by C3b molecules. Infect Immun. 61(8):3129-3136.

Christie P. J., Atmakuri K., Krishnamoorthy V., Jakubowski S., and Cascales E. (2005). Biogenesis, architecture, and function of bacterial type IV secretion systems. Annu Rev Microbiol. 59:451-485.

Clantin B., Delattre A. S., Rucktooa P., Saint N., Meli A. C., Locht C., Jacob-Dubuisson F., and Villeret V. (2007). Structure of the membrane protein FhaC: a member of the Omp85-TpsB transporter superfamily. Science. 317(5840):957-961.

Clantin B., Hodak H., Willery E., Locht C., Jacob-Dubuisson F., and Villeret V. (2004). The crystal structure of filamentous hemagglutinin secretion domain and its implications for the two-partner secretion pathway. Proc Natl Acad Sci U S A. 101(16):6194-6199.

Claude J. B., Suhre K., Notredame C., Claverie J. M., and Abergel C. (2004). CaspR: a web server for automated molecular replacement using homology modelling. Nucleic Acids Res. 32(Web Server issue):W606-9.

Cole L. E., Kawula T. H., Toffer K. L., and Elkins C. (2002). The Haemophilus ducreyi serum resistance antigen DsrA confers attachment to human keratinocytes. Infect Immun. 70(11):6158-6165.

82

Comanducci M., Bambini S., Brunelli B., Adu-Bobie J., Arico B., Capecchi B., Giuliani M. M., Masignani V., Santini L., Savino S., Granoff D. M., Caugant D. A., Pizza M., Rappuoli R., and Mora M. (2002). NadA, a novel vaccine candidate of Neisseria meningitidis. J Exp Med. 195(11):1445-1454.

Conners R., Hill D. J., Borodina E., Agnew C., Daniell S. J., Burton N. M., Sessions R. B., Clarke A. R., Catto L. E., Lammie D., Wess T., Brady R. L., and Virji M. (2008). The Moraxella adhesin UspA1 binds to its human CEACAM1 receptor by a deformable trimeric coiled-coil. EMBO J. 27(12):1779-1789.

Coombes B. K. (2009). Type III secretion systems in symbiotic adaptation of pathogenic and non-pathogenic bacteria. Trends Microbiol. 17(3):89-94.

Cope L. D., Thomas S. E., Latimer J. L., Slaughter C. A., Muller-Eberhard U., and Hansen E. J. (1994). The 100 kDa haem:haemopexin-binding protein of Haemophilus influenzae: structure and localization. Mol Microbiol. 13(5):863-873.

Cornelis G. R. (2002). Yersinia type III secretion: send in the effectors. J Cell Biol. 158(3):401-408. Cornelis G. R. (2006). The type III secretion injectisome. Nat Rev Microbiol. 4(11):811-825. Cotter S. E., Surana N. K., Grass S., and St. Geme J. W. r. (2006). Trimeric autotransporters require

trimerization of the passenger domain for stability and adhesive activity. J Bacteriol. 188(15):5400-5407.

Cotter S. E., Surana N. K., and St. Geme J. W. r. (2005a). Trimeric autotransporters: a distinct subfamily of autotransporter proteins. Trends Microbiol. 13(5):199-205.

Cotter S. E., Yeo H. J., Juehne T., and St. Geme J. W. r. (2005b). Architecture and adhesive activity of the Haemophilus influenzae Hsf adhesin. J Bacteriol. 187(13):4656-4664.

Coutte L., Antoine R., Drobecq H., Locht C., and Jacob-Dubuisson F. (2001). Subtilisin-like autotransporter serves as maturation protease in a bacterial secretion pathway. EMBO J. 20(18):5040-5048.

Cover T. L. (1996). The vacuolating cytotoxin of Helicobacter pylori. Mol Microbiol. 20(2):241-246. Cover T. L., and Aber R. C. (1989). Yersinia enterocolitica. N Engl J Med. 321(1):16-24. da Silva A. C., Ferro J. A., Reinach F. C., Farah C. S., Furlan L. R., Quaggio R. B., Monteiro-Vitorello C.

B., Van Sluys M. A., Almeida N. F., Alves L. M., do Amaral A. M., Bertolini M. C., Camargo L. E., Camarotte G., Cannavan F., Cardozo J., Chambergo F., Ciapina L. P., Cicarelli R. M., Coutinho L. L., Cursino-Santos J. R., El-Dorry H., Faria J. B., Ferreira A. J., Ferreira R. C., Ferro M. I., Formighieri E. F., Franco M. C., Greggio C. C., Gruber A., Katsuyama A. M., Kishi L. T., Leite R. P., Lemos E. G., Lemos M. V., Locali E. C., Machado M. A., Madeira A. M., Martinez-Rossi N. M., Martins E. C., Meidanis J., Menck C. F., Miyaki C. Y., Moon D. H., Moreira L. M., Novo M. T., Okura V. K., Oliveira M. C., Oliveira V. R., Pereira H. A., Rossi A., Sena J. A., Silva C., de Souza R. F., Spinola L. A., Takita M. A., Tamura R. E., Teixeira E. C., Tezza R. I., Trindade dos Santos M., Truffi D., Tsai S. M., White F. F., Setubal J. C., and Kitajima J. P. (2002). Comparison of the genomes of two Xanthomonas pathogens with differing host specificities. Nature. 417(6887):459-463.

Das A., Rangaraj N., and Sonti R. V. (2009). Multiple adhesin-like functions of Xanthomonas oryzae pv. oryzae are involved in promoting leaf attachment, entry, and virulence on rice. Mol Plant Microbe Interact. 22(1):73-85.

Dautin N., Barnard T. J., Anderson D. E., and Bernstein H. D. (2007). Cleavage of a bacterial autotransporter by an evolutionarily convergent autocatalytic mechanism. EMBO J. 26(7):1942-1952.

Dautin N., and Bernstein H. D. (2007). Protein secretion in Gram-negative bacteria via the autotransporter pathway. Annu Rev Microbiol. 61:89-112.

83

Davis I. W., Murray L. W., Richardson J. S., and Richardson D. C. (2004). MOLPROBITY: structure validation and all-atom contact analysis for nucleic acids and their complexes. Nucleic Acids Res. 32(Web Server issue):W615-9.

Durant L., Metais A., Soulama-Mouze C., Genevard J. M., Nassif X., and Escaich S. (2007). Identification of candidates for a subunit vaccine against extraintestinal pathogenic Escherichia coli. Infect Immun. 75(4):1916-1925.

Eitel J., and Dersch P. (2002). The YadA protein of Yersinia pseudotuberculosis mediates high-efficiency uptake into human cells under environmental conditions in which invasin is repressed. Infect Immun. 70(9):4880-4891.

Eitel J., Heise T., Thiesen U., and Dersch P. (2005). Cell invasion and IL-8 production pathways initiated by YadA of Yersinia pseudotuberculosis require common signalling molecules (FAK, c-Src, Ras) and distinct cell factors. Cell Microbiol. 7(1):63-77.

El Tahir Y., Kuusela P., and Skurnik M. (2000). Functional mapping of the Yersinia enterocolitica adhesin YadA. Identification Of eight NSVAIG - S motifs in the amino-terminal half of the protein involved in collagen binding. Mol Microbiol. 37(1):192-206.

Elkins C., Morrow K. J. J., and Olsen B. (2000). Serum resistance in Haemophilus ducreyi requires outer membrane protein DsrA. Infect Immun. 68(3):1608-1619.

Emödy L., Heesemann J., Wolf-Watz H., Skurnik M., Kapperud G., O'Toole P., and Wadstrom T. (1989). Binding to collagen by Yersinia enterocolitica and Yersinia pseudotuberculosis: evidence for yopA-mediated and chromosomally encoded mechanisms. J Bacteriol. 171(12):6674-6679.

Emsley J., Knight C. G., Farndale R. W., Barnes M. J., and Liddington R. C. (2000). Structural basis of collagen recognition by integrin alpha2beta1. Cell. 101(1):47-56.

Emsley P., Charles I. G., Fairweather N. F., and Isaacs N. W. (1996). Structure of Bordetella pertussis virulence factor P.69 pertactin. Nature. 381(6577):90-92.

Emsley P., and Cowtan K. (2004). Coot: model-building tools for molecular graphics. Acta Crystallogr D Biol Crystallogr. 60(Pt 12 Pt 1):2126-2132.

Escobar M. A., and Dandekar A. M. (2003). Agrobacterium tumefaciens as an agent of disease. Trends Plant Sci. 8(8):380-386.

Eslava C., Navarro-Garcia F., Czeczulin J. R., Henderson I. R., Cravioto A., and Nataro J. P. (1998). Pet, an autotransporter enterotoxin from enteroaggregative Escherichia coli. Infect Immun. 66(7):3155-3163.

Evans P. (2006). Scaling and assessment of data quality. Acta Crystallogr D Biol Crystallogr. 62(Pt 1):72-82.

Farndale R. W., Lisman T., Bihan D., Hamaia S., Smerling C. S., Pugh N., Konitsiotis A., Leitinger B., de Groot P. G., Jarvis G. E., and Raynal N. (2008). Cell-collagen interactions: the use of peptide Toolkits to investigate collagen-receptor interactions. Biochem Soc Trans. 36(Pt 2):241-250.

Feil H., Feil W. S., and Lindow S. E. (2007). Contribution of fimbrial and afimbrial adhesins of Xylella fastidiosa to attachment to aurfaces and virulence to grape. Phytopathology. 97(3):318-324.

Fekkes P., and Driessen A. J. (1999). Protein targeting to the bacterial cytoplasmic membrane. Microbiol Mol Biol Rev. 63(1):161-173.

Fernandez R. C., and Weiss A. A. (1994). Cloning and sequencing of a Bordetella pertussis serum resistance locus. Infect Immun. 62(11):4727-4738.

Fink D. L., Green B. A., and St. Geme J. W. r. (2002). The Haemophilus influenzae Hap autotransporter binds to fibronectin, laminin, and collagen IV. Infect Immun. 70(9):4902-4907.

Flügel A., Schulze-Koops H., Heesemann J., Kühn K., Sorokin L., Burkhardt H., von der Mark K., and Emmrich F. (1994). Interaction of enteropathogenic Yersinia enterocolitica with complex basement

84

membranes and the extracellular matrix proteins collagen type IV, laminin-1 and -2, and nidogen/entactin. J Biol Chem. 269(47):29732-29738.

Forman S., Wulff C. R., Myers-Morales T., Cowan C., Perry R. D., and Straley S. C. (2008). yadBC of Yersinia pestis, a new virulence determinant for bubonic plague. Infect Immun. 76(2):578-587.

Forsgren A., Brant M., Karamehmedovic M., and Riesbeck K. (2003). The immunoglobulin D-binding protein MID from Moraxella catarrhalis is also an adhesin. Infect Immun. 71(6):3302-3309.

Forsgren A., Brant M., Mollenkvist A., Muyombwe A., Janson H., Woin N., and Riesbeck K. (2001). Isolation and characterization of a novel IgD-binding protein from Moraxella catarrhalis. J Immunol. 167(4):2112-2120.

Forsgren A., Brant M., and Riesbeck K. (2004). Immunization with the truncated adhesin Moraxella catarrhalis immunoglobulin D-binding protein (MID764-913) is protective against M. catarrhalis in a mouse model of pulmonary clearance. J Infect Dis. 190(2):352-355.

Franzoso S., Mazzon C., Sztukowska M., Cecchini P., Kasic T., Capecchi B., Tavano R., and Papini E. (2008). Human monocytes/macrophages are a target of Neisseria meningitidis Adhesin A (NadA). J Leukoc Biol. 83(5):1100-1110.

Freund S., Czech B., Trulzsch K., Ackermann N., and Heesemann J. (2008). Unusual, virulence plasmid-dependent growth behavior of Yersinia enterocolitica in three-dimensional collagen gels. J Bacteriol. 190(12):4111-4120.

Galan J. E., and Wolf-Watz H. (2006). Protein delivery into eukaryotic cells by type III secretion machines. Nature. 444(7119):567-573.

Gangwer K. A., Mushrush D. J., Stauff D. L., Spiller B., McClain M. S., Cover T. L., and Lacy D. B. (2007). Crystal structure of the Helicobacter pylori vacuolating toxin p55 domain. Proc Natl Acad Sci U S A. 104(41):16293-16298.

Gelse K., Pöschl E., and Aigner T. (2003). Collagens - structure, function, and biosynthesis. Adv Drug Deliv Rev. 55(12):1531-1546.

Gentschev I., Dietrich G., and Goebel W. (2002). The E. coli alpha-hemolysin secretion system and its use in vaccine development. Trends Microbiol. 10(1):39-45.

Ghumra A., and Pleass R. J. (2007). Escherichia coli do not express Fc-receptors for human immunoglobulin G (IgG). Mol Immunol. 44(8):2144-2146.

Gilmore R. D. J., Bellville T. M., Sviat S. L., and Frace M. (2005). The Bartonella vinsonii subsp. arupensis immunodominant surface antigen BrpA gene, encoding a 382-kilodalton protein composed of repetitive sequences, is a member of a multigene family conserved among Bartonella species. Infect Immun. 73(5):3128-3136.

Gjörloff Wingren A., Hadzic R., Forsgren A., and Riesbeck K. (2002). The novel IgD binding protein from Moraxella catarrhalis induces human B lymphocyte activation and Ig secretion in the presence of Th2 cytokines. J Immunol. 168(11):5582-5588.

Goebel W., and Hedgpeth J. (1982). Cloning and functional characterization of the plasmid-encoded hemolysin determinant of Escherichia coli. J Bacteriol. 151(3):1290-1298.

Goldberg M. B., Barzu O., Parsot C., and Sansonetti P. J. (1993). Unipolar localization and ATPase activity of IcsA, a Shigella flexneri protein involved in intracellular movement. J Bacteriol. 175(8):2189-2196.

Grassl G. A., Bohn E., Muller Y., Buhler O. T., and Autenrieth I. B. (2003). Interaction of Yersinia enterocolitica with epithelial cells: invasin beyond invasion. Int J Med Microbiol. 293(1):41-54.

Greco D., Salmaso S., Mastrantonio P., Giuliano M., Tozzi A. E., Anemona A., Ciofi degli Atti M. L., Giammanco A., Panei P., Blackwelder W. C., Klein D. L., and Wassilak S. G. (1996). A controlled

85

trial of two acellular vaccines and one whole-cell vaccine against pertussis. Progetto Pertosse Working Group. N Engl J Med. 334(6):341-348.

Gripenberg-Lerche C., Skurnik M., and Toivanen P. (1995). Role of YadA-mediated collagen binding in arthritogenicity of Yersinia enterocolitica serotype O:8: experimental studies with rats. Infect Immun. 63(8):3222-3226.

Gripenberg-Lerche C., Skurnik M., Zhang L., Soderstrom K. O., and Toivanen P. (1994). Role of YadA in arthritogenicity of Yersinia enterocolitica serotype O:8: experimental studies with rats. Infect Immun. 62(12):5568-5575.

Grosskinsky U., Schutz M., Fritz M., Schmid Y., Lamparter M. C., Szczesny P., Lupas A. N., Autenrieth I. B., and Linke D. (2007). A conserved glycine residue of trimeric autotransporter domains plays a key role in Yersinia adhesin A autotransport. J Bacteriol. 189(24):9011-9019.

Grützkau A., Hanski C., Hahn H., and Riecken E. O. (1990). Involvement of M cells in the bacterial invasion of Peyer's patches: a common mechanism shared by Yersinia enterocolitica and other enteroinvasive bacteria. Gut. 31(9):1011-1015.

Gu J., Neary J. L., Sanchez M., Yu J., Lilburn T. G., and Wang Y. (2007). Genome evolution and functional divergence in Yersinia. J Exp Zoolog B Mol Dev Evol. 308(1):37-49.

Guédin S., Willery E., Locht C., and Jacob-Dubuisson F. (1998). Evidence that a globular conformation is not compatible with FhaC-mediated secretion of the Bordetella pertussis filamentous haemagglutinin. Mol Microbiol. 29(3):763-774.

Guyer D. M., Radulovic S., Jones F. E., and Mobley H. L. (2002). Sat, the secreted autotransporter toxin of uropathogenic Escherichia coli, is a vacuolating cytotoxin for bladder and kidney epithelial cells. Infect Immun. 70(8):4539-4546.

Hallström T., Müller S. A., Mörgelin M., Mollenkvist A., Forsgren A., and Riesbeck K. (2008). The Moraxella IgD-binding protein MID/Hag is an oligomeric autotransporter. Microbes Infect. 10(4):374-381.

Han Y. W., and Miller V. L. (1997). Reevaluation of the virulence phenotype of the inv yadA double mutants of Yersinia pseudotuberculosis. Infect Immun. 65(1):327-330.

Hansen J. K., and Forest K. T. (2006). Type IV pilin structures: insights on shared architecture, fiber assembly, receptor binding and type II secretion. J Mol Microbiol Biotechnol. 11(3-5):192-207.

He S. Y., Nomura K., and Whittam T. S. (2004). Type III protein secretion mechanism in mammalian and plant pathogens. Biochim Biophys Acta. 1694(1-3):181-206.

Heesemann J., and Grüter L. (1987). Genetic evidence that the outer membrane protein YOP1 of Yersinia enterocolitica mediates adherence and phagocytosis resistance to human epithelial cells. FEMS Microbiol Lett. 40(1):37-41.

Heise T., and Dersch P. (2006). Identification of a domain in Yersinia virulence factor YadA that is crucial for extracellular matrix-specific cell adhesion and uptake. Proc Natl Acad Sci U S A. 103(9):3375-3380.

Helminen M. E., Maciver I., Latimer J. L., Klesney-Tait J., Cope L. D., Paris M., McCracken G. H. J., and Hansen E. J. (1994). A large, antigenically conserved protein on the surface of Moraxella catarrhalis is a target for protective antibodies. J Infect Dis. 170(4):867-872.

Henderson I. R., Cappello R., and Nataro J. P. (2000). Autotransporter proteins, evolution and redefining protein secretion. Trends Microbiol. 8(12):529-532.

Henderson I. R., Meehan M., and Owen P. (1997). Antigen 43, a phase-variable bipartite outer membrane protein, determines colony morphology and autoaggregation in Escherichia coli K-12. FEMS Microbiol Lett. 149(1):115-120.

86

Henderson I. R., and Nataro J. P. (2001). Virulence functions of autotransporter proteins. Infect Immun. 69(3):1231-1243.

Henderson I. R., Navarro-Garcia F., Desvaux M., Fernandez R. C., and Ala'Aldeen D. (2004). Type V protein secretion pathway: the autotransporter story. Microbiol Mol Biol Rev. 68(4):692-744.

Henderson I. R., and Owen P. (1999). The major phase-variable outer membrane protein of Escherichia coli structurally resembles the immunoglobulin A1 protease class of exported protein and is regulated by a novel mechanism involving Dam and oxyR. J Bacteriol. 181(7):2132-2141.

Hendrixson D. R., and St. Geme J. W. r. (1998). The Haemophilus influenzae Hap serine protease promotes adherence and microcolony formation, potentiated by a soluble host protein. Mol Cell. 2(6):841-850.

Hernandez Alvarez B., Hartmann M. D., Albrecht R., Lupas A. N., Zeth K., and Linke D. (2008). A new expression system for protein crystallization using trimeric coiled-coil adaptors. Protein Eng Des Sel. 21(1):11-18.

Hirono I., Tange N., and Aoki T. (1997). Iron-regulated haemolysin gene from Edwardsiella tarda. Mol Microbiol. 24(4):851-856.

Hodak H., Clantin B., Willery E., Villeret V., Locht C., and Jacob-Dubuisson F. (2006). Secretion signal of the filamentous haemagglutinin, a model two-partner secretion substrate. Mol Microbiol. 61(2):368-382.

Hoiczyk E., Roggenkamp A., Reichenbecher M., Lupas A., and Heesemann J. (2000). Structure and sequence analysis of Yersinia YadA and Moraxella UspAs reveal a novel class of adhesins. EMBO J. 19(22):5989-5999.

Holland I. B., Schmitt L., and Young J. (2005). Type 1 protein secretion in bacteria, the ABC-transporter dependent pathway. Mol Membr Biol. 22(1-2):29-39.

Holm M. M., Vanlerberg S. L., Sledjeski D. D., and Lafontaine E. R. (2003). The Hag protein of Moraxella catarrhalis strain O35E is associated with adherence to human lung and middle ear cells. Infect Immun. 71(9):4977-4984.

Ilver D., Arnqvist A., Ogren J., Frick I. M., Kersulyte D., Incecik E. T., Berg D. E., Covacci A., Engstrand L., and Boren T. (1998). Helicobacter pylori adhesin binding fucosylated histo-blood group antigens revealed by retagging. Science. 279(5349):373-377.

Inatsuka C. S., Julio S. M., and Cotter P. A. (2005). Bordetella filamentous hemagglutinin plays a critical role in immunomodulation, suggesting a mechanism for host specificity. Proc Natl Acad Sci U S A. 102(51):18578-18583.

Jacob-Dubuisson F., Fernandez R., and Coutte L. (2004). Protein secretion through autotransporter and two-partner pathways. Biochim Biophys Acta. 1694(1-3):235-257.

Jacob-Dubuisson F., Locht C., and Antoine R. (2001). Two-partner secretion in Gram-negative bacteria: a thrifty, specific pathway for large virulence proteins. Mol Microbiol. 40(2):306-313.

Jain S., and Goldberg M. B. (2007). Requirement for YaeT in the outer membrane assembly of autotransporter proteins. J Bacteriol. 189(14):5393-5398.

Jani A. A., and Chen P. (2008). Pseudotuberculosis (Yersinia). eMedicine. http://emedicine.medscape.com/article/226871-overview.

Johnson T. L., Abendroth J., Hol W. G., and Sandkvist M. (2006). Type II secretion: from structure to function. FEMS Microbiol Lett. 255(2):175-186.

Jong W. S., ten Hagen-Jongman C. M., den Blaauwen T., Slotboom D. J., Tame J. R., Wickstrom D., de Gier J. W., Otto B. R., and Luirink J. (2007). Limited tolerance towards folded elements during secretion of the autotransporter Hbp. Mol Microbiol. 63(5):1524-1536.

87

Jose J., Jahnig F., and Meyer T. F. (1995). Common structural features of IgA1 protease-like outer membrane protein autotransporters. Mol Microbiol. 18(2):378-380.

Journet L., Hughes K. T., and Cornelis G. R. (2005). Type III secretion: a secretory pathway serving both motility and virulence. Mol Membr Biol. 22(1-2):41-50.

Juhas M., Crook D. W., and Hood D. W. (2008). Type IV secretion systems: tools of bacterial horizontal gene transfer and virulence. Cell Microbiol. 10(12):2377-2386.

Junker M., Besingi R. N., and Clark P. L. (2009). Vectorial transport and folding of an autotransporter virulence protein during outer membrane secretion. Mol Microbiol.

Junker M., Schuster C. C., McDonnell A. V., Sorg K. A., Finn M. C., Berger B., and Clark P. L. (2006). Pertactin beta-helix folding mechanism suggests common themes for the secretion and folding of autotransporter proteins. Proc Natl Acad Sci U S A. 103(13):4918-4923.

Kaiser P. O., Riess T., Wagner C. L., Linke D., Lupas A. N., Schwarz H., Raddatz G., Schäfer A., and Kempf V. A. (2008). The head of Bartonella adhesin A is crucial for host cell interaction of Bartonella henselae. Cell Microbiol.

Kajava A. V., Cheng N., Cleaver R., Kessel M., Simon M. N., Willery E., Jacob-Dubuisson F., Locht C., and Steven A. C. (2001). Beta-helix model for the filamentous haemagglutinin adhesin of Bordetella pertussis and related bacterial secretory proteins. Mol Microbiol. 42(2):279-292.

Kirjavainen V., Jarva H., Biedzka-Sarek M., Blom A. M., Skurnik M., and Meri S. (2008). Yersinia enterocolitica serum resistance proteins YadA and ail bind the complement regulator C4b-binding protein. PLoS Pathog. 4(8):e1000140.

Klauser T., Kramer J., Otzelberger K., Pohlner J., and Meyer T. F. (1993). Characterization of the Neisseria IgA beta-core. The essential unit for outer membrane targeting and extracellular protein secretion. J Mol Biol. 234(3):579-593.

Klauser T., Pohlner J., and Meyer T. F. (1990). Extracellular transport of cholera toxin B subunit using Neisseria IgA protease beta-domain: conformation-dependent outer membrane translocation. EMBO J. 9(6):1991-1999.

Klauser T., Pohlner J., and Meyer T. F. (1993). The secretion pathway of IgA protease-type proteins in Gram-negative bacteria. Bioessays. 15(12):799-805.

Klemm P., Vejborg R. M., and Sherlock O. (2006). Self-associating autotransporters, SAATs: functional and structural similarities. Int J Med Microbiol. 296(4-5):187-195.

Knight C. G., Morton L. F., Peachey A. R., Tuckwell D. S., Farndale R. W., and Barnes M. J. (2000). The collagen-binding A-domains of integrins α1β1 and α2β1 recognize the same specific amino acid sequence, GFOGER, in native (triple-helical) collagens. J Biol Chem. 275(1):35-40.

Konitsiotis A. D., Raynal N., Bihan D., Hohenester E., Farndale R. W., and Leitinger B. (2008). Characterization of high affinity binding motifs for the discoidin domain receptor DDR2 in collagen. J Biol Chem. 283(11):6861-6868.

Koretke K. K., Szczesny P., Gruber M., and Lupas A. N. (2006). Model structure of the prototypical non-fimbrial adhesin YadA of Yersinia enterocolitica. J Struct Biol. 155(2):154-161.

Kostakioti M., and Stathopoulos C. (2004). Functional analysis of the Tsh autotransporter from an avian pathogenic Escherichia coli strain. Infect Immun. 72(10):5548-5554.

Kostakioti M., and Stathopoulos C. (2006). Role of the alpha-helical linker of the C-terminal translocator in the biogenesis of the serine protease subfamily of autotransporters. Infect Immun. 74(9):4961-4969.

Kramer R. Z., Bella J., Mayville P., Brodsky B., and Berman H. M. (1999). Sequence dependent conformational variations of collagen triple-helical structure. Nat Struct Biol. 6(5):454-457.

88

Kramer R. Z., Vitagliano L., Bella J., Berisio R., Mazzarella L., Brodsky B., Zagari A., and Berman H. M. (1998). X-ray crystallographic determination of a collagen-like peptide with the repeating sequence (Pro-Pro-Gly). J Mol Biol. 280(4):623-638.

Laarmann S., Cutter D., Juehne T., Barenkamp S. J., and St. Geme J. W. (2002). The Haemophilus influenzae Hia autotransporter harbours two adhesive pockets that reside in the passenger domain and recognize the same host cell receptor. Mol Microbiol. 46(3):731-743.

Lafontaine E. R., Cope L. D., Aebi C., Latimer J. L., McCracken G. H. J., and Hansen E. J. (2000). The UspA1 protein and a second type of UspA2 protein mediate adherence of Moraxella catarrhalis to human epithelial cells in vitro. J Bacteriol. 182(5):1364-1373.

Lambais M. R., Goldman M. H., Camargo L. E., and Goldman G. H. (2000). A genomic approach to the understanding of Xylella fastidiosa pathogenicity. Curr Opin Microbiol. 3(5):459-462.

Lambris J. D., Ricklin D., and Geisbrecht B. V. (2008). Complement evasion by human pathogens. Nat Rev Microbiol. 6(2):132-142.

Lawrenz M. B., Lenz J. D., and Miller V. L. (2009). A novel autotransporter adhesin is required for efficient colonization during bubonic plague. Infect Immun. 77(1):317-326.

Leduc I., Olsen B., and Elkins C. (2009). Localization of the domains of the Haemophilus ducreyi trimeric autotransporter DsrA involved in serum resistance and binding to the extracellular matrix proteins fibronectin and vitronectin. Infect Immun. 77(2):657-666.

Leduc I., White C. D., Nepluev I., Throm R. E., Spinola S. M., and Elkins C. (2008). Outer membrane protein DsrA is the major fibronectin-binding determinant of Haemophilus ducreyi. Infect Immun. 76(4):1608-1616.

Lee H. W., and Byun S. M. (2003). The pore size of the autotransporter domain is critical for the active translocation of the passenger domain. Biochem Biophys Res Commun. 307(4):820-825.

Leininger E., Roberts M., Kenimer J. G., Charles I. G., Fairweather N., Novotny P., and Brennan M. J. (1991). Pertactin, an Arg-Gly-Asp-containing Bordetella pertussis surface protein that promotes adherence of mammalian cells. Proc Natl Acad Sci U S A. 88(2):345-349.

Leslie A. G. W. (1992). Recent changes to the MOSFLM package for processing film image plate data. Joint CCP4 + ESF-EAMCB Newsletter on Protein Crystallography.

Letley D. P., Rhead J. L., Bishop K., and Atherton J. C. (2006). Paired cysteine residues are required for high levels of the Helicobacter pylori autotransporter VacA. Microbiology. 152(Pt 5):1319-1325.

Li H., and Walker D. H. (1998). rOmpA is a critical protein for the adhesion of Rickettsia rickettsii to host cells. Microb Pathog. 24(5):289-298.

Li L. J., Dougan G., Novotny P., and Charles I. G. (1991). P.70 pertactin, an outer-membrane protein from Bordetella parapertussis: cloning, nucleotide sequence and surface expression in Escherichia coli. Mol Microbiol. 5(2):409-417.

Lian C. J., Hwang W. S., and Pai C. H. (1987). Plasmid-mediated resistance to phagocytosis in Yersinia enterocolitica. Infect Immun. 55(5):1176-1183.

Lindenthal C., and Elsinghorst E. A. (2001). Enterotoxigenic Escherichia coli TibA glycoprotein adheres to human intestine epithelial cells. Infect Immun. 69(1):52-57.

Linke D., Riess T., Autenrieth I. B., Lupas A., and Kempf V. A. (2006). Trimeric autotransporter adhesins: variable structure, common function. Trends Microbiol. 14(6):264-270.

Lipski S. L., Akimana C., Timpe J. M., Wooten R. M., and Lafontaine E. R. (2007). The Moraxella catarrhalis autotransporter McaP is a conserved surface protein that mediates adherence to human epithelial cells through its N-terminal passenger domain. Infect Immun. 75(1):314-324.

89

Long C. G., Braswell E., Zhu D., Apigo J., Baum J., and Brodsky B. (1993). Characterization of collagen-like peptides containing interruptions in the repeating Gly-X-Y sequence. Biochemistry. 32(43):11688-11695.

Lu Y., Iyoda S., Satou H., Satou H., Itoh K., Saitoh T., and Watanabe H. (2006). A new immunoglobulin-binding protein, EibG, is responsible for the chain-like adhesion phenotype of locus of enterocyte effacement-negative, shiga toxin-producing Escherichia coli. Infect Immun. 74(10):5747-5755.

Lupas A. N., and Gruber M. (2005). The structure of alpha-helical coiled coils. Adv Protein Chem. 70:37-78. Mack D., Heesemann J., and Laufs R. (1994). Characterization of different oligomeric species of the

Yersinia enterocolitica outer membrane protein YadA. Med Microbiol Immunol. 183(4):217-227. Mantle M., Basaraba L., Peacock S. C., and Gall D. G. (1989). Binding of Yersinia enterocolitica to rabbit

intestinal brush border membranes, mucus, and mucin. Infect Immun. 57(11):3292-3299. Mantle M., and Husar S. D. (1993). Adhesion of Yersinia enterocolitica to purified rabbit and human

intestinal mucin. Infect Immun. 61(6):2340-2346. Martinez R. J. (1983). Plasmid-mediated and temperature-regulated surface properties of Yersinia

enterocolitica. Infect Immun. 41(3):921-930. Martinez R. J. (1989). Thermoregulation-dependent expression of Yersinia enterocolitica protein 1 imparts

serum resistance to Escherichia coli K-12. J Bacteriol. 171(7):3732-3739. Maurer J., Jose J., and Meyer T. F. (1997). Autodisplay: one-component system for efficient surface display

and release of soluble recombinant proteins from Escherichia coli. J Bacteriol. 179(3):794-804. Maurer J., Jose J., and Meyer T. F. (1999). Characterization of the essential transport function of the AIDA-I

autotransporter and evidence supporting structural predictions. J Bacteriol. 181(22):7014-7020. Mazar J., and Cotter P. A. (2006). Topology and maturation of filamentous haemagglutinin suggest a new

model for two-partner secretion. Mol Microbiol. 62(3):641-654. Mazar J., and Cotter P. A. (2007). New insight into the molecular mechanisms of two-partner secretion.

Trends Microbiol. 15(11):508-515. McMichael J. C., Fiske M. J., Fredenburg R. A., Chakravarti D. N., VanDerMeid K. R., Barniak V., Caplan

J., Bortell E., Baker S., Arumugham R., and Chen D. (1998). Isolation and characterization of two proteins from Moraxella catarrhalis that bear a common epitope. Infect Immun. 66(9):4374-4381.

Meli A. C., Hodak H., Clantin B., Locht C., Molle G., Jacob-Dubuisson F., and Saint N. (2006). Channel properties of TpsB transporter FhaC point to two functional domains with a C-terminal protein-conducting pore. J Biol Chem. 281(1):158-166.

Meng G., St. Geme J. W. r., and Waksman G. (2008). Repetitive architecture of the Haemophilus influenzae Hia trimeric autotransporter. J Mol Biol. 384(4):824-836.

Meng G., Surana N. K., St. Geme J. W. r., and Waksman G. (2006). Structure of the outer membrane translocator domain of the Haemophilus influenzae Hia trimeric autotransporter. EMBO J. 25(11):2297-2304.

Meyaard L. (2008). The inhibitory collagen receptor LAIR-1 (CD305). J Leukoc Biol. 83(4):799-803. Mintz K. P. (2004). Identification of an extracellular matrix protein adhesin, EmaA, which mediates the

adhesion of Actinobacillus actinomycetemcomitans to collagen. Microbiology. 150(Pt 8):2677-2688. Morton L. F., Hargreaves P. G., Farndale R. W., Young R. D., and Barnes M. J. (1995). Integrin α2β1-

independent activation of platelets by simple collagen-like peptides: collagen tertiary (triple-helical) and quaternary (polymeric) structures are sufficient alone for α2β1-independent platelet reactivity. Biochem J. 306 (Pt 2):337-344.

Mota L. J., Journet L., Sorg I., Agrain C., and Cornelis G. R. (2005). Bacterial injectisomes: needle length does matter. Science. 307(5713):1278.

90

Moussatova A., Kandt C., O'Mara M. L., and Tieleman D. P. (2008). ATP-binding cassette transporters in Escherichia coli. Biochim Biophys Acta. 1778(9):1757-1771.

N'Guessan P. D., Vigelahn M., Bachmann S., Zabel S., Opitz B., Schmeck B., Hippenstiel S., Zweigner J., Riesbeck K., Singer B. B., Suttorp N., and Slevogt H. (2007). The UspA1 protein of Moraxella catarrhalis induces CEACAM-1-dependent apoptosis in alveolar epithelial cells. J Infect Dis. 195(11):1651-1660.

Natale P., Bruser T., and Driessen A. J. (2008). Sec- and Tat-mediated protein secretion across the bacterial cytoplasmic membrane - distinct translocases and mechanisms. Biochim Biophys Acta. 1778(9):1735-1756.

Nordström T., Blom A. M., Forsgren A., and Riesbeck K. (2004). The emerging pathogen Moraxella catarrhalis interacts with complement inhibitor C4b binding protein through ubiquitous surface proteins A1 and A2. J Immunol. 173(7):4598-4606.

Nordström T., Blom A. M., Tan T. T., Forsgren A., and Riesbeck K. (2005). Ionic binding of C3 to the human pathogen Moraxella catarrhalis is a unique mechanism for combating innate immunity. J Immunol. 175(6):3628-3636.

Nordström T., Forsgren A., and Riesbeck K. (2002). The immunoglobulin D-binding part of the outer membrane protein MID from Moraxella catarrhalis comprises 238 amino acids and a tetrameric structure. J Biol Chem. 277(38):34692-34699.

Nummelin H., El Tahir Y., Ollikka P., Skurnik M., and Goldman A. (2002). Expression, purification and crystallization of a collagen-binding fragment of Yersinia adhesin YadA. Acta Crystallogr D Biol Crystallogr. 58(Pt 6 Pt 2):1042-1044.

Nummelin H., Merckel M. C., Leo J. C., Lankinen H., Skurnik M., and Goldman A. (2004). The Yersinia adhesin YadA collagen-binding domain structure is a novel left-handed parallel β-roll. EMBO J. 23(4):701-711.

Ohnishi Y., Nishiyama M., Horinouchi S., and Beppu T. (1994). Involvement of the COOH-terminal pro-sequence of Serratia marcescens serine protease in the folding of the mature enzyme. J Biol Chem. 269(52):32800-32806.

Oliver D. C., Huang G., Nodel E., Pleasance S., and Fernandez R. C. (2003). A conserved region within the Bordetella pertussis autotransporter BrkA is necessary for folding of its passenger domain. Mol Microbiol. 47(5):1367-1383.

Oomen C. J., van Ulsen P., van Gelder P., Feijen M., Tommassen J., and Gros P. (2004). Structure of the translocator domain of a bacterial autotransporter. EMBO J. 23(6):1257-1266.

Otto B. R., Sijbrandi R., Luirink J., Oudega B., Heddle J. G., Mizutani K., Park S. Y., and Tame J. R. (2005). Crystal structure of hemoglobin protease, a heme binding autotransporter protein from pathogenic Escherichia coli. J Biol Chem. 280(17):17339-17345.

Otto B. R., van Dooren S. J., Nuijens J. H., Luirink J., and Oudega B. (1998). Characterization of a hemoglobin protease secreted by the pathogenic Escherichia coli strain EB1. J Exp Med. 188(6):1091-1103.

Paerregaard A., Espersen F., Jensen O. M., and Skurnik M. (1991). Interactions between Yersinia enterocolitica and rabbit ileal mucus: growth, adhesion, penetration, and subsequent changes in surface hydrophobicity and ability to adhere to ileal brush border membrane vesicles. Infect Immun. 59(1):253-260.

Pai C. H., and DeStephano L. (1982). Serum resistance associated with virulence in Yersinia enterocolitica. Infect Immun. 35(2):605-611.

91

Pallen M. J., Chaudhuri R. R., and Henderson I. R. (2003). Genomic analysis of secretion systems. Curr Opin Microbiol. 6(5):519-527.

Palmer K. L., and Munson R. S. J. (1995). Cloning and characterization of the genes encoding the hemolysin of Haemophilus ducreyi. Mol Microbiol. 18(5):821-830.

Paunola E., Mattila P. K., and Lappalainen P. (2002). WH2 domain: a small, versatile adapter for actin monomers. FEBS Lett. 513(1):92-97.

Peak I. R., Srikhanta Y., Dieckelmann M., Moxon E. R., and Jennings M. P. (2000). Identification and characterisation of a novel conserved outer membrane protein from Neisseria meningitidis. FEMS Immunol Med Microbiol. 28(4):329-334.

Pearson M. M., Lafontaine E. R., Wagner N. J., St. Geme J. W. r., and Hansen E. J. (2002). A hag mutant of Moraxella catarrhalis strain O35E is deficient in hemagglutination, autoagglutination, and immunoglobulin D-binding activities. Infect Immun. 70(8):4523-4533.

Pearson M. M., Laurence C. A., Guinn S. E., and Hansen E. J. (2006). Biofilm formation by Moraxella catarrhalis in vitro: roles of the UspA1 adhesin and the Hag hemagglutinin. Infect Immun. 74(3):1588-1596.

Perry R. D., and Brubaker R. R. (1983). Vwa+ phenotype of Yersinia enterocolitica. Infect Immun. 40(1):166-171.

Perry R. D., and Fetherston J. D. (1997). Yersinia pestis - etiologic agent of plague. Clin Microbiol Rev. 10(1):35-66.

Pilz D., Vocke T., Heesemann J., and Brade V. (1992). Mechanism of YadA-mediated serum resistance of Yersinia enterocolitica serotype O3. Infect Immun. 60(1):189-195.

Pohlner J., Halter R., Beyreuther K., and Meyer T. F. (1987). Gene structure and extracellular secretion of Neisseria gonorrhoeae IgA protease. Nature. 325(6103):458-462.

Poole K., Schiebel E., and Braun V. (1988). Molecular characterization of the hemolysin determinant of Serratia marcescens. J Bacteriol. 170(7):3177-3188.

Provence D. L., and Curtiss R. r. (1994). Isolation and characterization of a gene involved in hemagglutination by an avian pathogenic Escherichia coli strain. Infect Immun. 62(4):1369-1380.

Pukatzki S., Ma A. T., Sturtevant D., Krastins B., Sarracino D., Nelson W. C., Heidelberg J. F., and Mekalanos J. J. (2006). Identification of a conserved bacterial protein secretion system in Vibrio cholerae using the Dictyostelium host model system. Proc Natl Acad Sci U S A. 103(5):1528-1533.

Ray S. K., Rajeshwari R., Sharma Y., and Sonti R. V. (2002). A high-molecular-weight outer membrane protein of Xanthomonas oryzae pv. oryzae exhibits similarity to non-fimbrial adhesins of animal pathogenic bacteria and is required for optimum virulence. Mol Microbiol. 46(3):637-647.

Raynal N., Hamaia S. W., Siljander P. R., Maddox B., Peachey A. R., Fernandez R., Foley L. J., Slatter D. A., Jarvis G. E., and Farndale R. W. (2006). Use of synthetic peptides to locate novel integrin α2β1-binding motifs in human collagen III. J Biol Chem. 281(7):3821-3831.

Relman D. A., Domenighini M., Tuomanen E., Rappuoli R., and Falkow S. (1989). Filamentous hemagglutinin of Bordetella pertussis: nucleotide sequence and crucial role in adherence. Proc Natl Acad Sci U S A. 86(8):2637-2641.

Remaut H., Tang C., Henderson N. S., Pinkner J. S., Wang T., Hultgren S. J., Thanassi D. G., Waksman G., and Li H. (2008). Fiber formation across the bacterial outer membrane by the chaperone/usher pathway. Cell. 133(4):640-652.

Renn J. P., and Clark P. L. (2008). A conserved stable core structure in the passenger domain beta-helix of autotransporter virulence proteins. Biopolymers. 89(5):420-427.

92

Retamal C. A., Thiebaut P., and Alves E. W. (1999). Protein purification from polyacrylamide gels by sonication extraction. Anal Biochem. 268(1):15-20.

Reyrat J. M., Pelicic V., Papini E., Montecucco C., Rappuoli R., and Telford J. L. (1999). Towards deciphering the Helicobacter pylori cytotoxin. Mol Microbiol. 34(2):197-204.

Riesbeck K., Tan T. T., and Forsgren A. (2006). MID and UspA1/A2 of the human respiratory pathogen Moraxella catarrhalis, and interactions with the human host as basis for vaccine development. Acta Biochim Pol. 53(3):445-456.

Riess T., Andersson S. G., Lupas A., Schaller M., Schafer A., Kyme P., Martin J., Walzlein J. H., Ehehalt U., Lindroos H., Schirle M., Nordheim A., Autenrieth I. B., and Kempf V. A. (2004). Bartonella adhesin A mediates a proangiogenic host cell response. J Exp Med. 200(10):1267-1278.

Rigel N. W., and Braunstein M. (2008). A new twist on an old pathway - accessory secretion systems. Mol Microbiol. 69(2):291-302.

Roggenkamp A., Ackermann N., Jacobi C. A., Truelzsch K., Hoffmann H., and Heesemann J. (2003). Molecular analysis of transport and oligomerization of the Yersinia enterocolitica adhesin YadA. J Bacteriol. 185(13):3735-3744.

Roggenkamp A., Neuberger H. R., Flugel A., Schmoll T., and Heesemann J. (1995). Substitution of two histidine residues in YadA protein of Yersinia enterocolitica abrogates collagen binding, cell adherence and mouse virulence. Mol Microbiol. 16(6):1207-1219.

Roggenkamp A., Ruckdeschel K., Leitritz L., Schmitt R., and Heesemann J. (1996). Deletion of amino acids 29 to 81 in adhesion protein YadA of Yersinia enterocolitica serotype O:8 results in selective abrogation of adherence to neutrophils. Infect Immun. 64(7):2506-2514.

Rojas C. M., Ham J. H., Deng W. L., Doyle J. J., and Collmer A. (2002). HecA, a member of a class of adhesins produced by diverse pathogenic bacteria, contributes to the attachment, aggregation, epidermal cell killing, and virulence phenotypes of Erwinia chrysanthemi EC16 on Nicotiana clevelandii seedlings. Proc Natl Acad Sci U S A. 99(20):13142-13147.

Rose J. E., Meyer D. H., and Fives-Taylor P. M. (2003). Aae, an autotransporter involved in adhesion of Actinobacillus actinomycetemcomitans to epithelial cells. Infect Immun. 71(5):2384-2393.

Rosqvist R., Skurnik M., and Wolf-Watz H. (1988). Increased virulence of Yersinia pseudotuberculosis by two independent mutations. Nature. 334(6182):522-524.

Ruckdeschel K., Roggenkamp A., Schubert S., and Heesemann J. (1996). Differential contribution of Yersinia enterocolitica virulence factors to evasion of microbicidal action of neutrophils. Infect Immun. 64(3):724-733.

Ruiz T., Lenox C., Radermacher M., and Mintz K. P. (2006). Novel surface structures are associated with the adhesion of Actinobacillus actinomycetemcomitans to collagen. Infect Immun. 74(11):6163-6170.

Rutherford N., Charbonneau M. E., Berthiaume F., Betton J. M., and Mourez M. (2006). The periplasmic folding of a cysteineless autotransporter passenger domain interferes with its outer membrane translocation. J Bacteriol. 188(11):4111-4116.

Sakakibara S., Inouye K., Shudo K., Kishida Y., Kobayashi Y., and Prockop D. J. (1973). Synthesis of (Pro-Hyp-Gly) n of defined molecular weights. Evidence for the stabilization of collagen triple helix by hydroxypyroline. Biochim Biophys Acta. 303(1):198-202.

Sanchez-Pulido L., Devos D., Genevrois S., Vicente M., and Valencia A. (2003). POTRA: a conserved domain in the FtsQ family and a class of beta-barrel outer membrane proteins. Trends Biochem Sci. 28(10):523-526.

Sandkvist M. (2001). Biology of type II secretion. Mol Microbiol. 40(2):271-283.

93

Sandt C. H., and Hill C. W. (2000). Four different genes responsible for nonimmune immunoglobulin-binding activities within a single strain of Escherichia coli. Infect Immun. 68(4):2205-2214.

Sandt C. H., and Hill C. W. (2001). Nonimmune binding of human immunoglobulin A (IgA) and IgG Fc by distinct sequence segments of the EibF cell surface protein of Escherichia coli. Infect Immun. 69(12):7293-7303.

Sandt C. H., Hopper J. E., and Hill C. W. (2002). Activation of prophage eib genes for immunoglobulin-binding proteins by genes from the IbrAB genetic island of Escherichia coli ECOR-9. J Bacteriol. 184(13):3640-3648.

Sandt C. H., Wang Y. D., Wilson R. A., and Hill C. W. (1997). Escherichia coli strains with nonimmune immunoglobulin-binding activity. Infect Immun. 65(11):4572-4579.

Sauer F. G., Remaut H., Hultgren S. J., and Waksman G. (2004). Fiber assembly by the chaperone-usher pathway. Biochim Biophys Acta. 1694(1-3):259-267.

Scarselli M., Serruto D., Montanari P., Capecchi B., Adu-Bobie J., Veggi D., Rappuoli R., Pizza M., and Arico B. (2006). Neisseria meningitidis NhhA is a multifunctional trimeric autotransporter adhesin. Mol Microbiol. 61(3):631-644.

Schmid Y., Grassl G. A., Buhler O. T., Skurnik M., Autenrieth I. B., and Bohn E. (2004). Yersinia enterocolitica adhesin A induces production of interleukin-8 in epithelial cells. Infect Immun. 72(12):6780-6789.

Schönherr R., Tsolis R., Focareta T., and Braun V. (1993). Amino acid replacements in the Serratia marcescens haemolysin ShIA define sites involved in activation and secretion. Mol Microbiol. 9(6):1229-1237.

Schulze-Koops H., Burkhardt H., Heesemann J., Kirsch T., Swoboda B., Bull C., Goodman S., and Emmrich F. (1993). Outer membrane protein YadA of enteropathogenic yersiniae mediates specific binding to cellular but not plasma fibronectin. Infect Immun. 61(6):2513-2519.

Schulze-Koops H., Burkhardt H., Heesemann J., von der Mark K., and Emmrich F. (1992). Plasmid-encoded outer membrane protein YadA mediates specific binding of enteropathogenic yersiniae to various types of collagen. Infect Immun. 60(6):2153-2159.

Schulze-Koops H., Burkhardt H., Heesemann J., von der Mark K., and Emmrich F. (1995). Characterization of the binding region for the Yersinia enterocolitica adhesin YadA on types I and II collagen. Arthritis Rheum. 38(9):1283-1289.

Serruto D., Spadafina T., Scarselli M., Bambini S., Comanducci M., Hohle S., Kilian M., Veiga E., Cossart P., Oggioni M. R., Savino S., Ferlenghi I., Taddei A. R., Rappuoli R., Pizza M., Masignani V., and Arico B. (2009). HadA is an atypical new multifunctional trimeric coiled-coil adhesin of Haemophilus influenzae biogroup aegyptius, which promotes entry into host cells. Cell Microbiol.

Shere K. D., Sallustio S., Manessis A., D'Aversa T. G., and Goldberg M. B. (1997). Disruption of IcsP, the major Shigella protease that cleaves IcsA, accelerates actin-based motility. Mol Microbiol. 25(3):451-462.

Sherlock O., Vejborg R. M., and Klemm P. (2005). The TibA adhesin/invasin from enterotoxigenic Escherichia coli is self recognizing and induces bacterial aggregation and biofilm formation. Infect Immun. 73(4):1954-1963.

Sjölinder H., Eriksson J., Maudsdotter L., Aro H., and Jonsson A. B. (2008). Meningococcal outer membrane protein NhhA is essential for colonization and disease by preventing phagocytosis and complement attack. Infect Immun. 76(11):5412-5420.

Skillman K. M., Barnard T. J., Peterson J. H., Ghirlando R., and Bernstein H. D. (2005). Efficient secretion of a folded protein domain by a monomeric bacterial autotransporter. Mol Microbiol. 58(4):945-958.

94

Skurnik M. (1984). Lack of correlation between the presence of plasmids and fimbriae in Yersinia enterocolitica and Yersinia pseudotuberculosis. J Appl Bacteriol. 56(3):355-363.

Skurnik M. (1995). Role of YadA in Yersinia-enterocolitica-induced reactive arthritis: a hypothesis. Trends Microbiol. 3(8):318-319.

Skurnik M., Bölin I., Heikkinen H., Piha S., and Wolf-Watz H. (1984). Virulence plasmid-associated autoagglutination in Yersinia spp. J Bacteriol. 158(3):1033-1036.

Skurnik M., el Tahir Y., Saarinen M., Jalkanen S., and Toivanen P. (1994). YadA mediates specific binding of enteropathogenic Yersinia enterocolitica to human intestinal submucosa. Infect Immun. 62(4):1252-1261.

Skurnik M., and Toivanen P. (1992). LcrF is the temperature-regulated activator of the yadA gene of Yersinia enterocolitica and Yersinia pseudotuberculosis. J Bacteriol. 174(6):2047-2051.

Skurnik M., and Wolf-Watz H. (1989). Analysis of the yopA gene encoding the Yop1 virulence determinants of Yersinia spp. Mol Microbiol. 3(4):517-529.

Smethurst P. A., Onley D. J., Jarvis G. E., O'Connor M. N., Knight C. G., Herr A. B., Ouwehand W. H., and Farndale R. W. (2007). Structural basis for the platelet-collagen interaction: the smallest motif within collagen that recognizes and activates platelet Glycoprotein VI contains two glycine-proline-hydroxyproline triplets. J Biol Chem. 282(2):1296-1304.

St. Geme J. W. r., and Cutter D. (2000). The Haemophilus influenzae Hia adhesin is an autotransporter protein that remains uncleaved at the C terminus and fully cell associated. J Bacteriol. 182(21):6005-6013.

St. Geme J. W. r., Falkow S., and Barenkamp S. J. (1993). High-molecular-weight proteins of nontypable Haemophilus influenzae mediate attachment to human epithelial cells. Proc Natl Acad Sci U S A. 90(7):2875-2879.

St. Geme J. W. I. I. I., Cutter D., and Barenkamp S. J. (1996). Characterization of the genetic locus encoding Haemophilus influenzae type b surface fibrils. J Bacteriol. 178(21):6281-6287.

Stein M., Kenny B., Stein M. A., and Finlay B. B. (1996). Characterization of EspC, a 110-kilodalton protein secreted by enteropathogenic Escherichia coli which is homologous to members of the immunoglobulin A protease-like family of secreted proteins. J Bacteriol. 178(22):6546-6554.

Stevens J. M., Ulrich R. L., Taylor L. A., Wood M. W., Deshazer D., Stevens M. P., and Galyov E. E. (2005). Actin-binding proteins from Burkholderia mallei and Burkholderia thailandensis can functionally compensate for the actin-based motility defect of a Burkholderia pseudomallei bimA mutant. J Bacteriol. 187(22):7857-7862.

Stevens M. P., and Galyov E. E. (2004). Exploitation of host cells by Burkholderia pseudomallei. Int J Med Microbiol. 293(7-8):549-555.

Stevens M. P., Stevens J. M., Jeng R. L., Taylor L. A., Wood M. W., Hawes P., Monaghan P., Welch M. D., and Galyov E. E. (2005). Identification of a bacterial factor required for actin-based motility of Burkholderia pseudomallei. Mol Microbiol. 56(1):40-53.

Studier F. W. (2005). Protein production by auto-induction in high density shaking cultures. Protein Expr Purif. 41(1):207-234.

Suhr M., Benz I., and Schmidt M. A. (1996). Processing of the AIDA-I precursor: removal of AIDAc and evidence for the outer membrane anchoring as a beta-barrel structure. Mol Microbiol. 22(1):31-42.

Surana N. K., Cutter D., Barenkamp S. J., and St. Geme J. W. r. (2004). The Haemophilus influenzae Hia autotransporter contains an unusually short trimeric translocator domain. J Biol Chem. 279(15):14679-14685.

95

Suzuki T., Aono T., Liu C. T., Suzuki S., Iki T., Yokota K., and Oyaizu H. (2008). An outer membrane autotransporter, AoaA, of Azorhizobium caulinodans is required for sustaining high N2-fixing activity of stem nodules. FEMS Microbiol Lett. 285(1):16-24.

Suzuki T., Lett M. C., and Sasakawa C. (1995). Extracellular transport of VirG protein in Shigella. J Biol Chem. 270(52):30874-30880.

Szabady R. L., Peterson J. H., Skillman K. M., and Bernstein H. D. (2005). An unusual signal peptide facilitates late steps in the biogenesis of a bacterial autotransporter. Proc Natl Acad Sci U S A. 102(1):221-226.

Szczesny P., Linke D., Ursinus A., Bar K., Schwarz H., Riess T. M., Kempf V. A., Lupas A. N., Martin J., and Zeth K. (2008). Structure of the head of the Bartonella adhesin BadA. PLoS Pathog. 4(8):e1000119.

Szczesny P., and Lupas A. (2008). Domain annotation of trimeric autotransporter adhesins--daTAA. Bioinformatics. 24(10):1251-1256.

Tamm A., Tarkkanen A. M., Korhonen T. K., Kuusela P., Toivanen P., and Skurnik M. (1993). Hydrophobic domains affect the collagen-binding specificity and surface polymerization as well as the virulence potential of the YadA protein of Yersinia enterocolitica. Mol Microbiol. 10(5):995-1011.

Tan T. T., Forsgren A., and Riesbeck K. (2006). The respiratory pathogen Moraxella catarrhalis binds to laminin via ubiquitous surface proteins A1 and A2. J Infect Dis. 194(4):493-497.

Tan T. T., Nordström T., Forsgren A., and Riesbeck K. (2005). The respiratory pathogen Moraxella catarrhalis adheres to epithelial cells by interacting with fibronectin through ubiquitous surface proteins A1 and A2. J Infect Dis. 192(6):1029-1038.

Tertti R., Skurnik M., Vartio T., and Kuusela P. (1992). Adhesion protein YadA of Yersinia species mediates binding of bacteria to fibronectin. Infect Immun. 60(7):3021-3024.

Thanassi D. G., Stathopoulos C., Karkal A., and Li H. (2005). Protein secretion in the absence of ATP: the autotransporter, two-partner secretion and chaperone/usher pathways of gram-negative bacteria. Mol Membr Biol. 22(1-2):63-72.

(1994). The CCP4 suite: programs for protein crystallography. Acta Crystallogr D Biol Crystallogr. 50(Pt 5):760-763.

Tseng T. T., Tyler B. M., and Setubal J. C. (2009). Protein secretion systems in bacterial-host associations, and their description in the Gene Ontology. BMC Microbiol. 9 Suppl 1:S2.

Uphoff T. S., and Welch R. A. (1990). Nucleotide sequencing of the Proteus mirabilis calcium-independent hemolysin genes (hpmA and hpmB) reveals sequence similarity with the Serratia marcescens hemolysin genes (shlA and shlB). J Bacteriol. 172(3):1206-1216.

Vagin A. A., Steiner R. A., Lebedev A. A., Potterton L., McNicholas S., Long F., and Murshudov G. N. (2004). REFMAC5 dictionary: organization of prior chemical knowledge and guidelines for its use. Acta Crystallogr D Biol Crystallogr. 60(Pt 12 Pt 1):2184-2195.

Valle J., Mabbett A. N., Ulett G. C., Toledo-Arana A., Wecker K., Totsika M., Schembri M. A., Ghigo J. M., and Beloin C. (2008). UpaG, a new member of the trimeric autotransporter family of adhesins in uropathogenic Escherichia coli. J Bacteriol. 190(12):4147-4161.

van Ulsen P., van Alphen L., ten Hove J., Fransen F., van der Ley P., and Tommassen J. (2003). A Neisserial autotransporter NalP modulating the processing of other autotransporters. Mol Microbiol. 50(3):1017-1030.

van Wely K. H., Swaving J., Freudl R., and Driessen A. J. (2001). Translocation of proteins across the cell envelope of Gram-positive bacteria. FEMS Microbiol Rev. 25(4):437-454.

96

Veiga E., de Lorenzo V., and Fernandez L. A. (1999). Probing secretion and translocation of a beta-autotransporter using a reporter single-chain Fv as a cognate passenger domain. Mol Microbiol. 33(6):1232-1243.

Veiga E., Sugawara E., Nikaido H., de Lorenzo V., and Fernandez L. A. (2002). Export of autotransported proteins proceeds through an oligomeric ring shaped by C-terminal domains. EMBO J. 21(9):2122-2131.

Verduin C. M., Hol C., Fleer A., van Dijk H., and van Belkum A. (2002). Moraxella catarrhalis: from emerging to established pathogen. Clin Microbiol Rev. 15(1):125-144.

Visser L. G., Hiemstra P. S., van den Barselaar M. T., Ballieux P. A., and van Furth R. (1996). Role of YadA in resistance to killing of Yersinia enterocolitica by antimicrobial polypeptides of human granulocytes. Infect Immun. 64(5):1653-1658.

Voulhoux R., Bos M. P., Geurtsen J., Mols M., and Tommassen J. (2003). Role of a highly conserved bacterial protein in outer membrane protein assembly. Science. 299(5604):262-265.

Ward C. K., Lumbley S. R., Latimer J. L., Cope L. D., and Hansen E. J. (1998). Haemophilus ducreyi secretes a filamentous hemagglutinin-like protein. J Bacteriol. 180(22):6013-6022.

Welch M. D., and Mullins R. D. (2002). Cellular control of actin nucleation. Annu Rev Cell Dev Biol. 18:247-288.

Wilhelm S., Gdynia A., Tielen P., Rosenau F., and Jaeger K. E. (2007). The autotransporter esterase EstA of Pseudomonas aeruginosa is required for rhamnolipid production, cell motility, and biofilm formation. J Bacteriol. 189(18):6695-6703.

Wollmann P., Zeth K., Lupas A. N., and Linke D. (2006). Purification of the YadA membrane anchor for secondary structure analysis and crystallization. Int J Biol Macromol. 39(1-3):3-9.

Yen M. R., Peabody C. R., Partovi S. M., Zhai Y., Tseng Y. H., and Saier M. H. (2002). Protein-translocating outer membrane porins of Gram-negative bacteria. Biochim Biophys Acta. 1562(1-2):6-31.

Yeo H. J., Cotter S. E., Laarmann S., Juehne T., St. Geme J. W. r., and Waksman G. (2004). Structural basis for host recognition by the Haemophilus influenzae Hia autotransporter. EMBO J. 23(6):1245-1256.

Yu C., Ruiz T., Lenox C., and Mintz K. P. (2008). Functional mapping of an oligomeric autotransporter adhesin of Aggregatibacter actinomycetemcomitans. J Bacteriol. 190(9):3098-3109.

Zaleska M., Lounatmaa K., Nurminen M., Wahlstrom E., and Makela P. H. (1985). A novel virulence-associated cell surface structure composed of 47-kd protein subunits in Yersinia enterocolitica. EMBO J. 4(4):1013-1018.

Zhang P., Chomel B. B., Schau M. K., Goo J. S., Droz S., Kelminson K. L., George S. S., Lerche N. W., and Koehler J. E. (2004). A family of variably expressed outer-membrane proteins (Vomp) mediates adhesion and autoaggregation in Bartonella quintana. Proc Natl Acad Sci U S A. 101(37):13630-13635.