Bacterial Protein Systems at the Membrane Interface Structural and Biophysical Studies of the E

Bacterial protein systems at the membrane interface Structural and biophysical studies of the E. coli adhesion receptor intimin and the magnesium transporter MgtA

Dissertation for the degree of Ph.D. by Julia Anna Weikum

Centre for Molecular Medicine Norway Nordic EMBL Partnership for Molecular Medicine University of Oslo July 2020

Series of dissertations submitted to the Faculty of Mathematics and Natural Sciences, University of Oslo No. 2316

ISSN 1501-7710

Cover: Hanne Baadsgaard Utigard. Print production: Reprosentralen, University of Oslo.

Table of contents

Acknowledgments ...... III

List of publications ...... IV

Abbreviations ...... V

1. Introduction ...... 1

1.1 Pathogenic Escherichia coli ...... 2

1.1.1 Enteropathogenic E. coli (EPEC) and enterohemorrhagic E. coli (EHEC) ...... 2

1.2 Adhesion ...... 3

1.2.1 Initial adherence and colonization ...... 5

1.2.2 Translocation of bacterial signals into the host cell via a type three secretion system 5

1.2.3 Intimate adherence and pedestal formation of A/E lesions ...... 6

1.2.4 Intimin ...... 7

1.2.4.1 Structural features of intimin ...... 8

1.2.4.2 Inverse autotransporter ...... 9

1.2.4.3 Autotransport mechanism of inverse autotransporters ...... 10

1.3 E. coli cell envelope ...... 11

1.3.1 Glycerophospholipids ...... 12

1.3.1.1 Cardiolipin ...... 13

1.3.1.2 Adaption of lipid composition ...... 15

1.3.1.3 Lipid autooxidation ...... 15

1.3.2 Membrane proteins ...... 16

1.3.2.1 P-type ATPases ...... 17

1.3.3 Protein-lipid-interactions ...... 20

1.4 Magnesium-transport in E. coli ...... 22

1.4.1 Magnesium transporter A (MgtA) ...... 23

1.4.1.1 The role of cardiolipin for MgtA-mediated Mg2+ transport ...... 24

1.4.1.2 Phylogenetic distribution of MgtA ...... 25

1.4.1.3 Transcriptional and translational regulation of mgtA expression...... 26

1.4.1.4 Cellular functions of MgtA ...... 29

2. Aims of this thesis ...... 31

3. Synopses of publications ...... 33

Paper I: The extracellular juncture domains in intimin adopt a constitutively extended conformation and induce restraints in the intimin reach and sphere of action ...... 33

Paper II: The bacterial magnesium transporter MgtA reveals highly selective interaction with specific cardiolipin species ...... 34

Paper III: The Mg2+ sensing region of MgtA resides in the C-terminus and is dependent on pH ...... 35

4. Discussion ...... 36

5. Summary ...... 56

6. Future perspectives ...... 58

7. References...... 60

Annex: Paper I-III

Acknowledgments The work presented in this PhD thesis was carried out at the Centre for Molecular Medicine (NCMM), University of Oslo, Norway from April 2017 to January 2019 and at DTU Bioengineering, Danish Technical University, Denmark from February 2019 to June 2020. Financial support was provided by NCMM core funding, the Research Council of Norway, NordForsk and DTU funding. BioCat provided travel grants for courses and national conferences attended during the PhD studies. First, I would like to thank my primary supervisor, Jens Preben Morth, who has given me the opportunity to follow my interest in protein biochemistry and pursue my PhD in his research group. I always appreciate and value his input and comments, innovative ideas and support inside as well as outside of the laboratory. I would also like to thank my co-supervisors, Ole Andreas Løchen Økstad and Reidar Lund, for their support and taking interest in my progress. Secondly, I would like to thank all my colleagues in Oslo and Copenhagen, with whom I had the opportunity and pleasure to work with. I would especially like to thank Saranya Subramani for her contributions to my PhD thesis. Thank you for always being available when I needed guidance or support. Further, I would like to thank the former members of the Morth group, Bojana Sredic, Harmonie Perdreau-Dahl and Johannes Bauer, for the friendly welcome, great scientific input and cozy coffee breaks. I also thank Lisa Gerner for her feedback on my PhD thesis. Additionally, I would like to thank my group members at DTU, Lisa Merklinger and Emilie Müller, for interesting discussions during lunch breaks and fun-filled activities outside of work. Thank you for making my year in Copenhagen a great time. Lastly, I would also like to thank Line Vejby Jægerum, who contributed to my project as a master’s student and remained a friend after. I thank my office mates and co-workers at NCMM and DTU for the scientific and non- scientific discussions in the laboratory and at the coffee machine. Further, I would like to extend my thanks to the technical and administrative staff at DTU and NCMM for their support. I would like to especially thank Nina Modahl, Anita Skolem and Elisa Bjørgo, who helped me with the administrative struggles during my relocation from Oslo to Copenhagen. I would like to thank my family for their encouragement and support in going abroad to perform my PhD studies. Further, I would like to thank my father Gerhard, my mother Liz and my sister Maria for all the valuable feedback on my PhD thesis. I thank all my friends I met in Norway, who made my time outside of the laboratory delightful, allowed me to make great memories and fall in love with the country. Lastly, I would like to thank my partner, Magnus Wannebo, for always being there for me, cheering me up and motivating me. Thank you for being you.

III

List of publications

This thesis is based on the following scientific articles:

I. Weikum J, Kulakova A, Tesei G, Yoshimoto S, Jægerum LV, Schütz M, Hori K, Skepö M, Harris P, Leo JC, Morth JP (2020). The extracellular juncture domains in the intimin passenger adopt a constitutively extended conformation inducing restraints to its sphere of action. Accepted in Scientific Reports (Nature Publishing Group).

II. Weikum J, van Dyck J, Subramani S, Klebl DP, Storflor M, Muench SP, Abel S, Sobott F, Morth JP (2020). The bacterial magnesium transporter MgtA reveals highly selective interaction with specific cardiolipin species. Manuscript under revision.

III. Subramani S, Weikum J, Sredic B, Perdreau-Dahl H, Langbach-Hein K, Vilsen B, Morth JP (2020). The Mg2+ sensing region of MgtA resides in the C-terminus and is dependent on pH. Manuscript in preparation.

Abbreviations A/E Attaching and effacing Arp Actin-related protein AT Autotransporter ATP Adenosine triphosphate A-domain Actuator domain BAM β-barrel assembly machinery Big Bacterial immunoglobulin-like BFP Bundle-forming pili CL Cardiolipin Cls Cardiolipin synthase

C12E8 Octaethylene glycol monododecyl ether DDM n-Dodecyl β-D-maltoside ecMgtA E. coli magnesium transporter A E. coli Escherichia coli EF-P Elongation factor P EHEC Enterohemorrhagic E. coli EPEC Enteropathogenic E. coli

EspFU E. coli secreted protein F in prophage U GFP Green fluorescent protein HUS Hemolytic uremic syndrome IAT Inverse autotransporter Ig Immunoglobulin IRD Intrinsic ribosome destabilization IRTKS Insulin receptor tyrosine kinase substrate kb kilobase

Km Michaelis-Menten constant LEE Locus of enterocyte effacement LysM Lysin motif MC Monte Carlo MD Molecular dynamics

MgtA Magnesium transporter A MgtB Magnesium transporter B Nck Non-catalytic region of tyrosine kinase adaptor protein 1 N-domain Nucleotide binding domain N-WASP Neural Wiskott–Aldrich syndrome protein kDa Kilodalton PE Phosphatidylethanolamine PG Phosphatidylglycerol POPE 1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine P-domain Phosphorylation domain ROS Reactive oxygen species Salmonella Salmonella enterica serovar Typhimurium SAXS Small angle X-ray scattering SDS-PAGE Sodium dodecyl sulphate-polyacrylamide gel electrophoresis SEC Size exclusion chromatography Sec Secretion SERCA Sarco/endoplasmic reticulum Ca2+-ATPase SP Signal peptide stMgtA Salmonella magnesium transporter A stMgtB Salmonella magnesium transporter B T3SS Type III secretion system Tir Translocated intimin receptor TM Transmembrane

Tm Lipid transition temperature UPEC Uropathogenic E. coli

Vmax Maximal velocity V. cholerae Vibrio cholerae

1. Introduction

Escherichia coli (E. coli) is a versatile bacterial species and a common member of the human gut microbiota 1. However, as a pathogen E. coli remains a global health threat. Certain strains, including enteropathogenic E. coli (EPEC) and enterohemorrhagic E. coli (EHEC), are among the most common causes for foodborne diarrheal diseases 2, responsible for millions of acute illnesses and hundreds of deaths annually 3. A deep understanding of the bacterial infection process, ranging from host cell adhesion and colonization to the bacterial defense mechanisms against the human immunological response, is essential for the identification of novel bacterial drug targets and the development of treatments. In this thesis two E. coli proteins, which present potential drug targets, have been structurally and biophysically characterized. The first part of this thesis focused on the virulence factor intimin, essential for the bacterial adhesion process, while in the second part the bacterial magnesium transporter A (MgtA) and its regulation through lipid and magnesium interactions has been studied.

In the following, a short overview of pathogenic E coli (Section 1.1) with a focus on EPEC and EHEC (Subsection 1.1.2) will be given. As the first part of this thesis focuses on the adhesion receptor intimin, the bacterial adhesion process to the host cell (Section 1.2), including its three substages of initial adherence and colonization (Subsection 1.2.1), translocation of bacterial signals into the host cell (Subsection 1.2.2) and intimate adhesion and pedestal formation (Subsection 1.2.3), will be presented. Lastly, the virulence factor intimin will be introduced (Subsection 1.2.4). In the second part of the thesis the interaction between membrane proteins and lipids was investigated on the E. coli bacterial magnesium transporter A (MgtA) and cardiolipin. Therefore, the E. coli cell envelope (Section 1.3) and its main components, glycerophospholipids (Subsection 1.3.1) and membrane proteins (Subsection 1.3.2) with a focus on P-type ATPases (Subsection 1.3.2.1) will be shortly introduced. Additionally, an overview of the interaction between lipids and proteins will be presented (Subsection 1.3.3). Lastly, Section 1.4 will give an overview of Mg2+ transport in bacteria with a focus on MgtA (1.4.1).

1.1 Pathogenic Escherichia coli

E. coli is a Gram-negative bacterium from the family Enterobacteriaceae. Various strains of E. coli persist as commensal bacteria in the mucosa of the gastrointestinal tract 4. They colonize the mucus layer of the cecum and colon within hours of human birth and remain among the most abundant facultative anaerobic bacteria of the human intestinal microflora during human life 4. However, there are several known pathogenic E. coli strains, which have acquired virulence attributes that allow them to adapt to new environmental niches and cause a wide spectrum of diseases 1. Common intestinal pathogens are EPEC and EHEC, which adhere to the small bowel and colon, respectively 1. EPEC and EHEC will be described in detail below (Section 1.1.1). E. coli also plays a role as an extraintestinal pathogen. Uropathogenic E. coli (UPEC) is a common cause for urinary tract infections and meningitis-associated E. coli, which spreads through the blood circulation and translocate to the central nervous system, has been increasingly implicated with cases of meningitis and sepsis 1. Although these pathogenic E. coli strains use different mechanisms and virulence factors during their infection process, all exhibit a multi-step infection pathway. Classical steps include adhesion and colonization of the mucosal site, multiplication and finally, emergence of disease and host damage. A major obstacle for bacterial infection is the human immune system, therefore immune evasion is inevitable for a successful infection 1,5.

1.1.1 Enteropathogenic E. coli (EPEC) and enterohemorrhagic E. coli (EHEC)

Both, EHEC and EPEC, are among the most common foodborne pathogens and are regarded as a global health threat 6,7. EHEC is a highly infectious bacterium, responsible for hemorrhagic colitis (bloody diarrhea) or the potentially fatal hemolytic uremic syndrome (HUS) 8. Symptoms of the HUS are microangiopathic hemolytic anemia, reduced levels of thrombocytes and acute renal failure 9. The largest EHEC outbreak occurred in 2011 in Germany with more than 850 reported cases of HUS and 54 deaths 10. EPEC, on the other hand, is a major contributor to fatal infantile diarrhea 11,12. Both E. coli species belong to the attaching and effacing (A/E) family of gastrointestinal bacteria. A common characteristic of this family is the formation of A/E lesions on the host cell 5. These are characterized by cytoskeletal rearrangements and formation of actin-rich pedestals on which the adherent bacteria reside 2,13. Further, these lesions exhibit the disappearance of surface microvilli 13. The mechanism of bacterial-induced pedestal formation has been extensively

investigated, yet the physiological importance of the pedestal formation and subversion of the actin network is still not well understood 14. EHEC distinguishes from EPEC as its main virulence factor, the Shiga-toxin, is not present in EPEC 5. The toxin disrupts protein synthesis, leading to the death of intoxicated epithelial or endothelial cells 1.

1.2 Adhesion

Bacterial infection begins in most cases with the adhesion of bacteria to the host cell. In the case of EPEC and EHEC, we distinguish three essential stages within the adhesion process, which are required for a successful adhesion event 15. These stages are shown in Figure 1. First, initial adherence and colonization of the intestine occurs (Figure 1a) 15. Upon initial adherence, bacterial signals are translocated into the host cell via the type 3 secretion system (T3SS) as the second stage of adhesion (Figure 1b). Lastly, intimate adherence is mediated by the bacterial proteins intimin and the translocated intimin receptor (Tir) (Figure 1c) 15. Upon intimate adherence, cytoskeletal rearrangements and pedestal formation of A/E lesions occur in the host cell (Figure 1d) 15. In the following, these stages of bacterial adhesion will be described in detail (Section 1.2.1 - 1.2.3).

Figure 1: The adhesion process of EHEC and EPEC is divided into three substages, resulting in pedestal formation of A/E lesions The adhesion process of EHEC and EPEC can be divided into three substages: (a) Initial adherence of the bacteria to the mammalian host cell; (b) Translocation of bacterial signals into the host cell; (c) Intimate adherence of bacteria to the host cell through intimin-Tir interaction. (d) After the adhesion process cytoskeletal rearrangements occur in the host cell, leading to the pedestal formation of A/E lesions. Adapted from Kaper et al. (2004) and Croxen and Finlay (2010) 1,5.

1.2.1 Initial adherence and colonization

Initial adherence requires specific adhesion proteins, which allow close interaction between the bacteria and the epithelium (Figure 1a) 1. Whereas EPEC initially adheres to the epithelial cells in the small bowel, EHEC attaches to the epithelial cells in the colon 1. Although EHEC and EPEC colonize different intestinal regions, they both share several common adhesion proteins 15. However, pathotype-specific adhesins have also been identified. Among the shared adhesion proteins of EHEC and EPEC are the fimbria sorbitol- fermenting protein (Spf), type 1 fimbriae, long polar fimbria (LPF), curli and porcine A/E associated adhesin (Paa) 15. Fimbriae are rod-like structures with a large diameter of 5-10 nm that can interact with the host cell interface 1. Although fimbriae are the most distinct morphological structures, adhesins can take several forms, including as afimbrial surface proteins 16. Adhesins commonly interact with different binding partners on the host cell. In many cases interaction with extracellular matrix proteins has been observed, for example long polar fimbria 1 and curli can interact both with fibronectin and laminin. However, curli also responds to MHC class 1 molecules 17. An EPEC-specific adhesin is the bundle-forming pili (BFP) 15. These pili are rope-like filament bundles that mediate interbacterial adherence and support formation of bacterial networks as well as promote adherence to the host cell epithelium 15. Yet, as adhesins are essential for the initial adherence of bacteria to their host cells, the large overlap of shared adhesins indicates low contribution of adhesins for mediation of host specificity.

1.2.2 Translocation of bacterial signals into the host cell via a type three secretion system

The second step of the adhesion process involves translocation of bacterial signals into the host cell and subsequent interference with host signal transduction pathways. This step is highly dependent on T3SS, a specialized protein secretion apparatus located in the bacterial membrane, which injects effector proteins into the host 18 (Figure 1b). Genes for the T3SS, as well as regulators, chaperones and effector proteins, are encoded on a 35 kilobase (kb) chromosomally located pathogenicity island, called locus of enterocyte effacement (LEE) 14. This pathogenicity island is widely distributed among Gram-negative pathogens. Besides E. coli, it is found in Salmonella enterica serovar Typhimurium (hereafter referred to as Salmonella), Yersinia and

Shigella 14. Although the genes encoding T3SS structural proteins are conserved between species, large variations have been detected among secreted effector proteins 14. The T3SS is a complex machinery consisting of more than 20 proteins. It forms a syringe- shaped structure containing a central channel of 2-3 nm in diameter, which is protruding through the bacterial surface into the host cell 18. The syringe is attached to three ring structures embedded in the inner and outer bacterial membrane, which are connected through the periplasmic inner rod 18. The needle is essential for secretion of T3SS effector proteins. Interestingly, effector proteins are unfolded when they are transported through the channel as the channel size could not accommodate folded proteins 18. Effectors secreted by T3SS exhibit a large functional variety and affect multiple cellular pathways 19. Among these are effectors that are involved in subverting innate immune pathways, including phagocytosis, inflammatory signaling pathways and regulation of host cell survival by promotion or inhibition of apoptosis 19. For a more detailed overview of bacterial effectors, refer to Santos & Finlay (2015) 19 and Pinaud et al. (2018) 20. Among the injected effectors is also a receptor protein called Tir, which is essential for the third stage of adhesion, the intimate adherence of bacteria 21.

1.2.3 Intimate adherence and pedestal formation of A/E lesions

Tir plays an essential role for the third step of adhesion, the intimate adherence of the bacteria to the host. Upon injection into the mammalian cell by T3SS, Tir is presented outside the host cell membrane, where it binds to the virulence factor intimin on the E. coli surface (Figure 1c) 21. Both, intimin and Tir, are encoded on the LEE pathogenicity island 22,23. Tir harbors the intimin binding domain, consisting of two helices separated by a hairpin loop, which is presented on the surface of the host cell 24. There it interacts with the C-terminus of intimin protruding out of the bacterial outer membrane and, following, promotes the tight attachment of the bacteria to the host cell 24. Intimin-Tir interaction activates actin reorganization, which induces formation of pedestal structures on the host cell 25. However, EPEC and EHEC use different intracellular host mechanisms for actin pedestal formation (Figure 1d). In EPEC, Tir recruits host cell factor neural Wiskott–Aldrich syndrome protein (N-WASP) and the actin-related protein (Arp) 2/3 complex 26. These promote actin nucleation and eventually lead to actin polymerization 26. It has been hypothesized that the recruitment of both host cell factors is mediated through a tyrosine residue on Tir 27. Phosphorylation of the tyrosine induces Tir binding to the mammalian non-catalytic region of tyrosine kinase adaptor protein 1 (Nck). Nck

is involved in initiation of actin signaling 27. Additionally, the N-terminus of Tir directly binds to cytoskeletal protein α-actinin, mediating a stable anchor and contributing to the pedestal formation 28. However, this process occurs independently of the tyrosine phosphorylation on Tir 28. On the other hand, the Tir receptor specific for EHEC lacks the tyrosine residue involved in phosphorylation and does not require Nck for pedestal formation 29. Tir contains an Asn-Pro- Tyr (NPY458) sequence in the C-terminus that interacts with insulin receptor tyrosine kinase substrate (IRTKS), a key regulator of membrane and actin dynamics. IRTKS recruits secreted bacterial effector EspFU (E. coli secreted protein F in prophage U), which together form a ternary complex with N-WASP 30. N-WASP activates Arp2/3-complex mediated signaling processes for cytoskeletal rearrangements 31. Efficient downstream signaling is further promoted through dimerization of intimin and Tir, which promotes receptor clustering 32. Additionally, different intimin variants have been shown to interact with various eukaryotic proteins, such as nucleolin or β1-integrin, which contribute to the intimate adherence of the bacteria 33,34.

1.2.4 Intimin

Intimin, a 94 kilodaltons (kDa) surface protein, is an essential adhesin for EHEC and EPEC infection. Homologues are also found in other pathogens, such as UPEC 35, Citrobacter rodentium and Hafnia alvei 36. Intimin is uniformly expressed over the entire bacterial membrane 37. However, intimin expression is regulated by environmental and host cell factors. Its expression levels increase during the exponential growth phase and decrease following adhesion to the host cell 37. Intimin variants have shown variability in the exposed C-terminal region, allowing the generation of more than 20 variants and allele-specific subtypes of intimin. The most common clinically relevant types are the α- and β-type, found in several EPEC strains including the common strain O127:H6, and the γ-type, which is only detected in specific EPEC strains and EHEC O157:H7 17.

Figure 2: Intimin exhibits the classical structural composition of inverse autotransporters (a) Schematic of the structural composition of intimin containing a periplasmic, a transmembrane and an extracellular segment. (b) Structural comparison between an inverse and a classical autotransporter. Adapted from Leo et al. (2015) 36. (c) Hypothesized process of intimin passenger secretion according to the hairpin model described for classical autotransporters. Adapted from Leo et al. (2012) 38.

1.2.4.1 Structural features of intimin

Structurally, intimin can be divided into three segments. Intimin contains a small periplasmic sequence, including a lysin motif (LysM) domain and a signal peptide (SP), and a β-barrel located in the outer membrane 36. The third segment is the extracellular region consisting of four bacterial immunoglobulin-like (Big) domains (D00-D0-D1-D2) topped by a C-type lectin-like domain (D3)

at the C-terminus (Figure 2a) 36. In the following, we will refer to the extracellular region only as the passenger, as recommended by Drobnak et al. (2015) 39. The structural composition of intimin corresponds to the typical assembly of an inverse autotransporter, which will be discussed in detail further on (Section 1.2.4.2). The periplasmic domain of intimin was shown to mediate receptor dimerization, important for receptor clustering during the adhesion process 40. The bacterial outer membrane is spanned by the 12-stranded anti-parallel β-barrel of intimin, followed by an elongated linker that extends into the barrel pore 41. Intimin requires the β-barrel assembly machinery (BAM) complex for its proper insertion into the outer membrane 42, and a putative BAM signature sequence has been identified in the final strand of the β-barrel 41. The extracellular passenger consists of a chain of subdomains, which form a rod-shaped protruding extension 36. However, its exact composition has long been under debate as only structural information of the C-terminal subdomains D1-D2- D3 has been obtained 24. Intimin subdomains D2 and D3 form a superdomain, interacting directly with Tir 24. D3 exhibits a C-type lectin-like domain, while D1 and D2 exhibit the typical immunoglobulin (Ig)-fold of Big domains. The Ig fold is typically composed of 70-100 amino acids, which are arranged in seven anti-parallel β-strands, organized in two β-sheets and packed against each other in a β-sandwich 43. The β-strands are composed of alternating hydrophobic and hydrophilic residues with the hydrophobic side chains pointing towards the interior of the domain 43. Different subsets of Ig-like domains have been identified and are distinguished by their topology. The canonical feature of Ig-like domains is a disulfide bridge between two conserved cysteine residues 43, which is not present in intimin subdomains D1 and D2. An additional Big domain (D0) at the N-terminus of the intimin passenger has been predicted based on sequence similarity 24. The presence of another subdomain, termed D00, located at the interface of the β- barrel and the extracellular passenger has been suggested, yet no structural fold could be proposed 41. In 2016, Leo et al. predicted, based on homology-based structure prediction, that D00 also exhibits an Ig fold 44. However, no high-resolution structural information of D00 was obtained.

1.2.4.2 Inverse autotransporter

Intimin is the prototype of the type Ve secretion system, termed inverse autotransporter (IAT). The name derives as its extracellular passenger, located at the C-terminus, is exported with the help of an N-terminally located β-barrel 36. Therefore, it exhibits a reverse conformation in

comparison to the family of classical autotransporters (AT), which are characterized as type Va secretion systems 36. Figure 2b shows structural differences between the passenger of IATs and classical ATs. The passenger of classical ATs is located at the N-terminus followed by the membrane-embedded β-barrel 36. Additionally, while the passenger of ATs is formed by extended β-helices, the passenger of IATs typically consists of a chain of Big domains, often capped with a C-type lectin-like domain at the C-terminus 36. Besides intimin, many other IATs have been identified, including Yersinia invasin, E. coli FdeC and Salmonella virulence factor SinH 45,46. Especially, Yersinia pseudotuberculosis invasin has been in the focus of research as it exhibits a similar composition and function compared to intimin 41,47. In general, IATs share several common structural features. Most prominently is the β-barrel domain and its linker region 46 (Figure 2b). Large variations between the C-terminal passengers of IATs have been detected. Some IATs do not contain any Big domains or only a short C-terminal extension, while for others up to 47 Ig-like domains have been predicted 45,46. Interestingly, IATs are almost exclusively found in Gammaproteobacteria 46.

1.2.4.3 Autotransport mechanism of inverse autotransporters

Autotransporters secrete their passenger autonomously across the outer membrane, independent of adenosine triphosphate (ATP) hydrolysis or membrane potential as an energy source 38. However, the exact mechanism for the secretion through the β-barrel remains unclear. Figure 2c illustrates an initial model, adapted from classical ATs, which has been proposed for intimin passenger secretion. In the first step unfolded intimin is exported through the cytoplasmic membrane into the periplasm via the secretion (Sec) machinery dependent on a N-terminal signal peptide (Figure 2c – Step 1) 36. Periplasmic folding and outer membrane insertion of intimin is dependent on specific periplasmic chaperones (Figure 2c – Step 2) 42. DsbA catalyzes the formation of disulfide bonds in intimin subdomain D3 42. The general chaperone SurA prevents protein aggregation and assist in the insertion of the N-terminal region into the outer membrane 42. Chaperones Skp and DegP play a secondary role in intimin insertion 42. However, for IAT invasin DegP has been linked to quality control of the insertion process 48. Insertion of the β-barrel into the outer membrane is an essential step in the autotransport process as the passenger will be secreted through the pore into the extracellular space 49. Yet, β-barrel insertion is dependent on auxiliary factors, such as the BAM complex 41,42 (Figure 2c – Step 2). For the secretion of the passenger of IATs itself a

hairpin model has been proposed. A short linker region between the β-barrel and the passenger forms a hairpin intermediate inside the membrane pore, which initiates secretion (Figure 2c – Step 3) 36,48. Following, the passenger is pulled through the membrane pore in a way that the N- terminus of the passenger reaches the extracellular milieu first. Sequential folding of individual Big domains has been proposed as the main driving force for secretion, resulting in complete secretion and folding of the intimin passenger (Figure 2c –Step 4 & 5) 44. Leo et al. revealed that disturbance of the Ig fold through insertion of a tag or deletion of a β-strand results in stalling of passenger secretion 44. However, the hairpin model has been under debate. An involvement of the BAM complex for passenger export, next to its support for β-barrel insertion, has been proposed as the β-barrel remains associated with the BAM complex despite being fully folded and inserted into the membrane 36,44. Yet, no direct evidence for the involvement of the BAM complex in passenger secretion of IATs has been identified, but its role in classical ATs secretion has been revealed 50,51.

1.3 E. coli cell envelope

Biological membranes are key cellular components as they separate cells from the extracellular environment and allow functional compartmentalization. They consist of a lipid matrix and embedded and attached proteins 52. Yet, each type of cell membrane has distinct functions dependent on the complex lipid mixture and the unique set of associated proteins 52. The cell envelope of Gram-negative bacteria, like E. coli, consists of four compartments: the inner membrane, the periplasm, the peptidoglycan layer, also referred to as the cell wall, and the outer membrane 53. The inner, or cytoplasmic, membrane is the primary permeability barrier of the cell. It contains specific transport proteins and permeases as well as enzymes involved in ATP and phospholipid synthesis 53. The outer membrane exposes the antigenic determinants, lipopolysaccharides, to the environment and functions as a passive barrier against substrates with large molecular weights, for example antibiotics 53. However, both membranes consist mainly of the same building block, glycerophospholipids 53. The classical fluid mosaic model by Singer and Nicholson has long been the standard model for describing the lipid bilayer. It characterizes the bilayer as an unperturbed, homogenous surface with random distribution of lipids and few embedded membrane proteins 54. However, recent research has updated this model. It has been highlighted that local areas composed of

specific lipids are present, often described as regions of functional specialization, and that the membrane bilayer is packed with intercalating membrane proteins 54. In the following, the two main components of the bacterial membrane, glycerophospholipids (Section 1.3.1) and membrane proteins (Section 1.3.2), and their interaction with one another (Section 1.3.3) will be discussed in detail.

1.3.1 Glycerophospholipids

The most abundant lipids in E. coli membranes are glycerophospholipids 55. The glycerophospholipid structure can be divided into two functional parts: the hydrophilic headgroup and the hydrophobic tail consisting of two long hydrocarbon chains, referred to as acyl chains (Figure 3a). The lipid backbone is a sn-3-glycerol-3-phosphate, which is esterified to the acyl chains at the first and second position 52. The lipid head group is linked via its phosphoryl group at position three. Chemically diverse structures can be found as headgroups and their composition defines the different lipid classes. Examples of different lipid head groups are shown in figure 3a. The acyl chains of the hydrophobic tail consist of fatty acids or fatty alcohols, varying in their length and degree of saturation 56. In eubacteria, fatty acid chain length varies typically from 12 to 18 carbons and can be fully saturated or monounsaturated 56. Acyl chain properties are classically described by the symbol x:y, in which x refers to the number of carbon atoms and y to the number of double bonds present in the fatty acid chains 52. The modular composition of lipids is possible through variations of the polar headgroup or the composition of the acyl chains, which allows a large molecular variety of glycerophospholipids 57. The principal glycerophospholipid in the E. coli cytoplasmic membrane is the zwitterionic primary amine- containing phosphatidylethanolamine (PE), which makes up ~75-85 % of the total lipid content 53,55. It functions primarily as a structural lipid in membranes as it forms hydrogen bonds with neighboring polar groups 58. The remaining lipid components, phosphatidylglycerol (PG) (~10-20 %) and cardiolipin (CL) (~5-10 %), are both anionic lipids 53. However, CL exhibits a unique structure with distinct features, which will be discussed in detail below (Section 1.3.1.1). The properties of membrane bilayers are highly dependent on the physical and chemical characteristics of the lipids they contain (Figure 3b). Next to reactivity and charge of the polar headgroup, the composition of the hydrophobic tail also plays a major role 57. Acyl chain composition regulates membrane viscosity as lipids with saturated acyl chains pack with higher density and tend to form non-fluid gel phases 57. They are characterized by a high degree of acyl

chain order, reduced lateral lipid mobility and increased membrane viscosity 57. However, monounsaturated acyl chains exhibit a kinked shape, leading to less tight packing and increased mobility. Therefore, they tend to form fluid bilayers at physiological temperatures 57. Polyunsaturated lipids differ in their physical properties from both saturated and unsaturated lipids 57. However, bacteria lack polyunsaturated lipids 59. Lipid properties are also sensitive to temperature. The lipid transition temperature (Tm) is defined as the temperature, at which lipids transition from the ordered gel phase, which contains closely packed lipids, to the liquid disordered phase, in which acyl chains are randomly oriented and show increased fluidity 60.

1.3.1.1 Cardiolipin

Cardiolipin (1,3-diphosphatidyl-sn-glycerol) is a unique phospholipid dimer. As a phospholipid dimer, its major structural difference in comparison to classical glycerophospholipids is that a single headgroup glycerol is shared by two phosphatidate moieties 61. Therefore, the CL structure essentially consists of two glycerophospholipid molecules with four acyl chains in total, as shown in figure 3a. As both phosphodiester moieties should be negatively charged under physiologically relevant conditions, CL is classified as an anionic lipid 61. Additionally, CL has a very small headgroup in comparison to its hydrophobic tail. The small head size enhances the probability of CL to form inverted non-lamellar lipid phases and its presence in the lipid bilayer induces negative curvature stress 61. The cell poles of E. coli and other bacteria are often enriched in CL 62. Based on the conical lipid shape of CL, it was proposed that CL clusters at the cell poles due to spontaneous, transverse and lateral lipid microphase separation 63. However, the physical properties of CL-containing membranes depend on various factors including acyl chain composition, the solvent environment, the presence of counter cations and the composition of the remaining lipid environment 63. Although the percentage of CL normally remains low in the bacterial membrane, it is strongly involved in the cellular energy metabolism through interactions with enzymes involved in oxidative phosphorylation and ATP synthesis 64. Due to its negative charge and organization of microdomains, a function of CL as a proton sink has been proposed 64. Further, a role of CL as a “flexible linker” has been suggested, filling cavities at protein interfaces and allowing stabilization of individual subunits of oligomeric complexes 64. It was shown that the mobility of the CL headgroup is impaired, leading to reduced intra- and intermolecular interactions with other headgroup moieties. Therefore, the steric self-hindrance of CL in the lipid bilayer is likely reduced, which allows a higher accessibility of CL phosphate groups to interact with water,

metal ions and integral or surface bound membrane proteins 61. This highlights the specific significance of CL for lipid-protein interactions. As CL contains four acyl chains, more than 100 different CL species are theoretically possible through different combinations of acyl chains varying in their saturation degree and length. However, an unexpected symmetry in lengths of the four acyl chains of CL has been observed in biological membranes 59. It was proposed that an imbalance between chain lengths may disrupt the CL resonance structure and four acyl chains of the same length will symmetrically stabilize the CL headgroup 59. In E. coli, 56 species were identified with palmitic acid (16:0), palmitoleic acid (16:1) and oleic acid (18:1) as the most abundant acyl chains 65. Three enzymes are responsible for CL synthesis in E. coli. Cardiolipin synthase A (ClsA) synthesizes CL in the exponential phase, while cardiolipin synthase B and C (ClsB and ClsC) contribute to the lipid synthesis in the stationary growth phase 66. In eukaryotes CL is essentially only found in the mitochondrial membrane, where it contributes up to ~20 % to the total lipid content 67. Despite the large molecular diversity and number of potential CL species, CL is often limited to one or two major species in eukaryotic cells, which account for 60-90 % of the total CL fatty acid mass 68. As an example, CL extracted from bovine heart contains linoleic acid (18:2) as the dominant acyl chain 69.

Figure 3: Glycerophospholipids are a major building block of biological membranes (a) Schematic of the composition of classical glycerophospholipids in comparison to the phospholipid dimer cardiolipin. (b) Physicochemical properties of the lipid bilayer dependent on lipid composition. Adapted from Ernst et al. (2016) 57.

1.3.1.2 Adaption of lipid composition

The lipid composition of bacterial cells is highly adaptable to changes in the environment. Adaption is mediated through variations of the lipid class or the acyl chain composition of bacterial membranes 57. During the bacterial growth cycle the proportion of CL can increase from ~5 % up to ~30 % in the stationary phase 65, which has been linked to increased CL synthase activity 70. Additionally, changes in osmolarity lead to increased proportions of anionic phospholipids and decreased zwitterionic lipids 71. Especially, the proportion of CL has been observed to increase under osmotic stress or impaired energy metabolism 71. The function of CL under osmotic stress is not known. However, it was proposed that an upregulation of CL synthesis allows CL- dependent regulation of osmosensory transporter ProP 71. Additionally, fatty acid composition in E. coli adapts to environmental conditions. A classical phenomenon, homeoviscous adaptation, describes the adaptive response of the lipid environment to changing temperatures in order to maintain membrane fluidity 57. While the polar head group composition remains unchanged, increased percentages of saturated acyl chains are detected at elevated temperatures 72,73. Further, elevated temperatures also exhibit an increase in cyclopropane fatty acid formation 74. Cyclopropane increase also occurs under high salt conditions 75, acidic media and growth under anaerobic conditions 76. Additionally, the acyl chain composition also changes depending on the bacterial growth phases with decreasing levels of unsaturated fatty acids as the cultures progresses from the exponential growth phase to the stationary phase 77.

1.3.1.3 Lipid autooxidation

As lipids are essential for cell compartmentalization, damages to the lipid integrity can destabilize membranes, thus increasing membrane permeability and leading to cell lysis and death 78. A common cause for destruction of membrane integrity is lipid autooxidation 79. This process occurs through a radical chain mechanism upon interaction of lipids with reactive oxygen species (ROS). Lipid autooxidation is a self-propagating and self-accelerating process and it is divided in three distinct stages: initiation, propagation and termination 79. During initiation, unsaturated lipids lose a hydrogen atom and form free radicals in the presence of initiators, such as light, heat or metal ions. In the second stage, propagation, the lipid radical reacts with oxygen and forms a peroxyl radical, which can attack a new lipid molecule and induce a rapidly progressing reaction. This

reaction will be repeated until no hydrogen source is available anymore or when the propagation reaction is interrupted, for example by antioxidants 79. This represents the final stage, termination, and concludes the lipid autooxidation process. Although a variety of lipid oxidation products have been identified, hydroperoxides represent the primary products 80. However, they can decompose further into various oxygenated and aliphatic fatty acid scission products 80. Lipid stability depends on various factors, including the degree of unsaturation of fatty acids, making highly unsaturated lipids most susceptible to oxidation 80. Lipid oxidation has been shown to influence the integrity of biological systems altering membrane fluidity and membrane composition and disrupting lipid- protein interactions 78,81.

1.3.2 Membrane proteins

The membrane bilayer is packed with membrane proteins involved in various cellular functions ranging from transporters moving molecules across the membrane to receptors transmitting signals between the intra- and extracellular environment. Like lipids, membrane proteins are amphipathic. They can be classified into two general categories: integral (intrinsic) and peripheral (extrinsic) proteins 52. Integral proteins are tightly bound to the membrane through hydrophobic forces and they contain one or more segment that is embedded in the membrane 52. In most cases integral proteins even span the entire phospholipid bilayer, therefore being classified as transmembrane proteins. Peripheral proteins are not embedded into the lipid bilayer but only attach to the membrane surface. They do not interact with the hydrophobic core of the lipid bilayer, but rather with the surface exposing domains of lipid molecules or integral proteins 52. As membranes promote cellular compartmentalization, a major function of membrane proteins is maintaining ion homeostasis across the membrane. Proteins that transport ions across the lipid bilayer are separated into two classes: ion channels and ion pumps 82. The first class, ion channels, are passive conduits. Ions rush through channels dependent on concentration and electric potential gradients 82. The second class, ion pumps, require energy, released from ATP hydrolysis or another source, to actively transport ions against gradients 82. Therefore, they play an essential role in building up ion gradients and electric potentials across membranes 82. A major difference between ion channels and pumps is that the first class only requires the presence of a single gate, which restricts ion movement along the translocation pathways. An ion pump, on the other hand, needs at least two gates, which should never be open at the same time to avoid uncontrolled ion movement 82.

1.3.2.1 P-type ATPases

The P-type ATPase family constitutes a large class of ion transporters 83. P-type ATPases are involved in establishing and maintaining steep electrochemical gradients across biological membranes, making them essential for all eukaryotic organisms and most prokaryotes 84. P-type ATPase- mediated ion transport is coupled to ATP hydrolysis, which provides the energy required for ion transport across the respective membranes 83. A wide variety of ions are being transported by P-type ATPases as substrates 83. Among the most studied and well-known representatives of this transporter family is the Na+-K+-ATPase, responsible for maintaining the electrochemical gradient across the cytoplasmic membrane in most animal cells 83. In the focus of research has also been the sarco/endoplasmic reticulum Ca2+-ATPase (SERCA), which pumps Ca2+ back into the lumen of the sarcoplasmic reticulum during muscle relaxation 83. Further, the gastric H+-K+- ATPase, which maintains low pH in the stomach lumen by active transport of H+, has been extensively investigated 83. Lastly, plasma membrane H+-ATPases are essential for the establishment of membrane potentials in plants and fungi, which energize the plasma membrane for cellular processes 85. There are five branches of P-type ATPases with more than ten subgroups, which are mainly distinguished by their transported ions (Appendix A1) 84,86,87.

1.3.2.1.1 Structural features of P-type ATPases

All P-type ATPases are multi-domain membrane proteins with molecular masses between 70-150 kDa 88. P-type ATPases share a common structural composition, which is illustrated in figure 4a. The main catalytic unit (α-subunit) consists of six to twelve transmembrane (TM) α-helices and comprises the ion transport domain 84. In some cases this domain is divided into two subdomains, called T-domain, consisting of the first six, and S-domain, consisting of the remaining TM- helices 87 (Figure 4a). While the T-domain harbors the ion binding sites and is present in all P- type ATPases, the S-domain is an auxiliary unit, which can have specialized functions, such as ion-coordination or additional ion binding sites 87. Both, the C- and N-terminus of P-type ATPases, are on the membrane side facing the cytoplasm, therefore, all P-type ATPases have an even number of TM segments 88. Additionally, all P-type ATPases have three cytoplasmic domains, which confer the ATP- hydrolyzing activity 84. These are termed the N-domain (nucleotide binding domain), P-domain (phosphorylation domain) and A-domain (actuator domain) 84 (Figure 4a). The N-domain performs

ATP binding and phosphorylates the P-domain 87. During this process, the N-domain recognizes and positions the γ-phosphoryl of ATP for a nucleophilic attack 84. Additionally, coordination with Mg2+ allows to bring the ATP molecule closer to the phosphorylation site and promotes the nucleophilic attack and phosphoryl transfer 89. All P-type ATPases contain a highly conserved aspartate residue in the P-domain, which accepts the phosphoryl group and forms a high energy aspartyl-phosphate intermediate 84. The A-domain functions as a built-in phosphatase involved in dephosphorylation of the P-domain. A glutamate residue in the A-domain positions a water molecule for the subsequent hydrolysis and release of the phosphoryl group 84. Essential for ion transport is the conversion of the energy released through ATP hydrolysis in the cytoplasm to the physical translocation of ions through the TM segment 84. This is conversed through five linker regions that connect the cytoplasmic domains to the transmembrane section. It was revealed that the integrity and proper length of these linkers is highly relevant for P-type ATPase activity. The deletion of single residues in the linker region severely affected the catalytic activity of SERCA, even leading to complete inhibition 90,91. The ion binding sites are located in the TM domain between helices M4, M5, M6 and M8, where they are coordinated by negatively charged and polar residues 89. Interestingly, SERCA, the Na+-K+- and the H+-K+-ATPase show high similarity in their ion binding sites, although they transport different ions in different quantities in each direction 89. This indicates that the same binding sites reorient to accommodate different ions. Yet, distribution and number of charged amino acids in the sites likely have been adapted to the respective ions 89. Additionally, it proposes that ion selectivity might rather be mediated through a gating mechanism and a selectivity filter at the ion entrance pathway 84.

1.3.2.1.2 Catalytic cycle of P-type ATPases

To achieve active transport of ions against an electrochemical gradient, an open passage across the membrane needs to be avoided as this would result in rapid flow-back of the transported ions 89. P-type ATPases use an alternate-access model, in which both membrane sides are transiently closed and transported ions are occluded 89. Selected transport is then mediated through opening of a narrow pathway to either membrane side. This transport mechanism is only possible through extensive conformational changes of P-type ATPases, driven by ATP hydrolysis 89. The catalytic cycle follows the Post-Albers scheme and alternates between several

different conformational states. Two main conformations, termed E1 and E2 state, are distinguished, which differ in their affinity towards the transported ions 87 (Figure 4b). The first state, E1, exhibits high affinity for ion 1, which is the ion transported out of the cytoplasm 89. Ion binding triggers phosphorylation of the P-domain by Mg2+-ATP, forming the high- energy E1P·ADP state. Release of ADP induces conformational changes and the transporter reaches the low-energy state E2P 88,89. These conformational changes distort the ion binding site(s) and lower the affinity for ion 1. Simultaneous opening at the extracytoplasmic side results in exit of ion 1 to the extracytoplasmic space 88,89. Conformational state E2P binds ion 2, referred to as the counterion, with high affinity. Binding of ion 2 induces re-occlusion and dephosphorylation, resulting in conformational switching from the E2P to the E2 state. The release of the counterions into the cytoplasm completes the cycle and induces transition to the E1 state again 88,89. Exchange of ions occurs through half-channels oriented towards the cytoplasmic side (E1-ATP state) or the extracytoplasmic side (E2-P state) 87.

Figure 4: P-type ATPases are a large family of ion transporters moving ions across membranes according to the Post-Albers scheme (a) P-type ATPases classically consist of a transmembrane and a cytoplasmic segment with an actuator (A-), a phosphorylation (P-) and a nucleotide (N-) domain. (b) Schematic overview of the P-type ATPase transport cycle according to the Post-Albers scheme. Blue refers to E1 states with high affinity for ion 1 (light blue); Red refers to E2 states with high affinity for ion 2/ counterion (orange). Adapted from Palmgren and Nissen (2011) 87.

1.3.3 Protein-lipid-interactions

In the past the lipid bilayer was mostly considered to function only as a solvent media for membrane proteins 54. However today, the importance of lipid-protein interactions has been highlighted for hundreds of membrane proteins 92 and the characterization of lipid-protein interactions has allowed a better understanding of cell membrane organization. Lipids have been implicated with various membrane protein functions. These range from protein activity, stability and oligomerization as well as protein localization, folding and topology 92–94. Lipid-protein interplay can occur through interactions between selected lipid molecules binding to a specific site on the protein or through the general physicochemical properties of the lipid bilayer dependent on its composition 93. Figure 5 shows the three classical categories of lipid- protein interactions: bulk lipids, annular lipids and non-annular lipids 93. Bulk lipids are lipid molecules present in the bilayer, which interact non-specifically with the membrane protein. They exhibit fast diffusion rates and low resilience time at the protein surface 93. The second category, annular lipids, are comprised of lipids from a selected class or molecular species. They often interact with either hydrophobic or hydrophilic membrane protein surfaces 93. Simulations indicated that a membrane protein is typically engulfed by 50-100 lipid molecules that form a shell, termed annular lipid shell, around the TM segment 95. These lipids exhibit a reduced exchange rate in comparison to bulk lipids 93. The last category, non-annular lipids, are specific lipid molecules that often bind at selective site-specific binding sites on the protein. These lipids often reside within the membrane protein complex. They exhibit a low exchange rate and, due to the tight interaction, can in many cases be co-purified with membrane proteins 93. Additionally, physicochemical properties of the lipid bilayer can also affect membrane protein functions 93. The hydrophobic thickness of the bilayer, typically between 35-55 Å, is defined by its lipid composition. Hydrophobic mismatch describes the impaired compatibility between the hydrophobic surfaces of the membrane protein to the lipid bilayer 93. In case of incompatibility, local changes occur through recruitment of additional lipid molecules, deformation of the membrane or conformational changes of the protein. Hydrophobic mismatch has been revealed as a molecular mechanism for transporter regulation 93. Additionally, lateral pressure, which is defined as the pressure exerted by the membrane on its components located inside it, has been shown to affect membrane proteins. The highest pressure is typically at the interfacial region between hydrophobic and hydrophilic areas. However, lateral pressure is also dependent on local asymmetries across the bilayer, the lipid order and hydrophobic mismatch 93. In the

following, selected examples of lipid-protein interactions and their role for protein functions will be presented. The importance of the lipid environment or specific lipid interactions for protein activity has been observed for many membrane proteins, including several representatives of the P-type ATPase family. The ATPase activity of SERCA, including its affinity for Ca2+, is highly dependent on the membrane thickness and fluidity of the surrounding bilayer 96,97. SERCA requires a mismatch between hydrophobic thickness of the bilayer and its membrane embedded part for optimal flexibility 98. Norimatsu et al. (2017) performed an extensive study on the first layer of phospholipids of SERCA crystal structures in four different conformational states 99. They proposed that the lipid environment plays a more active role for SERCA function and is directly involved in the dynamics of its pump function. Local distortions of the lipid bilayer occur through TM helices movement, which can be used as an energy source contributing to conformational changes. Additionally, modulation of catalytic activity mediated through the lipid environment has been shown for several other members of the P-type ATPase family, including Cu(I)-transporter CopA and heavy metal transporter ZntA 100–102. Lipids have also been linked to protein stabilization. The high prevalence of membrane protein crystal structures with bound lipids has revealed a stabilizing effect early on 103. Today, native mass spectrometry (MS) has become a useful tool to analyse specific lipid-protein interactions and several cases of lipid-mediated stabilization of protein oligomers have been revealed 104,105. For the Na+-K+-ATPase, a specific lipid site was identified that only affected protein stabilization, but showed no effect on protein activity 106. A second lipid binding site only affecting protein activity was also characterized, but both effects are independently modulated by different lipid classes. Further, membrane protein topology has been shown to not only be determined by the amino acid sequence, but it can also be influenced by the lipid composition of the membrane. Most membrane proteins follow the positive-inside rule, in which they orient their positively charged amino acid residues facing the cytoplasm 107. Negatively charged phospholipids have been shown to inhibit translocation of positively charged domains, contributing to membrane protein topology 108. The topology of several E. coli transporter, containing multiple TM helices, was severely affected depending on the lipid composition of the membrane. The presence of PE has been shown to be required for the correct orientation of the transporters lactose permease LacY and γ-aminobutyric acid permease GabP 109,110. Additionally, membrane protein assembly was affected upon changes of membrane fluidity 111.

Lastly, lipids have also been implicated in protein localization. Many proteins have been identified at the bacterial cell poles, where they co-localize with CL 112. Therefore, it has been hypothesised that CL is involved in the polar recruitment or positioning of certain membrane proteins, including osmosensory protein ProP 113. However, as the mechanism and the cellular machinery for polar localization of lipids and proteins is not completely understood yet, showing direct effects of lipids on protein localization remains an obstacle.

Figure 5: Lipids can affect membrane protein function through different types of interaction Side view (a) or top view (b) of a membrane protein engulfed by the membrane bilayer, exhibiting different types of lipid interactions, which vary in their exchange rate. Adapted from Contreras et al. (2013) and Stangl and Schneider (2015) 93,114.

1.4 Magnesium-transport in E. coli

Magnesium is the most abundant divalent cation in cells and contributes to many cellular processes 115. These range from stabilization of macromolecules, including ribosomes and membranes, to neutralization of nucleic acids and nucleotides and a function as a co-factor in many enzymatic processes 115. Recently, Mg2+ homeostasis has also been shown to be important for bacterial survival and defense against the host innate immune response. Magnesium deprivation has been revealed to be the main resistance mechanism of host resistance factor

SLC11A1 against Salmonella infection in macrophages 116. However, the chemical properties of Mg2+ differ greatly from other cations. As an example, the ionic radius of Mg2+ is among the smallest of all cations (0.65 Å) while its hydrated radius is more than 400 times larger 117. In comparison, the hydrated radius of Ca2+ and K+ is only 25 or four times larger than their dehydrated from, respectively 117. This highlights the complex requirements of Mg2+ transporters in terms of recognition of the largely hydrated ion, removal of the hydration shell and transport of the bare ion 117. Bacteria contain three distinct transporter classes for magnesium uptake: CorA, magnesium transporter E (MgtE) and bacterial magnesium transporter A and B (MgtA/MgtB) 115. Many bacterial species express multiple Mg2+ transporters of the same or different classes 115. The constitutively expressed transporters CorA or MgtE exhibit a large phylogenetic distribution and are considered the primary Mg2+ transporters 115. Both use the electrochemical gradient across the cytoplasmic membrane as their energy source for Mg2+ transport, essentially mediating influx and efflux of Mg2+ 115. Therefore, they are considered channels rather than transporters 115. On the contrary, MgtA and its homologue MgtB only mediate Mg2+ influx 118. Initial research revealed that MgtA/MgtB mediate Mg2+ uptake only if cells are present in an environment containing low Mg2+ concentrations and expression of these transporters has been shown to be dependent on extracellular Mg2+ levels 119. In recent years a tight and complicated regulatory network of MgtA and MgtB expression and activation has been identified, which will be discussed in detail below (Section 1.4.1.3). Most of our understanding of bacterial Mg2+ homeostasis derives from the Gram-negative bacterium Salmonella 115, containing Mg2+ transporter CorA, MgtA and MgtB 119. However, large similarities between E. coli, which contains Mg2+ transporter CorA and MgtA, and Salmonella Mg2+ homeostasis have been described 117,120.

1.4.1 Magnesium transporter A (MgtA)

E. coli MgtA (ecMgtA) consists of 898 amino acids and has a molecular weight of 99.5 kDa 121. It is located in the bacterial inner membrane 122 and as a P-type ATPase its C- and N-terminus points into the cytoplasm. MgtA belongs to the PIIIB subfamily of P-type ATPases, closely related

84 to the yeast and plant proton transporters of the PIIIA subfamily . Further, MgtA is phylogenetically closer to eukaryotic P-type ATPases, for example SERCA, than prokaryotic ATPases 123. As a P- type ATPase, MgtA utilizes ATP hydrolysis for the Mg2+ transport from the periplasm to the cytoplasm. Interestingly, no direct evidence for Mg2+ transport by MgtA has been obtained yet.

This has led to the hypothesis that Mg2+ is not the primary transported ion, but rather the unknown counterion. Possible counterions being transported by MgtA are protons 124. No high-resolution structure of MgtA has been obtained so far, however structural prediction proposes the presence of classical P-type ATPase features. Ten transmembrane helices were predicted as well as a N-domain, a P-domain and an A-domain in the cytoplasm. MgtA is present as a monomer and no indication of oligomerization has been found 124. Although a high-resolution structure of the entire MgtA transporter is still missing, a crystal structure of the N-domain of MgtA has been resolved at 1.6 Å 125. Although it revealed similarity in comparison to the N-domain of SERCA, the N-domain of MgtA exhibited a more compact structure. Further, a unique motif to Mg2+ ATPases has been revealed in the ATP binding site and the crystal structure indicated that MgtA has a tighter binding site and higher selectivity for the adenine nucleotide base in comparison to SERCA 125. MgtA shares ~50 % sequence identity with its homologue, MgtB 115. While the mgtA locus consists of a single gene, mgtB is encoded on the mgtCBR operon 126. MgtC, which is encoded on the same operon as MgtB, is a 22 kDa virulence factor found in several intracellular pathogens and plays a role in intramacrophage survival 127. mgtCBR is located on the pathogenicity island SPI-3, while mgtA is located at a different chromosomal location 128.

2+ Purified ecMgtA showed substrate inhibition by Mg free, exhibiting maximum activity at a

2+ 124 concentration of 250 μM Mg free and strong inhibition at concentrations above 1 mM , which

2+ 129 2+ corresponds to physiological intracellular Mg free concentrations in E. coli . Mg free refers to Mg2+ that is not coordinated by ATP but is freely available for transport. Additionally, enzymatic studies revealed a pH dependency of MgtA activity as maximum activation occurred in a pH range between 7.0-7.2 124, corresponding to the native cytoplasmic pH of E. coli 130. At higher or lower

124 2+ pH MgtA ATPase activity was reduced . Additionally, at a higher pH the Mg free sensitivity of

2+ 124 MgtA was influenced and Mg free inhibition was shifted to higher concentrations . However, the mechanism of pH-dependency of MgtA activity remains unclear. Further, the ATPase activity of MgtA was inhibited by divalent cations, with Zn2+ showing the strongest inhibition, and Co2+, Ni2+, Ca2+ and Mn2+ exhibiting weaker inhibition 124.

1.4.1.1 The role of cardiolipin for MgtA-mediated Mg2+ transport

Subramani et al. (2016) showed that MgtA-mediated Mg2+ transport is highly dependent on anionic lipids 124. The presence of PG and, especially, CL induced strong MgtA ATPase

activation 124. Kinetic studies on MgtA ATPase activity in the presence of CL extracted from E. coli or bovine heart membranes proposed that CL-dependent MgtA activation is mainly mediated through the negatively charged lipid head group and the acyl chain composition of CL only plays a minor role 124. This was proposed as MgtA ATPase activity was detected in both lipid samples. However, in the presence of bovine heart CL the ATPase activity only reached 80 % in comparison to MgtA activity in the presence of E. coli CL. 124 It should be noted that bovine heart CL contains just two different CL species 69, while E. coli CL contains a large variety with over 50 different CL species 65. However, the function of CL for MgtA-mediated Mg2+ transport remains elusive. Additionally, co-localization of MgtA with CL at E. coli cell poles has been revealed 124. The N-terminus of MgtA, termed KEIF, is rich in positively charged residues and a DISOPRED3- based prediction proposed an intrinsically disordered region 124. Molecular Dynamics (MD) simulations and small angle X-ray scattering (SAXS) analysis confirmed that KEIF is a disordered peptide in aqueous solution 121. A role of the N-terminus in membrane trafficking was proposed 124. Yet, an N-terminal deletion mutant (NΔ31-MgtA) was unaffected in its polar localization 124. Interestingly, a recent study by Jephthah et al. (2020) revealed novel insight into the biophysical behavior of KEIF 121. In the presence of anionic lipid vesicles conformational changes of KEIF, including an increase in helical content, were observed 121. Further, KEIF adsorbed to the surface of anionic vesicles 121. Therefore, a role of KEIF as a bilayer anchor for MgtA was proposed.

1.4.1.2 Phylogenetic distribution of MgtA

MgtA is found in many enteric bacterial species, including E. coli, Salmonella, Shigella flexneri and Yersinia enterocolitica 128. Meanwhile MgtB and MgtC show a more narrow and sporadic distribution 128 and are found in Salmonella, but not E. coli. As mgtCBR is located on the pathogenicity island SPI-3 and mgtA is located at a different chromosomal location, the mgtCBR operon was likely acquired by horizontal gene transfer by selected bacterial species 128. Although MgtA homologues of Streptococcus pneumonia 131 and Shigella flexneri 132 have been selectively studied, most of our knowledge on MgtA comes from E. coli and Salmonella homologues 115,117,123,124,133. In recent years, a wider phylogenetic distribution of MgtA among bacteria has been identified 124. Additionally, MgtA homologues have been detected in archaea, fungi and plants 124. PH1, a MgtA homologue identified in Petunia, is the only characterized eukaryotic MgtA

homologue so far 134. PH1 has been implicated in proton transport in the plant vacuole affecting changes of the petal color 134. Although further research must be performed on eukaryotic MgtA homologues, this study proposes novel functions for eukaryotic MgtA representatives.

1.4.1.3 Transcriptional and translational regulation of mgtA expression mgtA expression is under an extensive regulatory network controlled by extracellular and

2+ intracellular Mg free concentrations among other factors. Figure 6 gives an overview of the tight regulation of MgtA expression and activity The two-component system PhoP/PhoQ represents the first regulatory step in mgtA

2+ 135 expression, upregulating mgtA transcription under low periplasmic Mg free levels (Figure 6a) . The dimeric PhoQ resides in the cytoplasmic membrane and acts as a sensor for external signals,

2+ 115 2+ for example low Mg free levels . In the presence of low periplasmic concentrations of Mg free PhoQ promotes PhoP phosphorylation (PhoP-P) 115. PhoP is the transcription regulator protein, which controls the expression of genes involved in Mg2+ homeostasis 115. PhoP phosphorylation increases its binding affinity to the promoters and activates expression of PhoP-activated target genes, including mgtA and mgtB 135. A model of PhoQ-dependent Mg2+ sensing was proposed 115, mediated by a patch of acidic residues in the sensing domain of PhoQ 136. These acidic residues

2+ 115 form metal bridges with Mg free and the surrounding phospholipid head groups , which stabilizes PhoQ in a rigid conformation and promotes the unphosphorylated form of PhoP. Removal of Mg2+ disrupts these metal bridges, which results in destabilization of PhoQ, and

2+ promotes the phosphorylated form of PhoP. Besides low Mg free levels, the PhoP/PhoQ system is also activated, to a lower extent, under mildly acidic pH 137,138 and in the presence of antimicrobial peptides 139 (Figure 6a). These are common conditions experienced for Salmonella in a macrophage environment. Different regulation mechanisms of PhoQ in the presence of acidic pH and antimicrobial peptides have been identified. While antimicrobial peptides compete with

2+ 139 Mg free for binding to the same highly acidic region of PhoQ , acidic pH leads to conformational changes of the PhoQ sensor domain 140. Numerous other genes are under the regulation of PhoP/PhoQ and the two-component system is required for virulence and intracellular survival of Salmonella in macrophages 141.

2+ MgtA expression is additionally controlled in response to cytoplasmic Mg free concentrations (Figure 6b). Although different regulation mechanisms of mgtA expression by low

2+ cytoplasmic Mg free have been identified in Salmonella and E. coli, the entire regulatory network

is still under investigation. Firstly, the 5’-untranslated region of the mgtA mRNA functions as a metal-sensing riboswitch 142. This leader region can adopt two mutually exclusive stem-loop

2+ 142 conformations depending on Mg free concentrations (Figure 6b). The first is favored by high

2+ Mg free and allows interaction with termination factor Rho, which functions as an effector and

143 2+ terminates transcription . Alternatively, in the presence of low Mg free the mgtA leader region will adopt a different conformation, interfering with Rho-interaction and permitting transcription into the mgtA coding region 143. Additionally, mgtA mRNA contains a proline-rich leader region, termed mgtL, that allows regulation of mgtA expression in response to cytosolic proline levels and high osmolarity 144 (Figure 6b). It was shown that low proline levels act synergistically with low

2+ 144 Mg free levels to increase mgtA mRNA levels . Gall et al. (2016) proposed that high intracellular

2+ 2+ concentrations of Mg free positively regulate translation efficiency of mgtL, as high Mg free concentrations might reduce translation interference presented by proline codons 145. The

2+ increased translation efficiency of mgtL in the presence of high Mg free affects the conformation of the 5’-untranslated region of mgtA and induces Rho-dependent transcription termination of mgtA 145. However, this model is under debate as Zhao et al. (2010) proposed an alternative

2+ 146 Mg free -dependent/proline-independent regulation model of mgtA transcription . In this model

2+ stem-loop switching of the 5’-untranslated region of mgtA is mediated only by Mg free levels,

2+ which allows mgtL translation under high Mg free concentrations and results in premature mgtA transcription termination 146. Nagao et al. (2017) further showed that mgtL contains a translation-

2+ aborting sequence that can lead to intrinsic ribosome destabilization (IRD) dependent on Mg free concentrations 147. IRD occurs when the ribosomal subunit dissociates as certain amino acid sequences in the nascent chain, containing proline-intermitted acidic residues, destabilize the

2+ ribosomal complex. However, high intracellular Mg free levels can counteract this process through

2+ 2+ the ribosomal stabilization effect of Mg . Under low Mg free levels the destabilized ribosomes abort mgtL translation and allow mgtA transcription 147. Additionally, the mgtA coding region itself contains two consecutive proline residues (Pro550, Pro551), which induce ribosome stalling during translation 148 (Figure 6b). Elongation factor P (EF-P) is required for the continuation of translation, thereby mgtA translation is dependent on the presence of EF-P. Mutational studies also revealed that these two consecutive proline residues do not play a role in MgtA-mediated Mg2+ transport 148, supporting their role as an additional step in the complex regulatory network of

2+ mgtA expression. Additionally, MgtA regulation is also mediated through Mg free substrate inhibition of MgtA ATPase activity as described in 1.4.1 (Figure 6c). In recent years, a large variety of small peptides with less than 50 amino acids in length have been identified in bacteria 149. Although their cellular functions require further investigation,

two peptides have been linked to the regulation of MgtA stability (Figure 6d). In Salmonella MgtR, which is encoded on the same operon as MgtB and MgtC, is located in the cytoplasmic membrane 150. MgtR has been shown to negatively regulate MgtA expression 150. It was proposed that MgtR directly interacts with MgtA in the inner membrane, although the exact regulatory mechanism remains unclear. MgtR has also been observed to directly bind to MgtC in vivo and negatively regulate MgtC expression 151. It was proposed that MgtR mediates MgtC unfolding and afterwards recruits FtsH protease for MgtC degradation 151. A similar mechanism for MgtR- mediated MgtA regulation was proposed (Figure 6d) 150. Recently, another small peptide, MgtS, has been linked to MgtA regulation in E. coli 152. MgtS is also localized to the cytoplasmic membrane. It has been shown to interact with MgtA in vivo and positively stabilize the Mg2+ transporter, likely through inhibition of FtsH-mediated degradation (Figure 6d). Thereby, MgtS promotes increasing intracellular Mg2+ concentrations 152. Coherent with this, mgtS expression is

2+ 152 also induced by low Mg free in a PhoP/PhoQ-dependent manner . Although the regulation of MgtA expression has been extensively studied, the necessity for this complex and tight regulatory network remains unclear. Interestingly, unregulated MgtA expression promoted Salmonella survival in macrophages and rendered it hypervirulent in mice 148, highlighting the consequences of interference with the regulation of MgtA expression. However, it remains unclear how Salmonella hypervirulence is mediated through increased MgtA expression.

2+ Figure 6: MgtA expression and activity is under a complex regulatory network dependent on Mg free concentrations MgtA expression is under a complex regulatory network regulated by external factors, most notably (a)

2+ 2+ extracellular and (b) intracellular Mg free levels. Additional control is mediated through (c) Mg free- dependent activation and inhibition of MgtA ATPase activity and (d) interaction with small membrane- embedded peptides, MgtS and MgtR.

1.4.1.4 Cellular functions of MgtA

The cellular function of MgtA has long been a subject for discussion as its requirement of ATP for intracellular Mg2+ transport has been unclear. The electrochemical gradient across the bacterial membrane is highly negative and should provide enough potential energy to drive Mg2+ import

2+ 117 without the need of ATP, even at micromolar extracellular Mg free concentrations . An explanation for this phenomenon was proposed as the PhoP/PhoQ regulatory system also promotes a reversal in membrane potential 115. This means that MgtA is likely expressed and operates under conditions, in which both, the electrical and the chemical gradient, is unfavorable for Mg2+ transport, providing a possible reason for the requirement of ATP hydrolysis for MgtA- mediated Mg2+ transport.

A role for MgtA as a “scavenger” transporter was hypothesized 117. Cells use scavenger transporters as secondary ion transport systems if extremely low substrate concentrations are present in the environment. However, MgtA exhibits the same Mg2+ affinity as the constitutively expressed CorA transporter 117. It should be noted though that the Mg2+ transporter CorA and MgtE are dependent on the membrane potential as their driving force. Therefore, their Mg2+ transport is likely influenced by changes in external pH, activity of the electron transport chain components and fluctuations in the concentrations of other ions 115. MgtA-mediated Mg2+ transport, on the other hand, is less dependent on environmental influences. In recent years novel cellular functions of MgtA, and its homologue MgtB, have been proposed. These are strongly linked to Mg2+ homeostasis, the cellular energy metabolism and survival during the bacterial infection process. In Salmonella, MgtA/MgtB as well as the virulence factor MgtC has been shown to be necessary for the assembly of functional ribosomes, regulating

2+ 153 2+ protein synthesis under low Mg free levels . Upon Mg deprivation ribosome synthesis is

2+ repressed to maintain cytosolic Mg free concentrations. This is required as the cations can be recycled for other processes, including association of ribosomal subunits 30S and 50S. Therefore, the cell induces expression of mgtA and mgtB to import Mg2+ to the cytosol in response to low

2+ 2+ 153 Mg free levels, and it has been revealed that MgtA functions as the main Mg transporter . The

2+ 153 inhibition of F1F0-ATP synthase by MgtC frees up additional Mg required for translation . A

2+ similar mechanism for maintaining protein synthesis under low cytosolic Mg free concentrations has been proposed in E. coli utilizing MgtA and additional unknown genes 153. In Salmonella MgtA and MgtB also have been shown to contribute to the accumulation of alternative sigma factor RpoS, which is involved in the cellular stress response 154. Further, an increased expression of the Mg2+ transporters strengthened the thermotolerance of Salmonella 155 and MgtA promoted survival of Salmonella in the presence of reactive nitrogen species, conditions found in the macrophage milieu 156. Additionally, magnesium deprivation has been shown to be the main resistance mechanism of host resistance factor SLC11A1 against Salmonella infection in macrophages. Inactivation of MgtA and its homologue MgtB led to severe fitness defects in a mouse model 116.

2. Aims of this thesis

Pathogenic E. coli remain a global threat, responsible for millions of acute illnesses each year 3. The intestinal pathogens, EHEC and EPEC, are major causes for foodborne gastrointestinal illnesses, leading to outbreaks of bloody diarrhea and HUS 5,8,12. The increasing prevalence of antibiotic resistance among pathogenic E. coli 157,158 demands a better understanding of the bacterial infection process as well as the human immunological defense mechanism in order to develop novel and efficient treatments.

Part I: A critical step in the bacterial infection process is the adhesion of bacteria to the host cell. The virulence factor intimin plays an essential role in EHEC and EPEC adhesion as it mediates intimate adherence of the bacterial cells to the host 5. Intimin localizes in the outer bacterial membrane, where it interacts with the bacterial Tir receptor presented on the host cell membrane. This interaction is mediated through the extracellular passenger of intimin, consisting of four Big domains (D00-D0-D1-D2) topped by a C-type lectin-like domain (D3) 36. Previous studies have structurally characterized intimin sections, including the periplasmic LysM domain 40, the membrane-bound β-barrel 41 and the C-terminal part of the extracellular passenger, consisting of subdomains D1, D2 and D3 24, which are directly involved in Tir binding. However, structural data is missing for the two remaining extracellular subdomains D00 and D0, which connect the β-barrel to the C-terminal passenger. Especially the structure of D00 has been under discussion and D00 folding has been proposed to play a critical role for the extraction of the C-terminal subdomains of the passenger 44. Therefore, the primary aims of this study were the following:

1. Structural and biochemical analysis of the elusive intimin passenger subdomains D00 and D0. 2. Investigation of the biological role of the passenger subdomain D00 in intimin-Tir interaction as the juncture domain between the membrane-bound β-barrel and the extracellular passenger.

Part II: Upon internalization of bacteria, the human immune response counteracts the bacterial infection and among the first line of defense are macrophages, which engulf and digest bacterial intruders. Recently, magnesium deprivation was revealed as a macrophage defense mechanism by the macrophage-located host resistance factor 116. Bacterial magnesium transporter MgtA has been shown to play a role in counteracting magnesium deprivation and protecting bacteria in the macrophage environment 116. Additionally, MgtA has been linked to the survival of Salmonella in the presence of reactive nitrogen species, conditions typically present in macrophages 156. These novel studies highlight MgtA as a potential drug target for the treatment of pathogenic E. coli and other pathogens. E. coli MgtA has been shown to be highly dependent on cardiolipin for its activity 124, yet the role of this lipid for MgtA-mediated Mg2+ transport needs to be elucidated. Additionally, a tight regulation of MgtA activity by Mg2+ levels has been discovered 124, yet no mechanism of the Mg2+- dependent substrate inhibition has been revealed. Therefore, the primary aims of this study were the following: 1. Biochemical analysis of MgtA-cardiolipin interaction involved in MgtA activity, stability and in vivo localization with a focus on cardiolipin specificity. 2. Investigation of specific functions of cardiolipin for MgtA-mediated Mg2+ transport.

2+ 3. Biochemical characterization of Mg free sensitivity of MgtA through mutational studies.

2+ 4. Characterization of lipid dependency and Mg free sensitivity of MgtA homologues from Salmonella.

3. Synopses of publications

Paper I: The extracellular juncture domains in intimin adopt a constitutively extended conformation and induce restraints in the intimin reach and sphere of action

An essential virulence factor involved in bacterial adhesion to the host cell, for example of EHEC and EPEC, is intimin. Localized on the outer bacterial membrane intimin binds to the Tir receptor presented on the host cell. Intimin belongs to the class of inverse autotransporters, exhibiting a classical composition with a small periplasmic domain, a membrane-bound β-barrel and an extracellular passenger. The passenger has been proposed to form a rod-like extension consisting of four bacterial immunoglobulin-like (Big) domains (subdomains D00-D0-D1-D2) capped by a C-terminal C-type lectin-like domain (D3). The structure of the passenger has long been debated as only the C-terminus of the passenger (D1-D2-D3) bound to Tir has been confirmed by a crystal structure. Although, bioinformatic and biophysical evidence suggested that subdomains D00 and D0 are also Big domains, no high-resolution structural information has been obtained to date. Our findings present for the first time the structures of the two elusive extracellular subdomains D00 and D0 determined by X-ray crystallography. These structures represent the last missing structural information of the entire EPEC intimin passenger. Both, D00 and D0, reveal a Big fold. While D0 shares high structural similarity to subdomain D1, D00 exhibits a more general Ig fold and topological resemblance to the extracellular domains of mammalian cadherins. Further, we highlight D00 as a common structural element found in many intimins homologues and a wide variety of other inverse autotransporters, selectively located at the interface of the bacterial membrane and the extracellular passenger. Structural and in silico data, obtained by SAXS and molecular dynamics simulations, show that D00-D0 exhibits an elongated and rigid conformation in solution, while D0-D1 reveals higher flexibility. Rigidification of D00-D0 is mediated through the connector region between both domains. Yet, the connector does not affect mechanical stability, which was assessed by atomic force microscopy. Monte Carlo simulations indicated that a stiff base of the passenger, represented by D00-D0, increases its radius of reach, likely allowing a more probable binding event of intimin and Tir. In summary, we revealed the entire structure of the intimin passenger and highlight structural and biophysical differences between subdomains, representing functional subunits along the passenger that could play a role in intimin-Tir interaction.

Paper II: The bacterial magnesium transporter MgtA reveals highly selective interaction with specific cardiolipin species

Understanding complex protein-lipid interactions has become the focus of membrane protein research to comprehend protein function and regulation in vivo. Bacterial magnesium transporter MgtA has a wide phylogenetic distribution among Gram-negative bacteria, including E. coli. As a P-type ATPase, MgtA transports Mg2+ from the periplasm to the cytoplasm using ATP hydrolysis as its energy source. Previously, the importance of cardiolipin for MgtA-mediated Mg2+ transport has been revealed. However, the function of the lipid for MgtA activity remains unclear. Here, we show that MgtA exhibits high selectivity to two specific cardiolipin species, cardiolipin 18:1 and cardiolipin 16:0. Although cardiolipin 16:0 by itself did not activate MgtA, at least 50 % cardiolipin 16:0 was required in combination with cardiolipin 18:1 for maximal MgtA ATPase activity, highlighting a dependence on the presence of both species to obtain maximal activity. MgtA activity was assessed by a colorimetric ATPase activity assay measuring release of inorganic phosphate from ATP hydrolysis. Further, activity profiles with increasing concentrations of cardiolipin revealed cooperative binding of more than one cardiolipin molecule, and native mass spectrometry confirmed two specific cardiolipin binding sites on MgtA. Nano differential scanning fluorimetry showed thermal stabilization of MgtA in the presence of E. coli cardiolipin. While cardiolipin 18:1 significantly increased the melting temperature of MgtA, the presence of cardiolipin 16:0 did not affect MgtA thermal stabilization. Therefore, we proposed a model of cardiolipin interaction with MgtA, in which two cardiolipin 18:1 bind cooperatively to MgtA as non-annular lipids while cardiolipin 16:0 plays a role as an annular lipid for MgtA activity. Additionally, the role of cardiolipin as a signal for MgtA localization to the E. coli cell poles was investigated, as polar co-localization of MgtA and cardiolipin was previously observed 124. However, no indication of cardiolipin as a localization signal for MgtA was detected, although it cannot be excluded that cardiolipin contributes to MgtA localization in a minor role. In summary, high selectivity of E. coli MgtA to specific cardiolipin species through site- specific interactions as well as interactions with the lipid bilayer environment was revealed. These results can contribute to our understanding of the role of cardiolipin for MgtA-mediated Mg2+ transport and the general knowledge of membrane protein-lipid interactions.

Paper III: The Mg2+ sensing region of MgtA resides in the C-terminus and is dependent on pH

Magnesium transporter MgtA is widely distributed among Gram-negative as well as Gram-positive bacteria. A tight regulation of MgtA expression dependent on low extracellular and intracellular

2+ Mg free concentrations has been revealed in Salmonella and E. coli. Additionally, ATPase activity

2+ of E. coli MgtA (ecMgtA) is also highly regulated by Mg free levels, with maximum activity at ~250

2+ 2+ μM Mg free and inhibition above 1 mM Mg free. To verify whether Salmonella MgtA (stMgtA) and its homologue MgtB (stMgtB) reveal similar Mg2+-dependent regulation of their function as ecMgtA, kinetic studies of MgtA/MgtB were performed using a colorimetric ATPase assay measuring ATP hydrolysis. Although stMgtA

2+ 2+ revealed lower affinity to Mg free than stMgtB or ecMgtA, stMgtA exhibited Mg -dependent inhibition, as observed for its E. coli counterpart. stMgtB retained its activity partially at higher

2+ Mg free concentrations. Additionally, stMgtA exhibited anionic lipid dependent activation and heterologously expressed stMgtA localized at E. coli cell poles and division septa, similar as detected for ecMgtA. In contrast, lipids did not play a crucial role for the ATPase activity of stMgtB and stMgtB localized over the entire E. coli membranes. As the lack of structural data hinders the understanding of MgtA-mediated Mg2+ transport, mutational studies were performed to investigate the role of the MgtA C-terminus in Mg2+ sensing. A tryptophan residue located at the penultimate position of the C-terminus is strongly conserved among MgtA homologues. Deletion (ΔWQ) or alanine substitution (W897A) of the tryptophan led to severely decreased ATPase activity, revealing an important role of the C-terminus in MgtA

2+ function. Further, the ΔWQ and W897A mutation impacted the ability of MgtA to sense Mg free concentration and analysis using an uncompetitive substrate inhibition model indicated that the

2+ mutants lost the ability to bind a specific number of Mg free ions at the inhibitor site. However, the loss of Mg2+ sensitivity of ΔWQ and W897A mutants could be rescued in a mildly acidic pH in the range between 6.2-6.5. In summary, similarities as the anionic lipid dependency and Mg2+-dependent inhibition were revealed between MgtA homologues from E. coli and Salmonella. Salmonella MgtB did not exhibit these properties, proposing different cellular functions of both Mg2+ transporters. Further, the role of the MgtA C-terminus for Mg2+ sensitivity was shown, contributing to our understanding of the complex regulatory mechanism of MgtA-mediated Mg2+ transport.

4. Discussion

Intestinal pathogenic E. coli, such as EHEC and EPEC, present a global threat and contribute to morbidity and mortality worldwide 3,5,12. With a rise in bacterial antibiotic resistance 157,158, the development of new therapies and treatments is needed. Understanding the bacterial infection process and the components involved allows for the identification and characterization of novel potential drug targets. In this thesis two proteins, the virulence factor intimin required for bacterial adhesion and the magnesium transporter MgtA involved in bacterial survival in a macrophage milieu, have been characterized as potential future drug targets. In the first part of this thesis, elusive domains of the adhesion receptor intimin were structurally and biophysically analyzed. The obtained results are discussed in Sections 4.1 through 4.3. The second part of this thesis focused on magnesium transporter MgtA and its interaction with the lipid environment as well as regulatory mechanisms via Mg2+ sensing. Results are discussed in Sections 4.4 to 4.12.

4.1 Alternating regions of rigidity and flexibility along the intimin passenger increases its sphere of reach and potentially promotes Tir receptor binding (Paper I)

The virulence factor intimin is essential for the intimate attachment of EHEC and EPEC to the host cell 5. Intimin binds to the Tir receptor presented on the host cell membrane via its extracellular passenger. Although the Tir interaction at the C-terminus of the intimin passenger (subdomains D1-D2-D3) has been studied 24, the structure of the entire passenger remains elusive. To understand the entire composition of the passenger, we solved the crystal structure of the missing N-terminal extracellular subdomains D00-D0 and D0-D1. Large differences in rigidity were detected between both with D00-D0 exhibiting a constitutively extended conformation, while D0-D1 displayed a bent structure with increased flexibility (Paper I – Figure 1 and 3). We hypothesized that the rigidity of D00-D0 is mediated through a large hydrogen bond network established through the connector between D00 and D0. A model of the intimin passenger consisting of subunits varying in flexibility was proposed. In this model the passenger consists of a rigid base composed of D00-D0, followed by a flexible segment mediated by D0-D1 and D1-D2 connectors and topped by a rigid module (D2-D3) (Paper I – Figure 5). The passengers of other characterized inverse autotransporters, for example Yersinia

pseudotuberculosis invasin 47 and E. coli FdeC 159, showed extended conformations in their crystal structures with no indication of flexible hinges between subdomains. The long repetitive chains of Big subdomains in passengers were proposed to play a role in spanning the bacterial surface structures in order to present the receptor subdomain for host cell interaction 45. Therefore, an extended structure of the passenger would be beneficial. Although the crystal structure of intimin C-terminal subdomains D1-D2-D3 also exhibited an elongated rod-shaped structure 24, the NMR structure showed a curved fold with a bend between subdomains D1 and D2 160. The bent structure resembles the kinked conformation observed for D0-D1. Therefore, structural data under solution conditions and biophysical data on passengers should be obtained to verify whether variations of flexibility and rigidity are a common feature in passengers as crystal structures in most cases only show selected conformations. The biological function of these variations in flexibility along the intimin passenger remains unclear. The attachment of bacteria to the human host cells occurs under fluid flow conditions, putting additional shear stress on the extracellular adhesins. We hypothesized that a rigid base promotes an increased radius of reach and sphere of action. The constitutively elongated base of the passenger, represented by D00-D0, forms a rigid unit that projects the passenger further from the outer bacterial membrane, likely increasing the probable interaction between Tir and intimin (Paper I – Supplementary Figure 7). Flexibility along D0-D1 and D1-D2, on the other hand, allows the Tir binding region of intimin to interact with Tir at a low angle, as seen in the crystal structure 24. MC simulations supported this hypothesis (Paper I – Figure 5). However, additional cellular studies must be performed to verify a biological role of the extended conformation of D00- D0 for the intimin-Tir interaction. It should be noted though that D00 is not required for intimin secretion and passenger surface display, and deletion of D00 did not affect adherence to mammalian HeLa cells 44.

4.2 The constitutively extended D00-D0 conformation might play a role in the secretion process of the intimin passenger (Paper I)

A second hypothesis on the functional role of variations in the flexibility along the intimin passenger is that the rigid base of D00-D0 contributes to the secretion process of the extracellular passenger. Intimin belongs to the class of IATs, which secret their extracellular passenger through their intrinsic N-terminal β-barrel 36. Leo et al. (2016) proposed that the secretion is driven by sequential folding of the Big domains, leading to passenger export from N- to C-terminal direction 44. A similar mechanism was proposed for classical ATs that would facilitate passenger

secretion independent of ATP hydrolysis. Sequential domain folding would function thereby as a Brownian ratchet mechanism and prevent backsliding of the polypeptide chain into the periplasm 161. Therefore, the extended conformation of D00-D0 could contribute to the secretion process as its extension could support pulling out of the subsequent Big domain, D1, which would promote a continuous secretion of the passenger. Further, it could play a role in preventing the passenger from sliding back through the β-barrel pore. As the juncture subdomain D00 is the first secreted domain, one might hypothesize that its secretion is the most critical step of the autotransport process. For classical ATs, a highly stable folding core, termed “autochaperone” domain, has been identified at the membrane- proximal C-terminus 162–164. This “autochaperone” domain is needed for an efficient secretion. The corresponding domain for the IAT intimin is subdomain D00. Therefore, a similar role as a folding core was previously proposed for D00 44. However, Leo et al. (2016) showed that no folding core is present in the intimin passenger 44. D00 is further not required for secretion but can be substituted by D0. Interestingly though, D00 exhibited reduced mechanical stability in comparison to D0 and D1 (Paper I – Figure 2) and Leo et al. (2016) proposed that the marginal instability of D00 could be beneficial for the initiation of the autotransport process 44. As a final note, the secretion process of IATs is not completely understood yet. A role of the BAM machinery in passenger secretion has been proposed, next to its role in the insertion of the β-barrel into the bacterial outer membrane. A subset of classical ATs has shown dependency on a BAM homologue, TamA, for efficient secretion of their passenger 165. Further, Leo et al. (2016) presented that an intimin mutant, stalled in its passenger secretion, remained linked to the BAM complex, even after complete insertion and folding of the β-barrel 44. Therefore, the secretion process of IATs needs further investigation to analyse a potential role of the rigid base D00-D0 for this process.

4.3 D00 is a common domain in many inverse autotransporters (Paper I)

The composition of the intimin passenger regarding quantity and structure of subdomains has long been under discussion. Only in 2012, the fifth subdomain, D00, was described by Fairman et al. 41. An like fold was predicted as its structure 44, yet this has been under debate as common structural prediction tools could not predict a structural motif. Here, we confirmed a modified Big domain fold for D00 (Paper I – Figure 1). D00 showed less structural similarities to intimin subdomains D0 and D1 but was structurally closer to the extracellular domains of mammalian cadherins. Cadherins mediate cell adhesion and classical cadherins typically consist of five

tandemly repeated extracellular domains, a single membrane-spanning segment and a cytoplasmic region, therefore showing structural resemblance to intimin (Paper I – Figure 6) 166. However, cadherins are dependent on Ca2+ for their function as Ca2+ binding in the connector regions of the extracellular domains leads to rigidification and dimerization 167. For the intimin passenger, no Ca2+-dependency for adhesion has been described and we propose a constitutively rigid conformation of D00-D0. The structural prediction of many IATs reveals a classical composition with a membrane- bound β-barrel and an extracellular passenger consisting of Big domains in varying numbers and combinations 45,46. Yet, for most IATs only poor structural domain prediction is possible in the ~100 amino acids following the β-barrel. This region corresponds to intimin subdomain D00. To verify if other IATs also contain a D00-like domain, we performed an alignment on selected sequences located at the juncture between β-barrel and passenger. We revealed conservation of several residues located in regions, which are not involved in the β-sandwich fold. This also includes a proline residue (P515) disrupting the β-sheet locally and introducing a small helix (Paper I – Figure 4). Due to the conservation among sequences, we hypothesized that the modified Ig fold exhibited by D00 is common among IATs, selectively located in the first extracellular subdomain following the β-barrel. Interestingly, passengers normally consist of chains of repetitive, highly similar domains. The necessity for a specific juncture domain between the membrane-embedded β-barrel and the remaining extracellular passenger remains unclear. Further, deletion of D00 did not affect intimin passenger surface display 44. We also detected conservation of mentioned residues in IATs with a minimal passenger domain, essentially only containing 100-200 amino acids following the β- barrel (Paper I – Figure 4). A conservation of a D00-like structure in these autotransporters raises the question whether D00 does not contribute to the structural stabilization of long extracellular passenger but fulfils another unknown function. However, the wide distribution of a D00-like domain at the specific position behind the β-barrel invites to speculate that D00 executes a specific role in IATs and further research could elucidate the biological function of D00.

4.4 MgtA shows high selectivity towards specific cardiolipin species (Paper II)

The variety of lipid-protein interactions, mediated either through site-specific binding of lipids or through lipid bilayer properties, and their various roles for protein functions have been described for many membrane proteins 92. Previously, the importance of CL for MgtA catalytic activity has been observed 124, its function for MgtA-mediated Mg2+ transport however remains elusive. Here,

we showed, using an ATPase activity assay, high specificity of MgtA for selected CL species (Paper II – Figure 1). While all tested CL species contained the same negatively charged head group, variations among the acyl chain composition regarding length and saturation degree were present. Interestingly, only the mixture of two specific species, CL18:1 and CL16:0, lead to maximum activity and only in the presence of at least 50 % CL16:0 maximum activity was observed (Paper II – Figure 2). It should be noted that although more than 50 different CL species are present in the E. coli membrane 65, only four selected CL species were tested as only few species of pure, synthetic CL were commercially available. However, CL species with the most common acyl chain composition present in E. coli, CL16:0, CL16:1 and CL18:1, were included in this study. Although we have shown the importance of CL18:1 and CL16:0 for MgtA activity, the interaction between MgtA and these lipids is not completely understood yet. The molecular shape of acyl chains affect lipid packing and, therefore, membrane bilayer properties 57. Monounsaturated lipids, like CL18:1, exhibit a kinked shape and tend to form fluid bilayers at physiological temperatures. Therefore, they exhibit higher flexibility and fluidity in comparison to lipids with saturated acyl chains 57. P-type ATPases undergo large conformational changes during their catalytic cycle, requiring flexibility regarding their lipid environment 168. Other members of the P-type ATPase family, including bacterial Cu2+-transporter CopA or heavy metal transporter ZntA, have shown higher activity levels upon increased lipid disorder and fluidity in the presence of unsaturated lipids 101,102,169. P-type ATPases have further been shown to depend strongly on the hydrophobic thickness of the surrounding membrane. SERCA showed activity peaks in the presence of lipids with C16-C18 tail lengths, corresponding to the most abundant lipids in the native environment 168. Additionally, MgtA exhibited activation in the presence of CL extracted from bovine heart membranes, which exhibits less variety of CL species than E. coli CL and mainly consists of CL18:2 69. In summary, a preference of MgtA for CL18:1 as an abundant monounsaturated lipid of the E. coli membrane corresponds to lipid interactions detected for other P-type ATPases. The interaction between MgtA and CL16:0 remains more elusive. We hypothesized that CL16:0 is required in the annular lipid belt of MgtA. However, it should be noted that all activation studies were performed in detergent-lipid micelles, which exhibit different physical properties regarding lateral pressure in comparison to a lipid bilayer. As a saturated lipid, CL16:0 forms ordered, non-fluid domains 57. These could stabilize MgtA in detergent-lipid micelles but might play a less important role under physiological conditions in a lipid bilayer. To verify if CL16:0 plays a role in micelle stabilization, activity assays at decreased temperatures, which would reduce lipid

fluidity and increase lipid order, could be performed in the future. Further, additional studies including other saturated CL species, for example CL18:0, should be performed to validate a specific effect of CL16:0 for MgtA activity. However, only a limited selection of saturated CL species was commercially available at the time of the experiment. The effect of CL on MgtA activity was additionally only tested on species containing four identical acyl chains. Although CL molecules show a large degree of symmetry regarding the length of their four acyl chains in vivo 59, CL species combining different acyl chains, for example containing two 16:0 and two 18:1 chains, have been detected in E. coli membranes 65. Additional experiments focusing on the effect of CL with a mixture of different acyl chains could give insight into lipid specificity of MgtA and the importance of physical properties of the lipid bilayer. To evaluate the role of CL18:1 and CL16:0 on MgtA activity under natural conditions and to analyze if CL selectivity of MgtA is dependent on the presence of the detergent octaethylene glycol monododecyl ether (C12E8), MgtA activity could be analyzed in a pure lipid bilayer in the absence of detergent. Although strong differences on MgtA ATPase activity were detected in the presence of different CL species, it cannot be excluded that variations among MgtA activation arose due to differences in solubility, and potentially aggregation, of the tested CL species in the presence of C12E8. Activity profiles of MgtA in a pure lipid bilayer could clarify these uncertainties. However, one needs to consider that CL exhibits a high transition temperature. The lipid transition temperature (Tm) is defined as the temperature, at which the physical state of lipids changes. Lipids convert from the ordered gel phase, which contains closely packed lipids, to the disordered liquid crystalline phase, in which acyl chains are randomly oriented and show increased fluidity 60. The lipid transition temperature is dependent on the lipid head group, but also increases with acyl

60 chain length and saturation . Shorter CL species, like CL14:1, already exhibit a Tm of 40 °C,

170 2+ 170 while CL16:0 exhibits a Tm of 58 °C . Furthermore, Mg has a stabilizing effect , which would likely increase the Tm of CL even further in MgtA ATPase activity studies. For SERCA, liquid crystalline bilayers were required for proper function 96. Unfortunately, MgtA activity assays cannot be performed at temperatures that would ensure the liquid crystalline phase of CLs as these high temperatures would affect MgtA stability and functionality. Therefore, activity assays should be performed in a lipid mixture of CL and additional lipids. These lipids should have reduced transition temperatures that would ensure liquid crystalline phase of the bilayer in the presence of CL. Activity profiles with increasing Mg2+ concentrations showed large variations among the

Michaelis-Menten constant (Km) values determined in the presence of different CL species. As Km indicates substrate binding affinity, large variations between Km propose an effect of specific CL

2+ species on Mg free affinity of MgtA (Paper II – Table 2). Higher Km values of MgtA were

determined in the presence of equimolar mixtures of CL18:1 and either CL16:0 (Km: 67 μM) or

CL16:1 (Km: 40 μM) in comparison to the one determined in the presence of E. coli CL (Km: 10 μM).

In the presence of CL18:1 alone a similar Km value, as obtained with E. coli CL, was observed

(Km: 12 μM). On the contrary, when the molar ratio of lipid species was shifted to a higher percentage or even 100 % of CL16:0 or CL16:1, even higher Km values were determined. In the presence of CL16:1 alone, a Km of 103 μM was observed. An equimolar mixture of CL16:1 and

CL16:0 revealed even a Km of 285 μM. Interestingly, for SERCA variations of the phospholipid chain length affected Ca2+ binding stoichiometry 171. In the presence of lipids with chain lengths between C16-C22 SERCA exhibited binding of two Ca2+-ions, reflecting the physiological state, while the presence of shorter (C12:0 or C14:1) or longer (C24:1) acyl chains led to binding of only one Ca2+-ion 171. It was proposed that differences in membrane thickness mediated by different lipid species affected closing of the outer Ca2+-entrance gate of SERCA, and therefore influenced Ca2+-binding stoichiometry. However, the variations in MgtA Mg2+ affinity in the presence of different CL species are likely mediated by another factor as CL18:1, CL16:0 and CL16:1 exhibit only small differences in acyl chain length. Additionally, as an anionic lipid CL interacts with Mg2+ and one might speculate that different CL species exhibit differences in Mg2+ affinity. CL has a propensity to form non-bilayer structures, proposed to form dynamic protein-lipid membrane domains 64. The phase behavior of CL has been shown to be dependent on divalent cations, including Mg2+ 172. Further, acyl chain length and composition can affect lipid phase behavior 173. Therefore, one might speculate that tested CL species exhibit small differences in their phase behavior, leading to variations in their interaction with Mg2+. Lastly, we also showed an activating effect of CL on MgtA homologues from Salmonella (Paper III – Figure 1) and Lactobacillus casei (Appendix A2). While stMgtA shares 91 % sequence identity with ecMgtA, Lactobacillus casei MgtA only shares 45 %. However, all MgtA homologues share a conserved C-terminal tryptophan residue, which is important for Mg2+ sensing of MgtA (Paper III – Figure 3). Unlike E. coli and Salmonella, Lactobacillus casei is a Gram-positive bacterium and, therefore, contains a different membrane lipid composition than E. coli 174. Analysis of CL specificity of Lactobacillus casei MgtA and the effect of additional native lipids on MgtA activation could shed light on the conservation of CL specificity among MgtA homologues, which could help to understand the role of CL for MgtA-mediated Mg2+ transport. Additionally, the characterization of eukaryotic MgtA homologues could contribute to our understanding of the lipid environment function for MgtA activity. In eukaryotic cells, CL is essentially only found in the mitochondria 67, and no evidence of mitochondrial location of eukaryotic MgtA homologues has been detected so far. Further, the sole characterized eukaryotic

MgtA representative, PH1, localizes to the vacuole in plant cells 134. No lipid dependency of PH1 has been detected yet and it remains unlikely that PH1 would exhibit specific CL interactions as CL is not present in the native vacuole membrane. However, an effect of anionic lipids in general cannot be excluded. Therefore, investigation of lipid interactions of eukaryotic MgtA homologues could contribute to our understanding of the role of lipids for MgtA function.

4.5 MgtA binds two cardiolipin molecules with positive cooperativity (Paper II)

Native MS revealed the selective binding of two CL molecules from a total E. coli lipid extract (Paper II – Figure 3) and nanoDSF revealed thermal stabilization of MgtA only in the presence of CL18:1 (Paper II – Figure 4). Therefore, we proposed a model of MgtA selectively binding two CL18:1 molecules while CL 16:0 is being required as an annular lipid (Paper II – Figure 6). Interestingly, lipid profiles of MgtA activity in the presence of E. coli CL and CL18:1 CL16:0 in an equimolar ratio indicated positive cooperative lipid binding. Positive cooperativity refers to the effect when ligand binding enhances binding of succeeding ligands, leading to a steeper saturation curve 175. Cooperativity is linked to allosteric modulation of biological macromolecules, mostly proteins, in which the effect of ligand binding is transferred to another distal site. This often

176 allows regulation of activity . Cooperativity can be quantified by the Hill coefficient (nH), which

177 measures the sigmoidal character of the activation curve . Binding curves with nH>1 are a direct measurement for cooperativity. MgtA exhibited a nH of 2.4-3 in the presence of E. coli CL or an equimolar lipid mixture of CL18:1 CL16:0. Although, nH is not a direct measurement of ligand binding sites, it is a useful index for the theoretical upper limit of the number of binding sites 177.

The determined nH of 2.4-3 of MgtA in the presence of E. coli CL and CL18:1 CL16:0 corresponds to our model of two CL binding sites on MgtA. Unfortunately, determination of nH of MgtA in the presence of other CL species was impossible due to the insufficient fit of the activation curves to the allosteric model. Allosteric modulation of lipids affecting protein-protein interactions have been detected previously 178,179. Recently, allosteric modulation through lipid binding by two different lipid types, PE and CL, was shown for E. coli ammonium channel AmtB by Patrick et al. (2018) 180. However, our data indicates allosteric modulation of MgtA through binding of two CL molecules of the same lipid type. Cong et al. (2017) showed that individual lipid binding affected allosteric interaction between an integral membrane protein, the E. coli ammonium channel AmtB, and a soluble regulatory protein, GlnK 179. The allosteric modulation was highly selective regarding lipid head group and acyl chain composition. It was proposed that binding of certain lipids stabilizes a

specific conformation of AmtB, which exhibits a higher binding affinity for GlnK and, therefore, results in increased AmtB-GlnK interaction. Negative allosteric modulation, observed in the presence of other lipid species, potentially stabilizes an alternative conformation with reduced GlnK affinity. A similar hypothesis was proposed by Patrick et al. (2018) 180. They found that the interaction between two selected residues on the AmtB surface was stabilized by CL, which locked AmtB in a specific conformational state that exhibited a higher binding affinity for PE 180. Mutational studies supported this hypothesis as disruption of the CL binding site also abolished the allosteric effect of the CL-PE pair. As CL has been shown to stabilize MgtA, a similar mechanism for positive allosteric modulation of MgtA-CL binding can be speculated. CL interaction at a specific binding site potentially stabilizes MgtA in a conformational state, which exhibits increased affinity for the second CL molecule. However, this is purely speculative as the role of CL for MgtA-mediated Mg2+ transport remains unclear.

4.6 The C-terminus of MgtA mediates Mg2+ sensitivity (Paper III)

2+ 124 Subramani et al. (2016) revealed a tight regulation of MgtA activity by Mg free levels . Strong

2+ inhibition of MgtA ATPase activity was detected at a Mg free concentration of 1 mM, which corresponds to the physiological intracellular concentration 124. However, the mechanism of Mg2+- dependent inhibition of MgtA remains unclear. Here, we showed that the penultimate amino acid at the MgtA C-terminus, W897, is involved in Mg2+ sensitivity (Paper III – Figure 3). MgtA mutants, in which residue W897 was substituted with either an alanine or phenylalanine, exhibited an

2+ altered Mg free binding capacity relative to the wild type. Further, residue W897 exhibited high conservation among MgtA homologues (Paper III – Figure 3). MgtA and MgtB from Salmonella, which contain the conserved tryptophan residue at their C-terminus, also exhibited Mg2+ sensitivity (Paper III – Figure 2). Yet, stMgtB showed a

2+ weaker inhibitory effect by Mg free in comparison to ecMgtA or stMgtA. These result support a common mechanism of Mg2+ sensitivity among MgtA homologues, likely mediated through selected conserved residues of the C-terminus. However, differences in Mg2+ sensing between

2+ MgtA and MgtB indicate additional components contributing to Mg free-dependent MgtA inhibition. Structural and biochemical investigation of these differences in Mg2+ sensing between MgtA and MgtB could contribute to our understanding of the complete mechanism. Interestingly, conservation of the C-terminal tryptophan was also detected among eukaryotic P-type ATPases, including the pig Na+-K+-pump (W1009). In this case, the residue was linked to interaction with the β-subunit, affecting the flexibility of the Na+-K+-pump and the equilibrium between

conformational states E1P and E2P dependent on β-subunit isoforms 181. However, as MgtA is present as a monomer, no indication of a C-terminal interaction between MgtA and another protein has been detected. Many P-type ATPases have been shown to be tightly regulated by accessory inhibitory proteins or by cis-acting autoinhibitory domains affecting ion transport 87. In the Na+-K+-ATPase the C-terminus has been shown to control Na+ affinity on both membrane sides 182. It further controls gating of a proton channel, which allows a cytoplasmic proton to enter during the K+- bound state 183. This stabilizes ion binding site III and allows an asymmetric stoichiometry of the

+ transported ions. For selected plant H -ATPases, belonging to P-type ATPase subfamily PIIIA, an autoregulatory function of both, the N- and C-terminal domains, has been detected 184. A regulatory mechanism was proposed, in which phosphorylation of the penultimate threonine residue induces binding of the regulatory 14-3-3 protein to the C-terminus. This leads to detachment of the autoinhibitory domain from an intramolecular binding site and transition of the enzyme from the low to the high activity state 184. Linked to this, the removal of the last 38 C- terminal residues transitioned the plant H+-ATPase AHA2 permanently into a high-affinity enzyme state 185. Although MgtA homologues exhibit strong conservation of their C-termini, their C-termini are shorter (10-12 amino acids) in comparison to the C-terminal regulatory domains of other P- type ATPases. Therefore, the mechanism how prokaryotic and eukaryotic P-type ATPases regulate protein function via their C-termini likely differs and the role of the C-terminus in autoregulation of MgtA-mediated Mg2+ transport requires further investigation. Interestingly, regulation of P-type ATPases has also been linked to interaction with small membrane-bound proteins 87. As an example, the ATPase activity of SERCA is regulated by the small protein sarcolipin, which inhibits SERCA through uncoupling of the pump and trapping it in the Ca2+-binding conformation 186. A role of small membrane-bound proteins in MgtA stability has also been revealed in vivo 150,152. The small peptide MgtS has been shown to positively affect Mg2+ influx by increasing MgtA concentrations in the membrane 152. To analyze the direct influence of MgtS on MgtA activity, we performed an ATPase assay of MgtA in the presence of increasing MgtS concentrations (Appendix A3). However, no activating or inhibitory effect of MgtS on MgtA activity was detected. Further, the presence of MgtS showed no effect on Mg2+ sensitivity of MgtA, highlighting no direct effect of MgtS on MgtA ATPase activity. It should be noted though that MgtS has a low solubility in aqueous solution as a membrane-bound peptide. Additional tests were performed to assure MgtS solubility, including MgtS solubilization in detergent, which also revealed no effect of MgtS on MgtA ATPase activity. However, it cannot be excluded that low

MgtS solubility affected MgtA interaction and MgtS influences MgtA activity through an unknown mechanism.

4.7 Mg2+ sensing mediated by the C-terminus of MgtA is dependent on pH (Paper III)

Subramani et al. (2016) revealed pH-dependency of MgtA ATPase activity with maximum activity detected only in a limited pH range of 7.0-7.2 124. This range corresponds to the cytoplasmic pH of E. coli, which ranges between 7.2 to 7.8 130. Interestingly, at a pH below 7.0 maximum MgtA activity was reduced but Mg2+ affinity remained unaffected. On the contrary, at a pH above 7.5 MgtA revealed reduced Mg2+ sensitivity. Here, we show that Mg2+ sensing, mediated by the highly conserved tryptophan residue (W897) at the C-terminus of MgtA, is affected by pH (Paper III – Figure 4). While mutants targeted in this residue were severely reduced in their Mg2+ sensitivity, the loss in Mg2+ susceptibility could be rescued at a mildly acidic pH between pH 6.2-6.5. However, the mechanism how acidic pH can counteract reduced Mg2+ sensitivity remains unclear. It was previously proposed that MgtA transports protons from the cytoplasm to the periplasm as counterions 124. Further, the plant MgtA homologue PH1 has been linked to proton transport 134. PH1 has been shown to form a heterocomplex with PH5, another P-type ATPase, which functions as a proton transporter involved in vacuole hyper-acidification 134. However, despite the close phylogenetic relationship to PIIIA-type ATPases, which mainly consist of yeast and plant proton pumps, neither MgtA nor PH1 contain the conserved membrane-embedded arginine residue shown to be involved in proton translocation of PIIIA-type proton ATPases Pma1 (Arg695) 187 and AHA2 (Arg655) 85. Therefore, low pH might rather facilitate protonation of certain amino acids directly involved in Mg2+ sensitivity of MgtA, thereby rescuing Mg2+ susceptibility of MgtA mutants affected in their C-termini. Further, pH has been shown to affect packing of CL-containing bilayers and the pH- dependent packing of CL domains has been proposed to play a role as a proton sink, for example necessary for ATP synthesis 188. It cannot be excluded that proton-dependent CL packing might also affect Mg2+-CL interactions, therefore influencing Mg2+-affinity of MgtA.

4.8 Lipid oxidation negatively affects MgtA activity (Paper II)

Unsaturated lipids are susceptible to oxidation, which can impair lipid integrity and destabilize membranes, resulting in cell lysis and death 78. Already low levels of lipid oxidation have been

shown to affect bilayer properties, such as membrane curvature and membrane permeability 78. As membrane proteins are highly dependent on their lipid environment and specific lipid-protein interactions, it is likely that lipid impairment can also negatively affect protein functions. To analyze the effect of lipid oxidation on protein activity we tested MgtA activity in the presence of CL stored under different conditions, ranging from oxidative-prone to protective lipid environments. Samples containing CL stored under oxidative-prone conditions without a protective argon layer showed the strongest decrease in MgtA activity. A decrease of 75 % in ATPase activity was observed after 30 days, promoting that lipid oxidation affects protein function (Paper II - Figure 2D). However, a decrease in MgtA ATPase activity, although to a lesser extent, was also detected in the presence of all other lipid samples over the time period of 60 days. Even though it cannot be excluded that the decrease in activity was due to or contributed by decreased protein or ATP quality over time, lipid oxidation remains the main potential source for decrease ATPase activity. Unsaturated lipids, and cyclopropane fatty acids to a lesser extent, are prone to lipid oxidation 189. During oxidation lipid hydroperoxides are produced as the primary products, which further decompose into various oxygenated and aliphatic fatty acid scission products 80. Lipid oxidation experiments were performed on CL extracted from E. coli membranes, which contain a wide variety of CL species 65. As we showed the importance of CL18:1 for MgtA activity, it is likely that oxidation leads to structural changes in the lipid structure, which interferes with the interaction between CL18:1 and MgtA. Simulation studies on CL peroxidation revealed conformational changes in the oxidized lipid structure, which lead to rearrangements of the acyl chains at to the membrane-water interface 81. Additionally, oxidation of only one acyl chain of CL was sufficient to induce tilting of the entire lipid molecule, likely due to interference with its molecular symmetry 81. As lipid oxidation affects bilayer properties, and the role of CL for MgtA-mediated Mg2+ transport remains elusive, variations in bilayer fluidity or thickness could additionally affect protein activity. However, as activity studies are performed in detergent-lipid micelles, which already differ in their physical properties compared to the lipid bilayer, this might only contribute partially. Lastly, one cannot exclude that lipid oxidation products, such as hydroperoxides or various oxygenated and aliphatic fatty acid scission products, affect MgtA activity through an unknown mechanism 80. Interestingly, the strongest protective effect was seen for CL stored with MgtA after solubilization in C12E8, which proposes a protective effect of lipid-protein interactions, likely through shielding of bound lipids. However, CL solubilized in C12E8 in the absence of MgtA was not affected in its activity either (Paper II – Supplementary Figure 1A). Therefore, it cannot be concluded whether MgtA had a protective effect or whether the presence of detergent contributed to lipid stability. It should be noted though that the effect of CL stored in the presence of C12E8

could only be tested in the time frame of seven days due to time limitations and the experiment should be repeated for a longer time period. Additionally, variations in MgtA ATPase activity between different E. coli CL batches were detectable (Paper II – Supplementary Figure 1B). Lipids are susceptible to oxidation in the presence of different catalytic systems, including light, heat as well as metals and metalloproteins 79. As it is unknown whether additional components, such as metal ions, are present in CL extracted from E. coli membranes, it cannot be excluded that additional catalysts for lipid oxidation are present in the lipid solution, which vary from lipid batch to lipid batch. Finally, the expression, purification and structural analysis of membrane proteins remains a bottleneck of protein research 190. As lipids have been revealed as essential components for membrane protein complexes, high lipid quality and integrity is essential for the successful characterization of membrane proteins. Protection of lipids through anti-oxidative measurements, for example during protein crystallization, should be considered as reduced lipid quality due to lipid oxidation will likely negatively affect the crystallization process and diffraction quality.

4.9 Cardiolipin thermally stabilizes MgtA (Paper II)

Next to protein activity, lipid-protein interactions have been shown to influence protein stability 92. To analyze the effect of CL on MgtA stability, nanoDSF experiments of MgtA in the absence and presence of selected lipids were performed. Thermal unfolding curves revealed a significant effect of E. coli CL on MgtA thermal stabilization (Paper II – Figure 4). Further, CL selectivity of MgtA thermal stabilization was observed as CL18:1 stabilized MgtA, while CL16:0 did not affect its thermal stabilization. Correspondingly, MgtA was not stabilized by 1-Palmitoyl-2-oleoyl-sn- glycero-3-phosphoethanolamine (POPE), which is commonly used to represents the native lipid environment in the E. coli cytoplasmic membrane. These results highlight the significance of CL- mediated MgtA stabilization. Stabilization of P-type ATPases by selected lipid species has been described for the plasma membrane Ca2+-pump 191. In this case, thermal stability has been studied in detergent- lipid micelles as well and Levi et al. showed that the detergent-lipid composition of the micelles is a major contributing factor for protein stabilization, not the total concentration of the micelles in the system. Further, they proposed that phospholipids differ in their protein stabilization properties as different lipids promote different conformational changes in the TM domain of the Ca2+- pump 191. Interestingly, for the Na+-K+-ATPase a specific lipid binding site was identified, which mediated protein stability but did not affect protein activity 106. A second lipid site stimulated activity

but did not influence protein stability. Both effects, at the separate sites, are independent and modulate distinct properties of Na+-K+-ATPase 106. Whether the stabilizing effect of CL on MgtA is mediated through specific lipid sites or through bilayer properties remains to be investigated. However, as CL18:1 selectively stabilized MgtA, while neither CL16:0 nor POPE supported MgtA stability, although both resemble the hydrophobic thickness of CL18:1, a site-specific binding site of CL18:1 is most likely mediating MgtA stability. It should be noted that differences in curve form between melting curves obtained in the absence or presence of CL were observed. MgtA melting curves of samples containing CL led to a decreased sigmoidal character of the curve, and therefore a less pronounced maximum in the first derivative. This effect might be due to differences in the final protein concentration between protein samples with and without CL. As all samples were incubated with lipids in excess, and additional lipids were removed by centrifugation prior to nanoDSF experiments, MgtA was partially removed in samples containing CL during the centrifugation step. This decreased the final protein concentration in comparison to the control samples without lipids. Interestingly, the selective pull- down of MgtA in the presence of CL additionally highlights the specific interaction between both components. nanoDSF measures the fluorescence intensity at 350 nm and 330 nm, which is mainly dependent on tryptophan residues present in the protein structure 192. Upon unfolding, buried tryptophan residues experience red shifts as the maximum emission intensity shifts to longer wavelengths during exposure to a polar environment 192. This is detected as a positive thermal transition signal. However, surface-exposed tryptophan residues exhibit a blue-shift or no shift at all upon protein unfolding, resulting in negative or unaffected thermal transition signals. Therefore, the measured thermal transition signal during nanoDSF experiments is a combination of signals from all tryptophan residues within the protein, which commonly results in signal interference. MgtA contains seventeen tryptophan residues, which are distributed all over the protein structure and exhibit a mixture of surface-exposed and buried residues. Additionally, the location of CL binding sites on MgtA is unknown and aromatic amino acids, such as tryptophan, are commonly involved in lipid interaction and stabilization 103. Therefore, CL binding to MgtA might burrow selected tryptophan residues and affect the thermal transition signal of MgtA, potentially contributing to the differences in curve shape observed between MgtA melting curves in the absence and presence of CL. Lastly, it should be noted that a mixture of heterogenous protein states can result in multiphasic unfolding curves. In the presence of ligands, this occurs due to the presence of a mixture of bound/ unbound states 192. MgtA revealed a shift in melting temperature in the presence

of CL. However, indication of a second melting temperature corresponding to the temperature detected for MgtA in the absence of CL has been observed. This likely indicates that the protein sample was not completely saturated with lipids and the presence of mixed states should be considered upon results interpretation.

4.10 Cardiolipin is not the main signal for MgtA localization to the E. coli cell poles (Paper II)

Localization of CL to the cell poles and septa has been observed for many bacterial species, including E. coli 64. Further, polar localization of MgtA has been revealed in E. coli where it co- localized with CL 124. For several E. coli membrane proteins CL has been implicated as a cellular localization signal, for example bacterial osmosensory transporter ProP 112. In this case, the polar localization of ProP was independent of its expression level, but correlated with CL polar localization 113. To determine whether CL functions as a localization signal for MgtA, we performed in vivo localization studies of fluorescently tagged MgtA (Paper II – Figure 5). Next to E. coli wild type strain MG1655, we tested MgtA localization in a CL-deficient strain MG1655Δcls, which is impaired in the CL synthesis. In both strains, MgtA localized to the cell poles and the localization pattern corresponds to previous observations 124. However, it should be noted that traces of CL have been detected in Δcls mutants, likely because residual CL synthesis is a side reaction of phosphatidylserine synthase 71. Therefore, it cannot be excluded that CL is completely absent from MG1655Δcls membranes. Further, another anionic lipid in the bacterial membrane, PG, has been shown to localize to the cell poles 193 and depletion of CL through genetic approaches even increased the concentration of PG at the cell poles 193. Subramani et al. (2016) showed that PG also activates MgtA ATPase activity, although to a lesser extent than CL 124. Therefore, it cannot be concluded whether CL is not involved in mediating MgtA polar localization or whether low levels of CL and the presence of PG can rescue MgtA localization to the cell poles in MG1655Δcls. MgtA localization was additionally analyzed in Vibrio cholerae (V. cholerae), which exhibits a similar membrane composition as E. coli, but does not contain a native MgtA homologue 194. In V. cholerae, no polar localization of MgtA was detectable, however the protein distributed evenly around the cell membrane. In summary, no indication that CL functions as a localization signal for the polar localization of MgtA has been obtained. Yet, a minor role of CL contributing to the cellular localization of MgtA cannot be exclude.

It should be additionally noted that the cellular mechanism responsible for the polar localization of proteins is generally not well understood. One hypothesis supports that the main driving force for the formation of CL-enriched domains at the bacterial cell poles and septa is curvature-mediated phase segregation, as CL is classified as a “high-curvature lipid”. 193. However, in recent years an alternative hypothesis has been proposed, in which recruitment of CL to the cell poles arises through the polar localization of proteins that influence lipid composition 195. In this model polarly localized proteins organize fluid lipids and induce lipid domain formation. The MgtA N-terminus is enriched in positively charged residues and harbors intrinsically disordered region 124 and as N-terminal interaction with lipids has been detected for P-type ATPase PH1 134, a role of the MgtA N-terminus for protein trafficking to the bacterial cell poles has been previously proposed. However, an N-terminal deletion mutant (NΔ1-31-MgtA) also co- localized with CL at the cell poles, indicating no function of the MgtA N-terminus as a trafficking signal 124. A recent study by Jephthah et al. (2020) indicated that MgtA anchoring to the anionic membrane surface is mediated by its intrinsically disordered N-terminus 121. Therefore, the role of the MgtA N-terminus for cellular localization of MgtA requires further investigation. Interestingly, PhoP, the regulator of the two-component system PhoP/PhoQ involved in MgtA upregulation under low Mg2+ concentrations, has been shown to localize to the cell poles in Salmonella upon phosphorylation 196. Although the organization of the lipid bilayer in cells remains under investigation, the presence of membrane microdomains containing embedded protein complexes and allowing the formation of regions of functional specialization has been proposed 54. Co-localization of MgtA with PhoP in CL microdomains would allow a tighter regulation of Mg2+ homeostasis in E. coli cells. However, this remains purely speculative.

4.11 Identification of cardiolipin binding sites on MgtA remains an obstacle (Paper II & Paper III)

Native MS revealed the presence of two specific CL binding sites in MgtA and the identification of these binding sites could give insight into the role of CL for MgtA-mediated Mg2+ transport. However, the localization of these sites on MgtA remains unknown. We identified a CL binding site motif, which consists of two positively charged residues followed by a polar one 103, in the MgtA sequence (RRY at residue 893-895). However, mutation of the site did not affect MgtA activity and cooperative lipid binding remained unaffected (Paper II – Supplementary Figure 4). This indicated that the site is likely not involved in direct CL binding. However, as MgtA requires a large molar excess of CL for activity and we proposed a function of CL16:0 as an annular lipid,

we cannot exclude that CL interacts with MgtA at the indicated site through less specific lipid interactions. Additionally, it should be noted that interpretation of mutational studies of MgtA using an ATPase assay remains intricate. Distinction whether decreased activity levels of MgtA mutants are due to disturbance of CL binding or whether important catalytic residues are affected by the mutation remains problematic. As an example, MgtA mutant W897A, affected in its C-terminus, exhibited a strongly decreased ATPase activity and we showed that this residue is involved in Mg2+ sensing (Paper III – Figure 3). However, one cannot exclude that residue W897 is also involved in CL binding as several factors can contribute to reduced MgtA activity levels. Additionally, ATPase activity of MgtA mutants W897A and ΔWQ, in the latter the two ultimate amino acids at the C-terminus of MgtA are deleted, were inhibited by higher concentrations of CL. Mildly acidic pH could alleviate CL-dependent inhibition of W897A and ΔWQ (Paper III – Figure 5). This led to the hypothesis that the MgtA C-terminus stabilizes CL interaction. Mutations at the C-terminus interfere with CL binding, which results in rapid CL exchange. As a mildly acidic pH has been shown to affect CL packing in membranes 197, we speculated that a lower pH induces tighter interaction of CL with the MgtA C-terminus. This could additionally play a role for the rescue of Mg2+ sensitivity of these MgtA mutants. Having said that, both C-terminal MgtA mutants were unaffected in their cooperative CL binding (Appendix A4). A nH of 2.2-2.6 was determined, which corresponds to MgtA wild type (nH: 2.6). Additionally, both revealed thermal stabilization in the presence of E. coli CL, which supports that the MgtA C-terminus is not strongly involved in site- specific CL binding. As we propose that CL also plays a role as an annular lipid for MgtA-mediated Mg2+ transport, interaction of the MgtA C-terminus with the annular lipid shell cannot be excluded though. Native MS, which has been successful in revealing specific binding of two CL molecules to MgtA (Paper II – Figure 3), has emerged as powerful tool for the investigation of dynamic and heterogeneous membrane protein structures and lipid interactions 198,199 and, therefore, provides a promising alternative technique to identify the CL binding sites on MgtA. However, native MS revealed limitations in obtaining a spectrum of MgtA without any bound CL. In most protein samples CL from natural E. coli membranes co-purified with MgtA during the purification process. Measurements to remove the bound lipid, through extensive washing with detergent or replacement with another anionic lipid, resulted in decreased protein stability and no spectrum could be obtained. As CL stabilizes MgtA and lipid removal strongly affected the protein sample quality, one can speculate that interference with the CL binding sites will likely influence MgtA stability and potentially no MS spectrum could be obtainable. However, optimization of native MS

measurement conditions might allow studying of MgtA mutants affected in their lipid binding sites. All native MS experiments were performed in the presence of n-Dodecyl β-D-maltoside (DDM). However, alternative detergents could be tested to improve MgtA stability for native MS measurements. In other studies, DDM was replaced with n-Udecyl-β-D-maltopyranoside (UDM), which has a higher critical micelle concentration. Therefore, it promotes acquisition of MS spectra at lower activation energies 106, which could promote obtaining a spectrum of less stable MgtA mutants. Additionally, reconstitution of MgtA in C12E8 was tested for native MS measurements. This detergent is actually of preferred choice as it supports MgtA activity in comparison to

124 DDM . However, no spectrum of MgtA in the presence of CL and C12E8 was measurable. Interestingly, no MS spectrum of the renal Na+-K+-ATPase was detectable in the presence of

C12E8 either. It was hypothesized that C12E8 is incompatible with MS due to sample inhomogeneity

+ + as C12E8, in comparison to DDM, led to the formation of higher oligomers of the Na -K -

106,200 124 ATPase . No oligomerization of MgtA has been detected in the presence of DDM or C12E8 and the incompatibility of MgtA and C12E8 for MS measurements remains unclear. Lastly, in recent years novel in silico studies and tools have contributed to our understanding of protein-lipid interplay, which could be beneficial for future studies on the interaction between MgtA and CL. A systematic study performed on crystal structures of proteins with bound CLs has identified characteristic amino acid residues and supersecondary structural motifs characteristic for CL-binding regions 201. Additionally, novel in silico tools using simulations have contributed to the identification of lipid interaction sites of membrane proteins 202. However, our limited high-resolution structural information of MgtA prevents the application of these novel insights and tools for the identification of CL-MgtA interaction sites. Obtaining the protein structure of MgtA remains therefore a major objective to study MgtA-CL interplay.

4.12 The role of cardiolipin in MgtA-mediated Mg2+ transport remains elusive (Paper II)

CL has been shown to be essential for MgtA activity and stability. However, the biological role of CL for MgtA-mediated Mg2+ transport remains elusive. For some membrane proteins, CL has been proposed to function as a proton sink, for example for the ATP-synthase function 64. As an anionic lipid CL can interact with protons, trap them in the ATP-synthase vicinity and, finally, facilitate proton entry to the channel 64. A similar role of CL in Mg2+ recruitment for MgtA-mediated transport was proposed 124. However, the specific interaction of MgtA with selected CL species suggest a more complex role of CL during Mg2+ transport.

We proposed a model of MgtA containing two specific CL18:1 binding sites with positive cooperative binding, while CL16:0 plays a role as an annular lipid (Paper II – Figure 6). However, some P-type ATPases have revealed changes of their cross-sectional area at the membrane interface dependent on the conformational state, which results in local differences in the bilayer thickness. Therefore, preferences for phospholipids varying in their acyl chain length for different protein conformations during the catalytic cycle have been proposed 203. The necessity of MgtA for both CL species, CL18:1 and CL16:0, for optimal activity might also indicate selective binding of each CL species in a specific conformational state. Therefore, CL specificity for different conformations was assessed by native MS. MgtA was locked into two conformational states, E2

- - 204 and E1, with inhibitors AlF4 and ADP-AlF4 respectively and, following, CL binding was determined (Paper II – Supplementary Figure 3). However, no difference in CL interaction was observed between both states and both samples revealed binding of two CL molecules. Although the low resolution of the MS measurements did not allow differentiation of CL species in the spectrum, binding of selected CL species during a specific conformational state of MgtA cannot be excluded. However, the experiment should be repeated, and experimental settings should be optimized. In future studies MgtA-CL interaction should be analyzed in the presence of higher inhibitor concentrations as tested conditions likely were not sufficient to assure complete transition

- of MgtA to the respective states. Additionally, other ATPase inhibitors, including BeF4 , could be tested to assure complete transition into the opposing conformational states 205. Additionally, the role of CL for MgtA-mediated Mg2+ transport could be investigated by fluorescence spectroscopy, as it has been successfully used to study the catalytic cycle of the Na+-K+-ATPase, among other P-type ATPases 206. However, addition of fluorescent labels to MgtA or CL affected protein expression and protein-lipid interactions. A MgtA construct containing a green fluorescent protein (GFP) tag at the C-terminus revealed severely decreased ATPase activity (Appendix A5). We hypothesized that reduced activation levels are due to interference of the GFP-tag with conserved residues at the C-terminus involved in Mg2+ sensing (Paper III – Figure 3). A N-terminal fluorescently tagged MgtA construct could not be expressed at all. It remains unclear why the N-terminal located GFP interfered with protein expression, however it seems likely that it affected membrane insertion of MgtA. It should be noted though that a N- terminal His-tagged MgtA variant produced high yields and was commonly used to express and purify wild type MgtA. Additionally, fluorescently tagged CL species, TopFluor-CL18:1 (Avanti Lipids, Cat# 810286) and TopFluor-CL16:0 (Avanti Lipids, Cat# 810246) were commercially acquired. However, MgtA-CL interaction was severely affected by the fluorescently labelled lipids.

As the fluorescent label was localized at the lipid head group for both species, it likely interfered with CL binding to MgtA. Interestingly, we revealed differences in CL-dependency between Salmonella MgtA and MgtB ATPase activity (Paper III – Figure 1). While stMgtA showed increased activation levels in the presence of CL, resembling its E. coli counterpart, activity of stMgtB was less affected by CL. stMgtA and stMgtB share only 50 % sequence identity 115 and differ in their genetic location with mgtB being encoded on pathogenicity island SPI-3 128. MgtB is additionally expressed under mildly acidic conditions in a PhoP/PhoQ-independent manner 207, which has not been detected for MgtA. Further, MgtB function is temperature-sensitive as MgtB-mediated Mg2+ uptake was observed at 37 °C, but not at 20 °C 118. Mg2+ uptake by MgtA was decreased at lower temperatures, but to a lesser extent. This led to the hypothesis that MgtB plays a role during infection of warm-blooded hosts, but not in non-host environments, while MgtA has another cellular function 115. However, recent findings highlighting the role of MgtA for Salmonella survival under macrophage-like conditions, propose a role of MgtA during host infection 116,156. Investigation of structural differences between MgtA and MgtB could contribute to the identification of CL interaction sites on MgtA and an understanding of the role of CL for Mg2+ transport might give an insight into the different cellular functions of MgtA and MgtB. Lastly, bacterial cells contain a large variety of lipids and the lipid composition of the membrane bilayer is highly adaptable to the cellular environment 57,71. However, the necessity and functional role of over thousand different lipid species in bacteria remains unclear. MgtA is

2+ under a tight Mg free-dependent regulatory network regarding its expression and activity. One might hypothesize that the specificity of MgtA for selected CL species contributes to the complex network of regulation by promoting MgtA activity only if selected CL species are present in the membrane bilayer. CL only constitutes 5 % of the lipid content in E. coli membranes under normal conditions, however more than 50 different CL species have been identified 65. CL18:1 and CL16:0 are among the most common CL species in E. coli, yet the percentage of CL and the composition of the acyl chains have been shown to change dependent on external factors, for example under osmotic pressure 71. Although the cellular function of MgtA remains elusive, it has been linked to survival under specific environmental conditions, including survival in macrophages during Salmonella infection 116,156. Lipidomic investigations of the composition of Salmonella or E. coli membranes in a macrophage environment could reveal whether lipid composition changes occur during bacterial infection that could modulate MgtA activity.

5. Summary

Pathogenic E. coli pose a global health threat and the intestinal pathogens, EHEC and EPEC, contribute as major causes for foodborne gastrointestinal illnesses. With rising antibiotic resistance, a better understanding of cellular components involved in the infection process of EHEC and EPEC is needed to develop better treatments. The following results, described in this thesis, contribute to our understanding of bacterial adhesion and ion homeostasis, affecting bacteria - host cell interaction.

Part I:

x The elusive structures of intimin passenger subdomains D00 and D0 have been revealed by X-ray crystallography, confirming a bacterial Ig-like (Big) domain for both. However, D00 presents structural differences to Big domains D0 and D1, while exhibiting similarities to the extracellular subdomains of mammalian cadherins. This highlights a novel type of Big domain fold of D00. In summary, these insights allow the modelling of the entire extracellular passenger of intimin and its connection to the membrane-bound β-barrel for the first time. x Functional subregions along the intimin passenger, characterized by variations in flexibility, have been revealed. The passenger consists of a rigid base (D00-D0), a flexible middle section (D0-D1-D2) and a rigid C-terminus (D2-D3). Monte Carlo simulations proposed that the rigid base, represented by D00-D0, increases the radius of reach of the intimin passenger, potentially promoting a more probable interaction between intimin and Tir. x In silico analysis highlighted D00 as a common module in many passengers of inverse autotransporters, located selectively at the interface between the membrane-embedded β-barrel and the C-terminal passenger.

Part II: x Analysis of the ATPase activity of E. coli MgtA revealed strong lipid selectivity towards specific cardiolipin species, varying only in the length and saturation degree of the acyl chains. The combination of two specific cardiolipin species, cardiolipin 18:1 and cardiolipin 16:0, induced maximum ATPase activity of MgtA. Two specific cardiolipin binding sites were observed by native mass spectrometry, and the analysis of kinetic studies of MgtA ATPase activity showed cooperative binding of more than one cardiolipin. MgtA exhibited thermal stabilization by the specific lipid species cardiolipin 18:1, while the presence of cardiolipin 16:0 did not affect MgtA stabilization. We propose a model of MgtA-cardiolipin interaction, in which two cardiolipin 18:1 bind specifically to MgtA as non-annular lipids, while cardiolipin 16:0 plays an unknown role for MgtA ATPase activity as an annular lipid. x In contrast to previous studies that revealed co-localization of MgtA and cardiolipin at E. coli cell poles, we highlighted that cardiolipin does not function as the main localization signal for the polar localization of MgtA. x The biochemical characterization of Salmonella MgtA and MgtB revealed the conservation

2+ of anionic lipid dependency and Mg free-dependent inhibition between Salmonella and E. coli MgtA. MgtB ATPase activity was not dependent on the presence of anionic lipids and

2+ MgtB exhibited only weak Mg free-dependent inhibition. These results highlight different biochemical properties and potentially different cellular functions of MgtA and MgtB. x Mutational studies revealed that the C-terminus of MgtA, especially a highly conserved tryptophan at the penultimate residue (W897), plays a role for Mg2+ sensitivity of MgtA. Further, the loss of Mg2+ sensitivity, instigated through the deletion of the MgtA C-terminus, could be restored in mildly acidic pH.

6. Future perspectives

In Paper I, we highlighted subregions along the intimin passenger varying in their flexibility with subdomains D00-D0 forming a rigid base at the membrane interface. MC simulations proposed that a rigid passenger base, represented by D00-D0, increases the radius of reach of the passenger. This could promote a more probable intimin-Tir interaction and potentially increase bacterial adhesion to the host cells. To prove a cellular role of a rigid passenger base in vivo, cell adhesion assays of intimin mutants need to be performed. However, no cellular adhesion could be performed so far as expression levels of intimin mutants, in which the D00-D0 connector was replaced by a glycine- serine sequence, were severely reduced. The connector region was proposed to mediate the constitutively extended conformation of D00-D0 through a tight hydrogen bond network. Therefore, future studies should continue with these cellular adhesion assays. Additional intimin mutants should be included, in which the hydrogen bond network of the D00-D0 connector is not completely abolished. This will hopefully allow higher expression levels of intimin mutants lacking the rigid passenger base represented by D00-D0 and the role of a rigid passenger base could be analyzed on a cellular level. Additionally, we highlighted D00 as a common element of inverse autotransporters, located between the membrane-embedded β-barrel and C-terminal passenger. Future studies should investigate whether D00 plays a role as a rigid base at the N-terminal part of the passenger of other inverse autotransporters as well and whether these passengers also contain functional subregions with varying degrees of flexibility. Therefore, the passenger structure of other inverse autotransporters, for example Yersinia pseudotuberculosis invasin, should be analyzed under solution conditions using SAXS or negative stain electron microscopy. This could give an overview of different conformational states of the passengers, which are present in vivo. Additionally, these studies could shed light on the cellular function of D00 as the juncture domain between the membrane-embedded β-barrel and extracellular passenger. In Papers II and III, insights into the interaction of the bacterial magnesium transporter MgtA with the lipid environment and its substrate Mg2+ have been revealed. However, the lack of high-resolution structural data of MgtA hampers our understanding of MgtA-cardiolipin or Mg2+ interaction on a molecular level. Therefore, obtaining a high-resolution structure of the entire Mg2+ transporter remains the major aim for future studies. Attempts using X-ray crystallography has been unsuccessful so far. However, novel insights into MgtA-cardiolipin interaction allow optimization of crystallization

conditions that support MgtA stabilization. Cryogenic electron microscopy should be additionally considered for future structural studies on MgtA as it has been successfully used for the structural

208 analysis of other P-type ATPases, for example the PIV-ATPase lipid flippase Drs2p . Obtaining a high-resolution structure of MgtA would contribute to the identification of cardiolipin binding sites, which further could shed light on the function of the lipid for MgtA-mediated Mg2+ transport. Additionally, novel in silico tools, like MemProtMD, have been used successfully to identify membrane protein-lipid interaction 202, however a high-resolution protein structure is required for these tools. The biochemical characterization of MgtA is mainly performed by an ATPase activity assay, which does not directly measure Mg2+ transport and, so far, no direct evidence of Mg2+ transport by MgtA has been obtained. In order to understand the electrogenic nature of MgtA- mediated Mg2+-transport and the presence of possible transported counterions, an assay should be established that directly measures MgtA-mediated Mg2+ transport in liposomes. Further, the analysis of Mg2+ transport by MgtA in a lipid bilayer in the absence of detergent would allow to study the role of specific cardiolipin species for MgtA activity in a native lipid environment. This could confirm results presented in Paper II and allow deeper insights into the role of cardiolipin for MgtA-mediated Mg2+ transport. Lastly, recent studies revealed a cellular function of MgtA for the survival of Salmonella and E. coli under ribosomal stress induced through antibiotics 209 or in a macrophage-like environment 116,156. Therefore, MgtA should be promoted as a potential drug target of pathogenic E. coli, among other bacterial species, and identification of MgtA inhibitors present a possible future treatment against bacteria under environmental stress conditions.

7. References

1. Kaper, J. B., Nataro, J. P. & Mobley, H. L. T. Pathogenic Escherichia coli. Nat. Rev. Microbiol. 2, 123–140 (2004). 2. Nataro, J. P. & Kaper, J. B. Diarrheagenic Escherichia coli. Clin. Microbiol. Rev. 11, 142– 201 (1998). 3. Majowicz, S. E. et al. Global incidence of human Shiga Toxin–producing Escherichia coli infections and deaths: A systematic review and knowledge synthesis. Foodborne Pathog Dis 11, 447–55 (2014). 4. Leimbach, A., Hacker, J. & Dobrindt, U. E. coli as an All-Rounder: The Thin Line Between Commensalism and Pathogenicity. Between Pathog. Commensalism 358, 3–32 (2013). 5. Croxen, M. A. & Finlay, B. B. Molecular mechanisms of Escherichia coli pathogenicity. Nature Reviews Microbiology vol. 8 26–38 (2010). 6. Roussel, C. et al. Increased EHEC survival and virulence gene expression indicate an enhanced pathogenicity upon simulated pediatric gastrointestinal conditions. Pediatr. Res. 80, 734–743 (2016). 7. Croxen, M. A. et al. Recent advances in understanding enteric pathogenic Escherichia coli. Clin. Microbiol. Rev. 26, 822–880 (2013). 8. Nguyen, Y. & Sperandio, V. Enterohemorrhagic E. coli (EHEC) pathogenesis. Front. Cell. Infect. Microbiol. 2, 1–7 (2012). 9. Webster, K. & Schnitzler, E. Hemolytic uremic syndrome. in Handbook of Clinical Neurology vol. 120 1113–1123 (Elsevier B.V., 2014). 10. Kampmeier, S., Berger, M., Mellmann, A., Karch, H. & Berger, P. The 2011 German Enterohemorrhagic Escherichia Coli O104:H4 Outbreak-The Danger Is Still Out There. Curr Top Microbiol Immunol 117–148 (2018) doi:10.1007/82_2018_107. 11. Hu, J. & Torres, A. G. Enteropathogenic Escherichia coli: Foe or innocent bystander? Clin. Microbiol. Infect. 21, 729–734 (2015). 12. Zhou, Y. et al. Characteristics of diarrheagenic Escherichia coli among children under 5 years of age with acute diarrhea: A hospital based study. BMC Infect. Dis. 18, 1–10 (2018). 13. Frankel, G. et al. Enteropathogenic and enterohaemorrhagic Escherichia coli: More subversive elements. Mol. Microbiol. 30, 911–921 (1998). 14. Schmidt, M. A. LEEways: Tales of EPEC, ATEC and EHEC. Cell. Microbiol. 12, 1544–1552 (2010). 15. Bardiau, M., Szalo, M. & Mainil, J. G. Initial adherence of EPEC, EHEC and VTEC to host cells. Vet. Res. 41, (2010). 16. Cassels, F. J. & Wolf, M. K. Colonization factors of diarrheagenic E. coli and their intestinal receptors. J. Ind. Microbiol. 15, 214–226 (1995). 17. McWilliams, B. D. & Torres, A. G. Enterohemorrhagic Escherichia coli Adhesins. Microbiol. Spectr. 2, (2014).

18. Gaytán, M. O., Martínez-Santos, V. I., Soto, E. & González-Pedrajo, B. Type three secretion system in attaching and effacing pathogens. Front. Cell. Infect. Microbiol. 6, 1– 25 (2016). 19. Santos, A. S. & Finlay, B. B. Bringing down the host: Enteropathogenic and enterohaemorrhagic Escherichia coli effector-mediated subversion of host innate immune pathways. Cell. Microbiol. 17, 318–332 (2015). 20. Pinaud, L., Sansonetti, P. J. & Phalipon, A. Host Cell Targeting by Enteropathogenic Bacteria T3SS Effectors. Trends Microbiol. 26, 266–283 (2018). 21. Kenny, B. et al. Enteropathogenic E. coli (EPEC) transfers its receptor for intimate adherence into mammalian cells. Cell 91, 511–520 (1997). 22. McDaniel, T. K., Jarvis, K. G., Donnenberg, M. S. & Kaper, J. B. A genetic locus of enterocyte effacement conserved among diverse enterobacterial pathogens. Proc. Natl. Acad. Sci. U. S. A. 92, 1664–1668 (1995). 23. McDaniel, T. K. & Kaper, J. B. A cloned pathogenicity island from enteropathogenic Escherichia coli confers the attaching and effacing phenotype on E. coli K-12. Mol. Microbiol. 23, 399–407 (1997). 24. Luo, Y. et al. Crystal structure of enteropathogenic Escherichia coli intimin–receptor complex. Nature 405, 1073–1077 (2002). 25. Frankel, G. et al. Intimin and the host cell - Is it bound to end in Tir(s)? Trends Microbiol. 9, 214–218 (2001). 26. Kalman, D. et al. Enteropathogenic E . coli acts through WASP and Arp2/3 complex to form actin pedestals. Nat Cell Biol 1, 389–391 (1999). 27. Campellone, K. G., Giese, A., Donald, J. & Leong, J. M. A tyrosine-phosphorylated 12- amino-acid sequence of enteropathogenic Escherichia coli Tir binds the host adaptor protein Nck and is required for Nck localization to actin pedestals. Mol. Microbiol. 43, 1227– 1241 (2002). 28. Goosney, D. L. et al. Enteropathogenic E . coli translocated intimin receptor, Tir, interacts directly with α-actinin. Curr. Biol. 735–738 (2000). 29. DeVinney, R. D. E. et al. Enterohemorrhagic Escherichia coli O157:H7 produces Tir, which is translocated to the host cell membrane but is not tyrosine phosphorylated. Infect. Immun. 67, 2389–2398 (1999). 30. Vingadassalom, D. et al. Insulin receptor tyrosine kinase substrate links the E . coli O157:H7 actin assembly effectors Tir and EspFU during pedestal formation. PNAS 106, 6754–6759 (2009). 31. Campellone, K. G., Robbins, D. & Leong, J. M. EspFU is a translocated EHEC effector that interacts with Tir and N-WASP and promotes Nck-independent actin assembly. Dev. Cell 7, 217–228 (2004). 32. Touzé, T., Hayward, R. D., Eswaran, J., Leong, J. M. & Koronakis, V. Self-association of EPEC intimin mediated by the β-barrel-containing anchor domain: A role in clustering of the Tir receptor. Mol. Microbiol. 51, 73–87 (2004). 33. Frankel, G. et al. The Cell-binding Domain of Intimin from Enteropathogenic Escherichia

coli binds to beta1 Integrins. J. Biol. Chem. 271, 20359–20364 (1996). 34. Sinclair, J. F. & O’Brien, A. D. Intimin types α, β, and γ bind to nucleolin with equivalent affinity but lower avidity than to the translocated intimin receptor. J. Biol. Chem. 279, 33751–33758 (2004). 35. Valiatti, T. B. et al. Draft Whole-Genome Sequence of a Uropathogenic Escherichia coli strain carrying the eae gene. Microbiol Resour Announc 8, (2019). 36. Leo, J. C., Oberhettinger, P., Schütz, M. & Linke, D. The inverse autotransporter family: Intimin, invasin and related proteins. Int. J. Med. Microbiol. 305, 276–282 (2015). 37. Knutton, S. et al. Down regulation of intimin expression during attaching and effacing enteropathogenic Escherichia coli adhesion. Infect. Immun. 65, 1644–1652 (1997). 38. Leo, J. C., Grin, I. & Linke, D. Type V secretion: Mechanism(s) of autotransport through the bacterial outer membrane. Philos. Trans. R. Soc. B Biol. Sci. 367, 1088–1101 (2012). 39. Drobnak, I. et al. Of linkers and autochaperones: An unambiguous nomenclature to identify common and uncommon themes for autotransporter secretion. Mol. Microbiol. 95, 1–16 (2015). 40. Leo, J. C. et al. The Intimin periplasmic domain mediates dimerisation and binding to peptidoglycan. Mol. Microbiol. 95, 80–100 (2015). 41. Fairman, J. W. et al. Crystal structures of the outer membrane domain of intimin and invasin from enterohemorrhagic E. coli and enteropathogenic Y. pseudotuberculosis. Structure 20, 1233–1243 (2012). 42. Bodelón, G., Marín, E. & Fernández, L. Á. Role of periplasmic chaperones and BamA (YaeT/Omp85) in folding and secretion of intimin from enteropathogenic Escherichia coli strains. J. Bacteriol. 191, 5169–5179 (2009). 43. Bodelón, G., Palomino, C. & Fernández, L. Á. Immunoglobulin domains in Escherichia coli and other enterobacteria: From pathogenesis to applications in antibody technologies. FEMS Microbiol. Rev. 37, 204–250 (2013). 44. Leo, J. C. et al. Secretion of the intimin passenger domain is driven by protein folding. J. Biol. Chem. 291, 20096–20112 (2016). 45. Tsai, J. C. et al. The Bacterial Intimins and Invasins: A Large and Novel Family of Secreted Proteins. PLoS One 5, (2010). 46. Heinz, E. et al. Conserved Features in the Structure, Mechanism, and Biogenesis of the Inverse Autotransporter Protein Family. Genome Biol. Evol. 8, 1690–1705 (2016). 47. Hamburger, Z. A., Brown, M. S., Isberg, R. R. & Bjorkman, P. J. Crystal Structure of Invasin: A Bacterial Integrin-Binding Protein. Science (80-. ). 286, 291–295 (1999). 48. Oberhettinger, P. et al. Intimin and Invasin Export Their C-Terminus to the Bacterial Cell Surface Using an Inverse Mechanism Compared to Classical Autotransport. PLoS One 7, (2012). 49. Oberhettinger, P., Leo, J. C., Linke, D., Autenrieth, I. B. & Schütz, M. S. The inverse autotransporter intimin exports its passenger domain via a hairpin intermediate. J. Biol. Chem. 290, 1837–1849 (2015).

50. Roman-Hernandez, G., Peterson, J. H. & Bernstein, H. D. Reconstitution of bacterial autotransporter assembly using purified components. Elife 3, 1–20 (2014). 51. Ieva, R., Tian, P., Peterson, J. H. & Bernstein, H. D. Sequential and spatially restricted interactions of assembly factors with an autotransporter β domain. Proc. Natl. Acad. Sci. U. S. A. 108, (2011). 52. Voet, D. & Voet, J. G. Biochemistry. (Wiley, 2011). 53. Raetz, C. R. H. Enzymology, genetics, and regulation of membrane phospholipid synthesis in Escherichia coli. Microbiol. Rev. 42, 614–659 (1978). 54. Engelman, D. M. Membranes are more mosaic than fluid. Nature 438, 578–580 (2005). 55. Raetz, C. R. H. & Dowhan, W. Biosynthesis and Function of Phospholipids in Escherichia coli. J. Biol. Chem. 265, 1235–1238 (1990). 56. Dowhan, W., Bogdanov, M. & Mileykovskaya, E. Functional Roles of Lipids in Membranes. in Biochemistry of Lipids, Lipoproteins and Membranes 1–40 (Elsevier, 2008). doi:10.1016/B978-0-444-63438-2.00001-8. 57. Ernst, R., Ejsing, C. S. & Antonny, B. Homeoviscous adaptation and the regulation of membrane lipids. J. Mol. Biol. 428, 4776–4791 (2016). 58. Stillwell, W. Membrane Polar Lipids. in An Introduction to Biological Membranes 63–87 (2016). doi:10.1016/b978-0-444-63772-7.00005-1. 59. Haines, T. R. A new look at Cardiolipin. Biochim. Biophys. Acta 1788, 1997–2002 (2009). 60. Oldfield, E. & Chapman, D. Dynamics of lipids in membranes: Heterogeneity and their role of cholesterol. FEBS Lett. 23, 285–297 (1972). 61. Lewis, R. N. A. H. & McElhaney, R. N. The physicochemical properties of cardiolipin bilayers and cardiolipin-containing lipid membranes. Biochim. Biophys. Acta 1788, 2069– 2079 (2009). 62. Mileykovskaya, E. & Dowhan, W. Visualization of phospholipid domains in Escherichia coli by using the cardiolipin-specific fluorescent dye 10-N-Nonyl Acridine Orange. J. Bacteriol. 182, 1172–1175 (2000). 63. Wood, J. M. Perspective: challenges and opportunities for the study of cardiolipin, a key player in bacterial cell structure and function. Curr. Genet. 64, 795–798 (2018). 64. Mileykovskaya, E. & Dowhan, W. Cardiolipin membrane domains in prokaryotes and eukaryotes. Biochim. Biophys. Acta 1788, 2084–2091 (2009). 65. Garrett, T. A., O’Neill, A. C. & Hopson, M. L. Quantification of cardiolipin molecular species in Escherichia coli lipid extracts using liquid chromatography/electrospray ionization mass spectrometry. Rapid Commun. Mass Spectrom. 26, 2267–2274 (2012). 66. Tan, B. K. et al. Discovery of a cardiolipin synthase utilizing phosphatidylethanolamine and phosphatidylglycerol as substrates. Proc. Natl. Acad. Sci. U. S. A. 109, 16504–16509 (2012). 67. Daum, G. Lipids of mitochondria. Biochim. Biophys. Acta 822, 1–42 (1985). 68. Schlame, M. Cardiolipin synthesis for the assembly of bacterial and mitochondrial

membranes. J. Lipid Res. 49, 1607–1620 (2007). 69. Schlame, M., Brody, S. & Hostetler, K. Y. Mitochondrial cardiolipin in diverse eukaryotes: Comparison of biosynthetic reactions and molecular acyl species. Eur. J. Biochem. 212, 727–733 (1993). 70. Hiraoka, S., Matsuzaki, H. & Shibuya, I. Active increase in cardiolipin synthesis in the stationary growth phase and its physiological significance in Escherichia coli. FEBS Lett. 336, 221–224 (1993). 71. Romantsov, T., Guan, Z. & Wood, J. M. Cardiolipin and the osmotic stress responses of bacteria. Biochim. Biophys. Acta 1788, 2092–2100 (2009). 72. Marr, A. G. & Ingraham, J. L. Effect of temperature on the composition of fatty acids in Escherichia coli. J Bacteriol 84, 1260–7 (1962). 73. Morein, S., Andersson, A. S., Rilfors, L. & Lindblom, G. Wild-type Escherichia coli cells regulate the membrane lipid composition in a ‘window’ between gel and non-lamellar structures. J. Biol. Chem. 271, 6801–6809 (1996). 74. McGarrity, J. T. & Armstrong, J. B. The effect of temperature and other growth conditions on the fatty acid composition of Escherichia coli. Can. J. Microbiol. 27, 835–840 (1981). 75. McGarrity, J. T. & Armstrong, J. B. The effect of salt on phospholipid fatty acid composition in Escherichia coli K-12. Biochim. Biophys. Acta (BBA)/Lipids Lipid Metab. 398, 258–264 (1975). 76. Knivett, V. A. & Cullen, J. Some factors affecting cyclopropane acid formation in Escherichia coli. Biochem. J. 96, 771–776 (1965). 77. Kito, M., Aibara, S., Kato, M. & Hata, T. Differences in fatty acid composition among phosphatidylethanolamine, phosphatidylglycerol and cardiolipin of Escherichia coli. Biochim. Biophys. Acta 260, 475–478 (1972). 78. Lee, H. & Malmstadt, N. Effect of low levels of lipid oxidation on the curvature, dynamics, and permeability of lipid bilayers and their interactions with cationic nanoparticles. J. Phys. D. Appl. Phys. 51, (2018). 79. Shahidi, F. & Zhong, Y. Lipid oxidation and improving the oxidative stability. Chem. Soc. Rev. 39, 4067–4079 (2010). 80. Vieira, S. A., Zhang, G. & Decker, E. A. Biological implications of lipid oxidation products. JAOCS, J. Am. Oil Chem. Soc. 94, 339–351 (2017). 81. Vähäheikkilä, M. et al. How cardiolipin peroxidation alters the properties of the inner mitochondrial membrane? Chem. Phys. Lipids 214, 15–23 (2018). 82. Gadsby, D. C. Ion channels vs ion pumps: the principal difference, in principle. Nat Rev Mol Cell Biol 10, 344–352 (2009). 83. Apell, H. J. How do P-Type ATPases transport ions? Bioelectrochemistry 63, 149–156 (2004). 84. Bublitz, M., Poulsen, H., Morth, J. P. & Nissen, P. In and out of the cation pumps: P-Type ATPase structure revisited. Curr. Opin. Struct. Biol. 20, 431–439 (2010). 85. Palmgren, M. G. Plant Plasma Membrane H + -ATPases: Powerhouses for Nutrient

Uptake. Annu. Rev. Plant Physiol. Plant Mol. Biol. 52, 817–845 (2002). 86. Axelsen, K. B. & Palmgren, M. G. Evolution of substrate specificities in the P-Type ATPase superfamily. J Mol Evol 46, 84–101 (1998). 87. Palmgren, M. G. & Nissen, P. P-Type ATPases. Annu. Rev. Biophys. 40, 243–266 (2011). 88. Kühlbrandt, W. Biology, structure and mechanism of P-type ATPases. Nat. Rev. Mol. Cell Biol. 5, 282–295 (2004). 89. Bublitz, M., Morth, J. P. & Nissen, P. P-type ATPases at a glance. J. Cell Sci. 124, 2515– 2519 (2011). 90. Daiho, T. et al. Deletions of any single residues in Glu40-Ser48 loop connecting A domain and the first transmembrane helix of sarcoplasmic reticulum Ca2+-ATPase result in almost complete inhibition of conformational transition and hydrolysis of phosphoenzyme intermediate. J. Biol. Chem. 278, 39197–39204 (2003). 91. Lenoir, G. et al. Functional properties of Sarcoplasmic Reticulum Ca2+-ATPase after proteolytic cleavage at Leu119-LYS120, close to the A-domain. J. Biol. Chem. 279, 9156– 9166 (2004). 92. Corradi, V. et al. Emerging Diversity in Lipid-Protein Interactions. Chem. Rev. 119, 5775– 5848 (2019). 93. Contreras, F., Ernst, A. M., Wieland, F. & Brügger, B. Specificity of Intramembrane Protein − Lipid Interactions. Cold Spring Harb Perspect Biol 3, (2013). 94. Lee, A. G. How lipids affect the activities of integral membrane proteins. Biochim. Biophys. Acta - Biomembr. 1666, 62–87 (2004). 95. Niemela, P. S. et al. Membrane Proteins Diffuse as Dynamic Complexes with Lipids. JACS Commun. 132, 7574–7575 (2010). 96. Gustavsson, M., Traaseth, N. J. & Veglia, G. Activating and deactivating roles of lipid bilayers on the Ca2+-ATPase/phospholamban complex. Biochemistry 50, 10367–10374 (2011). 97. Johannsson, A. et al. The effect of bilayer thickness and n-alkanes on the activity of the (Ca2+ + Mg2+)-dependent ATPase of sarcoplasmic reticulum. J. Biol. Chem. 256, 1643– 1650 (1981). 98. Sonntag, Y. et al. Mutual adaptation of a membrane protein and its lipid bilayer during conformational changes. Nat. Commun. 2, (2011). 99. Norimatsu, Y., Hasegawa, K., Shimizu, N. & Toyoshima, C. Protein-phospholipid interplay revealed with crystals of a calcium pump. Nature 545, 193–198 (2017). 100. Pignataro, M. F. et al. Modulation of plasma membrane Ca2+-ATPase by neutral phospholipids. J. Biol. Chem. 290, 6179–6190 (2015). 101. Autzen, H. E. et al. Interactions of a Bacterial Cu(I)-ATPase with a Complex Lipid Environment. Biochemistry 57, 4063–4073 (2018). 102. Zimmer, J. & Doyle, D. A. Phospholipid requirement and pH optimum for the in vitro enzymatic activity of the E. coli P-type ATPase ZntA. Biochim. Biophys. Acta 1758, 645– 652 (2006).

103. Palsdottir, H. & Hunte, C. Lipids in membrane protein structures. Biochim. Biophys. Acta 1666, 2–18 (2004). 104. Pyle, E. et al. Structural lipids enable the formation of functional oligomers of the eukaryotic purine symporter UapA. Cell Chem. Biol. 25, (2018). 105. Gupta, K. et al. The role of interfacial lipids in stabilizing membrane protein oligomers. Nature 541, 421–424 (2017). 106. Habeck, M., Kapri-Pardes, E., Sharon, M. & Karlish, S. J. D. Specific phospholipid binding to Na,K-ATPase at two distinct sites. Proc. Natl. Acad. Sci. 114, 2904–2909 (2017). 107. Von Heijne, G. Membrane-protein topology. Nat. Rev. Mol. Cell Biol. 7, 909–918 (2006). 108. Van Klompenburg, W., Nilsson, I., Von Heijne, G. & De Kruijff, B. Anionic phospholipids are determinants of membrane protein topology. EMBO J. 16, 4261–4266 (1997). 109. Wang, X., Bogdanov, M. & Dowhan, W. Topology of polytopic membrane protein subdomains is dictated by membrane phospholipid composition. EMBO J. 21, 5673–5681 (2002). 110. Zhang, W., Campbell, H. A., King, S. C. & Dowhan, W. Phospholipids as determinants of membrane protein topology. J. Biol. Chem. 280, 26032–26038 (2005). 111. Ito, K., Sato, T. & Yura, T. Synthesis and assembly of the membrane proteins in E. coli. Cell 11, 551–559 (1977). 112. Romantsov, T., Battle, A. R., Hendel, J. L., Martinac, B. & Wood, J. M. Protein localization in Escherichia coli cells: Comparison of the cytoplasmic membrane proteins ProP, LacY, ProW, AqpZ, MscS, and MscL. J. Bacteriol. 192, 912–924 (2010). 113. Romantsov, T. et al. Cardiolipin promotes polar localization of osmosensory transporter ProP in Escherichia coli. Mol. Microbiol. 64, 1455–1465 (2007). 114. Stangl, M. & Schneider, D. Functional competition within a membrane: Lipid recognition vs. transmembrane helix oligomerization. Biochim. Biophys. Acta - Biomembr. 1848, 1886– 1896 (2015). 115. Groisman, E. A. et al. Bacterial Mg2+ Homeostasis, Transport, and Virulence. Annu. Rev. Genet. 625–646 (2013). 116. Cunrath, O. & Bumann, D. Host resistance factor SLC11A1 restricts Salmonella growth through magnesium deprivation. Science (80-. ). 366, 995–999 (2019). 117. Maguire, M. E. Magnesium Transporters: Properties, Regulation and Structure. Front Biosci. 3149–3163 (2006). 118. Snavely, M. D., Florer, J. B., Miller, C. G. & Maguire, M. E. Magnesium Transport in Salmonella typhimurium : 28Mg2+ Transport by the CorA, MgtA, and MgtB Systems. J. Bacteriol. 171, 4761–4766 (1989). 119. Snavely, M. D., Gravin, S. A., Cheung, T. T., Miller, C. G. & Maguire, M. E. Magnesium transport in Salmonella typhimurium: Regulation of mgtA and mgtB expression. J. Biol. Chem. 266, 824–829 (1991). 120. Beth, M. & Michael, C. M. Magnesium transport in prokaryotes. 523–527 (1999).

121. Jephthah, S., Månsson, L. K., Belic, D., Morth, J. P. & Skepö, M. Physicochemical Characterisation of KEIF — The Intrinsically Disordered N-Terminal Region of Magnesium Transporter A. Biomolecules 1–22 (2020). 122. Snavely, M. D., Florer, J. B., Miller, C. G. & Maguire, M. E. Magnesium Transport in Salmonella typhimurium : Expression of Cloned Genes for Three Distinct Mg2+ Transport Systems. J. Bacteriol. 171, 4752–4760 (1989). 123. Maguire, M. E. MgtA and MgtB: Prokaryotic P-type ATPases that mediate Mg2+ influx. J. Bioenerg. Biomembr. 24, 319–328 (1992). 124. Subramani, S., Perdreau-Dahl, H. & Morth, J. P. The magnesium transporter A is activated by cardiolipin and is highly sensitive to free magnesium in vitro. Elife 5, 1–17 (2016). 125. Håkansson, K. O. The structure of Mg-ATPase nucleotide-binding domain at 1.6 Å reveals a unique ATP-binding motif. Acta Crystallogr. Sect. D Biol. Crystallogr. 65, 1181–1186 (2009). 126. Tao, T., Grulich, P. F., Kucharski, L. M., Smith, R. L. & Maguire, M. E. Magnesium transport in Salmonella typhimurium: Biphasic magnesium and time dependence of the transcription of the mgtA and mgtCB loci. Microbiology 144, 655–664 (1998). 127. Alix, E. & Blanc-Potard, A. B. MgtC: a key player in intramacrophage survival. Trends Microbiol. 15, 252–256 (2007). 128. Blanc-Potard, A. B. & Groisman, E. A. The Salmonella selC locus contains a pathogenicity island mediating intramacrophage survival. EMBO J. 16, 5376–5385 (1997). 129. Hurwitz, C. & Rosano, C. L. The Intracellular Magnesium Ions Concentration of Bound and Unbound Magnesium Ions in Escherichia coli. J. Biol. Chem. 242, 3719–3722 (1967). 130. Wilks, J. C. & Slonczewski, J. L. pH of the cytoplasm and periplasm of Escherichia coli: Rapid measurement by green fluorescent protein fluorimetry. J. Bacteriol. 189, 5601–5607 (2007). 131. Neef, J., Andisi, V. F., Kim, K. S., Kuipers, O. P. & Bijlsma, J. J. E. Deletion of a cation transporter promotes lysis in Streptococcus pneumoniae. Infect. Immun. 79, 2314–2323 (2011). 132. Lin, Z. et al. Virulence and Stress Responses of Shigella flexneri Regulated by PhoP/PhoQ. Front. Microbiol. 8, 1–22 (2018). 133. Tao, T., Snavely, M. D., Farr, S. G. & Maguire, M. E. Magnesium transport in Salmonella typhimurium: mgtA Encodes a P-type ATPase and is regulated by Mg2+ in a manner similar to that of the mgtB P- type ATPase. J. Bacteriol. 177, 2654–2662 (1995). 134. Faraco, M. et al. Hyperacidification of vacuoles by the combined action of two different P- ATPases in the tonoplast determines flower color. Cell Rep. 6, 32–43 (2014). 135. Véscovi, E. G., Soncini, F. C. & Groisman, E. A. Mg2+ as an extracellular signal: Environmental regulation of Salmonella virulence. Cell 84, 165–174 (1996). 136. Chamnongpol, S., Cromie, M. & Groisman, E. A. Mg2+ sensing by the Mg2+ sensor PhoQ of Salmonella enterica. J. Mol. Biol. 325, 795–807 (2003). 137. Foster, J. W. & Hall, H. K. Adaptive acidification tolerance response of Salmonella typhimurium. J. Bacteriol. 172, 771–778 (1990).

138. Alpuche Aranda, C. M., Swanson, J. A., Loomis, W. P. & Miller, S. I. Salmonella typhimurium activates virulence gene transcription within acidified macrophage phagosomes. Proc. Natl. Acad. Sci. U. S. A. 89, 10079–10083 (1992). 139. Bader, M. W. et al. Recognition of antimicrobial peptides by a bacterial sensor kinase. Cell 122, 461–472 (2005). 140. Prost, L. R. et al. Activation of the bacterial sensor kinase PhoQ by acidic pH. Mol. Cell 26, 165–174 (2007). 141. Miller, S. I., Kukral, A. M. & Mekalanos, J. J. A two-component regulatory system (phoP phoQ) controls Salmonella typhimurium virulence. Proc. Natl. Acad. Sci. U. S. A. 86, 5054– 5058 (1989). 142. Cromie, M. J., Shi, Y., Latifi, T. & Groisman, E. A. An RNA Sensor for Intracellular Mg2+. Cell 125, 71–84 (2006). 143. Hollands, K. et al. Riboswitch control of Rho-dependent transcription termination. Proc. Natl. Acad. Sci. U. S. A. 109, 5376–5381 (2012). 144. Park, S. Y., Cromie, M. J., Lee, E. J. & Groisman, E. A. A bacterial mRNA leader that employs different mechanisms to sense disparate intracellular signals. Cell 142, 737–748 (2010). 145. Gall, A. R. et al. Mg 2+ regulates transcription of mgtA in Salmonella Typhimurium via translation of proline codons during synthesis of the MgtL peptide. Proc. Natl. Acad. Sci. 113, 15096–15101 (2016). 146. Zhao, G., Kong, W., Weatherspoon-Griffin, N., Clark-Curtiss, J. & Shi, Y. Mg2+ facilitates leader peptide translation to induce riboswitch-mediated transcription termination. EMBO J. 30, 1485–1496 (2011). 147. Chadani, Y. et al. Intrinsic Ribosome Destabilization Underlies Translation and Provides an Organism with a Strategy of Environmental Sensing. Mol. Cell 68, 528–539 (2017). 148. Choi, E. et al. Elongation factor P controls translation of the mgtA gene encoding a Mg2+ transporter during Salmonella infection. Microbiologyopen 8, (2018). 149. Ramamurthi, K. S. & Storz, G. The small protein floodgates are opening; now the functional analysis begins. BMC Biol. 12, (2014). 150. Choi, E., Lee, K. Y. & Shin, D. The MgtR regulatory peptide negatively controls expression of the MgtA Mg2+ transporter in Salmonella enterica serovar Typhimurium. Biochem. Biophys. Res. Commun. 417, 318–323 (2012). 151. Alix, E. & Blanc-Potard, A. B. Peptide-assisted degradation of the Salmonella MgtC virulence factor. EMBO J. 27, 546–557 (2008). 152. Wang, H. et al. Increasing intracellular magnesium levels with the 31-amino acid MgtS protein. Proc. Natl. Acad. Sci. 114, 5689–5694 (2017). 153. Pontes, M. H., Yeom, J. & Groisman, E. A. Reducing ribosome biosynthesis promotes translation during low Mg2+ stress. Mol. Cell 64, 480–492 (2016). 154. Park, M., Nam, D., Kweon, D. H. & Shin, D. ATP reduction by MgtC and Mg2+ homeostasis by MgtA and MgtB enables Salmonella to accumulate RpoS upon low cytoplasmic Mg2+ stress. Mol. Microbiol. 110, 283–295 (2018).

155. O’Connor, K., Fletcher, S. A. & Csonka, L. N. Increased expression of Mg2+ transport proteins enhances the survival of Salmonella enterica at high temperature. Proc. Natl. Acad. Sci. 106, 17522–17527 (2009). 156. Bourret, T. J., Liu, L., Shaw, J. A., Husain, M. & Vázquez-Torres, A. Magnesium homeostasis protects Salmonella against nitrooxidative stress. Sci. Rep. 7, 15083 (2017). 157. Mir, R. A. & Kudva, I. T. AntibioticǦresistant Shiga toxinǦproducing Escherichia coli : An overview of prevalence and intervention strategies. Zoonoses Public Heal. 66, 1–13 (2019). 158. Erb, A., Stürmer, T., Marre, R. & Brenner, H. Prevalence of antibiotic resistance in Escherichia coli: Overview of geographical, temporal, and methodological variations. Eur J Clin Microbiol Infect Dis 26, 83–90 (2007). 159. Nesta, B. et al. FdeC, a Novel Broadly Conserved Escherichia coli Adhesin Eliciting Protection against Urinary Tract Infections. MBio 3, (2012). 160. Kelly, G. et al. Structure of the cell-adhesion fragment of intimin from enteropathogenic Escherichia coli. Nat. Struct. Biol. 6, 313–318 (1999). 161. Renn, J. P., Junker, M., Besingi, R. N., Braselmann, E. & Clark, P. L. ATP-independent control of autotransporter virulence protein transport via the folding properties of the secreted protein. Chem. Biol. 19, 287–296 (2012). 162. Oliver, D. C., Huang, G., Nodel, E., Pleasance, S. & Fernandez, R. C. A conserved region within the Bordetella pertussis autotransporter BrkA is necessary for folding of its passenger domain. Mol. Microbiol. 47, 1367–1383 (2003). 163. Peterson, J. H., Tian, P., Ieva, R., Dautin, N. & Bernstein, H. D. Secretion of a bacterial virulence factor is driven by the folding of a C-terminal segment. Proc. Natl. Acad. Sci. U. S. A. 107, 17739–17744 (2010). 164. Soprova, Z. et al. A conserved aromatic residue in the autochaperone domain of the autotransporter Hbp is critical for initiation of outer membrane translocation. J. Biol. Chem. 285, 38224–38233 (2010). 165. Selkrig, J. et al. Discovery of an archetypal protein transport system in bacterial outer membranes. Nat. Struct. Mol. Biol. 19, 506–510 (2012). 166. Brasch, J., Harrison, O. J., Honig, B. & Shapiro, L. Thinking outside the cell: How cadherins drive adhesion. Trends Cell Biol. 22, 299–310 (2012). 167. Nagar, B., Overduin, M., Ikura, M. & Rini, J. M. Structural basis of calcium-induced E- cadherin rigidification and dimerization. Nature 380, 360–364 (1996). 168. Thøgersen, L. & Nissen, P. Flexible P-type ATPases interacting with the membrane. Curr. Opin. Struct. Biol. 22, 491–499 (2012). 169. Moore, B. M., Lentz, B. R. & Meissner, G. Effects of Sarcoplasmic Reticulum Ca2+-ATPase on phospholipid bilayer fluidity: Boundary lipid. Biochemistry 17, 5248–5255 (1978). 170. Rainier, S. et al. Phase transition characteristics of diphosphatidylglycerol (cardiolipin) and stereoisomeric phosphatidyldiacylglycerol bilayers. Biochim Biophys Acta 558, 187–198 (1979). 171. Starling, A. P., East, J. M. & Lee, A. G. Effects of phosphatidylcholine fatty acyl chain length

on calcium binding and other functions of the (Ca2+−Mg2+)-ATPase. Biochemistry 32, 1593–1600 (1993). 172. Vail, W. J. & Stollery, J. G. Phase changes of cardiolipin vesicles mediated by divalent cations. Biochim Biophys Acta . 551, 74–84 (1979). 173. Sankaram, M. B., Powell, G. L. & Marsh, D. Effect of acyl chain composition on salt-induced lamellar to inverted hexagonal phase transitions in cardiolipin. Biochim Biophys Acta 980, 389–392 (1989). 174. Gandhi, A. & Shah, N. P. Effect of salt stress on morphology and membrane composition of Lactobacillus acidophilus, Lactobacillus casei, and Bifidobacterium bifidum, and their adhesion to human intestinal epithelial-like Caco-2 cells. J. Dairy Sci. 99, 2594–2605 (2016). 175. Neet, K. E. Chapter 7 - Cooperativity in Enzyme Function: Equilibrium and Kinetic Aspects. Methods in Enzymology vol. 64 (1980). 176. Motlagh, H. N., Wrabl, J. O., Li, J. & Hilser, V. J. The ensemble nature of allostery. Nature 508, 331–339 (2014). 177. Edelstein, S. J. & Le Novère, N. Cooperativity of allosteric receptors. J. Mol. Biol. 425, 1424–1432 (2013). 178. Ghosh, M. et al. Lipid-Mediated Regulation of Embedded Receptor Kinases via Parallel Allosteric Relays. Biophys. J. 112, 643–654 (2017). 179. Cong, X., Liu, Y., Liu, W., Liang, X. & Laganowsky, A. Allosteric modulation of protein- protein interactions by individual lipid binding events. Nat. Commun. 8, (2017). 180. Patrick, J. W. et al. Allostery revealed within lipid binding events to membrane proteins. Proc. Natl. Acad. Sci. U. S. A. 115, 2976–2981 (2018). 181. Hilbers, F. et al. Tuning of the Na,K-ATPase by the beta subunit. Sci. Rep. 6, (2016). 182. Toustrup-Jensen, M. S. et al. The C terminus of Na +, K + -ATPase controls Na + affinity on both sides of the membrane through Arg 935. J. Biol. Chem. 284, 18715–18725 (2009). 183. Poulsen, H. et al. Neurological disease mutations compromise a γ-terminal ion pathway in the Na+/K+-ATPase. Nature 467, 99–102 (2010). 184. Ekberg, K., Palmgren, M. G., Veierskov, B. & Buch-Pedersen, M. J. A novel mechanism of P-type ATPase autoinhibition involving both termini of the protein. J. Biol. Chem. 285, 7344–7350 (2010). 185. Regenberg, B., Villalba, J. M., Lanfermeijer, F. C. & Palmgren, M. G. C-terminal deletion analysis of plant plasma membrane H+-ATPase: Yeast as a model system for solute transport across the plant plasma membrane. Plant Cell 7, 1655–1666 (1995). 186. Sahoo, S. K., Shaikh, S. A., Sopariwala, D. H., Bal, N. C. & Periasamy, M. Sarcolipin protein interaction with Sarco(endo)plasmic reticulum Ca2+ ATPase (SERCA) is distinct from phospholamban protein,and only sarcolipin can promote uncoupling of the serca pump. J. Biol. Chem. 288, 6881–6889 (2013). 187. Dutra, M. B., Ambesi, A. & Slayman, C. W. Structure-function relationships in membrane segment 5 of the yeast plasma membrane Pma1 H+-ATPase. J. Biol. Chem. 273, 17411– 17417 (1998).

188. Lemmin, T., Bovigny, C., Lançon, D. & Dal Peraro, M. Cardiolipin models for molecular simulations of bacterial and mitochondrial membranes. J. Chem. Theory Comput. 9, 670– 678 (2013). 189. Reis, A. & Spickett, C. M. Chemistry of phospholipid oxidation. Biochim. Biophys. Acta - Biomembr. 1818, 2374–2387 (2012). 190. Seddon, A. M., Curnow, P. & Booth, P. J. Membrane proteins, lipids and detergents: Not just a soap opera. Biochim Biophys Acta 1666, 105–117 (2004). 191. Levi, V., Rossi, J. P. F. C., Echarte, M. M., Castello, P. R. & Gonza Flecha, F. L. Thermal stability of the plasma membrane calcium pump. Quantitative analysis of its dependence on lipid-protein interactions. J Membr Biol 173, 215–225 (2000). 192. Wen, J., Lord, H., Knutson, N. & Wikström, M. Nano differential scanning fluorimetry for comparability studies of therapeutic proteins. Anal. Biochem. 593, 113581 (2020). 193. Oliver, P. M. et al. Localization of anionic phospholipids in Escherichia coli cells. J. Bacteriol. 196, 3386–3398 (2014). 194. Giles, D. K., Hankins, J. V., Guan, Z. & Trent, M. S. Remodelling of the Vibrio cholerae membrane by incorporation of exogenous fatty acids from host and aquatic environments. Mol. Microbiol. 79, 716–728 (2011). 195. Strahl, H., Bürmann, F. & Hamoen, L. W. The actin homologue MreB organizes the bacterial cell membrane. Nat. Commun. 5, 3442 (2014). 196. Sciara, M. I., Spagnuolo, C., Jares-Erijman, E. & García Véscovi, E. Cytolocalization of the PhoP response regulator in Salmonella enterica: Modulation by extracellular Mg2+ and by the SCV environment. Mol. Microbiol. 70, 479–493 (2008). 197. Khalifat, N., Fournier, J. B., Angelova, M. I. & Puff, N. Lipid packing variations induced by pH in cardiolipin-containing bilayers: The driving force for the cristae-like shape instability. Biochim. Biophys. Acta - Biomembr. 1808, 2724–2733 (2011). 198. Konijnenberg, A., Van Dyck, J. F., Kailing, L. L. & Sobott, F. Extending native mass spectrometry approaches to integral membrane proteins. Biol. Chem. 396, 991–1002 (2015). 199. Van Dyck, J. F., Konijnenberg, A. & Sobott, F. Native Mass Spectrometry for the Characterization of Structure and Interactions of Membrane Proteins. in 205–232 (2017). doi:10.1007/978-1-4939-7151-0_11. 200. Cohen, E. et al. Purification of Na+,K+-ATPase expressed in Pichia pastoris reveals an essential role of phospholipid-protein interactions. J. Biol. Chem. 280, 16610–16618 (2005). 201. Planas-Iglesias, J. et al. Cardiolipin interactions with proteins. Biophys. J. 109, 1282–1294 (2015). 202. Stansfeld, P. J. et al. MemProtMD: Automated insertion of membrane protein structures into explicit lipid membranes. Structure 23, 1350–1361 (2015). 203. Hossain, K. R. & Clarke, R. C. General and specific interactions of the phospholipid bilayer with P-type ATPases. Biophys. Rev. 11, 353–364 (2019). 204. Missiaen, L., Wuytack, F., De Smedt, H., Vrolix, M. & Casteels, R. AlF4- reversibly inhibits

’P’-type cation-transport ATPases, possibly by interacting with the phosphate-binding site of the ATPase. Biochem. J. 253, 827–33 (1988). 205. Danko, S., Yamasaki, K., Daiho, T. & Suzuki, H. Distinct natures of beryllium fluoride- bound, aluminum fluoride-bound, and magnesium fluoride-bound stable analogues of an ADP-insensitive phosphoenzyme intermediate of sarcoplasmic reticulum Ca2+-ATPase: Changes in catalytic and transport sites during ph. J. Biol. Chem. 279, 14991–14998 (2004). 206. Habeck, M., Cirri, E., Katz, A., Karlish, S. J. & Apell, H. J. Investigation of electrogenic partial reactions in detergent-solubilized Na,K-ATPase. Biochemistry 48, 9147–9155 (2009). 207. Lee, E. J. & Groisman, E. A. Control of a Salmonella virulence locus by an ATP-sensing leader messenger RNA. Nature 486, 271–275 (2012). 208. Timcenko, M. et al. Structure and autoregulation of a P4-ATPase lipid flippase. Nature 571, 366–370 (2019). 209. Lee, D. D. et al. Magnesium Flux Modulates Ribosomes to Increase Bacterial Survival. Cell 177, 352–360 (2019). 210. Huang, C. S., Pedersen, B. P. & Stokes, D. L. Crystal structure of the potassium-importing KdpFABC membrane complex. Nature 546, 681–685 (2017). 211. Toyoshima, C., Nakasako, M., Nomura, H. & Ogawa, H. Structure determination of the calcium pump of sarcoplasmic reticulum at 2.6 Å resolution. Nature 405, 647–55 (2001). 212. Di Leva, F., Domi, T., Fedrizzi, L., Lim, D. & Carafoli, E. The plasma membrane Ca2+ ATPase of animal cells: Structure, function and regulation. Arch. Biochem. Biophys. 476, 65–74 (2008). 213. Morth, J. P. et al. Crystal structure of the sodium-potassium pump. Nature 450, 1043–1049 (2007). 214. Abe, K., Irie, K., Nakanishi, H., Suzuki, H. & Fujiyoshi, Y. Crystal structures of the gastric proton pump. Nature 556, 214–229 (2018). 215. Rodríguez-Navarro, A. & Benito, B. Sodium or potassium efflux ATPase. A fungal, bryophyte, and protozoal ATPase. Biochim Biophys Acta 1798, 1841–1853 (2010). 216. Sørensen, D. M., Buch-Pedersen, M. J. & Palmgren, M. G. Structural divergence between the two subgroups of P5 ATPases. Biochim Biophys Acta 1797, 846–855 (2010). 217. Subramani, S. & Morth, J. Heterologous expression and purification of the magnesium transporter A (MgtA) in Escherichia coli. Bio-Protocol 6, 1–9 (2016). 218. Edgar, R. C. MUSCLE: A multiple sequence alignment method with reduced time and space complexity. BMC Bioinformatics 5, 1–19 (2004). 219. Gassel, M., Mo, T., Puppe, W. & Altendorf, K. The KdpF Subunit Is Part of the K+- translocating Kdp Complex of Escherichia coli and Is Responsible for Stabilization of the Complex in Vitro. J. Biol. Chem. 274, 37901–37907 (1999).

ANNEX: PAPER I-III

1 The extracellular juncture domains in the intimin passenger adopt a

2 constitutively extended conformation inducing restraints to its sphere

3 of action

4 Julia Weikum,1,2 Alina Kulakova,3 Giulio Tesei,4 Shogo Yoshimoto,5 Line Vejby Jægerum,1 Monika 5 Schütz,6 Katsutoshi Hori,5 Marie Skepö,7 Pernille Harris,3 Jack C. Leo,8,9* and J. Preben 6 Morth1,2,10*

7 1 Membrane Transport Group, Centre for Molecular Medicine Norway (NCMM), Nordic EMBL 8 Partnership, University of Oslo, P.O. Box 1137 Blindern, 0318 Oslo, Norway 9 2 Enzyme and Protein Chemistry, Section for Protein Chemistry and Enzyme Technology, 10 Department of Biotechnology and Biomedicine, Technical University of Denmark, Søltofts Plads, 11 2800, Kgs. Lyngby, Denmark 12 3 Department of Chemistry, Technical University of Denmark, Kemitorvet building 207, 2800 Kgs. 13 Lyngby, Denmark 14 4 Structural Biology and NMR Laboratory, Linderstrøm-Lang Centre for Protein Science, 15 Department of Biology, University of Copenhagen, Ole Maaloes Vej 5, Copenhagen 2200, 16 Denmark 17 5 Department of Biomolecular Engineering, Graduate School of Engineering, Nagoya University, 18 Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan 19 6 Interfaculty Institute for Microbiology and Infection Medicine, University Clinics Tübingen, 72076 20 Tübingen, Germany 21 7 Division of Theoretical Chemistry, Department of Chemistry, Lund University, 221 00 Lund, 22 Sweden 23 8 Department of Biosciences, University of Oslo, P.O. Box 1137 Blindern, 0318 Oslo, Norway. 24 9 Department of Biosciences, Nottingham Trent University, Nottingham NG11 8NS, UK 25 10 Institute for Experimental Medical Research (IEMR), Oslo University Hospital, Ullevål PB 4956 26 Nydalen, NO-0424 Oslo, Norway. 27 *Correspondence: [email protected], [email protected] 28 Short title: Structural and in silico characterization of the intimin passenger 29

32 Graphical Abstract

33 34 Abstract

35 Enterohemorrhagic and enteropathogenic Escherichia coli are among the most important food- 36 borne pathogens, posing a global health threat. The virulence factor intimin is essential for 37 attachment of pathogenic E. coli to the intestinal host cells. Intimin consists of four extracellular 38 bacterial immunoglobulin-like (Big) domains, D00-D2, extending into the fifth lectin-like 39 subdomain (D3) that binds to the Tir-receptor on the host cell. Here, we present the crystal 40 structures of the elusive D00-D0 domains at 1.5 Å and D0-D1 at 1.8 Å resolution, which confirm 41 that the passenger of intimin has five distinct domains. We describe that D00-D0 exhibits a higher 42 degree of rigidity and D00 likely functions as a juncture domain at the outer membrane- 43 extracellular medium interface. We conclude that D00 is a unique Big domain with a specific 44 topology likely found in a broad range of other inverse autotransporters. The accumulated data 45 allows us to model the complete passenger of intimin and propose functionality to the Big 46 domains, D00-D0-D1, extending directly from the membrane. 47 48 Keywords: Intimin; Invasin; Inverse Autotransporter; X-ray Crystallography; SAXS; MD 49 simulations

50 Introduction

51 The gastrointestinal pathogens, enterohemorrhagic Escherichia coli (EHEC) and 52 enteropathogenic Escherichia coli (EPEC), pose a serious public health threat 1. EHEC is a food- 53 borne human pathogen leading to both sporadic infections and epidemics, in particular outbreaks 54 of hemorrhagic colitis (bloody diarrhea) and hemolytic uremic syndrome 23. The largest outbreak 55 occurred in Germany 2011 with 4000 cases of EHEC gastroenteritis and more than 850 reported 56 cases of hemolytic uremic syndrome, leading to the death of 54 people 3–5. EPEC is a major 57 contributor to diarrhea in low income countries, being a leading cause of child mortality through 58 diarrhea 6. EHEC and EPEC are highly related and share many virulence determinants and 59 features, many of which are encoded on a pathogenicity island named the locus of enterocyte 60 effacement (LEE) 7,8. Among them, an essential virulence factor for adherence to the host cells is 61 the outer membrane protein intimin, the gene product of the eaeA locus 9. In addition to EHEC 62 and EPEC, intimin variants are found in several other attaching and effacing pathogens, such as 63 Citrobacter rodentium and Hafnia alvei 10.

64 Intimin is a 94 kDa outer membrane protein, essential for the intimate attachment of 65 bacterial cells to the host cell surface, followed by actin pedestal formation 8. Prior to intimin 66 binding, the translocated intimin receptor (Tir) is injected into the host cell by the bacterial type 3 67 secretion system and subsequently inserted into the host plasma membrane to promote the 68 intimate attachment 11,12. Intimin is classified as an inverse autotransporter, a subclass of the type 69 5 secretion system that transports the C-terminal extracellular region or passenger through the 70 lumen of the N-terminal E-barrel located in the outer membrane 13. In the following, we will refer 71 to the extracellular region only as the passenger, as recommended by Drobnak et al. (2015) 14. 72 The secretion of the passenger proceeds through a hairpin-like intermediate, i.e. the membrane 73 proximal (N-terminal) part of the passenger is exported first, followed by the rest of the passenger 74 with the C-terminus reaching the surface last 15. Sequential folding of the individual subdomains 75 is the main driving force for the secretion process 16.

76 The intimin passenger forms a rod-like extension consisting of four tandem bacterial 77 immunoglobulin-like domains (Big) (subdomains D00-D0-D1-D2) capped by a C-terminal C-type 78 lectin-like domain (D3). This architecture is similar to other inverse autotransporters, such as 79 Yersinia pseudotuberculosis invasin 17. The C-type lectin-like domain (D3) at the C-terminus forms 80 a superdomain with D2, which combined create the functional Tir-binding region, described earlier 81 in complex with the binding domain of Tir (PDB: 1F02) 18. Although bioinformatic and biophysical

82 evidence suggested that subdomains D00 and D0 are also Ig-like domains 16,18, no structural 83 information has been available for these domains. Especially the structure of D00 and its 84 connection to the neighboring subdomains has been discussed as D00 is located at the juncture 85 between the membrane-bound β-barrel and the remaining extracellular passenger 86 subdomains 19. D00 represents the first subdomain upon excretion, and it was hypothesized to 87 represent the driving force for extracting the remaining Big domains in the intimin passenger 16,19.

88 Here, we describe the crystal structures of the combined intimin domains D00-D0 and D0- 89 D1 from the EPEC strain E2348/69 determined at 1.5 Å and 1.8 Å resolution, respectively. These 90 structures represent the last missing structural information for the complete passenger of intimin 91 from the EPEC strain E2348/69. We show that the D00 structure represents a Big domain. 92 However, in comparison with D0 and D1 it exhibits higher structural resemblance to the general 93 Ig folds and remarkable topological resemblance with the mammalian cadherins. The D0 domain 94 shows, as expected, high structural similarity with the D1 intimin domain. The crystal structures 95 in combination with small angle X-ray scattering (SAXS), and in silico simulation data, suggest 96 that D00-D0 adopts a more permanently extended and rigid conformation, likely stabilized by a 97 short connector region between the D00 and D0 domains. The D0-D1 construct adopts a largely 98 bent conformation and is likely more dynamic as predicted by simulation studies. The structural 99 integrity of intimin is independent of calcium, and thus differs from specific adhesins and cadherins 100 that commonly are regulated and rigidified upon calcium binding 20,21. We hypothesize that the 101 constitutively extended conformation of D00-D0 will increase the radius of probable interaction 102 between the passenger and the bacterial Tir receptor on the host cell plasma membrane, 103 positioning the Tir-binding region further away from the β-barrel translocator domain. This would 104 allow a faster binding to Tir following expression of intimin on the bacterial surface. 105

106 Results

107 Crystal structure of the E. coli intimin subdomains D00-D0 exhibits an elongated 108 conformation while D0-D1 shows a bent structure

109 The extended and flexible structure of the full-length intimin made structure determination with X- 110 ray crystallography a challenge in the past. Intimin is divided into three distinct regions, the 111 periplasmic lysin motif (LysM) domain, the membrane-embedded β-barrel and the extracellular 112 passenger (Fig 1A). Structural information of the three C-terminal subdomains (D1-D3), which 113 include the Tir binding module of the extracellular passenger, have been determined 18, as well

114 as a crystal structure of the outer membrane-located β-barrel 19 and a solution structure of the 115 LysM domain 22. To establish a structural model for the membrane-embedded and extracellular 116 part of the receptor, we determined the structures of the yet uncharacterized subdomains D00- 117 D0 and D0-D1, which link the β-barrel to the extracellular passenger domains. Each of the 118 constructs was purified separately and characterized by X-ray crystallography. Both constructs of 119 the extracellular subdomains, D00-D0 (PDB: 6TQD) and D0-D1 (PDB: 6TPL), crystallized in 120 space group P1 with six and two molecules per asymmetric unit, respectively.

121 The structure of D00-D0 has an elongated conformation, spanning 78 Å between the N- 122 and C-terminus (Fig 1B). By contrast, the crystal structure of D0-D1 exhibits a kink between the 123 two subdomains, leading to an approximately 45° angle between the two and a distance between 124 the N- and C-terminus of only 36 Å (Fig 1C). The latter conformation likely represents one of many 125 states, as most passengers from characterized inverse autotransporters exhibit extended 126 structures 10.

127 The crystal structures of each of the subdomains, D1, D0 and D00, show an expected Ig- 128 like fold consisting of a E-sandwich of seven antiparallel β-strands. Bodelón et al. (2013) predicted 129 for EPEC D1 a fold belonging to Ig superfamily (IgSF) type-I set with seven E-strands 23. Here, 130 we show that D0 and D1 exhibit a seven-stranded strand-switched type (s-type) fold. In this fold, 131 E-strand C’ belongs to sheet II instead of sheet I, as exhibited in the classical c-type subtype 24. 132 Canonical Ig domains contain a disulfide bridge between strands B and F, which is not present in 133 any of the intimin Big subdomains 25. The subdomain D00 also exhibits an IgSF fold with a E- 134 sandwich, yet it exhibits distinct structural differences when compared to the s-type fold seen in 135 subdomain D0 and D1. In addition to the two common E-sheets, E-strands A1 and B2 form a 136 separate small E-sheet due to shortened E-strands D and E. The presence of a third E-sheet in 137 an Ig fold is not novel as it has been previously described in Ig domain subtype C4 25. The E- 138 sandwich of D00 is additionally disrupted in strand C by a small helix induced by residue Pro515. 139 Sequence analysis was performed using the profile hidden Markov model program HMMER, via 140 the website (hmmer.org). The HMMER prediction readily detects a bacterial Ig-like domain 1 141 (Big1) motif in the sequence of intimin for subdomains D0 and D1, whilst no fold is predicted for 142 the subdomain D00. Additionally, intimin D0 is structurally highly similar to D1 with a root mean 143 square deviation (RMSD) of 0.68 Å and 0.84 Å with invasin subdomain D1 while sharing little 144 structural conservation with the intimin subdomain D00 (RMSD: 4.68 Å). Yet, D0 and D1 share 145 only 38 % sequence identity, while D00 shares 50 % sequence identity with D1 and only 25 % 146 with D0. A structural similarity search with the D00 structure on PDBeFold revealed similarity to

147 the outermost extracellular domains of several Cadherin types (EC). Cadherin domains also 148 consist of a E-structured IgSF fold. Similar to D00 and in contrast to D0, EC domains also exhibit 149 a disruption of the E-sandwich by a small helix in connection with β-strand C and shortened loop 150 between strand D and E 20. 151

152 The connector region between subdomains D00 and D0 shows no effect on mechanical 153 stability

154 All extracellular subdomains are connected by short sequences, which we term ‘connectors’. The 155 connector between subdomains D00 and D0 is longer than the other connectors in the intimin 156 passenger. The D00-D0 connector adopts a S-shaped conformation connecting both domains 157 through a hydrogen bond network and solvent-free dry interface (Fig 2A). Key hydrogen bonds 158 are formed between the amine nitrogen of Q553 and the carbonyl group of S550 backbone. The 159 side chain of Q553 further contributes to the hydrogen bond network as its amide group forms a 160 hydrogen bond to the hydroxy residue of the side chain of S550 and the carbonyl group of Q553 161 interacts with the S550 backbone. The tight interaction between these two key residues leads to 162 the formation of a classical type I turn in the connector. The turn is also stabilized through 163 hydrogen bonds of the neighboring amino acids including between the carboxy group of the E466 164 side chain to the N551 and Y517 backbone. Closer to subdomain D0, the hydrogen bond network 165 continues between residues D556 and N587.

166 To further analyze the stability of the connector region between D00 and D0, atomic force 167 microscopy was performed on passenger subdomains D00-D0-D1 of wild type intimin (Int-

168 WTConnector) and a construct, in which the connector region between D00 and D0 (residues 551-

169 557) was replaced by repetitive GS-substitutions (Int-GSConnector) (Fig 2B). Both constructs 170 showed, as expected, three unfolding events with three peaks at ~250 pN, ~290 pN, and ~310 171 pN, representing the unfolding of the three Ig domains D00, D0 and D1 (Fig 2C, Table S1). These 172 values are similar to previously obtained results by Leo et al. (2016) 16 and no significant 173 differences in force peaks upon unfolding were detected between both constructs, suggesting 174 that the connector sequence does not affect the mechanical stability of D00-D0. The unfolding 175 force of Ig domains was reported to be between 130-300 pN. Intimin Big domains revealed a force 176 between 250-300 pN and are considered to be relatively rigid among Ig domains 26. 177 To test the role of the connector between D00 and D0 in a cellular environment, Int-

178 WTconnector and Int-GSConnector was expressed for adhesion assays. However, E. coli BL21 showed

179 significantly decreased expression levels of Int-GSConnector in comparison to the wild type construct

180 (Fig S1). Adhesion assays were not performed as no conclusive results could be obtained due to 181 the drastically different expression levels. Interestingly, other mutations in the E-barrel and its 182 connector region to D00 domain 13,22 as well as deletion of a E-strand in the D00 domain did not 183 lead to comparably reduced expression levels 16. 184

185 D00-D0 shows an extended conformation in solution while D0-D1 is in equilibrium between 186 a bent and an elongated conformation

187 To verify that the elongated conformation of D00-D0 was not an artefact defined by the crystal 188 lattice, we analyzed the structures of D00-D0 and D0-D1 in solution. Small-angle X-ray scattering 189 (SAXS) was performed to analyze the overall shape characteristics and flexibility of D00-D0 and 190 D0-D1. Solution structures of D00-D0 and D0-D1 indicate no intermolecular interactions or 191 aggregation as the SAXS curves show no changes with increasing concentration of the protein 192 (Fig 3A and 3B). The pair-distance distribution functions derived for D00-D0 and D0-D1 clearly

193 have different shapes, indicative of conformational dissimilarities. The maximum dimension (Dmax) 194 of D0-D1 is 6.3-8.6, while D00-D0 is around 9.6-11 nm (Table S2), an indication that a more 195 elongated conformation exists for D00-D0 in solution. To analyze the possible conformations in 196 solution, both D00-D0 and D0-D1 were modeled using Ensemble Optimization Method (EOM), 197 which is able to account for flexibility.

198 Ab initio modelling of D0-D1 indicated a globular and close-packed conformation (Fig 3C) 199 with a good data fit (χ2 = 2.9) (Fig S2A). EOM analysis of D0-D1 revealed the presence of two 200 conformational states, a bent and an elongated conformation. Data showed a good fit to the 201 experimental SAXS data (χ2 = 1.708) (Fig 3E; Table S3). The bent conformation fits the kinked 202 crystal structure of D0-D1. The equilibrium is shifted towards the bent form in comparison to the 203 extended one with a volume fraction of 0.8:0.2. Further, the presence of two different 204 conformational states supports a higher flexibility of D0-D1.

205 Contrary to D0-D1, ab initio modelling of D00-D0 pointed to an elongated conformation 206 (Fig 3D) but did not have a good fit to experimental data (χ2 = 18.5) (Fig S2B). Likewise, D00-D0 207 EOM models did not have a good fit, especially at low q-values (χ2 = 11.6). Higher intensity at low 208 q-values for the experimental data is an indication of the presence of larger species. OLIGOMER 209 analysis of the monomer from the crystal structure and from the EOM analysis as well as a dimer 210 based on the crystal structure as input was included. OLIGOMER results had a better fit to 211 experimental data (χ2 = 2.45 for merged file) (Fig 3F) and revealed that the majority of D00-D0 is

212 present in the elongated conformation as revealed in the crystal structure (Fig S2C). However, 213 the individual SAXS curves could indicate that a dimer might form at a higher volume fraction at 214 higher concentrations (Table S4). As SAXS experiments were performed in a lower salt 215 concentration than present during the purification process, size exclusion chromatography (SEC) 216 was performed to analyze the effect of salt on the oligomerization or aggregation of D00-D0. SEC 217 in different sodium chloride concentrations revealed equivalent chromatograms with a major peak 218 corresponding to the monomer (Fig S3). We therefore conclude that D00-D0 is largely 219 monomeric, and that the apparent oligomerization of D00-D0 detected in SAXS analysis likely 220 only exists in a small fraction of the sample. 221

222 Molecular dynamics (MD) simulations support differences in flexibility between D00-D0 223 and D0-D1

224 The conformational ensembles sampled by microsecond-scale MD simulations agree with SAXS 225 findings. The conformations assumed by D00-D0 and D0-D1 are characterized in terms of the 226 angle between the principal axes of the two domains, θ, and by the largest distance between Cߙ

227 atoms, Dmax. The principal axis of a domain is defined as the eigenvector corresponding to the 228 smallest eigenvalue of the inertia tensor of the backbone. Based on this definition, the crystal 229 structures of D00-D0 and D0-D1 have θ ~ 170° and θ ~ 60°, respectively. Figure 3G shows that 230 D00-D0 adopts extended conformations characterized by a narrow, approximately bell-shaped 231 distribution of θ, centered around 140° whereas D0-D1 is considerably more flexible and explores 232 a wider range of domain-domain angles, including the strongly bent conformations. Accordingly,

233 Figure 3H indicates that the distributions of Dmax estimated from MD simulations display median 234 values that are 1 nm larger for D00-D0 than for D0-D1.

235 The root-mean-square fluctuation (RMSF) of residues in the D00 (K450–S550), D0 236 (G559–V653), and D1 (T658–F751) domains of D00-D0 and D0-D1 is calculated for the backbone 237 heavy atoms after alignment of the trajectory frames to the single-domain coordinates in the 238 corresponding crystal structures. In Figure S4, the RMSF values are superposed by shaded areas 239 indicating the secondary structure assignment into E-strands, H-bonded turns, bends, and D- 240 helices. Each residue is assigned to the secondary structural element identified in more than 50 % 241 of the analyzed trajectory frames according to the H-bonding and geometrical definitions of the 242 DSSP method 27. The most pronounced flexibility occurs in D00, particularly in regions of high 243 curvature connecting E-strands (Fig S4). In all domains, flexible loops are often accompanied by 244 H-bonded turns, i.e. H-bonds between backbone carbonyl and amide groups of consecutive

245 residues. The D00 domain shows enhanced flexibility in the loop (I475–I485) whereas the D-helix 246 and the loop region between E-strands E and F (H501–Y510) are comparatively less mobile. D1 247 is the most rigid domain and, interestingly, D0 is considerably more flexible in D0-D1 than in D00- 248 D0. 249

250 D00 is a building block found in passengers of many inverse autotransporters

251 In recent years, the range of the inverse autotransporter family in bacteria has been shown to be 252 significantly larger than previously recognized and also shows a broader variety of protein 253 domains and architectures than that of just close intimin and invasin homologues 28. Most 254 extracellular domains of inverse autotransporters contain certain subdomain types in different 255 amounts and orders, reminiscent of different combinations of the same building blocks (Fig 4A). 256 This is similar to what has been observed for type Vc secretion system or trimeric autotransporter 257 adhesins 29. Interestingly, for many of the (predicted) inverse autotransporters only a poor 258 structural domain prediction is available for the sequence (ca. 100 residues) directly after the 259 membrane-embedded β-barrel (Fig 4A) 30. This has led, until recently, to the conclusion that no 260 folded domain might be present 19. However, sensitive homology-based searches using 261 HHPred 31 suggested that multiple inverse autotransporters contain a D00-like IgSF domain at 262 the N-terminus of the passenger 16.

263 Due to the similarity to the domain prediction of intimin subdomain D00, we performed an 264 alignment on this region and identified conservation of several residues separated from the overall 265 Ig fold (Fig 4B). Two consecutive glycine residues (G496, G497) are conserved in close structural 266 proximity to a proline residue (P515), followed by a tyrosine (Y517) or tryptophan. An asparagine 267 (N523) and a second tyrosine (Y525)/ tryptophan are further conserved. Additionally, D00 268 contains a single tryptophan residue (W487), which is present among nearly all selected inverse 269 autotransporter sequences. The described residues are not conserved among all inverse 270 autotransporters, but most of the representatives contain them. Interestingly, the mentioned 271 residues, excluding the two consecutive glycine molecules, were also conserved among inverse 272 autotransporters with a minimal passenger, highlighted by the last three sequences in the 273 alignment.

274 In intimin D00, these conserved residues are located in the connector region of E-strands 275 E and F near the short helix element (Fig 4C). The double glycine stretch allows G496 to orient 276 the planar peptide bond with P515 in the neighboring E-strand E. Proline residue P515 is stabilized

277 through an interatomic aromatic-proline interaction with Y525 in E-strand F 32. The hydroxy group 278 of Y525 forms a hydrogen bond to a water molecule, which by itself creates a hydrogen bond to 279 the G-oxygen of the amido group on N523. The asparagine residue is hydrogen bonded to the 280 nitrogen in backbone peptide of V518 and further stabilized by a hydrogen bond between the G- 281 nitrogen of N523 and backbone carbonyl of V518. Thereby, the amido group of N523 is forced 282 into a planar conformation with the conjugated sidechain of Y517 kept within a 3.5 Å distance (Fig 283 4C). 284

285 Rigidification of D00-D0 increases the radius of reach of intimin

286 To investigate the influence of the conformational rigidity of D00-D0, we developed a coarse- 287 grained model of the extracellular passenger of intimin. The elongated protein is modelled by 288 spherical beads located at the mass centers of the folded domains and of the connector regions 289 (Fig 5A). The D00-L1, D2-L4 and D3-L4 bonds are modeled based on the all-atom data for D00 290 whereas the remaining bonds reproduce the all-atom distribution for the distance between D0 and 291 the connector region of D0-D1 (Fig S5B). Triplets of beads interact either via rigid or flexible 292 harmonic angular potentials, which are parametrized against the all-atom distributions of the 293 domain-domain angles of D00-D0 and D0-D1, respectively (FigS5A). D0-L2-D1 and D1-L3-D2 294 are flexible angles whereas L1-D0-L2, L2-D1-L3, L3-D2-L4 and D2-L4-D3 are rigid.

295 We performed Metropolis Monte Carlo simulations of the coarse-grained intimin, modelling 296 D00-L1-D0 either as a rigid or as a flexible angle. Figures 5B,C show probability distributions of 297 the distance between D3 and D00, ݎ, separation and of the angle between the D00–D3 vector 298 and the normal to the membrane surface, α, for the rigid (blue lines) and the flexible (red lines) 299 D00-D0 angles. Figures 5D,E show the corresponding 2D probability distribution as a function of 300 both the radius of reach, ݎ, and the of the multi-domain protein, α.

301 These results indicate that the stiffness of D00-D0 negligibly affects the orientation of the 302 extracellular passenger with respect to the surface whereas it significantly increases the radius of 303 reach of the receptor.

304

305

306

307 Discussion

308 The tight attachment to the host cell under fluid flow conditions is a major obstacle for bacterial 309 cells during adhesion. Most bacteria have therefore adhesion molecules, which are able to 310 withstand the shear forces present in fluid environment and maintain a strong attachment 33. 311 Intimin, an adhesion protein essential for EPEC and EHEC infections in the intestinal 312 environment, is required to also withstand these physical conditions, especially in its extracellular 313 passenger 34. In this study, we investigated the structure and organization of the subdomains of 314 the passenger region regarding physical properties that could play a role in promoting attachment 315 during the adhesion process. We propose a structural model of the entire passenger of E. coli 316 intimin and its connection to the bacterial outer membrane, sharing a common structural 317 organization with invasin (Fig 6). Structures of subdomains D00-D0 and D0-D1 obtained in this 318 study were modeled in connection to isolated structures of the C-terminal passenger domains 319 (PDB: 1F02) and the β-barrel translocation unit (PDB: 4E1S) to obtain a model for the complete 320 passenger. We hypothesize that the repeated Big domains of the extracellular passenger are 321 divided in different subregions depending on the connectors between the individual Big domains. 322 This domain organization would allow the passenger region to vary in rigidity and flexibility. 323 Although the functional role of these consecutive variations in flexibility is unclear, we hypothesize 324 that they promote increased binding probability of the passenger and intimate attachment (Fig 5).

325 We propose that a D00-like Ig domain is a common element present in most inverse 326 autotransporters, including the passenger of invasin. The invasin D0 domain would thus be 327 equivalent to intimin D00. The list includes, next to intimin and invasin homologues, other inverse 328 autotransporters, such as FdeC from E. coli 35 or inverse autotransporters yrInv and yrIlm from 329 the fish pathogen Yersinia ruckeri 36,37. Our study highlights that D00 should be included as a 330 building block in domain prediction tools. Previous results demonstrated that the D00 domain is 331 not crucial for the intimin autotransport process and an intimin mutant in which the D00 domain 332 was deleted showed similar adhesion to HeLa cells compared to the wild type protein 16. However, 333 its positioning directly after the translocator at the membrane interface supports our hypothesis 334 that D00 acts as a rigid stabilizing factor for the long rod-shaped extracellular passenger. Despite 335 an overall Ig fold, D00 exhibits differences to the other Big domains in the passenger region, while 336 we detected structural similarity to cadherin EC domains (Fig 6).

337 The following subdomains, D0-D1, extend the stalk of the passenger. In other inverse 338 autotransporters, the number of Big domains in this region can vary extensively, ranging from two

339 to 47 38. The reason for these variations is still unknown, but it was hypothesized that it is important 340 for spanning the bacterial surface structures, so that the functional domain (adhesin domain) can 341 be displayed for interaction with the host 38. The C-terminus of the extracellular passenger 342 consists of the actual receptor module, the superdomain D2-D3 in intimin, which directly interacts 343 with the Tir receptor on the host cell surface. Regarding the overall structure of the extracellular 344 passenger, our results show that the extended string of tandem Ig domains exhibits an internal 345 twist with each domain turned approximately 180° degrees in comparison to the previous one, 346 making the entire passenger resemble a helix. A similar formation of a higher order structure of a 347 superhelical-like rod consisting of hundreds of Ig-like subdomains is present in Type 1 pili, 348 important for bacterial cell adhesion 23. In the case of Type 1 pilus, the superhelical structure can 349 unwind, thus reducing shear stress upon receptor binding under fluid forces 39.

350 Our work indicated differences in rigidity and flexibility between the different extracellular 351 domains. D00-D0 revealed an elongated and more rigid conformation while D0-D1 displayed 352 increased flexibility with higher probability to be present in a bent conformation in solution, 353 equivalent to the obtained crystal structures. This rigidity is reminiscent of the inflexibility seen 354 between D3 and D2 in the solution structure of the C-terminus of intimin 40. By contrast, the D1- 355 D2 connector is flexible in the solution structure, similar to what we observed for D0-D1. 356 Therefore, we propose a model of the intimin passenger in which D00-D0 and D2-D3 form 357 functional units of high rigidity while the stalk subdomains D0-D1-D2 show higher flexibility, which 358 could be beneficial during the binding process (Fig 5 and Fig S7). MC simulations supported the 359 notion that a rigid base in the intimin passenger, represented by D00-D0, increases the radius of 360 reach and shifts the orientation of intimin in relation to the mammalian host cell. This could 361 accelerate the binding process of intimin and the Tir receptor, presented on the mammalian host 362 cell, and potentially increase the efficiency of a successful adhesion event (Fig S7). The flexibility 363 demonstrated by the D0-D1 and D1-D2 connectors is also likely to be functionally important. The 364 crystal structure of the intimin-Tir complex shows a large angle, close to 180 q, between the two 365 proteins. For the rigid D2-D3 domain to approach at this angle, the intimin passenger must be 366 able to bend. This configuration also brings the host cell membrane and the bacterial outer 367 membrane into close proximity. It promotes the two passengers of one intimin monomer to interact 368 with separate Tir molecules, rather than a single dimer. This is necessary for receptor clustering 369 and downstream signaling events leading to the formation of actin pedestals 22,41.

370 Interestingly, also members of the cadherin family, a family of adhesion proteins 371 containing long sequences of IgSF domains in their extracellular domains, contain both extended

372 and bent elements 42, similar to the structural conformation obtained for D00-D0 and D0-D1, 373 respectively. In their functional forms, the cadherin domains are rigidified by Ca2+-binding sites in 374 the domain connector regions. Yet, structural forms of the T-Cadherin domains without Ca2+- 375 bound showed a strong bend between neighboring domains. The presence of Ca2+-binding sites 376 allows regulation of the flexibility of the extracellular adhesion domains. For some members, like 377 Protocadherin-15, the overall rigid conformation was locally permanently disturbed through a 378 calcium-free linker between the two neighboring domains, in which the Ca2+ binding sites were 379 mutated. It was proposed that this structural feature confers soft elasticity in an otherwise rigid tip 380 link, allowing a smoother response to low forces applied as the molecule will rather extend than 381 unfold. Another example in which the presence of elastic elements disrupted a linear rod 382 conformation was found in titin, a muscle protein submitted to shear stress and involved in 383 mediating elasticity 43. In this case, elasticity was increased through bending and twisting of 384 neighboring IgSF domains.

385 In summary, our findings on structural and biophysical similarities and differences of the 386 extracellular passenger subdomains of intimin can contribute to our general understanding of 387 functional regions along the often neglected adhesin stalk and how receptor binding can be 388 affected through variations in rigidity and flexibility. 389

390 Material and Methods

391 Chemicals were obtained as grade BioUltra from Sigma-Aldrich unless otherwise stated. 392 393 Constructs of D00-D0 and D00-D1

394 The D00-D0 domains of intimin were amplified using Q5 polymerase (New England Biolabs) and 395 pIBA2-Int-Strep as the template 16. Primers were designed to introduce an N-terminal 396 hexahistidine tag. D00-D0 were then cloned into the expression vector pAKS-IBA3 (IBA GmbH) 397 using Gibson assembly 44. The reaction mix was transformed into chemically competent TOP10 398 cells (ThermoFisher) and transformants were selected on lysogeny broth (LB) medium 399 supplemented with ampicillin (100 μg/ml) 45. Positive colonies were screened for using colony 400 PCR, and a PCR-positive clone was verified by Sanger sequencing (Eurofins). 401 To introduce a tobacco etch virus (TEV) protease cleavage site after the His-tag, site- 402 directed mutagenesis was employed 46. Once the insertion had been verified by sequencing, the 403 plasmid was amplified from the TEV site until after the D00-D0 insert to produce a linearized

404 plasmid with a His-tag followed by a TEV site on one end and the stop codon on the other. Into 405 this, we inserted the sequence coding for the D0-D1 domains, amplified as above. The cloning 406 was performed as outlined above. These procedures resulted in two plasmids, pIBA3-His-TEV- 407 IntD00-D0 and pIBA3-His-TEV-IntD0-D1. 408 To make the connector flexible, the connector sequence (residues 551-557) was replaced 409 by a glycine-serine sequence of equal length (GSGSGSG). The exchange was done by site- 410 directed mutagenesis 46. This was done for both, the three-domain construct pIBA3-IntD00-

16 411 D1Cys to produce pIBA3-IntD00-D1Cys-GSconnector and the full-length intimin (pIBA2-Int-Strep)

412 construct to produce pIBA-Int-GSconnector-Strep. The mutations were verified by sequencing. All 413 primer sequences are given in Table S5. 414 415 Expression and purification

416 The expression and purification of all constructs were performed according to the protocol 417 described by Leo et al. (2016) 16. D00-D0 or D0-D1, inserted into pIBA3-vectors, were transformed 418 into E. coli BL21Gold(DE3) (Novagen) and plated on LB medium with 1.5% agar and 100 μg/ml 419 ampicillin. Colonies of transformed BL21Gold(DE3) cells were inoculated in LB medium 420 containing 100 μg/ml ampicillin and incubated at 37 °C for 16 hours (h). The following day, 1% 421 overnight culture was added to ZYP medium (10 g tryptone, 5 g yeast extract, 5 g sodium chloride

422 in 1 L ddH2O) enriched with 100 μg/ml ampicillin and incubated with shaking at 37 °C until the

423 optical density at 600 nm (OD600nm) reached 0.5-0.7. Protein expression was induced with 424 anhydrotetracycline (ATCH) (IBA GmbH) at a final concentration of 0.2 ug/ml. The cells were 425 further grown at 18 °C for 16 h. Cells were harvested by centrifugation at 7000 x g for 10 min. Cell 426 pellets were stored at -20 °C until use. Thawed cell pellets were resuspended at 1:10 ratio in 427 buffer A (10 mM HEPES-NaOH pH 7.5, 400 mM NaCl, 1 mM Phenylmethylsulfonyl fluoride 428 (PMSF), 1 μg/ml DNase) and lysed using a high pressure homogenizer (C5 model, Avestin, 429 Germany) (3 passes at 15,000 psi). Cellular debris and the membrane fraction were removed by 430 centrifugation at 100000 x g for 45 min. The clarified supernatant was applied to a 5 ml HisTrap 431 HP column (GE Healthcare). The column was washed with 10 column volumes (CV) in buffer A 432 supplemented with 30 mM imidazole. The protein was eluted with buffer A supplemented with 433 120 mM imidazole and the fractions containing the D00-D0 or D0-D1 protein were pooled 434 separately. His-tagged TEV protease purified in-house 47 was added to the protein solution in a 435 1:28.5 (w/w) ratio and dialyzed against 10 mM HEPES-NaOH pH 7.5, 400 mM NaCl, 5 mM β- 436 mercaptoethanol (βME) for 16 h at 4 °C. The dialyzed sample was passed through a Histrap HP 437 column equilibrated in buffer A. The flow-through was collected and concentrated (Centrifugal

438 concentrator, 10 kDa molecular weight cut off (VivaSpin)). The purity of the sample was assessed 439 by SDS-PAGE and size exclusion chromatography. For the latter, the Superdex 75 10/300 GL 440 column (GE, cat.#29148721) was equilibrated with 10 mM HEPES-NaOH pH 7.5, 400 mM NaCl.

441 For atomic force microscopy experiments, proteins D00-D1 and D00-D1-GSconnector with 442 C-terminal cysteines were produced and purified as described by Leo et al. (2016) 16.

443 444 Crystallization

445 D00-D0 or D0-D1 constructs were concentrated to 50 mg/ml and 80 mg/ml, respectively, to reach 446 supersaturation and induce crystal formation. Crystallization experiments were performed with 447 the IndexTM screen (Hampton Research), and inhouse screens. Drops were set up with 1 ul of 448 protein and 1 ul precipitant solution in 24-well plate as hanging drops. The wells were sealed with 449 immersion oil (Sigma cat.#56822). Plates were incubated at 18 °C. Initial crystals of D00-D0 450 appeared in three days. The best D00-D0 crystals were grown using a reservoir solution of 0.15 451 M potassium bromide and 30 % (w/v) polyethylene glycol monomethyl ether 2000 452 (PEG2000mme). The best crystals of D0-D1 were obtained in 0.1 M tris, pH 8.5; 0.2 M magnesium 453 chloride hexahydrate and 25 % polyethylene glycol 3350 (PEG3350). Crystallization had to be 454 induced by the addition of 4 % glycerol in the reservoir to promote vapor diffusion; crystals 455 appeared after one day. Crystals were harvested using mounted CryoLoops (Hampton Research) 456 and flash frozen in liquid nitrogen. Data sets were collected at DESY Hamburg, Germany using 457 the PETRAIII beam line P14. All experimental details are given in Table 1.

458 459 Data Processing

460 Molecular replacement using the program Phaser 48 was used to solve the structures of D00-D0 461 and D0-D1 with D1 domain (PDB: 1F02) 18 as an initial search model. The structures were refined 462 using the programs PHENIX 49 and REFMAC 50. Data collection and refinement statistics are 463 summarized in Table 1.

464

465 Small Angle X-ray Scattering (SAXS)

466 Purified D00-D0 and D0-D1 were dialyzed against 10 mM HEPES-NaOH, pH 7.5 and 100 mM 467 KCl and concentrated to ca. 28 mg/ml in a centrifugal concentrator (10 kDa molecular weight cut 468 off (VivaSpin)). Buffer that passed through the filter was collected and used as a blank for the

469 SAXS measurement. SAXS data was collected at the P12 beamline at the Petra III storage ring 470 (DESY, Hamburg DE) 51 (for experimental details see supplementary table 6). Structural

471 parameters, including radius of gyration (Rg) and maximum dimension (Dmax) were derived from 472 the experimental data with the graphical data analysis program PRIMUSQT 52. Ab initio models 473 for both proteins were obtained from GASBOR 53. Conformational polydispersity of each protein 474 was studied using Ensemble Optimization Method (EOM) 54,55, which consists of two programs 475 RANCH and GAJOE. RANCH was employed to create 10000 random conformations (genes) by 476 using two domains (D452-L549 and G559-Q656) from the crystal structure. GAJOE was used to 477 select ensembles of conformations that have a better fit to the experimental data and was run ten 478 times for each protein. The fit was done against merged SAXS curves in order to minimize noise 479 at high q-values and remove contributions of interparticle interactions at low q. For D00-D0 SAXS 480 data, OLIGOMER was performed, using in total 13 components: all the models from GAJOE, the 481 crystal structure, and two dimers extracted from the crystal structures. FFMAKER 52 was used to 482 create an input file for OLIGOMER with a form factor for each component. Experimental details 483 are shown in Table S2. Measurements were taken from distinct samples.

484 485 Molecular simulations

486 Molecular dynamics (MD) simulations are performed using the AMBER ff99SBnmr1 force field 56 487 and the GB7 implicit solvent model 57. The ionic strength of the solution is set to 0.15 M and 488 modelled within the Debye-Hückel approximation 58. After energy minimization of the respective 489 crystal structures, D00-D0 and D0-D1 are simulated for 3.0 μs using the AMBER16 MD package 490 in the NVT ensemble at 298 K maintained by Langevin dynamics with a collision frequency of 20 491 ps-1. The integration time step is 2 fs and all bonds involving hydrogen atoms are constrained 492 using the SHAKE algorithm 59. Secondary structure assignments are performed using the DSSP 493 algorithm 27 implemented in MDTraj 60. The first 200 ns of the trajectories are excluded from the 494 analyses based on the time evolution of the RMSD with respect to the crystal structure (see Fig 5). 495 Metropolis Monte Carlo (MC) simulations are performed using the Faunus framework 61. 496 The coarse-grained model of the extracellular passenger of intimin consists of five beads with a 497 diameter of 4 nm, separated by four smaller connector beads with a diameter of 1 nm (Fig 5A). 498 The beads are connected by harmonic bonds parametrized by fitting a Gaussian function to the 499 distributions of the mass-center distances between the domains and the connector in the all-atom

500 trajectory of D00-D0 (Fig S6B). The D00-D0 bond has force constant, kb, of 1 kJ/mol and

501 equilibrium distance, ݎ௘௤, of 2.6 nm, whereas all the other bonds have kb = 0.5 kJ/mol and req =

502 2.1 nm. Angular interactions are modeled by harmonic potentials derived by fitting a Gaussian 503 function to the probability distributions of the domain-domain angles of D00-D0 (rigid) and D0-D1

504 (flexible) (Fig S6A). Rigid angles have force constants kθ of 30 kJ/mol and equilibrium angle θeq

505 of 150° while flexible angles have kθ = 2 kJ/mol and θeq = 110°. The mass center of the D00 bead 506 is fixed at 5 nm above a hard wall, which represents the membrane surface located in the xy- 507 plane. To account for the excluded volume of the ߚ-barrel domain, an extra bead with a diameter 508 of 5 nm is positioned between the membrane and the D00 bead, at a distance of 0.5 nm from the 509 hard wall. The other beads translate in the volume above the surface, sampling the conformational 510 space of the coarse-grained adhesin model at 298 K. Nonbonded bead-bead interactions are 511 modeled via the Weeks-Chandler-Andersen potential 62. The probability density distributions of 512 the D3-D00 distance, P(r), and of the angle formed by the D3-D00 vector with the membrane 513 normal, P(α), as well as the 2D probability distribution P(r,α) are corrected to remove the Jacobian ל ஶ ሺ ሻ ଶ ଵ଼଴ ሺ ሻ ߙܲ ߙ ߙ ൌͳ and לൌͳ, ׬଴ ݎ ݎ ܲݎ entropy contributions and normalized so that ׬଴ 514 ל ஶ ଵ଼଴ ሺ ሻ ଶ ߙ ൌͳ. A Jupyter notebook in ipynb (Supplementary Data 1) and ݎ ǡ ߙݎ ߙܲݎ ל׬׬଴ ଴ 515 516 HTML formats (Supplementary Data 2) detailing simulations and analyses of the coarse-grained 517 model are provided as supplementary data.

518 519 Atomic force microscopy

520 Atomic force microscopy (AFM) was carried out as previously described 16 with slight 521 modifications. To prepare Ni-NTA functionalized probes, silicon nitrite gold-coated cantilevers 522 with spring constant of 0.08-0.12 N/m (OMCL-TR400PB: Olympus, Japan) were used. To 523 immobilize the intimin onto a gold substrate (ARGOLD-15mm: Asylum Research, CA), 10 mg/ml 524 of purified intimin reduced in 10 mM TCEP-PBS for 16 h was placed onto the gold substrate for 525 30 min.

526

527 Preparation of whole cell lysates, SDS PAGE and western blot to compare expression 528 levels of Strep-tagged Intimin variants

529 The plasmids pIBA2-Int-Strep and pIBA-Int-GSConnector-Strep were introduced into E. coli BL21 by 530 chemical transformation. Single clones were used for a second transformation with pACYC-EGFP 531 or pACYC-RedEx to obtain fluorescently labeled strains (originally intended to be used for 532 adhesion assays). These plasmids are based on pACYC-184, where first a constitutive pTac-

533 promoter via restriction sites BamHI/HindIII was inserted. The promoter was amplified from 534 pMK4 63 using the primers Ptac_For and Ptac_Rev. The coding sequences of EGFP (derived from 535 pEGFP-N1) or DsRedExpress (derived from pDsRedExpress-C1) was inserted. EGFP and 536 RedExwere amplified using the primers EGFP_For and EGFP_Rev or RedEx_For and

537 RedEx_Rev, respectively. To compare expression levels of Int-Strep and Int-GSConnector-Strep, 538 overnight cultures containing ampicillin (100 μg/ml) and chloramphenicol (25 μg/ml) in LB were 539 inoculated with a single colony from a selective agar plate. Bacteria were grown with shaking

540 overnight at 27 °C. The following day, the OD600 was measured and subcultures were inoculated 541 at an OD 0.1 in LB with the same supplements as in the overnight culture. After 2 hours of growth 542 at 27 °C, 1 ml samples were withdrawn to prepare whole cell lysates for western blot (“before

543 induction”). In the remaining cultures the expression of Int-Strep and Int-GSConnector-Strep was 544 induced by addition of AHTC (200 ng/ml final concentration), and the cultures were grown further 545 for 3 h at 27 °C with shaking. After 3 h, 1 ml samples were again withdrawn and prepared for 546 SDS-PAGE. Whole cell lysates were prepared with 10 μl of sample containing 2.5 x 106 cells. 10 547 μl samples were separated on a 4-20 % TGX gradient gel (BioRad) and blotted onto nitrocellulose. 548 The membrane was blocked overnight at 4 °C with 5 % milk in TBS-T. Strep-tagged proteins were 549 finally detected using an anti-Strep Tag II antibody (IBA Lifesciences) diluted 1:1000 in 5 % milk 550 in TBS-T for 2 h at room temperature. After two washes for 10 min each with TBS-T, the secondary 551 antibody (goat-anti mouse HRP, Dianova), diluted 1:1000 in 5 % milk in TBS-T was applied for 552 1 h at room temperature with gentle shaking. After two additional washing steps with TBS-T for 553 10 min each, detection was carried out using the Clarity Western ECL Kit (BioRad). Images were 554 taken using a chemiluminescence imaging system (Peqlab).

555 556 Data availability 557 Data that supports the findings of this study have been deposited in protein data bank with the 558 PDB identifier 6TQD and 6TPL. Additional data are implemented as Supplementary data. All other 559 relevant data are available from the corresponding author. 560 Identifier code 561 6TQD 562 6TPL 563 564 565 566

567 Author contributions 568 J.W., J.C.L. and J.P.M. conceived and planned the experiments, and analyzed the data. J.C.L. 569 prepared constructs used in this study. J.W. and L.V.J. expressed and purified the proteins, J.W. 570 performed crystallization, and analytical SEC. J.W. and J.P.M. carried out the crystal data 571 collection and structure determination. A.K. performed the SAXS data collection and analysis. 572 G.T. performed molecular simulations and S.Y. the atomic force microscopy. M. Schütz carried 573 out expression studies. K.H., M. Skepö, P.H., J.C.L. provided critical feedback. J.W. and J.P.M. 574 wrote the paper with input from all authors. 575 576 Acknowledgements 577 Computational resources were provided by Lunarc in Lund. We thank Prof. Dirk Linke (University 578 of Oslo) for support and valuable input during the course of the study. We thank Christopher J. 579 Stubenrauch (Monash University) for valuable critical feedback. 580 581 Declaration of Interests 582 The authors declare no Competing Financial or Non-Financial Interests. 583

584

585

586

587

588

589

590

591

592

593

594

595

596

597 Figure 1

598

599 Figure 1: 600 Crystal structure of the E. coli intimin subdomains D00-D0 exhibits an elongated 601 conformation while D0-D1 shows a bent structure 602 (A) Domain architecture of E. coli intimin is represented based on the PDB structures exhibiting 603 the periplasmic domain (PDB: 1E5U), including a signal peptide (SP), transmembrane β-barrel 604 (PDB: 4E1S) and an extracellular section, referred to as passenger (PDB: 1F02). 605 (B) Topology and ribbon diagram of D00-D0. D0 shows a classical bacterial Ig-like domain 1 606 (Big1) motif consisting of a β-sandwich. Opposing antiparallel beta-sheets are represented in cyan 607 and blue. D00 shows a modified β-sandwich with opposing beta-sheets represented in green and 608 yellow. β-strands A1 and B2 form a small separate β-sheet (dark green). The overall β-fold is 609 disrupted in the third β-strand by an α-helix (light purple) and shortened β-strands D and E. 610 (C) Topology and ribbon diagram of D0-D1. Both subdomains show a classical bacterial Ig-like 611 domain 1 (Big1) motif with opposing antiparallel β-sheets represented in cyan and blue. 612

613

614

615

616

617

618

619

620

621

622

623

624 Figure 2

625

626 627 628 629 630 631 632 633 634

635 Figure 2: 636 The connector region between subdomains D00 and D0 shows no effect on mechanical 637 stability 638 (A) The connector region (orange) between D00 (green) and D0 (cyan) is represented. A network 639 of hydrogen bonds between indicated amino acids stabilizes the extended connector, which forms 640 a type I turn. 641 (B) Scheme of intimin highlighting the connector region between D00 and D0. The connector 642 region (orange) was replaced by GS substitution in mutational studies to analyze the role of the 643 connector for mechanical stability. The construct used for AFM consisted of the three subdomains 644 D00-D0-D1, as highlighted, which are located in vivo between the β-barrel and the passenger 645 subdomains D2-D3 (transparent).

646 (C) AFM measurements on intimin passenger constructs Int-WTconnector and Int-GSconnector. 647 Representative force distance curves (up) and plotting of the force peaks from each unfolding

648 event from multiple experiments against contour length (LC) (down) of Int-WTconnector, containing

649 the wild type connector, (left) and Int-GSconnector (right), containing the mutated connector replaced

650 by GS substitution, are shown. The LC values were calculated for each force-distance curve by 651 fitting to a wormlike chain model. The clusters of unfolding events (UE) are indicated and the 652 colors match the schematic depictions in panel B. The unfolding events for D00, D0 and D1 are 653 assigned based on the results obtained by Leo et al. (2016), the unfolding events of D0 and D1 654 cannot be distinguished and are arbitrarily assigned. Detachment of sample from the cantilever 655 tip is indicated by (D). 656

657

658

659

660

661

662

663

664 Figure 3

665

666 Figure 3: In solution D00-D0 exhibits an extended conformation while D0-D1 shows 667 primarily a bent conformation 668 (A) SAXS scattering curves and pair-distance distribution function p(r) (inset) of D0-D1. 669 (B) SAXS scattering curves and pair-distance distribution function p(r) (inset) of D00-D0. 670 (C) Ab initio (grey beads) and Ensemble Optimization Method (EOM) model of bent (cyan), 671 corresponding to the crystal structure, and extended (dark grey) conformation for D0-D1. 672 (D) The fitting of the EOM models (cyan) is shown in the graph (χ2 = 1.708). 673 (E) Ab initio (grey beads) and Ensemble Optimization Method (EOM) model of extended crystal 674 structure (dark green) and EOM model (grey) of D00-D0. 675 (F) The fitting of the EOM models (green) is shown in the graph (χ2 = 2.45). 676 (G) Probability distributions of the domain-domain angle, θ, of D00-D1 (blue) and D0-D1 (red).

677 (H) Probability distributions of maximum dimension, Dmax, of D00-D1 (blue) and D0-D1 (red). 678 Figure 3G and 3H were calculated from implicit solvent molecular dynamics simulations. 679

680 Figure 4

681

682 Figure 4: D00 is a building block found in passengers of many inverse autotransporters 683 (A) Domain prediction by HMMER of (putative) intimin, invasin and other inverse autotransporter 684 sequences reveals repetition of the same domain types, equivalent to molecular building blocks. 685 For most, no domain is predictable for the 100 amino acids following the β-barrel. Each selected 686 sequence is shown as the Uniprot number or NCBI accession number followed by the 687 corresponding bacterial species. 688 (B) Alignment with Muscle of potential D00-like domains (100 amino acids following β-barrel) 689 reveals conservation among E. coli intimin D00 and the extracellular N-terminal region of other 690 inverse autotransporter, indicating D00 as another building block in passenger construction. 691 Highly conserved amino acids, located at the top of the domain and differing from the classical 692 Big1 fold, are highlighted. Sequences are selected with less than 70 % redundancy and 693 correspond to inverse autotransporters shown in (A).

694 (C) Amino acids in D00 are highlighted by an 2F0-Fc electron density map (blue mesh) countered 695 at 1.5 s. The double glycine stretch allows G496 to orient the planar peptide bond with P515 in 696 the neighbor β-strand E. Proline residue P515 is stabilized through an interatomic aromatic- 697 proline interaction with Y525 in β-strand F. The hydroxyl group of Y525 forms a hydrogen bond 698 to a water molecule, which by itself creates a hydrogen bond to the delta oxygen of the amido 699 group on N523. The asparagine residue is further hydrogen bonded to the nitrogen in backbone 700 peptide of V518 and further stabilized by a hydrogen bond between the delta nitrogen of N523 701 and backbone carbonyl of V518. Thereby, the amido group of N523 is forced into a planar 702 conformation with the conjugated sidechain of Y517 kept within a 3.5 Å distance. A similar 703 structural interaction is likely present in the listed sequences in (B). 704

705

706

707

708

709

710

711 Figure 5

712

713 Figure 5: Monte Carlo simulations indicate that the rigidity of D00-D0 increases the radius 714 of reach of intimin 715 (A) Cartoon representation (upper) and visualization of the coarse-grained model (lower) of the 716 extracellular passenger domain of intimin. In the cartoon, the rigid D00-D1 and D2-D3 domains 717 are depicted as rectangles, whereas the flexible D0-D1 and D1-D2 connectors are represented 718 by curves. The β-barrel, D00, D0-D1 and D2-D3 domains are shown in blue, green, cyan and 719 orange, respectively. The white beads of the coarse-grained model represent the linker regions. 720 The red and black arrows indicate the D3-D00 vector and the normal to the membrane surface, 721 respectively. Probability distributions of (B) the distance between D3 and D00 and (C) of the angle 722 between the D00-D3 vector and the normal to the membrane surface obtained from coarse- 723 grained simulations modelling the D00-L1-D0 angle as a flexible (red lines) or rigid (blue lines) 724 harmonic potential. 2D probability distributions as a function of radius of reach and orientation of 725 the D00-D3 vector obtained using a rigid (D) or a flexible (E) D00-L1-D0 angle.

726

727

728 Figure 6

729

730 Figure 6: Intimin and invasin passengers show an equivalent structural composition and 731 similarity to E-Cadherin extracellular extension 732 Structural model of intimin exhibiting the β-barrel domain located in the outer membrane (PDB: 733 4E1S; blue) connected to the extracellular passenger domain consisting of four Big subdomains: 734 D00-D0 (PDB: 6TQD; D0: cyan; D00: green) and D1-D2 connected to lectin-like domain D3 (PDB: 735 1F02; D1: cyan; D2-D3: yellow). Invasin exhibits high structural similarity to intimin, consisting of 736 membrane-located β-barrel (PDB: 4E1T, blue) and a long stalk of Ig domains with a lectin-like 737 receptor superdomain at the C-terminus (PDB:1CWV, D1-D3: cyan, D4-D5: red). Based on our 738 alignment, we propose an additional domain (D0) at the N-terminus of the invasin passenger, 739 equivalent to intimin subdomain D00. The structure of this domain is unsolved, but our work 740 suggests a similar structure to intimin D00. Further, D00 revealed structural similarity to the 741 extracellular domains of cadherins (EC), which are represented here by E-cadherin (PDB: 2QVF). 742 In comparison to intimin and invasin, there is no structural difference in the domain composition 743 along the extracellular stalk of cadherins. The rigidity and flexibility of cadherins are regulated by 744 calcium binding in the connector regions of EC domains.

745 Table 1: Data collection and refinement statistics

D00-D0 D0-D1 Data Collection Wavelength 0.9763 0.9762 Resolution range 24.83 - 1.48 24.77 - 1.797 (1.53 - 1.48) (1.86 - 1.797) Space group P1 P1 Cell dimensions a, b, c (Å) 56.0, 65.5, 79.4 33.7, 45.3, 67.6 α, β, γ (°) 77.8, 76.6, 84.8 97.1, 97.6, 99.9 Total reflections 809651 (49459) 125718 (12231) Unique reflections 175969 (16162) 32902 (3238) Completeness (%) 98.218 (90.218) 91.435 (89.24) I/σI 6.40 (1.54) 7.49 (2.09) Wilson B-factor 10.72 14.95 Rmerge 0.1014 (0.5213) Rmeas 0.14656 (0.6962) 0.1181 (0.6072) Rpim 0.066561 (0.3852) 0.06035 (0.31099) CC1/2 0.979 (0.354) 0.992 (0.821) CC* 0.995 (0.723) 0.998 (0.95) Refinement Reflections used in 175724 (16141) 32879 (3235) refinement Reflections used for Rfree 8656 (766) 1615 (160) Rwork 0.17436 (0.3364) 0.17768 (0.2326) Rfree 0.22198 (0.36594) 0.23437 (0.2998) CC(work) 0.969 (0.340) 0.938 (0.902) CC(free) 0.956 (0.575) 0.886 (0.848) Number of non-hydrogen 11495 3471 atoms macromolecules 9329 2849 ligands 30 7 solvent 2136 615 RMS(bonds) 0.008 0.009 RMS(angles) 0.91 1.02 Ramachandran favored (%) 98.52 97.43 Ramachandran allowed (%) 1.48 2.57 Ramachandran outliers (%) 0.00 0.00 Rotamer outliers (%) 0.48 0.31 Clashscore 0.96 2.43 Average B-factor 18.63 26.31 macromolecules 16.27 25.43 ligands 29.56 30.95 solvent 28.78 30.32 Number of TLS groups 36 13 746 a Statistics for the highest-resolution shell are shown in parentheses.

747

748 Supporting information

749 Figure S1

750

751 Figure S1: Expression of intimin constructs Int-Strep and Int-GSConnector-Strep 752 (A) Whole cell lysates corresponding to 2.5 x 106 cells per lane were analyzed by SDS-PAGE and

753 western blot. Int-Strep, containing a wild type D00-D0 connector, and Int-GSConnector-Strep, 754 containing a mutated D00-D0 connector replaced by GS substitution as described in the material 755 and methods section, were detected in samples harvested before and after 3 h of induction with 756 AHTC using an anti-Strep tag II antibody. For both intimin variants clones were tested that 757 additionally harbored either pACYC-EGFP or pACYC-RedEx. 758 (B) Ponceau staining of western blot shows equal loading for all samples.

759 Figure S2

760

761 Figure S2: Ab initio modeling and concentration dependent OLIGOMER fit on SAXS 762 scattering of D0-D1 and D00-D0 763 (A) Fitting curve (cyan) of ab initio model to merged and corrected SAXS data of D0-D1 764 (χ2 = 2.9). 765 (B) Fitting curve (green) of ab initio model to merged and corrected SAXS data of D00-D0 766 (χ2 = 18.5). 767 (C) Fitting curves (red) of OLIGOMER analysis on scattering curves of D00-D0 at different protein 768 concentrations. OLIGOMER was run against models of the EOM analysis, the protein crystal 769 structure and a dimer formation based on the crystal structure of D00-D0. 770 771

772

773

774 Figure S3

775

776 Figure S3: Salt-dependence of D00-D0 777 Analytical size exclusion profiles of subdomains D00-D0 on Superdex75 Increase 10/300 GL (GE) 778 in the presence of increasing sodium chloride concentrations. Analytical SEC revealed no salt- 779 dependence of D00-D0 oligomerization. D00-D0 is present as a monomer in all conditions. 780

781 Figure S4

782

783 Figure S4: Intimin subdomains vary in domain rigidity 784 Root-mean-square fluctuation (RMSF) of each residue in the D00 (K450–S550), D0 (G559– 785 V653), and D1 (T658–F751) domains of D00-D0 (left panels) and D0-D1 (right panels). Higher 786 RMSF indicates increased flexibility. Shaded areas indicate the secondary structure assignment 787 into β-strands (blue), H-bonded turns (orange), bends (green), and α-helices (red).

788

789 Figure S5

790

791 Figure S5: Comparison between intradomain probability distributions obtained from the 792 all-atom and the coarse-grained model.

793 (A) Probability distributions of the domain-domain angles in the all-atom (AA) simulations of D00- 794 D0 (black solid line) and D0-D1 (black dashed line). The colored lines are the D00-D0 angle 795 distributions obtained from the coarse-grained (CG) model using the flexible (red line) and the 796 rigid (blue line) harmonic potential. (B) Probability distributions of the mass-center separations 797 between the connector region L1 and the domains D00 (solid lines) and D0 (dashed lines). Black 798 and red lines are calculated from AA and CG simulations, respectively.

799

800

801

802

803

804 Figure S6

805

806 Figure S6: Time-resolved implicit solvent molecular dynamics simulations of D00-D0 and 807 D0-D1

808 Time evolution of (A) the domain-domain angle θ; (B) the maximum dimension Dmax; (C) the radius

809 of gyration Rg; (D) the RMSD with respect to the crystal structure of D00-D0 (blue) and D0-D1 810 (red) calculated from implicit solvent all-atom MD simulations. Dashed black vertical lines indicate 811 the last trajectory frame excluded from the analysis of the simulation data. 812

813

814

815

816 Figure S7

817

818 Figure S7: Model of how regions of rigidity and flexibility in the intimin passenger promote 819 adhesion and intimate attachment 820 (A) D00-D0 forms a rigid unit, which projects the intimin passenger further from the bacterial 821 surface, thus increasing the average radius of its reach (red arrow) and sphere of action (arc). 822 (B) In the hypothetical situation, in which the connector between D00-D0 allows similar flexibility 823 as seen experimentally for D0-D1, the average radius of reach and sphere of action are reduced.

824

825 S1 Table: Contour length increments and force peaks derived from AFM experiments

Int-WTconnector Int-GSconnector Number of measurements 40 47 Unfolding event 1 ΔLC (nm) 31.2 ± 2.4 30.8 ± 1.9 Max. Force (pN) 244.7 ± 45.2 256.1 ± 39.9 Unfolding event 2 ΔLC (nm) 30.2 ± 2.4 30.9 ± 1.9 Max. Force (pN) 292.9 ± 48.4 283.5 ± 31.9 Unfolding event 3 ΔLC (nm) 32.2 ± 2.0 31.9 ± 1.8 Max. Force (pN) 315.9 ± 28.4 305.9 ± 43.9

826 S2 Table: Structural parameters derived from SAXS experiments Guinier p(r) Porod Apparent MW (kDa) cprotein I(0)/c Rg I(0)/c Rg Dmax Volume Gui p(r) Porod (g/L) (nm) (nm) (nm) (nm3) nier Volu me D00- 1.00 4199.6 3.16 4070. 2.81 9.60 35 27 26 21 D0 3 00 2.27 4133.9 2.91 4060. 2.78 9.60 32 27 26 19 7 00 5.19 4176.6 2.96 4130. 2.88 11.0 32 27 27 19 1 00 0 7.55 4216.4 2.99 4178. 2.92 11.0 33 27 27 19 8 00 0 10.71 4250.9 2.94 4234. 2.96 11.0 33 28 27 20 7 00 0 D0-D1 1.09 2848.6 2.08 2799. 2.10 6.30 30 18 18 18 1 00 2.65 2984.0 2.66 2870. 2.23 8.60 34 19 19 20 5 00 5.95 2883.0 2.31 2852. 2.21 8.40 34 19 18 20 0 00 8.43 2880.0 2.25 2859. 2.23 8.70 33 19 19 20 0 00 12.43 2916.0 2.39 2870. 2.24 8.60 33 19 19 19 0 00 827

828 S3 Table: Fit parameters of D0-D1 with EOM

Model Rg (nm) Dmax (nm) Volume fraction ꭓ2 Bent 2.015 6.125 0.808 1.708 Extended 2.870 9.859 0.202 829

830 S4 Table: Fit parameters of D00-D0 with OLIGOMER

cprotein Bent Extended Dimer ꭓ2 (g/L) D00- 1.00 0.281 (±0.016) 0.536 (±0.016) 0.184 (±0.005) 1.04 D0 2.27 0.233 (±0.007) 0.561 (±0.007) 0.205 (±0.002) 1.47 5.19 0.270 (±0.004) 0.518 (±0.004) 0.212 (±0.001) 2.91 7.55 0.295 (±0.003) 0.478 (±0.003) 0.227 (±0.001) 3.65 10.71 0.330 (±0.003) 0.424 (±0.003) 0.246 (±0.001) 6.80 831

832

833 S5 Table: Primer sequences used in cloning Primers Sequences D00-D0 For: CGA GGG CAA AAA ATG CAT CAC CAT CAC CAT CAC AAG CAG GAT ATT CTT TCT CTG AAT ATT Rev: CAC AGG TCA AGC TTA TTA TTG ATC AAC AAA TAT AAC TGC ATT GG D0-D1 For: AAC CTG TAT TTT CAG AGC AAT GGT CAG GTG GTC GAC CA Rev: CAC AGG TCA AGC TTA TTA TGT AAA AAA TTC AAC TTC AGG TGC TTT pASK-IBA3 For: TAA TAA GCT TGA CCT GTG AAG TGA Rev: CAT TTT TTG CCC TCG TTA TCT AGA T TEV insertion For: CCTGTATTTTCAGAG C AAG CAG GAT ATT CTT TCT CTG AAT ATT Rev: CTC TGA AAA TAC AGG TTTTC GTG ATG GTG ATG GTG ATG CAT

pIBA3-His-TEV For: TAA TAA GCT TGA CCT GTG AAG TGA Rev: CTC TGA AAA TAC AGG ttttc GTG ATG GTG ATG GTG ATG CAT GSconnector For: GGT TCT GGC AGT GGT AGC GGC GTT GGG GTA ACG GAC TTT ACG Rev: CAC TGC CAG AAC C CGA CAG AAC GGT AAT AGT AAG CA pTac-promoter For: GCG AAG CTT GCC AGT GTG CTG GAA TTC G Rev: CCG GAT CCC CGG GAA TTC GTA ATC ATG GA EGFP For: CGC GGA TCC ATG GTG AGC AAG GGC GAG G Rev: GGG TCG ACT TAC TTG TAC AGC TCG TCC ATG C RedEx For: CGC GGA TCC ATG GCC TCC TCC GAG GAC G Rev: GGG TCG ACT TAC AGG AAC AGG TGG TGG CGG C 834

835

836

837 S6 Table: SAXS – experimental details

a) Sample details D00-D0 D0-D1 Organism E. coli O127:H6 E. coli O127:H6 E2348/69 (EPEC) E2348/69 (EPEC) Taxonomy ID: 574521 Taxonomy ID: 574521 -1 -1 Extinction coefficient (A280, M cm ) 14440 4470 Molecular mass M from chemical 21.1 20.9 composition (kDa) b) SAXS data collection Instrument P12 BioSAXS beamline (PETRAIII) Date 12th July 2019 Detector Pilatus6m Wavelength (nm) 0.123981 Beam size (mm2) 0.2 × 0.12 Detector distance (m) 4.0 q-measurement range 0.017- 5.506 (nm-1) Absolute scaling method Comparison with scattering from BSA Normalization To transmitted intensity by beam-stop counter Monitoring for radiation Frame-by-frame comparison damage Exposure time (s) 20 x 0.195 Sample configuration Quartz glass capillary Sample temperature (ºC) 20 c) Software employed for SAXS data reduction, analysis and interpretation SAS data reduction PRIMUSqt (Konarev et al., 2003) from ATSAS 2.8.3 (Franke et al., 2017) Extinction coefficient ExPaSy (Gasteiger et al., 2003) estimate Basic analyses: Guinier, PRIMUSqt (Konarev et al., 2003) p(r), VP Ab initio modelling GASBOR (Svergun, Petoukhov and Koch, 2001) 838

839

840 841 842 843

844 References 845 1. Roussel, C. et al. Increased EHEC survival and virulence gene expression indicate an 846 enhanced pathogenicity upon simulated pediatric gastrointestinal conditions. Pediatr. Res. 847 80, 734–743 (2016). 848 2. Nguyen, Y. & Sperandio, V. Enterohemorrhagic E. coli (EHEC) pathogenesis. Front. Cell. 849 Infect. Microbiol. 2, 1–7 (2012). 850 3. Soysal, N. et al. Enterohemorrhagic Escherichia coli hybrid pathotype O80:H2 as a new 851 therapeutic challenge. Emerg. Infect. Dis. 22, 1604–1612 (2016). 852 4. Kampmeier, S., Berger, M., Mellmann, A., Karch, H. & Berger, P. The 2011 German 853 Enterohemorrhagic Escherichia Coli O104:H4 Outbreak-The Danger Is Still Out There. 854 Curr Top Microbiol Immunol 117–148 (2018) doi:10.1007/82_2018_107. 855 5. Karmali, M. A. Factors in the emergence of serious human infections associated with highly 856 pathogenic strains of shiga toxin-producing Escherichia coli. Int. J. Med. Microbiol. 308, 857 1067–1072 (2018). 858 6. Hu, J. & Torres, A. G. Enteropathogenic Escherichia coli: Foe or innocent bystander? Clin. 859 Microbiol. Infect. 21, 729–734 (2015). 860 7. Wong, A. R. C. et al. Enteropathogenic and enterohaemorrhagic Escherichia coli: Even 861 more subversive elements. Mol. Microbiol. 80, 1420–1438 (2011). 862 8. Schmidt, M. A. LEEways: Tales of EPEC, ATEC and EHEC. Cell. Microbiol. 12, 1544–1552 863 (2010). 864 9. Jerse, A. E., Yu, J., Tall, B. D. & Kaper, J. B. A genetic locus of enteropathogenic 865 Escherichia coli necessary for the production of attaching and effacing lesions on tissue 866 culture cells. Proc. Natl. Acad. Sci. U. S. A. 87, 7839–7843 (1990). 867 10. Leo, J. C., Oberhettinger, P., Schütz, M. & Linke, D. The inverse autotransporter family: 868 Intimin, invasin and related proteins. Int. J. Med. Microbiol. 305, 276–282 (2015). 869 11. Frankel, G. et al. Intimin and the host cell - Is it bound to end in Tir(s)? Trends Microbiol. 9, 870 214–218 (2001). 871 12. Kenny, B. et al. Enteropathogenic E. coli (EPEC) transfers its receptor for intimate 872 adherence into mammalian cells. Cell 91, 511–520 (1997). 873 13. Oberhettinger, P. et al. Intimin and Invasin Export Their C-Terminus to the Bacterial Cell 874 Surface Using an Inverse Mechanism Compared to Classical Autotransport. PLoS One 7, 875 (2012). 876 14. Drobnak, I. et al. Of linkers and autochaperones: An unambiguous nomenclature to identify 877 common and uncommon themes for autotransporter secretion. Mol. Microbiol. 95, 1–16 878 (2015). 879 15. Oberhettinger, P., Leo, J. C., Linke, D., Autenrieth, I. B. & Schütz, M. S. The inverse 880 autotransporter intimin exports its passenger domain via a hairpin intermediate. J. Biol. 881 Chem. 290, 1837–1849 (2015). 882 16. Leo, J. C. et al. Secretion of the intimin passenger domain is driven by protein folding. J. 883 Biol. Chem. 291, 20096–20112 (2016).

884 17. Hamburger, Z. A., Brown, M. S., Isberg, R. R. & Bjorkman, P. J. Crystal Structure of Invasin: 885 A Bacterial Integrin-Binding Protein. Science (80-. ). 286, 291–295 (1999). 886 18. Luo, Y. et al. Crystal structure of enteropathogenic Escherichia coli intimin–receptor 887 complex. Nature 405, 1073–1077 (2002). 888 19. Fairman, J. W. et al. Crystal structures of the outer membrane domain of intimin and invasin 889 from enterohemorrhagic E. coli and enteropathogenic Y. pseudotuberculosis. Structure 20, 890 1233–1243 (2012). 891 20. Nagar, B., Overduin, M., Ikura, M. & Rini, J. M. Structural basis of calcium-induced E- 892 cadherin rigidification and dimerization. Nature 380, 360–364 (1996). 893 21. Griessl, M. H. et al. Structural insight into the giant Ca2+-binding adhesin siie: Implications 894 for the adhesion of salmonella enterica to polarized epithelial cells. Structure 21, 741–752 895 (2013). 896 22. Leo, J. C. et al. The Intimin periplasmic domain mediates dimerisation and binding to 897 peptidoglycan. Mol. Microbiol. 95, 80–100 (2015). 898 23. Bodelón, G., Palomino, C. & Fernández, L. Á. Immunoglobulin domains in Escherichia coli 899 and other enterobacteria: From pathogenesis to applications in antibody technologies. 900 FEMS Microbiol. Rev. 37, 204–250 (2013). 901 24. Bork, P., Holm, L. & Sander, C. The Immunoglobulin Fold. Journal of Molecular Biology 902 vol. 242 309–320 (1994). 903 25. Halaby, D. M., Poupon, A. & Mornon, J. P. The immunoglobulin fold family: Sequence 904 analysis and 3D structure comparisons. Protein Eng. 12, 563–571 (1999). 905 26. Li, H. & Fernandez, J. M. Mechanical design of the first proximal Ig domain of human 906 cardiac titin revealed by single molecule force spectroscopy. J. Mol. Biol. 334, 75–86 907 (2003). 908 27. Kabsch, W. & Sander, C. Dictionary of protein secondary structure: Pattern recognition of 909 hydrogenǦbonded and geometrical features. Biopolymers 22, 2577–2637 (1983). 910 28. Heinz, E. et al. Conserved Features in the Structure, Mechanism, and Biogenesis of the 911 Inverse Autotransporter Protein Family. Genome Biol. Evol. 8, 1690–1705 (2016). 912 29. Bassler, J., Hernandez Alvarez, B., Hartmann, M. D. & Lupas, A. N. A domain dictionary of 913 trimeric autotransporter adhesins. Int. J. Med. Microbiol. 305, 265–275 (2015). 914 30. El-Gebali, S. et al. The Pfam protein families database in 2019. Nucleic Acids Res. 47, 915 D427–D432 (2018). 916 31. Söding, J., Biegert, A. & Lupas, A. N. The HHpred interactive server for protein homology 917 detection and structure prediction. Nucleic Acids Res. 33, W244-8 (2005). 918 32. Zondlo, N. J. Aromatic-Proline interactions: Electronically tubale CH/pi interactions. Acc 919 Chem Res. 46, 1039–1049 (2013). 920 33. Persat, A. et al. The Mechanical World of Bacteria. Cell 161, 988–997 (2015). 921 34. Ross, N. T. & Miller, B. L. Characterization of the binding surface of the translocated intimin 922 receptor, an essential protein for EPEC and EHEC cell adhesion. Protein Sci. 16, 2677– 923 2683 (2007).

924 35. Nesta, B. et al. FdeC, a Novel Broadly Conserved Escherichia coli Adhesin Eliciting 925 Protection against Urinary Tract Infections. MBio 3, (2012). 926 36. Wrobel, A., Ottoni, C., Leo, J. C., Gulla, S. & Linke, D. The repeat structure of two 927 paralogous genes, Yersinia ruckeri invasin (yrInv) and a “Y. ruckeri invasin-like molecule”, 928 (yrIlm) sheds light on the evolution of adhesive capacities of a fish pathogen. J. Struct. Biol. 929 201, 171–183 (2018). 930 37. Wrobel, A. et al. The inverse autotransporters of Yersinia ruckeri , YrInv and YrIlm , 931 contribute to bio fi lm formation and virulence. Environ. Microbiol. (2020) doi:10.1111/1462- 932 2920.15051. 933 38. Tsai, J. C. et al. The Bacterial Intimins and Invasins: A Large and Novel Family of Secreted 934 Proteins. PLoS One 5, (2010). 935 39. Zakrisson, J., Wiklund, K., Axner, O. & Andersson, M. Helix-like biopolymers can act as 936 dampers of force for bacteria in flows. Eur. Biophys. J. 41, 551–560 (2012). 937 40. Kelly, G. et al. Structure of the cell-adhesion fragment of intimin from enteropathogenic 938 Escherichia coli. Nat. Struct. Biol. 6, 313–318 (1999). 939 41. Touzé, T., Hayward, R. D., Eswaran, J., Leong, J. M. & Koronakis, V. Self-association of 940 EPEC intimin mediated by the β-barrel-containing anchor domain: A role in clustering of 941 the Tir receptor. Mol. Microbiol. 51, 73–87 (2004). 942 42. Ciatto, C. et al. T-cadherin structures reveal a novel adhesive binding mechanism. Nat. 943 Struct. Mol. Biol. 17, 339–347 (2010). 944 43. Lee, E. H., Hsin, J., Von Castelmur, E., Mayans, O. & Schulten, K. Tertiary and secondary 945 structure elasticity of a six-lg titin chain. Biophys. J. 98, 1085–1095 (2010). 946 44. Gibson, D. G. et al. Enzymatic assembly of DNA molecules up to several hundred 947 kilobases. Nat. Methods 6, 343–345 (2009). 948 45. Bertani, G. Studies on lysogenesis. I. The mode of phage liberation by lysogenic 949 Escherichia coli. J. Bacteriol. 62, 293–300 (1951). 950 46. Liu, H. & Naismith, J. H. An efficient one-step site-directed deletion, insertion, single and 951 multiple-site plasmid mutagenesis protocol. BMC Biotechnol. 8, 91 (2008). 952 47. Wu, X. et al. A novel method for high-level production of TEV protease by superfolder GFP 953 tag. J. Biomed. Biotechnol. 2009, (2009). 954 48. McCoy, A. J. et al. Phaser crystallographic software. J. Appl. Crystallogr. 40, 658–674 955 (2007). 956 49. Adams, P. D. et al. PHENIX: A comprehensive Python-based system for macromolecular 957 structure solution. Acta Crystallogr. Sect. D Biol. Crystallogr. 66, 213–221 (2010). 958 50. Murshudov, G. N., Vagin, A. A. & Dodson, E. J. Refinement of macromolecular structures 959 by the maximum-likelihood method. Acta Crystallogr. Sect. D Biol. Crystallogr. 53, 240– 960 255 (1997). 961 51. Blanchet, C. E. et al. Versatile sample environments and automation for biological solution 962 X-ray scattering experiments at the P12 beamline (PETRA III, DESY). J. Appl. Crystallogr. 963 48, 431–443 (2015).

964 52. Franke, D. et al. New developments in the ATSAS program package for small-angle 965 scattering data analysis. J. Appl. Crystallogr. 45, 342–350 (2012). 966 53. Svergun, D. I., Petoukhov, M. V & Koch, M. H. Determination of domain structure of 967 proteins from X-ray solution scattering. Biophys. J. 80, 2946–2953 (2001). 968 54. Tria, G., Mertens, H. D. T., Kachala, M. & Svergun, D. I. Advanced ensemble modelling of 969 flexible macromolecules using X-ray solution scattering. IUCrJ 2, 207–217 (2015). 970 55. Bernadó, P., Mylonas, E., Petoukhov, M. V., Blackledge, M. & Svergun, D. I. Structural 971 Characterization of Flexible Proteins Using Small-Angle X-ray Scattering. J. Am. Chem. 972 Soc. 129, 5656–5664 (2007). 973 56. Li, D. W. & Brüschweiler, R. NMR-based protein potentials. Angew. Chemie - Int. Ed. 49, 974 6778–6780 (2010). 975 57. Mongan, J., Simmerling, C., McCammon, J. A., Case, D. A. & Onufriev, A. Generalized 976 born model with a simple, robust molecular volume correction. J. Chem. Theory Comput. 977 3, 156–169 (2007). 978 58. Srinivasan, J., Trevathan, M. W., Beroza, P. & Case, D. A. Application of a pairwise 979 generalized Born model to proteins and nucleic acids: Inclusion of salt effects. Theor. 980 Chem. Acc. 101, 426–434 (1999). 981 59. Ryckaert, J. P., Ciccotti, G. & Berendsen, H. J. C. Numerical integration of the cartesian 982 equations of motion of a system with constraints: molecular dynamics of n-alkanes. J. 983 Comput. Phys. 23, 327–341 (1977). 984 60. McGibbon, R. T. et al. MDTraj: A Modern Open Library for the Analysis of Molecular 985 Dynamics Trajectories. Biophys. J. 109, 1528–1532 (2015). 986 61. Stenqvist, B., Thuresson, A., Kurut, A., Vácha, R. & Lund, M. Faunus-A flexible framework 987 for Monte Carlo simulation. Mol. Simul. 39, 1233–1239 (2013). 988 62. Chandler, D., Weeks, J. D. & Andersen, H. C. Van Der Waals Picture of Liquids , Solids , 989 and Phase Transformation. Science (80-. ). 220, 787–794 (2014). 990 63. Köberle, M. et al. Yersinia enterocolitica Targets Cells of the Innate and Adaptive Immune 991 System by Injection of Yops in a Mouse Infection Model. PLOS Pathog. 5, e1000551 992 (2009).

993