Search

The type three secretion system (T3SS) is a bacterial weapon found in many Gram-negative bacteria. It consists out of a needle like structure called injectisome, which enables bacteria to inject effector proteins into eukaryotic cells. The injectisome is a very complex molecular machine, inserted in the inner and outer bacteria membrane, passing the periplasm with the peptidoglycan layer. We used an approach with fluorescent hybrid proteins to study the assembly order of the injectisome in Yersinia enterocolitica. First the outer membrane protein YscC was genetically fused to the red fluorescent protein mCherry. This construct was able to build fluorescent foci in the bacterial membrane by itself, without any other components of the T3SS. Then we engineered fusions between the green fluorescent protein EGFP and the inner membrane protein YscD, the putative C-ring protein YscQ or the ATPase YscN. All three constructs showed fluorescent foci at the bacterial membrane. Comparison of the different EGFP constructs with the YscC-mCherry construct in double mutants showed that the proteins co-localize. Thus we considered the foci to be a read out for injectisome assembly. By combining the EGFP constructs with different deletion mutants we found that the assembly occurs from the outside to the inside. Starting with the outer membrane protein YscC to the inner membrane. Then the ATPase and C-ring assemble together and finally the needle is formed. Different single particle structures of T3SS, needle complexes, purified from Shigella flexneri and Salmonella enterica are available. But in the course of purification the inner membrane export apparatus, the ATPase complex and the C-ring were lost. As well no such complex has been purified from Y. enterocolitica. Thus we investigated the Y. enterocolitica injectisomes in situ by cryo electron tomography (cryo-ET). Unfortunately, the 1-?m diameter of Y. enterocolitica is too large to obtain optimal resolution with cryo-ET. Thus we engineered a minD mutant that forms so called minicells, due to asymmetric septum placement. We collected tomograms of particles from wild type and minicells and constructed an average structure with a resolution of 3.7 nm. In addition the 6 nm resolution in situ structure of the injectisome of S. flexneri was made. We saw significant stretching of the in situ structure compared to the isolated particles. Moreover we saw flexibility of the basal body. We can only speculate that such flexibility might increase the stability of the structure and protect it of mechanical forces. In addition for the first time we could visualize a mass in cytoplasm just below the middle of the injectisome. Due to homology to the flagellum we can assume, that this is the ATPase. But to conclusively assign proteins to masses seen in the in situ structure we would need to label them. A question mark in the Y. enterocolitica injectisome assembly is, how the structure can pass the peptidoglycan layer. The flagellum, which is evolutionary closely related to the injectisome, as well as other types of injectisomes have a muramidase or more specifically lytic transglycosylase (LT) encoded within their loci. No gene encoding for a LT, can be found on the pYV plasmid that encodes otherwise for the entire T3SS. Thus we tested several genomic LT for their involvement in the assembly of the T3SS. It is possible that the injectisome assembles through temporary gaps generated during the synthesis of new peptidoglycan strands. This theory is very intriguing, as the pattern of the fluorescent foci resembled closely the arrangement of the bacterial cytoskeleton protein MreB, which is assumed to be involved in placing the peptidoglycan synthesis machinery. Comparing the localization of MreB and the injectisome showed that both constructs seem to be arranged on two different helical paths. This convinced us that injectisome arrangement is not stochastic but rather controlled. To find the underlying structure responsible for this arrangement, we compared the localization of other proteins with similar arrangement as MreB. In Bacillus subtilis different membrane compositions were shown to be helically distributed. Unfortunately staining of the inner membrane is difficult in Gram-negative bacteria. Therefore we compared injectisome location with a potential bacterial lipid raft marker, which showed injectisomes do not insert into the lipid rafts. Thus although our knowledge about the assembly, the structure and the function of the T3SS is improving enormously, the question of how injectisomes are localized remains to be answered.

NF-κB is a transcription factor involved in the regulation of inflammation and innate immunity. The IκB kinase (IKK) complex contains two catalytic subunits, IKKα and IKKβ, and plays an essential role in the activation of NF-κB through the phosphorylation and degradation of the NF-κB inhibitor IκBα, thereby allowing translocation of NF-κB into the nucleus. Numerous evidences indicate that IKKβ mediates NF-κB activation in response to pro-inflammatory cytokines and microbial products, but the role of IKKα in inflammation and innate immunity is unknown. In the first part of dissertation, we focus on understanding the previously unknown function of IKKα in the canonical NF-κB pathway, associated with inflammation and innate immunity. We show that silencing of IKKα by RNA interference (RNAi) significantly reduced phosphorylation and degradation of IκBα, and nuclear translocation of NF-κB, and secretion of the pro-inflammatory chemokine interleukin-8 (IL-8) during Shigella flexneri infection of human epithelial HeLa cells. This suggests that IKKα like IKKβ plays a pivotal role in inflammation and innate immunity by mediating NF-κB activation in response to microbial infection. Proper control of NF-κB activation is essential for inflammation and innate immunity triggered by microbial infection, but the dysregulation of NF-κB is associated with various diseases such as chronic inflammatory diseases and cancers. Thus, the NF-κB pathway has been a target of therapeutic drug development. Although constitutive and excessive NF-κB activation has been detected in many inflammation-related diseases, the cause of the constitutive NF-κB activation in non-stimulated cells is largely unknown. In the second part of dissertation, we focus on clathrin heavy chain (CHC), a well-known regulator of endocytosis that plays a novel endocytosis-independent function as an inhibitor of basal NF-κB activation. We show that silencing of CHC induced constitutive NF-κB nuclear translocation and high level of IL-8 secretion in resting cells. We revealed that constitutive NF-κB nuclear translocation was mediated through the constant IκBα degradation in an IKKα-dependent mechanism. We further showed that CHC depletion-induced constitutive IκBα degradation and high level of IL-8 secretion in resting cells was independent of the inhibition of clathrin-mediated endocytosis (CME) as silencing of μ2 subunit of AP2 complex (AP2M1), an adaptor protein essential for CME failed to induce the constitutive IκBα degradation and high level of IL-8 secretion. Therefore, the results presented may suggest a potential link between a defect in CHC expression and chronic inflammatory disorders and cancers.

Many pathogenic bacteria harbor a type III secretion system to translocate effector proteins from the bacterium into the host cell cytosol. This one-step translocation requires a nanomachinery which was termed injectisome. It consists of a basal body, spanning the two bacterial membranes, and a needle-like structure, bridging the distance between the bacterium and the target cell. Control of the length of the type III secretion injectisome needle is crucial for a correct function. In Yersinia, the YscP protein is involved in needle length control: the number of YscP residues directly correlates with needle length. In this thesis, this correlation was shown to be dependent on the secondary structure of YscP. By substitution of individual residues, needle length could be altered without changing the number of residues in YscP. The molecular ruler model was proposed for length control of the Yersinia injectisome needle. There are, however, two possibilities for the molecular ruler model regarding the amount of YscP needed for regulation of the needle length of one injectisome. In the static model, only one molecule of YscP and in a more dynamic model, several proteins are required for length control of one needle. Here, it was demonstrated that partially diploid bacteria, expressing a short and a long YscP simultaneously assemble distinct sets of short and long needles. These results suggest that only one YscP molecule is required for length control of one needle. In Yersinia, the YscU protein (a member of the export machinery) was suggested to be involved in the substrate specificity switch. Here, YscU was demonstrated to play a role in substrate recognition but not in substrate switching. Taken together, a refined model for length control of the Yersinia injectisome needle is proposed in this thesis, confirming the role of YscP as a molecular ruler.

We investigated the assembly of the Yersinia enterocolitica type III secretion injectisome by grafting fluorescent proteins onto several components, YscC (outer-membrane (OM) ring), YscD (forms the inner-membrane (IM) ring together with YscJ), YscN (ATPase), and YscQ (putative C ring). The recombinant injectisomes were functional and appeared as fluorescent spots at the cell periphery. Epistasis experiments with the hybrid alleles in an array of injectisome mutants revealed a novel outside-in assembly order: whereas YscC formed spots in the absence of any other structural protein, formation of YscD foci required YscC, but not YscJ. We therefore propose that the assembly starts with YscC and proceeds through the connector YscD to YscJ, which was further corroborated by co-immunoprecipitation experiments. Completion of the membrane rings allowed the subsequent assembly of cytosolic components. YscN and YscQ attached synchronously, requiring each other, the interacting proteins YscK and YscL, but no further injectisome component for their assembly. These results show that assembly is initiated by the formation of the OM ring and progresses inwards to the IM ring and, finally, to a large cytosolic complex. Further, we analyzed the assembly kinetics and dynamics of the cytosolic components of the Yersinia injectisome in detail, and showed that the needle subunit as well as parts of the IM export machinery can be copurified with the membrane rings using a specific combination of crosslinking and affinity purification.

Neutrophils are among the most important components of the innate immune response, which provides the first line of host defense. The antimicrobial potential of neutrophils has been traditionally divided into either non-oxidative or oxidative mechanisms. Two of the most important antimicrobial systems of these mechansims are granule-associated antimicrobial proteases and peptides and the nicotinamid adenine dinucleotide phosphate (NADPH) oxidase generating reactive oxygen species (ROS). In the past, studies often focused on the effects of either non-oxidative or oxidative mechanisms and decades of research have provided a detailed understanding of the regulation, generation and actions of these processes alone. Recent evidence challenged the established view of two independent mechanisms and proposed the cooperation of the NADPH oxidase with granule proteases in the killing of microorganisms. Furthermore, a novel phagocytosis-independent antimicrobial mechanism was found by the discovery of neutrophil extracellular traps (NETs). The formation of NETs was found dependent on NADPH oxidase activation and the production of ROS but NETs are believed to kill entrapped pathogens by NETsassociated granule proteases and peptides. Consequently, NETs present additional evidence for the interaction of non-oxidative and oxidative killing mechanisms. Together, these findings opened a new field of investigation with many controversies to be elucidated and underscored that we need further insight into the mechanisms by which neutrophils specifically recognize and overcome pathogens. In this thesis, we followed the question whether the murine antimicrobial peptide cathelin-related antimicrobial peptide (CRAMP) is an important component of the non-oxidative arm of neutrophil defense against S. aureus. This was motivated because little is known about cathelicidin function and activity in neutrophils and seems of crucial interest since mice lack the major constituent of human neutrophils - the -defensins. We further aimed to specify the relationship between the NADPH oxidase and CRAMP with focus on the antimicrobial activity of CRAMP in association with NETs and in NADPH oxidase-deficient mice (gp91phox-/-). As a result, we could demonstrate a previously unknown intracellular antimicrobial activity of CRAMP against S. aureus. Specifically, CRAMP colocalized with S. aureus in phagolysosomes and we showed first evidence for the presence of intracellular active CRAMP. Most interestingly, phagolysosomal localization and intracellular activity of CRAMP was found independent of a functional NADPH oxidase controversially to our expectations. Investigation of NET-dependent killing of S. aureus revealed a negligible role for CRAMP due to inactivation of the peptide in association with NETs. This point is of particular relevance and should be considered in the current opinion of NETs-mediated antimicrobial activity. In summary, our data provided deeper knowledge about one specific member of the non-oxidative killing mechanism and gives reason to reconsider the controversial results about interaction of NADPH oxidase in activating non-oxidative mechanisms. In addition, we followed the question whether recognition of S. aureus by TLR2 regulates the induction of non-oxidative and oxidative killing responses as well as the induction of NETs. The background of this study is based on several previous reports. First, results of our group evidenced a relationship between TLR2-mediated staphylococcal killing by neutrophils and the susceptibility of S. aureus to cationic antimicrobial peptides. Additionally, TLR2 activation has been shown to up-regulate 7 cathelicidin expression. Second, TLR2 was demonstrated to induce phosphorylation of p47phox and up-regulation of p47phox mRNA in macrophages. However, little attention has been paid to similar studies in neutrophils. Third, recognition of pathogens by TLRs was hypothesized to induce formation of NETs but there is as yet no evidence. First results unraveled a role for TLR2 in rapid induction of the NADPH oxidase, whereas TLR2 signaling had no influence on CRAMP activity. Interestingly, pathogen sensing for the induction of NETs formation did not depend on TLR2- MyD88 signaling. Taken together, the results demonstrate a role for TLR2 in mediating rapid killing of S. aureus by accelerating the activation of the NADPH oxidase complex possibly by influencing assembly. Further studies of the mechanisms underlying the relationship between pathogen sensing and non-oxidative and oxidative killing mechanisms would contribute greatly to our understanding of how the innate immune system resolves bacterial infections and will help in the development of therapeutic strategies to assist in clearance of pathogenic bacteria.

We show that Capnocytophaga canimorsus strain 5 (Cc5) is even more resistant to phagocytosis and killing by murine macrophages (J774.1) and human polymorphonuclear neutrophils (PMNs) than Yersinia enterocolitica, which is known as a model bacterium for resistance against phagocytosis due to its type 3 secretion system (Grosdent et al., 2002). We observed that Cc5 even becomes completely resistant to phagocytosis at high multiplicity of infection (moi of 50). In addition, we demonstrate that the Cc5 transposon mutant Y1C12, identified during a serum sensitivity screen, has an increased sensitivity to phagocytosis and killing by either murine macrophages or human PMNs even in the unopsonized state. This indicated that not an increased susceptibility for antibody binding or complement deposition led to an increased phagocytosis of the mutant, but that rather the outer surface was more readily recognized by the phagocytes. Furthermore, we demonstrate that Cc5 induces the formation of neutrophil extracellular traps upon infection of human PMNs in vitro and that Cc5 is trapped and killed within neutrophil extracellular traps, indicating sensitivity of Cc5 towards antimicrobial peptides present in PMN granules. Analysis of serum resistance in Cc5 revealed that serum resistance is probably linked to its lipopolysaccharide, which prevents deposition of the membrane attack complex on the bacterial surface. Moreover, we have observed that upon growth in the presence of cells, Cc5 releases or modifies factor(s) in the medium, which interfere with the killing ability of macrophages. Investigating the underlying mechanism, we could show that Cc5 does not affect phagosome maturation, but blocks the oxidative burst. This capacity was shown to depend on the release of the zinc metallopeptidase pitrilysin by Cc5. First analyses on the prevalence of the hypothetical virulence factors serum resistance and interference with the oxidative burst indicated that C. canimorsus strains might display strain variability. While 59% of the strains (50% of case strains, 61% of dog isolates) were able to block the killing ability of macrophages, 60% of the strains were highly serum resistant (100% of case strains, 54% of dog isolates). However, serum resistance could not be directly linked to a specific polysaccharide structure in C. canimorsus. November 2009, Salome Casutt-Meyer

We show that C. canimorsus (Cc) can serve as a recipient for RP4 mediated conjugation but there is neither replication of broad host range plasmid vectors nor expression of commonly used E. coli markers in C. canimorsus. We identified three selection markers, ermF, tetQ and cfxA leading to resistance against erythromycin, tetracycline and cefoxitin, respectively, that can be used in C. canimorsus. We engineered expression shuttle vectors using the replicon of a endogenous plasmid found in strain Cc7 and the promoter of one of the selection markers for gene expression. We developed a transposon mutagenesis strategy based on Tn4351 from Bacteroides fragilis and protocols for allelic exchange and electrotransformation. We carried out an extensive transposon mutagenesis and screened these mutants for different properties. We demonstrate that presence of mammalian cells, including phagocytes, favors growth of C. canimorsus 5 and this property was found to be dependent on direct cellular contacts. We isolated a Tn mutant unable to grow in presence of mammalian cells. The mutation occurred in a gene encoding a sialidase. The surface-exposed sialidase allows Cc5 to feed on internal aminosugars of glycan chains from host cell glycans. In addition, sialidase confers resistance to complement by promoting the binding of factor H. We developed an experimental mouse infection in which the read-out is bacterial persistence. In this infection model, Cc5, but not the sialidase deficient mutant, grew and persisted, showing the importance of this metabolic pathway in vivo. C. canimorsus by itself does not elicit the onset of an inflammatory response from macrophages. One strain, Cc5 turned out to have a mechanism that actively blocks the pro-inflammatory signaling of macrophages upon stimulation with endotoxic LPS. We screened the Tn mutant library for clones of Cc5 affected in this active mechanism. Isolated mutants have been mapped, characterized and complemented. The function of the mutated genes is presently under investigation as well as the mode of action of its gene product(s). The prevalence of C. canimorsus in dogs has not been clarified at present. We therefore sampled dog swabs to isolate C. canimorsus strains in Swiss dogs. We could identify 61 C. canimorsus isolates from 103 dogs, which represents 59.22% of the dogs tested. Besides this I also contributed to the analysis of LPS, to the study of resistance of Cc5 to complement mediated lysis, to sequencing of the genome, the assembly of the reads and the annotation.

Type III secretion systems are used by many pathogenic bacteria to deliver effector proteins into host cells. The secretion machinery, also known as the injectisome, is composed of a basal body anchored in the bacterial membranes and an external needle. Delivery of the effectors across the host cell membrane requires, in addition, a set of proteins called the translocators that are exported by the injectisome itself. The interaction between the needle of the injectisome and the translocators is currently not understood in detail. In Yersinia, the translocators are a group of three proteins called YopB, YopD and LcrV. The two hydrophobic proteins YopB and YopD form a pore in the target cell membrane, while LcrV assists the assembly of the pore. LcrV is known since the mid-fifties as the major protective antigen against plague. Using Scanning Transmission Electron Microscopy (STEM) we showed that LcrV forms a distinct structure at the tip of the injectisome needle, the tip complex. The tip complex could be specifically labelled with antibodies directed against LcrV and could be crosslinked to the YscF needle, demonstrating that LcrV forms the tip complex. The unique localization of LcrV at the tip of the injectisome needle explains its crucial role in the translocation process and its role as the main protective antigen against plague. The orientation of LcrV within the tip complex was determined by analysis of tips formed by hybrids between LcrV and its orthologues in Pseudomonas and Aeromonas, PcrV and AcrV, respectively. The N-terminal globular domain and the C-terminus of LcrV were found to form the base structure of the tip complex. By quantitative immunoblot analyses and STEM, we determined that the needle of the Yersinia enterocolitica E40 injectisome consists of 139+/-19 YscF subunits and that the tip complex is formed by three to five LcrV monomers. A pentamer represented the best fit for an atomic model of the tip complex. In addition, the N-terminal globular domain of the V-antigen is crucial for pore formation as it mediates the interaction with the translocator YopB.

Adaptation of bacteria to their environment is a necessity to keep their fitness and survival under limiting conditions. In chapter I of this work we demonstrated that adaptation of glycopeptide-resistant S. aureus (GISA) happens at the cost of resistance loss. An in vitro generated step selected teicoplanin-resistant mutant was phenotypically and genotypically characterized and compared to its glycopeptide-susceptible parent S. aureus strain SA113, prior to and after challenging the host in a foreign body infection model. Slower growth, increased thickness of the bacterial cell wall, increased N-acetylglucosamine incorporation, decreased hemolysis, and downregulation of some virulence-associated genes were determined upon in vitro step selection. In vivo host pressure together with the absence of antibiotics were shown to impair this GISA phenotype and to result in the selection of fitter variants at the cost of resistance loss. In a further study we compared the virulence and pathogenicity of isolates of two epidemic ST8 and ST5 health-care associated methicillin-resistant S. aureus (MRSA) clones with MSSA isolates of variable STs that are prevalent in French clinical settings and with MSSA of the same ST in a murine sepsis model. Mouse mortality, induced by the different groups of MRSA ST8 and ST5 and MSSA ST8, ST5 and var. ST was related to phenotypic and genotypic features. MRSA isolates of both clones ST8 and ST5 caused higher mortality than MSSA expressing var. ST. However, higher mortality was not related to the presence of SCCmec, since only ST5 but not ST8 isolates showed a higher virulence in the presence of this resistance determinant. Despite the same genetic background individual clonal MRSA isolates had heterogeneous virulence properties. No specific virulence factor determined in vitro, was found related to mortality in mice. In conclusion, in a bacteraemic model, lethality varied with the sequence type, and was modulated by SCCmec. Within the same ST, clonality was not associated with a homogenous outcome in the murine sepsis model. Whole genome sequencing of clonal lethal and non-lethal isolates will allow identifying virulence factors relevant to the outcome in this model. In the third part of this thesis we investigated phenotypic properties of the clonal MRSA ST8 isolates that had a variable outcome in the murine sepsis model and measured innate and adaptive immune responses of 7 lethal and 3 non-lethal isolates. Clonality of the 10 MRSA ST8 isolates was certified by MLVA. In vitro assays, simulating steps that occur during the course of infection were performed in order to detect host response discriminating differences between lethal and non-lethal isolates. While chemotaxis, bactericidal activity and apoptosis of PMN were similar for lethal and non-lethal isolates, a preliminary study yielded a different specific immune response induced by a lethal compared to a non-lethal isolate. The role of the adaptive immunity in the outcome of an infection with lethal and non-lethal isolates will be investigated in further.

Bacteria are able to switch between two mutually exclusive lifestyles, motile single cells and sedentary multicellular communities, known as biofilms, that colonize surfaces. Recent studies demonstrated that the global bacterial second messenger c-di-GMP orchestrates the developmental transition between both lifestyles. In a wide variety of bacterial species high intracellular c-di-GMP levels provoke excretion of protective and adhesive exopolymeric substances and inhibit flagella and pili based cell motility. Synthesis and degradation of c-di-GMP is catalyzed by diguanylate cyclases (DGC’s) and c-di-GMP-specific phosphodiesterases (PDE), respectively. Although the enzymes responsible for the synthesis of c-di-GMP have been recently identified, little information is available on general regulatory principles of the c-di-GMP signaling circuitry. Here we present genetic and biochemical approaches in combination with structural analysis to elucidate the molecular mechanisms of signal transduction, signal modulation and signal inactivation. In (Christen and Christen et al 2007, PNAS) we describe the isolation of several c-di-GMP binding proteins from Caulobacter crescentus by affinity chromatography. One of these proteins, DgrA, is a PilZ homolog involved in mediating c-di-GMP-dependent control of C. crescentus cell motility. Biochemical and structural analysis of DgrA and homologs from C. crescentus, Salmonella typhimurium and Pseudomonas aeruginosa identified this protein family as the first class of specific diguanylate receptors. Our studies suggested a general mechanism for c-di-GMP binding and signal transduction whereby increased concentrations of c-di-GMP are sensed by DgrA through direct binding and induce conformational changes of the diguanylate receptor that block motility by interfering with motor function rather than flagellar assembly. In (Christen and Christen et al 2006, JBC) we demonstrate that an allosteric binding site for c-di- GMP (I-site) is responsible for non-competitive product inhibition of DGC’s. The I-site was mapped in both multi- and single domain DGC proteins and shown to be fully contained within the GGDEF domain itself. In vivo evolution experiments combined with kinetic analysis of the obtained I-site mutants led to the definition of an RXXD motif as the core allosteric binding site for c-di-GMP. Based on these results and based on the observation that the I-site is conserved in a majority of known and potential DGC proteins, we propose that product inhibition of DGC’s is of fundamental importance for c-di-GMP signaling and cellular homeostasis. The definition of the I-site binding pocket provides an entry point into unraveling the molecular mechanisms of ligand-protein interactions involved in c-di-GMP signaling, makes DGC's a valuable target for drug design and offers new strategies against biofilm-related diseases. In (Christen et al 2005, JBC) we show biochemically that CC3396, a GGDEF-EAL composite protein from C. crescentus, is a soluble PDE. The PDE activity, rapidly converts c-di-GMP into the linear dinucleotide pGpG is confined to the C-terminal EAL domain of CC3396, depends on the presence of Mg2+ ions and is strongly inhibited by Ca2+ ions. Remarkably, the associated GGDEF domain, which contains an altered active site motif (GEDEF), lacks detectable DGC activity. Instead, this domain is able to bind GTP and in response activates the PDE activity in the neighboring EAL domain. PDE activation is specific for GTP (KD 4 μM) and operates by lowering the KM for c-di-GMP of the EAL domain to a physiologically significant level (420 nM). Mutational analysis suggested that the substrate-binding site (A-site) of the GGDEF domain is involved in the GTP-dependent regulatory function, arguing that a catalytically inactive GGDEF domain has retained the ability to bind GTP and in response can activate the neighboring EAL domain. Based on this we propose that the c-di-GMP-specific PDE activity is confined to the EAL domain, that GGDEF domains can either catalyze the formation of c-di-GMP or can serve as regulatory domains and that c-di-GMP-specific phosphodiesterase activity is coupled to the cellular GTP level in bacteria. In addition to the contribution in understanding the c-di-GMP signaling circuitry we characterized in (Stephens et al 2007, JBac) the metabolic enzymes and regulators of D-xylose catabolism in C. crescentus by genetic and biochemical methods. A saturated transposon screen was used to define the xylXABCD operon consisting of five genes, essential for xylose degradation. Subsequently biochemical and bioinformatical approaches were applied to provide enzymatic functions and predict possible conversion pathways for xylose catabolism. We demonstrated that the xylXABCD operon is tightly control via a LacI like repressor and defined determinants of the xylose operator, critical for negative control of xylXABCD transcription.

During its biphasic life cycle Caulobacter crescentus switches from a planktonic to surface attached life style. This transition requires the continuous remodeling of the cell poles through the temporally and spatially coordinated assembly and disassembly of polar organelles like the flagellum, pili, and an adhesive holdfast. A genetic screen for mutants affected in surface binding and colonization led to the identification of various genes required for motility, pili, and holdfast biogenesis, suggesting a specific role for all three organelles in C. crescentus surface colonization. Several novel holdfast genes were identified, which are potentially involved in the synthesis and regulation of the polysaccharidic component of the holdfast. Quantitative surface binding studies during the C. crescentus cell cycle revealed that optimal attachment coincides with the presence of flagellum, pili, and holdfast at the same pole. This indicated that accurate temporal control of polar appendices is critical for surface colonization of C. crescentus and represents the first example for developmentally controlled bacterial surface adhesion. We have used genetic and biochemical analyzes to demonstrate that di-cyclic guanosine monophosphate (c-di-GMP) is a central regulatory compound involved in the timing of C. crescentus pole development. Mutants lacking the diguanylatecyclase PleD show a dramatic delay of holdfast formation during swarmer cell differentiation. In contrast, cells lacking the GGDEF-EAL composite protein CC0091 show premature holdfast formation, while overexpression of CC0091 also leads to a delayed appearance of holdfast. The observation that CC0091 is a c-di-GMP specific phosphodiesterase indicated that the antagonistic activities of PleD and CC0091 could be responsible for the correct timing of holdfast formation and flagellum ejection. Finally, our genetic screen identified a candidate for the c-di-GMP effector protein, which mediates holdfast synthesis in response to fluctuating levels of c-di-GMP. The glycosyltransferase CC0095 is strictly required for holdfast formation and its overexpression leads to premature holdfast synthesis. This and the observation that CC0095 is able to bind c-di-GMP lead to the hypothesis that holdfast synthesis is regulated via allosteric control of the CC0095 glycosyltransferase. These data provide the first example of a developmental process being regulated by the bacterial second messenger, c-di-GMP.

The facultative intracellular bacterium Bartonella henselae enters human endothelial cells either passively by conventional phagocytosis or actively by a pathogen-triggered process known as invasome-mediated internalization. The latter involves the formation of a cell-surface-associated bacterial aggregate, which is subsequently engulfed by host cell membranes eventually resulting in its complete internalization. Recent work indicated that invasome formation of B. henselae depends on its VirB/VirD4 T4SS. This work describes that the VirB/VirD4 T4SS of B. henselae injects a cocktail of seven effector proteins into endothelial host cells to subvert cellular functions and that one of these translocated effector proteins, BepG, mediates the sustained cytoskeletal changes leading to invasome formation. Moreover, this work indicates the existence of two non-redundant pathways to promote invasome formation, one of which is governed by the action of BepG and another one, involving further Bep proteins, which is BepG-independent. On the host cell side, Rho-family small GTPases Cdc42 and Rac1, but not RhoA are shown to be required for invasome-mediated internalization. Furthermore, it is shown that invasome formation leads both to the rearrangement of pre-existing F-actin fibers and to localized actin polymerization enriched for Arp2/3, which occurs in a Scar1/WAVE-dependent manner. Finally, this work provides evidence that after complete internalization the invasome membranes do not fuse with Lamp-1 positive lysosomes, indicating that invasome-mediated invasion represents a novel mechanism allowing the uptake of bacteria without entering the endocytic-lysosomal pathway. Importantly, this is the first report, which attributes the orchestrated action of more than one effector protein of B. henselae to a known VirB/VirD4 T4SS-dependent phenotype, namely invasome formation, which represents a multifacetted example for the complexity of host cell subversion by a bacterial pathogen.

Many pathogenic bacteria use type III secretion systems (TTSS) to deliver effector proteins into the cytosol of target cells to subvert host cell functions. The actual secretion apparatus, called injectisome, consist of a basal body embedded in the bacterial membranes and a needle. The needle is thought to serve as a conduit for protein secretion. However, to cross the target cell membrane an additional translocation step is necessary. This translocation involves the formation of a pore in the target cell membrane, which is presumably connected to the needle. Three proteins are required for the assembly of this pore. In Yersinia, the three “translocators” are YopB and YopD, two putative membrane proteins, and LcrV a hydrophilic protein. LcrV is also known, since the mid-fifties, to represent the major protective antigen against plague infections. The aim of my thesis was to characterize the structure and function of the translocators. Infection of erythrocytes with wildtype Yersiniae causes hemolysis due to the formation of the translocation pore in the red blood cell membrane. We showed that the isolated membranes of these erythrocytes contain the hydrophobic translocators YopB and YopD, but not LcrV. Bacteria deprived of LcrV did not assemble a functional pore, but were still able to insert reduced amounts of YopB and YopD into the target cell membrane. This is in agreement with reports showing that purified YopB and YopD can oligomerize and insert into artificial membranes independently of LcrV. We showed further that polyclonal antibodies directed against LcrV interfere with the formation of a functional translocation pore by live bacteria. Based on these results, we hypothesized that LcrV acts as a platform or scaffold onto which the YopBD translocation pore assembles (Goure, Broz et al. 2005, Journal of Infectious Diseases 192:218-25). We purified needles and detected LcrV as well as YscF, the needle subunit, in these preparations. In parallel we analyzed these purified needles by STEM (scanning transmission electron microscopy) and found that the needle ends with a defined tip complex, that comprises a head, a neck and a base. We then showed that the tip complex is missing in lcrV mutant bacteria and can be restored after the mutation is complemented in trans. These results indicated that LcrV is involved in the formation of the tip complex. In addition, crosslinking of purified needles indicated that LcrV and YscF interact and thus the V-antigen might form the tip complex. Immunolabelling of wildtype needles with anti-LcrV antibodies showed a strong binding to the tip complex, anti-YscF antibodies bound to the bottom of the needle. Together these results demonstrate that LcrV forms the observed tip complex and explain why anti-LcrV antibodies can inhibit pore formation. In addition, these data reinforce the assembly platform hypothesis (Mueller, Broz et al. 2005, Science 310: 674-676). P. aeruginosa and A. salmonicida possess an injectisome closely related to that of Yersinia. Their respective LcrV orthologs, PcrV (32.3 kDa) and AcrV (40.2 kDa) are slightly different in size to LcrV (37.6 kDa). We demonstrated that PcrV as well as AcrV can functionally complement a lcrV deletion in Y. enterocolitica . The needles exhibited distinct tip complexes similar to those of wildtype needles but they were smaller in the case of PcrV and larger with AcrV (Mueller, Broz et al. 2005, Science 310: 674-676). Hybrids between the three proteins LcrV, PcrV and AcrV could complement an lcrV deletion in Y. enterocolitica in the hemolysis assay, but the level of complementation varied. The amino-terminus seemed to play an important role in the function of the protein. STEM analysis of tip complexes formed by different hybrid proteins allowed us to show that the aminoterminal domain of LcrV forms the base while the second globular domain forms the head of the tip complex. In addition we determined the stoichiometry of YscF and LcrV in purified needles and found that between three to six molecules of LcrV form the tip complex. Together, these results allowed us to propose an atomic modeling of the LcrV tip complex on top of the injectisome needle.

The Type Three Secretion (T3S) system is a device used by many Gram-negative pathogens that allows bacteria to deliver effector proteins straight into the eukaryotic cell cytosol. These effectors interfere with various signaling pathways to subvert the host cell functions. The secretion machinery of the T3S system consist of a basal body spanning the bacterial inner and outer membrane followed by a stiff hollow needle outside the bacterium. The fully assembled secretion apparatus constitute a continuous hollow conduit that connects the bacteria to the eukaryotic target cell. After cell contact, virulence proteins -called effectors- are injected directly into the cytosol of the host cell via the T3S apparatus. Several effectors of the T3S system require the assistance of specific cytosolic chaperones to be efficiently exported. There are three classes of T3S chaperones. Effector proteins are assisted by Class I chaperones. Although Class I chaperones are well characterized, their main function is still a matter of controversy. In this thesis, we demonstrate that orf155 encodes a specific chaperone for the effector YopO that we called SycO. We showed that SycO enhances YopO secretion in vitro and is required for translocation of YopO into infected cells. By pulldown assay we demonstrated that residues 20 to 77 of YopO are required and sufficient for SycO binding. Using crosslinking experiments and size exclusion chromatography analysis, we determined the stoichiometry of purified SycO and YopO-SycO complexes. SycO alone forms dimers in solution and the YopO-SycO complex has a 1:2 stoichiometry. These results suggested that SycO is a typical chaperone of the Class I. YopO is a serine/theronine kinase that interacts with Rho and Rac and disrupts the cytoskeleton of the target cells. YopO has been shown to localize at the cell plasma-membrane. By transfection of YopO-EGFP hybrid proteins into HEK293T cells, we demonstrated that the chaperone-binding domain (CBD) coincides with the membrane localization domain of YopO. Nevertheless, the CBD was not needed for the kinase activity of YopO. By ultracentrifugation, we also showed that the CBD causes YopO aggregation in the bacteria, when SycO does not cover it. Further, we show that the CBD of YopE and YopT also caused aggregation in the bacteria in the absence of SycE and SycT respectively. YopE, YopT and T3S effectors in other systems also act at the membrane of the eukaryotic host cell. We propose a new hypothesis concerning the role of T3S chaperones. The sub-cellular localization domain of effectors is aggregation-prone and creates the need for a chaperone inside bacteria. We propose that masking such aggregation-prone localization domains may be a general function for type III effector chaperones.

Conclusions What are the substrates of the VirB/VirD4 T4SS of Bartonella henselae? Starting point for the first published report (“A bipartite signal mediates the transfer of type IV secreted substrates of Bartonella henselae into human cells”, Chapter 3.1) in my Ph.D. thesis was the finding that the VirB/VirD4 T4SS of B. tribocorum was essential for the pathogen to establish an intra-erythrocytic infection in an animal model (1). Additionally, at that time, unpublished data indicated that many phenotypes which were already known for B. henselae infecting HUVECs, as anti-apoptosis (2), cytoskeletal rearrangements (3) and pro-inflammatory activation (4), were dependent on a intact VirB/VirD4 T4SS (5). These findings were suggestive of substrates being translocated through the VirB/VirD4 T4SS of Bartonella into the host cells. Sequencing 23 kb downstream of the virB locus of B. henselae revealed a coupling protein (virD4), and seven genes encoding at least one common Cterminal domain. The proteins encoded by these seven genes were later termed Bartonella exported proteins (Beps), the common C-terminal domain they contain Bartonella intracellular delivery (BID). By constructing a Hidden Markov Model from these BID domains and querying protein databases, we found similar domains in the C-terminus of relaxases from conjugative plasmids in the α-proteobacteria. We showed exemplarily that the C-terminus of the TraA relaxase from the AvhB/TraG conjugation system in A. tumefaciens – one of the top hits in the database search - could still be translocated by the VirB/VirD4 T4SS of B. henselae. Both these findings support that the BID domain evolved from conjugative relaxases, in parallel with the T4SS of B. henselae. Full-length relaxases bind covalently to plasmid DNA and mediate its transfer through T4SSs. This allows the fascinating speculation that the VirB/VirD4 T4SS of Bartonella could be used to export DNA by those means in vivo into host cells. To demonstrate exemplarily the translocation of the Beps through the VirB/VirD4 T4SS, we fused a FLAG-tag to the N-terminus of BepD and could show that BepD is translocated into infected endothelial cells in a VirB/VirD4 T4SS-dependent manner, whereupon it localizes to the cytoplasm of these cells and is tyrosine-phosphorylated by host-cell kinases. The precise experimental delineation of the domain needed for translocation was made possible by the development of the Cre-recombinase reporter assay for translocation (CRAFT), which showed the translocation domain to be bipartite. In addition to the BID domain, a short, positively charged C-terminal amino acid sequence was needed for an effective delivery of proteins. Non-polar deletions of all the ORFs encoding the Beps abolished the ability of Bh to induce a variety of host-cell phenotypes, suggesting that these proteins elicit biological effects in their eukaryotic target cells. This finding was the fundament for the two Manuscripts presented in Chapter 3.3, "A translocated protein of the vascular-tumor inducing pathogen Bartonella protects human vascular endothelial cells from Apoptosis" and “Subversion of host cell cytoskeletal function during invasome-mediated uptake of Bartonella henselae into human endothelial cells”. In each of these two manuscripts, a major phenotype of the infection of HUVECs by B. henselae is shown to depend on a single Bep. While BepA inhibits the apoptosis of endothelial cells, BepG does induce massive cytoskeletal rearrangements. Interestingly, it is the BID domain of BepA which mediates the anti-apoptotic activity in the host cell as well as the localization of this protein to the plasma membrane. This involvement of the BID in localization and function in the host cell holds also true for BepE, as we showed that the two BID domains of this protein are crucial for its localization (Chapter 3.2), and to inhibit the fragmentation of infected endothelial cells (Chapter 3.4). Many interfaces for interactions over phosphotyrosines Of the seven Bep proteins, three contain putative tyrosine-phosphorylation motifs in their N-terminus. We showed that both BepD and BepE are tyrosinephoshorylated by the c-Src kinase. Because of their similar substrate specificities and their variable expression levels depending on the cell type, any member of the SFK could potentially be the kinase of BepD and BepE in vivo. Bioinformatics and preliminary experiments suggest c-Abl as the kinase of BepF. As Crk was shown to bind BepF, and one of the functions of Crk has been described to activate c-Abl, one might speculate that a positive activation loop takes place once BepF has been phosphorylated, which would bind and activate increasing amounts of the c-Abl kinase to BepF through Crk. Whereas c-Abl has been implicated in the invasion process of other pathogens (6), the exact function of BepF in Bartonella still remains to be uncovered. BepD localizes in immunofluorescence stainings to a cytoplasmic vesicular-like compartment. Additionally, it localizes upon its tyrosine-phosphorylation to a Triton X-1oo insoluble fraction, and binds SHP2 and Csk. Lipid rafts are prominently associated to Triton X-100 insoluble fractions, and future studies might show BepD interfering with the signaling in these rafts. I subsequently focused on BepE, for which bioinformatics revealed an intriguing similarity to inhibitory immune receptors of mammals. This resulted in the second manuscript “Molecular mimicry of inhibitory immune receptors by the bacterial pathogen Bartonella” (Chapter 3.2). BepE contains two C-terminal BID domains which mediate the localization of this protein to the plasma membrane of HELA cells and to the plasma membrane and cell-cell contacts in endothelial cells, as shown by a co-localizing immunofluorescence staining with VE-Cadherin in HUVECs. This peculiar localization is very intriguing, and might modulate the cell-cell contact strength or the contact inhibition which is crucial in endothelial cells. Recently, a publication demonstrated that VE-Cadherin contains in its intracellular domain a binding site for the SH2 domain of Csk, the binding of Csk to this site being crucial for the contact inhibition of cell growth (7). We identified a very similar motif in the N-terminus of BepE which did also bind Csk. The co-localization and a similar tyrosine-phosphorylation motif binding the same protein makes this finding a good starting point for investigating the functions of BepE in the endothelium. BepE contains in addition to this Csk-binding motif two immunotyrosine inhibitory motif – immunotyrosine-based switch motif (ITIM/ITSM) tandems. These motifs are commonly found in the intracellular domain of inhibitory immune receptors. In vitro phosphorylation with the c-Src kinase and subsequent mapping by mass spectrometry analysis indicated that c-Src tyrosine-phosphorylates the Csk-binding site and both ITIMs of BepE. No phosphorylation of an ITSM could be detected, which might be due to the fact that another kinase phosphorylates these motifs in vivo. Both the inhibitory immune receptors and BepE do contain ITIMs, ITSMs, Csk-binding sites, localize to the plasma membrane, are tyrosine-phosphorylated by Src family kinases and bind SHP2 and Csk. Additionally, BepE is constitutively tyrosine-phosphorylated in HEK293T cells as well as in the primary HUVECs (Chapter 3.2), in contrast to most inhibitory immune receptors, which are thought to only become phosphorylated upon engagement of their extracellular ligands. All these lines of evidence indicate BepE mimicking inhibitory immune receptors. Csk contains one SH2 domain, SHP2 two. To elucidate the binding sites of these two proteins, and also to be able to use mutants lacking these interactions, we generated a panel of tyrosine-to-phenylalanine exchange mutants in BepE. While Csk binds to one motif (the one with the similarity to the Csk-binding site of VE-Cadherin), SHP2 interacts with motifs in the two ITIM-ITSM tandems. While this interaction study was carried out in HEK293T cells, more potential ITIM/ITSM-binding proteins such as SHP-1 and SHIP, EAT-2 and SAP might increase the complexity of the picture in myeloid and lymphoid cells. The two adapter proteins SAP and EAT-2 have been shown to bind to ITSMs. While some cells such as NK express both adapters, other cells as for example T-cell do only express SAP, and other such as DCs only EAT-2. While SAP recruits Fyn and can lead to an increase in cell activation, EAT-2 recruits phosphatases and Csk, inhibiting the activity. For BepE, this opens the possibility of switching its mode of action, depending on the cell type it is translocated into. Collaborations have been initiated to assess the immunomodulatory potential of BepE and we are currently studying the impact of this protein in our HUVEC models. To summarize, my Ph.D. thesis aimed at investigating the VirB/VirD4 T4SS of Bartonella henselae for the presence of secreted substrates and signals mediating this secretion. Additionally, to describe the functions and interaction partners of these substrates in the host cell, with an emphasis on the putative tyrosinephosphorylated effectors. The core findings of this thesis are a.) The discovery of seven modular substrates secreted by the apparatus b.) Description of the BID domain mediating the secretion of proteins and protein-DNA complexes, c.) Csk and SHP2 being interaction partners for BepD and BepE on the host cell side d.) BepE mimicking inhibitory immune receptors.

Bartonellae are bacterial pathogens uniquely adapted to cause intraerythrocytic infection in their mammalian reservoir hosts. In the case of human-specific Bartonella bacilliformis and Bartonella quintana, the intraerythrocytic bacteremia leads to the clinical manifestations of Oroya fever and trench fever, respectively. Here, we adapted large-scale signature-tagged mutagenesis (STM) for the first time to Bartonella, allowing us to screen for pathogenicity factors required for infection of the mammalian reservoir host in vivo. A total of 3084 STM mutants of rat-specific B. tribocorum were screened in a rat infection model for these criteria. After two rounds of screening, 130 mutants showed severe attenuation compared to wild-type B. tribocorum. We mapped the transposon insertion sites of these mutants to 80 different genes, and categorized them according to their putative function. Besides already described pathogenicity factors responsible for interaction with the host, like the two type IV secretion systems VirB-D4 and Trw, we discovered factors previously unlinked to pathogenesis. These belong to diverse functional classes, like transport, gene-expression regulation, cell envelope integrity, or metabolism. A quarter of the identified genes are (conserved) hypothetical coding for novel pathogenicity factors. We have used an additional PCR-screening approach on the entire mutant library to test for the level of mutational saturation and to identify non-essential genes in a pathogenicity island encoding 18 gene products related to the process of type IV secretion.

The Yersinia injectisome needle has a constant length of around 60nm. Mutants affected in the yscP gene display abnormally long needles, suggesting that needle length is genetically controlled and that YscP is involved in this process. First, both N- and C-term of YscP were shown to be required for proper needle length control. In contrast, the central part could be shortened or lengthened without loss of function, but giving a needle length strictly proportional to the number of residues of YscP. This led us to assign the function of a ruler to YscP. In addition to its role in the determination of the needles length, YscP is required for Yop secretion. The C-term part of YscP, required for proper needle length control, was shown to be also crucial for Yop secretion. Thus it is probably involved in the substrate-specificity switch of the machinery, from the needle component YscF to Yops. The study of this region by Hydrophobic Cluster Analysis led to a characterization of a new domain that we called T3S4 (TypeSecretion SubstrateSpecificity Switch) and which can be found in all counterparts of YscP in other injectisomes but also in flagellum. The conservation of the T3S4 structure suggests a similar function for all these proteins. The T3S4 domains found in injectisomes happened to be partially exchangeable. The overall structure of the domain seems to be the critical point. This is further supported by the fact that single alaninereplacement of the few conserved amino acids not affecting the structure, did not impact the functions. Since YscP is secreted by the injectisome, we analyzed whether its export is necessary for its functions. Two original export signals were found in YscP: one in the N-term (aa 1-35) and one more central (aa 97-137). These two signals exactly correspond to the N-term regions involved in needle length control. When YscP is deprived from its export signals, it still switches the substrate specificity but it cannot control needle length any more. These data suggest a model in which YscP export is linked to needle length control. Having evolved such a complex process for length control stresses the necessity of a certain length. Indeed, the length has evolved to match specific structures at the bacterial and host cell surfaces.