Toxin Inhibitors for the Treatment of Clostridium Difficile Infection

Home , Cadazolid, Tolevamer, Urtoxazumab

Research Collection

Doctoral Thesis

Toxin Inhibitors for the Treatment of Clostridium difficile Infection

Author(s): Ivarsson, Mattias E.

Publication Date: 2014

Permanent Link: https://doi.org/10.3929/ethz-a-010345630

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For more information please consult the Terms of use.

ETH Library

DISS. ETH NO. 22210

Toxin Inhibitors for the Treatment of Clostridium difficile Infection

A thesis submitted to attain the degree of

DOCTOR OF SCIENCES of ETH ZURICH

(Dr. sc. ETH Zurich)

presented by

Mattias Emanuel IVARSSON

MSc in Biomedical Engineering, ETH Zurich

born on 10.3.1987

citizen of Zurich, Switzerland

accepted on the recommendation of

Prof. Jean-Christophe Leroux, examiner

Prof. Bastien Castagner, co-examiner

Prof. Wolf-Dietrich Hardt, co-examiner

2014

Abstract

Clostridium difficile is a bacterial pathogen causing life-threatening infections that are the leading cause of hospital-acquired diarrhea. Recommended treatments for C. difficile infection (CDI) are limited to three antibiotics, which have unsatisfactory cure rates and lead to unacceptably high recurrence. The aim of the doctoral work presented herein was to explore the development of two novel therapeutic approaches against CDI. A variety of innovative therapeutic and prophylactic options against CDI are currently already in clinical trials, ranging from intestinal microbiota regeneration therapies to vaccines. These are presented and discussed in Chapter 1 of this thesis.

Protein toxins constitute the main virulence factors of several species of bacteria and have proven to be attractive targets for drug development. Lead candidates that target bacterial toxins range from small molecules to polymeric binders, and act at each of the multiple steps in the process of toxin- mediated pathogenicity. Despite recent and significant advances in the field, a rationally designed drug that targets toxins has yet to reach the market. Chapter 2 presents the state of the art in bacterial toxin- targeted drug development with a critical consideration of achieved breakthroughs and withstanding challenges. The discussion focuses on A–B-type protein toxins secreted by four species of bacteria, namely C. difficile (toxins A and B), Vibrio cholerae (cholera toxin), enterohemorrhagic Escherichia coli (Shiga toxin), and Bacillus anthracis (anthrax toxin), which are the causative agents of diseases for which treatments need to be improved.

The virulence of C. difficile toxins A and B (TcdA and TcdB) is modulated by intracellular auto- proteolysis following allosteric activation of their protease domains by inositol hexakisphosphate (IP6). In Chapter 3, we explore the possibility of inactivating the toxins by triggering their auto-proteolysis in the gut lumen prior to cell uptake using gain-of-function small molecules. The high calcium concentrations typically found in the gut precipitate IP6, precluding it from pre-emptively inducing toxin auto-proteolysis if administered exogenously. We therefore designed IP6 analogs with reduced susceptibility to complexation by calcium, which maintained allosteric activity at physiological calcium concentrations. We found that oral administration of IP2S4, a tetra-sulfated IP6 analog, attenuated inflammation in a mouse model of CDI.

The first step in the uptake mechanism of TcdA and TcdB is binding to the epithelial cell surface. In Chapter 4, we sought to develop a polymeric binder for TcdA that would inhibit the toxin’s receptor binding and cellular uptake, thereby preventing its cytotoxicity. The design was based on multivalently grafting a known ligand for TcdA, αGal(1-3)βGal(1-4)Glc, to poly(hydroxypropyl methacrylamide), a flexible, biocompatible and water-soluble polymer extensively used for biomedical applications. The polymers we synthesized did not inhibit TcdA cytopathy in cellular assays. We hypothesized that this was most likely because they failed to prevent binding of the toxin to the cell surface, as evidenced by erythrocyte hemagglutination assays. We attributed the lack of activity observed to the low binding

ABSTRACT affinity of the ligand (∼ 1 mM) and the concurrent failure of the polymer to generate positive binding cooperativity between the multivalently displayed ligand and the toxin. Attempts to identify novel high affinity peptide ligands for the receptor binding domain of TcdA through phage display were not fruitful.

The findings presented in this thesis have helped us to gain a deeper understanding of the opportunities and pitfalls of exploiting C. difficile toxins as therapeutic targets for the treatment of CDI. Although we were not able to create a polymeric binder effective at inhibiting the activity of TcdA, we succeeded in designing a small molecule capable of reducing CDI-associated symptoms in a mouse model of the infection. The latter result warrants further evaluation of the proposed molecule scaffold as a therapeutic option for treating CDI.

Résumé

Clostridium difficile est une bactérie pathogène causant des infections potentiellement mortelles qui sont la cause principale de diarrhée nosocomiale. Le traitement contre une infection à C. difficile (ICD) est limité à un choix de trois antibiotiques, ayant des taux de guérison insuffisants et conduisant à un taux de récidive de l’infection inacceptable. L'objectif de la recherche présentée dans cette thèse était de mettre au point deux nouvelles approches thérapeutiques contre l’ICD. A l’heure actuelle, un nombre d'options thérapeutiques et prophylactiques novatrices contre l’ICD sont evaluées en essais cliniques. Celles-ci vont des thérapies de régénération du microbiote intestinal aux vaccins et sont discutées dans le chapitre 1 de cette thèse.

Les toxines protéiques sont les principaux facteurs de virulence de plusieurs espèces de bactéries et se sont révélées être des cibles intéressantes pour le développement de médicaments. Les agents ciblant ces toxines bactériennes existent sous diverses formes, allant de petites molécules aux liants polymères, et agissent à différentes étapes du processus de pathogénicité des toxines. Malgré de récents progrès dans ce domaine, aucun principe actif conçu de manière rationnelle n'a encore été commercialisé. Le chapitre 2 décrit le développement de médicaments ciblant les toxines bactériennes en examinant de façon critique les percées réussies et les défis à surmonter. La discussion porte sur les toxines de type A-B sécrétée par quatre espèces de bactéries, à savoir C. difficile (toxines A et B), Vibrio cholerae (la toxine du choléra), Escherichia coli entérohémorragique (Shiga-toxines), et Bacillus anthracis (toxines létales et œdémateuses), qui sont les agents responsables de maladies pour lesquelles les traitements restent à être améliorées.

La virulence des toxines A et B de C. difficile (TcdA et TcdB) est modulée par la protéolyse intracellulaire faisant suite à l’auto-activation allostérique de leurs domaines de protéase par l'inositol hexakisphosphate (IP6). Dans le chapitre 3, nous explorons la possibilité d’utiliser des petites molécules pour provoquer l’auto-protéolyse des toxines dans la lumière du gros intestin afin de les inactiver avant qu’elles ne soient absorbées par l’épithélium. Les concentrations élevées de calcium habituellement présentes dans l'intestin font précipiter l’IP6, l'empêchant d'induire l’auto-protéolyse des toxines suite à une administration exogène. Nous avons donc conçu des analogues d’IP6 ayant une susceptibilité réduite à complexer le calcium tout en maintenant une activité allostérique à des concentrations de calcium physiologiques. De plus, nous avons démontré que l'administration orale d’un de ces analogues (IP2S4), atténuait l’inflammation du colon dans un modèle murin d’ICD.

La première étape du mécanisme d'absorption de TcdA et TcdB est leur liaison à la surface de l’épithélium intestinal. Dans le chapitre 4, nous avons cherché à développer un liant polymère pour la TcdA qui empêcherait la toxine de s’attacher aux cellules, inhibant ainsi sa cytotoxicité. La conception de ce polymère a consisté à greffer un ligand de TcdA, αGal(1-3)ßGal(1-4)Glc à un polymère d‘hydroxypropyle méthacrylamide de manière multivalente. Ce polymère est flexible, biocompatible, III

RÉSUMÉ soluble dans l'eau et largement utilisé dans des applications biomédicales. Malheureusement, les polymères ainsi synthétisés se sont avérés inefficaces pour attenuer la cytopathie de la TcdA dans des essais cellulaires. Nous émettons l'hypothèse que les polymères ne sont pas parvenus à empêcher la liaison de la toxine à la surface des cellules, comme en ont témoigné les tests d'hémagglutination d’érythrocytes. Nous avons attribué le manque d'activité observé à la faible affinité du ligand pour la toxine (∼ 1 mM) et l'échec simultané du polymère à engendrer une coopérativité de liaison positive entre le ligand et la toxine. Nos tentatives visant à identifier par expression phagique de nouveaux ligands peptidiques de plus haute affinité pour la TcdA n’ont pas été fructueux.

Les résultats présentés dans cette thèse nous ont aidés à mieux comprendre les possibilités et les défis liés à l’utilisation de toxines de C. difficile en tant que cibles thérapeutiques pour le traitement de l'ICD. Bien que nous n’ayons pas été en mesure de créer un liant polymère efficace pour inhiber l'activité de la TcdA, nous avons réussi à concevoir une petite molécule capable de réduire les symptômes associés à l’ICD dans un modèle murin de l'infection. Ce dernier résultat justifie une évaluation plus approfondie de la famille de molécules proposée en tant qu’option thérapeutique pour le traitement de l‘ICD.

Table of Contents

Abstract ...... I

Résumé ...... III

1 Background and Purpose ...... 1

1.1 Clostridium difficile Infection ...... 1

1.2 Antibiotics for the Treatment of CDI ...... 2

1.3 Microbiological Approaches for CDI Treatments ...... 4

1.4 Non-Microbial Biological Approaches for CDI Treatment and Prevention ...... 7

1.5 Antibiotic Inactivation for CDI Prevention ...... 9

1.6 Synthetic Toxin-Targeted Approaches for CDI Treatment ...... 9

2 Targeting Bacterial Toxins ...... 11

2.1 Introduction ...... 11

2.1.1 Clostridium difficile (Toxin A and Toxin B) ...... 11

2.1.2 Vibrio cholerae (Cholera Toxin) ...... 14

2.1.3 Enterohemorrhagic Escherichia coli (Shiga-like Toxin) ...... 16

2.1.4 Bacillus anthracis (Anthrax Toxin) ...... 19

2.2 Non-antibiotic Therapeutic Approaches ...... 21

2.2.1 Approaches Targeting Toxin Binding/Assembly ...... 21

2.2.2 Approaches Targeting Toxin Trafficking ...... 31

2.2.3 Approaches Targeting Toxin Translocation ...... 31

2.2.4 Approaches Targeting Toxin Processing ...... 32

2.2.5 Approaches Targeting Toxin Enzymatic Activity ...... 34

2.3 Conclusion ...... 41

3 Therapeutic Potential of Triggering Pre-Emptive Clostridium difficile Toxin B Auto- Proteolysis ...... 43

3.1 Introduction ...... 43

3.2 Materials and Methods ...... 44

3.3 Results ...... 48

3.4 Discussion ...... 55

TABLE OF CONTENTS

3.5 Remarks ...... 56

4 Development of a Polymeric Binder Targeting the Receptor-Binding Domain of Clostridium difficile Toxin A ...... 58

4.1 Introduction ...... 58

4.2 Materials and Methods ...... 59

4.3 Results ...... 65

4.4 Discussion ...... 76

4.5 Remark ...... 80

5 Conclusions and Outlook ...... 81

6 Appendix ...... 84

7 References ...... 90

List of Abbreviations ...... 114

Curriculum Vitae ...... 116

Acknowledgements ...... 119

1 Background and Purpose

1.1 Clostridium difficile Infection

C. difficile is an anaerobic, Gram-positive, spore-forming and toxin-secreting bacterium first identified by Hall and O’Toole in 1935 (Fig. 1.1A) (1). Its modern name stems from the difficulty in isolating it from other Clostridia because of its slow doubling time and high sensitivity to air. The first link between the bacterium and infections was made in the late 70s, when it was observed that broth cultures of Clostridium species isolated from stool of patients with pseudomembranous colitis due to antibiotic treatment caused cytotoxicity in cell cultures and enterocolitis in hamsters (2). Confirmation that C. difficile, and more specifically the toxins that it secretes, were the causative agents of disease symptoms in these patients followed shortly (3).

Clostridium difficile is now widely recognized as the leading cause of nosocomial diarrhea worldwide and is associated with substantial morbidity (4, 5). The reported incidence of C. difficile infection (CDI) has exploded in the last 15 years and is of the order of hundreds of thousands of cases per year in the United States and Europe (Fig. 1.1) (6, 7). The number of community-acquired cases is also on the rise (8). An important factor causing these increases is the emergence of so-called hypervirulent strains such as NAP1/BI/ribotype 027 that are more resistant to antibiotics and produce more toxin. In addition, they have also been suggested to form spores more readily and to adhere to intestinal epithelium better than wild-type strains (9). Reports of CDI cases outside of the United States and Europe have also been on the rise in recent years (10, 11). Accurate numbers for CDI prevalence are difficult to obtain for several reasons that complicate reporting of cases. These include a high rate of cases in patients with co-morbidities, lack of widespread testing of at-risk patients and variable diagnostic methods (12). Furthermore, reporting of CDI cases is not mandatory in all countries (13).

CDI typically occurs when a patient is given broad-spectrum antibiotics, which deplete the gut flora, and is exposed to C. difficile spores that germinate in the colon after ingestion. C. difficile’s resistance to several commonly used antibiotics such as clindamycin and moxifloxacin is thus the root cause of infections (14). Once established in the colon, C. difficile secretes two toxins, toxin A (TcdA) and toxin B (TcdB) that are the bacterium’s main virulence factors. These toxins are large proteins containing four domains, three of which (the binding, translocation and cysteine protease domains) act sequentially to deliver the fourth (the glucosyltransferase domain) into the cytosol of target cells, namely cells of the intestinal epithelium (15). The glucosyltransferase domain disrupts the actin cytoskeleton, ultimately causing cell death. At the macroscopic level, this results in diarrhea and pseudomembranous colitis, which can lead to severe and fatal complications such as toxic megacolon, bowel perforation, renal failure, systemic inflammatory response syndrome and sepsis (16). Chapter 1 offers a critical

CHAPTER 1: BACKGROUND AND PURPOSE overview of chemical, biological and microbiological agents currently in development for the treatment of CDI (Fig. 1.2), with a strong focus on drug candidates in clinical trials (Table 1.1).

FIGURE 1.1 Number of diagnosed cases of C. difficile infection in the United States between 1997 and 2012 (data from hcupnet.ahrq.gov). This figure is an update of the data presented in reference (6). The data were obtained by searching using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code for “Intestinal infection due to Clostridium difficile”: 008.45. All listed diagnoses for all patients in all hospitals in the US were searched.

1.2 Antibiotics for the Treatment of CDI

The antibiotics metronidazole and vancomycin have been the treatments of choice for CDI for over 30 years and are to this day still the recommended drugs in guidelines to physicians in Europe and the United States (metronidazole has not been approved for this indication but is nevertheless widely used) (17–19). Metronidazole is a nitroimidazole prodrug that, once taken up and reduced by bacterial cells, binds DNA covalently to inhibit nucleic acid synthesis (20). Vancomycin is a glycopeptide that inhibits the synthesis of peptidoglycans necessary for bacterial cell wall integrity and is a so-called antibiotic of last-resort (21). Metronidazole and vancomycin are both broad-spectrum antibiotics, with the former being the reference therapy for infections with anaerobic bacteria and some parasites, while the latter displays activity against the majority of Gram-positive bacteria (22). As such, they are active against C. difficile but also cause significant damage to the endogenous gut microbiota, which exposes patients to the risk of re-infection due to reduction of the protective barrier function of the microbiota. Indeed, a recent systematic review of treatment failures with the two antibiotics revealed that complete cure was 2

CHAPTER 1: BACKGROUND AND PURPOSE achieved in 77.6 % and 85.8 % of cases, with 27.1 % and 24.0 % recurrence for metronidazole and vancomycin, respectively (23). The fact that metronidazole is less efficacious for treatment of severe infections than vancomycin (but equivalent for mild disease) is the main reason for the small discrepancies in the figures obtained for each antibiotic. This, and the fact that metronidazole is roughly 10 times less costly than vancomycin, is also the reason why metronidazole is used to treat mild infections whereas vancomycin is recommended in cases of severe infection or of a lack of response to metronidazole (19). Indeed, the recommendations for CDI management state that metronidazole 500 mg should be administered three times daily in cases of mild to moderate CDI, and vancomycin 125 mg should be administered four times daily for severe CDI, both orally for a duration of ten days. Since vancomycin is not absorbed, it tends to be associated with minimal side-effects. Metronidazole, on the other hand, can be absorbed and lead to mild-to-moderate systemic side effects such as nausea, vomiting and in rare cases, peripheral and optic neuropathy (18).

Emerging strains of C. difficile are becoming less susceptible to both metronidazole and vancomycin (14, 24). Exacerbating the harmful consequences of their use, it has also been shown that these antibiotics enhance overgrowth of vancomycin-resistant enterococci, another healthcare- associated pathogen of serious concern (25). Taken together with the incomplete efficacy and high rates of recurrence attained with metronidazole and vancomycin, it is clear that there is significant upside potential for novel therapies to treat CDI.

In 2011, fidaxomicin (Optimer Pharmaceuticals) became the second drug to gain approval from the U.S. Food and Drug Administration (FDA) for the treatment of CDI (26). Fidaxomicin, whose discovery was first described in 1975, is a macrolide displaying a narrow-spectrum of activity against C. difficile and a few other Gram-positive anaerobes, which acts through inhibition of RNA polymerase. Importantly, fidaxomicin spares certain key species of the endogenous microbiota such as Bacteroides and Bifidobacterium and thus induces lower selection pressure for overgrowth of harmful bacteria (27, 28). Minimizing damage to the microbiota is also cardinal to protecting patients from disruptive, dangerous and costly recurrences, and thus constitutes the key value proposition of fidaxomicin and indeed all novel antibiotics in development for CDI treatment. Fidaxomicin resulted in 13.3 % recurrence of infection versus 24.0 % with vancomycin on a per-protocol basis for patients taking part in a phase 3 clinical trial (29). A notable outcome of this study is that the difference in recurrence rates for patients infected with non-NAP1 (i.e. non-hypervirulent) strains was large (7.8 % with fidaxomicin versus 25.5 % with vancomycin), whereas the recurrence rates were indistinguishable in the 35.9 % of patients infected with NAP1 strains.

Despite the fact that fidaxomicin has clinically proven advantages over metronidazole and vancomycin, its use has remained limited because of high pricing: 10-day regimens of metronidazole, vancomycin and fidaxomicin pills respectively cost approximately $ 22, $ 680 and $ 2’800 (19). The majority of recent cost-effectiveness studies conclude that fidaxomicin’s price is not justified despite

CHAPTER 1: BACKGROUND AND PURPOSE the recurrences that it prevents in non-hypervirulent strains, and that it would need to be priced at less than half of its current price to be deemed cost-effective (30–32). The hype and media storm surrounding fidaxomicin’s approval thus has yet to result in the patient benefit that society needs.

There are currently two antibiotics in phase 3 clinical trials for CDI: cadazolid (Actelion) and surotomycin (Cubist Pharmaceuticals), both of which have received Qualified Infectious Disease Product designations and Fast Track status by the FDA to accelerate their development (see: www.actelion.com and www.cubist.com). Cadazolid is a chimeric quinolonyl-oxazolidinone inhibitor of protein synthesis that also shows weak inhibition of DNA synthesis whereas surotomycin is a cyclic lipopolypeptide that acts by inducing cytoplasmic membrane depolarization (33, 34). Although structurally heterogeneous and possessing different mechanisms of action, the pharmacodynamics of both antibiotics appear to be similar to that of fidaxomicin in that they display strong bactericidal activity against C. difficile, are minimally absorbed and therefore readily reach and sustain therapeutic concentrations at their site of action in the colon, and largely spare key components of the gut flora such as Bacteroides species (35, 36). Both cadazolid and surotomycin appear to be more detrimental than fidaxomicin to Bifidobacterium species and it will be interesting to see if this translates into a difference in recurrence rates once clinical trials are complete (28).

The only other antibiotics currently in clinical trials for treating CDI are SMT-19969 (Summit Corporation in collaboration with the Wellcome Trust) in clinical phase 2 and CRS-3123 (Crestone) in phase 1. Both are non-absorbable narrow-spectrum agents and SMT-19969 has been shown to give rise to lower rates of recurrence than fidaxomicin in hamsters infected with hypervirulent strains such as ribotype 012 (37). After several clinical trials, a broad spectrum glycolipodepsipeptide antibiotic, ramoplanin (Nanotherapeutics), is being re-purposed for CDI prophylaxis (see: www.nanotherapeutics.com/ramoplanin). Phase 2 trials have also been completed with LFF-571 (Novartis) but its development for CDI treatment appears not to be actively pursued anymore (see: www.novartis.com/innovation/research-development/clinical-pipeline/index.shtml). Clinical testing of NVB-302 (Novacta Biosystems) also seems to have been recently discontinued after completion of phase 1 trials but the reasons for this are unclear.

1.3 Microbiological Approaches for CDI Treatments

A number of microbiological and biological approaches for treating CDI have emerged in recent years. Although highly different from a mechanistic point of view, they share the common property of allowing the endogenous gut flora to flourish (38). Their development has been spurred by the high rates of recurrence, and particularly debilitating multiple recurrences, observed with standard antibiotic treatment: the cure rate for recurrent CDI using vancomycin or metronidazole drops to < 30 % (39). By recently designating recurrent CDI as an orphan indication, the FDA has also provided a regulatory incentive for the development of new treatments. 4

CHAPTER 1: BACKGROUND AND PURPOSE

Among microbiological approaches, fecal microbiota transplantation (FMT) has only emerged in the last few years as having remarkably high success rates (> 90 %) for treating recurrent CDI, despite that it was first tested clinically in the 1950s (40, 41). The first randomized clinical trial using FMT for recurrent CDI was carried out in 2013 and was interrupted early for ethical reasons, since the patient group receiving vancomycin had a 31 % response to treatment whereas the FMT group had an 81 % cure rate (with a 94 % response after the second administration of FMT) (42). As its name suggests, FMT consists of transplanting processed stool from a healthy donor to a patient’s colon via enema, colonoscopic or nasogastric tube administration. The procedure rapidly restores the diversity of the intestinal microbiota and thus prevents proliferation of C. difficile. To date, there are at least one company (Rebiotix) and one non-profit organization (OpenBiome) that provide clinicians with filtered and frozen pathogen-free stool for FMT. The sharp rise in number of reports of successful treatments using FMT and the concomitant lack of regulatory oversight for the procedure prompted the FDA to announce in May 2013 that it considered human feces as a drug, hence requiring the submission of an Investigational New Drug (IND) application (43). Although this requirement would make FMT safer by standardizing donor screening and treatment administration, the FDA withdrew the requirement 6 weeks after the announcement in order not to restrict access to a much-needed therapy. The regulatory status of FMT thus remains in limbo and the final decision will have a strong impact on whether or not FMT will become a widespread therapy (44). The combination of high efficacy, regulatory uncertainty, limited access leading to numerous postings of do-it-yourself procedures on the Internet, and the unconventional esthetics of FMT have led to a media storm around the treatment that has had the beneficial effect of increasing public, patient and doctor awareness about CDI, a condition that until recently was referred to in the media as a “silent killer” because so few people had heard of it.

Although the advantages of FMT over antibiotics have been widely lauded, significant hurdles for its widespread use remain, not least of which is safe and standardized large scale production and formulation. Based on the fact that a key active ingredient in the stool samples used for FMT are the bacteria contained in it, efforts are now being made to define minimal consortia of a few tens of bacterial strains that can be formulated for oral administration as a replacement for FMT (41). Identifying which of the ca. 5,000 species of bacteria found in the gut are key to providing colonization is by no means a trivial task and these efforts are being led by the venture capital-backed companies Seres Health (clinical stage) and Vedanta Biosciences (pre-clinical stage) (45, 46). Their approaches deviate from probiotics development programs of large food manufacturers such as Nestlé and Danone that have focused heavily on a limited number of strains of Lactobacillus and Bifidobacterium that are readily amenable to inclusion in foodstuffs because of their resistance to harsh conditions in production processes and storage but are not key members of the endogenous human microbiota. Isolated examples of randomized, placebo-controlled double-blind studies reporting positive results for probiotics do however exist, such as for Bio-K+ products (47). Vedanta focuses on (non-toxigenic) bacteria from the Clostridium genus, which the founders have shown to play a key role in inducing regulatory T cells that 5

CHAPTER 1: BACKGROUND AND PURPOSE maintain immune homeostasis (48). By rebalancing activity of the immune system, Vedanta’s therapy also has potential for treating conditions such as Crohn’s disease, inflammatory bowel disease and allergies. Seres’ core technology is an algorithm to analyze differences between the healthy and diseased state microbiome, based on which consortia of a few tens of bacteria species capable of equilibrating the microbiota can be defined. The company’s lead candidate resulted in clinical cure of CDI in 29 out of 30 patients in a recently completed single-arm, open-label phase 1/2 trial (see: www.sereshealth.com/news).

It should be noted that there is an increasing amount of evidence showing the important role that non-bacterial components of fecal extracts playin conferring the protective effects observed through FMT. Experiments using the hamster model have demonstrated that sterile-filtered fecal extracts confer protection from CDI, which is abrogated upon ultrafiltration or heating (49). Taken together with results demonstrating the role of taurocholate and its derivatives on C. difficile sporulation and toxin activity, it would appear that bile salts and enzymes that metabolize them into secondary bile salts are key in mediating the therapeutic effects of FMT (50, 51). Taking these observations into consideration may be key to ensuring the efficacy of the microbiota supplementation therapies.

ViroPharma (acquired by Shire in January 2014) has pursued the development of another microbiological approach that consists of administering a non-toxigenic strain of C. difficile (VP20621) to recurrent CDI patients (52). The therapeutic concept is based on the observation that hospital patients asymptomatically colonized with C. difficile have a significantly reduced risk of contracting CDI, possibly because the niche in the intestinal ecosystem that harmful C. difficile would occupy is already taken (53). A phase 2 clinical trial to investigate safety, colonization and efficacy in reducing recurrence when administering VP20621 after a standard antibiotic treatment was completed in 2013 (see: clinicaltrials.gov/ct2/show/NCT01259726). Although the results were positive, the therapeutic use of non-toxigenic strains of C. difficile is questionable because it has been shown that the genetic region coding for C. difficile toxins, the pathogenicity locus, can be acquired by non-toxigenic strains by horizontal gene transfer from toxigenic strains, turning them into toxin producers (54). Furthermore, the microbiota repopulation technologies described above offer a safer and more balanced means of re- establishing microbiota homeostasis than a single non-toxigenic strain of C. difficile and are therefore likely to be preferred modalities from a regulatory and efficacy perspective.

Bacteriophages constitute an entirely different class of microbiological treatment with potential as a therapy for CDI (55). This approach has appealing properties such as high specificity but on the reverse side it is also susceptible to the development of resistance since C. difficile, similarly to other bacteria, have very effective natural defense mechanisms against phages, like the CRISPR/Cas system, that need to be thwarted (56). The pre-clinical development of therapeutic C. difficile-specific phages has recently been financed by UK-based AmpliPhi Biosciences Corporation and it will be interesting to see whether or not this treatment modality will enter clinical testing (see: http://www.ampliphibio.com/news.html).

CHAPTER 1: BACKGROUND AND PURPOSE

Given the rapid mutation rates and sensitivity to growth parameters of bacteria, a common challenge that all the microbiological approaches described above will ultimately face will be to develop scalable, GMP-compliant production processes with low batch-to-batch variability. Companies currently venturing into the microbiome-modulation market can benefit from the regulatory challenges and pitfalls that pioneering companies such as Osel and Actogenix have already faced in the last decade in other disease areas (41).

FIGURE 1.2 Schematic representation of the mode of action in the colon of clinical-stage drug candidates for CDI treatment and prevention.

1.4 Non-Microbial Biological Approaches for CDI Treatment and

Prevention

The non-microbial biological approaches closest to market entry are passive immunization and vaccination. They are fundamentally different from antibiotics and microbiological therapies in that they target the symptom-causing toxins secreted by C. difficile. In addition, they do not possess antibiotic properties and therefore also avoid causing damage to the intestinal microbiota.

Passive immunization consists of the systemic administration of antibodies to a non-immune individual and was used as an anti-toxin treatment as early as 1890 by Emil Behring and Shibasaburo Kitasato to treat diphtheria, earning them a Nobel prize (57). The rather crude so-called serum therapy, using blood serum from immunized animals for treatment, has evolved dramatically since then and the state-of-the-art are now highly purified antibodies engineered for high affinity and low non-specific 7

CHAPTER 1: BACKGROUND AND PURPOSE activity. In the context of CDI, phase 3 clinical trials of a mixture of fully human monoclonal anti-TcdA and anti-TcdB antibodies administered intravenously to patients receiving either metronidazole or vancomycin for CDI are currently underway with recurrence of infection within 12 weeks as the primary endpoint. This drug candidate, developed by Merck (MSD) and Medarex, showed positive results in a phase 2 trial with the same endpoint where patients receiving the treatment had a recurrence rate of 7 %, whereas recurrence occurred in 25 % of patients in the placebo group. The exact mechanism of protection is unclear in terms of where in the body the neutralization effect is taking place: TcdA and TcdB are secreted in the colon lumen and exert their toxic effects locally, except in severe cases where the damage to the intestines is such that endothelial barriers are destroyed and the toxins can circulate systemically (58). Antibodies, however, are commonly accepted to be unable to passively cross epithelia, although a few reports of increased fecal IgG after systemic administration of IgG at high doses do exist (59, 60). Nevertheless, the clinical data are very encouraging and other companies such as MicroPharm or UCB Pharma are also actively developing passive immunization drug candidates for CDI (61, 62).

With mean half-lives of circulation in humans of 26 and 22 days for anti-TcdA and anti-TcdB respectively, passive immunization offers transient boosts of circulating anti-toxin IgG (63). Several groups are now developing vaccines that can trigger sustained increases of serum anti-toxin. Sanofi Pasteur is carrying out a phase 3 trial with a bi-valent toxoid vaccine generated from formalin- inactivated TcdA and TcdB adsorbed on alum that increases serum IgG concentrations against the toxins (64). The endpoint of the trial is efficacy in preventing onset of primary symptomatic CDI and is seeking to enroll 15,000 people (see: clinicaltrials.gov/ct2/show/study/NCT01887912). Development of the vaccine has also been granted Fast Track designation by the FDA. The rationale for the development of a vaccine against CDI is similar to that for passive immunization and stems from studies showing that there is a strong correlation between asymptomatic carriage of C. difficile and high serum anti-TcdA IgG titers, and that patients that produce an anti-TcdA IgG response during a first episode of CDI are less likely to suffer recurrence (65, 66). Although data on the studies carried out to date are sparse, Sanofi Pasteur’s vaccine has been reported to increase serum anti-TcdA IgG levels for at least 60 days (64). Immune response to TcdB was consistently lower than for TcdB. The development of vaccines tailored for the elderly has been encouraged as an approach to reduce morbidity in the aging population but improvements in formulation are needed to obtain the strong and sustained immune responses seen in young and healthy individuals (67). The Sanofi Pasteur study will provide valuable information about the quality of the immune response in a large, elderly (> 50 years old) population, and its correlation with protection from CDI. Multiple other companies such as Pfizer, Merck and Valneva are also developing vaccine candidates using recombinant fragments of the toxins or different adjuvants to obtain stronger immunes responses. One should note that all of these vaccines elicit a systemic immune response and the anti-toxin antibodies produced do not inhibit colonization by C. difficile. A radically different approach developed by Prof. Simon Cutting of Royal Holloway University of London and 8

CHAPTER 1: BACKGROUND AND PURPOSE sponsored by the European Union 7th Framework Programme that should be entering clinical trials shortly consists of orally-administered inactivated bacterial spores that raise a mucosal response (68). If successful, such a vaccine would be likely to offer a more thorough protection from CDI than injected toxoid vaccines due to the co-localization of the immune response and the infection site.

1.5 Antibiotic Inactivation for CDI Prevention

Addressing the root cause of C. difficile, namely antibiotic treatment leading to destruction of the endogenous microbiota, is also an attractive option for reducing incidence of CDI. Improved antibiotic stewardship programs are an obvious and necessary way of reducing exposure to antibiotics but very often, antibiotics cannot be avoided. In these cases, an indirect approach to spare the microbiota is to co-administer agents that inactivate or sequester antibiotics present in the large intestine, thus allowing systemic circulation of the antibiotic but preventing damage to the intestinal microbiota. Such approaches for prevention of CDI are being led by two companies, Synthetic Biologics and Da Volterra. Synthetic Biologics is entering phase 1 in the fourth quarter of 2014 with SYN-004, an oral formulation of β-lactamase that digests intravenously administered β-lactam antibiotics to ampicilloic acid (see: www.syntheticbiologics.com). Da Volterra has completed phase 1 trials with an activated charcoal sorbent coated with a pH-dependent enteric polymer that acts as a sequestrant in the colon (69). This approach has the advantage of in principle being more universal than SYN-004. Pre-clinical testing was carried out with a clinical but non-epidemic strain of C. difficile (UNT-103-1) (70). One possible negative consequence of reducing the concentration of antibiotic in the gut without complete elimination is that it could increase exposure of bacteria to sub-minimum inhibitory concentrations of antibiotic, which encourages emergence of resistance.

1.6 Synthetic Toxin-Targeted Approaches for CDI Treatment

As discussed in chapter 2, bacterial toxins are amenable to targeting by synthetic approaches in a multitude of different ways. However, there are currently no synthetic toxin-targeted agents in clinical development for CDI treatment. Such an orally administered agent would offer the advantages of being rapid acting, non-antibiotic, cheap to produce relative to biologics, and could also potentially be used prophylactically. This thesis describes two approaches to develop synthetic toxin-inactivating molecules. In chapter 3, we present the development of a small molecule that pre-emptively triggers auto-proteolysis of the toxins, thereby inactivating them. The key to the success of this strategy was to circumvent the chelation of the toxins’ endogenous co-factor, IP6, by the high concentrations of calcium found in the gut by substituting phosphate groups for sulfate groups that have much lower affinity for calcium. Chapter 3 gives details on the rationale behind the molecule’s development, its synthesis, characterization, and in vitro and in vivo testing. Chapter 4 describes attempts to develop a novel polymeric binder that would sequester the toxins in the colon. We first grafted a reported trisaccharide 9

CHAPTER 1: BACKGROUND AND PURPOSE ligand for the receptor binding domain of the toxins onto a hydrophilic polymer to obtain multivalent display of the ligand. This polymer failed to show reduction of cytopathy in a cell-based assay, most likely because of the weak affinity of the ligand for the toxin. This provided the motivation to try to identify new ligands with higher affinity using phage display of heptapeptides but no high-affinity ligands could be identified. Both chapters provide insights into the opportunities and pitfalls of targeting these toxins at different sites.

TABLE 1.1 Products currently in development for treating and preventing CDI. Details about the clinical trials can be found using the NCT number on http://clinicaltrials.gov. The data were retrieved on the 15th of September 2014.

Treatment/prophylactic type* Product Sponsor Clinical phase Antibiotic Cadazolid Actelion 3 (NCT01987895) Surotomycin Cubist 3 (NCT01597505) SMT19969 Summit Corp. 2 (NCT02092935) CRS3123 Crestone 1 (NCT02106338) Microbiota supplement FMT† Rebiotix 2 (NCT01925417) SER109 Seres Health 2 (not registered) VP20621 ViroPharma 2 (NCT01259726) Passive immunization MK3415A Merck (MSD) 3 (NCT01513239) Vaccine CDiffense Sanofi Pasteur 3 (NCT01887912) Adjuvanted vaccine Pfizer 2 (not registered) IC84 Valneva 1 (NCT01296386) Synthetic Antibiotic inactivation SYN004 1 (not registered) Biologics DAV132 Da Volterra 1 (NCT02176005)

* Antibiotics and microbiota supplements are treatments (in the latter case only for recurrences); passive immunization prevents recurrence, vaccines and antibiotic inactivation can prevent both first cases and recurrences. † A total of 11 non-commercial hospital-sponsored phase 1 and 2 clinical trials are also underway as revealed by a search for “FMT Clostridium difficile” on clinicaltrials.gov. 10

2 Targeting Bacterial Toxins‡

2.1 Introduction

Protein toxins constitute an important part of the virulence factors that mediate the harmful effects of pathogenic bacteria. Since the discovery of diphtheria toxin in 1888, over 300 bacterial toxins have been identified and many of them are now recognized as being the causative agents of a multitude of bacterial diseases (71). These include widely known diseases such as cholera and anthrax, as well as emerging threats such as CDI.

Antibiotics have long been the preferred treatment for bacterial diseases. However, there are several advantages in targeting virulence factors rather than the bacteria themselves (72, 73). First and foremost, such treatments apply less evolutionary pressure on bacteria and are therefore less likely to induce emergence of resistant strains – a problem that has been recognized as one of the most important challenges of the 21st century (74). Second, toxins can continue to cause symptoms even after bacteria have been cleared from the host (75, 76). Third, non-antibiotic treatments circumvent the disruption of the normal microbiota that is typically associated with antibiotic treatments.

Bacterial toxins have an impressive variety of mechanisms of action: they can act on receptors at the cell surface, disrupt the plasma membrane, or enzymatically modify intracellular targets (71). The latter type are known as A-B toxins, with the binding (B) moiety mediating receptor binding and toxin translocation, and the active (A) moiety containing an enzymatic domain, which disrupts cellular homeostasis (77). The detailed understanding of these mechanisms has provided scientists with fertile grounds to devise creative approaches to inhibit bacterial toxins. The steps in the molecular mechanism of pathogenicity of secreted A-B toxins are shown in Figure 2.1. This review presents different chemical approaches targeting these steps and provides context to underline the unique challenges to translate concepts into effective therapies. The first part of this manuscript presents an overview of the existing knowledge about pathogenesis and current therapies for the toxin-mediated diseases caused by four well-documented species of bacteria (Clostridium difficile, Vibrio cholerae, enterohemorrhagic Escherichia coli and Bacillus anthracis). The second part describes selected approaches to inhibit the pathogenicity of the toxins.

2.1.1 Clostridium difficile (Toxin A and Toxin B)

CDI is responsible for a spectrum of nosocomial diseases ranging from mild diarrhea to life threatening conditions such as toxic megacolon, septic shock and intestinal perforation (78). The three main risk factors of CDI are antibiotic treatment, hospitalization, and age (79). Treatment with broad-

‡ This chapter has been published: Ivarsson ME, Leroux J-C, Castagner B (2012) Targeting Bacterial Toxins. Angewandte Chemie (International Edition) 51:4024–4045. 11

CHAPTER 2: TARGETING BACTERIAL TOXINS spectrum antibiotics against anaerobes alters the balance of bacterial species found in the gut. This dysbiosis facilitates the colonization of the large intestine by C. difficile (78). In the last 15 years, C. difficile has gained attention in North America and Europe because of significant increases in the frequency and severity of infection (80). These observations have been traced back to the emergence of an epidemic, hypervirulent strain of C. difficile termed toxinotype III ribotype 027 (81). The increase in severity of C. difficile-associated diseases is characterized by higher incidence of toxic megacolon, more cases in younger patients, and higher relapse and death rates (82, 83).

FIGURE 2.1 Schematic representation of the steps of bacterial toxin pathogenesis in eukaryotic cells, which can potentially be targeted by therapeutic molecules. Such molecules include polymeric binders, insoluble sorbents, peptides, peptide analogs and non-peptidic small molecules.

C. difficile colonizes the gut by using flagella and proteases to penetrate down to the enterocyte layer to which it adheres through adhesins (78). There, it secretes three toxins with varying contributions to virulence: TcdA, TcdB and C. difficile transferase (CDT) (84). CDT is a binary actin-ADP- ribosylating toxin that has been shown to induce the formation of microtubule projections in cells, but its role in C. difficile virulence remains unclear (85). TcdA and TcdB are glucosyltransferases (GTs) that modify the actin cytoskeleton by glucosylating the Rho GTPases RhoA, Rac, and Cdc42 found in mammalian cells (86). The action of both toxins on the cytoskeletal regulatory machinery causes cell rounding and ultimately cell death (87). TcdA and TcdB are responsible for the pathogenesis, however, recent literature offers conflicting reports on which of the two is more injurious. TcdB is more important than TcdA in infection models, but strains producing only the latter are still capable of inducing disease and death in hamsters (88, 89).

The closely related TcdA and TcdB can be divided into four functional domains: a C-terminal receptor-binding domain, a hydrophobic translocation domain, an autoprocessing cysteine protease domain (CPD) and a glucosyltransferase domain (Figure 2.2) (15, 87). TcdA and TcdB bind enterocytes via their receptor-biding domain (90). TcdA has been shown to bind glycoprotein96 (gp96) on the surface of human colonic epithelial cells, although details of the interaction remain unknown (91). TcdA binds to various glycans, including the trisaccharide αGal(1-3)βGal(1-4)Glc but this sugar is not found 12

CHAPTER 2: TARGETING BACTERIAL TOXINS in human tissue (92). The characterization of the receptor for TcdB remains even more elusive despite the fact that TcdB is toxic to a broad range of cell types, suggesting that its receptor is ubiquitously expressed (87). Once bound to the cell surface, TcdA and TcdB are internalized through endocytosis (Figure 2.3). Acidification of the endosomal compartment induces a conformational change that exposes the hydrophobic region in the putative pore-forming domain and leads to membrane insertion, with the N-terminus protruding into the cytosol and the C-terminus remaining inside the endosomal compartment (15).

FIGURE 2.2 The structure of TcdA revealed by negative stain electron microscopy with the glucosyltransferase domain (red), the cysteine protease (blue), the translocation domain (yellow) and the receptor-binding domain (green). The image on the right is an overlay of the image on the left with known crystal structures (93).

IP6, a signaling molecule ubiquitously found in the cytosol of mammalian cells, binds to the CPD after it is transferred to the cytosol, and activates the CPD resulting in auto-processing of the toxins (94–96). Toxin cleavage releases the glucosyltransferase domain into the cytosol, where it induces pathogenesis.

CDI is a significant and growing problem with at least 250,000 hospitalized cases a year, a mortality rate of 1 – 2.5% and estimated costs of several hundred million dollars annually in the United States (97). Oral vancomycin or metronidazole have proven to be the most effective treatments for CDI since the 1980s (16, 98). Recent studies with fidaxomicin, a narrow-spectrum antibiotic with high selectivity for C. difficile over other bacteria, showed similar cure rates as vancomycin but reduced rate of relapse (29, 99). Fidaxomicin was approved by the FDA in May 2011 and could replace vancomycin as a first- line treatment for CDI (100). The treatment of CDI with antibiotics has several inherent drawbacks. First, there are growing concerns over the emergence of antibiotic-resistant strains that complicate treatment. Second, high recurrence rates have been reported after treatment with antibiotics (98). Third, an increasing number of cases that are refractory to antibiotics have been reported (101). Developing treatments that target C. difficile toxins rather than the bacterium itself has the advantage of maintaining homeostasis of natural gut flora, rather than further disrupting it. This is of key importance because in individuals with a healthy gut flora, the presence of endogenous bacteria acts as a protective barrier against C. difficile proliferation. 13

CHAPTER 2: TARGETING BACTERIAL TOXINS

FIGURE 2.3 Molecular mechanism of action of TcdA and TcdB. The toxins bind to the surface of enterocytes via the receptor-binding domain and are endocytosed. Acidification of the endosome leads to translocation of the enzymatic domain and the CPD into the cytosol. Cytosolic IP6 then binds to the CPD, thus activating it and initiating autoprocessing. The released enzymatic domain catalyzes the transfer of a glucose moiety to a conserved threonine residue on Rho/Ras proteins, which inhibits downstream signalling events.

2.1.2 Vibrio cholerae (Cholera Toxin)

Cholera is a diarrheal disease caused by V. cholerae, a bacteria transmitted by contaminated food and water (102). Cholera is endemic in many regions, and the World Health Organization estimates the number of deaths attributed to cholera to be 120,000 each year (103). The symptoms are often hard to distinguish from other diarrheal diseases, a fact that complicates its diagnosis and reporting. Cholera symptoms include voluminous watery diarrhea and vomiting and will appear after an induction period ranging from 1 – 5 days. The treatment is simply parenteral and/or oral rehydration. However, treatment has to be administered rapidly, since severe dehydration can lead to death within a few hours of onset. Because of this, most of the deaths caused by cholera occur in countries with limited access to healthcare. Therefore, even a successful toxin-targeted therapy would have limited impact in such settings. The large amounts of water being secreted from the small intestine also make the development of oral therapies more challenging. Prevention through sanitary measures, increased access to healthcare and the development of new oral vaccines are the most likely initiatives to avoid deaths due to cholera.

CT is an AB5 type toxin composed of an enzymatic A subunit and a ring-like pentameric B subunit that is involved in cellular uptake (Figure 2.4) (104, 105). Individually, neither the A nor the B5 subunits are toxic. The A subunit is divided in two parts. A1 is the enzymatic portion of the toxin,

CHAPTER 2: TARGETING BACTERIAL TOXINS

responsible for the toxicity, and is linked to the pentameric B5 subunit by the A2 domain. The A1 and A2 domains are linked via a disulfide bond (106).

FIGURE 2.4 A, The cholera toxin AB5 structure (PDB:1XTC) (107). B, the B5 subunit co-crystallized with GM1 (PDB:3CHB) (108).

The B5 subunit binds to GM1 ganglioside on the cell surface (Figure 2.5) (109, 110). After endocytosis, the whole toxin is transported to the endoplasmic reticulum (ER) by retrograde transport via the trans-Golgi network (TGN) (111). At this point the A1 domain separates from the A2-B5 complex, unfolds and translocates to the cytosol by hijacking the ER-associated degradation pathway (106, 112, 113). The A1 subunit is an ADP-ribosyltransferase that targets the trimeric Gsα component of adenylyl cyclase (AC). Once ADP-ribosylated, AC remains in its active GTP-bound state, which results in increases in the concentrations of cyclic AMP (cAMP) and chloride ions, which then leads to massive fluid secretion from the small intestine. CT shares roughly 80% homology with the E. coli heat- labile toxin (LT) and shares the same cell-surface receptor. Enterotoxigenic E. coli (ETEC), which produces LT and/or the heat-stable toxin (ST) is responsible for the so-called “traveller’s diarrhea” that is clinically indistinguishable from cholera.

Because of its lack of toxicity and high mucosal antigenicity, recombinant CT B subunit is part of the oral cholera vaccine Dukoral, along with various inactivated V. cholerae strains (114). Treatment with antibodies seems unlikely in situations where basic healthcare facilities are lacking because of cost issues, and because of the harsh conditions of the gastro-intestinal (GI) tract. However, egg yolk IgY obtained from hen immunized with inactivated bacterial and CT B subunit was protective to suckling pigs against challenge with V. cholerae (115). IgY from egg has the advantages of being cheaper to produce, can be lyophilized, and administered orally.

CHAPTER 2: TARGETING BACTERIAL TOXINS

FIGURE 2.5 Molecular mechanism of action of CT. The whole toxin binds to GM1 ganglioside, followed by endocytosis. After retrograde transport via the TGN, the A1 subunit reaches the cytosol where it ADP- ribosylates the trimeric Gsα component of adenylyl cyclase.

2.1.3 Enterohemorrhagic Escherichia coli (Shiga-like Toxin)

The enterohemorrhagic E. coli (EHEC) is an important pathogen transmitted through food and water that causes severe gastroenteritis (so-called “hamburger disease”) (116). It produces the phage- encoded Shiga-like toxin (SLT, also Shiga toxin or verotoxin) as its main virulence factor. Shiga toxin, the namesake of SLTs, was initially discovered while studying Shigella dysenteriae that produces the toxin, but the name was kept to describe the very closely related SLTs produced by E. coli. The S. dysenteriae Stx and the Stx1 secreted by E. coli are virtually identical. Stx1 and Stx2, another SLT produced by some strains of EHEC, share approximately 60% sequence homology. The main cause of outbreaks in parts of Europe and North America is usually E. coli O157:H7, although some recent outbreaks in Europe were caused by the O104:H4 strain (116, 117). SLTs are released by the bacteria in the lumen of the GI tract, causing diarrhea, and then penetrate the intestinal submucosa and access systemic circulation. This can cause acute kidney injuries and central nervous system complications, especially in children (118). EHEC is responsible for most of the hemolytic uremic syndrome (HUS) in children in regions where the bacteria is endemic. Although HUS is a potentially fatal condition, current treatments are mostly supportive, and antibiotics are potentially damaging (119), perhaps due to bacterial lysis or increased SLT production (120). The usual course of an E. coli O157:H7 infection results in diarrhea 3 days after ingestion of the bacteria, followed by bloody diarrhea 1 to 3 days later (121). This is usually when the patient seeks medical attention and a culture test reveals the presence of

CHAPTER 2: TARGETING BACTERIAL TOXINS

EHEC. The disease can then resolve spontaneously or deteriorate to HUS. The latter is characterized by microvascular thrombi and swollen endothelial kidney cells caused by circulating SLTs. By the time the presence of EHEC is confirmed, the toxin-related damage is usually well underway, making a rapid intervention crucial for the success of any toxin-targeted therapy (121). Recent literature suggests that the damage to kidney by SLTs during HUS is in part mediated via the complement system and that the monoclonal C5 antibody Eculizumab might be useful for the management of HUS in children (122, 123).

SLTs are AB5 type toxins, similarly to CT (Figure 2.6) (124). The assembled toxin interacts with the globotriaosylceramide Gb3 on the cell surface via its B5 subunit, followed by endocytosis (Figure 7) (125, 126). It is transported via a retrograde pathway from the endosome to the ER (127). The A subunit is composed of two parts: the N-terminal A1 domain is the enzymatic domain that has to be released by a protease from the C-terminal A2 domain interacting with the B pentamer (Figure 2.6A). Proteolytic cleavage is mediated by proprotein convertases such as furin in the Golgi and endosome (128). The reduction of a disulfide bond also has to take place to fully release the A1 subunit, which eventually reaches the cytosol via chaperone-mediated transfer, where it acts on the 28S RNA of the 60S ribosomal subunit (129, 130). The A subunit of SLTs and the closely related ricin toxin (RT, Ricinus communis) are both N-glycosidases that ultimately prevent protein synthesis (130, 131). Another AB5 toxin, subtilase cytotoxin (SubAB), secreted by E. coli has recently been discovered (132). SubAB can induce pathological features similar to that of HUS in mice, but the extent of its role in HUS in the context of an EHEC infection is still unclear.

FIGURE 2.6 A, Structure of Stx2. The B pentamer is multicolored, the A2 subunit is in cyan and the A1 subunit is in black. The enzymatic pocket is highlighted in magenta and the disulfide bond that links the A1 and A2 subunits is highlighted in orange. The loop that undergoes proteolytic cleavage is missing from the crystal structure and is represented by a dotted black line (PDB:1R4P) (107, 124). B, Binding sites on Stx1. Sites 1, 2 and 3 are highlighted on one B subunit in green, blue and magenta respectively and the co-crystallized carbohydrates are shown in the respective colors (PDB:1BOS) (133).

Of the two toxins, Stx2 has been linked to more severe clinical outcomes and is therefore the most important target for therapeutic interventions (134). Stx1 and Stx2 differ slightly in their carbohydrate-binding domain. Stx2 has not been crystallized in the presence of bound carbohydrate, but

CHAPTER 2: TARGETING BACTERIAL TOXINS

Stx1 has (134). Stx1 shows three different binding sites that seem all important in mediating cytotoxicity

(Figure 2.6B) (135). However, in solution, the Gb3 trisaccharide seems to bind mostly to site 2 and, to some extent site 1, but not site 3 (133, 136). It was recently shown that Stx2 has stronger binding affinity to an N-acetylgalactosamine-terminated Gb3 and that this preference was linked to site 2 (137). The bacteria can secrete SLTs in outer membrane vesicles, but the significance of this in pathogenesis is still unclear (138).

Three humanized mouse IgG1κ antibodies against SLTs have been tested in phase I clinical studies (139). Furthermore, human monoclonal antibodies are currently being evaluated: cαStx1 and cαStx2 (Shigamabs), directed against the Stx1B subunit and Stx2A subunit respectively and Urtoxazumab (TMA-15), directed against Stx2B subunit (118, 140, 141). They were all tested in healthy adults, and Urtoxazumab was also evaluated for safety in pediatric patients infected with EHEC. Both formulations were found to be well-tolerated with half-lives exceeding 9 days, so a single injection at the beginning of symptoms should hypothetically last until the onset of HUS. It remains to be seen if these antibodies will offer protection when administered to EHEC-positive patients that are diagnosed after a few days of diarrhea. Another serious development problem for therapies targeting such rare and sporadic diseases is to design an appropriate clinical study with clear primary end-points, and to have a population group large enough to achieve statistical significance. This is certainly a real problem for HUS, although countries where there is a higher incidence such as Argentina, could provide large study groups (72). Such countries, where EHEC infections are endemic, could also potentially benefit from a vaccine against EHEC or its toxins. The advantage of such widespread immunization in other countries is unclear.

CHAPTER 2: TARGETING BACTERIAL TOXINS

FIGURE 2.7 Molecular mechanism of action of E. coli SLTs. The B5 subunit binds to the cell-surface receptor. The toxin then undergoes retrograde transport to the ER, during which it is processed by furin, liberating the enzymatic A1 subunit from the A2-B5 complex. The A1 subunit is then translocated into the cytosol where it ultimately prevents protein synthesis.

2.1.4 Bacillus anthracis (Anthrax Toxin)

Anthrax toxin is a tripartite toxin secreted by the spore-forming bacterium Bacillus anthracis (142). Infection with B. anthracis can be cutaneous, gastrointestinal, or pulmonary (inhalational), all of which can lead to fatal systemic anthrax disease if not treated adequately (143). Anthrax disease in humans (inhalational anthrax in particular) has attracted significant research attention because B. anthracis spores are particularly resilient and readily aerosolized, and are therefore amenable to dissemination as a biological weapon (144).

The three proteins that constitute anthrax toxin are lethal factor (LeF), edema factor (EF) and protective antigen (PA) (142). LeF is a zinc protease that cleaves mitogen-activated protein kinase kinases (MKKs) at the amino terminus (145). EF is a Ca2+/calmodulin-dependent adenylate cyclase that increases cytosolic cAMP concentrations to pathological concentrations (146, 147). PA is secreted as an 83-kDa protein that can bind either one of two cellular receptors: tumor endothelial marker 8 (TEM8 or ANTXR1) or capillary morphogenesis protein 2 (CMG2 or ANTXR2) (148). PA83 is cleaved to 63 kDa by a protease at the cell surface or in the extracellular milieu, which allows its heptamerization into a ring-like structure (Figures 2.8 and 2.9) (149). PA83 is particularly sensitive to cleavage by furin and other proprotein convertases (PCs) (150). Once cleaved, cell-surface-bound and heptamerized, PA binds LeF and/or EF, and the complex undergoes clathrin-mediated endocytosis. Acidification of the 19

CHAPTER 2: TARGETING BACTERIAL TOXINS

endosome leads to membrane insertion of (PA63)7 and formation of a pore through which bound LeF and/or EF eventually translocate into the cytosol (148, 151). PA-associated LeF and EF are respectively known as lethal toxin (LeT) and edema toxin (ET), with LeT considered the primary virulence factor (152). None of the three proteins is toxic individually (153).

FIGURE 2.8 Crystal structure of the PA63 heptamer pre-pore. A, (PA63)7 viewed from the top. B, Side-view of the same heptamer. Two monomers have been removed for clarity (PDB: 1TZO) (154).

Antibiotics, vaccines, and passive immunization are all used for anthrax prophylaxis (144). Although these approaches are effective in preventing onset of anthrax, access to effective therapies rather than prophylactic measures would be necessary in the event of sudden and unexpected dissemination of anthrax spores. The recommended treatment for anthrax is currently limited to administration of antibiotics (155). However, toxin accumulation in infected patients may still lead to death despite effective clearance of B. anthracis. This is especially true in cases of inhalational anthrax, where disease progression is very rapid (76). The completion of phase III clinical trials and of animal studies in rabbits and monkeys with Raxibacumab, a human monoclonal antibody directed against PA, supports that targeting anthrax toxins alone is sufficient to treat infection with B. anthracis (156). However, the highly malleable nature of PA implies that it can easily be engineered to become antibody- resistant (157). This fact, together with the antibody’s high cost and dosage requirements (40 mg/kg), encourages the development of non-antibody, toxin-targeted approaches to treating anthrax. The rarity of inhalational anthrax, however, poses significant hurdles in evaluating the efficacy of new therapeutics (158, 159).

CHAPTER 2: TARGETING BACTERIAL TOXINS

FIGURE 2.9 Molecular mechanism of action of anthrax toxin. Protective antigen 83 kDa in size (PA83) binds either TEM8 or CMG2 at the cell surface and is processed to PA63 by PCs such as furin. PA63 heptamerizes, allowing it to bind EF and/or LeF. The complex formed is endocytosed and subsequent acidification of the endosome leads to translocation of the bound EF and/or LeF. Once in the cytosol, EF catalyzes the formation of cAMP from ATP and LeF mediates the proteolytic cleavage of MKKs, which disrupts the mitogen-activated protein kinase (MAPK) pathway.

2.2 Non-antibiotic Therapeutic Approaches

2.2.1 Approaches Targeting Toxin Binding/Assembly

All A-B type toxins presented here act on intracellular proteins and therefore need to penetrate the cell membrane to cause pathogenesis. The binding of the toxin to the cell surface, prior to internalization, constitutes one of the earliest steps in the pathogenesis, which makes it an attractive target. Several approaches have been proposed to disrupt the binding of the toxin to its receptor on the cell surface, ranging from insoluble sorbents and polymers, to optimized multivalent presentations of ligand.

2.2.1.1 Insoluble Sorbents

For toxins released in the GI tract, insoluble sorbents can, in principle, be administered orally, bind to the toxin and get excreted in the feces without systemic absorption. Synsorb is a sorbent that was produced by linking a toxin-binding carbohydrate ligand to Chromosorb P, a diatomite used in chromatography. Synsorb-90, directed against C. difficile TcdA, and Synsorb-Pk, directed against E.

CHAPTER 2: TARGETING BACTERIAL TOXINS coli SLTs, were evaluated in clinical trials but ultimately failed (1-2, Figure 2.10) (160, 161). In a phase II study, Synsorb-Pk did not diminish the severity of HUS in pediatric patients. It would be tempting to conclude that this study confirms that targeting SLTs in the GI tract after hospitalization is too late, but it is also possible that efficacy of Synsorb-Pk in vivo was not sufficient to observe an effect, perhaps because of its lower neutralizing activity against Stx2 compared to Stx1 (162). It is, however, more likely that a therapy targeting SLTs in the circulation would be more effective.

Synsorb-90 displays the trisaccharide αGal(1-3)ßGal(1-4)βGlc known to bind the receptor- binding domain of TcdA (163). At 200 mg/kg in a mouse model, Synsorb-90 offered protection against TcdA enterotoxicity. However, at 400 mg/kg, even non-functionalized Chromosorb P showed activity, suggesting that the non-specific interactions were playing a role (164). Synsorb-90 failed in phase 3 clinical trial for possibly two reasons (161). First, it is now known that TcdB is at least as important as TcdA in causing disease whereas Synsorb-90 was only active against TcdA. This is an intriguing observation since the carbohydrate ligand should, in principle, be capable of binding to the receptor- biding domain of TcdB as well (165). A second possible factor is the fact that the Chromosorb particles are too large to diffuse in the mucus layer that protects enterocytes, whereas C. difficile is known to penetrate this layer (78, 166, 167). Presumably, a successful therapeutic agent will have to penetrate the mucus to inhibit internalization of the toxins.

In the previous examples, carbohydrates were used to obtain specific binding of the decoy to the toxin. Screening campaigns can be used to identify new peptides or peptoid moieties that bind selectively to a toxin, even if the native receptor is a carbohydrate. Peptoids are unnatural peptide analogs where side-chains are attached to the amide nitrogen on a polyglycine backbone. Since peptoids are resistant to enzymatic digestion, they constitute an attractive choice for an oral decoy. A peptoid- functionalized tentagel bead that binds V. cholerae CT was identified by screening a library of up to 105 candidates (3, Figure 10) (168). Two hits were obtained and shared a consensus sequence that was re- synthesized without the last two, non-conserved residues. Beads functionalized with this consensus sequence protected intestinal cell monolayers against CT, albeit with preincubation of the beads with the toxin. Ultimately, oral therapy targeting the toxin for cholera could be complicated by voluminous diarrhea.

CHAPTER 2: TARGETING BACTERIAL TOXINS

FIGURE 2.10 Structure of oral, insoluble decoys against C. difficile TcdA (1), E. coli Stx (2) and V. cholerae CT (3) respectively.

2.2.1.2 Soluble Decoys

Soluble forms of native receptors involved in binding and internalization of toxins can be used as decoys. This approach was explored against B. anthracis PA with the advantage being that the toxin cannot be nefariously engineered to avoid binding to the receptor without influencing its activity (169). Soluble fragments of the PA receptors TEM8 and CMG2 were expressed in mammalian cells and tested in vitro and in vivo. The sCMG2 decoy protected rats when co-injected at or near equimolar ratio with LT. The sTEM8 was less effective in vitro and did not protect rats in the challenge experiments. The difference in activity was attributed to differences in binding affinity of the soluble decoy and PA. Indeed, sCMG2 has a binding affinity of 170 pM for PA compared to 130 nM for sTEM8. Recently, a mutated form of sTEM8 (L56A) showed improved affinity for PA and in vivo protection similar to that of sCMG2 (170). In this study, the L56A soluble decoy showed increased plasma half-life and increased targeting to the lungs compared with sCMG2, which could explain its similar in vivo efficacy to sCMG2 despite lower affinity to PA. The sCMG2 soluble decoy was recently evaluated in a mouse challenge using toxins engineered to be antibody resistant. As expected, the soluble decoy was still effective in protecting mice (157). The circulation half-life of the sCMG2 decoy was increased dramatically to 5 days by fusing it to an antibody Fc region (171). This approach may constitute an interesting adjunct or alternative therapy to antibody treatment, especially in the context of weaponized anthrax exposure (170).

2.2.1.3 Polymeric/Multimeric Binders

Polymeric binders are macromolecules that are designed to bind molecules in the GI tract and prevent their absorption (172, 173). These polymers are themselves not absorbed and thus generally non-toxic. For example, cholestyramine and colesevelam are used to sequester bile acid salts and prevent 23

CHAPTER 2: TARGETING BACTERIAL TOXINS their reabsorption, which results in a reduction of plasma cholesterol (174). Other polymers are employed in the management of hyperkalemia (Kayexalate) and hyperphosphatemia (Sevelamer), and more are being evaluated, notably in the management of celiac disease (175). Tolevamer is a polymeric binder that was designed to sequester C. difficile toxins (176). This polymer showed promising results in phase 2, especially for recurrent CDI, but failed in phase 3 clinical trial in a non-inferiority test against metronidazole and vancomycin (161). The polymer had to be administered in high doses (9 g/day), perhaps because of unspecific interaction with mucins or other proteins. Tolevamer is a high molecular weight polystyrene sulfonate that was shown to bind TcdA and, to a lesser extent, TcdB via electrostatic interactions and multiple weak hydrophobic interactions (177). The efficiency of the polymer to sequester TcdB was perhaps not sufficient in light of the renewed importance of TcdB in the pathogenesis.

Polymers can be grafted with recognition moieties such as carbohydrates to improve their specificity. Such glycopolymers have been used extensively to bind bacterial toxins and other virulence factors such as adhesins (178). For example, Gb3 trisaccharide, the receptor for SLTs, bound to a polyacrylamide backbone (4, Figure 2.11) was shown in one report to bind both Stx1 and Stx2 and inhibit their cytotoxicity (179, 180). Moreover, the polymer was effective when administered intragastrically at day 3 to mice challenged with a fatal dose of E. coli O157:H7 (179). However, these results contrast with that of a previous report showing that a Gb3 polymer could bind Stx1 but not Stx2

(181). The authors suggested that the aliphatic linker used between the Gb3 and the polymer backbone played an active role in the successful binding to Stx2, thus explaining the different results (182). They suggested that this polymer could be a useful prophylactic agent for people who have been in contact with EHEC patients or who might have been exposed to EHEC in an outbreak (183).

Due to their unique pentameric structure, SLTs and CT have been the subject of very elegant studies where multivalent displays of carbohydrates exhibited high affinity binding to toxins. One of the most notable examples is a pentavalent display of Gb3 trisaccharides for the binding of SLTs (Starfish 5, Figure 2.11) (184). The design is based on a glucose core with 5 radiating spacers terminated by two

Gb3 trisaccharides, linked via their central galactose residues (185). Starfish interacted with sub- nanomolar affinity with Stx1 and low nanomolar affinity with Stx2. The 5x2 carbohydrate design was originally planned to accommodate two binding sites on each B subunit. However, crystallography showed that Starfish was sandwiched by two B5 pentamers, with each Gb3 bound to only one site.

Starfish and Daisy (6, Figure 2.11), a related construct where the Gb3 trisaccharide is linked via its reducing end, were tested in vivo (186). When co-injected subcutaneously with a lethal dose of either Stx1 or Stx2, Daisy was shown to offer protection, whereas Starfish was only effective against Stx1. The higher activity of Daisy against Stx2 was attributed to the flexibility of the linker, although the positioning and hydrophobic nature of the linker could also have played a role, as it was the case for the

Gb3 polyacrylamide. In a mouse model challenged with E. coli O91:H21 strain, daily subcutaneous

CHAPTER 2: TARGETING BACTERIAL TOXINS doses of Daisy 24 h after inoculation with bacteria protected 50% of the mice from death. Overall, Daisy had lower activity in vitro than Starfish but outperformed it in vivo, possibly because of different pharmacokinetic profiles. Alternatively, the biodistribution or excretion of the toxin could be different with Daisy. The authors only tracked the effect of Starfish on the biodistribution of 125I-labeled Stx1 and Stx2. It was found that Starfish reduces radioactivity found in the kidney, brain, spleen, and liver for Stx1 but not Stx2.

FIGURE 2.11 Multivalent Gb3 presentation for SLT inhibition.

A dendritic carbosilane core was also used to present the Gb3 trisaccharide in a multivalent fashion (187). The hexavalent compound nicknamed Super twig(1)6 (7, Figure 2.11) provided complete protection to mice when co-injected with a lethal dose of Stx2 (188). Importantly, Super twig was able to protect mice when injected twice daily, starting on day 3, in an E. coli O157:H7 infection model. Here again, the in vitro binding affinities of various constructs were not necessarily correlated with in vivo efficacy. The authors identified the ability of the complex to promote uptake and degradation by macrophages, perhaps through scavenger receptors, as being an important factor for in vivo activity. Indeed, Super twig was able to increase the amount of radioactivity recovered from the liver and spleen when co-injected with 125I-labeled Stx2, when compared with the labeled toxin injected alone. These examples demonstrate that in vitro binding is not always predictive of in vivo activity, and that parameters such as pharmacokinetic profile and fate of the toxin-binder complex are important as well.

More recently, a fullerene pentavalent Gb3 glycoconjugate has been synthesized, but no binding assay to the toxins was reported (189). The presentation of Gb3 on surfaces or gold nanoparticles also gives good affinities that could be exploited for detection and analysis (190, 191).

CHAPTER 2: TARGETING BACTERIAL TOXINS

FIGURE 2.12 GM1 pentasaccharide LT and CT receptor and simplified analogs.

V. cholerae CT and the closely related E. coli LT are both AB5 toxins that have also been the subject of numerous studies aimed at finding inhibitors of their interaction with the pentasaccharide GM1, their natural receptor. Dendrimers have been functionalized with GM1 (8, Figure 2.12) (192– 194). The multivalent display offered significant improvement over the affinity of the monomeric ligand alone. Given the complexity of the GM1 pentasaccharides, simpler moieties were explored for multivalent display. The GM1 mimic 9 (Figure 2.12), where the sialic acid is replaced by a lactic acid moiety and the branching lactose is simplified, has been presented on dendrimers and yielded more potent inhibition than GM1 (195). The terminal galactose being the most buried sugar in the toxin binding pocket, a simple β-galactose residue radiating from a pentacyclen core (10) was also shown to have good binding affinity, albeit still lower than the GM1 natural ligand (Figure 2.12) (196). The length of the linker in this case was optimized so that it would match the size of the toxin in order to have optimal binding. More recently, dendrimers decorated with 4 or 8 galactose residues were found to have inhibition potency against CT similar to GM1 itself (193). Comparable dendrimers decorated with carbohydrates have also been used on surfaces to probe multivalency effects on microarrays (197). A small structure-activity relationship study to identify other improved ligands has revealed that m- nitrophenyl-α-D-galactoside (MNPG 11, Figure 2.12) was 100 times more active than galactose in a displacement assay with LT and CT (198). While this compound is still 4 orders of magnitude away from the GM1, it constitutes a good starting point for designing a multivalent binder. The presentation of this moiety in a pentavalent fashion (12) was indeed accompanied by an affinity increase of 260-fold over the monomeric ligand in a displacement assay with CT (199). However, the optimal length of the linker arm that was previously optimized on the galactose construct could not be reproduced with the MNPG because of solubility problems, so the construct length was sub-optimal, but active nonetheless. A later study explored the activity of “nonspanning” bivalent MNPG molecules, where the distance 26

CHAPTER 2: TARGETING BACTERIAL TOXINS between the two ligands is too short to accommodate two binding sites at the same time (200). In this case, the improved inhibitory activity of the dimer, compared to its monovalent counterpart, cannot be explained by multivalency. Steric inhibition was proposed as a possible explanation. One strategy to increase the binding affinity of a carbohydrate, beyond multivalency, is to gain additional binding with a different, non-sugar moiety as was shown by the addition of a nitrophenol group on the galactose for binding to CT and LT. A fragment-based approach was recently used where a multivalent galactose residue with a pendant azide was reacted with a library of simple alkynes via “click chemistry” (201). Screening of the resulting library in a displacement assay with CT yielded a low nanomolar multivalent inhibitor (13, Figure 2.12). The monomeric ligand had similar affinity to the monomeric MNPG. Glycopolymers with a well-defined and controlled distance between ligands are difficult to obtain if the distance is large. The only controllable parameter is the degree of functionalization of the polymer backbone, which gives a distribution of distances between ligands. Synthetic glycopeptides, where a helical or random coil peptide is decorated at specific amino acid with a galactose residue, were used in order to control with precision the distance between the ligands (202). These glycopeptides displayed up to 340-fold inhibition enhancement over galactose in a displacement assay with CT. An optimal distance of 35 Å between galactose residues and a helical conformation of the peptide both had a positive effect on the inhibition. More recently, random-coil bivalent glycopeptides were synthesized to evaluate the influence of the charge of the peptide backbone on the binding (203). As expected from the presence of basic residues on the protein surface surrounding the carbohydrate binding sites, negative charges on the backbone improved the inhibition of CT interaction with its receptor.

Ultimately, none of the previous molecules were tested in cell culture or in vivo, so their therapeutic potential against CT or LT is unclear. As described above, the binding and inactivation of CT or LT with an oral binder during an infection will be very challenging. The interest of these studies lies in the lessons learned in multivalent design of potent protein binders.

An elegant solution for the controlled ordering of random polymers is to use a template to pre- organize the ligand in a defined orientation. Serum amyloid P component (SAP) is a pentameric protein present in blood that was found to be the endogenous neutralizing factor for Stx2 in humans (204). This protein circulates with concentrations of around 30 mg/L but only in humans. In fact, mice receiving twice daily injections of human SAP were completely protected against a challenge with a lethal dose of Stx2, but not against Stx1 (205). Transgenic mice expressing human SAP at levels similar to that of humans survived significantly longer than wild type mice in the Stx2 challenge, but ultimately succumbed. The authors of this study suggested that administration of SAP could be a potential therapy for HUS. SAP itself is a therapeutic target because of its presence in amyloid deposits. A small molecule that binds to SAP and inhibits its binding to amyloid fibrils was identified (206). The dimerization of this molecule resulted in a ligand capable of assembling the pentavalent protein into a dimer sandwich that was quickly removed from circulation by the liver. A similar concept was applied to induce the

CHAPTER 2: TARGETING BACTERIAL TOXINS complexation of SAP with either SLTs or CT (Figure 2.13). A small molecule combining a SAP ligand and MNPG (14, Figure 2.13) was shown to inhibit the binding of CT to its receptor and, more importantly, the presence of SAP increased the potency of 14 by almost 3 orders of magnitude (207). A similar type of ligand was also designed for inhibition of SLTs, with Gb3 as the SLT ligand and a pyruvate ketal as the SAP ligand (15, Figure 2.13) (208). In this case a pentavalent presentation of bifunctional ligand 15 was also evaluated (15-pentamer, Figure 2.13). As expected, both ligands were active in an ELISA assay, with the pentamer showing lower IC50, and both displaying a significant decrease in IC50 in the presence of SAP. Perhaps not surprisingly, the effect of SAP was more dramatic for 15 (280-fold) than for 15-pentamer (35-fold), as this compound is already pre-organized into a pentamer.

An improved bifunctional ligand design led to compound 16, that merges the Gb3 and the pyruvate ketal without rotable bonds in between the two ligands (209). This change led to a 50-fold improvement in the IC50 with an enhancement of inhibition of around 4 orders of magnitude in the presence of SAP. Remarkably, this small molecule was as active as Starfish (Figure 2.11) in a Vero cell cytotoxicity assay. Unfortunately, compound 16 was inactive in a mouse model because of its rapid clearance by the kidney. Acrylamide polymers functionalized with either 15 or 16 showed increased inhibition of Stx1 compared to the monomeric ligand, and a large enhancement in the presence of SAP (15-polymer, 16-polymer, Figure 2.13) (210). Interestingly, a polymer displaying randomly both a SAP ligand and Gb3 did not show any improvement in inhibition in the presence of SAP. Unlike the monomeric 16, 16-polymer showed activity in transgenic mice expressing human SAP. The polymer has a higher hydrodynamic volume and probably exists as a complex with SAP, which would explain its increased circulation time, compared to monomeric 16. The polymer was able to protect 100% of mice against Stx1-induced lethality in a model where Starfish was only able to delay death. No in vivo experiments were performed with Stx2 because of its binding to human SAP in the absence of ligand. 16-polymer was shown to influence the biodistribution of 125I-Stx1, notably reducing the amount of radioactivity in the kidneys and brain. Most of the toxin appeared to be metabolized in the liver. Clarification of the precise role of SAP in the context of human infection will be crucial in determining the clinical potential of this approach (211).

CHAPTER 2: TARGETING BACTERIAL TOXINS

FIGURE 2.13 A, Templating strategy where a circulating pentameric protein is used to pre-organize a monomeric or polymeric bifunctional ligand to effectively bind the B5 subunit of a toxin. B, Different bifunctional ligands for SAP-templated inhibition of CT and SLT.

2.2.1.4 Multivalent Peptide Binders

Whereas carbohydrates represent an obvious starting point in the design of molecules intended to inhibit the binding of toxins to their carbohydrate receptor, they offer certain disadvantages. First, the affinity of the monovalent ligand for the toxin is almost invariably low, thus requiring the multivalency described so far to obtain high affinity binders. Second, some of the more complex carbohydrates are still difficult to synthesize in large quantities. One alternative is to use peptides, which can mimic carbohydrate epitopes. Furthermore, peptides are easier to synthesize and libraries of peptides for screening are readily assembled by solid-phase synthesis. A library of tetravalent peptides, presented on a lysine trimer, was screened for its interaction with the Stx2 B5 subunit (212). Peptide 17 (Figure 2.14)

CHAPTER 2: TARGETING BACTERIAL TOXINS protected mice even when given intragastrically 3 days after infection with E. coli O157:H7. Surprisingly, the peptide did not prevent cellular uptake of the toxin to the Golgi, but disrupted the transfer to the ER. This system worked only intragastrically in mice, and not intravenously (i.v.) (213). However, when evaluated in a non-human primate toxemia model that displays a pathophysiology closer to human HUS, the tetravalent peptide 17 was effective at rescuing baboons from an otherwise lethal dose of Stx2 after i.v. dosing (214). When 17 was administered at 5 mg/kg 24 h after challenge with Stx2, followed by daily 1 mg/kg dose up to day 4, three out of four baboons survived, and showed protected kidney functions and preserved urine output. While closer to human HUS, this model is not perfect and does not involve an infection with bacteria. Further evaluation in an infection model will certainly be of interest for bringing such a therapy closer to the clinical stage.

FIGURE 2.14 Tetravalent presentation of the SLT-binding peptide.

Multivalent peptides can also be used as synthetic equivalents of soluble decoys (see section 2.2.1.2) to prevent protein-protein interaction between the toxin and its receptor or prevent toxin assembly. For anthrax, multivalent peptides were used to inhibit both PA-receptor and PA-LeF interactions. A phage display library was employed to identify a dodecameric peptide that bound to

PA63, albeit modestly, and weakly inhibited its interaction with LeF (215). When this peptide was conjugated to a polyacrylamide backbone, the multivalency resulted in a lowering of IC50 to 20 nM in a cytotoxicity assay, a drop of almost 4 orders of magnitude. The polymer protected rats against LeT, when injected 3 – 4 min after the challenge.

A similar approach was applied to identify a peptide that binds the cell surface receptors TEM8 and CMG2 (216). A phage display selection experiment identified a peptide that prevented association between both receptors and PA. Multivalency was achieved by decorating liposomes with multiple copies of the peptide. While this peptide was technically designed to target the receptors of the toxin, and not the toxin itself, the inhibition on cells turned out to work by steric inhibition of the LeF binding to the heptameric PA63, not by inhibiting the interaction between PA and the receptor. Presumably, the peptide-binding site on the receptor does not overlap with the PA-binding site, so the binding and assembly of PA63 are not inhibited, but the complex is then unable to bind to the LeF. To explain the discrepancies with the phage display results, the authors speculated that steric inhibition from the phage might have disrupted PA binding. Nevertheless, co-injection of LeT with the liposomes prevented rats from becoming moribund.

CHAPTER 2: TARGETING BACTERIAL TOXINS

2.2.2 Approaches Targeting Toxin Trafficking

Once endocytosed, A-B toxins reach the cytosol by one of two possible routes: either directly from endosomes (e.g., anthrax, diphtheria toxin, TcdA, TcdB) or from the ER after undergoing retrograde transport from early endosomes through the TGN (e.g., CT, SLTs) (217). The latter route of transport appears to be unique to toxin proteins, offering a potential target for inhibitory molecules.

The only molecules that have been shown to inhibit retrograde toxin trafficking in vivo to date are Retro-1 and Retro-2 (18 and 19 respectively, Figure 2.15A) (218). Indeed, the intraperitoneal (i.p.) injection of Retro-2 at doses as low as 2 mg/kg were found to protect mice from a nasal RT challenge

(21-day LD90) when administered 1 h prior to the toxin. Furthermore, doses as high as 400 mg/kg were not toxic to the animals. Retro-1 and Retro-2 appear to act on the cellular trafficking machinery rather than on the toxins and as such also have the potential to inhibit other toxins that rely on retrograde trafficking. The authors found that relocalization of the SNARE protein syntaxin 5 correlated with blockage of SLTs in early endosomes. A better understanding of the mechanism of action of Retro-1 and Retro-2 may reveal their actual target and allow for their further optimization.

2.2.3 Approaches Targeting Toxin Translocation

A common feature of A-B type toxins is their ability to transfer an enzymatic domain to the cytosol of target cells. Translocation is the series of conformational changes that toxins undergo which result in a shift of the catalytic domain from the non-cytosolic to the cytosolic side of the plasma membrane. The translocation of toxins that enter the cytosol directly from endosomes is triggered by acidification of the local environment, in contrast to the translocation of toxins that enter the cytosol following retrograde transport, which is mediated by chaperones and translocators (217). Indirect inhibition of A-B toxins that depend on endosome acidification for translocation can be achieved by administration of drugs that block endosomal acidification but this does not constitute a viable method of treatment given the potential toxicity issues associated with such a therapy (219).

FIGURE 2.15 A, Inhibitors of the retrograde route of trafficking employed by e.g. SLTs. B, β-cyclodextrin derivative that blocks EF and LeF translocation through the PA-heptamer pore. 31

CHAPTER 2: TARGETING BACTERIAL TOXINS

The translocation of anthrax EF and LeF has been studied in particular detail and a number of therapeutic leads have been developed that block the translocation process, thereby inhibiting pathogenesis (220). An especially successful approach has been to inhibit translocation of EF and LeF by blocking the channel of the heptamerized PA pore. Indeed, β-cyclodextrin derivatives with a seven- fold symmetry that matches that of the PA pore have been shown to have in vivo efficacy in anthrax infection models (221–223). The most recently reported development of this idea consists in functionalizing cyclic β-cyclodextrin with peptides that specifically bind PA using poly(ethylene glycol)

(PEG) linkers (20, Figure 2.15B) (224). The authors reported an IC50 value of about 10 nM for the heptavalent peptidic inhibitor, representing a significant improvement compared to amine-decorated β- cyclodextrin derivatives. In vivo results with the peptidic inhibitor showed that it protected rats from anthrax toxin-mediated morbidity when co-injected with toxin.

A-B toxins other than anthrax do not form stable pore structures through which the A moiety can translocate, making it impossible to target them in an analogous manner. However, translocation via the ER-associated degradation pathway requires an unfolding step that can be targeted. For example, 4- phenylbutyric acid was shown to prevent the thermal unfolding, and thus translocation of the A1 subunit of CT (113). This finding provides a new avenue to explore for the inhibition of toxin translocation.

2.2.4 Approaches Targeting Toxin Processing

2.2.4.1 Proprotein Convertase Inhibitors

Proprotein convertases (PCs) are a nine-member family of Ca2+-dependent serine endoproteases, seven of which target multibasic motifs (furin, PC1/3, PC2, PACE4, PC4, PC5/6, PC7) (225). Found in the secretory and endocytic pathways and at the surface of mammalian cells, these PCs are involved in activating protein precursors by cleaving them at conserved recognition sites (e.g. R–X–K/R–R↓–X) (226). Furin, a 96-kDa type-I transmembrane glycoprotein has been studied in particular detail because of its involvement in a multitude of developmental, homeostatic and disease-related cellular processes (227). Furin and other PCs process key molecules in diseases ranging from Alzheimer’s disease to cancer and viral infections, as well as bacterial toxins. Key examples of toxins that PCs activate include anthrax toxin, SLTs and diphtheria toxin (DT, C. diphtheriae). Controlled inhibition of furin and other PCs has been explored as a possible approach to limit the progression of diseases caused by these toxins (228–230). Anthrax toxin in particular is an attractive target for such an approach because it is processed by PCs at the cell surface (or in the extracellular space), in contrast to SLTs and DT, whose processing is intracellular (149). These approaches will only briefly be considered here, as they target toxins indirectly.

The PC inhibitors that have shown the most promise to date are short peptides (and their analogs). The crystal structure of furin covalently bound to the potent PC inhibitor decanoyl-RVKR- chloromethylketone (21, Ki = 1 nM, Figure 2.16) has greatly contributed to the design of novel inhibitory 32

CHAPTER 2: TARGETING BACTERIAL TOXINS peptides by revealing the key features of the active site cleft (231). An approach that has recently shown promise in further increasing stability of peptidic inhibitors was to replace the P1 arginine of 21 by a decarboxylated peptidomimetic residue such as 4-amidino-benzylamide (22, Ki = 0.8 nM, Figure 2.14) (232). This reduces the residue’s susceptibility to degradation by carboxypeptidases and enables tighter binding with the target enzyme’s active cleft. The lack of a C-terminal ketone group also makes the molecule less prone to racemization and nucleophile attack.

A comparison of the residues of the catalytic cleft subsites based on homology modeling of the seven PCs reveals that the S1 to S5§ subsites are highly conserved (228). This implies that peptides must be at least 6 residues long to have any degree of selectivity between PCs. Recently developed peptidic PC inhibitors are based on the extended cleavage motif of a viral substrate, the avian influenza H5N1 hemagglutinin (233, 234). The 9-residue RARRRKKRT peptide with an N-terminal 8-amino-octanoyl group (23, Ki = 8 nM, Figure 2.16) was found to reduce fatality in an inhalational anthrax model in mice at low dosage (5 mg/kg i.p.). Molecule 23 was not toxic, possibly because it selectively inhibited furin, thus allowing for endogenous processing by other PCs required for homeostasis.

All therapeutic approaches based on PC inhibitors share a common challenge: managing the functional redundancy of the PC family in processing both pathogenic and endogenous substrates. It also remains to be seen to what extent inhibition of PCs can hinder the pathogenicity of toxins other than anthrax such as SLTs, where penetration of the active molecule into the cell is required for inhibition.

FIGURE 2.16 Small molecule inhibitors of autoprocessing and PC-mediated processing of toxins.

2.2.4.2 Autoprocessing Inhibitors

TcdA and TcdB from C. difficile also require processing to release the enzymatic domain into the cytosol after endocytosis but rather than undergoing cleavage by PCs, TcdA and TcdB both autocleave themselves through the action of their own CPD (235). A similar CPD is also found in the V. cholerae MARTX toxin (236). A peptide derivative has recently been designed to covalently inhibit

§ A nomenclature commonly used to describe the sequence of peptide residues relative to the cleavage site of protease substrates denotes the amino acid that is amino-terminal to the cleavage site as P1. Subsequent residues are named P2, P3 and so on; the amino acids that are carboxy-terminal to the cleavage site are denoted P1´, P2´and so on. The corresponding protease binding sites are denoted with an S rather than a P. 33

CHAPTER 2: TARGETING BACTERIAL TOXINS the TcdB CPD (24, Figure 2.16) (237). The dipeptide contains the same amino acids as those in the P1 and P2 positions of the CPD’s natural cleavage site and the authors observed that adding a bulky hydrophobic group at the N-terminus increased potency of inhibitors in comparison to a smaller acetyl group. Interestingly, tripeptide inhibitors were less potent than dipeptide ones. Molecular docking studies suggest that minimal interactions exist between the inhibitor’s peptide backbone and the TcdB CPD, which implies that the development of non-peptidic inhibitors that bind the CPD in a similar fashion to 24 is feasible. Given the similarity of the TcdB cleavage site with that of TcdA, it is conceivable that 24 also inhibits TcdA autocleavage.

An alternative approach to inhibit TcdA and TcdB autoprocessing is to induce S-nitrosylation of the active site cysteine of the toxins, a process believed to be an endogenous mechanism of host protection from the toxins(238). The S-nitrosylation of TcdB by S-nitrosoglutathione (GSNO), an endogenous source of nitric oxide, inhibits IP6-mediated cleavage of TcdB in vitro. The therapeutic potential of this route of inhibition was confirmed in a mouse model of CDI where oral treatment with 10 mg/kg GSNO significantly increased probability of survival. Administration of GSNO intracecally and co-administration with 0.25 mg/kg IP6 resulted in a further significant increase in the protective effect. However, it is unclear to what extent GSNO may be degraded in the GI tract, or to what extent it may interact with other proteins in the GI tract.

Small molecule leads to inactivate TcdA and TcdB before they are internalized are particularly attractive because as the molecules need only be active in the lumen of the GI tract, membrane permeability is not a must. Furthermore, high concentrations of active compound in the GI tract can be reached.

2.2.5 Approaches Targeting Toxin Enzymatic Activity

Another approach to inhibit the action of bacterial toxins is to target the ultimate causative agent of cellular disruption, that is to say the toxins’ enzymatic “warhead”. Bacterial toxins can be grouped according to the cellular targets and/or the mechanism of action of their enzymatic domains. The enzymatic domains presented in this section are not exhaustive; examples of enzymatic warheads found in toxins that are either not important virulence factors or not A-B type toxins (and are therefore outside of the scope of this review) are phospholipases, deamidases, proteases (other than metalloproteases) and deoxyribonucleases (71).

2.2.5.1 Adenylyl Cyclase Inhibitors

Class II adenylyl cyclase (AC) toxins (also called adenylate cyclase toxins) include EF (B. anthracis) and adenylate cyclase toxin (CyaA, B. pertussis) (239, 240). ACs catalyze the formation of cAMP from ATP (Figure 2.17A) (241). cAMP being an important second messenger, an increase in its cytosolic concentration beyond physiological levels has a variety of potentially harmful effects that include increased ion fluxes and fluid secretion, which can cause edema and diarrhea, and impairment 34

CHAPTER 2: TARGETING BACTERIAL TOXINS of phagocytic functions (153, 242). Ca2+ and calmodulin (CaM) modulate the AC activity of EF and CyaA (243–245).

The inhibitors of the enzymatic activity of AC toxins developed to date inhibit either the binding of substrate (ATP) to the active pocket (competitive inhibitors) or the activation by CaM (non- competitive inhibitors) (246, 247). The crystal structure of EF with and without CaM has been extensively used to guide the design of EF inhibitors, and the structural homology that EF’s AC domain shares with CyaA implies that inhibitors specifically designed to target EF’s AC binding pocket are also likely to inhibit CyaA (147, 248).

Anthraniloyl-substituted nucleotides possess inhibitory activity against bacterial ACs (249).

One compound in particular, bis-Cl-ANT-ATP (25, Figure 2.17B) combines very high potency (Ki = 16 nM) against CyaA with 100- to 150-fold lower activity against mammalian ACs (250). Bis-substitution of nucleotides resulted in a large improvement in the selectivity of inhibitors for CyaA over mACs compared to mono-substitution because bulky groups are more readily accommodated in the larger CyaA substrate binding pocket.

Another compound that has recently been found to inhibit activation of EF and CyaA by CaM

(IC50 = 2 µM for EF) is thiophen uredoacil dichloride (26, Figure 2.17B) (247). The screening approach used to identify this compound involved constructing a plausible conformational path between EF’s inactive and CaM-activated states in order to identify a putative pocket that could serve as a binding site for an inhibitor of activation by CaM. Inhibitors targeting binding pockets other than the ATP-binding site of AC toxins have the advantage that they are likely to be more specific and less toxic than competitive inhibitors, given the fact that ATP is a common substrate for a multitude of enzymes.

A possible reason for the lack of in vivo validation of inhibitors of AC toxins is that they are always secreted along with other toxins that may be more important for virulence. In the case of anthrax, for example, only very high doses of ET are lethal to mice (≈ 2 mg/kg) in comparison to LeT (≈ 50 µg/kg) (245, 251). Consequently, AC toxin inhibitors would appear to have greater potential as adjunct therapeutics rather than as standalone curative agents.

CHAPTER 2: TARGETING BACTERIAL TOXINS

FIGURE 2.17 A, Mechanism of action of adenylyl cyclases. B, Inhibitors of toxin adenylyl cyclases.

2.2.5.2 ADP-ribosyltransferase Inhibitors

ADP-ribosyltransferase (ADPRT) toxins are a group of roughly 20 toxins that have mono-ADP- ribosyltransferase enzymatic activity but are otherwise not structurally related (252, 253). ADPRT toxins bind NAD+, from which the ADP-ribose moiety is transferred to an acceptor protein (Figure 2.18A). DT (C. diphtheriae), CT (V. cholerae), LT (E. coli), exotoxin-A (P. aeruginosa) and pertussis toxin (PT, B. pertussis) are noteworthy examples of ADPRT toxins (254). Although ADP-ribosylation modulates the functions of a host of endogenous eukaryotic enzymes, ADPRT toxins target only a small subset of them: actin and members of the G-protein family including heterotrimeric G-proteins (e.g. Gs and Gi), eukaryotic elongation factor 2 (eEF2) and Ras proteins (e.g. Rho family of GTPases). DT, exotoxin-A and cholix toxin (V. cholerae) comprise a distinct subset of ADPRT toxins in that they ADP- ribosylate a modified amino acid residue (diphthamide) on their common target (eEF2), a property that can be exploited to tune inhibitor selectivity (255). ADPRT toxins can be particularly cytotoxic, a classic example being that one molecule of DT fragment containing the ADPRT domain is sufficient to kill a cell (256). Although the enzymatic domains of ADPRT toxins are attractive therapeutic targets, only a limited number of lead candidates with therapeutic potential have been identified so far.

The small molecule PJ34 (27, Figure 2.18B) is a non-specific competitive inhibitor of P. aeruginosa exotoxin-A ADPRT activity (257). Although PJ34 inhibits exotoxin A potently (Ki = 140 nM), it is also an inhibitor of mammalian poly(ADP-ribose) polymerases (PARPs) in mouse models of stroke and therefore likely lacks the specificity to ADPRT toxins required for therapeutic use. The crystal structure of PJ34 bound to cholix toxin has recently been used as a template to carry out virtual screening with the aim of identifying novel inhibitors of ADPRT toxins (255, 258). The screen included a large library of drug-like commercial compounds and a much smaller library of known PARP inhibitors. Two compounds, 28 and 29 (Figure 2.18B) were found to combine high activity against

CHAPTER 2: TARGETING BACTERIAL TOXINS cholix toxin and ExoA in cell-based assays with low toxicity and could form the basis for efforts to rationally design inhibitors with improved binding affinity.

FIGURE 2.18 A, Mechanism of action of ADPRTs. B, Inhibitors of toxin ADPRTs.

A bisubstrate analog incorporating benzamide and guanidine moieties that mimic substrate + NAD and arginine respectively (30, Figure 2.18B) was found to have a Ki of 8 µM against CT but it remains to be seen how this compound performs in cellular assays (259). Designing inhibitors that selectively target the ADPRT domain of toxins is a challenge because despite the limited sequence identity between them and endogenous ADPRTs, they display similar structures in the NAD+ binding region designated the “scorpion” motif (260). Interaction with less conserved residues beyond the scorpion motif is likely necessary for molecules to discriminate between endogenous ADPRTs and ADPRT toxins. Novel fluorescence-based assays for high throughput screening of lead compounds should contribute to the expansion of the currently restrained number of ADPRT toxin inhibitors (261).

2.2.5.3 Metalloprotease Inhibitors

Bacteria secrete a large and diverse number of zinc-metalloproteases but only a handful of these are classified as A-B type toxins (Clostridium tetani neurotoxin (TeNT) and Clostridium botulinum neurotoxins (BoNTs), Bacillus anthracis LeT and Bacteroides fragilis fragilysin) (71, 262–264). It is worth noting that the two most potent bacterial toxins known, the clostridial neurotoxins (CNTs) BoNT 37

CHAPTER 2: TARGETING BACTERIAL TOXINS serotype A (BoNT/A) and TeNT, are metalloproteases (251, 265). Recombinant B. anthracis producing LeF with a catalytic point mutation in the metalloprotease site does not kill mice, suggesting that inhibiting metalloprotease activity is a promising therapeutic approach for anthrax treatment (266). Anthrax LeT and BoNTs (both Category A bioterrorism agents) are the metalloprotease toxins that have been the most popular targets for inhibitor design in recent years and are the focus of this section (267, 268). Metalloprotease toxins share a strictly conserved zinc-binding HEXXH motif but do not share a high degree of sequence homology outside of the zinc-binding motif, except in the case of CNTs (262, 269). In fact, the CNTs BoNT/B and TeNT have over 50% sequence identity and CNTs all target proteins composing the neuronal SNARE complex by similar mechanisms, ultimately inhibiting acetylcholine release at neuromuscular junctions (270, 271). Although C. botulinum produces 7 different serotypes of BoNT (named A-G), only four of these (A, B, E and F) are causative agents of botulism disease in humans. Efforts to design inhibitors of BoNTs tend to focus on BoNT/A because it is the most potent and has the longest half-life of neurotransmitter release inhibition (t1/2 > 31 days) (272–274).

Molecules containing a metal-chelating hydroxamic acid group have proven to be particularly efficient inhibitors of the metalloprotease activity of BoNTs and LeF (267, 272). The sulfonamide-based hydroxamate 31 (Figure 2.19) protected all rabbits tested in a so-called “point of no return” model when co-administered sub-cutaneously with ciprofloxacin, whereas ciprofloxacin alone protected 50% of rabbits from death (76). The molecule binds the LeF metalloprotease domain competitively (Ki = 24 nM), with the hydroxamic acid moiety interacting with the zinc ion via a bidentate interaction (76). The binding pocket of LeF at the S1´ site is smaller and tighter than that of matrix metalloproteases and accommodates the aromatic moiety of the molecule snuggly, conferring it with selectivity for LeF over matrix metalloproteases and several other endogenous proteases and protease-like enzymes (275). Pharmacokinetic parameters for 31 have also been determined in several preclinical species and the results suggest that it could even be administered orally. Recent efforts have investigated structural modifications to the 31 that confer it with increased in vivo efficacy (276). Optimization resulted in compound 32 (Figure 2.19) with a significantly improved Ki (0.24 nM) that protected all rats tested from death after LeT challenge at doses of 2.5 mg/kg (276).

Hydroxamates have also yielded encouraging in vivo results as inhibitors of the metalloprotease activity of BoNTs. A computer-aided optimization using a cation dummy atom (CaDA) approach was used to design a micromolar inhibitor of BoNT/A, F4H (33, Figure 2.19) (277, 278). I.p. pretreatment with F4H 30 min before toxin challenge protected 100% of mice from death during 12 h after being challenged with BoNT/A, while 40% of untreated mice survived during the same time lapse (278). F4H also has a relatively long half-life in mice (t1/2 = 6.5 h). However it is unclear whether F4H protects mice from BoNT/A by binding extracellular toxin or by acting on the toxin directly in intoxicated neurons. It also remains to be shown to what extent F4H may inhibit endogenous MMPs, although the molecule does not appear to elicit any acute toxicity in vivo.

CHAPTER 2: TARGETING BACTERIAL TOXINS

FIGURE 2.19 Inhibitors of metalloprotease toxins.

The use of rhodanine derivatives as inhibitors of LeF and BoNT/A has also been explored (279, 280). A pyrazole derivative of rhodanine (34, Figure 2.19) has been shown to inhibit BoNT/A and LeF competitively in the low micromolar range and the compound increased survival of mice infected with B. anthracis Sterne spores with treatment starting 24 h post-exposure. However, it must be noted that rhodanines have been recognized as “pan assay interference compounds” because of their promiscuity and reactivity (281–283). The potential of rhodanines as therapeutic molecules therefore needs to be validated with rigorous assaying of selectivity and toxicity.

Peptide-like compounds modeled on residues around the BoNT/A substrate (SNAP-25) scissile bond have also been shown to inhibit bacterial metalloproteases in vitro (e.g. 35, Figure 2.19) (284, 285). The insights into the conformation and chemical contacts with the target gained from structural studies can be used for database mining of virtual small molecule libraries with the aim of obtaining hits that combine the potency of peptide-like compounds with the drug-like qualities of non-peptidic small molecules (75). An example of leads optimized in this fashion are the regioisomers 36 and 37 (Figure 2.19), the first non-zinc-coordinating non-peptidic small molecule inhibitors of BoNT/A active in the sub-micromolar range in vitro (286).

A systematic study of the effect of peptide length on inhibitory activity against BoNT/A has newly demonstrated that a simple tetrapeptide is sufficient to competitively inhibit BoNT/A with high potency (287). The tetrapeptide RRGC (Ki = 158 nM) mimics the first four residues of the endogenous 39

CHAPTER 2: TARGETING BACTERIAL TOXINS substrate, QRATKM, and therefore acts in a truly competitive fashion. The tetrapeptide was shown to be specific to BoNT/A and inhibited BoNT/A-mediated SNAP-25 cleavage in multiple assays using chick motor neurons, primary rat and mouse cerebral neurons at micromolar concentrations. Furthermore, the ∼500 Da tetrapeptide displayed favorable drug-like characteristics such as high solubility in water, resistance to intracellular proteases, high cell penetration due to its cationic character and low toxicity (at 600 µM), making it a promising therapeutic lead.

2.2.5.4 RNA N-glycosidase Inhibitors

Shiga toxins and SLTs produced by S. dysenteriae and E. coli, respectively, are the most studied bacterial toxins having RNA N-glycosidase activity (131, 288). They are ribosome-inhibiting proteins (RIPs) that act by cleaving an adenine group in a highly conserved sequence of 28S ribosomal RNA, thereby inhibiting elongation factor 1-dependent aminoacyl-tRNA binding to ribosomes and ultimately inhibiting protein translation (Figure 2.20A) (289). The mode of action is very similar to that of the highly structurally related RT (290).

The challenge in developing competitive inhibitors of RNA N-glycosidase toxin activity is that mimicking the endogenous substrate, adenine, yields molecules that lack the solubility and potency required for therapeutic use. Furthermore, the active site of RNA N-glycosidase toxins is very large making it a difficult target for drug development (291). A recent study showed that the most potent inhibitor of Stx1A1 and RT in vitro found to date (38, IC50 ≈ 30 µM, Figure 2.20B) was only moderately active in cell-based assays, an observation that could be attributed to the compound’s poor solubility (291). This molecule acts in a manner similar to 39 (Figure 2.20B), another inhibitor of Stx1A1 and RT, in that it interacts with its target when the enzyme is in a closed or inactive conformation (292). Binding in the closed conformation offers a more limited surface for specific interactions than binding in the open conformation, in which adenine is endogenously accommodated (291). Virtual screening approaches that exploit structural differences between open and closed conformations of SLTs and RT have recently been used to identify novel inhibitors that were subsequently tested in cellular assays

(291). One of the most potent compounds identified, 40 (IC50 ≈ 200 µM, Figure 2.20) inhibited RT activity in a cellular assay but failed to inhibit Stx1A1 and was cytotoxic at concentrations above 30 µM.

The authors of a study using a similar approach for lead validation warn that luciferase assays sometimes used for cell-free screening of lead compounds can result in false positives because of the similarity in structure between leads and D-luciferin, the substrate for firefly luciferase (293). The development of more robust high throughput screens for RNA N-glycosidase toxin inhibitors could hence contribute to the discovery of new leads. Clearly, there is a trade-off between maximizing inhibitor-target interactions to increase affinity and obtaining a drug-like, small molecule inhibitor of RNA N-glycosidase toxins but the aforementioned leads provide a basis for further developments.

CHAPTER 2: TARGETING BACTERIAL TOXINS

FIGURE 2.20 A, Mechanism of action of RNA N-glycosidases. B, Inhibitors of toxin RNA N-glycosidases.

2.3 Conclusion

Despite the unquestionable success of antibiotics in treating bacterial infections, effective therapies are plagued by a concomitant increase in resistance against antibiotics, and decrease in number of new antibiotics reaching the market. Furthermore, antibiotics are not always effective against toxin- producing bacteria such as EHEC and V. cholerae. The value of targeting bacterial toxins has long been recognized since the early days of toxin antiserum administration and the more recent success of antibodies and vaccines. Nevertheless, there has been a renewed interest in the field stemming from the knowledge gained in the mechanisms of action of bacterial toxins. So far, the targeting of toxins has yielded important knowledge on the interaction of molecules with proteins, in particular with respect to multivalency. However, several challenges remain to be addressed if any of these strategies are to progress further towards therapies.

In academic research, the drug delivery aspects such as permeability, solubility, and elimination are too often ignored when designing new molecules with a therapy in mind. The lack of in vitro / in vivo correlation of activity often observed for bacterial toxin binders is a reminder that the ultimate fate of the drug-toxin complex has to be considered. For instance, actively targeting clearance pathway such as the macrophage scavenger receptors might constitute an interesting line of research. In the case of molecules designed to bind the toxins in the GI tract, excretion is not an issue but the harsh environment and the physical barrier of the mucus layer that protects enterocytes should be taken into account early on in the design of the drug.

Some classes of enzymes are inherently difficult targets. For example, while the glucosyltransferase domains of C. difficile TcdA and TcdB are viable targets, the design of inhibitors is challenging because mimicking the UDP-glucose substrate yields molecules with poor drug-like characteristics. Therefore, a solution to the general problem of glycosyltransferase inhibitor development could yield new candidates against TcdA and TcdB. The use of peptides to bind to the 41

CHAPTER 2: TARGETING BACTERIAL TOXINS active site may constitute a breakthrough in this field (294). Similarly, RNA N-glycosidases are attractive but challenging targets. In the case of ADPRT toxins, the main difficulty is to obtain selectivity over endogenous ADPRTs. The continued increase in throughput and reduction in false-positives of the enzymatic assays used for inhibitor screening will play a key role in the discovery of potent inhibitors of toxin enzymatic moieties.

In addition, as it has been alluded to throughout this review, not all toxins are clinically sound targets. For example, EF from B. anthracis is probably not as important as LeF in causing disease symptoms and death. Similarly, we now know that TcdB of C. difficile cannot be ignored since it is as least as important as TcdA in the pathology. The socio-economical context of the disease also requires consideration. Although science should ideally be free of such constraints, it is clear that a complex and expensive therapy against cholera is unlikely to make an impact on disease burden in the face of simpler and cheaper alternatives like rehydration therapy. Similarly, elaborate syntheses pose a significant hurdle for commercial development of complex macromolecules, even for diseases affecting the developed world. Recent failures of advanced candidates demonstrate that it is also of crucial importance to consider the timing and location of toxin release when considering therapeutic opportunities. For instance, targeting SLTs for the treatment of HUS is challenging in part because the therapeutic window of opportunity is very small.

We anticipate that new mechanistic insights into the mode of action of toxins will open the door for novel and more effective therapies. For instance, identification of the actual receptors for C. difficile TcdA and TcdB could provide useful information for the design of inhibitors. Chemists are in a position to contribute to new understanding by designing chemical tools, and to exploit this knowledge to devise innovative strategies against these diseases. The myriad of toxins produced by bacteria, coupled with our increasing knowledge of their pathogenesis should provide ample grounds for academic research to thrive.

3 Therapeutic Potential of Triggering Pre-Emptive Clostridium difficile Toxin B Auto-Proteolysis

3.1 Introduction

As our reliance on the intensive use of antibiotics becomes increasingly precarious, alternative strategies to manage bacterial infections are urgently needed. Bacterial virulence factors, including toxins, are gaining interest as attractive drug targets to avoid putting selective pressure on bacteria and thereby reduce the likelihood of further emergence of resistance (73, 74, 295). The multidrug-resistant C. difficile is one of the most common nosocomial pathogens, causing a relentlessly growing number of life-threatening infections in hospitalized patients (6, 10). Worrying reports of community-acquired infections are also on the rise (8, 13). The large, monomeric, and closely related multi-domain proteins, TcdA and TcdB (Fig. 3.1A) secreted by C. difficile in the colon lumen, are responsible for pathogenesis, with TcdB being particularly virulent (88, 296). After binding to epithelial cells via the receptor binding domain and endocytosis of the toxin, a pH-dependent conformation change allows the N-terminal glucosyltransferase domain and the adjacent CPD to translocate into the cytosol (297). Intracellular IP6 (Fig. 3.1B), found ubiquitously in eukaryotic cells, acts as an allosteric activator of the CPD and triggers toxin auto-cleavage (93, 95, 96, 298). The thus freed glucosyltransferase domain glucosylates small GTPases, ultimately causing cytoskeleton disruption, cell death, inflammation and intestinal injury (299, 300).

IP6 exerts its allosteric effect by binding to a highly positively charged binding pocket on the CPD, thus stabilizing its active conformation via a crucial β-flap (Appendix Fig. 6.1) (301). IP6 has an unusually high charge density, and excellent calcium chelating properties (302). For this reason, despite being present at high concentrations in some food such as grains (rice in particular) and nuts, IP6 is not soluble in the gut lumen (303, 304). Although IP6 can trigger the auto-processing of TcdB in vitro, the cleavage is abolished if performed in the presence of luminal concentrations of calcium (Fig. 3.1C). Therefore, the basis for the intracellular nature of IP6-induced toxin auto-proteolysis is likely to stem in large part from the extracellular interference of calcium with IP6 and the constant intracellular availability of soluble IP6 (304). Other groups have recognized the attractiveness of the CPD as a therapeutic target and two approaches to inhibit the protease of TcdB have recently been described, but both require prior binding of IP6 to the protease for full inhibition, limiting their therapeutic potential (235, 238, 305). Furthermore, protease inhibition does not block uptake of the harmful enzymatic domain but merely prevents its release into the cytosol.

In this study, we describe gain-of-function molecules designed to trigger the toxin auto-proteolysis prior to its cell uptake (306, 307), thus preventing the internalization of the noxious enzymatic domain. Since the target site-of-action is the intestinal lumen, the molecule does not need to be absorbed in order 43

CHAPTER 3: PRE-EMPTIVE TOXIN B AUTO-PROTEOLYSIS to be active. We synthesized sulfated analogs of IP6 rationally designed to activate the CPD in the presence of calcium. Although replacing phosphates by sulfate groups provided additional solubility in calcium-containing media, only IP2S4, an inositol with four sulfates and two phosphates, retained allosteric activity on TcdB in physiologically relevant extracellular concentrations of calcium. In vitro assays were used to confirm that TcdB cleavage prior to uptake reduces cytopathy. Mice infected with C. difficile that received oral IP2S4 exhibited significantly less colitis than control mice. Our findings provide a proof-of-principle that pre-emptive toxin auto-proteolysis could be a viable therapeutic approach.

FIGURE 3.1 TcdB and its cleavage by IP6. A, Structural organization of TcdA, TcdB and the recombinant TcdB constructs used in this study. B, Structure of IP6. C, TcdB autoprocessing induced by incubation of 150 ng TcdB for 3 h at 37ºC with 1 mM IP6 in the presence and absence of 10 mM Ca2+.

3.2 Materials and Methods

3.2.1 Study design

The goal of this pilot study was to assess the effect of administration of IP2S4 on colonic inflammation in mice infected with TcdA- TcdB+ C. difficile. We used a germ-free mouse model of CDI that closely resembles pathology in humans to demonstrate IP2S4 efficacy in reducing colitis, as evidenced by measuring the activity of myeloperoxidase (MPO), a marker of inflammation, in colon explants. Group size was estimated from preliminary results. The investigators carrying out animal studies were blinded to group allocation during the experiment and when assessing outcomes. Mice were randomly assigned to each of the experimental groups and no outliers were excluded from the analysis. The experiments were carried out as two sets of individual experiments (each set comprising n = 3 – 5 animals per group) with two different synthetic batches of IP2S4.

3.2.2 Animal care

Germ-free NIH Swiss mice were rederived at the McMaster University Axenic Gnotobiotic Unit (AGU) by an axenic two-cell embryo transfer technique previously described (308). Germ-free mouse colonies were maintained in flexible film isolators at the AGU, and germ-free status was routinely 44

CHAPTER 3: PRE-EMPTIVE TOXIN B AUTO-PROTEOLYSIS confirmed by a combination of culture- and non-culture-based techniques in faecal and caecal contents (309). Periodic serological testing was also performed for viruses, parasites, and known pathogens (Charles Rivers Laboratories). All mice had unlimited access to autoclaved food and water. Both male and female mice were used at the age of 8 to 12 weeks. All experiments were carried out in accordance with the McMaster University animal utilization protocols.

3.2.3 Bacterial culture for monocolonization

LB broth was aseptically inoculated from single colonies of Escherichia coli JM83 and incubated with shaking at 160 rpm at 37°C for 12 hours. Bacteria were harvested by centrifugation (15 min, 3500X g) in a 400ml sterile flask, washed in sterile PBS and concentrated to a density of about 109 CFU/ml in PBS, all performed aseptically under a sterile laminar flow hood. Bacterial suspensions were sealed in sterile tubes, with the outside surface kept sterile, and imported into gnotobiotic isolators.

3.2.4 Testing of IP2S4 in mouse model of C. difficile infection

In order to prime the immune system, but limit colonization resistance provided by a complex microbiota, germ-free NIH Swiss mice were monocolonised with E. coli JM83 by intragastric gavage using 109 CFU in 200μl of PBS per mouse. Three weeks post mono-colonization, mice were gavaged with faecal content from a patient with recurrent CDI (C. difficile NAP1, 200 μL of a 1:10 dilution in saline per mouse). Mice then received IP6, IP2S4 or myo-inositol in drinking water (12.3 μmol/kg in 5 ml of water/mouse) starting 6 hours after C. difficile inoculation (n = 7 – 9). In addition, all mice received one dose of IP6, IP2S4 or myo-inositol by intragastric gavage (12.3 μmol/kg) at noon on days 2, 3, 4 and 5. On the afternoon of day 5, the mice were sacrificed and the colon excised for histological evaluation and myeloperoxidase activity as described before (310).

3.2.5 Analogue solubility measurements by ICP-MS

One hundred micromolar solutions of inositol hexakisphosphate (IP6) analogues with or without 10 mM CaCl2 were prepared in 10 mM tris pH 7.4 and incubated with agitation for 2 h at 37 °C. The solutions were immediately filtered through 0.2 μm nylon filters equilibrated to 37 °C. The phosphorous content in each filtrate was determined by inductively coupled plasma-mass spectrometry (ICP-MS). The values obtained were divided by the number of phosphates in each IP6 analogue to determine the concentration of the compound in the solutions.

3.2.6 Free calcium ion quantification

A fresh 1 mM solution of murexide (Merck, Germany) was prepared in 10 mM tris pH 7.4. For each IP6 analogue, samples containing 0.5 mM analogue, murexide and CaCl2 in 50 μL 10 mM tris were prepared in triplicate. After 5 min incubation at room temperature, the samples were centrifuged at 20’000 g for 2 min and the upper 40 μL of the supernatant transferred to a 384-well plate. Samples without IP6 analogue containing CaCl2 ranging from 1 mM to 20 μM, and 20 mM were also prepared

CHAPTER 3: PRE-EMPTIVE TOXIN B AUTO-PROTEOLYSIS and used for calibrating each experiment. The absorbance was measured at 474 nm and 544 nm and the data analyzed as reported by Ohnishi (311). The experiment was repeated in triplicate.

3.2.7 Plasmid construction

To generate recombinant cysteine protease domain from TcdB (rCPD, His6-TcdB511-843), the nucleotide sequence coding for amino acids 511—843 of TcdB from ribotype 078 C. difficile was amplified from genomic DNA (kindly donated by Prof. Roger Stephan, University of Zurich, Switzerland) by PCR with the forward primer GGCTAGCTGGACATTTGACGATGCAAG (NheI restriction site underlined) and the reverse primer CGAAGCTTTCAGGCTTCTTCAATCCTTTCCTCAA (HindIII restriction site underlined, STOP codon in italic). The DNA was subsequently cloned into pET28a plasmid (Novagen, Germany) using standard techniques and the resulting plasmid transformed into BL21(DE3). To generate truncated recombinant cysteine protease domain from TcdB (tCPD, His6-TcdB545-797), the nucleotide sequence coding for amino acids 545—797 was amplified by PCR using the forward primer GCCGCGGCTGAAGATGATAATCTTGATTT (SacII restriction site underlined) and the reverse primer CGAAGCTTTCACAGCTCAGGTAAATTTTT (HindIII restriction site underlined, STOP codon in italic). The DNA was cloned into pET47b (Novagen, Germany) and the resulting plasmid also transformed into BL21(DE3). All primers were purchased from Microsynth (Switzerland).

3.2.8 Recombinant protein expression and purification

rCPD and tCPD were both expressed and purified according to the following protocol. Overnight cultures of transformed BL21(DE3) were diluted 1:100 in 1 L Terrific Broth (Sigma-Aldrich, USA) and grown at 37 °C until an OD600 of 0.6—0.8 was reached. IPTG was added to a final concentration of 0.5 mM and the cultures grown for 3 h at 30 °C. The cultures were pelleted by centrifugation at 4’000 g for

15 min at 4 °C and resuspended in sonication buffer (20 mM NaH2PO4, 100 mM NaCl, 1 mM MgCl2) with 1 mg/mL lysozyme from chicken egg white (Sigma-Aldrich, USA) and 10 U/mL benzonase (Sigma-Aldrich, USA). The lysates were incubated for 30 min on ice and then sonicated with a probe sonicator before centrifugation at 4’000 g for 20 min. Recombinant proteins were purified from the cleared lysate by immobilized metal-ion affinity chromatography using 50 % Ni-NTA resin (Qiagen, Germany) according to the manufacturer’s instructions for batch purification. Eluted fractions containing protein were buffer exchanged into 20 mM tris, 60 mM NaCl, 250 mM sucrose pH 7.4 using PD Midi Trap G-25 columns (GE Healthcare Life Sciences, UK) and supplemented with 10 % glycerol before storage at -80 °C.

3.2.9 Intrinsic tryptophan fluorescence assay

The intrinsic fluorescence of the lone tryptophan residue in tCPD was exploited to determine the Kd of the interaction between IP6 analogues and the toxin fragment using an assay based on that described by Shen et al.(301) 200 μL of 20 μM tCPD in 20 mM tris, 60 mM NaCl, 250 mM sucrose pH

CHAPTER 3: PRE-EMPTIVE TOXIN B AUTO-PROTEOLYSIS

7.4 were added to three wells per IP6 analogue in a 96-well plate. Small aliquots of IP6 analogue stock solutions were added up to a maximal concentration of 200 μM. At each step, the samples were excited at 295 nm and the emission recorded between 325 nm and 360 nm on a plate reader (Tecan infinite M200, Tecan, Switzerland). All measurements were made at 25 °C. The maximum responses (obtained at an emission wavelength of 333 nm) were plotted against analogue concentration and the data fitted with a rectangular hyperbolic function. The data were normalized and the Kd determined for each IP6 analogue using OriginPro© (OriginLab©, MA, USA).

3.2.10 Trypsin digestion assay

tCPD (0.2 μg/μL final concentration) was added to 1 mM IP6 analogue in 10 μL 100 mM tris pH 7.4 and incubated at 37 °C for 30 min. Trypsin (1 μL) was added to a final concentration of 0.3 ng/μL and incubated at 37 °C for 10 min. Laemmli sample buffer (5X) was added to each reaction before heating for 3 min at 95 °C to stop the reactions. Controls with and without trypsin (both without IP6 analogues) were included in every experiment. The cleavage products were visualized by SDS-PAGE followed by Coomassie staining. The gels were imaged and band intensities were quantified using ImageJ (http://rsbweb.nih.gov/ij/). The extent of digestion as normalized to the controls.

3.2.11 Cleavage assays with recombinant toxin

Serial dilutions of the IP6 analogues were prepared in 100 mM pH 7.4 tris buffer to which 0.85

μM rCPD was added with or without 10 mM CaCl2 in a total volume of 13.3 μL. After a 2 h incubation at 37 °C, the reactions were put on ice and 10 mM EDTA added to the samples containing calcium. Laemmli sample buffer (5X) was added before heating the samples at 95 °C for 3 minutes. The cleavage products were visualized by SDS-PAGE followed by Coomassie staining. The gels were imaged and band intensities were quantified using ImageJ. Every gel included a negative control (no IP6) and a positive control (maximum cleavage, induced by 1 mM IP6 in the absence of calcium), which were used as internal references for minimal and maximal cleavage, respectively. All experiments were made in triplicate. Molar extent of cleavage (EoC) was first calculated for each lane according to the following formula, which takes into account the difference in molecular weight of the uncleaved and cleaved protein fragments: EoC = (I34/34)/(I34/34 + I40/40), where I34 and I40 correspond to the intensity of the cleaved and uncleaved bands, respectively (rCPD has a molecular weight of 40 kDa, which decreases to 34 kDa after release of the N-terminus). The EoC values obtained were then normalized to the internal references for minimal and maximal cleavage according to the following formula: EoCnorm = (EoC –

EoCmin)/( EoCmax - EoCmin). Half-maximal effective concentrations (EC50) values were calculated for each analogue in presence and absence of 10 mM CaCl2 from logistic fitting of plots of EoCnorm values versus concentration using OriginPro.

3.2.12 Cleavage assays with holotoxin 47

CHAPTER 3: PRE-EMPTIVE TOXIN B AUTO-PROTEOLYSIS

One millimolar IP6 analogues were equilibrated with 10 mM CaCl2 in 100 mM tris pH 7.4 for 15 min at 37 °C before addition of 150 ng TcdB (TgcBiomics, Germany) in a total volume of 20 μL. A negative control (no IP6, 10 mM CaCl2) and a positive control (1 mM IP6) were also included on every gel. The reaction mixtures were incubated for 3 h at 37 °C and then placed on ice. Laemmli sample buffer (5X) was added to stop the reactions and 10 mM EDTA was added to the samples containing

CaCl2 before heating at 95 °C for 3 minutes. The cleavage products were visualized by SDS-PAGE (using 15-well 8 % acrylamide PreciseTM Tris-Glycine gels, ThermoScientific, USA) followed by silver staining according to a modified Vorum protocol(312) with the thiosulphate sensitization step extended to 10 min. The linearity of the staining protocol was verified with serial dilutions of TcdB starting at 160 ng/lane down to 20 ng/lane. The band intensities were quantified as described for the cleavage assays with recombinant toxin. The experiment was done in triplicate.

3.2.13 Inhibition of TcdB toxicity in cell assays

3T3-L1 fibroblasts (gift from Prof. Michael Detmar) were maintained in DMEM with 10 % bovine serum. Cell lines are regularly assayed for mycoplasma contamination in our laboratory. For experiments, cells were seeded at a density of 2 x 103 cells per well in 90 μL on an E-plate View 16 and the cell index monitored using an xCelligence RTCA DP (ACEA Biosciences) impedance monitoring system. 100 ng TcdB was incubated alone, with 1 mM IP6 or with 1 mM IP2S4 in 10 μL 100 mM tris buffer pH 7.4 for 1 h at 37°C before dilution by 1:100 in culture medium. Ten μL of the diluted reaction mixtures were added to cells grown to sub-confluence for 24 h. The cell index of all wells was normalized to 1.0 immediately after addition of the toxin. Two wells were used for each reaction mixture per experiment, and the experiment was repeated on three different days.

3.2.14 Statistical analysis

For in vitro assays, data were assumed to be normally distributed and Tukey’s two-tailed test in conjunction with one-way analysis of variance was used to determine statistically significant differences (P < 0.05). For the MPO assay, the distributions were skewed and therefore a two-tailed Mann-Whitney U test on ranks was used for evaluating statistically significant differences (P < 0.05).

Further experimental details are given in the Appendix.

3.3 Results

3.3.1 Synthesis of IP6 analogs

We synthesized and assayed a series of IP6 analogs where phosphate groups were progressively replaced by sulfates, ranging from three phosphates and three sulfates (IP3S3, Fig. 3.2), to inositol hexasulfate (IS6, Fig. 3.2). The synthesis strategy relies on a series of intermediates exhibiting ortho- xylene-protected phosphate groups (43, 44 and 46, Fig. 3.2) (313). The inositol orthoformate 41 served

CHAPTER 3: PRE-EMPTIVE TOXIN B AUTO-PROTEOLYSIS as a starting point for the differentiation of the various inositol hydroxyl groups (314). We chose para- methoxybenzyl and tert-butyl-dimethylsilyl ethers as additional temporary protecting groups that could be removed by mild acidic hydrolysis in the presence of the relatively sensitive protected phosphate groups. The final compounds were purified using reverse-phase solid-phase extraction columns, ion- exchange and/or size-exclusion chromatography. The concentration of the compounds was determined by 1H NMR using an internal standard or by inductively coupled plasma-mass spectrometry (ICP-MS). For cellular assays and in vivo studies, the final compounds were treated with ion-exchange resin to obtain the sodium salt. In order to ensure stability in the acidic gastric environment, IP2S4 was incubated 1 in D2O acidified with deuterated phosphoric acid (pH 2-3) at room temperature. The H NMR spectra revealed no major decomposition even after 8 days (Appendix Fig. 6.2).

FIGURE 3.2 Synthesis of inositol phosphate and sulfate hybrids. PMB, para-methoxybenzyl; DMF, dimethylformamide; DEABP, 3-(Diethylamino)-1,5-dihydro-2,4,3-benzodioxa-phosphepin; DCM, dichloromethane; mCPBA, meta-chloroperbenzoic acid; TFA, trifluo-roacetic acid; TBDMS, tert-butyl- dimethylsilyl; TfOH, trifluoromethanesulfonic acid. aPurchased from Sigma-Aldrich.

3.3.2 Characterization of binding of IP6 analogs to calcium and toxin

Dynamic light scattering (DLS) measurements detected aggregates in solutions containing 100

μM IP6 and 10 mM CaCl2, confirming that IP6 precipitates in the presence of luminal concentrations of calcium. None of the hybrids formed precipitates that could similarly be detected by DLS. To corroborate this result, we incubated 100 μM of the analogs with and without calcium before filtering the solutions and measuring the filtrates’ phosphorous content by ICP-MS. All of the compounds except

CHAPTER 3: PRE-EMPTIVE TOXIN B AUTO-PROTEOLYSIS for IP6 were soluble in the presence of calcium (Fig. 3.3A). The relative affinity of the analogs for calcium was further investigated by determining the concentration of free calcium ions in an equimolar mixture of analogs and calcium, using murexide as a colorimetric calcium indicator (Fig. 3.3B). As expected, the solution containing IP6 had the lowest concentration of free calcium, which increased with analogs displaying more sulfates. This result shows that the sulfated derivatives are not only more soluble in the presence of calcium, but also have weaker interactions with calcium.

FIGURE 3.3 Determination of IP6 analog interaction with calcium and toxin. A, Solubility of analogs in presence of calcium. The concentration of phosphorous in filtered 100 µM solutions of analogs with and without 10 mM

CaCl2 was measured by ICP-MS, and adjusted to the number of phosphates per molecule to obtain analog concentrations in the filtrates. Mean ± s.d. of triplicate measurements of one sample. B, Free calcium quantification. 0.5 mM of analog, CaCl2 and murexide were mixed, incubated and centrifuged. Absorbance of the supernatant at 474 nm and 544 nm was used to determine free calcium concentration. Mean ± s.d., n = 3;

Asterisk indicates statistical difference compared with IP6 (P < 0.05). C, Intrinsic fluorescence assay for Kd determination. tCPD was titrated against increasing concentrations of analogs. The curves of fluorescence emission at 333 nm (excitation at 295 nm) versus concentration were fitted with a rectangular hyperbolic function from which Kd values were determined. Mean ± s.d., n = 3. D, Limited trypsin digestion (0.3 ng/µL trypsin) of tCPD after incubation with 1 mM analogs on Coomassie-stained SDS-PAGE gels. E, Extent of digestion determined by densitometry of the top band. Error bars show s.d.; Asterisk indicates statistical difference compared to IP6 (P < 0.05); n = 3.

The binding affinity of the analogs to the CPD was determined using an intrinsic fluorescence assay. We cloned and recombinantly expressed a truncated portion of the CPD (tCPD, Fig 3.1A), lacking 50

CHAPTER 3: PRE-EMPTIVE TOXIN B AUTO-PROTEOLYSIS the cleaved N-terminal portion from C. difficile ribotype 078. The tCPD contains a single tryptophan on the β-flap that is a mobile element of the allosteric circuit, rendering the amino acid’s fluorescence sensitive to the presence of an allosteric co-factor (Appendix Fig. 6.1) (301). In this assay, all hybrids containing phosphates had dissociation constants in the low micromolar range, indistinguishable from

IP6 (Fig. 3.3C). The Kd value obtained for IP6 is in line with reported values obtained by others (298,

315). In contrast, IS6 had a higher Kd of 30 μM, which can be rationalized by its lower negative charge. The substitution of phosphate groups by sulfates was surprisingly well tolerated with regard to the effect on Kd. Since allosteric activation involves stabilizing the active conformation of the enzyme, we also sought to determine the conformational stability of the resulting complex as a surrogate measure of folding. IP6 interacts with the TcdB CPD via cationic residues that are unlikely to form a pocket in the absence of IP6. Past studies have shown that the CPD is more flexible and thus more susceptible to proteolysis in the absence of IP6 (95, 298). Limited digestion of tCPD by trypsin in the presence of the various hybrids revealed that IP6 protected tCPD from digestion almost completely in contrast to IS6, which afforded only limited protection (Fig. 3.3D-E). The addition of phosphate groups gradually increased the protection up until IP3S3, which was indistinguishable from IP6. This result confirms that analogs with more negative charges induce more rigid folding of the protease.

3.3.3 Induction of toxin auto-proteolysis

In order to investigate the effect of the small molecules on toxin auto-processing, we recombinantly expressed a fragment of TcdB encompassing the CPD and a portion of the glucosyltransferase domain (rCPD, Fig. 3.1A). The resulting fragment undergoes cleavage when incubated with IP6, which was quantified by SDS-PAGE densitometry (Fig. 3.4A, Appendix 6.3) (95). We used this assay to determine the molecules’ half maximal effective concentrations (EC50, Fig. 3.4C). The mixed hybrids each had an

EC50 between 25 nM and 93 nM, whereas IS6 stood out from the other analogs with an EC50 of 590 nM, in a similar trend to that observed for the Kd values. The reduced activity we observed for IS6 is also in line with previous reports showing reduced holotoxin cleavage (298). The absence of substrate amino acids Ser542 and Leu543 in tCPD, which stabilize the active rCPD conformer, explains the large differences (roughly 3 orders of magnitude) between Kd and EC50 values we determined (301). When rCPD cleavage was performed in the presence of 10 mM calcium ions, large differences in activity were observed between the different hybrids (Fig. 4B, S4). In this assay, the EC50 values reflect not only the compounds’ allosteric activity, but also their solubility and the competition between their binding to the protein and to calcium. As expected, the activity of IP6 decreased by 25,000 fold (from EC50 = 25 nM to 600 μM) in 10 mM calcium (Fig. 3.4C). However, the replacement of phosphates by sulfates gradually restored the activity until the EC50 reached a value of 350 nM for IP1S5. When all phosphates were replaced by sulfates (IS6), the EC50 increased to 780 nM. This result suggests that the optimal balance between allosteric activity and interference by calcium is reached by the hybrid IP1S5, although the differences in EC50 for IP2S4, IP1S5 and IS6 with calcium are small. Given that the phosphate/sulfate

CHAPTER 3: PRE-EMPTIVE TOXIN B AUTO-PROTEOLYSIS hybrids have similar allosteric activity on the rCPD in the absence of calcium, the reduced interaction of IP1S5 with calcium due to its single phosphate group is most probably the determining factor.

FIGURE 3.4 Recombinant toxin fragment and holotoxin cleavage by IP6 analogs. A, Extent of cleavage of rCPD induced by analogs in absence of calcium and B, in the presence of 10 mM CaCl2. rCPD was incubated with varying concentrations of analogs for two hours at 37ºC before calculating extent of cleavage by densitometry of Coomassie-stained SDS-PAGE gels; n = 3. C, EC50 values obtained from plots A and B. D, TcdB cleavage in presence of 10 mM CaCl2 on silver-stained SDS-PAGE gels. 1 mM analogs were equilibrated with 10 mM CaCl2 for 15 min at 37°C before addition of 150 ng TcdB and a three hour incubation at 37°C. E, Extent of holotoxin cleavage in presence of calcium determined by densitometry and normalized to the positive and negative controls. Error bars show s.d.; Asterisk indicates statistical difference compared to IP2S4 (P < 0.05); n = 3. F,

TcdB cleavage in absence of CaCl2 on silver-stained SDS-PAGE gels. 1 mM analogs were added to 150 ng TcdB and incubated for three hours at 37°C. G, Extent of holotoxin cleavage in absence of calcium determined by densitometry and normalized to positive and negative controls. Error bars show s.d.; Asterisk indicates statistical difference compared to IP6 (P < 0.05); n = 3.

We then performed a cleavage assay with TcdB isolated from C. difficile strain VPI 10463 in the presence and absence of calcium. In juxtaposition to IP1S5 having the strongest activity with rCPD, IP2S4 led to the largest extent of holotoxin cleavage in the presence of calcium (Fig. 3.4D-E). It is likely that a higher charge density is necessary to mediate the large conformational change required for protease activation in the holotoxin relative to rCPD due to the almost ten-fold larger molecular weight of the former. In support of this hypothesis, hybrid compounds containing fewer phosphates led to slightly lower holotoxin cleavage in absence of calcium (Fig. 3.4F-G), a trend we observed to a much smaller extent with rCPD (Fig. 3.4A). A possible reason for the surprisingly low holotoxin cleavage observed with IP1S5 in the presence of calcium is that the kinetics of cleavage are slow compared to the rate of loss in activity of the protease when it is incubated in buffer at 37°C. Indeed this cleavage assay is limited due to a gradual inactivation of the TcdB protease over time, resulting in a 37 % decrease in cleavage when incubated at 37°C for one hour prior to the addition of IP6 (Appendix Fig. 6.5). The 52

CHAPTER 3: PRE-EMPTIVE TOXIN B AUTO-PROTEOLYSIS analogs with fewer phosphates are likely to induce cleavage with slower kinetics due to their weaker stabilization of the active toxin form regardless of calcium concentration (Fig. 3.3D-E). Binding competition with calcium, albeit relatively low for IP1S5 (Fig. 3.3B), further reduces the cleavage rate. In this assay, it is unclear if the incomplete holotoxin cleavage we obtained with IP6 without calcium (even after extended incubations and addition of DTT) is a result of cysteine protease inactivation during the TcdB purification process, and how this reflects the cysteine protease activity of TcdB produced by C. difficile in vivo. Nevertheless, the clear superiority of IP2S4 over other molecules in inducing holotoxin cleavage in high calcium conditions led us to select it for further study.

3.3.4 In vitro inhibition of TcdB cytopathy

In order to confirm the effect of pre-emptive TcdB auto-processing on cytopathic effect, we incubated TcdB alone, with IP6 or IP2S4 in presence or absence of 10 mM CaCl2, and added the mixtures to sub-confluent 3T3-L1 fibroblasts grown on a real-time impedance monitoring system (Fig. 3.5A). The impedance (cell index) recorded reflects the cells’ adherence to the culture plate well, and therefore effectively corresponds to the extent of cell rounding. Significant differences in cell index were recorded between the various treatments after two hours of incubation (Fig. 3.5B, Appendix 6.6), with a high cell index (low cell rounding) observed when TcdB was pre-treated with IP6 versus a null one (high cell rounding) with untreated TcdB. When 10 mM CaCl2 was present in the reaction mixture, the cell index was null with IP6, whereas IP2S4 retained partial inhibitory activity. Microscopy images of the cells clearly show cell rounding and corroborate the quantitative cell index results (Fig. 3.5A). The cell indexes recorded correlated very closely with the extent of holotoxin cleavage induced by IP6 and IP2S4 we observed by SDS-PAGE (Fig. 3.4G), leading to the conclusion that TcdB cleavage prior to cell intake does indeed reduce TcdB cytopathy significantly.

Highly charged molecules such as IP2S4 are not expected to be able to cross cell membranes and induce toxicity. Incubating cells with concentrations of IP2S4 up to 1 mM in culture medium did not cause any noticeable effect on cell proliferation within 24 h, as measured by evolution of cell index (Appendix Fig. 6.7). At the same concentration, IP6 prevented cell proliferation, most likely due to the visible precipitation caused by chelation of calcium in the cell culture medium.

CHAPTER 3: PRE-EMPTIVE TOXIN B AUTO-PROTEOLYSIS

FIGURE 3.5 In vitro cytopathy inhibition by IP2S4 and IP6. A, 100 ng TcdB was pre-incubated alone, with 1 mM IP6 or 1 mM IP2S4, and with or without 10 mM CaCl2 for one hour at 37°C before addition to sub-confluent

3T3-L1 fibroblasts at a final concentration of 10 ng/mL TcdB (1 µΜ IP6/IP2S4, 100 µΜ CaCl2 added to cells). Images were taken after incubating for two hours at 37°C. The black circles are gold electrodes that allow impedance measurement. B, Cell index of cells incubated under conditions described in a, measured by real- time impedance-monitoring system. Error bars show s.d.; Asterisk indicates statistical difference compared to TcdB (P < 0.05); Duplicate measurements were repeated in triplicate.

3.3.5 In vivo attenuation of colitis

To establish a disease model resembling human CDI, germ-free mice were monocolonized with Escherichia coli JM83 to prime the immune system and then inoculated with fecal matter from a patient suffering from recurrent infection by a TcdA-TcdB+ C. difficile strain (NAP1) three weeks later. Within five days, the mice developed moderate to severe colitis with elevated activity of the inflammation marker MPO and histological features that are the hallmark of CDI. The MPO activity measured after inoculating mice with fecal matter from a healthy donor in this model is low (less than 1 unit/mg of tissue). Animals were randomly assigned to receive IP6, IP2S4 (from two synthetic batches) or myo- inositol as a negative control, administered twice daily at 12.3 μmol/kg (corresponding to 10 mg/kg IP2S4) for 4.5 days starting 6 h post-inoculation. Colonic MPO activity measurements were obtained from two series of blinded experiments with 3 – 5 mice per group in each series. Oral administration of IP2S4 but not IP6 significantly reduced the acute inflammatory component of colitis compared to administration of myo-inositol (Fig. 3.6A). Histologically, mice treated with myo-inositol had abnormal architecture of the mucosa, with focal epithelial necrosis, ulceration and overlying fibrin-rich exudate containing neutrophils (Fig. 3.6B). IP2S4, and to a lesser degree IP6-treated mice, had decreased mucosal damage and inflammatory infiltrate. As evident from the MPO activity of the myo-inositol control group, a moderate degree of variability in disease outcome following inoculation exists, as also reported by others (316). The results of this animal study provide the first evidence that IP2S4 modulates pathology in vivo and provide impetus for further validation through dose-optimized studies in other established CDI models.

CHAPTER 3: PRE-EMPTIVE TOXIN B AUTO-PROTEOLYSIS

FIGURE 3.6 In vivo IP2S4 activity. A, Colonic MPO activity. Germ-free NIH Swiss mice were inoculated with 200 µL of a 1:10 dilution in saline of fecal matter from a patient with recurrent CDI. IP6, IP2S4 or inositol (negative control) were administered as doses of 12.3 µmol/kg (corresponding to 10 mg/kg IP2S4) in a blinded fashion given in 5 mL drinking water starting 6 h after inoculation and for three nights thereafter, as well as per intragastric gavage once a day for four days starting the day after infection. The animals were sacrificed at the end of day five post-infection and the colon excised for histological evaluation and myeloperoxidase activity measurement. Boxes represent interquartile ranges, dividing lines in boxes median values, and whiskers maximal and minimal values (n(inositol) = 8, n(IP6) = 7, n(IP2S4) = 9). Asterisk indicates statistical difference compared to inositol according to two-tailed Mann-Whitney test (P < 0.05). B, Histological sections of excised colons. Inositol-treated mice (negative control) displayed overt colonic structural changes characterized by mucosal ulceration and overlying exudate, marked acute and chronic inflammatory infiltrate and submucosal edema.

3.4 Discussion

Previously published efforts to develop non-antibiotic therapies against C. difficile infection include antibodies (317), vaccines (64), fecal microbiota transplantation (42), polymeric binders (161), inhibitors of spore germination (318), and non-toxigenic C. difficile (319). The most clinically advanced of these are all aimed at preventing infection recurrence. A strong unmet need for effective and safe first-line, combination or prophylactic agents therefore persists.

The present study describes the synthesis, characterization, and in vitro and in vivo testing of allosteric triggers capable of inducing TcdB cleavage in high calcium conditions mimicking those in the colon lumen. In order to achieve this, we modified IP6 to tune its affinity to calcium while maintaining the uniquely high charge density that mediates its interaction with TcdB, as well as TcdA and numerous other proteins such as Vibrio cholerae MARTX(320) and haemoglobin (321). Indeed, lower inositol phosphates (e.g. IP3, IP2) at 100 μM are known to be unable to induce TcdB cleavage presumably because they cannot provide the ionic cohesion necessary for allosteric activation (96). We synthesized a series of IP6 analogs where phosphate groups are progressively replaced by sulfate groups. We favored the positioning of sulfates in positions 2 and 3 on the inositol ring, which are less buried in the binding pocket, and therefore less likely to be crucial for ionic cohesion (Appendix Fig. 6.1) (95, 305). Vicinal phosphates were also avoided to disfavor chelation with calcium. In order to avoid difficult manipulations of charged molecules, we devised a synthesis where the sulfation and final hydrogenolysis of the phosphate protecting group are the last steps (Fig. 3.2).

CHAPTER 3: PRE-EMPTIVE TOXIN B AUTO-PROTEOLYSIS

The partial replacement of phosphates by sulfates dramatically improves solubility in the presence of calcium and reduces interactions with calcium (Fig 3.3). Moreover, the inositol phosphate/sulfate hybrids are capable of binding the TcdB CPD, and inducing its cleavage at micromolar concentrations in the presence of calcium (Fig. 3.4). In this series of analogs, IP2S4 strikes the optimal balance between allosteric activity on TcdB and competition by calcium. It is safe to assume that all phosphate groups do not play an equal role in the allosteric activation depending on their position in the CPD’s binding pocket. Therefore, a more in-depth structure-activity-relationship study is warranted to probe the impact of the substitution position on the activity of the analog.

Furthermore, we show that TcdB auto-proteolysis by IP2S4 reduces TcdB cytopathy in cultured cells, as expected given that intracellular delivery of the glucosyltransferase domain in the cytosol is known to be necessary for full toxin cytopathy (297, 322–326). TcdB produced by hypervirulent strains of C. difficile are reported to auto-cleave more efficiently in vitro than historical strains, suggesting that glucosyltransferase domain uptake after exposure to IP2S4 would be even more limited with these strains (324). The extent to which auto-proteolytic release of the glucosyltransferase domain after holotoxin uptake plays a role in the context of an infection has recently come under scrutiny but is of lesser relevance to the situation where the glucosyltransferase is not taken up in the first place due to its separation from the remainder of the toxin in the lumen (327, 328). It is well established that TcdB is a mediator of strong immune responses and that reduction in inflammation markers such as MPO activity is associated with attenuated pathology (299, 329). To investigate the effect of toxin-induced inflammation, we used a mouse model of CDI where monocolonized mice were challenged with stool samples from a patient suffering from recurrent CDI. We show that IP2S4 attenuates colitis by quantitative measure of MPO activity after oral dosing. These observations support the hypothesis that the TcdB glucosyltransferase domain plays an important role in pathogenesis and that preventing its uptake influences CDI symptoms.

Our study provides an example that combining consideration of molecular pathogenesis with bottom-up modulation of natural cofactor chemistry can yield new therapeutic approaches. Modulating the physicochemical properties of IP6 circumvents competition by calcium thus allowing allosteric activity against TcdB in the gut lumen. Our findings provide a starting ground for investigating the therapeutic potential of IP6 analogs through structure-activity relationship studies and generalization of the results presented herein to other animal models and strains of C. difficile.

3.5 Remarks

We are grateful to Stefan Feurer, and Anna Pratsinis for technical assistance and thank the Mass Spectrometry Service of the Institute for Organic Chemistry at ETH Zurich for mass determinations. Mattias Ivarsson designed and carried out biological and biochemical experiments. Estelle Durantie designed and carried out the chemical synthesis. Corina Hüberli performed biochemical experiments. 56

CHAPTER 3: PRE-EMPTIVE TOXIN B AUTO-PROTEOLYSIS

Samuel Heuwiler performed chemical synthesis. Jun Lu, Elena Verdu and Premysl Bercik (McMaster University, Hamilton, Ontario, Canada) designed and performed the in vivo experiments. Mattias Ivarsson, Jean-Christophe Leroux and Bastien Castagner analyzed the data and designed experiments. Bastien Castagner initiated and directed the project.

4 Development of a Polymeric Binder Targeting the Receptor-Binding Domain of Clostridium difficile Toxin A

4.1 Introduction

Preventing damage to the colon caused by TcdA and TcdB in the context of CDI has the potential to markedly reduce disease symptoms and to prevent infection recurrence by avoiding the use of antibiotics. The receptor binding domains (RBD) of C. difficile toxins mediate the first contact between the toxins and the cells of the colonic epithelium and are therefore attractive targets for the design of inhibitors that prevent the toxins’ uptake and cytotoxicity (330). Indeed, two toxin sorbents, tolevamer (Genzyme) and Synsorb 90 (Synsorb Biotech) have been tested and failed in phase 3 clinical trials, as described in chapter 2. As a reminder, tolevamer is a ca. 600 kDa polystyrene sulfonate and Synsorb 90 consists of a trisaccharide ligand for TcdA (αGal(1-3)βGal(1-4)Glc) immobilized on Chromosorb- P, an insoluble solid support (295).

The RBD of TcdA and TcdB are situated at the proteins’ C-termini (165). Both consist of either 8 (for TcdA) or 5 (for TcdB) series of 3 – 5 short repeating sequences (19 – 24 amino acids) separated by one long repeating sequence (30 – 31 amino acids) as depicted in Figure 4.1 (330). Each repeat is separated by a hairpin and rotated by 120° relative to its neighbor. The carbohydrate binding pockets are formed by long repeats and the hairpin turn of the consecutive short repeat. The RBDs adopt a β- solenoid tertiary structure, as is common for proteins with repetitive sequences, which increases the surface area available for protein-protein and protein-carbohydrate interactions (15). Despite the structural similarity of RBD-A and RBD-B, their sequences differ substantially, which is also the reason why efforts to identify monoclonal antibodies that bind to both domains with high affinity and that neutralize both toxins have not been successful to date (61).

In this study, we sought to design a novel polymeric binder for TcdA, with the ultimate aim of applying a similar strategy for TcdB. We hypothesized that tolevamer failed in phase 3 clinical trials because along with insufficient specificity, and poor affinity for TcdB, its net negative charge may have prevented it from penetrating into the similarly charged mucus layer to reach TcdA and TcdB. At around 600 kDa, tolevamer’s size is also in a range (ca. 20 – 200 nm) where the mesh-spacing of mucus inhibits rapid diffusion (331). Similarly, we reasoned that Synsorb 90 was ultimately inefficacious because the large size of the insoluble particles (hundreds of micrometers) also prevented their diffusion through the mucus layer, and that the precise presentation of the ligand was not optimized. We therefore decided to display the same trisaccharide ligand as used in Synsorb 90 on a smaller polymer scaffold (20 – 55 kDa)

CHAPTER 4: POLYMERIC BINDER FOR TCDA with a neutral charge: poly(hydroxypropyl methacrylamide) or poly(HPMA). Poly(HPMA) also has high aqueous solubility, is inert and highly flexible, making it a widely used polymer in biomedical applications (332). We assumed that its physico-chemical properties would prevent unfavorable interactions with mucus that can arise due to charge and large size, and also offer sufficient flexibility to simultaneously engage all 7 carbohydrate binding sites on TcdA. The latter would result in a strong multivalent effect and thus prevent the toxin from binding to its endogenous receptor. As evident from the polymeric binders presented in chapter 2, a large body of knowlegde exists on the optimal presentation of carbohydrates for binding certain bacterial toxins and lectins. No such rigorous investigations have been reported for TcdA and TcdB to date, however, and we hoped that our findings would contribute towards a better understanding of the ideal display of ligands for these toxins.

FIGURE 4.1 Structural organization of RBD-A and crystal structure of αGal(1-3)βGal(1-4)Glc bound to 127 C- terminal residues of RBD-A. PDB: 2G7C (333).

4.2 Materials and Methods

4.2.1 Monomer and chain transfer agent synthesis

Synthesis of hydroxypropyl methacrylamide (HPMA, 47) was performed as reported by Ulbrich et al. (334). 47 was obtained as a white powder (2.4 g, 52 mmol, 22 %). Synthesis of N-methacryloyl hydroxysuccinimide (NMS, 48) was performed as reported by Shunmugam et al. (335). 48 was obtained as a white powder (1.3 g, 7.0 mmol, 11 %). Synthesis of pentafluorophenyl methacrylate (PFMA, 49) was performed as reported by Eberhardt et al. (336). 49 was obtained as a colorless liquid (3.5 g, 14 mmol, 43 %). Synthesis of azido-triethyleneglycol methacrylamide (ATEGMA, 50) was performed as 59

CHAPTER 4: POLYMERIC BINDER FOR TCDA reported by Dubacheva et al. (337). 50 was obtained as a colorless oil (0.62 g, 2.2 mmol, 48 %). The chain transfer agent (CTA) 4-(cyanopentanoic acid) dithiobenzoate was prepared as described by Vosloo et al. (338). CTA was obtained as a red solid (1.2 g, 4.4 mmol, 68 %). All 1H NMR spectra were in accordance with literature.

4.2.2 Spectroscopic and chromatographic analysis

1H NMR spectra were obtained on a Bruker AV 400 MHz spectrometer. Gel permeation chromatography (GPC) analysis to determine molecular weights of polymers was carried out in water with 0.03 % sodium azide using a Viscotek TDAmax system with differential refractive index and light scattering detectors at room temperature with a flow rate of 0.7 mL/min using three Viscotek CLM3016 columns in series. Absolute molecular weights were determined using narrow polydispersity poly(ethylene glycol) (PEG, 19 kDa) and broad polydispersity dextran (65 kDa) standards. Fourier transform infrared (FTIR) spectroscopy was carried out on a Perkin-Elmer Spectrum 65 infrared spectrophotometer using attenuated total reflectance.

4.2.3 Synthesis of poly(HPMA)

HPMA 47 (1990 mg, 13.9 mmol) and tert-butanol (1.4 mL) were added to a Schlenk tube. CTA (39 mg, 0.14 mmol), azobisisobutyronitrile (AIBN, Sigma Aldrich, USA – 11 mg, 0.069 mmol) and 1,3,5-trioxane (Sigma Aldrich – 13 mg, 0.14 mmol) (50 μL) were added from stock solutions of dimethylformamide (DMF). The contents of the reaction vessel were degassed by bubbling argon gas through the reaction mixture for 30 min. The reaction mixture was placed at 80°C for 9 h and then placed at 4°C to stop the reaction. The product was subsequently obtained by precipitation with diethyl ether and acetone (1:1, v:v) three times using centrifugation to recover the product. The supernatant was decanted and the residue dried on a rotavap and then under vacuum to yield a pink powder (210 mg, 70 1 %). H NMR (400 MHz, CD3SOCD3) δ (ppm): 0.74 (2H, CH2, backbone), 0.96 (3H, CH3), 1.5 (3H,

CH3, backbone), 2.85 (2H, CH2), 3.61 (1H, CH), 4.66 (1H, OH), 7.09 (1H, NH).

4.2.4 Synthesis of poly(HPMA-ran-NMS) 11

The procedure for synthesis of 51 was based on the procedure reported by Yanjarappa et al. (339). HPMA 47 (250 mg, 1.7 mmol) and tert-butanol (1.75 mL) were added to a Schlenk tube. CPDB CTA (1.5 mg, 5.5 µmol), AIBN (0.5 mg, 2.7 µmol) and 1,3,5-trioxane (13 mg, 0.14 mmol) (50 μL) were added from stock solutions of DMF. NMS 48 (80 mg, 0.4 mmol) was dissolved in 0.9 mL DMF in a 5 mL reaction vessel. The contents of both reaction vessels were degassed with argon gas for 15 minutes. The Schlenk tube valve was closed and the reaction was heated to 75°C for 8 h. The NMS was put in an air tight syringe and the contents emptied into the Schlenk tube at a flow rate of 0.15 mL/h over 6 hours using a syringe pump. The reaction was continued for 2 h after all the NMS was added to the Schlenk tube and then stopped by placing the reaction vessel at 4°C. The product was obtained by precipitation from diethyl ether and acetone (1:1, v:v) three times using centrifugation to recover the

CHAPTER 4: POLYMERIC BINDER FOR TCDA product. The product was dried on a rotavap and under vacuum to yield a pink powder (90 mg, 27 %). The percentage contents of HPMA 47 and NMS 48 were calculated from NMR spectra. 1H NMR (400

MHz, CD3SOCD3) δ (ppm): 0.75 (2H, CH2, backbone), 0.97 (3H, CH3), 1.62 (3H, CH3, backbone), 2.74

(4H, CH2, succinimide), 2.86 (2H, CH2), 3.62 (1H, CH), 4.66 (1H, OH), 7.13 (1H, NH).

4.2.5 Synthesis of poly(PFMA) 52

The procedure for synthesis of 52 was based on the procedure reported by Gibson et al. (340). PFMA 49 (0.8 g, 3 mmol) was added to a 5 mL pear flask followed by 1.5 mL dioxane. CPDB CTA (4.4 mg, 0.016 mmol) in 40 µL dioxane and 4,4'-azobis(4-cyanovaleric acid) (ACVA) (2.2 mg, 7.9 µmol) in 40 µL dioxane were added to the vessel and the solution was degassed by bubbling the reaction mixture with argon gas for 30 min. The flask was then heated to 90°C for 90 min before being cooled in an ice-water bath to stop the reaction. The product was obtained by precipitation in pentane three times using centrifugation to recover the product. The dried product was a pink powder (700 mg, 27 %). 1 H NMR (400 MHz, CDCl3) δ (ppm): 0.84 – 1.54 (3H, CH3), 2.41 (2H, CH2).

4.2.6 Synthesis of poly(ATEGMA-ran-HPMA) 53

HPMA 47 (82 mg, 0.57 mmol) and ATEGMA 50 (41 mg, 0.14 mmol), CPDB (0.5 mg, 1.8 µmol), AIBN (0.5 mg, 3.1 µmol) and anisole (40 µL, 0.37 mmol) were added to a 5 mL pear flask. 0.7 µL DMF was added to the flask. The DMF was degassed by bubbling argon through the reaction mixture while the reagents were weighed out. The contents of the reaction vessel were degassed by bubbling argon through the reaction mixture for 1 h. The reaction was heated to 65°C for 24 h and then stopped by placing the reaction vessel at 4°C. The product was purified by precipitation in ethyl acetate four times using centrifugation to recover the product. The supernatant was decanted and the residue dried on a rotavap and then under vacuum to yield a white powder. The amounts of HPMA 47 and ATEGMA 50 used provided 20 mol% ATEGMA 50. For 10 mol% ATEGMA, the amounts used were: HPMA 47 (92 mg, 0.644 mmol) and ATEGMA 50 (20.5 mg, 0.072 mmol). For 5 mol% ATEGMA, the amounts used were: HPMA 47 (97 mg, 0.680 mmol) and ATEGMA 50 (10.25 mg, 0.036 mmol). The percentage contents of each monomer were calculated from NMR spectra of the products. 1H NMR (400 MHz,

D2O) δ (ppm): 0.99 (2H, CH2, backbone), 1.21 (3H, CH3), 1.78 (3H, CH3, backbone), 3.17 (2H, d, CH2),

3.36 (2H, CH2, ATEGMA), 3.55 – 3.76 (14H, OCH2CH2), 3.93 (1H, CH).

4.2.7 Aminolysis of polymers containing activated esters

Poly(PFMA) 52 (10 mg, 0.39 µmol) was dissolved in 80 µL DMF. 1-aminopropan-2-ol (6 µL, 0.08 mmol) and triethylamine (11 µL, 0.079 mmol) were added to an HPLC vial and the vial flushed with argon gas before stirring at room temperature (RT). After 14 h, 2 mL water was added to the reaction mixture before transferring it to a dialysis bag (Spectra/Por 3 membrane from Spectrum Labs, MWCO: 3,500 kDa). Dialysis was carried out for 2 days against 2 L of Ultrapure water with 3 changes of water and the samples lyophilized to obtain the final products.

CHAPTER 4: POLYMERIC BINDER FOR TCDA

4.2.8 Lactose-amine synthesis 56-NH2

56-NH2 was synthesized as previously described (341).

4.2.9 Lactose-alkyne synthesis 56-alkyne

2,5-dioxopyrrolidin-1-yl pent-4-ynoate was synthesized as described elsewhere (342). 2,5- dioxopyrrolidin-1-yl pent-4-ynoate (44 mg, 0.22 mmol) was dissolved in 1.5 mL acetonitrile and added to 56-amine (66 mg, 0.15 mmol) dissolved in 0.1 mL NaHCO3 (0.3 M) and 0.3 mL MeOH. The solution was agitated overnight. The acetonitrile was evaporated with a stream of nitrogen. The mixture was diluted with H2O, poured on a Sep-Pak cartridge and eluted with a gradient of 0-100 % MeOH-H2O. 1 Yielded 24 mg of 56-alkyne (0.046 mmol, 31% yield). H-NMR (400 MHz; D2O): δ 4.40 (d, J = 8.0 Hz, 1H), 4.38 (d, J = 7.8 Hz, 1H), 3.90 (dd, J = 12.2, 2.1 Hz, 1H), 3.87-3.81 (m, 2H), 3.75-3.50 (m, 9H), 3.47 (dd, J = 10.0, 7.8 Hz, 1H), 3.25-3.21 (m, 1H), 3.13 (t, J = 6.8 Hz, 2H), 2.45-2.40 (m, 2H), 2.37- 2.33 (m, 2H), 2.31 (t, J = 2.6 Hz, 1H), 1.59-1.52 (m, 2H), 1.48-1.42 (m, 2H), 1.30-1.26 (m, 4H). 13C

NMR (101 MHz; D2O): δ 174.5, 102.9, 102.0, 83.3, 78.4, 75.4, 74.8, 74.5, 72.9, 72.5, 71.0, 70.6, 70.2, 68.6, 61.0, 60.1, 39.3, 34.6, 28.6, 28.2, 25.7, 24.7, 14.7.

4.2.10 Trisaccharide-amine synthesis 57-amine

The synthesis was based on that reported by Ratcliffe et al. (343). Yielded 10 mg of 57-amine 1 (0.014 mmol, 48% yield). H-NMR (400 MHz; D2O): δ 5.10 (d, J = 3.8 Hz, 1H), 4.47 (d, J = 7.7 Hz, 1H), 4.44 (d, J = 8.1 Hz, 1H), 4.17-4.12 (m, 2H), 3.97-3.54 (m, 17H), 3.28-3.25 (m, 1H), 2.95 (t, J = 7.5 13 Hz, 2H), 1.66-1.57 (m, 4H), 1.40-1.33 (m, 4H). C-NMR (101 MHz; D2O): δ 102.9, 102.1, 95.5, 78.8, 77.3, 75.1, 74.79, 74.59, 72.9, 70.9, 70.5, 69.6, 69.34, 69.18, 68.2, 64.9, 61.04, 60.98, 60.2, 39.5, 28.5, 26.6, 25.3, 24.6.

4.2.11 Trisaccharide-amine synthesis 57-alkyne

2,5-dioxopyrrolidin-1-yl pent-4-ynoate (5 mg, 0.026 mmol) was dissolved in 0.05 mL acetonitrile and added to 57-amine (10 mg, 0.014 mmol) dissolved in 0.25 mL NaHCO3 (0.3 M). The solution was agitated overnight. The acetonitrile was evaporated with a stream of nitrogen. The mixture was diluted with H2O, poured on a Sep-Pak cartridge and eluted with a gradient of 0-40 % MeOH-H2O. 1 Yielded 7.6 mg of 57-alkyne (0.0111 mmol, 80% yield). H-NMR (400 MHz; D2O): δ 5.07 (d, J = 3.8 Hz, 1H), 4.45 (d, J = 7.8 Hz, 1H), 4.41 (d, J = 8.0 Hz, 1H), 4.14-4.10 (m, 2H), 3.95-3.49 (m, 17H), 3.25- 3.21 (m, 1H), 3.13 (t, J = 6.8 Hz, 2H), 2.44-2.40 (m, 2H), 2.37-2.33 (m, 2H), 2.31 (t, J = 2.5 Hz, 0.2H partially exchanged to deuterium), 1.59-1.52 (m, 2H), 1.49-1.41 (m, 2H), 1.34-1.26 (m, 4H). 13C-NMR

(101 MHz; D2O): δ 174.5, 102.8, 102.0, 95.4, 78.6, 77.2, 75.1, 74.73, 74.53, 72.8, 70.8, 70.6, 69.6, 69.29, 69.13, 68.2, 64.8, 60.99, 60.92, 60.2, 39.3, 34.6, 28.6, 28.2, 25.7, 24.7, 14.7.

4.2.12 Functionalization of poly(ATEGMA-ran-HPMA) 53

CHAPTER 4: POLYMERIC BINDER FOR TCDA

Poly(ATEGMA-ran-HPMA) was functionalized with lactose-(linker)-alkyne 56 or αGal(1-

3)βGal(1-4)Glc-(linker)-alkyne 57 by copper-catalyzed azide-alkyne cycloaddition. Five milligrams of polymer to be functionalized and a 1.1 molar equivalent of 56 or 57 relative to polymer azide content were added to a 1.5 mL reaction tube with 560 µL Ultrapure water, 70 µL fresh 50 mM sodium ascorbate solution and 70 µL 10 mM CuSO4-TBTA solution (added last). The reaction was allowed to proceed for 16 h at RT with agitation. The reactions were dialysed for 24 h with one change of water and then purified on a PD MidiTrap G-25 column. The column was equilibrated with 3 volumes of Ultrapure water, the reaction mixture added and allowed to enter the packed bed. The sample was eluted into fractions each containing 6 drops. Fractions 1-10 were freeze-dried and a solid was present in fractions 3-10. Fractions 3-7 were pooled. Product masses obatined after lyophilization varied from 1.9 mg to 4.9 mg (28 % – 52 %).

4.2.13 Functionalization of PAMAM-COONa

PAMAM COONa G5.5 (3 mg, 0.06 µmol) (AndrewsChemServices, USA) was dissolved in 30 µL methanol and mixed with αGal(1-3)βGal(1-4)Glc-(linker)-amine 57 (10 mg, 0.015 mmol) previously dissolved in 170 uL H2O (1 eq. amine to carboxylate). The pH was adjusted to 7.0 and the mixture incubated for 10 min with shaking. 4-(4,6-dimethoxy-1,3,5-triazin-2-yl)-4- methylmorpholinium chloride (DMTMM, Sigma Aldrich) (20 mg, 0.073 mmol) was added and the pH adjusted to 7.0. The reaction mixture was incubated at RT with vigorous shaking for 4 h and the pH was measured and adjusted every 30 min. The product was purified using VivaSpin2 columns by washing

3x with 2 mL H2O and 1 x with 500 uL D2O. After lyophilization: 3.5 mg (27 %).

4.2.14 Recombinant expression RBD-A

To generate recombinant RBD-A (6His-TcdA1875–2710), the nucleotide sequence coding for amino acids 1739 – 2170 of TcdA from strain VPI10463 C. difficile was amplified from genomic DNA (TgcBiomics, Germany) by PCR with the forward primer GGCTAGCTGGTCTACAGAAGGAAGTGAC (NheI restriction site underlined) and the reverse primer CGAAGCTTCTAGCCATATATTCCAGGGGCTTTTAC (HindIII restriction site underlined, STOP codon in italic). The DNA was subsequently inserted into pET28a plasmid (Novagen, Germany) and transformed in NovaBlue E. coli (Novagen) using standard techniques. To generate a truncated RBD-A still containing all 7 binding sites but that led to more efficient expression, pET28a-RBD-A was digested with SpeI/HindIII, the digestion products separated on an agarose gel and the truncated insert excised and purified using a NucleoSpin Extract II kit (Machery-Nagel, Germany). The insert was subsequently ligated into SpeI/HindIII-digested pET28a and transformed into BL21(DE3). All primers were purchased from Microsynth (Switzerland).

RBD-A was expressed and purified according to the following protocol. Overnight cultures of transformed BL21(DE3) were diluted 1:33 in 1 L LB Broth (Sigma Aldrich) supplemented with 30 μg/L

CHAPTER 4: POLYMERIC BINDER FOR TCDA kanamycin and grown at 25 °C until an OD600 of 0.6 – 0.8 was reached. IPTG was added to a final concentration of 0.1 mM and the cultures grown for 16 h at 16 °C. The cultures were pelleted by centrifugation at 4’000 g for 15 min at 4 °C and resuspended in sonication buffer (20 mM NaH2PO4,

100 mM NaCl, 1 mM MgCl2) with 10 U/mL benzonase (Sigma Aldrich) and protease inhibitor cocktail for His-tagged protein purification at a dilution of 1:200 (Sigma Aldrich). The lysates were sonicated with a probe sonicator before centrifugation at 4’000 g for 20 min. Recombinant proteins were purified from the cleared lysate by immobilized metal-ion affinity chromatography using 50 % Ni-NTA resin (Qiagen, Germany) according to the manufacturer’s instructions for column purification. Eluted fractions containing protein were buffer exchanged into 50 mM MOPS, 100 mM NaCl, pH 7.0 using Zeba Spin columns (ThermoScientific, Switzerland) before storage at -80 °C.

4.2.15 Cell assays

HT-29 (ATCC-HTB-38) were maintained in McCoy’s 5A modified medium (Sigma Aldrich) supplemented with GlutaMax (Invitrogen, USA), penicillin, streptomycin and 10 % fetal bovine serum. For experiments, cells were seeded at a density of 4 x 104 cells per well in 100 μL on an E-plate View 16 and the cell index monitored using an xCelligence RTCA DP (ACEA Biosciences, USA) impedance monitoring system. After incubation for 24 h at 37°C, the medium was exchanged with TcdA (2 ng/mL) alone, or with 1 mg/mL test polymer (the mixtures were added to cells immediately), in 90 μL complete medium plus 10 μL PBS. The cell index of all wells was normalized to 1.0 immediately after addition of the toxin-polymer mixtures. Two wells were used for each reaction mixture per experiment.

4.2.16 Hemagglutination assay

The hemagglutination assays were performed as reported by Dingle et al. (165). 12.5 µL of 1 μM RBD-A in PBS (lowest concentration of two-fold dilutions that led to full hemagglutination in preliminary experiments) was added to 12.5 μL of serial two-fold dilutions of test polymer (starting at 10 mg/mL, across one row on a polypropylene 96-well plate with U-shaped wells) and to 25 μL rabbit erythrocyte suspension (Fitzgerald, USA) and mixed by pipetting up and down. The plate was incubated at 4°C overnight and the extent of hemagglutination in each well was then scored visually according to the following scale: none (0), moderate (1) and complete (2).

4.2.17 Thermal denaturation assay

Thermal denaturation assays were carried out as described by Koch et al. on a StepOnePlus real-time PCR system (Applied Biosystems) (344). The final concentration of the components were: 0.5 μM protein and 20 mg/mL test polymer (corresponding to 10 mM ligand) in 20 mM tris pH 7.4.

4.2.18 Phage display selection 64

CHAPTER 4: POLYMERIC BINDER FOR TCDA

Phage display selections were carried out using the Ph.D. 7 library (New England Biolabs USA). The procedure that follows is for solution capture of biotin-labelled RBD-A. Biotin labelling was carried out using a 20-fold molar excess of biotin-X-NHS (Calbiochem, USA) over RBD-A according to the manufacturer’s instructions. For round 1, 10 μg biotin-RBD-A were added to 1.2x1011 phages in 200 μL phage binding buffer (50 mM tris, 0.5 M NaCl, 2 mM CaCl2, 2 % BSA, 0.1 % polysorbate 20 pH 7.5) and incubated for 2 h at 4°C a low-binding 1.5 mL reaction tube (Maxymum Recovery, Axygen, USA). In parallel, 200 μL Neutravidin agarose (Piercenet, USA) were blocked with 2 % BSA, 0.1 % polysorbate 20 in tris-buffered saline (TBS) pH 7.5 for 2 h at RT. The blocked Neutravidin agarose was then used to capture biotin-RBD-A for 15 min at RT. The resin was washed 4 times with 1 mL high-salt TBS with tween 20 (HS-TBS-T, phage binding buffer without BSA). Bound phages were eluted by incubating the washed resin with 200 μL 0.2 M glycine pH 2.2 with 1 mg/mL BSA for 7 min. The supernatant was then neutralized with 30 μL 1 M tris pH 9.1.

Round 2 was carried out similarly to round 1 using 1x1011 phages from round 2 and 5 μg biotin- RBD-A. For round 3, 2 μg biotin-RBD-A were used and the capturing step was done using a pre-blocked high-binding streptavidin plate (Piercenet) and the bound phages were washed 10 times with HS-TBS- T. Round 4 was identical to round 3. Phage amplification, titering after elution and DNA sequencing (Synergene, Switzerland) was carried out according to the library manufacturer’s instructions. Concentrations of amplified phage were determined spectroscopically (345).

4.2.19 Direct phage ELISA

Phage ELISAs were carried out essentially as described by Tonikian et al. (346). Briefly, 47 plaques from rounds 3 and 4 of titering (solution capture and surface immobilized RBD-A) were amplified overnight (one control well did not contain phage) in a deepwell plate (Eppendorf, Germany) and the phage precipitated using a solution of 20 % (w/v) PEG8000 and 2.5 M NaCl, then resuspended in 100 μL TBS. The rows on a 96-well Nunc MaxiSorp plate (eBioscience, USA) were alternatively coated with 20 μg/mL RBD-A or buffer overnight at 4°C. After blocking the plate with 2 % BSA in TBS, 70 μL phage binding buffer was added to all the wells and 30 μL phage in TBS was added to each row. After a 2 h incubation at RT, phages were detected using an anti-M13-HRP conjugate (SinoBiological, China) and 3, 3’, 5, 5’-tetramethylbenzidine supersensitive liquid substrate (Sigma Aldrich). The signal from rows containing RBD-A to that from rows without RBD-A were divided to obtain a signal:background ratio.

4.3 Results

4.3.1 Polymer Synthesis

The poly(HPMA)-based polymers used in this study were synthesized using reversible addition- fragmentation chain transfer (RAFT) polymerization because of the method’s versatility and the lack of 65

CHAPTER 4: POLYMERIC BINDER FOR TCDA requirement of a catalyst that is difficult to remove and can influence cellular assays (347). Our first efforts to synthesize polymers with reactive groups that could be functionalized in a post-polymerization reaction were centered on the established use of monomers containing activated esters that readily react with primary amine-containing ligands (348). We synthesized a random copolymer of HPMA and N- methacryloyl hydroxysuccinimide (NMS) (Scheme 4.1, 51) containing 30 mol% NMS with Mn = 24,000 g/mol and Mw/Mn = 1.04 (349). The degree of functionalization was selected so as to achieve an activated ester density above the spacing of ligand-binding pockets on RBD-A (∼ 30 Å) as evident from the 3D structure of RBD-A bound to its trisaccharide ligand: assuming a stretch linear polymer conformation, a tetrahedral arrangement around C atoms and a 0.15 nm C-C bond length gives a theoretical optimum for degree of functionalization of 12.5 % (333). Our reasoning was that, similarly to the GM1 pentasaccharides discussed in chapter 2, having similar spacing of ligand binding sites found on the target and ligands on the polymer would maximize affinity.

SCHEME 4.1 Synthesis of polymers containing activated esters or an azide group by reversible addition- fragmentation transfer (RAFT) polymerization. Copolymers of 47 with 48 and 50, and homopolymers of 49 were prepared. 4-Cyanopentanoic acid dithiobenzoate served as the chain transfer agent (CTA) for all reactions.

Either azobisisobutyronitrile (AIBN) or 4,4′-azobis(4-cyanovaleric acid) (ACVA) were used as the initiator.

Before attempting to graft the desired ligands onto the polymer, optimization of the postpolymerization modification reaction was carried out using 1-amino-2-propanol to convert 51 to poly(HPMA). 1H NMR spectroscopy revealed the apparition of an unexpected peak at 2.6 ppm, and a peak ratio of protons b to a of less than the expected two (Fig. 4.2). 66

CHAPTER 4: POLYMERIC BINDER FOR TCDA

FIGURE 4.2 1H NMR spectra of poly(HPMA) from homopolymerization (top) and aminolysis of 51 by 1-amino-

2-propanol (bottom) taken in D2O. The peaks at 3.6 ppm in the top spectrum correspond to a small impurity that is not incorporated in the polymer, as concluded from the sharpness of the peaks compared to the more rounded peaks typical of polymers.

Optimizing reaction time, temperature and reagent equivalents all failed to ameliorate the reaction outcome, as did screening for different nucleophilic catalysts (triethylamine, hydroxybenzotriazole, 4-dimethylaminopyridine, pentafluorophenol). A literature search revealed plausible side reactions that could explain our observations (350). The unexpected peak at 2.6 ppm was most probably due to succinimide ring opening whereas the peak ratio could be affected by the formation of glutarimide between neighboring active esters, such that the resulting polymer contained a mixture of the side reactions as shown in 54 (Scheme 4.2A). Although modifying reaction conditions could push the reaction more towards one or the other side reaction, they could never be eliminated or reduced to a satisfactory degree (e.g. < 5 %).

CHAPTER 4: POLYMERIC BINDER FOR TCDA

SCHEME 4.2 Side reactions occurring during aminolysis of 51 (A) and 52 (B) by 1-amino-2-propanol.

As an alternative to 48, we synthesized pentafluorophenyl methacrylate (49) and homopolymerized it to yield 52. The acrylate group of 49 is more reactive than 48 and could therefore readily be homopolymerized. Aminolysis reactions with 52 have been reported to proceed with high conversions and an absence of side reactions; 49 cannot undergo a ring-opening reaction analogous to that observed with 48 (340). However, as with 51, we consistently observed a high proportion of glutarimide formation (Scheme 4.2B), which we also observed when we attempted to graft the control ligand 16 onto the polymer.

SCHEME 4.3 Amine or alkyne-terminated lactose (56) and αGal(1-3)βGal(1-4)Glc trisaccharide ligand (57) for grafting onto polymers containing pendant active esters or azides.

We therefore proceeded to the synthesis of 50, a monomer containing an azide group rather than an activated ester, to which an alkyne-containing molecule could be conjugated by copper-catalyzed azide-alkyne cycloaddition (“click”) chemistry (351). Co-polymerization of 47 with 50 yielded 53 with

Mn = 22,000 – 29,000 g/mol and Mw/Mn = 1.5 – 1.6, having degrees of incorporation of 50 of 5 % – 19 % (polymers A – C, Table 4.1). Larger polymers (and with greater polydispersity indices) could only be obtained by omitting the CTA from reactions to carry out uncontrolled polymerizations (polymers D

CHAPTER 4: POLYMERIC BINDER FOR TCDA

& E, Table 4.1). We observed probable loss of the dithiobenzoate end group originating from the CTA during polymerization through a change in color from pink to white, and hence we did not make an effort to remove the end group prior to functionalization (352).

TABLE 4.1 Polymer precursors (53) for functionalization with 56 and 57. The columns indicate the molar percent of incorporation of monomer 50, the average number of 50 monomers incorporated per polymer chain, the number average molecular weight (determined by gel permeation chromatography) and the polydispersity index of the polymers.

3 Identifier % of 10 # of 10 / chain Mn (10 x g/mol) Mw/Mn A 5 10 29 1.5 B 10 14 22 1.6 C 19 28 25 1.5 D 10 34 54 3.2 E 20 38 33 2.8

Polymers A – C were all functionalized with 56-alkyne and 57-alkyne to give 58, as shown in Scheme 4.4 for 57. A peak corresponding to the triazole of 58 was visible at 7.8 ppm by 1H NMR and the conversion of the reactions were full as evidenced by complete disappearance of the azide peak in Fourier transform infrared (FTIR) spectroscopy (Fig. 4.3).

SCHEME 4.4 Conjugation of trisaccharide ligand (57) to 53 by “click“ chemistry.

CHAPTER 4: POLYMERIC BINDER FOR TCDA

FIGURE 4.3 FTIR spectrum of A before and after conjugation with 56. The arrow indicates the position of the azide peak, which is, not evident in the spectrum of A-56.

4.3.2 In vitro testing of polymer efficacy

In order to obtain a rapid and direct measure of the ability of our functionalized polymers to reduce TcdA-mediated cytopathy, we directly carried out cellular assays on HT-29 colon adenocarcinoma cells. Similarly to TcdB, TcdA induces cell rounding indicative of cytopathy in a variety of cell lines (353). After attempting to use an image-based manual counting of cell rounding, which was unreliable, and a fluorescence-activated cell sorting (FACS) method, which failed to yield clear-cut differences in signals between intoxicated and healthy cells, we decided to use an automated impedance monitoring system for measuring cell rounding, as described in chapter 3. Cells were grown to semi-confluence for 24 h before addition of serial dilutions of TcdA. Incubation of HT-29 cells with TcdA led to dose-dependent variations in cell index (Fig. 4.4). Based on these results, a concentration of 2 ng/mL TcdA leading to moderate intoxication was selected for cytopathy inhibition experiments.

CHAPTER 4: POLYMERIC BINDER FOR TCDA

FIGURE 4.4 Evolution of cell index of TcdA-intoxicated HT-29 measured by impedance-monitoring. Cells were seeded 24 h before the 0 h time point. TcdA was added and the cell index normalized to 1.0 at the 0 h time point (mean, n = 2).

To test the cytopathy inhibition efficacy of A – C coupled to 57 (and 56 as a control), 10 μL of 10 mg/mL polymer in PBS was added to 90 μL complete medium with a final concentration of 2 ng/mL TcdA and immediately added to semi-confluent HT-29 cells and incubated for 24 h. We first tested polymer C conjugated to 56 and 57 respectively (Fig. 4.5A). No differences were observed between the control polymer C-56 and the ligand-functionalized polymer C-57, and both of them failed to reduce TcdA-mediated cytopathy. Similar lack of efficacy was observed with polymers A-57 and B-57, with 5 % and 10 % functionalization versus the 20 % of C-57 (Fig. 4.5B). Pre-incubating the toxin-polymer mixture before addition to cells did not influence the outcome and neither did pre-incubation in serum- free conditions (data not shown). Further increasing the concentration of polymer was not feasible due to material restraints but the 1 mg/mL concentration used already corresponded to a 106-fold molar excess of polymer over TcdA. With the aim of increasing steric hindrance around the bound toxin, 57-

NH2 was conjugated to a generation 5.5 poly(amidoamine) dendrimer with a carboxylate surface (47 kDa) using DMTMM as a coupling agent with 15 % loading but this conjugate also did not show any efficacy in cell assays (data not shown).

CHAPTER 4: POLYMERIC BINDER FOR TCDA

FIGURE 4.5 Evolution of cell index of HT-29 cells after addition of TcdA-polymer mixtures. Cells were seeded 24 h before the 0 h time point. TcdA-polymer mixtures were added and the cell index normalized to 1.0 at the 0 h time point. A, Assay using functionalized polymer C. B, Assay using functionalized polymers A and B (mean, n = 2).

To determine whether or not the lack of efficacy observed in cellular assays was due to lack of binding of the polymers to TcdA we expressed recombinant receptor binding domain of toxin A (RBD- A), which could be used in binding assays requiring larger amounts of protein that the cellular assays.

4.3.3 Recombinant expression of RBD-A

The recombinant expression of the full RBD-A (amino acids 1739 – 2710) in E. coli with a hexahistidine tag has been published in isolated reports (90, 354). The expression is challenging because the protein is relatively large (ca. 100 kDa) and the codon usage of E. coli is different to that of C. difficile (355). Indeed, codon usage analysis revealed that 16 % of the codons in the DNA coding for RBD-A are “rare” to E. coli (i.e. occur at a frequency of < 10 %), and that these are particularly densely situated towards the N-terminal portion of the protein. E. coli expresses low levels of tRNAs corresponding to these “rare” codons and their expression is therefore hampered (356). As a consequence, Bacillus megaterium is often used as an expression host for C. difficile proteins because it has a similar codon usage to C. difficile. To overcome this issue, we attempted to use E. coli BL21 expression hosts that contained a plasmid coding for additional tRNAs for 6 “rare” codons. This approach did not result in expression of protein of acceptable purity or yields, possibly because of the metabolic burden of the additional plasmid in the expression host. We ultimately found that expressing a slightly N-terminal truncated fragment of RBD-A (amino acids 1875 – 2710) that still contained all

CHAPTER 4: POLYMERIC BINDER FOR TCDA seven trisaccharide binding sites and carrying out expression at low temperatures (16 °C) using standard BL21 cells gave satisfactory, albeit low, yields of pure and soluble protein (2 – 3 mg/L bacterial culture) (Fig. 4.6). The expressed protein also gave a positive signal in an ELISA using anti-RBD-A antibody.

FIGURE 4.6 Coomassie-stained SDS-PAGE of RBD-A purified by nickel affinity chromatography.

A significant improvement to the expression of C. difficile toxins in E. coli that was recently reported consists of using a synthetic GC-enriched (from G + C = 27 % to G + C = 45 %) and codon- optimized copy of the gene encoding the toxin for insertion into the expression plasmid (357). Using this approach, the investigators were able to express TcdB holotoxin in E. coli with activity indistinguishable from that of TcdB produced in B. megaterium.

4.3.4 RBD-A Binding Assays

We used rabbit erythrocyte hemagglutination assays and thermal denaturation assays to qualitatively assess the binding of our polymers to RBD-A. The former exploits the fact that rabbit erythrocytes have abundant αGal(1-3)βGal(1-4)Glc on their surface (erythrocytes from other mammals do not) and that incubation of the erythrocytes with TcdA leads to hemagglutination (92). One would therefore expect that it could be possible to compete for binding with the trisaccharide displayed on the erythrocytes using a synthetic trisaccharide display vehicle. We found that C-57 did not reduce hemagglutination more than C-56 or C, implying that the trisaccharide as displayed on C-57 could not inhibit RBD-A binding to the erythrocytes (Fig. 4.7A). The thermal denaturation assay indicates the thermal stability of a protein by allowing determination of its melting point (Tm, the midpoint of thermal denaturation) (358). Binding of a ligand to a protein typically stabilizes its conformation, rendering it more resistant to thermal denaturation, which results in an increase in Tm. In this assay too, C-57 did not show any stronger binding to RBD-A than C did, with an identical Tm recorded when RBD-A was

CHAPTER 4: POLYMERIC BINDER FOR TCDA

incubated with both polymers (Fig. 4.7B). Free 57 did not affect the Tm of RBD-A either (data not shown). We also attempted to graft 57-NH2 to amine-binding 96-well plates, and biotin-conjugated 57 to streptavidin-coated 96-well plates, so as to quantitatively determine RBD-A binding to 57 by competitive enzyme-linked immunosorbent assays (ELISA) but were not able to observe binding using these setups either (data not shown).

FIGURE 4.7 A, Rabbit erythrocyte hemagglutination assay. The minimal concentration of RBD-A to cause full hemagglutination (18.7 ug/mL) was added to rabbit erythrocyte solutions containing a dilution series of polymers. The concentration of polymer used was 2.5 mg/mL. Each well was visually scored for hemagglutination: 0 – none, 1 – moderate or 2 – complete hemagglutination. The data points for C and C-57 are superimposed. B, Thermal denaturation assay. RBD-A at 1 μM (100 μg/mL) was added to 20 mg/mL polymer or buffer and SYPRO orange dye solution (1X). The fluorescent signal was measured as temperature was increased in 1 °C increments using an RT-PCR device. The minimum of the derivative of fluorescence with respect to temperature indicates the midpoint of thermal denaturation (Tm). The Tm of RBD-A in buffer was measured to be 49.5 °C, that of RBD-A + C and of RBD-A + C-57 49.9 °C.

4.3.5 Identification of TcdA-binding peptides

We hypothesized that the general lack of activity and binding that we observed with 57 conjugated to polymers A – C was due to the low binding affinity of 57 to TcdA, which is of the order of Kd = 1 mM (165). Screens for carbohydrates that bind TcdA carried out by the Consortium for Functional Glycomics (www.functionalglycomics.org) have revealed that a large degree of variation in carbohydrate structures is tolerated by TcdA and stronger binders than 57 have recently been identified (359, 360). Although repeatedly suggested and even patented, use of these ligands to develop novel binding inhibitors has yet to be published. We therefore decided to attempt a completely different approach using phage display technology to identify short peptide ligands for TcdA, which has to date not been reported in the literature for RBD-A or RBD-B (361). The use of phage display to generate carbohydrate mimetics is not unprecedented and has the advantage that phages with carbohydrate mimetic peptides displayed on their surface can readily be amplified whereas the synthesis of complex carbohydrates remains challenging (362, 363). We selected a well-established and extensively used

CHAPTER 4: POLYMERIC BINDER FOR TCDA library of M13 phages that display heptapeptides fused to the minor coat protein pIII in a pentavalent manner: New England Biolab’s Ph.D. 7 library (364). Libraries of phages displaying 12mer peptides and cyclic heptapeptides also exist but we reasoned that the size and conformation of linear heptapeptides resemble oligosaccharides more closely.

Affinity maturation of binding sequences was carried out either by surface panning or solution panning with solid support affinity capture (using Ni-NTA or streptavidin coated on agarose beads or 96-well plates). Our first attempts employed solution panning using Ni-NTA capture since solution panning offers the advantage of maintaining the target protein in solution and therefore in its native conformation, and because the RBD-A we expressed already contained a 6His tag and hence did not require additional modification. After four rounds of panning, individual phage clones were amplified and tested for binding to RBD-A by a direct phage ELISA, and the ratio of the signals of phage incubated with and without RBD-A were calculated (346). Typically, a signal:background ratio cut-off of at least three is used for classifying a given phage as a “binder” (365). As is evident from the results (Fig. 4.8A), the majority of signal:background ratios obtained were between 1.0 and 1.5, indicating that there were no strong binders. A variety of modifications to the panning procedure were made in an attempt to encourage the selection of binding sequences. These included: (i) using thrombin** to elute bound phages rather than low pH so as to increase elution specificity, (ii) omitting negative selection to prevent possible loss of small populations of strong binders, (iii) adjusting phage binding buffer pH, NaCl content and blocking protein concentration and nature to reduce non-specific interactions, (iv) alternating Ni-NTA bound to agarose beads with Ni-NTA bound to 96-well plates to reduce amplification of matrix binders, (v) increasing or reducing RBD-A concentrations and (vi) adjusting incubation times. These adjustments did not enable us to obtain strong binders and sequencing of phages selected at the end of the various panning experiments revealed that 95 % of phages sequenced (the DNA of 96 phages was sequenced) contained two or more histidine residues, suggesting that affinity for Ni-NTA was consistently being preferentially selected.

We subsequently employed surface panning and solution panning using biotinylated RBD-A with streptavidin capture. Direct phage ELISA of selected clones from these rounds revealed 5 clones yielding signal:background ratios above 1.5 but none were above 3 (Figs. 4.8B and 4.8C). We nevertheless sequenced the 20 phages giving the highest signals and did not observe any consensus sequence. The only two sequences that occurred more than once were TVNFKLY, which appeared 5 times, and HYIDFRW, which appeared twice. The only common factor that the remaining sequences had with these two was a high degree of hydrophobicity, which is associated with non-specific interactions with e.g. plastic (366). The peptide HYIDFRW had also been isolated on several occasions during our Ni-NTA capture solution panning experiments and we therefore sought to verify if this was

** The pET28a plasmid used for expressing RBD-A encodes a thrombin cleavage site between the N-terminal 6His sequence and the RBD-A insert. Incubating 6His-RBD-A bound to Ni-NTA with thrombin should therefore release the protein from the Ni-NTA support. 75

CHAPTER 4: POLYMERIC BINDER FOR TCDA due to a genuine interaction with RBD-A or some target-unrelated selection. We were, however, not able to observe any displacement of phage displaying HYIDFRW by free peptide in a competition ELISA where increasing concentrations of free peptide (up to 100 μM, precipitation observed at 1 mM) were added to RBD-A mixed with phage (5.5 x 1011 phages corresponding to a peptide concentration of 46 nM). This suggests that interactions between phage and RBD-A were probably not mediated by the peptide sequence but non-specific electrostatic or hydrophobic interactions.

FIGURE 4.8 Direct phage ELISA. A, Results with phages selected after 4th round of solution panning using Ni- NTA capture. B, Results with phages selected after 3rd and 4th rounds of solution panning with streptavidin capture. E1-5 from 4th round, E6-12 & F1-10 from 3rd round of panning. C, Results with phages selected after 3rd and 4th rounds of surface panning. F11-12 and G1-12 from 4th round surface panning, H1-11 from 3rd round. H12 control without phage.

4.4 Discussion

We have described the synthesis and in vitro testing of polymeric binders designed to inhibit binding and uptake of TcdA, and a phage display campaign aimed at identifying novel, high affinity 76

CHAPTER 4: POLYMERIC BINDER FOR TCDA peptide ligands for RBD-A. The end results of both of these endeavors fell short of anticipated outputs but nevertheless offer valuable insights into the development of polymeric binders and of novel ligands for RBD-A. The first thing we learned was that caution needs to be used when trying to functionalize poly(HPMA) containing activated esters because avoiding side-reactions is very challenging indeed. To try to explain the lack of activity with the “click”-functionalized polymers we finally obtained, we can consider in turn the three key components of the target-inhibitor interaction: the target itself (RBD-A), the support for the targeting ligand (the poly(HPMA) backbone) and the targeting ligand (17).

The RBD has long been shown to mediate interactions with glycan structures and that these are involved in binding of TcdA to cells – similar conclusions are drawn for TcdB based on the structural similarity of its RBD with that of TcdA (92, 367, 368). The last few years have, however, seen a surge in reports suggesting that the domain adjacent to the C-terminal RBD, the translocation domain, which is the least well characterized and understood domain of TcdA and TcdB, is also involved in binding and uptake. The evidence is based on experiments showing that RBD-truncated TcdA is still capable of entering host cells and that although recombinant RBD-A is able to compete for binding with TcdA, a protein fragment encompassing the RBD-A and part of the translocation domain shows stronger competition (369–371). Just how the cryptic regions of the translocation domain interact with cellular membranes to mediate endosomal uptake, pore-formation and translocation is just beginning to be understood (357, 372). The conclusion to draw from these observations is that aiming to inhibit merely inhibiting the RBD’s ability to bind carbohydrates may not be sufficient for effectively inhibiting uptake, due to the putative involvement of portions of the translocation domain and non-carbohydrate binding regions of the RBD in toxin binding. Indeed, point mutations in the translocation domain that prevent translocation and/or pore formation of TcdB were recently shown to reduce toxicity over 1,000-fold, suggesting a strong potential for novel inhibitors that act beyond the RBD (357). However, the inhibitory effect of a strong binder to the RBD should certainly be investigated before a polymeric binder that interacts with both the RBD and the traslocation domain is designed.

The fact remains that antibodies that specifically target the RBD of TcdA and TcdB are able to inhibit toxin binding. Indeed, based on crystal structure data, binding of inhibitory antibodies to the toxins does not lead to the formation of large immune complexes but that it creates steric hindrance or induces a conformational change, which inhibit the toxin’s interaction with its receptor (368). Surprisingly, direct occlusion of the carbohydrate binding sites is not a pre-requisite for an antibody to effectively neutralize toxin activity, supporting the involvement of areas other than the carbohydrate binding sites in receptor binding (373). A consideration of the physical parameters of antibody-toxin interactions can help to identify limitations of and potential improvements to the polymer backbone that we employed. First, the molecular weights of the polymers we synthesized were smaller (Mn < 77 kDa after functionalization) than that of IgG antibodies (∼ 150 kDa), which implies that the total steric hindrance afforded by one polymer chain was less than that due to a single antibody molecule, although

CHAPTER 4: POLYMERIC BINDER FOR TCDA the extent to which this plays a role also depends on the polymer-toxin binding stoichiometry. The unbranched and highly flexible polymer is furthermore likely to exist as a random coil with very low structural stability, in contrast to antibodies, which are tetramers of polypeptides with defined tertiary structures. This renders the latter more rigid and therefore better suited to inhibit toxin-receptor interactions than the former, which is more likely to yield and make room for the toxin to engage with its receptor rather than to prevent the two from getting into close proximity from one another. One could speculate that the larger size and structural stability of tolevamer (600 kDa) and Synsorb-90 (large microparticles) compared to our polymers contributed to both of the former’s reported in vitro efficacy.

Although the binding affinity of 57 is in the millimolar range, we anticipated that the multivalent display afforded by our polymer would boost affinity, as has been observed with other carbohydrate display vehicles, such as those presented in chapter 2. In addition to providing weaker steric hindrance, having a highly flexible polymer backbone has a high entropic cost that leads to negative cooperativity of multivalent carbohydrate-protein interactions (374–376). Indeed, scaffold rigidity is a pre-requisite for positive cooperativity in any multivalent interaction. In addition to our polymer backbone having high flexibility, we also needed to add a linker to the ligand to enable grafting onto the polymers and as a result, the ligand was positioned a full 45 atoms away from the polymer backbone (cf. 18), resulting in a further increase in flexibility. In addition, RBD-A being a linear monomer that adopts an elongated serpentine structure (377), the density of the seven carbohydrate binding sites (each site ∼ 30 Å apart) is much lower than for e.g. Shiga toxin, where the pentameric B subunits display 15 carbohydrate binding sites within a radius of 40 Å: a 10 million-fold increase in affinity has been observed with multivalently displayed carbohydrate ligands in this case (184). What makes the problem even more challenging in the case of RBD-A is the fact that the carbohydrate binding sites are not displayed on a single plane but differ in rotation around the protein’s axis by 90° or 120°(330). This implies that there is a trade-off between the flexibility required to reach all of the binding sites and the rigidity required for effective multivalency.

The highest density of ligands on the polymers that we tested was 20 mol%, which was already higher than what we calculated would give roughly the same spacing of ligands on the polymer as that of ligand binding sites on the target. However, having a much higher density of ligands (e.g. close to 100 %) could have the obvious effect of increasing local ligand concentration and additionally increasing rigidity, which could synergistically contribute to achieving a stronger “glycocluster effect” resulting in higher affinity (378). To achieve such high ligand densities using the azide-containing monomer we synthesized (10) would require optimization of the reaction conditions, since we already observed that when carrying out a non-controlled polymerization with 47 and 50, increasing the molar proportion of 50 from 10 % to 20 % resulted in a reduction in polymer size from 54 kDa to 33 kDa. Increasing the monomer concentration in the reaction may for example be needed to enable the synthesis of high

CHAPTER 4: POLYMERIC BINDER FOR TCDA molecular weight polymers using high contents of 50. An alternative would be to synthesize a methacrylamide monomer with 57, which could then be homopolymerized.

The third component of the target-polymer interaction that is amenable to optimization is the binding moiety. We reasoned that having a ligand with high affinity for RBD-A could offset some of the weaknesses of our polymer discussed in the previous paragraphs. As already mentioned, attempts have been made to find carbohydrate ligands with affinity superior to that of 57 with limited success, which is why we tried to identify short peptide binders for RBD-A through phage display.

The phage display strategies we employed were unsuccessful in yielding strong binders for RBD- A. The strongest contributor to this shortfall was likely to be high background binding due to RBD-A’s high isoelectric point (pI) of 8.4, which confers it with a positive charge at neutral pH (379). Since the phage coat is negatively charged, strong non-specific electrostatic interactions render selecting for phages that bind to RBD-A due to the affinity of the peptides displayed on its coat difficult, as has been observed by others for multiple high pI proteins (380). Indeed, proteins with a pI > 9 are considered “nonstarters” for phage display experiments. Neither the use of various blocking proteins (BSA, casein, SuperBlock from ThermoScientific), nor high detergent concentrations (0.5 %) in the binding buffer were able to eliminate non-specific binding in our case, although high NaCl concentrations (0.5 M) did reduce them significantly.

Two approaches have been suggested to overcome these issues: genetic and chemical phage wrapping to shield from electrostatic interactions. Genetic wrapping consists of introducing a sequence coding for a peptide that contains lysine into the phage coat DNA (380). This offsets the otherwise net negative charge of the coat and was shown to reduce or eliminate non-specific binding of phage to high pI proteins. A disadvantage of this approach is that the mutant phages generated are less stable and have lower infectivity than wild type phage. Chemical wrapping consists of incubating phage particles with a cationic polymer before carrying out selection. Oligolysine can be used for this purpose but synthetic polymers such as poly(ethylamine methacrylate) more effectively shield the negative charges without affecting infectivity, and are more stable and less costly to produce (381).

Regardless of the phage display approach used, the carbohydrate binding sites on RBD-A are such that it is fundamentally difficult for any small molecule, be it a carbohydrate, a peptide or other, to bind strongly to them. The crystal structure of αGal(1-3)βGal(1-4)Glc bound to a fragment of RBD-A (Fig. 4.1) clearly shows that the binding pockets are open and shallow troughs, where only ∼ 50 % of the carbohydrate surface is buried when bound. As such, these binding sites inherently mediate weak binding relative to narrow and deep pockets. One way to increase the probability of success of a phage display campaign against RBD-A could therefore also be to use commercially available 12-mer phage libraries, which would have a larger surface of interaction than the peptides in the 7-mer library we used.

CHAPTER 4: POLYMERIC BINDER FOR TCDA

In conclusion, the poly(HPMA)-based polymers functionalized with αGal(1-3)βGal(1-4)Glc we synthesized did not inhibit cytotoxicity of TcdA. In addressing the probable weaknesses in vivo of tolevamer and Synsorb 90, namely high negative charge and large size limiting diffusion through the mucus layer, our use of a highly flexible and soluble poly(HPMA) backbone was likely detrimental to achieving strong binding through multivalency and effective steric hindrance. Furthermore, our attempts to identify peptide binders for RBD-A highlighted the challenges of applying standard phage display techniques to high pI target proteins without deep binding pockets such as RBD-A.

4.5 Remark

We are grateful to Simon Bussmann, who carried out phage display experiments with Ni-NTA capture during his Master’s thesis work, and to Max Pillong and Professor Gisbert Schneider for assistance with thermal denaturation assays. Prof. Bastien Castagner carried out the synthesis of all carbohydrates. All other experiments were carried out by Mattias Ivarsson.

5 Conclusions and Outlook

The work presented in this thesis has allowed us to gain a better understanding of key design considerations for the development of inhibitors of C. difficile TcdA and TcdB, and has resulted in the generation of a lead candidate with potential for treating CDI.

In our first study (chapter 3), we developed a strategy to induce TcdB auto-proteolysis in the colon lumen. We showed that the pre-emptive release of the glucosyltransferase domain from the remainder of the toxin thus achieved reduced cytopathy in a tissue culture assay and diminished inflammation in a germ-free mouse model of CDI. The cornerstone of our approach was the realization that the endogenous activator of intracellular cleavage of the toxins, IP6, is not soluble at the high concentrations of calcium typically found in the gut lumen, and that replacement of some of its phosphate groups by sulfate groups restores the molecule’s solubility while maintaining its ability to activate the toxin’s CPD.

Our efforts to solubilize IP6 in calcium solutions first consisted of replacing the 2-phosphate, the phosphate least involved in binding according to the crystal structure of the CPD-IP6 complex, by poly(ethylene glycol) (PEG) chains of 500 Da or 2,000 Da (95). These modifications only led to slight improvements in activity and we postulated that although the PEG chains could sterically prevent aggregation and precipitation of the molecules, calcium would still compete with the toxin for binding to the phosphates. We therefore changed our strategy to that presented in chapter 3, namely replacing the phosphate groups of IP6 by sulfate groups, based on the knowledge that calcium sulfate is 10 times more soluble in water than dicalcium phosphate. By synthesizing IP6 analogs with 3 to 6 substitutions of phosphates by sulfates and characterizing their solubility, binding affinity and activity, we identified four substitutions as giving the best tradeoff between in vitro solubility and activity. The lead compound thus identified has phosphates in the 4 and 6 positions (SSSPSP). Future work will explore what influence the positioning of the substitutions has on activity, and preliminary data suggests that having the phosphate groups in other positions either decreases or maintains similar activity, although further experiments are necessary to confirm these observations. A recent study investigating the conformation and geometry of IP6-calcium complexes suggests that as pH is lowered from alkaline to neutral, the first phosphates to become protonated are those in the 5 and 4/6 positions, suggesting that the metal ion tends to position itself towards the 2 and 1/3 phosphates. Although it is unclear to what extent these results may be extrapolated to mixed sulfate/phosphate analogs of IP6, the data would suggest that the disposition of phosphates in our lead molecule already disfavors interactions with calcium more than other isomers, and concurrently favors interactions with the toxin (382).

The in vivo data we obtained in a mouse model of CDI support the hypothesis that our lead molecule has therapeutic potential. However, neither the inoculum, the dose nor the dosing schedule was optimized and hence further in vivo studies are clearly needed to establish a more rigorous proof- of-concept. Indeed, relative to the anticipated complete lack of toxicity of our lead compound at doses 81

CHAPTER 5: CONCLUSIONS AND OUTLOOK below 1,000 mg/kg, the dose of 10 mg/kg we used was very small indeed and could readily be increased by a factor of e.g. 5 – 10. Furthermore, the number of daily doses administered could be increased from two to three or even four to enable constant exposure to the drug candidate. We hope that the ongoing scale up of the synthesis of the lead compound will enable the establishment of ideal dosing regimens without needing to take into consideration compound scarcity. We also aim to obtain data in other species, such as the hamster, which is particularly susceptible to CDI and constitutes the industry gold standard for testing novel therapies for CDI. Such studies, in combination with in vivo confirmation of the lack of toxicity of our lead compound will provide a strong case for further pre-clinical and clinical development of the molecule.

One aspect that we have not covered in characterizing our compounds is their susceptibility to digestion by phytases. Phytases are inositol phosphate-specific phosphatases and can be present in the gastro-intestinal tract following oral intake of plant-based foods, which contain large amounts of phytase, or due to microbial phytases produced by the gut microbiota (383). Given the structural similarity of our IP6 analogs to IP6 it is to be expected that phytases are also able to hydrolyze the analogs, and indeed we have observed that when incubated with high concentrations of phytase, the IP6 analogs lose the ability to cleave CPD-B. Phytases have also been shown to bind IS6 (384). However, it is very difficult indeed to make an estimate of what concentration and what kind of phytase are relevant in the context of CDI as reports on phytase activity in healthy humans are already scarce. Oral intake of phytases can easily be controlled and eliminated e.g. by making sure to eat only cooked/heat treated plant-based foods. On the other hand, it is much more difficult to reduce the activity of microbial phytases in the digestive tract, which is thought to be low in humans. Since CDI patients have a compromised gut microbiota it is plausible that if anything, they exhibit reduced microbial phytase activity (385). These hypotheses will certainly need to be verified in the future development of our lead compound as a drug candidate, for example by incubating the molecule in supernatants of stool from CDI patients and assessing degradation using post-column derivatization-based high pressure liquid chromatography (HPLC) methods (386).

In order to fully understand the scope of the activity of our lead compound, it will also be important to characterize its effect on TcdA. TcdA undergoes IP6-induced cleavage much less effectively than TcdB in buffers and it has recently been shown that this is due to interactions of the RBD domain with the N-terminal domains in TcdA (387). It is unclear to what extent these observations reflect the in vivo situation since the gut lumen is rich with a plethora of molecules that could disrupt this intramolecular interaction. Therefore, in vivo experiments using genetically engineered TcdA+TcdB- strains of C. difficile are probably the most reliable way of characterizing the effect of our compounds on TcdA. Ideally, an assay would be developed to show that luminal toxin cleavage is taking place. We attempted to use Western blotting and immunoprecipitation to demonstrate toxin cleavage in the cecal contents from the mice used in our in vivo study but were not able to detect any toxin, most probably because the

CHAPTER 5: CONCLUSIONS AND OUTLOOK amount of toxin in the samples was very low. Carrying out a similar procedure using stool from hamsters, where larger quantities would be available may turn out to be more fruitful.

In our second study (chapter 4) we attempted to develop a polymeric binder capable of binding to TcdA, thereby inhibiting its uptake and cytopathy. The design of the inhibitor was based on attaching a known ligand for TcdA, αGal(1-3)βGal(1-4)Glc, to a flexible, soluble and biocompatible polymer commonly used in biomedical applications: poly(hydroxypropyl methacrylamide) (poly(HPMA)). We showed that these polymers, at molecular weights up to 77 kDa and degrees of functionalization of up to 20 % were unable to inhibit TcdA-mediated cytopathy. Assuming that the main reason for this lack of activity was the very low affinity (1 mM) of the ligand for the toxin, we attempted to identify novel peptide binders for the receptor-binding domain of toxin A (RBD-A) using phage display of heptameric peptides. Using this approach, we were unable to identify strong binders specific for RBD-A. The high pI of RBD-A (8.4) is believed to have contributed to the challenge because at neutral pH, RBD-A was likely to bind non-specifically to the negatively charged phage particles, despite the use of high concentrations of salt, detergent and blocking proteins. It may be that RBD-B, which has an acidic pI, would be more readily amenable to successful phage display panning. Use of cationic polymers to sheath the phages’ negative charge before carrying out selection could also provide an alternative to increase the probability of success of a phage display selection against RBD-A. In addition, based on their crystal structures, neither RBD-A nor RBD-B contain deep binding pockets that are favorable for the identification of strong binders.

Based on our experience, several key factors should be taken into consideration for future efforts to develop an effective polymeric binder for TcdA and TcdB. First, the use of larger, more rigid and branched scaffolds than the poly(HPMA) we employed are likely to offer more effective steric hindrance of the toxin-receptor interaction and a more effective multivalent effect. Nanogels of hyperbranched polyglycerols are emerging scaffolds that could offer such properties (388). Using a higher ligand density than what we did could also contribute to more effective binding, although this is limited by the large scale availability of the trisaccharide ligand. Combining these changes with the use of a stronger ligand would increase the likelihood of a polymeric binder being effective at inhibiting the toxins.

To conclude, the work carried out in the context of this doctoral thesis has allowed us to learn (i) that the intracellular activation of TcdA and TcdB is due to chelation of otherwise abundant IP6 in the gastro-intestinal tract, (ii) that replacement of phosphate by sulfate groups on IP6 can be used to tune its affinity for a divalent cation and a protein co-factor, (iii) that pre-emptive cleavage of TcdB reduces cytopathy, (iv) that a sulfate-phosphate hybrid analog of IP6 reduces toxin-mediated inflammation in a murine model of CDI and that (v) multivalent display of the well characterized αGal(1-3)βGal(1-4)Glc ligand for TcdA on a highly flexible and soluble linear polymer does not prevent cytopathy caused by TcdA. The knowledge acquired through this work is a significant contribution to our understanding of the mechanisms of action of TcdA and TcdB, and provides a drug lead for further pre-clinical testing. 83

6 Appendix

6.1 IP6 Analog Characterization

All IP6 analogs were synthesized in the Laboratory of Prof. Leroux by Prof. Bastien Castagner and Estelle Durantie, who describes the synthetic methods in her doctoral thesis. The characterization of the final molecules used are provided below for reference.

6.1.1 Characterization of IP3S3

+ + The product IP3S3 was a white solid (ammonium salt, 16 mg, as PSPSPS·2Et3NH ·6NH4 (by 1 1 H NMR) MW 963.84, 17 µmol, 50 %). H NMR (400 MHz, D2O) δ (ppm) 4.98 (1H, bd, H-C2), 4.90

(2H, bd, H-C4/6), 4.47-4.60 (3H, m, H-C5 and H-C1/3), 3.22 (14H, q, J 7.3 Hz, CH3CH2NH), 1.30 13 (21H, t, J 7.3 Hz, CH3CH2NH); C NMR (125 MHz, D2O) δ (ppm) 76.9-77.2 (m, C4/6), 71.3 (d, 2 31 1 JCP 5.1 Hz, C1/3), 46.7 (CH3CH2NH), 8.2 (CH3CH2NH); P NMR (162 MHz, H-decoupled, D2O) δ 9- + + - (ppm) -0.39 (2P, s, P-C1/3), -0.81 (1P, s, P-C5); [m/z (ESI) [IP3S3 ·7H ·3NH4 ] C6H25N3O24P3S3 calculated 711.9197, found 711.9199]. C2 and C5 were neither visible on the 13C NMR nor on the HSQC spectra.

6.1.2 Characterization of IP1S5

The product IP1S5 was a white solid (ammonium salt, 10.29 µmol, 40 % over 2 steps, the concentration of the final product was determined by 1H NMR with dioxane as internal reference). 1H

NMR (400 MHz, D2O) δ (ppm) 5.31 (1H, br, H-C2), 4.85 (2H, br), 4.68 (2H, br), 4.42-4.50 (1H, m, H- 13 C5), 3.22 (42H, q, J 7.3 Hz, CH3CH2NH), 1.30 (63H, t, J 7.3 Hz, CH3CH2NH); C NMR 500 MHz

(125 MHz, D2O) δ (ppm) 76.0 (br, C4/6), 73.7 (s, C1/3), 46.7 (CH3CH2NH), 8.3 (CH3CH2NH); 31 1 7- + + + P NMR (162 MHz, H-decoupled, D2O) δ (ppm) -0.60 (P-C5); [m/z (ESI) [IP1S5 ·3H ·5NH4 ] 13 C6H29N5O24P1S5 calculated 745.9538, found 745.9541]. C2 and C5 were neither visible on the C NMR nor on the HSQC spectra.

6.1.3 Synthesis of IP2S4

The product IP2S4 was a white solid (sodium salt, 15.5 mg, assuming IP2S4·6Na MW: 814.24, 1 19 µmol, yield 86 %, over two steps). H NMR (400 MHz, D2O) δ (ppm) 5.47 (1H, bd, H-C2), 4.66

(2H, bd, H-C4/6), 4.58 (2H, bd, H-C1/3), 3.22 (36H, q, J 7.3 Hz, CH3CH2NH), 1.30 (54H, t, J 7.3 Hz, 13 3 CH3CH2NH), H-C5 under H-C1/3 and H-C4/6; C NMR (125 MHz, D2O) δ (ppm) 74.6 ( JCP 3.0 Hz, 2 31 C1/3), 72.7 (d, JCP 4.8 Hz, C4/6), 46.7 (CH3CH2NH), 8.2 (CH3CH2NH); P NMR (162 MHz, 1 8- + + + H-decoupled, D2O) δ (ppm) -0.84 (bd, P-C4/6); [m/z (ESI) [IP2S4 ·6H ·3NH4 ] C6H24N3O24P2S4 calculated 711.9102, found 711.9100]. (1) C2 and C5 were neither visible on the 13C NMR nor on the HSQC spectra.

APPENDIX

6.2 Figures

FIGURE 6.1. Crystal structure of the CPD of TcdB. A, Crystal structure of the CPD of TcdB displaying the bound IP6, the three active site residues Asp587, His653, and Cys698, as well as the β-flap involved in the allosteric circuit and the fluorescent Trp761 residue. B, Close-up of the CPD IP6 binding site with the position 1-3 of the inositol numbered in gray circles. PDB: 3PA8 (301).

FIGURE 6.2. Stability of IP2S4 in acidic medium. IP2S4 was incubated in D2O (600 µL) acidified 1 with D3PO4 to pH 2-3, and the H NMR was recorded at various time points.

APPENDIX

FIGURE 6.3. SDS-PAGE gels showing rCPD cleavage. Exemplary Coomassie-stained SDS-PAGE gels showing rCPD cleavage by IP6 analogs in absence of calcium.

APPENDIX

FIGURE 6.4. SDS-PAGE gels showing rCPD cleavage in the presence of calcium. Exemplary

Coomassie-stained SDS-PAGE gels showing rCPD cleavage by IP6 analogs in 10 mM CaCl2.

APPENDIX

FIGURE 6.5. Kinetics of TcdB activity loss. 150 ng TcdB in 100 mM tris pH 7.4 was incubated at 37°C for the indicated times prior to addition of 1 mM IP6 and a further incubation at 37°C for 1 h. Extent of cleavage was determined by band densitometry of silver-stained SDS-PAGE gels as described for holotoxin cleavage by IP6 analogs; Error bars show s.d., n = 3.

FIGURE 6.6. Cell Index curves. Inhibition of TcdB toxicity by IP6 and IP2S4 with and without 10 mM CaCl2 on 3T3-L1 fibroblasts. The experiment was interrupted after 2 h to image the wells. Cell index reached 0 within 4 h wherever TcdB was present. Error bars show s.d.; Duplicate measurements were repeated in triplicate.

APPENDIX

FIGURE 6.7. Toxicity of IP6 and IP2S4 on 3T3-L1 fibroblasts. Cells were seeded at a density of 2 x 103 cells per well in 90 μL on an E-plate View 16 and the cell index monitored using an xCelligence RTCA DP (ACEA Biosciences) impedance monitoring system. After growing for 24 h to sub- confluence, IP6 and IP2S4 was added to the culture medium to the concentrations indicated on the graph. Error bars show s.d.; n = 3.

7 References

1. Carroll KC, Bartlett JG (2011) Biology of Clostridium difficile: Implications for Epidemiology and Diagnosis. Annu Rev Microbiol:501–521.

2. Bartlett JG, Chang TW, Gurwith M, Gorbach SL, Onderdonk AB (1978) Antibiotic-associated pseudomembranous colitis due to toxin-producing clostridia. N Engl J Med 298:531–4.

3. Lance George W, Goldstein EC, Sutter V, Ludwig S, Finegold S (1978) Aetiology of antimicrobial-agent-associated colitis. Lancet 311:802–803.

4. Peery AF et al. (2012) Burden of gastrointestinal disease in the United States: 2012 update. Gastroenterology 143:1179–87.e1–3.

5. Wiegand PN et al. (2012) Clinical and economic burden of Clostridium difficile infection in Europe: a systematic review of healthcare-facility-acquired infection. J Hosp Infect 81:1–14.

6. Ananthakrishnan AN (2011) Clostridium difficile infection: epidemiology, risk factors and management. Nat Rev Gastroenterol Hepatol 8:17–26.

7. Bauer MP et al. (2010) Clostridium difficile infection in Europe: a hospital-based survey. Lancet 6736:1–5.

8. Chitnis AS et al. (2013) Epidemiology of community-associated Clostridium difficile infection, 2009 through 2011. JAMA Intern Med 173:1359–67.

9. Cartman ST, Heap JT, Kuehne S a, Cockayne A, Minton N (2010) The emergence of “hypervirulence” in Clostridium difficile. Int J Med Microbiol 300:387–395.

10. He M et al. (2012) Emergence and global spread of epidemic healthcare-associated Clostridium difficile. Nat Genet 45:109–13.

11. Burke KE, Lamont JT (2014) Clostridium difficile infection: a worldwide disease. Gut Liver 8:1–6.

12. Burnham C-AD, Carroll KC (2013) Diagnosis of Clostridium difficile Infection: an Ongoing Conundrum for Clinicians and for Clinical Laboratories. Clin Microbiol Rev 26:604–30.

13. Jones AM, Kuijper EJ, Wilcox MH (2013) Clostridium difficile: a European perspective. J Infect 66:115–28.

14. Tickler I a et al. (2014) Strain types and antimicrobial resistance patterns of Clostridium difficile isolates from the United States, 2011 to 2013. Antimicrob Agents Chemother 58:4214– 8.

15. Jank T, Aktories K (2008) Structure and mode of action of clostridial glucosylating toxins: the ABCD model. Trends Microbiol 16:222–9.

16. Cohen SH et al. (2010) Clinical practice guidelines for Clostridium difficile infection in adults: 2010 update by the society for healthcare epidemiology of America (SHEA) and the infectious diseases society of America (IDSA). Infect Control Hosp Epidemiol 31:431–55.

REFERENCES

17. Teasley DG et al. (1983) Prospective randomised trial of metronidazole versus vancomycin for Clostridium-difficile-associated diarrhoea and colitis. Lancet 2:1043–6.

18. Debast SB, Bauer MP, Kuijper EJ (2014) European Society of Clinical Microbiology and Infectious Diseases: update of the treatment guidance document for Clostridium difficile infection. Clin Microbiol Infect 20 Suppl 2:1–26.

19. Surawicz CM et al. (2013) Guidelines for diagnosis, treatment, and prevention of Clostridium difficile infections. Am J Gastroenterol 108:478–98; quiz 499.

20. Kedderis GL, Argenbright LS, Miwa GT (1989) Covalent interaction of 5-nitroimidazoles with DNA and protein in vitro: mechanism of reductive activation. Chem Res Toxicol 2:146–9.

21. Reynolds PE (1989) Structure, biochemistry and mechanism of action of glycopeptide antibiotics. Eur J Clin Microbiol Infect Dis 8:943–50.

22. Löfmark S, Edlund C, Nord CE (2010) Metronidazole is still the drug of choice for treatment of anaerobic infections. Clin Infect Dis 50 Suppl 1:S16–23.

23. Vardakas KZ et al. (2012) Treatment failure and recurrence of Clostridium difficile infection following treatment with vancomycin or metronidazole: a systematic review of the evidence. Int J Antimicrob Agents 40:1–8.

24. Freeman J, Baines SD, Saxton K, Wilcox MH (2007) Effect of metronidazole on growth and toxin production by epidemic Clostridium difficile PCR ribotypes 001 and 027 in a human gut model. J Antimicrob Chemother 60:83–91.

25. Al-Nassir WN et al. (2008) Both oral metronidazole and oral vancomycin promote persistent overgrowth of vancomycin-resistant enterococci during treatment of Clostridium difficile- associated disease. Antimicrob Agents Chemother 52:2403–6.

26. Lancaster JW, Matthews SJ (2012) Fidaxomicin: The Newest Addition to the Armamentarium Against Clostridium difficile Infections. Clin Ther 34:1–13.

27. Louie TJ, Emery J, Krulicki W, Byrne B, Mah M (2009) OPT-80 eliminates Clostridium difficile and is sparing of bacteroides species during treatment of C. difficile infection. Antimicrob Agents Chemother 53:261–3.

28. Tannock GW et al. (2010) A new macrocyclic antibiotic, fidaxomicin (OPT-80), causes less alteration to the bowel microbiota of Clostridium difficile-infected patients than does vancomycin. Microbiology 156:3354–9.

29. Louie TJ et al. (2011) Fidaxomicin versus vancomycin for Clostridium difficile infection. N Engl J Med 364:422–31.

30. Konijeti GG, Sauk J, Shrime MG, Gupta M, Ananthakrishnan AN (2014) Cost-effectiveness of competing strategies for management of recurrent Clostridium difficile infection: a decision analysis. Clin Infect Dis 58:1507–14.

31. Wagner M, Lavoie L, Goetghebeur M (2014) Clinical and economic consequences of vancomycin and fidaxomicin for the treatment of Clostridium difficile infection in Canada. Can J Infect Dis Med Microbiol 25:87–94.

REFERENCES

32. Bartsch SM, Umscheid C a, Fishman N, Lee BY (2013) Is fidaxomicin worth the cost? An economic analysis. Clin Infect Dis 57:555–61.

33. Locher HH et al. (2014) Investigations of the Mode of Action and Resistance Development of Cadazolid, a New Antibiotic for Treatment of Clostridium difficile Infections. Antimicrob Agents Chemother 58:901–8.

34. Mascio CTM et al. (2012) In vitro and in vivo characterization of CB-183,315, a novel lipopeptide antibiotic for treatment of Clostridium difficile. Antimicrob Agents Chemother 56:5023–30.

35. Chilton CH et al. (2014) In vitro activity of cadazolid against clinically relevant Clostridium difficile isolates and in an in vitro gut model of C. difficile infection. J Antimicrob Chemother 69:697–705.

36. Chilton CH et al. (2014) Efficacy of surotomycin in an in vitro gut model of Clostridium difficile infection. J Antimicrob Chemother 69:2426–33.

37. Vickers R., Teague J. TP, P.A. W (2014) SMT 19969 for Clostridium difficile infection (CDI): comparative efficacy compared to fidaxomicin and vancomycin in the hamster model of CDI. ECCMID, Barcelona, Spain.

38. Chang JY et al. (2008) Decreased diversity of the fecal Microbiome in recurrent Clostridium difficile-associated diarrhea. J Infect Dis 197:435–8.

39. Youngster I et al. (2014) Fecal microbiota transplant for relapsing Clostridium difficile infection using a frozen inoculum from unrelated donors: a randomized, open-label, controlled pilot study. Clin Infect Dis 58:1515–22.

40. Brandt LJ et al. (2012) Long-term follow-up of colonoscopic fecal microbiota transplant for recurrent Clostridium difficile infection. Am J Gastroenterol 107:1079–87.

41. Olle B (2013) Medicines from microbiota. Nat Biotechnol 31:309–15.

42. Van Nood E et al. (2013) Duodenal infusion of donor feces for recurrent Clostridium difficile. N Engl J Med 368:407–15.

43. Smith MB, Kelly C, Alm EJ (2014) Policy: How to regulate faecal transplants. Nature 506:290–1.

44. Ratner M (2014) Fecal transplantation poses dilemma for FDA. Nat Biotechnol 32:401–2.

45. Blumberg R, Powrie F (2012) Microbiota, disease, and back to health: a metastable journey. Sci Transl Med 4:137rv7.

46. Dethlefsen L, Huse S, Sogin ML, Relman D a (2008) The pervasive effects of an antibiotic on the human gut microbiota, as revealed by deep 16S rRNA sequencing. PLoS Biol 6:e280.

47. Beausoleil M et al. (2007) Effect of a fermented milk combining Lactobacillus acidophilus Cl1285 and Lactobacillus casei in the prevention of antibiotic-associated diarrhea: a randomized, double-blind, placebo-controlled trial. Can J Gastroenterol 21:732–6.

48. Atarashi K et al. (2011) Induction of colonic regulatory T cells by indigenous Clostridium species. Science 331:337–41. 92

REFERENCES

49. Miezeiewski M et al. (2014) An in vitro culture model to study the dynamics of colonic microbiota in Syrian golden hamsters and their susceptibility to infection with Clostridium difficile. ISME J:1–12.

50. Darkoh C, Brown EL, Kaplan HB, Dupont HL (2013) Bile salt inhibition of host cell damage by clostridium difficile toxins. PLoS One 8:e79631.

51. Howerton A, Patra M, Abel-Santos E (2013) A new strategy for the prevention of Clostridium difficile infection. J Infect Dis 207:1498–504.

52. Merrigan MM, Sambol SP, Johnson S, Gerding DN (2009) New approach to the management of Clostridium difficile infection: colonisation with non-toxigenic C. difficile during daily ampicillin or ceftriaxone administration. Int J Antimicrob Agents 33 Suppl 1:S46–50.

53. Shim JK, Johnson S, Samore MH, Bliss DZ, Gerding DN (1998) Primary symptomless colonisation by Clostridium difficile and decreased risk of subsequent diarrhoea. Lancet 351:633–6.

54. Brouwer MSM et al. (2013) Horizontal gene transfer converts non-toxigenic Clostridium difficile strains into toxin producers. Nat Commun 4:2601.

55. Hargreaves KR, Clokie MRJ (2014) Clostridium difficile phages: still difficult? Front Microbiol 5:184.

56. Pennisi E (2013) The CRISPR craze. Science 341:833–6.

57. Winau F, Winau R (2002) Emil von Behring and serum therapy. Microbes Infect 4:185–8.

58. Steele J et al. (2012) Systemic dissemination of Clostridium difficile toxins A and B is associated with severe, fatal disease in animal models. J Infect Dis 205:384–91.

59. Holmgren J, Czerkinsky C (2005) Mucosal immunity and vaccines. Nat Med 11:S45–53.

60. Westerman LE, McClure HM, Jiang B, Almond JW, Glass RI (2005) Serum IgG mediates mucosal immunity against rotavirus infection. Proc Natl Acad Sci U S A 102:7268–73.

61. Davies NL et al. (2013) A mixture of functionally oligoclonal humanized monoclonal antibodies that neutralize Clostridium difficile TcdA and TcdB with high levels of in vitro potency shows in vivo protection in a hamster infection model. Clin Vaccine Immunol 20:377– 90.

62. Roberts A et al. (2012) Development and evaluation of an ovine antibody-based platform for treatment of Clostridium difficile infection. Infect Immun 80:875–82.

63. Lowy I et al. (2010) Treatment with monoclonal antibodies against Clostridium difficile toxins. N Engl J Med 362:197–205.

64. Foglia G, Shah S, Luxemburger C, Pietrobon PJF (2012) Clostridium difficile: development of a novel candidate vaccine. Vaccine 30:4307–9.

65. Kyne L, Warny M, Qamar A, Kelly CP (2000) Asymptomatic carriage of Clostridium difficile and serum levels of IgG antibody against toxin A. N Engl J Med 342:390–7.

REFERENCES

66. Kyne L, Warny M, Qamar a, Kelly CP (2001) Association between antibody response to toxin A and protection against recurrent Clostridium difficile diarrhoea. Lancet 357:189–93.

67. Boraschi D et al. (2013) The gracefully aging immune system. Sci Transl Med 5:185ps8.

68. Permpoonpattana P et al. (2011) Immunization with Bacillus spores expressing toxin A peptide repeats protects against infection with Clostridium difficile strains producing toxins A and B. Infect Immun 79:2295–302.

69. Ducher A., Modess C., Weitschies W., Siegmund W., Dressman J., Augustin V., Feger C., Andremont A. de GJ (2014) DAV132, developed to prevent side effects of antibiotics in the gut flora: results of a pilot cross-over study in healthy volunteers. ECCMID, Barcelona, Spain.

70. Miossec C et al. (2014) DAV131 , an oral adsorbent-based product , exerts a dose-dependent protection of hamsters against moxifloxacin-induced Clostridium difficile lethal infection. ECCMID, Barcelona, Spain:2014.

71. Alouf JE (2006) in The Comprehensive Sourcebook of Bacterial Protein Toxins (Third Edition), eds Joseph EA, Michel RP (Academic Press, London), pp 3–21.

72. Cegelski L, Marshall GR, Eldridge GR, Hultgren SJ (2008) The biology and future prospects of antivirulence therapies. Nat Rev Microbiol 6:17–27.

73. Clatworthy AE, Pierson E, Hung DT (2007) Targeting virulence: a new paradigm for antimicrobial therapy. Nat Chem Biol 3:541–548.

74. Rasko DA, Sperandio V (2010) Anti-virulence strategies to combat bacteria-mediated disease. Nat Rev Drug Discov 9:117–128.

75. Burnett JC et al. (2006) Inhibition of Metalloprotease Botulinum Serotype A from a Pseudo- peptide Binding Mode to a Small Molecule That Is Active in Primary Neurons. J Biol Chem 282:5004–5014.

76. Shoop WL (2005) Anthrax lethal factor inhibition. Proc Natl Acad Sci 102:7958–7963.

77. Schiavo G, van der Goot FG (2001) The bacterial toxin toolkit. Nat Rev Mol Cell Biol 2:530– 537.

78. Denève C, Janoir C, Poilane I, Fantinato C, Collignon A (2009) New trends in Clostridium difficile virulence and pathogenesis. Int J Antimicrob Agents 33 Suppl 1:S24–8.

79. Rupnik M, Wilcox MH, Gerding DN (2009) Clostridium difficile infection: new developments in epidemiology and pathogenesis. Nat Rev Microbiol 7:526–36.

80. Gerding DN (2010) Global epidemiology of Clostridium difficile infection in 2010. Infect Control Hosp Epidemiol 31 Suppl 1:S32–4.

81. McDonald LC et al. (2005) An epidemic, toxin gene-variant strain of Clostridium difficile. N Engl J Med 353:2433–41.

82. Pepin J (2004) Clostridium difficile-associated diarrhea in a region of Quebec from 1991 to 2003: a changing pattern of disease severity. Can Med Assoc J 171:466–472.

REFERENCES

83. Bartlett JG (2006) Narrative Review: The New Epidemic of Clostridium difficile–Associated Enteric Disease. Ann Intern Med 145:758–764.

84. Sullivan NM, Pellett S, Wilkins TD (1982) Purification and characterization of toxins A and B of Clostridium difficile. Infect Immun 35:1032–40.

85. Schwan C et al. (2009) Clostridium difficile toxin CDT induces formation of microtubule- based protrusions and increases adherence of bacteria. PLoS Pathog 5:e1000626.

86. Barbieri JT, Riese MJ, Aktories K (2002) Bacterial toxins that modify the actin cytoskeleton. Annu Rev Cell Dev Biol 18:315–44.

87. Voth DE, Ballard JD (2005) Clostridium difficile Toxins : Mechanism of Action and Role in Disease. Society 18:247–263.

88. Lyras D et al. (2009) Toxin B is essential for virulence of Clostridium difficile. Nature 458:1176–9.

89. Kuehne S a., Cartman ST, Minton N (2011) Both, toxin A and toxin B, are important in Clostridium difficile infection. Gut Microbes 2:0–1.

90. Demarest SJ et al. (2005) Structural characterization of the cell wall binding domains of Clostridium difficile toxins A and B; evidence that Ca2+ plays a role in toxin A cell surface association. J Mol Biol 346:1197–206.

91. Na X, Kim H, Moyer MP, Pothoulakis C, LaMont JT (2008) gp96 is a human colonocyte plasma membrane binding protein for Clostridium difficile toxin A. Infect Immun 76:2862–71.

92. Krivan HC, Clark GF, Smith DF, Wilkins TD (1986) Cell surface binding site for Clostridium difficile enterotoxin: evidence for a glycoconjugate containing the sequence Gal alpha 1-3Gal beta 1-4GlcNAc. Infect Immun 53:573–81.

93. Pruitt RN, Chambers MG, Ng KK-S, Ohi MD, Lacy DB (2010) Structural organization of the functional domains of Clostridium difficile toxins A and B. Proc Natl Acad Sci U S A 107:13467–72.

94. Egerer M, Satchell KJF (2010) Inositol hexakisphosphate-induced autoprocessing of large bacterial protein toxins. PLoS Pathog 6:e1000942.

95. Pruitt RN et al. (2009) Structure-function analysis of inositol hexakisphosphate-induced autoprocessing in Clostridium difficile toxin A. J Biol Chem 284:21934–40.

96. Reineke J et al. (2007) Autocatalytic cleavage of Clostridium difficile toxin B. Nature 446:415–9.

97. Ghantoji SS, Sail K, Lairson DR, DuPont HL, Garey KW (2010) Economic healthcare costs of Clostridium difficile infection: a systematic review. J Hosp Infect 74:309–18.

98. Aslam S, Hamill RJ, Musher DM (2005) Treatment of Clostridium difficile-associated disease: old therapies and new strategies. Lancet Infect Dis 5:549–57.

99. Louie TJ, Miller M, Donskey C, Mullane K, Goldstein EJC (2009) Clinical outcomes, safety, and pharmacokinetics of OPT-80 in a phase 2 trial with patients with Clostridium difficile infection. Antimicrob Agents Chemother 53:223–8. 95

REFERENCES

100. Dolgin E (2011) “Game changer” antibiotic and others in works for superbug. Nat Med 17:10.

101. Pépin J, Valiquette L, Cossette B (2005) Mortality attributable to nosocomial Clostridium difficile-associated disease during an epidemic caused by a hypervirulent strain in Quebec. CMAJ 173:1037–42.

102. Sack D, Sack R, Nair G, Siddique A (2004) Cholera. Lancet 363:223–233.

103. Zuckerman JN, Rombo L, Fisch A (2007) The true burden and risk of cholera: implications for prevention and control. Lancet Infect Dis 7:521–530.

104. Zhang JS, Liu F, Huang L (2005) Implications of pharmacokinetic behavior of lipoplex for its inflammatory toxicity. Adv Drug Deliv Rev 57:689–698.

105. Haan L de, Hirst TR (2004) Cholera toxin: A paradigm for multi-functional engagement of cellular mechanisms (Review). Mol Membr Biol 21:77–92.

106. Sánchez J, Holmgren J (2008) Cholera toxin structure, gene regulation and pathophysiological and immunological aspects. Cell Mol Life Sci 65:1347–1360.

107. Zhang RG et al. (1995) The three-dimensional crystal structure of cholera toxin. J Mol Biol 251:563–573.

108. Merritt EA et al. (1998) The 1.25 Å resolution refinement of the cholera toxin B-pentamer: evidence of peptide backbone strain at the receptor-binding site. J Mol Biol 282:1043–1059.

109. Holmgren J, Lonnroth I, Mansson J, Svennerholm L (1975) Interaction of cholera toxin and membrane GM1 ganglioside of small intestine. Proc Natl Acad Sci U S A 72:2520–2524.

110. Holmgren J, Lonnroth I, Svennerholm L (1973) Tissue receptor for cholera exotoxin: postulated structure from studies with GM1 ganglioside and related glycolipids. Infect Immun 8:208–214.

111. Feng Y et al. (2004) Retrograde transport of cholera toxin from the plasma membrane to the endoplasmic reticulum requires the trans-Golgi network but not the Golgi apparatus in Exo2- treated cells. EMBO Rep 5:596–601.

112. Teter K, Holmes RK (2002) Inhibition of endoplasmic reticulum-associated degradation in CHO cells resistant to cholera toxin, Pseudomonas aeruginosa exotoxin A, and ricin. Infect Immun 70:6172–6179.

113. Taylor M et al. (2011) A Therapeutic Chemical Chaperone Inhibits Cholera Intoxication and Unfolding/Translocation of the Cholera Toxin A1 Subunit. PLoS One 6:e18825.

114. Lopez-Gigosos RM, Plaza E, Diez-Diaz RM, Calvo MJ (2011) Vaccination strategies to combat an infectious globe: oral cholera vaccines. J Glob Infect Dis 3:56–62.

115. Hirai K et al. (2010) Passive oral immunization by egg yolk immunoglobulin (IgY) to Vibrio cholerae effectively prevents cholera. Acta Med Okayama 64:163–170.

116. Croxen MA, Finlay BB (2010) Molecular mechanisms of Escherichia coli pathogenicity. Nat Rev Microbiol 8:26–38.

REFERENCES

117. Bielaszewska M et al. (2011) Characterisation of the Escherichia coli strain associated with an outbreak of haemolytic uraemic syndrome in Germany, 2011: a microbiological study. Lancet Infect Dis 11:671–676.

118. Bitzan M (2009) Treatment options for HUS secondary to Escherichia coli O157:H7. Kidney Int 75:S62–S66.

119. Wong CS, Jelacic S, Habeeb RL, Watkins SL, Tarr PI (2000) The risk of the hemolytic-uremic syndrome after antibiotic treatment of Escherichia coli O157:H7 infections. N Engl J Med 342:1930–1936.

120. Zhang X et al. (2000) Quinolone antibiotics induce Shiga toxin-encoding bacteriophages, toxin production, and death in mice. J Infect Dis 181:664–670.

121. Tarr P, Gordon C, Chandler W (2005) Shiga-toxin-producing Escherichia coli and haemolytic uraemic syndrome. Lancet 365:1073–1086.

122. Orth D et al. (2009) Shiga toxin activates complement and binds factor H: evidence for an active role of complement in hemolytic uremic syndrome. J Immunol 182:6394–6400.

123. Lapeyraque AL et al. (2011) Complement Blockade in Severe Shiga-Toxin-Associated HUS. N Engl J Med 364:2561–2563.

124. Fraser ME (2004) Structure of Shiga Toxin Type 2 (Stx2) from Escherichia coli O157:H7. J Biol Chem 279:27511–27517.

125. Jacewicz M, Clausen H, Nudelman E, Donohue-Rolfe A, Keusch GT (1986) Pathogenesis of shigella diarrhea. XI. Isolation of a shigella toxin-binding glycolipid from rabbit jejunum and HeLa cells and its identification as globotriaosylceramide. J Exp Med 163:1391–1404.

126. Sandvig K, van Deurs B (2000) Entry of ricin and Shiga toxin into cells: molecular mechanisms and medical perspectives. EMBO J 19:5943–5950.

127. Sandvig K et al. (1992) Retrograde Transport of Endocytosed Shiga Toxin to the Endoplasmic- Reticulum. Nature 358:510–512.

128. Garred O, van Deurs B, Sandvig K (1995) Furin-induced cleavage and activation of Shiga toxin. J Biol Chem 270:10817–10821.

129. Yu M, Haslam DB (2005) Shiga toxin is transported from the endoplasmic reticulum following interaction with the luminal chaperone HEDJ/ERdj3. Infect Immun 73:2524–2532.

130. Endo Y, Mitsui K, Motizuki M, Tsurugi K (1987) The mechanism of action of ricin and related toxic lectins on eukaryotic ribosomes. The site and the characteristics of the modification in 28 S ribosomal RNA caused by the toxins. J Biol Chem 262:5908–5912.

131. Endo Y et al. (1988) Site of action of a Vero toxin (VT2) from Escherichia coli O157:H7 and of Shiga toxin on eukaryotic ribosomes. Eur J Biochem 171:45–50.

132. Paton AW, Paton JC (2010) Escherichia coli Subtilase Cytotoxin. Toxins (Basel) 2:215–228.

133. Ling H et al. (1998) Structure of the shiga-like toxin I B-pentamer complexed with an analogue of its receptor Gb3. Biochemistry 37:1777–1788.

REFERENCES

134. Boerlin P et al. (1999) Associations between virulence factors of Shiga toxin-producing Escherichia coli and disease in humans. J Clin Microbiol 37:497–503.

135. Soltyk AM et al. (2002) A Mutational Analysis of the Globotriaosylceramide-binding Sites of Verotoxin VT1. J Biol Chem 277:5351–5359.

136. Shimizu H, Field RA, Homans SW, Donohue-Rolfe A (1998) Solution structure of the complex between the B-subunit homopentamer of verotoxin VT-1 from Escherichia coli and the trisaccharide moiety of globotriaosylceramide. Biochemistry 37:11078–11082.

137. Flagler MJ, Mahajan SS, Kulkarni AA, Iyer SS, Weiss AA (2010) Comparison of Binding Platforms Yields Insights into Receptor Binding Differences between Shiga Toxins 1 and 2. Biochemistry 49:1649–1657.

138. Ellis TN, Kuehn MJ (2010) Virulence and immunomodulatory roles of bacterial outer membrane vesicles. Microbiol Mol Biol Rev 74:81–94.

139. Sheoran AS et al. (2005) Human Antibody against Shiga Toxin 2 Administered to Piglets after the Onset of Diarrhea Due to Escherichia coli O157:H7 Prevents Fatal Systemic Complications. Infect Immun 73:4607–4613.

140. Dowling TC et al. (2005) Phase 1 Safety and Pharmacokinetic Study of Chimeric Murine- Human Monoclonal Antibody c Stx2 Administered Intravenously to Healthy Adult Volunteers. Antimicrob Agents Chemother 49:1808–1812.

141. Lopez EL et al. (2010) Safety and pharmacokinetics of urtoxazumab, a humanized monoclonal antibody, against Shiga-like toxin 2 in healthy adults and in pediatric patients infected with Shiga-like toxin-producing Escherichia coli. Antimicrob Agents Chemother 54:239–243.

142. Young JAT, Collier RJ (2007) Anthrax Toxin: Receptor Binding, Internalization, Pore Formation, and Translocation. Annu Rev Biochem 76:243–265.

143. Mock M, Fouet A (2001) Anthrax. Annu Rev Microbiol 55:647–671.

144. Inglesby T V (2002) Anthrax as a Biological Weapon, 2002: Updated Recommendations for Management. JAMA J Am Med Assoc 287:2236–2252.

145. Duesbery NS (1998) Proteolytic Inactivation of MAP-Kinase-Kinase by Anthrax Lethal Factor. Science (80- ) 280:734–737.

146. Leppla SH (1982) Anthrax toxin edema factor: a bacterial adenylate cyclase that increases cyclic AMP concentrations of eukaryotic cells. Proc Natl Acad Sci 79:3162–3166.

147. Drum CL et al. (2002) Structural basis for the activation of anthrax adenylyl cyclase exotoxin by calmodulin. Nature 415:396–402.

148. Van der Goot G, Young JAT (2009) Receptors of anthrax toxin and cell entry. Mol Aspects Med 30:406–412.

149. Buckle AM et al. (2010) A Femtomol Range FRET Biosensor Reports Exceedingly Low Levels of Cell Surface Furin: Implications for the Processing of Anthrax Protective Antigen. PLoS One 5:e11305.

REFERENCES

150. Remacle AG et al. (2008) Substrate Cleavage Analysis of Furin and Related Proprotein Convertases: A COMPARATIVE STUDY. J Biol Chem 283:20897–20906.

151. Abrami L et al. (2010) Endocytosis of the Anthrax Toxin Is Mediated by Clathrin, Actin and Unconventional Adaptors. PLoS Pathog 6:e1000792.

152. Moayeri M (2003) Bacillus anthracis lethal toxin induces TNF- -independent hypoxia-mediated toxicity in mice. J Clin Invest 112:670–682.

153. Leppla SH (2006) in The Comprehensive Sourcebook of Bacterial Protein Toxins (Third Edition), eds Joseph EA, Michel RP (Academic Press, London), pp 323–347.

154. Lacy DB, Wigelsworth DJ, Melnyk RA, Harrison SC, Collier RJ (2004) Structure of heptameric protective antigen bound to an anthrax toxin receptor: A role for receptor in pH- dependent pore formation. Proc Natl Acad Sci U S A 101:13147–13151.

155. Chen Z et al. (2009) Potent neutralization of anthrax edema toxin by a humanized monoclonal antibody that competes with calmodulin for edema factor binding. Proc Natl Acad Sci 106:13487–13492.

156. Migone T-S et al. (2009) Raxibacumab for the Treatment of Inhalational Anthrax. N Engl J Med 361:135–144.

157. Sharma S, Thomas D, Marlett J, Manchester M, Young JAT (2008) Efficient Neutralization of Antibody-Resistant Forms of Anthrax Toxin by a Soluble Receptor Decoy Inhibitor. Antimicrob Agents Chemother 53:1210–1212.

158. Jernigan DB et al. (2002) Investigation of bioterrorism-related anthrax, United States, 2001: epidemiologic findings. Emerg Infect Dis 8:1019–28.

159. Abramova FA, Grinberg LM, Yampolskaya O V, Walker DH (1993) Pathology of inhalational anthrax in 42 cases from the Sverdlovsk outbreak of 1979. Proc Natl Acad Sci 90:2291–2294.

160. Trachtman H (2003) Effect of an Oral Shiga Toxin-Binding Agent on Diarrhea-Associated Hemolytic Uremic Syndrome in Children: A Randomized Controlled Trial. JAMA 290:1337– 1344.

161. Weiss K (2009) Toxin-binding treatment for Clostridium difficile: a review including reports of studies with tolevamer. Int J Antimicrob Agents 33:4–7.

162. Armstrong GD, Fodor E, Vanmaele R (1991) Investigation of Shiga-like Toxin Binding to Chemically Synthesized Oligosaccharide Sequences. J Infect Dis 164:1160–1167.

163. Heerze LD, Kelm MA, Talbot JA, Armstrong GD (1994) Oligosaccharide sequences attached to an inert support (SYNSORB) as potential therapy for antibiotic-associated diarrhea and pseudomembranous colitis. J Infect Dis 169:1291–1296.

164. Castagliuolo I, LaMont JT, Qiu B, Nikulasson ST, Pothoulakis C (1996) A receptor decoy inhibits the enterotoxic effects of Clostridium difficile toxin A in rat ileum. Gastroenterology 111:433–8.

165. Dingle TC et al. (2008) Functional properties of the carboxy-terminal host cell-binding domains of the two toxins, TcdA and TcdB, expressed by Clostridium difficile. Glycobiology 18:698. 99

REFERENCES

166. Cone RA (2009) Barrier properties of mucus. Adv Drug Deliv Rev 61:75–85.

167. Lai SK, Wang Y-Y, Hanes J (2009) Mucus-penetrating nanoparticles for drug and gene delivery to mucosal tissues. Adv Drug Deliv Rev 61:158–71.

168. Simpson LS, Burdine L, Dutta AK, Feranchak AP, Kodadek T (2009) Selective toxin sequestrants for the treatment of bacterial infections. J Am Chem Soc 131:5760–5762.

169. Scobie HM et al. (2005) A soluble receptor decoy protects rats against anthrax lethal toxin challenge. J Infect Dis 192:1047–1051.

170. Cai C et al. (2011) Tumor endothelium marker-8 based decoys exhibit superiority over capillary morphogenesis protein-2 based decoys as anthrax toxin inhibitors. PLoS One 6:e20646.

171. Wycoff KL et al. (2011) Recombinant Anthrax Toxin Receptor-Fc Fusion Proteins Produced in Plants Protect Rabbits against Inhalational Anthrax. Antimicrob Agents Chemother 55:132– 139.

172. Kiick KL (2007) Materials science. Polymer therapeutics. Science (80- ) 317:1182–1183.

173. Bertrand N et al. (2011) New pharmaceutical applications for macromolecular binders. J Control Release 155:200–10.

174. Dhal PK, Holmes-Farley SR, Huval CC, Jozefiak TH (2006) Polymers as drugs. Adv Polym Sci 192:9–58.

175. Pinier M et al. (2009) Polymeric binders suppress gliadin-induced toxicity in the intestinal epithelium. Gastroenterology 136:288–298.

176. Kurtz CB et al. (2001) GT160-246, a toxin binding polymer for treatment of Clostridium difficile colitis. Antimicrob Agents Chemother 45:2340–7.

177. Braunlin W et al. (2004) Toxin binding of tolevamer, a polyanionic drug that protects against antibiotic-associated diarrhea. Biophys J 87:534–9.

178. Roy R, Tropper FD, Romanowska A (1992) New strategy in glycopolymer syntheses. Preparation of antigenic water-soluble poly(acrylamide-co-p-acrylamidophenyl .beta.- lactoside). Bioconjug Chem 3:256–261.

179. Watanabe M et al. (2004) Oral therapeutic agents with highly clustered globotriose for treatment of Shiga toxigenic Escherichia coli infections. J Infect Dis 189:360–368.

180. Gargano JM, Ngo T, Kim JY, Acheson DWK, Lees WJ (2001) Multivalent inhibition of AB(5) toxins. J Am Chem Soc 123:12909–12910.

181. Dohi H et al. (1999) Synthesis of an artificial glycoconjugate polymer carrying Pk-antigenic trisaccharide and its potent neutralization activity against Shiga-like toxin. Bioorg Med Chem 7:2053–2062.

182. Miura Y, Sasao Y, Dohi H, Nishida Y, Kobayashi K (2002) Self-assembled monolayers of globotriaosylceramide (Gb3) mimics: surface-specific affinity with shiga toxins. Anal Biochem 310:27–35.

100

REFERENCES

183. Karmali MA (2004) Prospects for preventing serious systemic toxemic complications of Shiga toxin-producing Escherichia coli infections using Shiga toxin receptor analogues. J Infect Dis 189:355–359.

184. Kitov PI et al. (2000) Shiga-like toxins are neutralized by tailored multivalent carbohydrate ligands. Nature 403:669–72.

185. Dubber M, Lindhorst TK (1998) Synthesis of octopus glycosides: core molecules for the construction of glycoclusters and carbohydrate-centered dendrimers. Carbohydr Res 310:35– 41.

186. Mulvey GL et al. (2003) Assessment in mice of the therapeutic potential of tailored, multivalent Shiga toxin carbohydrate ligands. J Infect Dis 187:640–649.

187. Nishikawa K et al. (2002) A therapeutic agent with oriented carbohydrates for treatment of infections by Shiga toxin-producing Escherichia coli O157:H7. Proc Natl Acad Sci 99:7669– 7674.

188. Nishikawa K et al. (2005) Identification of the optimal structure required for a Shiga toxin neutralizer with oriented carbohydrates to function in the circulation. J Infect Dis 191:2097– 2105.

189. Isobe H et al. (2007) Synthesis of fullerene glycoconjugates via a copper-catalyzed Huisgen cycloaddition reaction. Org Lett 9:4611–4614.

190. Chien YY et al. (2008) Globotriose-functionalized gold nanoparticles as multivalent probes for Shiga-like toxin. ChemBioChem 9:1100–1109.

191. Uzawa H, Ito H, Neri P, Mori H, Nishida Y (2007) Glycochips from polyanionic glycopolymers as tools for detecting Shiga toxins. ChemBioChem 8:2117–2124.

192. Thompson JP, Schengrund CL (1997) Oligosaccharide-derivatized dendrimers: defined multivalent inhibitors of the adherence of the cholera toxin B subunit and the heat labile enterotoxin of E. coli to GM1. Glycoconj J 14:837–845.

193. Branderhorst HM, Liskamp RM, Visser GM, Pieters RJ (2007) Strong inhibition of cholera toxin binding by galactose dendrimers. Chem Commun:5043–5045.

194. Sisu C et al. (2009) The influence of ligand valency on aggregation mechanisms for inhibiting bacterial toxins. ChemBioChem 10:329–337.

195. Arosio D et al. (2004) Synthesis and cholera toxin binding properties of multivalent GM1 mimics. Org Biomol Chem 2:2113–2124.

196. Fan EK et al. (2000) High-affinity pentavalent ligands of Escherichia coli heat-labile enterotoxin by modular structure-based design. J Am Chem Soc 122:2663–2664.

197. Parera Pera N et al. (2010) Rapid screening of lectins for multivalency effects with a glycodendrimer microarray. ChemBioChem 11:1896–1904.

198. Minke WE, Roach C, Hol WG, Verlinde CL (1999) Structure-based exploration of the ganglioside GM1 binding sites of Escherichia coli heat-labile enterotoxin and cholera toxin for the discovery of receptor antagonists. Biochemistry 38:5684–5692.

101

REFERENCES

199. Merritt EA et al. (2002) Characterization and crystal structure of a high-affinity pentavalent receptor-binding inhibitor for cholera toxin and E. coli heat-labile enterotoxin. J Am Chem Soc 124:8818–24.

200. Pickens JC et al. (2004) Nonspanning bivalent ligands as improved surface receptor binding inhibitors of the cholera toxin B pentamer. Chem Biol 11:1205–1215.

201. Tran HA, Kitov PI, Paszkiewicz E, Sadowska JM, Bundle DR (2011) Multifunctional multivalency: a focused library of polymeric cholera toxin antagonists. Org Biomol Chem 9:3658–3671.

202. Liu S, Kiick KL (2008) Architecture Effects on the Binding of Cholera Toxin by Helical Glycopolypeptides. Macromolecules 41:764–772.

203. Maheshwari R, Levenson EA, Kiick KL (2010) Manipulation of Electrostatic and Saccharide Linker Interactions in the Design of Efficient Glycopolypeptide-Based Cholera Toxin Inhibitors. Macromol Biosci 10:68–81.

204. Kimura T (2001) Serum Amyloid P Component Is the Shiga Toxin 2-neutralizing Factor in Human Blood. J Biol Chem 276:41576–41579.

205. Armstrong GD et al. (2006) Human serum amyloid P component protects against Escherichia coli O157:H7 Shiga toxin 2 in vivo: therapeutic implications for hemolytic-uremic syndrome. J Infect Dis 193:1120–1124.

206. Pepys MB et al. (2002) Targeted pharmacological depletion of serum amyloid P component for treatment of human amyloidosis. Nature 417:254–259.

207. Liu JY et al. (2005) Protein heterodimerization through ligand-bridged multivalent pre- organization: Enhancing ligand binding toward both protein targets. J Am Chem Soc 127:2044– 2045.

208. Solomon D et al. (2005) Heterobifunctional multivalent inhibitor-adaptor mediates specific aggregation between Shiga toxin and a pentraxin. Org Lett 7:4369–4372.

209. Kitov PI et al. (2008) An Entropically Efficient Supramolecular Inhibition Strategy for Shiga Toxins. Angew Chemie Int Ed 47:672–676.

210. Kitov PI et al. (2008) From the Cover: In vivo supramolecular templating enhances the activity of multivalent ligands: A potential therapeutic against the Escherichia coli O157 AB5 toxins. Proc Natl Acad Sci 105:16837–16842.

211. Griener TP et al. (2011) Lipopolysaccharide renders transgenic mice expressing human serum amyloid P component sensitive to Shiga toxin 2. PLoS One 6:e21457.

212. Nishikawa K et al. (2006) A multivalent peptide library approach identifies a novel Shiga toxin inhibitor that induces aberrant cellular transport of the toxin. FASEB J 20:2597–2599.

213. Watanabe-Takahashi M et al. (2010) An orally applicable Shiga toxin neutralizer functions in the intestine to inhibit the intracellular transport of the toxin. Infect Immun 78:177–183.

214. Stearns-Kurosawa DJ et al. (2011) Rescue from lethal Shiga toxin 2-induced renal failure with a cell-permeable peptide. Pediatr Nephrol 26:2031–9.

102

REFERENCES

215. Mourez M et al. (2001) Designing a polyvalent inhibitor of anthrax toxin. Nat Biotechnol 19:958–961.

216. Basha S et al. (2006) Polyvalent inhibitors of anthrax toxin that target host receptors. Proc Natl Acad Sci 103:13509–13513.

217. Sandvig K, van Deurs B (2005) Delivery into cells: lessons learned from plant and bacterial toxins. Gene Ther 12:865–872.

218. Stechmann B et al. (2010) Inhibition of Retrograde Transport Protects Mice from Lethal Ricin Challenge. Cell 141:231–242.

219. Sanchez AM et al. (2007) Amiodarone and Bepridil Inhibit Anthrax Toxin Entry into Host Cells. Antimicrob Agents Chemother 51:2403–2411.

220. Collier RJ (2009) Membrane translocation by anthrax toxin. Mol Aspects Med 30:413–422.

221. Karginov VA (2005) Blocking anthrax lethal toxin at the protective antigen channel by using structure-inspired drug design. Proc Natl Acad Sci 102:15075–15080.

222. Karginov VA et al. (2006) Search for Cyclodextrin-Based Inhibitors of Anthrax Toxins: Synthesis, Structural Features, and Relative Activities. Antimicrob Agents Chemother 50:3740– 3753.

223. Moayeri M, Robinson TM, Leppla SH, Karginov VA (2008) In Vivo Efficacy of - Cyclodextrin Derivatives against Anthrax Lethal Toxin. Antimicrob Agents Chemother 52:2239–2241.

224. Joshi A et al. (2011) Structure-Based Design of a Heptavalent Anthrax Toxin Inhibitor. Biomacromolecules 12:791–796.

225. Seidah N et al. (2008) The activation and physiological functions of the proprotein convertases. Int J Biochem Cell Biol 40:1111–1125.

226. Komiyama T et al. (2009) Inhibition of Furin/Proprotein Convertase-catalyzed Surface and Intracellular Processing by Small Molecules. J Biol Chem 284:15729–15738.

227. Thomas G (2002) Furin at the cutting edge: From protein traffic to embryogenesis and disease. Nat Rev Mol Cell Biol 3:753–766.

228. Fugère M, Day R (2005) Cutting back on pro-protein convertases: the latest approaches to pharmacological inhibition. Trends Pharmacol Sci 26:294–301.

229. Chrétien M, Seidah NG, Basak A, Mbikay M (2008) Proprotein convertases as therapeutic targets. Expert Opin Ther Targets 12:1289–1300.

230. Basak A (2005) Inhibitors of proprotein convertases. J Mol Med 83:844–855.

231. Henrich S et al. (2003) The crystal structure of the proprotein processing proteinase furin explains its stringent specificity. Nat Struct Biol 10:520–526.

232. Becker GL et al. (2010) Potent Inhibitors of Furin and Furin-like Proprotein Convertases Containing Decarboxylated P1 Arginine Mimetics. J Med Chem 53:1067–1075.

103

REFERENCES

233. Shiryaev SA et al. (2007) Targeting Host Cell Furin Proprotein Convertases as a Therapeutic Strategy against Bacterial Toxins and Viral Pathogens. J Biol Chem 282:20847–20853.

234. Remacle AG et al. (2010) Selective and potent furin inhibitors protect cells from anthrax without significant toxicity. Int J Biochem Cell Biol 42:987–995.

235. Egerer M, Satchell KJ (2010) Inositol hexakisphosphate-induced autoprocessing of large bacterial protein toxins. PLoS Pathog 6:e1000942.

236. Shen A et al. (2009) Mechanistic and structural insights into the proteolytic activation of Vibrio cholerae MARTX toxin. Nat Chem Biol 5:469–478.

237. Puri AW et al. (2010) Rational design of inhibitors and activity-based probes targeting Clostridium difficile virulence factor TcdB. Chem Biol 17:1201–11.

238. Savidge TC et al. (2011) Host S-nitrosylation inhibits clostridial small molecule-activated glucosylating toxins. Nat Med DOI: 10.10.

239. Ahuja N, Kumar P, Bhatnagar R (2004) The Adenylate Cyclase Toxins. Crit Rev Microbiol 30:187–196.

240. Gallagher DT et al. (2006) Structure of the Class IV Adenylyl Cyclase Reveals a Novel Fold. J Mol Biol 362:114–122.

241. Sunahara RK, Dessauer CW, Gilman AG (1996) Complexity and diversity of mammalian adenylyl cyclases. Annu Rev Pharmacol Toxicol 36:461–480.

242. Collier RJ, Young JAT (2003) Anthrax Toxin. Annu Rev Cell Dev Biol 19:45–70.

243. Shen Y et al. (2002) Physiological calcium concentrations regulate calmodulin binding and catalysis of adenylyl cyclase exotoxins. EMBO J 21:6721–6732.

244. Guo Q et al. (2005) Structural basis for the interaction of Bordetella pertussis adenylyl cyclase toxin with calmodulin. EMBO J 24:3190–3201.

245. Tang W-J, Guo Q (2009) The adenylyl cyclase activity of anthrax edema factor. Mol Aspects Med 30:423–430.

246. Chen D et al. (2008) Novel inhibitors of anthrax edema factor. Bioorg Med Chem 16:7225– 7233.

247. Laine E et al. (2010) Use of allostery to identify inhibitors of calmodulin-induced activation of Bacillus anthracis edema factor. Proc Natl Acad Sci 107:11277–11282.

248. Soelaiman S (2003) Structure-based Inhibitor Discovery against Adenylyl Cyclase Toxins from Pathogenic Bacteria That Cause Anthrax and Whooping Cough. J Biol Chem 278:25990– 25997.

249. Gottle M et al. (2007) Molecular Analysis of the Interaction of Bordetella pertussis Adenylyl Cyclase with Fluorescent Nucleotides. Mol Pharmacol 72:526–535.

250. Geduhn J et al. (2010) Bis-Halogen-Anthraniloyl-Substituted Nucleoside 5’-Triphosphates as Potent and Selective Inhibitors of Bordetella pertussis Adenylyl Cyclase Toxin. J Pharmacol Exp Ther 336:104–115. 104

REFERENCES

251. Burnett JC, Henchal EA, Schmaljohn AL, Bavari S (2005) The evolving field of biodefence: therapeutic developments and diagnostics. Nat Rev Drug Discov 4:281–296.

252. Masignani V, Pizza M, Rappuoli R (2006) in The Comprehensive Sourcebook of Bacterial Protein Toxins (Third Edition), eds Joseph EA, Michel RP (Academic Press, London), pp 213– 244.

253. Holbourn KP, Shone CC, Acharya KR (2006) REVIEW ARTICLE: A family of killer toxins. FEBS J 273:4579–4593.

254. Hottiger MO, Hassa PO, Lüscher B, Schüler H, Koch-Nolte F (2010) Toward a unified nomenclature for mammalian ADP-ribosyltransferases. Trends Biochem Sci 35:208–219.

255. Jørgensen R et al. (2008) Cholix Toxin, a Novel ADP-ribosylating Factor from Vibrio cholerae. J Biol Chem 283:10671–10678.

256. Yamaizumi M, Mekada E, Uchida T, Okada Y (1978) One molecule of diphtheria toxin fragment a introduced into a cell can kill the cell. Cell 15:245–250.

257. Yates SP et al. (2005) Structure-function analysis of water-soluble inhibitors of the catalytic domain of exotoxin A from Pseudomonas aeruginosa. Biochem J 385:667–675.

258. Turgeon Z et al. (2010) Newly Discovered and Characterized Antivirulence Compounds Inhibit Bacterial Mono-ADP-Ribosyltransferase Toxins. Antimicrob Agents Chemother 55:983–991.

259. Zhang G (2008) Design, synthesis, and evaluation of bisubstrate analog inhibitors of cholera toxin. Bioorg Med Chem Lett 18:3724–3727.

260. Lee Y-M et al. (2010) Conserved Structural Motif for Recognizing Nicotinamide Adenine Dinucleotide in Poly(ADP-ribose) Polymerases and ADP-Ribosylating Toxins: Implications for Structure-Based Drug Design. J Med Chem 53:4038–4049.

261. Maurer B et al. (2011) From cosubstrate similarity to inhibitor diversity-inhibitors of ADP- ribosyltransferases from kinase inhibitor screening. Mol Biosyst 7:799–808.

262. Hase CC, Finkelstein RA (1993) Bacterial extracellular zinc-containing metalloproteases. Microbiol Rev 57:823–837.

263. Miyoshi S, Shinoda S (2000) Microbial metalloproteases and pathogenesis. Microbes Infect 2:91–98.

264. Holton J (2008) Enterotoxigenic Bacteroides fragilis. Curr Infect Dis Rep 10:99–104.

265. Schiavo GG et al. (1992) Tetanus and botulinum-B neurotoxins block neurotransmitter release by proteolytic cleavage of synaptobrevin. Nature 359:832–835.

266. Brossier F, Weber-Levy M, Mock M, Sirard JC (2000) Role of toxin functional domains in anthrax pathogenesis. Infect Immun 68:1781–1786.

267. Turk BE (2008) Discovery and Development of Anthrax Lethal Factor Metalloproteinase Inhibitors. Curr Pharm Biotechnol 9:24–33.

105

REFERENCES

268. Roxas-Duncan V et al. (2009) Identification and Biochemical Characterization of Small- Molecule Inhibitors of Clostridium botulinum Neurotoxin Serotype A. Antimicrob Agents Chemother 53:3478–3486.

269. Li B et al. (2010) Synthesis and Biological Evaluation of Botulinum Neurotoxin A Protease Inhibitors. J Med Chem 53:2264–2276.

270. Turton K, Chaddock JA, Acharya KR (2002) Botulinum and tetanus neurotoxins: structure, function and therapeutic utility. Trends Biochem Sci 27:552–558.

271. Lacy DB, Tepp W, Cohen AC, DasGupta BR, Stevens RC (1998) Crystal structure of botulinum neurotoxin type A and implications for toxicity. Nat Struct Biol 5:898–902.

272. Montal M (2010) Botulinum Neurotoxin: A Marvel of Protein Design. Annu Rev Biochem 79:591–617.

273. Foran PG (2002) Evaluation of the Therapeutic Usefulness of Botulinum Neurotoxin B, C1, E, and F Compared with the Long Lasting Type A. J Biol Chem 278:1363–1371.

274. Poulain B, Stiles BG, Popoff MR, Molgo J (2006) in The Comprehensive Sourcebook of Bacterial Protein Toxins (Third Edition), eds Joseph EA, Michel RP (Academic Press, London), pp 348–389.

275. Xiong Y et al. (2006) The discovery of a potent and selective lethal factor inhibitor for adjunct therapy of anthrax infection. Bioorg Med Chem Lett 16:964–968.

276. Kim S et al. (2011) Antidotes to anthrax lethal factor intoxication. Part 2: Structural modifications leading to improved in vivo efficacy. Bioorg Med Chem Lett 21:2030–2033.

277. Tang J, Park JG, Millard CB, Schmidt JJ, Pang YP (2007) Computer-aided lead optimization: improved small-molecule inhibitor of the zinc endopeptidase of botulinum neurotoxin serotype A. PLoS One 2:e761.

278. Pang YP et al. (2010) Small molecules showing significant protection of mice against botulinum neurotoxin serotype A. PLoS One 5:e10129.

279. Johnson SL et al. (2008) Rhodanine Derivatives as Selective Protease Inhibitors Against Bacterial Toxins. Chem Biol Drug Des 71:131–139.

280. Forino M (2005) Efficient synthetic inhibitors of anthrax lethal factor. Proc Natl Acad Sci 102:9499–9504.

281. Baell JB, Holloway GA (2010) New Substructure Filters for Removal of Pan Assay Interference Compounds (PAINS) from Screening Libraries and for Their Exclusion in Bioassays. J Med Chem 53:2719–2740.

282. Carlson EE, May JF, Kiessling LL (2006) Chemical Probes of UDP-Galactopyranose Mutase. Chem Biol 13:825–837.

283. Voss ME et al. (2003) Both 5-arylidene-2-thioxodihydropyrimidine-4,6(1H,5H)-diones and 3- thioxo-2,3-dihydro-1H-imidazo[1,5-a]indol-1-ones are light-Dependent tumor necrosis factor- [alpha] antagonists. Bioorg Med Chem Lett 13:533–538.

106

REFERENCES

284. Zuniga JE et al. (2010) Iterative Structure-Based Peptide-Like Inhibitor Design against the Botulinum Neurotoxin Serotype A. PLoS One 5:e11378.

285. Zuniga JE et al. (2008) A Potent Peptidomimetic Inhibitor of Botulinum Neurotoxin Serotype A Has a Very Different Conformation than SNAP-25 Substrate. Structure 16:1588–1597.

286. Burnett JC et al. (2009) Pharmacophore-guided lead optimization: The rational design of a non- zinc coordinating, sub-micromolar inhibitor of the botulinum neurotoxin serotype a metalloprotease. Bioorg Med Chem Lett 19:5811–5813.

287. Hale M, Oyler G, Swaminathan S, Ahmed SA (2010) Basic Tetrapeptides as Potent Intracellular Inhibitors of Type A Botulinum Neurotoxin Protease Activity. J Biol Chem 286:1802–1811.

288. Sandvig K (2006) in The Comprehensive Sourcebook of Bacterial Protein Toxins (Third Edition), eds Joseph EA, Michel RP (Academic Press, London), pp 310–322.

289. Monzingo AF, Robertus JD (1992) X-ray analysis of substrate analogs in the ricin A-chain active site. J Mol Biol 227:1136–1145.

290. Endo Y, Tsurugi K (1988) The RNA N-glycosidase activity of ricin A-chain. The characteristics of the enzymatic activity of ricin A-chain with ribosomes and with rRNA. J Biol Chem 263:8735–8739.

291. Bai Y, Watt B, Wahome PG, Mantis NJ, Robertus JD (2010) Identification of new classes of ricin toxin inhibitors by virtual screening. Toxicon 56:526–534.

292. Miller DJ et al. (2002) Structure-based design and characterization of novel platforms for ricin and shiga toxin inhibition. J Med Chem 45:90–98.

293. Pang Y-P et al. (2011) Small-Molecule Inhibitor Leads of Ribosome-Inactivating Proteins Developed Using the Doorstop Approach. PLoS One 6:e17883.

294. Abdeen SJ, Swett RJ, Feig AL (2010) Peptide inhibitors targeting Clostridium difficile toxins A and B. ACS Chem Biol 5:1097–103.

295. Ivarsson ME, Leroux J-C, Castagner B (2012) Targeting Bacterial Toxins. Angew Chemie (International Ed 51:4024–4045.

296. Kuehne S a. et al. (2010) The role of toxin A and toxin B in Clostridium difficile infection. Nature 467:711–3.

297. Pfeifer G et al. (2003) Cellular uptake of Clostridium difficile toxin B. Translocation of the N- terminal catalytic domain into the cytosol of eukaryotic cells. J Biol Chem 278:44535–41.

298. Egerer M, Giesemann T, Herrmann C, Aktories K (2009) Autocatalytic processing of Clostridium difficile toxin B. Binding of inositol hexakisphosphate. J Biol Chem 284:3389–95.

299. Ng J et al. (2010) Clostridium difficile toxin-induced inflammation and intestinal injury are mediated by the inflammasome. Gastroenterology 139:542–52, 552.e1–3.

300. Just I et al. (1995) Glucosylation of Rho proteins by Clostridium difficile toxin B. Nature 375:500–3.

107

REFERENCES

301. Shen A et al. (2011) Defining an allosteric circuit in the cysteine protease domain of Clostridium difficile toxins. Nat Struct Mol Biol 18:364–71.

302. Shears SB (2001) Assessing the omnipotence of inositol hexakisphosphate. Cell Signal 13:151–8.

303. Schlemmer U, Frølich W, Prieto RM, Grases F (2009) Phytate in foods and significance for humans: food sources, intake, processing, bioavailability, protective role and analysis. Mol Nutr Food Res 53 Suppl 2:S330–75.

304. Veiga N et al. (2006) The behaviour of myo-inositol hexakisphosphate in the presence of magnesium(II) and calcium(II): protein-free soluble InsP6 is limited to 49 microM under cytosolic/nuclear conditions. J Inorg Biochem 100:1800–10.

305. Puri AW et al. (2010) Rational design of inhibitors and activity-based probes targeting Clostridium difficile virulence factor TcdB. Chem Biol 17:1201–11.

306. Zorn J a, Wells J a (2010) Turning enzymes ON with small molecules. Nat Chem Biol 6:179– 188.

307. Wolan DW, Zorn J a, Gray DC, Wells J a (2009) Small-molecule activators of a proenzyme. Science 326:853–8.

308. Slack E et al. (2009) Innate and adaptive immunity cooperate flexibly to maintain host- microbiota mutualism. Science 325:617–20.

309. Natividad JMM et al. (2013) Differential induction of antimicrobial REGIII by the intestinal microbiota and Bifidobacterium breve NCC2950. Appl Environ Microbiol 79:7745–54.

310. Natividad JMM et al. (2012) Commensal and probiotic bacteria influence intestinal barrier function and susceptibility to colitis in Nod1-/-; Nod2-/- mice. Inflamm Bowel Dis 18:1434–46.

311. Ohnishi ST (1978) Characterization of the murexide method: dual-wavelength spectrophotometry of cations under physiological conditions. Anal Biochem 85:165–79.

312. Mortz E, Krogh TN, Vorum H, Görg A (2001) Improved silver staining protocols for high sensitivity protein identification using matrix-assisted laser desorption/ionization-time of flight analysis. Proteomics 1:1359–63.

313. Watanabe Y, Komoda Y, Ebisuya K, Ozaki S (1990) An efficient phosphorylation method using a new phosphitylating agent, 2-diethylamino-1,3,2-benzodioxaphosphepane. Tetrahedron Lett 31:255–256.

314. Lee HW, Kishi Y (1985) Synthesis of mono- and unsymmetrical bis-orthoesters of scyllo- inositol. J Org Chem 50:4402–4404.

315. Guttenberg G et al. (2011) Inositol hexakisphosphate-dependent processing of Clostridium sordellii lethal toxin and Clostridium novyi alpha-toxin. J Biol Chem 286:14779–86.

316. Hutton ML, Mackin KE, Chakravorty A, Lyras D (2014) Small animal models for the study of Clostridium difficile disease pathogenesis. FEMS Microbiol Lett 352:140–9.

317. Lowy I et al. (2010) Treatment with monoclonal antibodies against Clostridium difficile toxins. N Engl J Med 362:197–205. 108

REFERENCES

318. Howerton A, Patra M, Abel-Santos E (2013) A new strategy for the prevention of Clostridium difficile infection. J Infect Dis 207:1498–504.

319. Natarajan M, Walk ST, Young VB, Aronoff DM (2013) A clinical and epidemiological review of non-toxigenic Clostridium difficile. Anaerobe 22:1–5.

320. Lupardus PJ, Shen A, Bogyo M, Garcia KC (2008) Small molecule-induced allosteric activation of the Vibrio cholerae RTX cysteine protease domain. Science 322:265–8.

321. Arnone A, Perutz MF (1974) Structure of inositol hexaphosphate–human deoxyhaemoglobin complex. Nature 249:34–36.

322. Donald RGK et al. (2013) A novel approach to generate a recombinant toxoid vaccine against Clostridium difficile. Microbiology 159:1254–66.

323. Kreimeyer I et al. (2011) Autoproteolytic cleavage mediates cytotoxicity of Clostridium difficile toxin A. Naunyn Schmiedebergs Arch Pharmacol 383:253–62.

324. Lanis JM, Hightower LD, Shen A, Ballard JD (2012) TcdB from hypervirulent Clostridium difficile exhibits increased efficiency of autoprocessing. Mol Microbiol 84:66–76.

325. Rupnik M, Pabst S, von Eichel-Streiber C, Urlaub H, Söling H-D (2005) Characterization of the cleavage site and function of resulting cleavage fragments after limited proteolysis of Clostridium difficile toxin B (TcdB) by host cells. Microbiology 151:199–208.

326. Egerer M, Giesemann T, Jank T, Satchell KJF, Aktories K (2007) Auto-catalytic cleavage of Clostridium difficile toxins A and B depends on cysteine protease activity. J Biol Chem 282:25314–21.

327. Chumbler NM et al. (2012) Clostridium difficile Toxin B Causes Epithelial Cell Necrosis through an Autoprocessing-Independent Mechanism. PLoS Pathog 8:e1003072.

328. Farrow M a et al. (2013) Clostridium difficile toxin B-induced necrosis is mediated by the host epithelial cell NADPH oxidase complex. Proc Natl Acad Sci U S A 110:18674–9.

329. Madan R, Jr W a P (2012) Immune responses to Clostridium difficile infection. Trends Mol Med 18:658–66.

330. Ho JGS, Greco A, Rupnik M, Ng KK-S (2005) Crystal structure of receptor-binding C-terminal repeats from Clostridium difficile toxin A. Proc Natl Acad Sci U S A 102:18373–8.

331. Lai SK et al. (2007) Rapid transport of large polymeric nanoparticles in fresh undiluted human mucus. Proc Natl Acad Sci 104:1482–1487.

332. Ulbrich K, Subr V (2010) Structural and chemical aspects of HPMA copolymers as drug carriers. Adv Drug Deliv Rev 62:150–66.

333. Greco A et al. (2006) Carbohydrate recognition by Clostridium difficile toxin A. Nat Struct Mol Biol 13:460–1.

334. Ulbrich K et al. (2000) Polymeric drugs based on conjugates of synthetic and natural macromolecules. I. Synthesis and physico-chemical characterisation. J Control Release 64:63– 79.

109

REFERENCES

335. Shunmugam R, Tew GN (2005) Efficient route to well-characterized homo, block, and statistical polymers containing terpyridine in the side chain. J Polym Sci Part A Polym Chem 43:5831–5843.

336. Eberhardt M, Mruk R, Zentel R, Théato P (2005) Synthesis of pentafluorophenyl(meth)acrylate polymers: New precursor polymers for the synthesis of multifunctional materials. Eur Polym J 41:1569–1575.

337. Dubacheva G V. et al. (2010) Unlimited growth of host–guest multilayer films based on functionalized neutral polymers. Soft Matter 6:3747.

338. Aamer KA, Tew GN (2007) RAFT polymerization of a novel activated ester monomer and conversion to a terpyridine-containing homopolymer. J Polym Sci Part A Polym Chem 45:5618–5625.

339. Yanjarappa MJ, Gujraty K V, Joshi A, Saraph A, Kane RS (2006) Synthesis of copolymers containing an active ester of methacrylic acid by RAFT: controlled molecular weight scaffolds for biofunctionalization. Biomacromolecules 7:1665–70.

340. Gibson MI, Fröhlich E, Klok H (2009) Postpolymerization modification of poly(pentafluorophenyl methacrylate): Synthesis of a diverse water-soluble polymer library. J Polym Sci Part A Polym Chem 47:4332–4345.

341. Limberg G et al. (1996) A New Assay For Sialyltransferases Using Fluorescein-Labelled Acceptors. Liebigs Ann 1996:1773–1784.

342. Chen G, Ning X, Park B, Boons G-J, Xu B (2009) Simple, clickable protocol for atomic force microscopy tip modification and its application for trace ricin detection by recognition imaging. Langmuir 25:2860–4.

343. Ratcliffe RM et al. (2004) Large-Scale Chemical and Enzymatic Synthesis of a Clostridium difficile Toxin A Binding Trisaccharide. Synthesis (Stuttg) 2004:2293–2296.

344. Koch CP et al. (2013) Exhaustive proteome mining for functional MHC-I ligands. ACS Chem Biol 8:1876–81.

345. Phage concentration calculator Available at: http://www.abdesignlabs.com/technical- resources/phage-calculator/ [Accessed July 31, 2014].

346. Tonikian R, Zhang Y, Boone C, Sidhu SS (2007) Identifying specificity profiles for peptide recognition modules from phage-displayed peptide libraries. Nat Protoc 2:1368–86.

347. Boyer C et al. (2009) Bioapplications of RAFT polymerization. Chem Rev 109:5402–36.

348. Theato P (2008) Synthesis of well-defined polymeric activated esters. J Polym Sci Part A Polym Chem 46:6677–6687.

349. Godwin A, Hartenstein M, Müller AHE, Brocchini S (2001) Narrow Molecular Weight Distribution Precursors for Polymer-Drug Conjugates We are grateful for funding from the EPSRC and the School of Pharmacy (A.G.) and Mr. Derek Marley for conducting the atomic absorption analysis. Angew Chem Int Ed Engl 40:594–597.

350. Wong SY, Putnam D (2007) Overcoming limiting side reactions associated with an NHS- activated precursor of polymethacrylamide-based polymers. Bioconjug Chem 18:970–82. 110

REFERENCES

351. Hong V, Presolski SI, Ma C, Finn MG (2009) Analysis and optimization of copper-catalyzed azide-alkyne cycloaddition for bioconjugation. Angew Chem Int Ed Engl 48:9879–83.

352. Allmeroth M et al. (2011) Modifying the Body Distribution of HPMA-Based Copolymers by Molecular Weight and Aggregate Formation. Biomacromolecules 12:2841–2849.

353. Torres J, Camorlinga-Ponce M, Muñoz O (1992) Sensitivity in culture of epithelial cells from rhesus monkey kidney and human colon carcinoma to toxins A and B from Clostridium difficile. Toxicon 30:419–26.

354. Letourneur O (2003) Molecular cloning, overexpression in Escherichia coli, and purification of 6×his-tagged C-terminal domain of Clostridium difficile toxins A and B. Protein Expr Purif 31:276–285.

355. Burger S et al. (2003) Expression of recombinant Clostridium difficile toxin A using the Bacillus megaterium system. Biochem Biophys Res Commun 307:584–588.

356. E. coli Codon Usage Analyzer 2.1 Available at: http://www.faculty.ucr.edu/~mmaduro/codonusage/usage.htm [Accessed July 9, 2014].

357. Zhang Z et al. (2014) Translocation domain mutations affecting cellular toxicity identify the Clostridium difficile toxin B pore. Proc Natl Acad Sci U S A 111:3721–6.

358. Zhang R, Monsma F (2010) Fluorescence-based thermal shift assays. Curr Opin Drug Discov Devel 13:389–402.

359. Zhang P, Ng K, Ling C-C (2010) Total synthesis of Le(A)-LacNAc pentasaccharide as a ligand for Clostridium difficile toxin A. Org Biomol Chem 8:128–36.

360. Zhang P et al. (2011) Unexpected structure of a C. difficile toxin A ligand necessitates an annotation correction in a popular screening library. Chem Commun (Camb) 47:12397–9.

361. Smith GP, Petrenko VA (1997) Phage Display. Chem Rev 97:391–410.

362. Fukuda MN (2006) Screening of peptide-displaying phage libraries to identify short peptides mimicking carbohydrates. Methods Enzymol 416:51–60.

363. Hatakeyama S et al. (2011) Targeted drug delivery to tumor vasculature by a carbohydrate mimetic peptide. Proc Natl Acad Sci U S A 108:19587–92.

364. Scott J, Smith G (1990) Searching for peptide ligands with an epitope library. Science (80- ) 249:386–390.

365. Stanger K et al. (2012) Allosteric peptides bind a caspase zymogen and mediate caspase tetramerization. Nat Chem Biol 8:655–60.

366. Vodnik M, Zager U, Strukelj B, Lunder M (2011) Phage display: selecting straws instead of a needle from a haystack. Molecules 16:790–817.

367. Von Eichel-Streiber C, Sauerborn M (1990) Clostridium difficile toxin A carries a C-terminal repetitive structure homologous to the carbohydrate binding region of streptococcal glycosyltransferases. Gene 96:107–13.

111

REFERENCES

368. Orth P et al. (2014) Mechanism of Action and Epitopes of Clostridium difficile Toxin B- neutralizing Antibody Bezlotoxumab Revealed by X-ray Crystallography. J Biol Chem 289:18008–21.

369. Frisch C, Gerhard R, Aktories K, Hofmann F, Just I (2003) The complete receptor-binding domain of Clostridium difficile toxin A is required for endocytosis. Biochem Biophys Res Commun 300:706–711.

370. Gerhard R, Frenzel E, Goy S, Olling A (2013) Cellular uptake of Clostridium difficile TcdA and truncated TcdA lacking the receptor binding domain. J Med Microbiol 62:1414–22.

371. Olling A et al. (2011) The Repetitive Oligopeptide Sequences Modulate Cytopathic Potency but Are Not Crucial for Cellular Uptake of Clostridium difficile Toxin A. PLoS One 6:e17623.

372. Zhang Y et al. (2013) A Segment of 97 Amino Acids within the Translocation Domain of Clostridium difficile Toxin B Is Essential for Toxicity. PLoS One 8:e58634.

373. Murase T et al. (2013) Structural Basis for Antibody Recognition in the Receptor-Binding Domains of Toxins A and B from Clostridium difficile. J Biol Chem:0–24.

374. Chabre YM, Roy R (2013) Multivalent glycoconjugate syntheses and applications using aromatic scaffolds. Chem Soc Rev 42:4657–708.

375. Dam TK, Roy R, Pagé D, Brewer CF (2002) Thermodynamic binding parameters of individual epitopes of multivalent carbohydrates to concanavalin a as determined by “reverse” isothermal titration microcalorimetry. Biochemistry 41:1359–63.

376. Dam TK, Roy R, Pagé D, Brewer CF (2002) Negative cooperativity associated with binding of multivalent carbohydrates to lectins. Thermodynamic analysis of the “multivalency effect”. Biochemistry 41:1351–8.

377. Pruitt RN, Chambers MG, Ng KK-S, Ohi MD, Lacy DB (2010) Structural organization of the functional domains of Clostridium difficile toxins A and B. Proc Natl Acad Sci U S A 107:13467–72.

378. Ting SRS, Chen G, Stenzel MH (2010) Synthesis of glycopolymers and their multivalent recognitions with lectins. Polym Chem 1:1392.

379. Kozlowski L (2012) Calculation of protein isoelectric point. Available at: http://isoelectric.ovh.org [Accessed July 1, 2013].

380. Lamboy J a et al. (2008) Chemical and genetic wrappers for improved phage and RNA display. Chembiochem 9:2846–52.

381. Lamboy J a et al. (2009) Phage wrapping with cationic polymers eliminates nonspecific binding between M13 phage and high pI target proteins. J Am Chem Soc 131:16454–60.

382. Veiga N et al. (2014) Coordination, microprotonation equilibria and conformational changes of myo-inositol hexakisphosphate with pertinence to its biological function. Dalton Trans 43:16238–51.

383. Sandberg A-S, Andlid T (2002) Phytogenic and microbial phytases in human nutrition. Int J Food Sci Technol 37:823–833.

112

REFERENCES

384. Oakley AJ (2010) The structure of Aspergillus niger phytase PhyA in complex with a phytate mimetic. Biochem Biophys Res Commun 397:745–9.

385. Iqbal TH, Lewis KO, Cooper BT (1994) Phytase activity in the human and rat small intestine. Gut 35:1233–6.

386. Chen Q-C, Li BW (2003) Separation of phytic acid and other related inositol phosphates by high-performance ion chromatography and its applications. J Chromatogr A 1018:41–52.

387. Olling A et al. (2014) The Combined Repetitive Oligopeptides of Clostridium difficile Toxin A Counteract Premature Cleavage of the Glucosyl-Transferase Domain by Stabilizing Protein Conformation. Toxins (Basel) 6:2162–2176.

388. Steinhilber D et al. (2011) Hyperbranched polyglycerols on the nanometer and micrometer scale. Biomaterials 32:1311–6.

113

List of Abbreviations

AC adenylyl cyclase

ADP adenosinediphosphate

ADPRT ADP-ribosyltransferase

BoNT botulinum neurotoxin

CaM calmodulin

CDI C. difficile infection

CDT C. difficile transferase

CMG2 capillary morphogenesis protein 2

CNT clostridial neurotoxin

CPD cysteine protease domain

CT cholera toxin

CyaA adenylate cyclase toxin

DT diphtheria toxin eEF2 eukaryotic elongation factor 2

EF edema factor

EHEC enterohemorrhagic E. coli

ET edema toxin

ETEC enterotoxigenic E. coli

GI gastro-intestinal

GT glucosyltransferase

HPMA hydroxypropyl methacrylamide

HUS hemolytic uremic syndrome i.p. intraperitoneal

IP6 myo-inositol hexakisphosphate i.v. intravenous

114

LIST OF ABBREVIATIONS

LeF lethal factor

LeT lethal toxin

LT heat-labile toxin

MKK mitogen-activated protein kinase kinase

MNPG m-nitrophenyl-α-D-galactoside

MPO myeloperoxidase

NMR nuclear magnetic resonance

PA protective antigen

PARP poly(ADP-ribose) polymerase

PC proprotein convertase

RBD receptor binding domain

RT ricin toxin

SAP serum amyloid P component

SLT Shiga-like toxin

Stx Shiga toxin

TcdA C. difficile toxinA

TcdB C. difficile toxin B

TEM8 tumor endothelial marker 8

TeNT tetanus neurotoxin

TGN trans-Golgi network

115

Curriculum Vitae

EDUCATION 2010 - Oct 2014 Swiss Federal Institute of Technology (ETHZ) Zurich, Switzerland Ph.D. thesis: “Toxin Inhibitors for the Treatment of Clostridium difficile Infection” Institute of Pharmaceutical Sciences, Department of Chemistry and Applied Biosciences 2008 - 2010 Swiss Federal Institute of Technology (ETHZ) Zurich, Switzerland M.Sc. Biomedical Engineering (GPA: 5.5/6.0) 2007 - 2008 Cornell University Ithaca, NY, USA B.Sc. Exchange Year: Biomedical and Environmental Engineering 2005 - 2008 Swiss Federal Institute of Technology (EPFL) Lausanne, Switzerland B.Sc. Life Sciences and Technologies (GPA: 5.2/6.0) 1996 - 2004 International School of Geneva Founex, Switzerland Bilingual International Baccalaureate (English/French) (GPA: 43/45)

EXPERIENCE Nov 2010 - Oct 2014 Swiss Federal Institute of Technology (ETHZ) Zurich, Switzerland Research and Teaching Assistant . Developed and patented an innovative new molecular entity for the treatment of Clostridium difficile infection . Teaching assistant for 3rd year Pharmacy students . Supervised three Master’s thesis projects

Jul - Sep 2010 Siemens Management Consulting Munich, Germany Temporary Junior Consultant . Supported the strategic development of a newly-acquired industry software business unit . Developed a comprehensive Excel tool to model financial evolution of a business unit . Participated in developing a business case for a new partner program

May - Jun 2010 World Health Organization Geneva, Switzerland Intern . Produced material to be publicized for the 1st “Call for Innovative Technologies” . Analyzed data from the “Baseline Country Survey on Medical Devices” for publication

Jun - Aug 2009 McKinsey & Company Zurich, Switzerland Intern . Worked on a strategic development plan for a leading company in the specialty chemicals sector . Conducted research and analyses including product portfolio assessment, evaluation of possible M&A partners and auditing of financial statements during due diligence

2006 - 2007 Laboratory of Neural Microcircuitry, EPFL Lausanne, Switzerland 116

CURRICULUM VITAE

Student Assistant . Made computerized 3D reconstructions of rat brain neurons for the Blue Brain Project using Neuroleucida

2006 - 2007 Life Sciences IT Help Desk, EPFL Lausanne, Switzerland Student Assistant . Provided IT support for the Life Sciences department for Mac OS X and Windows users LANGUAGES French Fluent (mother tongue) English Fluent Swedish Fluent German C1 Polish A2

Croatian A1 AWARDS . Venture Leaders 2014, VentureLab & Swiss Commission for Technology and Innovation (Business development training course in Boston & New York for top 20 promising entrepreneurs in Switzerland, worth CHF 10’000) . EMBO travel grant, The EMBO Meeting, Amsterdam, Netherlands 2013 . 3rd best presentation: Global Pharmaceutical Education Network Conference 2012, Melbourne, Australia . Venture Kick 2012, winner Stage I (CHF 10’000 prize for a brilliant business idea) . Best Presentation: Swiss Galenic Meeting 2012 . ETHZ Excellence Scholarship and Opportunity Program 2008 (Merit-based prize awarded to 30 Masters students every year)

∼ EXTRACURRICULAR ACTIVITIES . President of the Pharmaceutical Scientist’s Association (PSA) at ETHZ (2013 - 2014) . ETH Zurich study trip to Stanford University (Nov 2013) . President of the Biomedical Engineering ETH Zurich (BEEZ) student association (2008 - 2009) . Student League of Nations participant (2002) . Saxophone (www.lyingeight.ch), golf, football, skiing, snowboarding

PUBLICATIONS Ivarsson M. E., Leroux J.-C., Castagner B., “Targeting Bacterial Toxins”, Angew. Chem., Int. Ed. 2012, 51, 4024

Ivarsson M. E. ; Durantie E., Hüberli C., Huwiler S., Verdu E., Bercik P., Leroux J.-C., Castagner B., “Therapeutic Potential of Triggering Pre-Emptive Clostridium difficile Toxin B Auto-Proteolysis”, submitted.

Castagner, B., Leroux, J.-C., Ivarsson M. E., Schneider G., Pratsinis, A. “Pharmaceutical Compounds for Clostridium difficile Infection.” International Patent Application: PCT/EP2012/004088

CONFERENCES “Toxin-targeted polymeric binders for the treatment of Clostridium difficile infection” Globalization of Pharmaceutics Education Network Meeting, Melbourne, Australia Nov. 28- Dec. 1, 2012. (Presentation)

117

CURRICULUM VITAE

“Licensing Opportunity: Small Molecules for the Treatment of Clostridium difficile Infection” BIO Europe, Hamburg, Germany, Nov. 13, 2012. (Presentation)

“Polymeric Binders for the Treatment of Clostridium difficile Infection” PharmTox Day; Zurich, Switzerland Oct. 4, 2012. (Poster)

“Alternatives to antibiotics for the treatment of Clostridium difficile infection” 4th International Clostridium difficile Symposium, Bled, Slovenia, Sep. 20-22, 2012. (Poster)

“Non-Antibiotic Therapeutic Approaches for the Treatment of Clostridium difficile Infection” ETH Industry Day, Zurich, Switzerland, Sep. 7, 2012. (Poster)

“Toxin-Targeted Allosteric Triggers for the Treatment of Clostridium difficile Infection”, Novartis Day ETH Zurich, June 2013, Zurich, Switzerland. (Poster)

“Rationally Designed Allosteric Activators of Clostridium difficile Toxin B Auto-Proteolysis”, Swiss Pharma Day, August 2013, Bern, Switzerland. (Poster)

“Small-Molecule Activators of Clostridium difficile Toxin B Auto-Proteolysis”, The EMBO Meeting, September 2013, Amsterdam, Netherlands. (Presentation & Poster)

118

Acknowledgements

The research presented in this thesis was financially enabled by an ETH grant (ETH-08 10-3) and a Swiss National Science Foundation grant (205321_141263), for both of which we are very grateful.

My gratitude goes out to all the people that have supported me during my time as a doctoral student. In particular, I would like to thank Prof. Jean-Christophe Leroux, who provided me with an exceptional environment in which to carry out my doctoral research. The balance between giving me the freedom to pursue my endeavors and providing advice to maintain a high degree of focus was very conducive to ensuring that my output was of the highest caliber that I could deliver.

I am most grateful to Prof. Bastien Castagner for always taking the time to listen to me, to fine- tune ideas, and for generously sharing his insights. I would also like to thank Bastien for his unwavering trust and honesty, which made for seamless communication and exchange of ideas between us. It has been a true privelege and a pleasure for me to get to know and to work with Bastien.

Many thanks to Prof. Wolf-Dietrich Hardt for agreeing to take the time to be my co-examiner.

I would like to express my gratitude for Estelle Durantie’s contribution to the work presented in Chapter 3. Her extensive efforts in synthesizing IP6 analogs were key in enabling us to make the findings presented.

Thanks to Prof. Premysl Bercik, Prof. Elena Verdu and Jun Lu, our collaborators at the Farncombe Family Digestive Health Research Institute at McMaster in Ontario, Canada, for generously carrying out the in vivo work described in Chapter 3.

I am indebted to my Master students Corina Hüberli, Simon Bussmann and Nina Romantini for helping me to execute my research. Thanks also to Prof. Dario Neri and Prof. Michael Detmar, and their respective group members for the use of their instruments.

I have also been lucky to rub shoulders with exceptional colleagues throughout my PhD. The discussions we have had and the time spent with you inside and outside of the lab have been incredibly enriching and have always made me feel like I was in the right place. Thanks to Vincent Forster and Arnaud Felber for their amazing support right from the start; to Lorine Brülisauer, Elena Moroz and Maurizio Roveri for being awesome office-mates; and to Prof. Marc Gauthier, Dr. Paola Luciani, Dr. Davide Brambilla, Dr. Ander Estella, Jessica Schulz, Anna Połomska, Nancy Dasargyri, Diana Andina, Virginie Schmid and Monica Langfritz for their help, advice and friendship.

I would like to thank my parents and sister for always being patient and optimistic throughout my entire studies. Finally, I am grateful to my wife, Marija, for being who she is and for always believing in me.

119