Development of chemical protein glycosylation, multivalent carbohydrates, and folded peptide-PNA catalyst

Thesis

Reference

Development of chemical protein glycosylation, multivalent carbohydrates, and folded peptide-PNA catalyst

MACHIDA, Takuya

Abstract

This Ph.D. thesis is composed by three chapters based on different interests in chemical biology. The first chapter describes my work on the chemical glycosylation of proteins by cell compatible biorthogonal ligation. Current existing methods for the chemical glycosylation of proteins are powerful, however they are limited for purified proteins. Also, multiple synthetic steps are required to functionalize glycans of interest. The author has developed new synthetic methods to functionalize a series of native bioactive oligosaccharides in 1-2 steps.

These neo glycans have enabled the glycosylation of proteins and other biomolecules in the ways that are applicable in live cells, which are tetrazine-strain alkene/alkyne cycloaddition and chelation assisted CuAAC. The second chapter describes my work on inhibition of multivalent interactions between glycans and pathogenic bacterial lectin. While multivalent interactions have been successfully blocked by glycopolymer and glycodendrimer approaches, a structurally well-designed ligand is required to enhance the potency and selectivity. A PNA encoded heteroglycan library was [...]

MACHIDA, Takuya. Development of chemical protein glycosylation, multivalent

carbohydrates, and folded peptide-PNA catalyst. Thèse de doctorat : Univ. Genève, 2017, no. Sc. 5118

DOI : 10.13097/archive-ouverte/unige:97072 URN : urn:nbn:ch:unige-970725

Available at:

http://archive-ouverte.unige.ch/unige:97072

Disclaimer: layout of this document may differ from the published version.

1 / 1

1

Development of chemical protein

glycosylation, multivalent carbohydrates, and folded peptide-PNA catalyst

THÈSE

présentée à la Faculté des sciences de l’Université de Genève pour obtenir le grade de Docteur ès sciences, mention chimie

par

Takuya Machida

de

Saitama (Japon)

Thèse N° 5118

GENÈVE

Centre d’impression de l’Université de Genève, ReproMail, Unimail 2017

UNIVERSITE DE GENEVE Section de chimie et biochimie Département de chimie organique

FACULTE DES SCIENCES

Professeur Nicolas Winssinger

2

3

Acknowledgement

First of all, I would like to give my special thanks to my mentor Prof. Nicolas Winssinger.

Thank you for giving me the great opportunities to work for Ph.D study here in Geneva. It was life-changing experience and I’m glad to learn your passion, intelligence and philosophy for both science and life.

I would like to thanks to Dr. Sofia Barluenga for your wisdoms and great discussions. Your logical thinking is very straightforward and always leads to solve problem. It was great pleasure to learn a lot from you.

Thanks for Patrick Romanens, Mireille Heimendinger and Sonia Candolfi for all the support for the research and administrative. I was quite naive for Swiss life. Thanks to all of your helps, I could focus to my work.

For the Winssinger’s group member, I would like to give special thanks to Dr. Som Dutt, Dr.

Alexandre Novoa, Lluc Farrera Soler and Dalu Chang to share the project with me and work together. It was my great pleasure.

I would like to thank all the former and current member of Winssinger’s group (Kumar, Gaomai, Daniela, Claudio, Shenal, Kalyan, Pierre, Christelle, Marcello, Subrata, Eric, Jacques, Roman, Sarah, Igor, Arunava, Bala, Jessica, Simona, Rémi, Ki Tae, Manuel,

Miryam). I enjoyed a lot spending time together with you both personally and professionally.

In terms of project, I would like to thank Dr. Christoph Bauer, Jérôme Bosset, Prof. Jean Gruenberg, Dr. Dimitri Moreau, Dr. Cameron Scott, Marion Pupier, Dr. Miwa Umebayashi, Prof. Jason W. Chin, Prof. Kathrin Lang, Dr. Lin Xue, Émilie Gillon, Shuangshuang Zheng,

4

Dr. Julie Claudinon, Dr. Thorsten Eierhoff, Dr. Anne Imberty, Prof. Winfried Römer for all the collaboration and kind help for the critical data in my Ph.D thesis.

I would like to thank to all of my friends in university of Geneva. Thanks to your kind friendship, my life in Geneva got excited.

At the end, I would like to give my heartfelt thanks to my family (Osamu, Yoshie, Ryu-ichi, Kazuhiko) and dedicate my entire work to you and me.

5

Summary

This Ph.D. thesis is composed by three chapters based on different interests in chemical biology.

The first chapter describes my work on the chemical glycosylation of proteins by cell

compatible biorthogonal ligation. Current existing methods for the chemical glycosylation of proteins are powerful, however they are limited for purified proteins. Also, multiple synthetic steps are required to functionalize glycans of interest. The author has developed new

synthetic methods to functionalize a series of native bioactive oligosaccharides in 1-2 steps.

These neo glycans have enabled the glycosylation of proteins and other biomolecules in the ways that are applicable in live cells, which are tetrazine-strain alkene/alkyne cycloaddition and chelation assisted CuAAC.

The second chapter describes my work on inhibition of multivalent interactions between glycans and pathogenic bacterial lectin. While multivalent interactions have been

successfully blocked by glycopolymer and glycodendrimer approaches, a structurally well- designed ligand is required to enhance the potency and selectivity. A PNA encoded

heteroglycan library was investigated to discover various sets of divalent glycan inhibitors for corresponding pathogenic lectins and DNA display showed structural information of ligands for preferential lectins. Furthermore, the interaction was also inhibited by PNA-programed dynamic glycan assembly. The dynamic glycan assembly was stabilized only when

multivalent lectin is exists and showed nanomolar affinity to pathogenic lectin BambL. The assembly also successfully blocked the invasion of BambL to human lung epithelial cell H1299 in 723 fold effectively compared with monomeric glycan.

6

The third chapter describes my work on the development of constrained peptide catalysts folded through PNA hybridization. Folded short peptide catalysts have great potential to imitate enzyme active site and perform enzyme-like functions and ground breaking peptide catalysts have been reported to perform the diverse stereo and regio selective reactions with rate acceleration. The 1,000 member folded peptide-PNA library was screened to observe enhanced catalytic activity for phosphate bond hydrolysis. Identified sequence had >25 fold increased activity when the peptide structure was folded through PNA hybridization and the activity was moderated by strand displacement. Furthermore, to apply this concept for more broad chemistry, metallo peptide-PNA catalyst was designed and synthesized to hydrolyse metal chelating ester.

7

Résumé

Cette thèse est composée de trois chapitres adressant different sujets à l’interface chimie- biologie.

Le premier chapitre concerne la glycosilation chimique de proteine par ligation bioorthogonal de cellule compatible. Les méthodes actuelles pour la glycosilation chimique de protéines sont efficaces bien que limitées aux protéines purifiées. Des synthèses multi-étapes sont nécessaires pour fonctionnaliser les glycanes d’intérêt. Mon travail a porté sur le

developpment de nouvelles méthodes synthétiques pour fonctionnaliser des séries

d’oligosaccharides bioactifs en une à deux étapes. Ces nouveaux glycoconjugs sont capables de glycosiler la protéine d’intérêt ou une biomolécule au sein d’un système applicable aux cellules vivantes, qui implique une cycloaddition entre une ‘tetrazine et un alcène/alcyne contraint via une CuAAC assisté par une chelation.

Le second chapitre concerne l’inhibition d’interaction multivalente entre un glycane et une lectine bactérienne pathogène. Tandis que les interactions multivalentes ont été bloquées avec succès via l’approche des glycopolymer et des glycodendrimères, un design structurel de ligand est nécessaire dans le but d’améliorer l’affinité et la sélectivité. Des librairies de ‘PNA encoded heteroglycan’ ont été utilisé pour découvrir différentes séries d’inhibiteurs glycanes divalents correspondants aux lectines pathogènes et l’ADN montre l’information structurale des ligands pour les lectines préférentielles. De plus, l’interaction fût aussi inhibée par l’assemblage ‘PNA-programmed dynamic glycan’. L’assemblage de glycanes dynamique fût stabilisé seulement lorsque de la lectine multivalente existe et montre une affinité

8

nanomolaire sur la lectine BambL pathogène. L’assemblage bloque également l’invasion de BambL sur les cellules épithéliales H1299 723 fois mieux comparé aux glycanes

monomériques.

Le troisième chapitre concerne un peptide jouant le rôle de catalyseur plié par le biais d’une hybridation de PNA. Les catalyseurs peptidiques courts et pliés possèdent un grand potentiel pour imiter les sites actifs des enzymes et jouer le rôle d’enzyme-like et rompt avec les catalyseurs peptidiques publiés dans la littérature pour réaliser diverses réactions stéréo et régiosélective avec une cinétique plus élevée. La librairie de 1000 folded peptides-PNA a été screené pour observer l’activité catalytique d’hydrolyse de liaison phosphate. Des séquences identifiées ont augmenté l’activité de plus de 25 fois lorsque la structure du peptide a été plié par le biais d’une hybridation de PNA et l’activité a été modéré par ‘strand displacement’. De plus, dans le but d’appliquer ce concept dans des domaines plus large de la chimie, des catalyseurs PNA metallopeptides ont été designé et synthétisé pour hydrolyser l’ester chelatant le métal.

9

Table of Contents

1. Chemical glycosylation of protein by cell-compatible bioorthogonal ligation ... 11

1.1 Introduction ... 11

1.2 Results and discussion-protein glycocojugation by tetrazine ligation ... 16

1.3 Results and discussion-protein glycoconjugation by chelation assisted CuAAC ligation ... 26

1.4 Conclusion and future aspect ... 36

2. Inhibition of multivalent interaction between glycan and pathogenic bacterial lectin ... 37

2.1 Introduction ... 37

2.2 Results and discussion-PNA encoded heteroglycoconjugate library ... 39

2.3 Results and discussion-PNA programmed dynamic cooperative glycan assembly ... 44

2.4 Conclusion and future aspect ... 52

3. Short peptide catalyst folded through PNA hybridization ... 53

3.1 Introduction ... 53

3.2 Results and discussion-phosphatase activity screening ... 56

3.3 Result and discussion-metallo peptide-PNA catalyst for expanding scope of reaction by mass spectrum based screening. ... 66

3.4. Conclusion and future aspect ... 72

4. Supporting information ... 73

4.1 Supporting information for chapter 1 ... 73

10

4.2 Supporting information for chapter 2 ... 171

4.3 Supporting information for chapter 3 ... 202

List of abbreviation ... 242

References ... 244

11

1. Chemical glycosylation of protein by cell-compatible bioorthogonal ligation

1.1 Introduction

Figure 1. Protein-glycan interaction involved in cellular recognition for important biological events.

Protein glycosylation is an important post-translational modification in living systems and the glycans of glycoproteins have been associated with numerous biological processes¹. For instance, cell surface glycoproteins are important in cellular recognition and have been

12

implicated in embryonic development, lymphocyte trafficking, and cancer metastasis²

Figure 2.Schematic representation of chemoselective glycoconjugation. a: Approach of thioether linkage from Davis and co-workers b: Approach of triazole linkage from Berti and co-workers. c:

Approach of oxime linkage from Bertozzi and co-workers.

(Figure 1), so that there is resurging interest in glycoconjugates as therapeutics³. Therefore, there is strong interest in accessing homogeneous glycoproteins and several methods have beenreported for the post translation of glycans by chemical ligation⁴. Davis and co-workers

13 pioneered to development powerful protein

Figure 3.a: Second order rate constant of various bioorthogonal reaction.⁵ b: equation for half-life of second order reaction or pseudo first order reaction.

14

glycosylation methods utilizing cysteine residues on proteins to form thioether linkages (Figure 2a)⁶. Berti and co-workers reported that tyrosine was selectively labelled with an alkyne moiety which is efficiently conjugated with azide-derivatized glycans via CuAAC (Figure 2b). Bertozzi and co-workers reported a chemoenzymatic method to convert a cysteine residue to an aldehyde which leads to the formation of a stable oxime (Figure 2c)⁷. Despite the robustness of these methodologies, they all present the main disadvantage that the chemistry scope is limited to purified proteins due to poor reaction kinetics or toxic

conditions for living cells. To understand the role of glycans at the molecular level for a phenotype of interest, a method which would allow direct conjugation of glycans to proteins in cell compatible conditions would be a highly attractive tool^2c.

The term ‘bioorthogonal chemistry’ was first raised by Bertozzi in 2003⁸. This concept makes reference to a chemical reaction which occurs in living systems without interfering with biochemical contents. To fulfill this condition, the reaction requires stable reactants in water, fast kinetics, inert conditions to live cells or animals, and the lack of unwanted side reactions with biomolecules. Staudinger ligation was first reported by Bertozzi as

‘bioorthogonal chemistry’ because of the fact that both the azide and phosphine as a reactants are abiotic functional groups, so the reaction is completely orthogonal to biological contents.

This reaction was applied to modify cell surface proteins⁹. Since then, with increasing demand of labelling biomolecules in living organisms (for visualizing proteins^{5, 10}, lipids¹¹, glycans¹², and nucleic acids¹³, for instance), diverse bioorthogonal reactions with different orthogonal partners have been recently developed^{5, 14}.

To choose bioorthogonal reactions, second order rate constants act as important indicators in calculating the efficiency of conversion (Figure 3)^{5, 14}. In many cases, the concentration of target biomolecules in living organisms is not easy to be determined and excess labelling reagents are used to perform the reaction in a pseudo first order manner. Those rate constants

15

help to understand ideal reaction conditions such as reaction time, conversion and

concentration, to achieve the ideal outcome. Among the various bioorthogonal reactions, inverse electron demanding Diels-Alder reactions between 1,2,4,5-tetrazines and strained alkene or alkynes is distinguished for exhibiting a remarkable rate constant in a completely additive-free environment.

Genetic code expansion technology opens the window to apply the tetrazine ligation reaction for labelling of proteins in live cells through the incorporation of tetrazine or strained alkyne bearing unnatural amino acids to proteins (Figure 4)¹⁵. Herein the author has developed new methodologies to conjugate glycans to proteins through live cell compatible ligations.

Figure 4. a: Unnatural amino acid incorporation to protein by orthogonal amino acyl-tRNA synthase and tRNACUA pair for amber codon suppression. b: Structures of unnatural amino acid incorporated to protein for tetrazine ligation.

16

1.2 Results and discussion-protein glycoconjugation by tetrazine ligation

This project was carried out in collaboration with Dr. Kathrin Lang et al.¹⁶

Figure 5. a: Tetrazine-glycan conjugate. b: Shoda’s activation to modify native oligosaccharides.

Initially, the functionalization of glycans for application in tetrazine ligation was investigated (Figure 5a). Glycans have several hydroxyl groups within the same molecule at different positions and with specific stereochemistry so the main approach to modify one position of the glycan selectively is to carry out systematic protection-deprotection of hydroxyl groups.

But ideally, the glycan should be converted from the native oligosaccharide to the tetrazine adduct via simple transformation since most of bioactive oligosaccharides are precious and therefore not tolerant of multistep synthesis (Figure 5a).

17 Scheme 1.Proposed mechanism of Shoda’s activation.

Shoda and co-workers pioneered the activation of the anomeric hydroxyl group of glycans without interfering with other hydroxyl groups by utilizing the small difference in pKa of the hemiacetal hydroxyl group (Figure 5b)¹⁷.

Mild bases such as triethylamine were used to increase the electron negativity of the anomeric hydroxyl group for reaction with 2-Chloro 1,3-dimethylimidazolinium chloride (DMC), converting it to a leaving group. Several nucleophiles (sodium azide,

mercaptopyridine) can be substituted at that position to modify the glycan. This method can be applicable for complex oligosaccharide containing sialic acid.

From the proposed mechanism, the stereochemistry of the anomeric position is highly reliant on the stereochemistry of the 2’ hydroxyl group (Scheme 1). The anomer which has anti- configuration at the 2’ position is suppressed by forming 1,2-anhydride to increase selectivity. The intramolecular 1,6-anhydride formation is a considerable side reaction so high concentration of nucleophile is required to outcompete this problem¹⁸.

18 Scheme 2.Synthesis of alkyne-bearing tetrazine 1.3.

Since glycosyl azide is easily accessible, copper catalysed alkyne-azide cycloaddition

(CuAAC) can be employed to introduce the tetrazine moiety to the glycan. Substitution of the 3- and 6- position of the tetrazine dramatically changes the kinetics of ligation as well as the stability of the tetrazine in serum and various substituents have already been reported with corresponding second order rate constants¹⁹. Followed by Devaraj’s protocol²⁰, formamidine was chosen as an amidine partner and 4-cyanophenyl acetic acid 1.1 was converted to tetrazine 1.2 which has a proton at 6-substitution and has decent reactivity and stability.

Propargyl amine was coupled to the acid to obtain alkyne 1.3 (Scheme 2).

Next, the compatibility of CuAAC between glycosyl azide and alkyne 1.3 was evaluated (Figure 6). Based on the potential reduction of tetrazine under the excess sodium ascorbate and copper, copper iodide (I) was first chosen. The reaction proceeded smoothly with glucosyl azide affording the desired product in 81% isolated yield. However, applying the same conditions to disaccharides proved more problematic with significant formation of side product arising from oxidative degradation (iodo derivative b and homodimer c)²¹. On the other hand, investigation of CuSO4/sodium ascorbate conditions revealed that controlling the molar equivalents of the reagents was critical to suppress the tetrazine reduction and

condition C was found to yield the desired product without tetrazine reduction and oxidative degradation.

19

Figure 6. Optimization of CuAAC conjugation. Condition A: CuI (200 mM), TBTA (10 mM) in DMF. Condition B: CuSO4 (250 mM), TBTA (2.5 mM), NaAsc (250 mM), DMF/t-BuOH/H2O = 1/1/1, 0 °C, 5 h. Condition C: CuSO4 (12 mM), TBTA (12 mM), NaAsc (120 mM), DMF/t- BuOH/H2O = 1/1/1, 0 °C, 5 h.

Next, the compatibility of optimized chemistry for a panel of glycans was investigated (Figure 7). Glycosyl azides were prepared from native carbohydrates by following Shoda’s protocol with consistent yield and stereoselectivity (details in the supporting

information)^17a. The reaction was performed in deuterium oxide to monitor the progress of reaction by NMR. The CuAAC reaction according to the optimized procedure afforded the desired product in 50-80% isolated yield for mono and disaccharides after HPLC purification (in the supporting information). This procedure was compatible with low glycan

concentration (10-30 mM) for complex oligosaccharides like hexa-saccharide or sialic acid and still afforded useful isolated yields after HPLC purification.

20

Since substitution of the 3’- and 6’-positions of the tetrazine effect reactivity, second- order rate constant of ligation between the glycan-tetrazine conjugate and trans-cyclooctene (TCO) Figure 7. Scope of glycans for CuAAC conjugation of alkyne-tetrazine (concentration for second step). 1) 2,6-lutidine was used as base instead of Et3N.

21

was assessed (Figure 8a, supporting information). It has been shown that tetrazine chromophore acts as both a fluorescence quencher and a bioorthogonal reactant and is therefore utilized as a turn-on probe²². Cy3 labelled glucosamine-tetrazine conjugate 1.4 was prepared to measure the turn-on fluorescence for kinetic studies and a second order rate constant of 8649 M^-1s^-1 was calculated which is consistent with previous analysis¹⁹. Figure 8.Evaluation of glycan-tetrazine conjugate in vitro. a: Turn-on measurement of cy3

fluorescence for calculation of second order rate constant for ligation. b: Glycoconjugation on purified sfGFP with TCO-lysine or Boc-Lys. c: SDS-PAGE analysis of reactions after 12 h. d: MALDI-TOF mass analysis of the reaction from lane 1.

22

Figure 9.Glycoconjugatin of sfGFP-BCNK with panel of glycan-tetrazine conjugate.

23

Next, site-selective glycoconjugation to purified protein was evaluated (Figure 8b). sfGFP- TCOK and sfGFP-BocK were prepared (both proteins were prepared by incorporation of an unnatural amino acid bearing TCO and Boc by MbPylRS/tRNACUA pair into sfGFP

overexpressed in E. coli)^15b at 13.5 M and were incubated with 1.4 (10 eq.) for 12 h to test the selectivity and the stability. SDS-PAGE analysis showed a strong fluorescence band corresponding to the conjugate of sfGFP-TCOK with 1.4 but not in the other controls (Figure 8c). Mass spectrum analysis of the crude mixture of lane 1 by MALDI-TOF showed a

complete conversion with a mass gain corresponding to the single cycloaddition product and N2 extrusion which illustrates the high chemoselectivity of the reaction and the stability of the product (Figure 8d).

Next, the glycoconjugation with genetically encoded bicyclononyne²³ protein (sfGFP- BCNK, prepared according to the same procedure as sfGFP-TCOK)^15b was investigated to further probe the reactivity and scope of the glycan-tetrazine conjugate (Figure 9). Tetrazine conjugation with bicyclononyne was reported to be 10-15 times slower than with TCO^15b. The conjugation of sfGFP-BCNK with 1.4 at the same concentration as used in sfGFP-TCOK afforded the desired glycoconjugate after 10 min (entry 1, the reaction was quenched with 100 equivalent of TCO before mass spectrum analysis). Reducing protein concentration to 1

M or 100 nM with the corresponding glycan equivalent still afforded the desired product (entry 2-4). At 100 nM concentration of sfGFP-BCNK (entry 4), traces of starting material are still present after 10 min indicating that these conditions approach the limit of reactivity.

BCN is also known to undergo cycloaddition with azide at a slower rate than with tetrazine²³. To compare the reactivity of the two conjugation methods, Cy3 labelled glycosyl azide 1.7 was investigated for ligation with sfGFP-BCNK under the same forceful conditions as used in entry 1. After 10 min, the reaction was quenched with tetrazine 1.3 yielding the quenched product without notable glycoconjugation. Extension of the reaction to 3 h afforded ca. 30%

24

of the conjugation product which clearly showed benefit of reactivity between glycan- tetrazine conjugate and glycosyl azide.

Finally, the specificity of the glycoconjugation was investigated in E coli expressing sfGFP- TCOK (Figure 10). Cells were pelleted, washed with PBS to remove excess TCOK present in the medium, and incubated with 1.4 for 8 h. After the reaction, pellet was lysed and crude mixture was analysed by SDS-PAGE showing that single protein migrating at the molecular weight of GFP underwent conjugation (lane 1 for Cy3 fluorescence), whereas the control reaction showed no conjugation (lane 2-4). The data demonstrated the highly specificity of conjugation by developed glycan-tetrazine conjugate in cellular contents.

25

Figure 10. SDS-PAGE analysis of glycoconjugation of 1.4 in E coli pellet expressing sfGFP-TCOK or sfGFP-BocK.

26

1.3 Results and discussion-protein glycoconjugation by chelation assisted CuAAC ligation

Figure 11. Chelation assisted CuAAC as alternative strategy for glycoconjugation²⁴.

Although the glycan-tetrazine conjugate successfully functionalized protein, there are several drawbacks that remain (Figure 11). 1) The synthetic protocol involves 2 steps. A single step procedure would be preferable for precious complex oligosaccharide. 2) The modified moiety

is big which might structurally change the effect of the glycan. 3) The method to label strained alkenes or alkynes is limited. CuAAC stands out for the simplicity of the reactive partners but the high concentration of copper required is a limitation for use as a biorthogonal

reaction due to the toxicity and oxidative degradation of biomolecules. Recently copper chelating azide was reported as an alternative approach to reduce the copper concentration

27

significantly for the cycloaddition²⁵. The system developed by Ting and co-workers is based on a picolyl azide substrate and was shown to be compatible with

Scheme 3. Synthesis of thiopicolyl azide 1.8.

highly sensitive neuronal cell lines. More recently a related technique was used in live vertebrates²⁶. Because alkyne-tagged biomolecules are readily available through metabolic labelling, a carbohydrates functionalized with picolyl azide could be an alternative approach for live cell compatible glycoconjugation. Moreover, Shoda’s protocol which used 2-

thiopyridine as the nucleophile for direct functionalization of glycans^17b arose the idea that thiopicolyl azide 1.8 might be an appropriate substrate to access the functionalized glycan in a single step.

The synthesis of 1.8 was investigated (Scheme 3). Commercially available alcohol 1.9 was protected with TBS group. The bromide was exchanged to lithium and the afforded

organolithium intermediate was reacted with sulfur to introduce a mercapto group which was then protected with MOM group in one-pot to obtain compound 1.10. The TBS group was de-protected with TBAF and the resulting alcohol was transformed to azide by MsCl/NaN3 to obtain compound 1.11. Since the azide group is sensitive to reductive conditions, MOM group was deprotected under acidic conditions to prevent deprotonation of the thiol and

28

freshly prepared product 1.8 was immediately used for the subsequent investigations without further purification.

Figure 12. a: Optimization of direct functionalization of glucose. b: Possible mechanistic detail of glycosylation for thiopyridine.

Next, direct functionalization of glucose by prepared compound 1.8 was investigated (Figure 12a). Firstly, 3 equivalents of both freshly prepared 1.8 and DMC were used, however no desired product was detected (entry 1). Large excess of 1.8 was required to obtain the desired product but the yield was still poor and the reaction was not reproducible (entry 2). It was rationalized by major oxidation of the deprotonated thiol and reduction of the

29

azide moiety. To improve the yield, the possible mechanism of reaction between thiopyridine

Figure 13.Panel of glycan for thiopycolylazide conjugation.

and glycose in shoda’s report^17b was proposed (Figure 12b). Considering the nucleophilicity of the thiol and alcohol starting materials, it is reasonable that thiol would first react with

30

DMC to form thioimidazolinium intermediate (i). The glucose would then react with this intermediate to activate the anomeric position and release the free thiol (ii) which reacts with activated glucose to form thioglycoside. Notably, the thioimidazolinium intermediate (i) is still reactive and able to activate the glucose. The imidazolinium protected thiopicolyazide 1.12 was expected to be a possible solution which avoids major oxidation of thiol but retains enough reactivity to facilitate glucose activation and release the active thiol for continuous substitution. The compound 1.12 was prepared from reaction with 1.8 and DMC and a stable crystal was obtained after crystallization by ether. Gratifyingly, 3 equivalents of 1.12 were sufficient to activate the glucose and provide the desired adduct in 88% yield. Hünig’s base was proved to be effective for better conversion¹⁸.

Once the reaction was optimized, a panel of glycans were submitted to this protocol to

demonstrate the compatibility and scope of the reaction (Error! Reference source not found.).

After the reaction, the mixture was treated with 1M NaOH after reaction to decompose the excess reagent for purification purposes and cleave any thioester formed from the carboxylic acid on sialic acid. In terms of 2’-N-acylglucosamine as a glycan source, acidification of the reaction mixture by 1 M HCl was required to enable the reaction between thiol and oxazoline formed during the reaction. The optimized conditions turned out to be compatible with a range of glycans

including a complex oligosaccharide containing sialic acid.

The reactivity of glycothiopicolyl azide 1.13 at low copper concentrations was investigated (Figure 14, in the supporting information). The reaction was performed with 7-

alkynylcoumarin²⁷ 1.15, which becomes fluorescent upon cycloaddition and thus provides a simple method to monitor the reaction kinetics. In the presence of 10 M of copper/THPTA the reaction between 20 M of 1.13 and 1.15 had a second order rate constant of 193 M^-1s^-1.

31

Figure 14.Second order rate constant measurement of glycoconjugation and comparison with glycosyl azide 1.14.

This is ca. 200 times faster than the rates reported for the fastest copper-free cycloaddition with strained alkynes²⁸. Under the same reaction conditions, glucosyl azide 1.14 did not afford measurable reaction. In the presence of 1 mM copper/THPTA (50 equivalent), 1.14 afforded 50% conversion after 5 h.

Next, the enhanced reactivity of the prepared functional glycan was validated (Figure 15).

MUC1 core peptide domain²⁹ containing a single propargly glycine residue was treated at 10

M with three different thiopicolyl azide-functionalized glycans. After 1 h, each reaction was

32

quenched with a large excess of propargyl alcohol (1000 equivalents) and analysed by MALDI-TOF mass. In all cases the reactions reached completion within 1 h (MALDI-TOF analysis caused fragmentation at C-N bond next to triazole, in the supporting information).

Similarly, an oligonucleotide substrate (DNA: CAE GCG CTA TGA CTC G, E: 5’-

ethynyluridine) bearing a single alkyne-modified nucleobase was investigated (Figure 16a).

There is increasing interest in combining DNA and glycans to create antigen-like structures of glycoproteins for therapeutic applications by taking advantage of the PCR based powerful screening method³⁰. But in many cases nucleic acids cause problems for CuAAC conjugation probably due to the strong chelation of copper to nucleobases. However, prepared

glycosylthiopycolyazides were successfully conjugated to DNA with lower concentration of copper at ambient temperature in 50-70% conversion analysed by PAGE (Figure 16c). The band for the product was isolated and the product was extracted. The afforded product was analysed by MALDI-TOF mass to give the desired mass of product (Figure 16b).

Finally, the compatibility of glycothiopicolyl azide conjugation in live cells was investigated (Figure 17). Hela cells were incubated with homopropargylglycine (HPG) to incorporate an alkyne into protein biosynthesis³¹. It has been shown that incorporation of HPG into proteins does not affect cell viability. After removal of the excess HPG by washing, cells were treated with Cy3-derivatized glycothiopicolyl azide 1.18 (Cy3 is not sulphated so its adduct is cell permeable)³² or Cy3-derivatized glycosyl azide 1.7, and the levels of glycoconjugation were assessed by fluorescence microscopy. As a control the same experiment was performed omitting the treatment with HPG. A bright fluorescence is observed only when the cells were metabolically labelled with HPG and 1.18 was used as a glycan source supporting that developed glycoconjugation was compatible with live cells.

33

Figure 15. Conjugation of diverse oligosaccharides to MUC1 core peptide domain. Condition:

peptide (10 M), glycan (100 M), CuSO4 (50 M), THPTA (250 M), NaAsc (2.5 mM), H2O, r.t., 1 h, then quenched with 100 mM propargyl alcohol.

G

V T S A P D R P

A P G S T A P P A H

G

V T S A P D R P

A P G S T A P P A H

G

V T S A P D R

P A

P G S T A P P A H

G

V T S A P D R P

A P G S T A P P A H

G

V T S A P D R P

A P G S T A P P A H

34

Figure 16. a: Conjugation of diverse oligosaccharides to DNA containing ethynyluridine. Condition:

DNA 10 M, glycan 100 M, CuSO4 50 M, THPTA 250 M, NaAsc 500 M, H2O, r.t., 3 h. b:

panel of glycan and MALDI-TOF mass analysis of purified DNA-glycan conjugate. c: PAGE analysis of DNA glycosylation.

35

Figure 17. Glycoconjugation in live cells. Cells were incubated with methionine-free medium supplemented with HPG for 4 h (A and C) or without HPG (B). Cells were then treated with 1.18 (A and B) or 1.7 (B) (20 M), as well as CuSO4 (50 M), THPTA (250 M), NaAsc (2.5 mM) in DPBS for 5 min.

36

1.4 Conclusion and future aspects

In this chapter, two new approaches of chemical protein glycosylation were successfully developed; 1) tetrazine ligation and 2) chelation assisted CuAAC ligation. In both cases, a functionalized glycan was efficiently prepared in 1-2 steps starting from a native

oligosaccharide containing therapeutically relevant sialic acid. Both glycoconjugation

methods had remarkable second order rate constants and were demonstrated to be compatible with E coli pellets or even live cells. The developed methodologies contribute to

understanding the molecular precision of glycans for many glycan binding proteins.

37

2. Inhibition of multivalent interaction between glycan and pathogenic bacterial lectin

2.1 Introduction

Figure 18. Ribbon representation for series of multivalent glycan binding protein (white structure represent protein, blue structure represent glycan. PDB ID: from left, 1OKO, 1BOS, 1OP5) Protein-Glycan interactions for ubiquitous biological events often rely on multivalency (Figure 18)³³. For instance, on the bacterial surface, there is a glycan binding protein called lectin which has multiple glycan binding sites for the recognition of host cell surface glycans for infection (Figure 18a as example of tetravalent galactose binding lectin LecA)³⁴. As well, certain bacterial toxins like AB5 toxins share the B subunit which is responsible for glycan binding to enter the cell and other A subunit to perform catalytic machinery to take over the host cell’s regular function (Figure 18b as example of Shiga like toxin B subunit)³⁵. HIV neutralizing antibody 2G12 have been known for multiple binding sites to recognize HIV surface glycoproteins (Figure 18c)³⁶. Therefore ligands which have multivalent interaction for those proteins provide beneficial outcomes such as inhibition of pathogen invasion and vaccination strategy. Numerous studies have already been established in which polymers or dendrimers with tailored ligand density can recapitulate the function of oligomeric glycan interaction or effectively outcompete host-pathogen interactions (Figure 19a,b)³⁷. More recently, it has been demonstrated that DNA can be used to organize and display multiple

38

nucleic acid (DNA or PNA)³⁸ tagged glycans, which provide a programmable platform to vary the interligand distance (Figure 19c)³⁹. This method allows the creation of small libraries with different distances and ligands depending on DNA sequence and glycan conjugates to determine optimal configuration. However, a structurally well-designed ligand is required to enhance the potency and selectivity.

Figure 19. Established strategies to recapitulate the function of complex glycan assemblies. a:

dendrimer based approach for LecA inhibition. b: polymer based approach for SLT-1 inhibition. c:

DNA programmed approach for 2G12 binding.

39

2.2 Results and discussion-PNA encoded heteroglycoconjugate library

This project was carried out in collaboration with Dr. Alexandre Novoa et al.⁴⁰

As well as glycan-glycan distance, secondary interactions of non-carbohydrate moieties is also important to increase the affinity to glycan binding protein⁴¹. A PNA encoded library allowed the synthesis of a structurally diverse hetero-glycoconjugate library with individual PNA barcodes representing each structure of the library for DNA microarray readout (Figure 20)⁴². A 10’000 member hetero-glycoconjugate library was constructed with 33 different glycans with three different spacers between two glycans and three terminal structures. A total of 7 different lectins were investigated to identify hit binders and two examples were shown (Figure 21). For instance, BambL, a fucose-specific lectin from Burkholderia

ambifaria, showed preferential binding to H-type 2 epitope (Fuc1-2Gal1-4GlcNAc) and a combination of fucose or digalactose increased binding affinity (Figure 21a). The ligand with the strongest affinity was resynthesized in the solution form and ITC measurement revealed that this compound was shown to have good affinity for BambL (K_d = 1.9 M, which is 3 times superior to that for H-type2 alone, Figure 21b). It is important to note that the compound was found to bind bambL (which has two binding sites) with 2:1

stoichiometry. This indicates that the improved affinity achieved in this case with the hetero- glycoconjugate compared to H type2 was reasoned from secondary interactions around the binding site rather than occupying two binding sites simultaneously. Such an information could be obtained from other lectins too. LecA from Pseudomonas aeruginosa is known to bind with -linked galactoside⁴³, but has stronger affinity for galactosides with -linked aromatic aglycons (Figure 21a,c).

40

This information could be transferred for the synthesis of a focused library for specific lectin to improve affinity (Figure 22)⁴⁴. Novoa and Winssinger and co-workers investigated the inhibition of LecA through the construction of a focused library for this protein which

included thiophenylgalactose moiety which was determined to have better affinity to LecA in the last experiment. Also, this library was designed to consider the distance between the two binding sites of LecA and structurally diverse linkers were included in the library contents.

These efforts resulted in finding an excellent binder 2.4 (K_D=82 nM, Figure 22b). The compound 2.4 showed inhibition of Pseudomonas aeruginosa invasion of human lung epithelial cell.

41

Figure 20. a: PNA encoded library for DNA display b: 10’000 members hetero-glyconconjugate library

42

1 2

3 1

2 3

1

Figure 21. DNA display analysis of 10’000 member hetroglycoconjugate library for different lectins.

a: fluorescence image of microarray. b: ITC measurement of identified most potent glycan conjugate 2.1. c: structures of identified glycoconjugate.

43

Figure 22. Focused library for Lec A inhibition. a: space-filling representation of the crystal structure of LecA and library construction including thiophenylgalactoside moietiy. b: fluorescent image of microarray (left) and inhibition of P. aeruginosa invasion of human lung epithelial cells (right).

44

2.3 Results and discussion-PNA programmed dynamic cooperative glycan assembly

This project was carried out in collaboration with Dr. Alexandre Novoa et al.⁴⁵

Figure 23. PNA-programmed dynamic cooperative glycan assembly.

Although covalent bond assembly has been utilized to create a potent multivalent inhibitor, the noncovalent interactions, especially nucleic acid programmed assembly, are attractive since nucleic acids are able to form multiple different architectures (Figure 23)⁴⁶. Reversible interactions have already been widely used to discover diverse ligands in the field of dynamic combinatorial chemistry and DNA encoded libraries⁴⁷. It is feasible that nucleic acid-ligand complexes have the potential to form more complex 3D architectures to fit diverse

multivalent proteins.

For that, Ralstonia solanacearum lectin (RSL) and Burkholderia ambifaria lectin (BambL) were investigated to find multivalent inhibitor based on PNA-glycan conjugate (Figure 24).

The Burkholderia cepacia complex is a group of closely related bacteria that cause lung

45

infections in immunocompromised patients as well as in patients with granulomatous disease or cystic fibrosis⁴⁸. BambL and RSL are bacterial lectins that share six fucose binding sites formed by trimerization of monomers carrying two fucose binding sites⁴⁹. It has been shown

Figure 24. a: Space-filling and ribbon representation of Ralstonia solanacearum lectin (RSL). b:

schematic representation of PNA programed hexavalent binder.

that the interaction of RSL with fucosylated epitopes such as histo-blood group

oligosaccharides exposed on glycolipids is sufficient to induce membrane invaginations in giant liposomes⁵⁰. To block those interactions, a fucose ligand with two 8 mer PNA tags was

46

designed (Figure 24b). Using the appropriate complementary sequences, the fucose ligand can be oligomerized to form a hexagon architecture around the protein. The 8-mer PNA duplex

Scheme 4. Solid phase synthesis of PNA-fucose conjugates.

stretches over 34 Å with a small curvature which is sufficient to bridge two binding sites (PDB ID: 3PA0)⁵¹. The fucose ligand is linked to each side of the PNA with a 10 Å flexible PEG spacer that leaves some plasticity to fit the multivalent interaction. The 8 mer PNA have Tm > 60 °C thus yielding stable assemblies at room temperature.

47

To do that, all the PNA-fucose conjugates were prepared by solid phase synthesis (Scheme 4, details in the supporting information). PNA monomer, either serine modified at the  position or non-modified, were used to synthesise the PNA sequences by Mtt chemistry⁵². PEG spacer and propargylglycine was coupled by Fmoc chemistry. Acetylated 2-

azidoethoxyfucose was coupled with the sequence on resin by

Figure 25. KD values of PNA-fucose assemblies based on SPR measurement.

48

Figure 26. a: Schematic representation of the dynamic fucose dimer interaction with RSL (red triangle represent fucose). b: structure and corresponding KD of different dynamic assemblies based on SPR measurement.

49

CuAAC chemistry. Finally, the conjugates were cleaved from resin by TFA and the acetyl group was deprotected with NaOH solution to obtain the desired sequences after reverse phase HPLC purification.

The affinities of the different assemblies were measured by SPR (Figure 25). Compared with fucose-triazole and monomeric fucose-PNA conjugate, dimeric fucose-PNA conjugate showed a significant gain of affinity (>500 fold from fucose-triazole). This suggested that the conjugate is able to bridge two binding sites. However, progression to higher-order oligomers only brought a small incremental benefit. Enthalpy-entropy compensations might be

considered as a potential reason for this diminishing return in affinity gain.

Next, a dimeric assembly with shorter PNA (4 mer) sequences was investigated (Figure 26).

In this case, the shorter PNA would not form stable duplexes in solution but only in the presence of the lectin through the cooperative interaction of the PNA duplex and the ligand- target interaction (Figure 26a). The 4 mer PNA sequences have self-complementary palindromic sequence (TTAA: KD = 13 , Tm < 15 °C at 2 M; GGCC: KD = 3.8 , Tm

23 °C at 2 M, in the supporting information). As a result, both TTAA and GGCC showed a dramatic gain of affinity compared with the fucose ligand (>100 fold, Figure 26b). The SPR sensorgrams showed a significant difference in koff between fucose ligand 2.5 and dimeric ligand 2.13 that is characteristic of multivalent interactions (k_off = 2 × 10^-1 s^-1 for 2.5, koff = 5 × 10^-4 s^-1 for 2.13, in the supporting information). As further controls, glucose-PNA conjugate 2.14 and PNA omitting carbohydrate 2.15 were prepared and neither of these controls had measurable affinities for the lectin (in the supporting information). To further assess the cooperative dynamic assembly of 2.13, SPR measurements of 2.13 with increasing amounts of control 2.15 which competes for hybridization (up to 5 equivalent) was performed (in the supporting information). The presence of control 2.15 had no impact on the affinity

50

of 2.13 at stoichiometric amounts (at this concentration, the probability of a 2.13 dimer from random hybridization is 25%). At higher concentrations (2.5 and 5 equivalents), the binding curves were slightly altered due to non-specific binding of 2.15 in the control channel, however, the same slow dissociation was observed as in the binding of 2.13 alone (at 5 equivalent of 2.15, the statistical probability of a 2.13 dimer is 2.7 %), thus suggesting a self- sorting of 2.13 dimer on the protein. The dramatic gain in affinity observed for the dynamic assembly of 2.12 or 2.13, as well as the lack of influence of 2.15 in competing for

hybridization, can only be rationalized by the cooperativity of PNA hybridization and lectin interaction wherein the rebinding of ligand in the dimeric chelate is faster than PNA-duplex dissociation, and likewise, duplex formation at the protein surface is faster than koff of the fucose ligand.

Next, the efficacy of the more potent assembly 2.13 to block the binding of BambL to human lung epithelial cells H1299 was investigated (Figure 27, in the supporting information).

Both anomeric configurations of the fucose linked to the PNA were prepared. Assemblies arising from both anomeric configurations were competent in inhibiting BambL binding to epithelial cells in a dose-dependent manner with IC₅₀ values of 0.56 M and 0.94 M for the

-anomer and -anomer of fucose, respectively, which was 723-fold more effective than the fucose alone. The gain of efficacy observed in this assay with the dynamic assemblies concurs with affinity measurements on RSL. It should be noted that the assays were carried out at a RSL concentration of 178 nM with a capacity to bind up to 1 M of fucose ligand.

While occupancy of the six binding sites is not necessary for adhesion inhibition, the IC50

measurement points to the fact that the inhibition is achieved with nearly stoichiometric quantities of ligand relative to the lectin. Furthermore, these cellular assays were performed at 37 °C, which further displaces the equilibrium away from duplex formation. The fact that

51

assemblies 2.13 were more efficacious than the ligand alone at this temperature indicates that the cooperativity between ligand binding and duplex formation is still operative.

Figure 27. Inhibition of BambL binding to human lung epithelial cells H1299.

52

2.4 Conclusion and future aspect

In this chapter, multivalent interactions of lectin were successfully blocked by divalent heteroglycoconjugate and PNA programmed glycan assembly. A DNA display of

heteroglycoconjugate library demonstrated the importance of secondary interactions in the binding of lectin as well as glycans to increase affinity. PNA programed glycan assembly showed very potent affinity to hexavalent lectin and blocked the binding to human lung epithelial cells. The dynamic assembly was stabilized in the presence of equivalent of complementary PNA sequence supporting the assertion that a self-sorting system was operating. This work offers a general strategy to discover inhibitors for lectins which have unique distances between glycan binding pockets.

53

3. Short peptide catalyst folded through PNA hybridization

Figure 28. a: Ribbon representation of crystal structure of trypsine triad (PDB ID: 2PTC). b: stereo and regioselective reactions by folded short peptide catalyst.

Catalysis in living systems is generally achieved by proteins that fold in unique three-

dimensional structures. The ideal orientation of each amino acid within the enzyme creates an active site to stabilize the transition state of reaction for catalysis. Various methods such as computational⁵³, antibody⁵⁴ and directed evolution⁵⁵ have been applied to improve ideal orientations for transition state stabilization. Generally, only a few residues participate in catalysis, as exemplified by the catalytic triad of proteases (Figure 28a). However, a relatively large molecular weight (> 20 kDa) is required to achieve a stable, folded

54

conformation. The prospect of recapitulating a given catalytic reaction using minimal peptides or a rigid scaffold to position the required functionalities has attracted significant attention, and catalysis with oligopeptides has delivered ground-breaking results (Figure 28b)⁵⁶. For catalysis with short peptides, selected amino acids are frequently included to create a conformational bias (for instance,

Figure 29. Ribbon representation of crystal structure of ribozyme active site (PDB ID: 3ZD5).

Pro-Aib or C-tetrasubstituted amino acids have been used to induce a  turn) and position the required side chains in close proximity.

On the other hand, oligonucleotides adopt stable tertiary structures with relatively short sequences. However, relatively few reactions are efficiently catalysed by structured

oligonucleotides, phosphodiester cleavage by the hammerhead ribozyme being one example

55

(Figure 29)⁵⁷. Clearly, the larger pallette of functional groups present on amino acid side chains endows proteins with a broader scope of chemistries.

56

3.2 Results and discussion-phosphatase activity screening

This project was carried out in collaboration with Dr. Som Dutt.⁵⁸

Figure 30. a: peptide-oligonucleotide structure folded through hybridization. b: screening of one-bead one compound catalyst library for phosphate hydrolysis based on precipitating dye.

Short peptides conjugated to flanking oligonucleotides should allow for structuring through hybridization, resulting in a productive disposition of amino acid side chains for catalysis (Figure 30a). For ease of conjugate synthesis, PNA employed and synthesized by standard solid-phase peptide synthesis (SPPS). Hybridization of PNA-peptide conjugates has already

57

been used to modulate the conformation of peptides involved in biologically relevant peptide- receptor interactions⁵⁹, however, this approach has not been used in the context of catalysis.

Scheme 5. Synthesis of immolative linker triggered by azide reduction.

Several methods have been reported to screen libraries of catalysts on beads⁶⁰. Screening for the hydrolysis of a phosphate ester was chosed based on the fact that the background

hydrolysis is very slow and cleavage can be monitored using a fluorogenic precipitating dye (quinazolinone precipitating dye: QPD, Figure 30b) which is routinely used to image alkaline phosphatase activity⁶¹.

To do that, the immolative linker 3.4 which is orthogonal with acidic deprotection of peptides was designed (Scheme 5). For sequence detection, the linker cleavage can be triggered by the reduction of the para-azidobenzyloxy moiety to cause 1,6-elimination. This type of elimination was shown to proceed with a half-life of approximately 20 min (k = 0.035 min^-1)⁶².

58

For the synthesis, the azide group was introduced to the commercially abailable 6- aminophthalide 3.1 and the lactone ring was opened by ethylene diamine to afforded the amine which was subsequently protected with an Fmoc group to obtain compound 3.2. The hydroxyl group of 3.2 was converted to p-nitrophenylcarbonate to obtain 3.3. This reaction resulted in low yield with a long reaction time because significant intramolecular cyclization raised from the neighboring amide to the benzyl position was occurring in the

Figure 31. a: Stability of immolative linker for acid deprotection condition. b: NMR study for kinetic of acid degradation of linker.

59

presence of base or solvent such like DMF. Therefore the reaction was carried out in a suspension of DCM without any base. This linker was loaded onto resin through a carbamate linkage to obtain resin 3.4.

The NMR study was performed to investigate the stability of the designed linker (Figure 31).

In the deuterium trifluoroacetic acid, carbamate 3.5 had ca. 50% of degradation in 60 min which was sufficient for peptide deprotection.

Figure 32. Split and mix synthesis of 1’000 member one-bead one-compound (OBOC) library.

The library was designed to contain a bias for residues frequently found in hydrolytic enzymes (His, Ser, Asp, Lys) in a hairpin loop resulting from a constant CGGC sequence prior to the peptide and a complementary GCCG sequence at the end of the peptide (Figure 32). A preliminary experiment showed that a tetraglycine hairpin with this sequence had a Tm

of 70 °C, ensuring predictable folding. A library of 1000 PNA-peptide adducts was prepared

60

by split-and-mix synthesis using 3 × 10 sets of peptide fragments with variable

Figure 33.a: Microscope image of fluorescence on bead from 1,000 member library with different time point. condition: 2 mM QPD phosphate in PBS buffer pH 8.0. b: Statistic analysis of

fluorescence on each individual bead after 5 h. c: MALDI-TOF mass spectrum of hit sequence cleaved from bright fluorescence bead.

lengths (0- to 4-mer) affording the library as 10 pools of 100 putative catalysts with a hairpin composed of 4-10 mer peptides.

61

The library was screened by exposing 400-700 beads of each pool (i.e. 4-7 copies of each library member) to the QPD-phosphate (2 mM) in PBS buffer (pH 8.0). The development of fluorescence was monitored as a function of time by fluorescence microscopy using a high content screening microscope at 37 °C over a period of 5 h (Figure 33a and details in the supporting information). A clear discrepancy in the number of highly fluorescent beads was shown amongst the 10 pools, with the pool corresponding to the fragment C8 showing some of the brightest beads (Figure 33b). The brightest beads were picked up with a pipette tip, distributed on a MALDI plate and treated with 2 l of a 2 mM PMe₃ solution in H₂O/THF followed by matrix for direct MALDI analysis (details in the supporting information).

Three of the beads had the mass corresponding to the sequence Lys-His- Ile-Ser-Glu-Ser (Figure 33c).

Figure 34. Catalytic activity of identified catalyst and control sequences measured in PBS (25 mM, pH 8.0) at 37 °C. The fluorescence was measured at ex = 345 nm and em = 530 nm (Y-axis shows arbitrary fluorescent units). Each data point is the average of 3 experiments.

To validate this result, the hit sequence was resynthesized on a Rink amide resin to obtain the product in solution following TFA cleavage. To ascertain the contribution of PNA

62

hybridization, control sequences lacking the PNA 3.13, with only one side of the PNA 3.11 and 3.12 (C and N terminal, respectively), or different hairpin forming sequences 3.7 and 3.8 were prepared (Figure 34). As further controls, a sequence containing the linear catalytic peptide emanating from a hexaglycine hairpin duplex 3.9 and a conjugate that cannot fold 3.10 were prepared. Analysis of the phosphate hydrolysis using the fluorogenic substrate (100

M catalyst, 2 mM QPD-phosphate in 25 mM PBS buffer, pH 8.0) confirmed the activity of the selected sequence 3.7 and showed that the hybridization –enforced hairpin was necessary for catalysis. The peptide alone 3.13, the peptide with only one side of the PNA 3.11 and 3.12, the PNA-hairpin conjugate to the linear peptide 3.9 and unfolded PNA-peptide adduct 3.10 showed dramatically less activity. Furthermore, specific nucleobases are unlikely to be involved in the catalysis since comparable phosphate hydrolysis activity was obtained for the GC-based hairpin as with the AT-based hairpin (3.7 vs. 3.8). Comparing the slope of the reaction for compound 3.7 vs. 3.9 indicates a rate acceleration of >25-fold upon hairpin formation. Performing the reaction with 2 mM EDTA did not alter the rate of reaction, suggesting metal chelation is not involved.

63

Figure 35. Alanine-scan of the catalyst measured in PBS (25 mM, pH 8.0) at 37 °C. The fluorescence was measured at ex = 345 nm and em = 530 nm (Y-axis shows arbitrary fluorescent units). Each data point is the average of 3 experiments.

Next, the relative contribution of each residue with a heteroatom was investigated using an Ala scan of the sequence (Figure 35). Lys or His substitution 3.17 and 3.18 led to loss of activity, whereas the other positions (compound 3.14-3.16) yielded products with comparable activity. It should be noted that compound 3.7 failed to cleave other aryl phosphates

(phosphorylated form of 4-nitrophenol, 2-nitrophenol, 7-hydroxycoumarin and salicylamide), suggesting some substrate specificity. Taken together, these results support the hypothesis that the observed activity results from catalysis mediated by a constrained conformation of the Lys-His dipeptide within a hairpin loop that provides a favourable interaction with the substrate.

Based on the impact of the hairpin on catalysis, it was considered whether allosteric control of activity could be engineered by toehold strand displacement switching the hairpin

configuration to an open configuration (Figure 36a). To this end, compound 3.19 with a

64

hairpin made up of a 3 mer PNA with a 3 mer overhang was prepared. It is interesting to note that at 37 °C compound 3.19 had 50% of the activity of compound 3.7. However, at 25 °C, both compounds showed comparable activity. This is consistent with the measured Tm of the hairpin (Figure 36b); at 25 °C, both 3.7 and 3.19 are predominantly in the hairpin

Figure 36. a: Allosteric switch of catalytic activity of 3.20 measured in PBS (25 mM, pH 8.0) at 25

°C. The fluorescence was measured at ex = 345 nm and em = 530 nm (Y-axis shows arbitrary fluorescent units). Each data point is the average of 3 experiments. b: Tm value measurement of different intramolecular hybridization.

65

configuration, whereas at 37 °C a significantly larger portion of compound 3.19 is in the non- productive open form than compound 3.7. The catalytic activity of the compound 3.19 was indeed dramatically attenuated in the presence of complementary PNA (3.20) that opens the hairpin, demonstrating that a programmable allosteric control is possible. Concurringly, the activity of 3.19 could also be attenuated during the course of the reaction by the addition of the complementary PNA (3.20).

At the end, the enzymatic parameters of catalyst 3.7 were investigated by Lineweaver-Burk plot (Km=10 mM, kcat=0.045 h^-1, Figure 37).

Figure 37. Lineweaver-Burk plot of catalyst 3.7 at 37 °C (5, 3.75, 3, 2.5, 2 mM of QPD-phosphate, respectively). Km=10 mM, kcat=0.045 h^-1

66

3.3 Result and discussion-metallo peptide-PNA catalyst for expanding scope of reaction by mass spectrum based screening.

Figure 38. a: metallopeptide-PNA catalyst. b: synthesis of metal-binding monomer.

Transition metals have been demonstrated to play important roles in diverse organic transformations such as cross coupling⁶³, photoredox reaction⁶⁴, cycloaddition⁶⁵, and C-H activation⁶⁶. Recently, metalloenzyme, metallopeptide or DNA metal complex have been shown to be attractive scaffolds to perform diverse reactions with good selectivity and rate acceleration (for metalloenzyme⁶⁷, metallopeptide⁶⁸ and DNA metal catalyst⁶⁹). To expand the reaction scope of peptide-PNA catalyst, a metal binder domain was introduced to peptide (Figure 38a).