Synthesis, structure and conformation of oligo- and polysaccharides

E. Andreas Larsson

Department of Organic Chemistry Arrhenius Laboratory Stockholm University

Stockholm, Sweden 2004

ii

iii

Doctoral Dissertation 2004 Department of Organic Chemistry Stockholm University

Abstract

Carbohydrates are a complex group of biomolecules with a high structural diversity. Their almost omnipresent occurrence has generated a broad field of research in both biology and chemistry. This thesis focuses on three different aspects of carbohydrate chemistry, synthesis, structure elucidation and the conformational analysis of carbohydrates. The first paper describes the synthesis of a penta- and a tetrasaccharide related to the highly branched capsular polysaccharide from Streptococcus pneumoniae type 37. In the second paper, the structure of the O-antigenic repeating unit from the lipopolysaccharide of E. coli 396/C1 was determined along with indications of the structure of the biological repeating unit. In addition, its structural and immunological relationship with E. coli O126 is discussed. In the third paper, partially protected galactopyranosides were examined to clarify the origin of an intriguing 4JHO,H coupling, where a W-mediated coupling pathway was found to operate. In the fourth paper, the conformation of methyl α-cellobioside is studied with a combination of molecular dynamics simulations and NMR spectroscopy. In addition to the expected syn-conformation, detection and quantification of anti-φ and anti-ψ conformers was also possible. Andreas Larsson 2004 ISBN 91-7265-872-X, pp 1-40 Akademitryck AB, Edsbruk

iv

v

Contents List of papers vi

List of Abbreviations vii

1 General introduction 1 1.1 Synthesis of oligosaccharides 2 1.2 Experimental methods for structure determination of carbohydrates. 4 1.3 Conformational analysis of carbohydrates 8

2 Synthesis of oligosaccharides related to the repeating unit of Streptococcus pneumoniae type 37 (paper I) 13

2.1 Introduction 13 2.2 Strategy for the synthesis 15 2.3 Synthesis of tetrasaccharide 1 15 2.4 Synthesis towards hexasaccharide 2 - an abandoned approach 17 2.5 Synthesis of pentasaccharide 13 18 2.6 Future possibilities 21

3 Structural and immunochemical relationship between the O-antigenic polysaccharides from enteroaggregative Escherichia coli strain 396/C1 and Escherichia coli O126 (paper II) 22

3.1 Introduction 22 3.2 The E. coli 396/C1 O- antigen repeating unit 22 3.3 Structural and immunological relationship with E. coli O126 23 3.4 The biological repeating unit 23

4 Analysis of NMR J couplings in partially protected galactopyranosides (paper III). 25

4.1 Introduction 25 4.2 The model substances and their NMR spectra 25 4.3 Optimization and analysis of the different rotamers 27 4.4 W-coupling 29 4.5 Comparison of experimental and calculated properties 29

5 Determination of the conformational flexibility of methyl α-cellobioside in solution by NMR spectroscopy and molecular simulations (paper IV) 33

5.1 Introduction 33 5.2 1D DPFGSE T-ROESY 33 5.3 Deuterium isotope effects 34 5.4 Computer simulations 35 5.5 Population analysis 35

Acknowledgements 36

References 37

vi

List of papers

This thesis is based on the following papers and manuscripts, referred to by their Roman numerals I-IV.

I Synthesis of oligosaccharides related to the repeating unit of the capsular polysaccharide from Streptocuccus pneumoniae type 37 E. A. Larsson, M. Sjöberg and G. Widmalm In manuscript

II Structural and immunochemical relationship between the O-antegenic polysaccharides from the enteroaggregative Escherichia coli strain 396/C1 and Escherichia coli O126# E. A. Larsson, F. Urbina, Z. Yang, A. Weintraub and G. Widmalm Carbohydr. Res. 2004, article in press

III Analysis of NMR J couplings in partially protected galactopyranosides* E. A. Larsson, J. Ulicny, A. Laaksonen and G. Widmalm Org. Lett. 2002, 4, 1831-1834

IV Determination of conformational flexibility of methyl α-cellobioside in solution by NMR spectroscopy and molecular simulations* E. A. Larsson, M. Staaf, P. Söderman, C. Höög and G. Widmalm J. Phys. Chem. A 2004, 108, 3932-3937

Reprinted with kind permission of Elsevier# and ACS publications.*

vii

List of Abbreviations

AgOTf silver trifluoromethanesulfonate (triflate)

CPS capsular polysaccharide

DDCCR dipole dipole cross correlated relaxation

DDQ 2,3-dichloro-5,6-dicyano-1,4-benzoquinone

DMTST dimethyl(methylthio)sulfonium triflate

DPFGSE double pulse field gradient spin echo

EAEC enteroaggregative E. coli hJ hydrogen bond mediated J-coupling

HSEA hard sphere exo anomeric

ISPA isolated spin pair approximation

LPS lipopolysaccharide

MALDI-TOF matrix assisted laser desorption ionization – time of flight

MC monte carlo

MD molecular dynamics

MM molecular mechaniscs

NIS N-iodosuccinimide

PMB p-methoxybenzyl

QM quantum mechanics

R2 transverse relaxation rate, 1/T2

sym-collidine 2,4,6-trimethylpyridine

TFA trifluoroacetic acid

TfOH trifluoromethanesulfonic acid (trifflic acid)

T-ROESY transverse rotating frame Overhauser effect spectroscopy

viii

1

1 General introduction

Carbohydrates form a complex group of biomolecules with diverse biological functions. These include roles such as being the primary energy source for biological systems, the main structural elements of plants in the form of cellulose and the K- and O-antigens – an information carrying immunologically active identifier tag, coating cell walls of bacteria in the form of capsular and lipopolysaccharides.[1] The possibility of forming a large number of possible structures with just a few monosaccharide units gives rise to an unsurpassed information density in polysaccharides, as compared to DNA. This makes it possible to form unique combinations of small repeating units, consisting of very few sugars. However, the same structural diversity makes both the analysis and synthesis of carbohydrate structures challenging. To be able to understand the carbohydrates themselves or the interactions involving carbohydrates and other biomolecules, pure oligosaccharides are needed. Such substances can be obtained, either by isolation from natural sources or by chemical or enzymatic synthesis. If the aim is to use analogues or isotopically labelled versions of the native structure, chemical synthesis is usually the only option. Biological interactions such as molecular recognition are based on the three-dimensional shape of molecules. This shape is in turn determined by the chemical structure and the conformation of the molecule. Determining the structure of carbohydrates includes several steps: finding out which monosaccharides are present, elucidating the actual linkage positions between the monosaccharide units and determining the stereochemistry in the structure. This information corresponds to the sequence of amino acids in proteins, with the additional complication that it does not necessarily have to be linear and that all linkages have two possible stereoisomers. This analysis is usually performed with a variety of NMR experiments. In addition, fragment analysis with mass spectrometry lends further support to the proposed structures. The conformation is typically determined by a combination of computational techniques and NMR spectroscopy but X-ray crystallography is also useful if crystals can be grown.[2] In paper I, the synthesis of penta- and tetrasaccharides structurally related to the highly branched capsular polysaccharide from S. pneumoniae type 37 is described. In Paper II, the structure of the O-antigenic lipopolysaccharide repeating unit of E. coli 396/C1 was determined. In addition, its structural and immunological

2

relationship with E. coli O126 is discussed. In paper III, partially protected galactopyranosides are examined to clarify the origin of an observed 4JHO,H coupling. In paper IV, the conformation of methyl α-cellobioside is studied with a combination of molecular mechanics and NMR. In addition to the expected syn-conformation, detection and quantification of anti-φ and anti-ψ conformers was also possible.

1.1 Synthesis of oligosaccharides

1.1.1 The glycosylation reaction

A glycosylation is described as the reaction between the hydroxyl group of a glycosyl acceptor and the anomeric centre of a glycosyl donor (Figure 1.1). The glycosyl donor carries an anomeric leaving group, which reacts under the influence of a promoter, commonly in a SN1 fashion.

OPO

PO

OP

LGPO

OPO

PO

OP

PO

OPO

POPO

POOR

+Promoter

- LG− HO-R

OPO

POPO

POOR+

-H+Glycosyl donor

α− and β−glycosides

Oxocarbenium ion

P = Non participating protecting groupLG = Leaving group

Figure 1.1 Typical mechanism of a glycosylation.

The oxocarbenium intermediate (Figure 1.1) is accessible from both the α and the β face, typically generating a mixture of anomers. A weak stereoelectronic effect termed the anomeric effect[3, 4] stabilizes products with axially oriented electron withdrawing substituents. By employing equilibrating conditions,[5] the thermodynamically stable α-anomer can often be the dominant product in the

3

reaction mixture[6]. In addition, the influence of the solvent[7, 8] can also have an effect on both the yield and stereochemistry.

1.1.2 Neighbouring group participation

The presence of a participating protecting group (i.e., a protecting group with an electron rich carbonyl oxygen which reversibly can stabilize the positive charge at the anomeric centre) influences the stereoselectivity greatly. The acyloxonium intermediate, (see Figure 1.2) shields one of the faces from incoming nucleophiles and thereby directs the stereochemistry of the reaction. If present on the vicinal 2-position, a high degree of selectivity can be achieved, producing 1,2-trans glycosides. In addition, participating protecting groups in the more distant 4-position of galactose can also influence the anomeric selectivity.[9]

OPO

PO

OP

LGO

R O

OPO

PO

OP

O

+

R O

OPO

PO

OP

O

R

O

+

HO-R

OPO

POPO

POOR

1,2-trans glycosides

-H+

Promoter- LG−

P = Protecting groupLG = Leaving group

Figure 1.2 Glycosylation in presence of a participating protecting group.

Protecting groups may also affect the reaction in a more passive way. Steric interactions can make the approach to one side less favourable, and thus alter the reaction path.[10] In the case of mannose configured pyranoses, neither the anomeric effect nor participation of protecting groups work in favour of the formation of β-glycosides. Synthesis of such glycosidic linkages requires different methods.[11-13] A wide range of leaving group/promoter systems have been developed.[5, 14-20] A general disadvantage of most of the leaving groups is their low stability towards many of the conditions involved in protecting group manipulations. One of the few

4

exceptions is thioglycosides[21] which in general are relatively inert, but reacts in the presence of soft electrophiles and oxidizers. Thus they can be used as an anomeric protecting group as well.

1.2 Experimental methods for structure determination of carbohydrates.

1.2.1 Introduction

To define the structure of a carbohydrate polymer, several aspects must be determined. Three techniques dominate the field of structure determination: X-ray crystallography, mass spectroscopy and NMR spectroscopy. X-ray crystallography is a very powerful technique, invaluable for protein and inorganic chemists. However, it has so far found a relatively limited use with carbohydrates, mainly due to difficulties obtaining crystals. Mass spectroscopy is also a very useful technique which has grown in importance for bio-molecules since the development of soft ionization techniques such as MALDI[22] and electrospray.[23] However, the amount of information that can be extracted is relatively limited when compared to NMR. In addition, carbohydrate polymers can be difficult to ionize. This is particularly true for lipopolysaccharides. NMR spectroscopy has for many years been an important method for structural analysis of carbohydrates.[24] With the aid of highfield magnets, sensitive cryo-probes and modern pulse sequences, NMR structural determination is becoming more and more the tool of choice.[25] Sensitivity, the major drawback of NMR is gradually becoming less of an issue. However, solubility, polymer heterogeneity[26] and polymer size[27] can still complicate the procedure considerably. In principal, before employing physical methods (e.g. NMR), chemical methods are used to gather information about the identity and quantity of present monosaccharides components[28] and their absolute configuration.[29]

1.2.2 NMR spectroscopy

First a handle is identified for all spin systems (residues). In most cases, the anomeric protons serve well since they often are sufficiently resolved. The next

5

task is to assign the individual sugar residues. When all atoms in all residues (optimally) are assigned, the inter-glycosidic connectivities may be established. During this process, a range of mainly 2D-experiments are used. I have chosen to divide the description of these experiments in two sections: those mainly useful for determining intra-glycosidic correlations and those deducing the glycosylation pattern, respectively. A typical structure determination starts by acquiring a proton spectrum. Usually, the amount of information that can be extracted is limited: However, some important aspects can be addressed. Impurities or heterogeneity may be visible in the one dimensional proton spectra. In addition, the number of unique anomeric protons and the CH3 protons of N- and O-acetyl groups and 6-deoxy sugars may also be determined (see Figure 1.3). Following this, a 1H,13C correlated HSQC (Heteronuclear Single-Quantum Correlation)[30] spectrum is run and the anomeric proton/carbon atom pairs can be detected. It also provides the 1JCH coupling constants which, in combination with the 3JHH coupling constants extractable from the proton spectrum, are invaluable in determining the anomeric configuration.[31-33] The 1H,13C HSQC spectrum commonly resolves the majority of the 1H,13C pairs. Although not assigned, they provide a “pool” of 1H,13C pairs which can be expected to be found in later experiments.

1H (ppm)

1.01.52.02.53.03.54.04.55.05.56.0

Figure 1.3 1H NMR of the E. coli 396/C1 polysaccharide. The signals from the anomeric, N-acetyl and H6 protons from the fucosyl residues protrude from the bulk region at ~3.2-4.4 ppm where most protons are found. The “hump” at ~4.6 ppm is the residual HDO peak after pre-saturation and post-processing.

6

1.2.3 Intra-glycosidic correlations

Using an anomeric proton as a reference point, a 1H,1H-COSY (COrrelated SpectroscopY)[34] provides access to H2. In a 1H,1H-TOCSY (TOtal Correlation SpectroscopY)[35, 36] a spin lock allows the detection of correlations to protons which do not couple directly to H1, i.e. H3-H6. How far into the spin system that correlations will be detected depends on the mixing time and the size of the coupling constants. For example, in the case of β-glucose, where all the coupling constants are large, a mixing time of 90 ms typically provides correlations to all protons in the residue. In contrast, for galactose, transfer to protons beyond H4 is rarely observed, due to the smaller coupling constants associated with H4. Performing multiple TOCSY spectra with increasing mixing times (typically 30-100 ms) consequently allows the user to “walk through” the spin systems starting from H1. A complement to 1H,1H-COSY and 1H,1H-TOCSY, which are J-coupling mediated experiments, 1H,1H-NOESY (Nuclear Overhauser Effect SpectroscopY)[37] can also be useful, in particular for systems containing protons with very small coupling constants such as H1-H2 in β-mannosides. In contrast to COSY and TOCSY, the correlations in a NOESY experiment arise from through space interactions of two protons. Hence, care must be taken, to avoid confusing intra- and inter-residue interactions (see Figure 1.4). NOESY and J-coupled spectra well complement each other and when combined, most proton resonances can be assigned. In cases of overlap in the 1H-dimension, combinations of the above mentioned techniques such as HSQC-TOCSY[38] can greatly facilitate the assignment. Although, accompanied with some loss of sensitivity, the fact that it transfers the correlations to the usually much less crowded 13C dimension can be hugely beneficial.

1.2.4 Inter-glycosidic correlations

At this point in the analysis, all 1H,13C pairs of the individual residues are usually assigned. Then, information on the sequence of the involved residues is gathered. Performing this task, one mainly relies upon two techniques: HMBC (Heteronuclear Multiple-Bond Correlation)[39] and NOESY. Most often, the anomeric proton is in the proximity of the proton of the substituted position on the other side of the glycosidic linkage. Hence, a crosspeak can be observed in a NOESY experiment. However, care must be taken since occasionally

7

more than one inter-glycosidic correlation can be found and sometimes to more than one residue (Figure 1.4). These multiple NOEs can provide an insight on the conformation of the polymer. In contrast, HMBC produces correlations of long-range J-coupled proton and carbon atoms. Generally, with a mixing time of ~50 ms, 2JCH and 3JCH correlations are observed, such as those found at inter-glycosidic sites. For very large polymers, (many thousands of repeating units) a mixing time of 50 ms is not feasible. The R2 increases with the size of the polymer, and when sufficiently large, the magnetization decays too rapidly for use of the HMBC experiment. Until recently, in such cases one had to rely on NOESY experiments from which, if spectral overlap was present in the proton dimension, sequence information is difficult to obtain. The DDCCR (Dipole Dipole Cross Correlated Relaxation)[40] which was designed to counter these problems needs a mixing time of only 10-15 ms to produce a spectrum similar to HMBC (see Figure 1.4).

a

68

70

72

74

76

78

80

66

64

13 C

(p

pm

)

b

1H (ppm)

5.07 5.07

B5

B3

E2

B2

E2

D2

1H

(p

pm

)

4.5

3.7

5.07

c

3.9

4.1

4.3

Figure 1.4 (a) HMBC, (b) DDCCR, and (c) NOESY showing the correlations from H1 of the α-L-fucosyl residue B at 5.07 ppm in the polysaccharide from E. coli 396/C1 (see Section 3). Connectivity to position 2 of residue E is established with all methods. Note the additional NOE correlation to residue D which is not linked to residue B.

8

1.3 Conformational analysis of carbohydrates

Conformational analysis is performed to gain access to information about the three dimensional shape of a molecule. The goal of a conformational analysis is to be able to highlight significantly populated conformers. In the case of a carbohydrate oligomer, the main degrees of freedom are found at the exocyclic bonds (see Figure 1.5 and Section 5). The flexibility of the ring itself is inconsequential for the most commonly occurring pyranoses.[41] However, if more detailed information is required, the hydroxyl torsions should also be monitored (see Figure 1.3 and Section 4).

C1'

OOH

O C4 O

H1' H4

C O

OH

φ ψ

θa b

ω

H

OMe

Figure 1.5 A picture of two simplified systems showing the principal degrees of freedom for carbohydrates: (a) φ (phi), ψ (psi) and ω (omega) and (b) the rotation of a hydroxyl group θ (theta). The glycosidic torsion angles are in this work defined as φ = H1´-C1´-O4-C4 and ψ = C1´-O4-C4-H4 and the hydroxyl torsion as θ = HO4-O4-C4-H4.

1.3.1 Factors influencing the conformation of carbohydrates

The 3D structure of carbohydrates is influenced by a number of factors. One of the most important is the exo-anomeric effect.[3] It is the same type of stereo-electronic effect that influences the stereoselectivity of glycosylations. The exo-anomeric effect stabilizes particular conformations of the φ-torsion. This leads to a conformational preference for α- and β-glycosides at about −60° and +60°, respectively.[42] But as demonstrated in paper IV, other conformations are also possible. Steric interactions also restrict the conformational space. In addition, inter-residue hydrogen bonds can also stabilize specific conformers and thereby affect the conformation.[43] Solvent interactions can also be important to the conformational preference at the glycosidic linkage.[44]

9

1.3.2 Computational methods for conformational analysis

Two conceptually very different methods, Quantum Mechanics (QM) and Molecular Mechanics (MM) are commonly employed in conformation analysis. Which one to implement is based on which properties are being studied and the available computing power. QM is much more calculation intensive and as such it limits the size of the system that can be studied. Despite the ongoing development in the speed of computers, QM calculations can still be very time consuming. QM calculations generate information about electronic interactions within a molecule. From the results of such calculations, intrinsic physical properties can be described, such as NMR J-coupling constants, IR spectra etc. (see Section 4 and paper III). When properties not directly dependent on the orientation of orbitals or electron densities are studied, MM is a more efficient method to use. MM is less computer intensive and allows for a more generous inclusion of specific solvent molecules which can prove more valuable than the precision of QM. The basis of MM is a force field,[45] a set of equations from which, a potential energy can be calculated. The force field simulates the forces that act upon molecules by optimizing the potential energy of bond angles, bond lengths, torsions etc. Upon minimization of the potential energy the molecule relaxes to lower potential energy conformers. A number of force fields exist. Some are optimized for use with carbohydrates.[46,

47] One example is HSEA[48] (Hard Sphere Exo Anomeric), which essentially treats the pyranose moiety as a rigid sphere. As the name explains, it is parametrized for the exo-anomeric effect. Perhaps, the most important technique is Molecular Dynamics (MD) simulations. By adding heat (in the form of kinetic energy), the simulated molecule is allowed to stochastically pass over energy barriers and over time, optimally populate all low energy conformations accordingly. The results are often viewed as trajectories of a distance or angle of interest plotted against time (see paper IV).

1.3.3 Conformational analysis employing NMR spectroscopy

The cycle time for a typical NMR experiment is on the order of seconds. Processes such as molecular motion and conformational dynamics usually equilibrate at a timescale several orders of magnitude shorter (i.e. in the nano- to milli-second range). The enormous difference in timescale between sampling and the processes observed in most cases means that the data extracted from the NMR experiments

10

represents a population weighted average of the different possible states a certain molecule can possess. Under such conditions, it is not possible to study or measure the presence of different conformations directly from NMR data, since the proportions and composition of the different states are usually not known. However, if NMR is used in combination with modelling techniques, the information from this averaged data can be converted into populations.

1.3.4 NOE derived distances

NOEs are transmitted through space and provide distance related information on nearby protons.[49] By determining the inter-proton cross-relaxation rates, σij, distances can be calculated. The derived distances are time averaged, but are very useful in the study of the three-dimensional structure of carbohydrates. The sign and the magnitude of the NOE is dependent on the magnetic field and the correlation time and thus of the size of the molecule. As a consequence, for certain combinations of the correlation time and the magnetic field, the NOE is minuscule. In such conditions, experiments based on ROE (Rotating frame Overhauser effect) or T-ROE[50, 51] (Transverse Rotating frame Overhauser effect), which give positive and significant ROE or T-ROE for all combinations of correlation-time and magnetic field, can be employed. Both ROESY and T-ROESY utilize a spin-lock. In a ROESY experiment the spin-lock can give rise to unwanted TOCSY-transfer. T-ROESY however, does not suffer from such problems. A convenient way to obtain the relaxation rates is to monitor the build-up of NOE with time. The NOE build-up can be mapped by a series of experiments with a range of mixing-times. By fitting this data to a second-order polynomial the cross-relaxation rates can be determined. Cross-relaxation rates can be interpreted with the Isolated Spin Pair Approximation (ISPA)[52]. ISPA requires the knowledge of a reference distance rref which can be obtained from molecular mechanics simulations. The distance of the unknown spin-pair rij, can then be calculated as rij = rref (σref / σij)1/6.

11

If the correlation time and the dynamics of the molecule are known, the cross-relaxation rates can be interpreted with Lipari and Szabo’s model-free approach[53,

54].

1.3.5 NMR Spin-spin couplings

For two nuclei to be able to exchange information in the form of scalar or indirect coupling their electrons must be able to interact.[55] Theoretically, it is usually described as a presence of electron density at the bond critical point, that is, the point where the orbitals of the two nuclei meet. Covalent bonds obviously fulfill this criterion, and normally scalar couplings can be considered to be bond mediated. The 3J spin-spin coupling constants are sensitive to the bond geometry and can be interpreted by a Karplus-type relationship. The relationship between coupling constants and geometry was formulated in an equation by Karplus in 1959.[56] The first equation was designed for 3J H-C-C-H, H-C-C-F and F-C-C-F systems, but other variants of the equation have been developed such as that for H-O-C-H torsion angles[57] (shown in the equation below and in Figure 1.6), which was suitable for the HO4-O4-C4-H4 torsion investigated in Section 4.

2.0cos5.1cos4.10 23 +−= θθHCOHJ

This is a useful tool but one must use the relation with caution. Due to the periodicity of the function, at least two, but normally four conformations are possible for every value of the coupling constant with the exception of the trans conformation.

12

Figure 1.6 The Karplus relationship for H-O-C-H torsion angles.

In addition, these J spin-spin coupling constants will be time averaged and one must realize that the measured coupling constant is an average of all conformers, weighted by their relative population. Similar equations exist for many other systems including the trans-glycosidic C-O-C-H torsion.[58] In that case 3JCH long range couplings provide information on the conformational preference of the glycosidic linkage.

1.3.6 Hydrogen bond mediated J-couplings

As mentioned in Section 1.3.5, J-coupling is seen as bond mediated. There are however other possibilities. Fluorine nuclei, in close proximity in space, but separated by multiple bonds can sometimes be seen to couple to each other (an effect of overlap of their electrons).[59] There are also examples in which hydrogen bonds have mediated couplings, apparently when sufficient electron density between the hydrogen bond donor and acceptor is found. Hydrogen bond mediated scalar couplings have recently been described by several groups, and are known to exist in between the base pairs of DNA and RNA[60-62] as well as in structures resembling carbohydrates.[63] The hJ couplings are usually small, but are extremely valuable in the refinement of protein and nucleic acid structures.[55, 64-66]

13

2 Synthesis of oligosaccharides related to the repeating unit of Streptococcus pneumoniae type 37 (paper I)

2.1 Introduction

Streptococcus pneumoniae is a serious human pathogen. It can among other things, cause community-acquired pneumonia and meningitis.[67] The surface of the bacterium is covered with capsular polysaccharides (CPS), which play an important role in its interaction with its host. The presence of a capsule has been shown to be an important virulence factor,[68, 69] in particular for disease development after primary infection.[70] The capsule offers protection for the bacterium against phagocytosis, and might also be involved in adhesion during infection.[70]

2.1.1 Streptococcus pnemoniae type 37

Of the 90 known types of S. pneumoniae, type 37 is the only one whose CPS is a homopolymer, containing only glucose residues. The genetic regulation for the biosynthesis of the type 37 capsule differs from all other known species of S. pneumoniae in another aspect; only one gene directs the synthesis of the capsular polysaccharide.[67, 71] Interestingly, it is located far from the clusters where these genes normally reside.[72]

2.1.2 Structural features of the capsular polysaccharide

The type 37 capsular homopolymer of S. pneumoniae is built from repeating units of →3)[β-D-Glcp-(1→2)]-β-D-Glcp-(1→ (see Figure 2.1). Earlier studies have indicated that this comb-like structure is most likely to be very sterically crowded.[73] The proton NMR chemical shifts of the CPS deviate considerably from what are expected compared to non-branched analogues. This indicates that O2 of a neighboring residue is in close proximity with the anomeric proton, since the calculated chemical shift[74] should be observed at 4.82 ppm in a non-perturbed structure compared to 5.12 ppm in the native CPS.[73]

14

OO

HOHO

OOH

HO

HO

OH

O

n

Figure 2.1 The repeating unit of the CPS from S. pneumoniae type 37.

A number of structurally related trisaccharides have previously been investigated by our group with interesting results. The trisaccharide β-D-Glcp-(1→2)-β-D-Glcp-(1→3)-α-D-Glcp-OMe showed an increased flexibility at the (1→2)-linkage.[75] Oligomers of the repeating unit of the type 37 CPS might also show an interesting tertiary structure and serve as a representative model of the native polymer in NMR studies.

2.1.3 Target substances

The question of how large an oligosaccharide needs to be to attain the tertiary structure of its native polymer has recently been addressed, both with spectroscopic[76] and immunologic methods. Such knowledge is of great importance for example in the development of polysaccharide vaccines.[77] The epitope involved in biological interactions may require that the saccharide adopts conformations not populated by an oligomer with to few repeats. In the case of the type 37 related structures, the synthesis has also been performed to generate substances with interesting structural features. Hence, a dimer, tetrasaccharide 1 is also an interesting target. However in terms of mimicking the steric pressure of the polymer, a hexasaccharide 2 is probably needed (Figure 2.2). The choice of an α-configured methyl glycoside as anomeric protecting group for the reducing end of the oligosaccharide is to provide a spectroscopic handle and reduce spectral overlap as much as possible.

15

O

OOMe

OOHO

O

HO HOHOHO

O OH

OHHO

HOO OH

OHHO

HO

21

O

OOMe

OO

O

HO HOHOHO

O OH

OHHO

HOO OH

OHHO

HO

O

OHO

HOHO

O OH

OHHO

HO

O

Figure 2.2 The two initial target substances.

2.2 Strategy for the synthesis

The type 37 polymer is a (1→3)-linked β-D-glucopyranosyl backbone, substituted with β-(1→2)-linked residues. A strategy based on simultaneous introduction of the β-(1→2)-linked residues to a (1→3)-linked diol/triol acceptor therefore seemed an efficient and logical protocol. If successful, it could potentially be extended for synthesis of larger analogues. The type 37 structures are not only homopolymers, they contain exclusively β-linkages as well. Consequently, participating and temporary protecting groups were needed at the 2-positions of the donors. These protecting groups could then be removed simultaneously to form an acceptor. In addition, an orthogonal protecting group at position 3 is needed for the backbone residues to facilitate its elongation. Although it worked for the synthesis of 1, the strategy was unsuccessful for the synthesis of 2, which is discussed in more detail in Section 2.4.

2.3 Synthesis of tetrasaccharide 1

2.3.1 Formation of the disaccharide acceptor

In our initial efforts to synthesize the first β-(1→3)-linkage, a NIS/TfOH-promoted[19] regioselective glycosylation of donor 3[78] and methyl 4,6-O-benzylidene-α-D-glucopyranoside[79] was employed (Scheme 2.1). This sequence reduces the number of protecting group manipulations. The regio- and stereoselectivity was good. Similar results have been reported by Zeng and co-

16

workers.[80] However, solubility and separation problems were a major drawback for this route when scaled up.

OBnO

AcO

OOPh

SEtO

HOHO

OOPh

OMe

O

HO

OOPh

OMe

OBnO

AcO

OOPh

O

3

i+

(73%)

Scheme 2.1 Regioselective glycosylation. i. NIS/TfOH, CH2Cl2, −35 °C.

Acceptor 4[81] is accessible in one step from methyl 4,6-O-benzylidene-α-D-glucopyranoside utilizing tin-activated benzoylation in almost quantitative yield.[81] NIS/TfOH-promoted glycosylation of 3 and 4 gave 5 in 83% yield (Scheme 2.2). In addition, this route was more effective if more labile protecting groups such as allyl and p-methoxybenzyl (PMB) ethers was present at the 3-postiton. The ester protecting groups were removed with sodium methoxide in methanol to form disaccharide acceptor 6.

OBnO

AcO

OOPh

SEtO

HOBzO

OOPh

OMe

O

R2O

OOPh

OMe

OBnO

R1O

OOPh

O

3 45 R1 = Ac, R2 = Bz (83%)

6 R1 = H, R2 = H (99%)

i

ii

+

Scheme 2.2 i. NIS/TfOH, CH2Cl2, 25 °C; ii. NaOMe, MeOH.

2.3.2 Formation and deprotection of tetrasacharide 1

Diglycosylation of 6 with thioglycoside donors employing NIS/TfOH or DMTST-promoted[15] conditions resulted in poor yields due to incomplete glycosylation forming a mixture of tri- and tetrasaccharides. However, the tetrasaccharide 8 could be produced by a AgOTf-promoted[20] glycosylation (Scheme 2.3) with 4.5 equivalents of 2,3,4,6-tetra-O-benzoyl-α-D-glucosyl bromide 7 in the presence of of sym-collidine at −55 °C.

17

O

HO

OOPh

OMe

OBnO

OH

OOPh

O

O

BzO

BzO

OBz

Br

BzO

OOOPh

OMe

OBnOO

OPh

O

OOBz

BzO

BzO

OBz

OOBz

BzO

BzO

OBz

OO

8 (84%)

1 (79%)ii, iii

+i

6

7

Scheme 2.3 i. AgOTf, CH2Cl2, sym-collidine, −55 °C, 5 h; ii. NaOMe, MeOH , 25 °C, 12 h; iii. H2, Pd/C, MeOH/HOAc/H2O: (6/1/1), 100 psi, 12 h.

Repeated additions of donor 7 and AgOTf throughout the glycosylation pushed the yield to 83%. However, the use of larger excesses of donor and promoter eventually resulted in the removal of the 4,6-benzylidene groups on the acceptor so the reaction must be carefully monitored. Subsequent deprotection of 8 by sodium methoxide and catalytic hydrogenation gave the deprotected tetrasaccharide 1 in 79% yield.

2.4 Synthesis towards hexasaccharide 2 - an abandoned approach

Encouraged by the success for the synthesis of 1, we tried to synthesize the hexasaccharide trimer with the polyglycosylation strategy. Hence, an analogous trisaccharide triol acceptor was constructed. By exchanging the benzyl ether at position 3 of donor 3 for a PMB group, elongation of the backbone is possible. Glycosylation of such a donor in analogy with the synthesis of 5, disaccharide 9 could be produced. By removing the PMB protecting group with DDQ[82] in water saturated chloroform, acceptor 10 is formed (Scheme 2.4). NIS/TfOH-promoted glycosylation of 3 with 10 followed by O-deacetylation with sodium methoxide in methanol gave triol 12.

18

O

AcO

OOPh

OMe

OPMBO

OAc

OOPh

O

O

AcO

OOPh

OMe

OHO

OAc

OOPh

O

O

RO

OOPh

OMe

OO

OR

OOPh

OO

BnOOR

OOPh

11 R= Ac

12 R= H

9 10

i

ii

iii

(87%)

(97%)

(99%)

+ 3

Scheme 2.4 i. DDQ, CHCl3-H2O; ii. NIS/TfOH, −30 °C, CH2Cl2; iii. NaOMe, MeOH.

The tri-glycosylation of 12 proved to be difficult. A number of glycosyl donors, promoters and conditions were explored leading only to either incomplete glycosylation or removal of the 4,6-O-benzylidene protecting groups on the acceptor. An O-benzylated analogue of acceptor 12 was synthesized, with the idea that a more durable acceptor could solve these problems. However, glycosylations employing this acceptor rendered similar results. Polyglycosylation proved unsuccessful for the synthesis of 2.

2.5 Synthesis of pentasaccharide 13

We decided to try a reversed order of glycosylation and introduce a β-(1→3) linked glucose residue to generate a pentasaccharide. Such a derivative, 13 (Figure 2.3), could be structurally interesting in and of it self. In addition, with proper protecting groups it could also provide a path to larger compounds such as hexasaccharide 2.

19

OHOHO

OMe

OO

HOHO

O

OOH

HO

HO

OH

OOH

HO

HO

OH

OO

OHOHO

HOHO

13

Figure 2.3

2.5.1 Synthesis of tetrasaccharide acceptors

Suitably protected acceptors were prepared by O-deacetylation of 9 and its 3´-O-allyl analogue. However, the results from the glycosylations of these acceptors were not satisfactory. NIS/TfOH-promoted conditions resulted in low yields and low reproducibility. In addition, AgOTf-promoted conditions which were successfully used in the formation of tetrasaccharide 8 were even worse and led mainly to side reactions. Instead, we turned to a strategy of reprotecting 8 at tetrasaccharide level. If the 4,6-O-benzylidene acetals were replaced with suitable protecting groups, the 3´-O-benzyl ether could be used as an orthogonal protecting group (Scheme 2.5). The change of protecting groups from benzoyl to acetyl esters on the two terminal residues is not without reason. Initially, benzoyl protecting groups were introduced directly after the removal of the 4,6-O-benzylidenes. However, upon treatment with H2, Pd-C, only small amounts of tetrasaccharide acceptor were formed. Reactions involving heterogeneous catalysts can be sensitive to steric interactions. As a consequence, a method employing NaBrO3/Na2S2O4 that should be more compatible with steric hindrance was also explored.[83] However, yields were low. Similar results from both methods were also obtained with the 4,6,4´,6´-tetra-O-acetylated analog. In addition, none of the substances formed during these reactions were functional as glycosyl acceptors. Eventually, The 4,6-O-benzylidene acetals in 8 were found to be efficiently cleaved with 90% TFA and treatment with sodium methoxide in methanol removed the benzoyl protecting groups. Subsequent O-acetylation gave 14 in 73% overall yield

20

(Scheme 2.4). On 14 however, the 3´-O-benzyl group was readily removed by treatment with H2, Pd-C producing acceptor 15 in 95% yield.

OOO

OMe

OBnOO

O

O

OOBz

BzO

BzO

OBz

OOBz

BzO

BzO

OBz

OO

OAcOAcO

OMe

ORO

AcOAcO

O

OOAc

AcO

AcO

OAc

OOAc

AcO

AcO

OAc

OO

Ph Ph

14 R = Bn (73%)

15 R = H (95%)iv

i,ii,iii

8

Scheme 2.5 i. 90% TFA 0.1 h; ii. NaOMe, MeOH; iii. Ac2O, pyridine, DMAP, CH2Cl2, 20 h; iv. H2, 5% Pd/C, THF, 100 psi, 30 h.

2.5.2 Synthesis and deprotection of the pentasaccharide

More importantly, when 15 were used as an acceptor in a NIS/TfOH-promoted glycosylation with a small donor, 16[84] the per-O-acetylated pentasaccharide 17 was formed in 58% yield (Scheme 2.6). Deprotection utilizing sodium methoxide in methanol gave 13 in 93% yield.

21

OAcOAcO

SEtAcOAcO

OAcOAcO

OMe

OO

AcOAcO

O

OOAc

AcO

AcO

OAc

OOAc

AcO

AcO

OAc

OO

OAcOAcO

AcOAcO

i

OHOHO

OMe

OO

HOHO

O

OOH

HO

HO

OH

OOH

HO

HO

OH

OO

OHOHO

HOHO

ii

13 (93%)

17 (58%)

15+

16

Scheme 2.6 i. NIS/TfOH, MS 4Å, CH2Cl2, 25°C; ii. NaOMe, MeOH.

2.6 Future possibilities

Evidently, the choice of small protecting groups is important. If a donor with a suitable orthogonal protecting group can be introduced instead of 16, a pentasaccharide acceptor is within reach, possibly allowing the synthesis of hexasaccharide 2. Whether or not these structures display structurally interesting properties in general or as mimics for the CPS of S. pneumoniae type 37 are left for future investigators.

22

3 Structural and immunochemical relationship between the O-antigenic polysaccharides from enteroaggregative Escherichia coli strain 396/C1 and Escherichia coli O126 (paper II)

3.1 Introduction

396/C1 is an E. coli strain which belongs to the enteroaggregative E. coli (EAEC) virotype. EAEC strains are believed to cause persistent diarrhea among infants, especially in the developing world[85]. Like many EAEC strains E. coli 396/C1 autoagglutinates. This initially precluded serotyping by standard methods such as slide agglutination. In such cases, characterization must be done by determination of the structure of the O-antigenic polysaccharide.

3.2 The E. coli 396/C1 O- antigen repeating unit

The 1H NMR spectrum of the polysaccharide from E. coli 396/C1 showed five significant resonances in the anomeric region. Assignment of 1H and 13C nuclei for the five residues were accomplished using an array of 1H,1H and 1H,13C 2D NMR techniques (see Section 1.2.4 and paper II). The sequence and glycosylation pattern i.e., the repeating unit, was determined by 1H,13C-HMBC and corroborated by 1H,13C-DDCCR and 1H,1H-NOESY (see Section 1.2.5). With the NMR data and a determination of the absolute configuration, the 396/C1 polysaccharide repeating unit was found to be a pentasaccharide (Figure 3.1).

→2)-β-D-Manp-(1→3)-β-D-Galp-(1→3)-α-D-GlcpNAc-(1→3)-β-D-GlcpNAc-(1→ │ α-L-Fucp-(1→2)

Figure 3.1 The repeating unit of the polysaccharide from E. coli 396/C1.

Component and methylation analysis were also performed and were in agreement with the NMR data.

23

3.3 Structural and immunological relationship with E. coli O126

Eventually, by employing a different agglutination method (see paper II) slide agglutination tests were possible. The result showed that E. coli 396/C1 was positive towards monospecific O126 serum and no selectivity was found in competitive tests using purified LPS from O126 reference strains and E. coli 396/C1. When submitted to the International Escherichia and Klebsiella Centre (WHO), the E. coli 396/C1 sample was found to be E. coli O126:K+:H27. The structure of the O-antigen from O126 has previously been described as a structurally very similar tetrasaccharide repeating unit (see Figure 3.2).[86]

→2)-β-D-Manp-(1→3)-α-D-Galp-(1→3)-β-D-GlcpNAc-(1→ │ β-L-Fucp-(1→2)

Figure 3.2 The previously described repeating unit for E. coli O126.

A brief inspection of the 1H NMR spectrum of the reference O126 LPS showed five anomeric resonances with very similar chemical shifts and coupling constants as compared to the E. coli 396/C1 sample. In addition, other protruding features such as NHAc and fucose-H6 resonances were also present. The reason for the discrepancy in the identification of the repeating unit of the E. coli O126 strains is yet to be explained.

3.4 The biological repeating unit

A more careful analysis of the 1H spectrum revealed additional signals of low intensity in the anomeric region. In combination with data primarily from TOCSY and HMBC, it was probable that these signals belonged to unsubstituted fucosyl and mannosyl[87] residues located at the terminus of the polymer. Consequently, we can deduce that the biological repeating unit of E. coli 396/C1 is initiated with a GlcpNAc residue (Figure 3.1). In many other cases if GlcpNAc is found, it often forms the first residue of the biological repeating unit[88].

24

5.05.15.2

1H (ppm)

B´

B

a b

Figure 3.3 (a) Part of the 1H NMR spectrum of E. coli 396/C1 PS showing the anomeric proton resonance of fucosyl residue B and the tentatively assigned terminal fucosyl residue here denoted as B’. (b) SDS/PAGE of an LPS isolate of E. coli 396/C1.

In addition, integration of the two anomeric fucosyl signals B and B´ allow us to comment on the length of the polymer. A ratio of 12:1 was found suggesting a length of ~13 repeating units. SDS/PAGE Gel electrophoresis was also performed on the E. coli 396/C1 LPS. The length of the polymer was found to be fairly dispersed, however a majority of the population contained 10-15 repeats; consistent with the NMR data (Figure 3.3).

25

4 Analysis of NMR J couplings in partially protected galactopyranosides (paper III).

4.1 Introduction

4.1.1 Synthesis

As mentioned in Section 1.1.2, protecting groups are of great importance in the stereochemical outcome of glycosylation reactions. For donors in the gluco- and manno-series the O2 protection groups are able to participate and thus control the anomeric configuration in the product. Galactose however, which has an axial hydroxyl at C4, is different since the protection group on O4 can also interfere.[9] A free hydroxyl group at this position can also have a directing effect.[89] In addition, galactosyl acceptors show a different reactivity depending on their anomeric configuration.[90] An interaction between the hydroxyl proton and the ring oxygen was judged to be important in both of these cases.

4.1.2 Spectroscopy

A 1H,1H-COSY crosspeak between HO4 and H5 was observed for several galactose derivatives.[89] The authors stated that the origin of this crosspeak is a hydrogen bond mediated scalar coupling from the hydroxyl proton via the ring oxygen (O5). We found these results interesting and they prompted us to investigate the role of the hydroxyl proton further. We were able to confirm a 1H,1H-COSY crosspeak between HO4 and H5 in two model substances, but the interpretation was not straightforward.[91]

4.2 The model substances and their NMR spectra

Two partially protected galactosides were selected as model substances: ethyl 2,3,6-tri-O-benzyl-1-thio-β-D-galactopyranoside 18[92] and methyl 2,3,6-tri-O-benzyl-α-D-galactopyranoside 19.[93] (Figure 4.1) They were synthesized via reductive benzylidene opening of the corresponding 2,3-di-O-benzyl-4,6-O-benzylidene derivatives in analogy with previously described methods.[93]

26

OO

O

HO

S

OH

OO

O

HO

O

OH

OO

O

HO

O

OH

O O

O

O

H

O

MeO

H

O O

O

O

O

H

θ

θ

18 19

20 21 22

Figure 4.1 The structure of the investigated molecules. The torsion angle θ = HO4-O4-C4-H4.

4.2.1 Evaluation of the experimental J-coupling constants

1D 1H and 2D 1H,1H-COSY NMR spectra were collected for both 18 and 19 in CDCl3 solution. (See Figure 4.2 for an example) The 1D spectra clearly showed a complex coupling pattern for the H4, HO4 and H5 resonances and a crosspeak between HO4 and H5 was unmistakably present in the 1H,1H-COSY spectrum. The hydroxyl proton in 18 for example, was a dd with JH4,HO4 = 2.6 Hz and JHO4,H5 = 1.0 Hz. (Figure 4.3)

Figure 4.2 The 1H NMR spectrum of 18 in CDCl3 at 25°C. The HO4 resonance is present at 2.50 ppm.

27

Using the relevant Karplus relationship, gauche rotamers show a reasonable agreement with the measured value of about 2 Hz, whereas if the trans rotamer (θ ≈ 180°, with the hydroxyl proton pointing towards the ring oxygen) was present to a large extent, a much higher value of the coupling constant would be expected. (Compare in Figure 1.6) In a freely rotating hydroxyl group, all angles of the θ torsion will be equally populated and hence averaged to a JH4,HO4 value of about 5 Hz. Thus, the low experimental value of this coupling (shown below) suggests that a gauche conformation is preferred and that a structure similar to 20t (see Figure 4.4) is not predominant in solution.

Figure 4.3 Part of the 1H NMR spectrum at 25 °C showing the HO4 resonance, δ 2.50, with JH4,HO4 = 2.6 Hz and JHO4,H5 = 1.0 Hz.

4.3 Optimization and analysis of the different rotamers

To map the source of the observed J coupling between HO4 and H5, we decided to calculate the properties for three static rotamers of 20, a simplified system more suitable for computational calculations. We used the three staggered rotamers of the hydroxyl group for optimization, namely, trans with the hydroxyl proton oriented towards the ring oxygen O5, gauche+ and gauche–. These rotamers are referred to as 20t, 20g+ and 20g– (see Figure 4.4 for the optimized structures of 20t, 20g+) as defined by the torsion angle θ = HO4-O4-C4-H4 (Figure 4.1). The ab initio method used was Density Functional Theory calculations (at the B3LYP/6-31G* level of theory).

28

By analyzing the electron density at the bond critical point for a selected interaction of an optimized structure such as 20t, information on whether a hydrogen bond mediated coupling pathway is likely or not can be obtained[55]. When structure 20t was analyzed with the Atoms In Molecules package,[94] no proof in terms of electron density was found supporting a hydrogen bond mediated coupling pathway. This is also supported by the large equilibrium interatomic distance of about 2.3 Å between the ring oxygen and the hydroxyl proton, too long to favour a strong hydrogen bond interaction. Looking at the two gauche structures, 20g+ and 20g–, the latter rotamer repeatedly converged to the 20g+ rotamer, despite the fact that several starting conformations were used. Interestingly, the relative energy of structure 20g+ was 3.5 kcal/mole lower than 20t due to more favourable hydrogen bonding of the hydroxyl group to O3 rather than to O5 (the ring oxygen). This suggests that 20g+ should be a major conformation also in solution.

Figure 4.4 The three optimized structures 20t (left), 20g+ (middle) and 22 (right).

To validate our results from the calculations above, a molecule for which experimental 3JH,HO couplings have been reported[95] was selected (21). The NMR data strongly suggested a trans conformation of the H4-C4-O4-HO4 torsion – as opposed to our results on galactopyranosides 18-20. A simplified model - 22 (see Figures 4.1 and 4.4) was then created. The computational results showed good agreement with the spectroscopic data, predicting trans geometry to be dominating in this case.

29

4.4 W-coupling

Four bond spin-spin couplings are normally small and therefore seldom observed. But if two protons, separated by four bonds are arranged as a W, then these couplings are often significant.[96, 97] Most organic chemists have seen examples of this as “meta-couplings” when interpreting spectra of aromatic compounds. However, 4J-couplings may also exist in nonaromatic compounds such as α-bromocyclohexanone (shown in Figure 4.5) as long as it fulfils the above mentioned criteria.[96]

O

OH

a bHc

Br

Ha

HbO H

Figure 4.5 (a) α-Bromocyclohexanone where Jab, Jac and Jbc are all about 1.1 Hz. (b) Schematic fragment of the investigated galactopyranosides highlighting the atoms in the W-arrangement which mediate the coupling between HO4 and H5.

The four atoms defining the torsion O4-C4-C5-H5 in galactopyranosides are practically anti-periplanar. As a result, when the hydroxyl proton is oriented in the gauche+ conformer, a W geometry exists between HO4 and H5. This pathway rather than a hydrogen bond mediated one is a plausible explanation for the observed couplings and crosspeaks in the 1H-1H-COSY spectra of our model substances and the related structures that initiated our interest in this subject.[89]

4.5 Comparison of experimental and calculated properties

4.5.1 IR spectra

The IR absorption frequencies of hydroxyl protons are very sensitive to hydrogen bonding. Variation in the O-H stretching frequencies is a classic indication of the presence of such interactions. IR can therefore provide additional information which can help in determining geometry and the direction of hydrogen bonds. In particular, since IR spectra can be calculated via ab initio methods for distinct conformers, the magnitudes of red- or blue shift, i.e., the decrease or increase,

30

respectively, of the O-H absorption frequencies can be compared to experimental data.

H2O 6-membered ring5-

Figure 4.6 Selected region of the IR spectrum of 18 in CH2Cl2 at ambient temperature with relevant absorptions marked.

IR spectra were recorded for 18 (Figure 4.6) and 19. Both showed a distinct O-H stretching band at 3565 cm-1 in CH2Cl2. This suggests that a cis five-membered ring is present to a large extent. The shoulder around 3500 cm-1 indicates that a six-membered ring also is present.[97] From the NMR JH5,H6 couplings the geometry of the hydroxymethyl group can be estimated. The population of a gg conformation of ω (O5–C5–C6–O6) required for a six-membered hydrogen bonded ring is low as predicted with IR spectroscopy. This explains the shoulder at 3500 cm-1. IR spectra for the different conformers of 22 and 20 were calculated. As expected, the two supposedly hydrogen-bonded conformers, i.e., 20g+ and 22t showed the largest relative redshift as compared to the the other conformers. However, for the two analyzed conformers of 20, the difference is small. As a result, it is difficult to experimentaly differentiate 20g+ from 20t via IR.

4.5.2 NMR J couplings

Spin-spin coupling constants were calculated for 20g+ and 20t at the UB3LYP/6-311G** level of theory. The results are compiled in Figure 4.7 together with the

31

corresponding experimental data of 18. Excellent agreement is observed for JH4,H5 as well as for the postulated large value of JH4,HO4 in 20t. It is shown that JHO4,H5 is significant in 20g+ but only marginal in 20t, thereby supporting our line of reasoning. Despite the fact that only static conformers are evaluated, i.e., no conformational averaging is accounted for as in NMR data, the agreement between simulation and experiment is quite satisfactory.

O

OH

H

HRO

ORSEt

OR2.6

1.01.0

O

OH

H

HRO

OROMe

OR0.04

1.81.6

O

O H

H

HRO

OROMe

OR12.2

1.6

-0.4

18

20g+ 20t

Figure 4.7 A comparison of the J coupling constants from the recorded NMR spectrum of 18 (R = Bn) and the calculated coupling constants from the static rotamers of 20 (R = Me).

32

4.5.3 Conclusion

We do not find it necessary to invoke a “virtual” bond and strong hydrogen bonding via the ring oxygen as proposed by Kwon and Danishefsky.[89] A straightforward interpretation of the observed JHO4,H5 as a classic W-mediated coupling is to be preferred. In essence, due to these five points: • There is no support for electron density at the hydrogen bond critical point as

deduced by ab initio calculations • The relative energies between conformations • The magnitude of 3JHO4,H4 • The likelihood of significant 4J couplings when W arrangements are present,

also in saturated systems • The calculated J couplings, in particular 4JHO4,H5 The discussion of these galactopyranosides concerns their conformation in solution. The picture may however be different for the high-energy transition state species formed as new glycosidic bonds are constructed. Even so, the JH4,HO4 coupling cannot be considered as an argument for a strong hydrogen bond interaction via the ring oxygen.

33

5 Determination of the conformational flexibility of methyl α-cellobioside in solution by NMR spectroscopy and molecular simulations (paper IV)

5.1 Introduction

The cellobiose disaccharide is a fragment from the nearly omnipresent cellulose polymer. As a consequence conformational studies of cellobiosides have reoccurred many times in the literature.[98-103] In an earlier investigation, an increased flexibility of the ψ-torsion was found.[99] In a recent study, employing ab initio calculations of β-cellobiose, hydrogen bonds were found to have a great impact on the conformation.[100, 101] Interestingly, in this case the anti-φ conformation was found to be the most stable. It has however, not before been confirmed by NMR spectroscopy.

OHO

HO

HOOH

O O

OMeHO

HO

OHH R

ψφ

23 23-d

R = HR = D

Figure 5.1 Methyl α-cellobioside (23) and its deuterated analogue (23-d).

In the present investigation, the conformational flexibility in solution was studied with 1H,1H T-ROESY experiments in combination with molecular dynamics simulations. Detection and quantification of anti-φ and anti-ψ conformations was performed.

5.2 1D DPFGSE T-ROESY

Employing 1D DPFGSE T-ROESY[104] with a selective excitation of H1´ generated the proton-proton cross relaxation rates to H3, H4 and H5 in D2O and DMSO-d6. The cross relaxation rates were interpreted using both ISPA[52] and the model-free

34

approach[53, 54], both methods generating very similar distances. In addition, via excitation of H2´, rH2´,H4 were also calculated using rH2´,H4´ as reference distance. To avoid possible misinterpretation of the H1´-H3 and H1´-H5 relaxation rates, deuterium labeling was performed[105] (23-d) to suppress three-spin effects between H1´ and H4.

OHO

HO

HOOH

O O

OMeHO

HO

OHH H

OHO

HO

HOOH

O

H OHO

HO

OH

OMe

HO

HOHO

HO

OHO

OH

OD

O

HOOMe

HH

syn anti-ψ anti-φ

Figure 5.2 The three conformations and the key inter-glycosidic proton-proton pairs for the determination of their relative populations.

5.3 Deuterium isotope effects

In addition to the measurement of inter-glycosidic proton-proton distances, methyl α-cellobioside was also screened for deuterium isotope effects.[106, 107] In aprotic solvents such as DMSO, observation of hydroxyl-protons is possible.

ODO

O

DOOD

O

OO

DO

OD

OMe

D

HO

DOO

DOOD

O

O

OOD

D

HOMe

DO

23 24

Figure 5.3 Methyl α-cellobioside 23 and β-D-Galp-(1→3)-β-D-Glcp-OMe, 24. Partially deuterium exchanged (H/D 2/1) samples were prepared in DMSO-d6. Substance 24 displayed deuterium isotope effects whereas none were observed for 23.

Partially deuterium exchanged samples can, if a hydrogen-bond is present, give rise to deuterium isotope effects and thereby act as a sensor for specific conformations. We were able to reproduce the results of Dabrowski and co-workers[106, 107] on β-D-Galp-(1→3)-β-D-Glcp-OMe, 24 but could not detect such effects for 23.

35

5.4 Computer simulations

To provide support for the interpretation of the NMR data and to investigate the conformational space of 23, computer simulations were performed. Following initial Monte Carlo (MC) simulation using the HSEA approach, a molecular dynamics simulation employing the PARM22 force field with explicit water as solvent was performed. In addition to the syn conformation, the simulation also revealed the presence of anti-φ and anti-ψ conformations.

5.5 Population analysis

The T-ROE derived distances from the experiments performed in D2O were evaluated in combination with data from the MD simulation with respect to the three conformations (see table 5.5). Employing a three-state analysis, relative populations of 93%, 5% and 2% of the syn, anti-ψ and anti-φ conformer respectively were present.

Table 5.5 Distances obtained from simulation, minimization and NMR experiments.

Proton pair <r>syn <r>anti-ψ ranti-φ rexp H1',H4 2.39 3.58 3.53 2.29 H1',H3 4.45 2.26 4.45 3.58 H2',H4 4.41 4.71 2.22 3.90

<r>syn and <r>anti-ψ are average distances extracted from the MD simulation. ranti-φ is obtained by energy minimization. rexp is derived from NMR data with ISPA. All distances are given in Å.

36

Acknowledgements

Many people have helped me in various ways since I started here back in ´99. Hence, I would

like to acknowledge:

My supervisor, Professor Göran Widmalm for the knowledge, encouragement, and patience.

Professor Jan Erling Bäckvall and professor emeritus Per J. Garegg for kind intrest in our work.

Andrei Weintraub for good collaboration.

Past and present members of the Widmalm group; Mange, Tina, Classe, Micke, Chrille, Peter,

Glen, Malin, Torgny, Smühl (for the “Berlin” wall), Rob, Jennie, Ulrika and Tara.

Tjockisarna Mia, Peter and Rolle For nice lunch company and stimulating discussions of

chemistry, football and other important things.

Micke, Classe, Tina, Smühl and Jennie for the crash course in NMR

The former habitants of Lab 7.

Glen for linguistic improvement.

And the rest of the department for providing a pleasant environment to work in!

Dr. Jan Lindberg and Professor Peter Konradsson for the introduction to carbohydrate synthesis.

Britt, Pia, Kristina, Robin and Olle for help in various ways.

All off-campus friends

My family.

Anna

I would also like to acknowledge the financial support from: Nils Löfgrens fond, Svenska

kemistsamfundet, Wallenbergsstiftelsens jubileumsfond and C. F. Liljevalch Jr:s stipendiefond.

37

References [1] A. Varki, Glycobiology 1993, 3, 97-130. [2] M. Färnbäck, L. Eriksson, G. Widmalm,

Acta Crystallographica Section C 2003, C59, 171-173.

[3] I. Tvaroska, T. Bleha, Advances in Carbohydrate Chemistry and Biochemistry 1989, 47, 45-123.

[4] P. P. Graczyk, M. Mikolajczyk, in Topics in Stereochemistry, Vol 21, Vol. 21, 1994, pp. 159-349.

[5] R. U. Lemieux, K. B. Hendriks, R. V. Stick, K. James, Journal of the American Chemical Society 1975, 97, 4056-4062.

[6] P. Söderman, E. A. Larsson, G. Widmalm, European Journal of Organic Chemistry 2002, 1614-1618.

[7] A. Demchenko, T. Stauch, G. J. Boons, Synlett 1997, 818-820.

[8] I. Braccini, C. Derouet, J. Esnault, C. H. Dupenhoat, J. M. Mallet, V. Michon, P. Sinay, Carbohydrate Research 1993, 246, 23-41.

[9] A. V. Demchenko, E. Rousson, G. J. Boons, Tetrahedron Letters 1999, 40, 6523-6526.

[10] N. M. Spijker, C. A. A. Vanboeckel, Angewandte Chemie-International Edition in English 1991, 30, 180-183.

[11] D. Crich, W. L. Cai, Journal of Organic Chemistry 1999, 64, 4926-4930.

[12] K. H. Jung, M. Muller, R. R. Schmidt, Chemical Reviews 2000, 100, 4423-4442.

[13] R. W. Binkley, M. G. Ambrose, Journal of Carbohydrate Chemistry 1984, 3, 1.

[14] R. R. Schmidt, W. Kinzy, in Advances in Carbohydrate Chemistry and Biochemistry, Vol 50, Vol. 50, 1994, pp. 21-123.

[15] F. Andersson, P. Fügedi, P. J. Garegg, M. Nashed, Tetrahedron Letters 1986, 27, 3919-3922.

[16] G. H. Veeneman, J. H. van Boom, Tetrahedron Letters 1990, 31, 275-278.

[17] P. Fügedi, P. J. Garegg, Carbohydrate Research 1986, 149, C9-C12.

[18] H. Lönn, Journal of Carbohydrate Chemistry 1987, 6, 301-306.

[19] G. H. Veeneman, S. H. van Leeuwen, J. H. van Boom, Tetrahedron Letters 1990, 31, 1331-1334.

[20] S. Hanessian, J. Banoub, Methods in Carbohydrate chemistry 1980, 8, 247-250.

[21] P. J. Garegg, in Advances in Carbohydrate Chemistry and Biochemistry, Vol 52, Vol. 52, 1997, pp. 179-205.

[22] D. J. Harvey, Mass Spectrometry Reviews 1999, 18, 349-450.

[23] C. V. Robinson, S. E. Radford, Structure 1995, 3, 861-865.

[24] R. U. Lemieux, R. K. Kullnig, J. H. Bernstein, W. G. Schneider, Journal of the American Chemical Society 1958, 80, 6098-6105.

[25] J. Ø. Duus, C. H. Gotfredsen, K. Bock, Chemical Reviews 2000, 100, 4589-+.

[26] M. Staaf, F. Urbina, A. Weintraub, G. Widmalm, European Journal of Biochemistry 1999, 262, 56-62.

[27] E. J. Faber, J. P. Kamerling, J. F. G. Vliegenthart, Carbohydrate Research 2001, 331, 183-194.

[28] J. S. Sawardeker, J. H. Sloneker, A. Jeanes, Analytical Chemistry 1965, 37, 1602-1604.

[29] G. J. Gerwig, J. P. Kamerling, J. F. G. Vliegenthart, Carbohydrate Research 1979, 77, 1-7.

[30] L. E. Kay, P. Keifer, T. Saarinen, Journal of the American Chemical Society 1992, 114, 10663-10665.

[31] K. Bock, I. Lundt, Tetrahedron Letters 1973, 13, 1037-1040.

[32] K. Bock, C. Pedersen, Journal of the Chemial Society, Perkin Transactions 2 1974, 3, 293-297.

[33] G. Widmalm, in Carbohydrate Chemistry (Ed.: G. J. Boons), Blackie Academic and Proffesional, London, 1998, pp. 450-451.

[34] W. P. Aue, E. Bartholdi, R. R. Ernst, The Journal of Chemical Physics 1976, 64, 2229-2246.

[35] C. Griesinger, G. Otting, K. Wüthrich, R. R. Ernst, Journal of the American Chemical Society 1988, 110, 7870-7872.

38

[36] L. Braunschweiler, R. R. Ernst, Journal of Magnetic Resonance 1983, 53, 521-528.

[37] J. Jeener, B. H. Meier, P. Bachman, R. R. Ernst, Journal of Chemical Physics 1979, 71, 4546-4553.

[38] T. Domke, Journal of Magnetic Resonance 1991, 95, 174-177.

[39] A. Bax, M. F. Summers, Journal of the American Chemical Society 1986, 108, 2093-2094.

[40] S. J. F. Vincent, C. Zwahlen, Journal of the American Chemical Society 2000, 122, 8307-8308.

[41] S. Perez, M. Kouwijzer, K. Mazeau, S. B. Engelsen, Journal of Molecular Graphics 1996, 14, 307-321.

[42] A. J. Kirby, N. H. Williams, The Anomeric effect and associated stereoelectronic effects, American Chemical Society, Washington, 1993.

[43] M. Färnbäck, L. Eriksson, G. Widmalm, Acta Crystallographica Section C-Crystal Structure Communications 2003, 59, O171-O173.

[44] K. J. Naidoo, J. W. Brady, Journal of the American Chemical Society 1999, 121, 2244-2252.

[45] J. E. Williams, P. J. Stand, P. R. Schleyer, Annual Review of Physical Chemistry 1968, 19, 531-558.

[46] S. Perez, A. Imberty, S. B. Engelsen, J. Gruza, K. Mazeau, J. Jiménez-Barbero, A. Poveda, J.-F. Espinosa, B. P. van Eyck, G. Johnson, Carbohydrate Research 1998, 314, 141-155.

[47] R. Eklund, G. Widmalm, Carbohydrate Research 2003, 338, 393-398.

[48] R. Stuike-Prill, B. Meyer, European Journal of Biochemistry 1990, 194, 903-919.

[49] D. Neuhaus, M. P. Williamson, The nuclear Overhauser effect in structural and conformational analysis, Weinheim: VCH, New York, 1989.

[50] T. L. Hwang, A. J. Shaka, Journal of Magnetic Resonance Series B 1993, 102, 155-165.

[51] T. L. Hwang, A. J. Shaka, Journal of the American Chemical Society 1992, 114, 3157-3159.

[52] J. W. Keepers, T. L. James, Journal of Magnetic Resonance 1984, 57, 404-426.

[53] G. Lipari, A. Szabo, Journal of the American Chemical Society 1982, 104, 4546-4559.

[54] G. Lipari, A. Szabo, Journal of the American Chemical Society 1982, 104, 4559-4570.

[55] A. J. Dingley, F. Cordier, S. Grzesiek, Concepts in Magnetic Resonance 2001, 13, 103-127.

[56] M. Karplus, Journal of Chemical Physics 1958, 30, 11-15.

[57] R. R. Fraser, M. Kaufman, P. Morand, Canadian Journal of Chemistry 1969, 47, 403-409.

[58] F. Cloran, I. Carmichael, A. S. Serianni, Journal of the American Chemical Society 1999, 121, 9843-9851.

[59] T. Y. Lin, H. C. Chang, R. J. Lagow, Journal of Organic Chemistry 1999, 64, 8127-8129.

[60] B. Luy, U. Richter, E. S. DeJong, O. W. Sorensen, J. P. Marino, Journal of Biomolecular Nmr 2002, 24, 133-142.

[61] G. Gemmecker, Angewandte Chemie-International Edition 2000, 39, 1224-1226.

[62] A. J. Dingley, S. Grzesiek, Journal of the American Chemical Society 1998, 120, 8293-8297.

[63] M. Fierman, A. Nelson, S. I. Khan, M. Barfield, D. J. O'Leary, Organic Letters 2000, 2, 2077-2080.

[64] F. Cordier, M. Rogowski, S. Grzesiek, A. Bax, Journal of Magnetic Resonance 1999, 140, 510-512.

[65] A. J. Dingley, J. E. Masse, J. Feigon, S. Grzesiek, Journal of Biomolecular Nmr 2000, 16, 279-289.

[66] F. Lohr, S. G. Mayhew, H. Ruterjans, Journal of the American Chemical Society 2000, 122, 9289-9295.

[67] D. Llull, R. Munoz, R. Lopez, E. Garcia, Journal of Experimental Medicine 1999, 190, 241-251.

[68] M. Sicard, A. M. Gasc, P. Giammarinaro, J. Lefrancois, F. Pasta, M. Samrakandi, Research in Microbiology 2000, 151, 407-411.

39

[69] Y. Mizrachi Nebenzahl, N. Porat, S. Lifshitz, S. Novick, A. Levi, E. Ling, O. Liron, S. Mordechai, R. K. Sahu, R. Dagan, FEMS Microbiology Letters 2004, 233, 147-152.

[70] A. D. Magee, J. Yother, Infection and Immunity 2001, 69, 3755-3761.

[71] D. Llull, E. Garcia, R. Lopez, Journal of Biological Chemistry 2001, 276, 21053-21061.

[72] E. Garcia, R. Lopez, FEMS Microbiology Letters 1997, 149, 1-10.

[73] A. Adeyeye, P.-E. Jansson, B. Lindberg, Carbohydrate Research 1988, 180, 295-299.

[74] P. Söderman, P. E. Jansson, G. Widmalm, Journal of the Chemical Society-Perkin Transactions 2 1998, 639-648.

[75] C. Landersjö, R. Stenutz, G. Widmalm, Journal of the American Chemical Society 1997, 119, 8695-8698.

[76] J. R. Brisson, S. Uhrinova, R. J. Woods, M. van der Zwan, H. C. Jarrell, L. C. Paoletti, D. L. Kasper, H. J. Jennings, Biochemistry 1997, 36, 3278-3292.

[77] M.-J. Clement, A. Imberty, A. Phalipon, S. Perez, C. Simenel, L. A. Mulard, M. Delepierre, J. Biol. Chem. 2003, 278, 47928-47936.

[78] P. J. Garegg, L. Olsson, S. Oscarson, Journal of Organic Chemistry 1995, 60, 2200-2204.

[79] P. Calinaud, J. Gelas, in Preparative carbohydrate chemistry (Ed.: S. Hanessian), Marcel Deller, New York, 1997, pp. 3-33.

[80] Y. Zeng, F. Z. Kong, Carbohydrate Research 2003, 338, 843-849.

[81] S. David, A. Thieffry, Journal of the Chemical Society Perkin transactions 1 1979, 1568-1573.

[82] Oikawa, Yuji, T. Yoshioka, O. Yonemitsu, Tetrahedron Letters 1982, 23, 885-888.

[83] M. Adinolfi, G. Barone, L. Guariniello, A. Iadonisi, Tetrahedron Letters 1999, 40, 8439-8441.

[84] R. J. Ferrier, R. H. Furneaux, Carbohydrate Research 1976, 52, 63-68.

[85] J. P. Nataro, T. Steiner, R. L. Guerrant, in Emerging Infectious Diseases, Vol. 4,

Centers for Disease Control & Prevention (CDC), 1998, pp. 251-261.

[86] T. Bhattacharyya, S. Basu, Carbohydrate Research 1994, 254, 221-228.

[87] P.-E. Jansson, L. Kenne, G. Widmalm, Carbohydrate Research 1989, 188, 169-191.

[88] D. Alexander, M. Valvano, J. Bacteriol. 1994, 176, 7079-7084.

[89] O. Kwon, S. J. Danishefsky, Journal of the American Chemical Society 1998, 120, 1588-1599.

[90] D. Magaud, R. Dolmazon, D. Anker, A. Doutheau, Y. L. Dory, P. Deslongchamps, Organic Letters 2000, 2, 2275-2277.

[91] E. A. Larsson, J. Ulicny, A. Laaksonen, G. Widmalm, Organic Letters 2002, 4, 1831-1834.

[92] P. J. Garegg, I. Kvarnström, A. Niklasson, G. Niklasson, S. C. T. Svensson, Journal of Carbohydrate Chemistry 1993, 12, 933-953.

[93] P. J. Garegg, H. Hultberg, S. Wallin, Carbohydrate Research 1982, 108, 97-101.

[94] R. F. W. Bader, Atoms in Molecules. A Quantum Theory, Claredon Press: Oxford, 1990.

[95] F. J. Luque, J. M. Lopez, M. L. de la Paz, C. Vicent, M. Orozco, Journal of Physical Chemistry A 1998, 102, 6690-6696.

[96] H. Günther, in NMR spectroscopy, 2nd ed., Wiley: Chichester, 1992, p. 124.

[97] P. R. Muddasani, E. Bozo, B. Bernet, A. Vasella, Helvetica Chimica Acta 1994, 77, 257-290.

[98] R. Bentley, Journal of the American Chemical Society 1959, 81, 1952-1956.

[99] G. M. Lipkind, A. S. Shashkov, N. K. Kochetkov, Carbohydrate Research 1985, 141, 191-197.

[100] G. L. Strati, J. L. Willett, F. A. Momany, Carbohydrate Research 2002, 337, 1833-1849.

[101] G. L. Strati, J. L. Willett, F. A. Momany, Carbohydrate Research 2002, 337, 1851-1859.

[102] B. J. Hardy, A. Gutierrez, K. Lesiak, E. Seidl, G. Widmalm, Journal of Physical Chemistry 1996, 100, 9187-9192.

40

[103] M. M. Kuttel, Doctoral thesis, University of Cape Town (Cape Town), 2003.

[104] A. Kjellberg, G. Widmalm, Biopolymers 1999, 50, 391-399.

[105] P. Söderman, G. Widmalm, Journal of Organic Chemistry 1999, 64, 4199-4200.

[106] J. Dabrowski, H. Grosskurth, C. Baust, N. E. Nifant'ev, Journal of Biomolecular Nmr 1998, 12, 161-172.

[107] J. Dabrowski, T. Kozar, H. Grosskurth, N. E. Nifantev, Journal of the American Chemical Society 1995, 117, 5534-5539.