Macrolide antibiotics as anti-bacterial and potential anti ...

Macrolide antibiotics as anti-bacterial andpotential anti-malarial medicines

A thesis submitted to The University of Manchester for the degree of Doctor ofPhilosophy in the Faculty of Medical and Human Sciences

2012

Biljana Arsic

School of Pharmacy and Pharmaceutical Sciences

2

Contents

ABSTRACT 4DECLARATION 5COPYRIGHT STATEMENT 6ACKNOWLEDGEMENTS 7

I Introduction 8

Chapter 1: Macrolide antibiotics in the treatment of bacterial and protozoal diseases 81.1 Macrolide antibiotics: discovery and structures 81.1.1. 14-Membered macrolide antibiotics: erythromycin A, 6-O-methyl erythromycin Aand their derivatives 8

1.1.2. 15-Membered macrolide antibiotics: 9a-methyl-9-deoxo-9-dihydro-9a-aza-9a-homoerythromycin A, 6-O-methyl homoerythromycins and their derivatives 10

1.1.3. 16-Membered macrolide antibiotics: tylosin A, josamycin and their derivatives 121.2. Macrolide antibiotics as antibacterial agents 151.2.1. The mode of action of 14-membered macrolide antibiotics (erythromycin A,clarithromycin) and effect of substitution on their mode of action 15

1.2.2. Mode of action of 15-membered macrolide antibiotics (azithromycin) and theeffect of substitution on their mode of action 18

1.2.3. Modifications of erythromycin A and their anti-tuberculosis activity 211.2.4. Spiramycin, tylosin A and tylosin B derivatives and their mode of action 221.3. Malaria and other protozoal diseases 251.3.1. Introduction to protozoa 251.3.1.1. Toxoplasmosis 251.3.1.2. Amoebic meningoencephalitis 261.3.1.3. Human African trypanosomiasis 271.3.1.4. Cutaneous leishmaniasis 271.3.1.5. Malaria 281.3.1.5.1. Anti-malarial medicines and drug resistance 301.4. Macrolide antibiotics used in the treatment of malaria 311.4.1. Azithromycin alone and in the combination with existing anti-malarial medicines 311.5. References 32Chapter 2: Computational and other supporting techniques in drug discovery andevaluation 35

2.1. Computational techniques in drug discovery 352.1.1. Molecular mechanics and dynamics 352.1.1.1. Fundamental concepts of molecular dynamics and mechanics 352.1.2.Molecular modelling of proteins 45Sequence analysis 45Molecular modelling softwares and resources 45Secondary structure prediction 50Tertiary structure prediction 512.1.2.1. Template-based modelling (homology modelling) 522.1.2.2. Free modelling (ab initio or de novo modelling) 522.1.3.Molecular modelling of RNAs 59Typical RNA secondary structures 59Secondary Structure 60Homology principle 602.1.3.1. Molecular modelling of the secondary structure of 23S rRNA 602.1.3.2. Molecular modelling of the 3D structure of 23S rRNA 612.1.4. Molecular docking 632.2. Supporting techniques used for the determination of macrolide antibiotics inchemical and biological media 68

2.2.1. Nuclear magnetic resonance techniques 682.2.2. Mass spectrometry 762.2.3. Microbiological techniques for the determination of drug activity 792.2.3.1. Minimum inhibitory concentration (MIC) technique 802.3. References 82Aims 89

3

II Results and Discussion 90

Chapter 3: Theoretical and experimental investigation on clarithromycin, erythromycin Aand azithromycin and descladinosyl derivatives of clarithromycin and azithromycin with3-O-substitution as anti-bacterial agents

90

Chapter 4: Free and bound state structures of 6-O-methyl homoerythromycins andepitope mapping of their interactions with ribosomes 91

Chapter 5: Structures and antibacterial activity of tylosin A and tylosin B 92Chapter 6: Anti-malarial activity of macrolide antibiotics: in silico study on the apicoplastribosomal exit tunnel of Plasmodium falciparum 93

6.1. Introduction 936.2. Results and Discussion 966.2.1. Modelling on apicoplast-encoded P. falciparum L4 and nuclear genome-encodedP. falciparum L22 ribosomal proteins 96

6.2.1.1.Construction of a segment of L4 ribosomal protein 966.2.1.2. Construction of the L22 ribosomal protein 996.2.2. Modelling on apicoplast 23S rRNA from P. falciparum 1006.2.3. Construction of the exit tunnel of the apicoplast ribosomal exit tunnel ofPlasmodium falciparum 101

6.3. Discussion 1036.4. Experimental 1046.5. Conclusion 1056.6. References 106

III Conclusion 108

IV Appendix 110

Appendix 1 110Appendix 2 1111D and 2D NMR spectra of descladinosyl clarithromycin 1111D and 2D NMR spectra of descladinosyl azithromycin 1201D and 2D NMR spectra of tylosin A in phosphate buffered D2O, apparent pH=7 1301D and 2D spectra of tylosin A in CDCl3 1391D and 2D spectra of tylosin B in phosphate buffered D2O (apparent pH=7) 1471D and 2D spectra of tylosin B in CDCl3 153Acquisition and processing NMR parameters 160

Total word count: 53 687

4

The University of ManchesterBiljana ArsicDoctor of PhilosophyMacrolide antibiotics as anti-bacterial and potential anti-malarial medicines2012

ABSTRACT

Macrolide antibiotics are known as anti-bacterial agents. Erythromycin A,14-membered macrolide antibiotic is known to exist in two forms-ketone(active) and hemiacetal form (inactive). It shows mild flexibility in silico. Itsderivative, clarithromycin, 6-O-methyl erythromycin A, shows rigidity andactivity against Gram-positive bacteria. The semisynthetic derivative oferythromycin A, azithromycin, 15-membered macrolide antibiotic, showsflexibility in silico and activity against Gram-negative bacteria. The combinationof molecular modelling (molecular mechanics and/or molecular dynamics) withTRNOESY NMR data give us the active conformation of flexible molecules.Constraining the strong intramolecular hydrogen bonds can be helpful in thedetermination of the active conformation of the drug. We have developedmodelling strategy for the construction of new 14- and 15-membered macrolideantibiotics with desired activity.

Tylosin A and tylosin B, 16-membered macrolide antibiotics, showrigidity in silico. However, tylosin A is very unstable in aqueous solutions, soprecise determination of hydrogen and carbon chemical shifts is extremelydifficult. Nobody else before us tried to publish the full assignments of thiscompound in D2O. Accurate determination of hydrogen and carbon chemicalshifts is necessary in order to further explore the properties of this compound.Anti-bacterial activity investigation of tylosin A and its derivative, tylosin B,shows lower activity both against Gram-positives and Gram-negativescompared to clarithromycin and azithromycin. Superposition of two moleculesof azithromycin with one molecule of tylosin A reveals that two molecules ofazithromycin actually occupy the space of one tylosin A molecule, which canexplain found anti-malarial activity of tylosin A (both azithromycin and tylosin Ashow similar contacts to bacterial ribosomes).

Clinical trials show that azithromycin has an anti-malarial activity. Inorder to investigate the potential anti-malarial activity of macrolide antibiotics,we had to construct the exit tunnel of the apicoplast ribosome from Plasmodiumfalciparum. Because of the unavailability of the crystal structure of P. falciparumribosome (it is impossible to separate mitochondria and apicoplast, and both ofthem contain ribosomes), we used different computational methods andsoftwares in order to construct it. We used both homology modelling and abinitio modelling server for the construction of L4 (apicoplast-encoded P.falciparum ribosomal protein) and L22 (nuclear genome-encoded P. falciparumribosomal protein) and RNA_2D3D software to construct the 23S rRNA fromapicoplast ribosome of Plasmodium falciparum. Using Pymol software andMOE we have constructed the exit tunnel of apicoplast ribosome from P.falciparum. The model shows that it can bind one azithromycin molecule. It isthe first model of the exit tunnel of the apicoplast ribosome from Plasmodiumfalciparum. Further work can be extended to the docking of other moleculesthan azithromycin into the modelled exit tunnel of Plasmodium falciparum.

5

DECLARATION

No portion of the work referred to in the thesis has been submitted in support of

an application for another degree or qualification of this or any other University or other

institute of learning.

6

COPYRIGHT STATEMENT

i. The author of this thesis (including any appendices and/or schedules to this thesis)

owns certain copyright or related rights in it (the “Copyright”) and s/he has given The

University of Manchester certain rights to use such Copyright, including for

administrative purposes.

ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic

copy, may be made only in accordance with the Copyright, Designs and Patents Act

1988 (as amended) and regulations issued under it or, where appropriate, in

accordance with licensing agreements which the University has from time to time. This

page must form part of any such copies made.

iii. The ownership of certain Copyright, patents, designs, trade marks and other

intellectual property (the “Intellectual Property”) and any reproductions of copyright

works in the thesis, for example graphs and tables (“Reproductions”), which may be

described in this thesis, may not be owned by the author and may be owned by third

parties. Such Intellectual Property and Reproductions cannot and must not be made

available for use without the prior written permission of the owner(s) of the relevant

Intellectual Property and/or Reproductions.

iv. Further information on the conditions under which disclosure, publication and

commercialisation of this thesis, the Copyright and any Intellectual Property and/or

Reproductions described in it may take place is available in the University IP Policy

(see http://www.campus.manchester.ac.uk/medialibrary/policies/intellectual-

property.pdf), in any relevant Thesis restriction declarations deposited in the University

Library, The University Library’s regulations (see

http://www.manchester.ac.uk/library/aboutus/

7

ACKNOWLEDGEMENT

Firstly, I would like to express my deep gratitude to my supervisor, Dr Jill

Barber who gave me advices and suggestions throughout my thesis. I would like to

give my special thanks to my advisor, Dr Kaye Williams.

Also, I would like to thank to Professor Gareth Morris and his NMR group at

School of Chemistry, The University of Manchester for their help in recording and

processing complicated NMR spectra. I carried out most of my molecular modelling

work in Molecular modelling lab under the guidance of Dr Richard Bryce, whom I

would like to express special thanks. Special thanks also to all current and previous

members of Drug Design group for their valuable help and discussions. Work with

Macromodel would be puzzle for me without the help of Dr Andrew Regan to whom I

owe my special thanks. Microbiological work will be extremely difficult for me without

the help and instructions given by Dr Andrew McBain and his research group. I carried

out microbiological work in their lab and they provided me free bacterial samples. Also,

I would like to express my thanks to Mass Spectrometry group at School of Chemistry,

The University of Manchester for recording mass spectra and providing computers for

processing of some NMR spectra.

Also, I owe special thanks to Dr Elena V. Bichenkova and her previous and

current group members. I did part of my synthetic work in the lab of Dr Sally Freeman

to whom I owe special thanks and her group members for valuable help and

suggestions.

I owe special gratitude to The University of Manchester for providing Overseas

Research Scholarship (ORS) and School of Pharmacy and Pharmaceutical Sciences

for granting (URS) towards my living expenses.

Special thanks to all past and present colleagues from my research group led

by Dr Jill Barber.

Here, I also would like to thanks to Zahra Hamrang who went through my

Chapter 1 and Chapter 6 and corrected English, Dr Nicholas Rattray who corrected

English in the draft of Chapter 1, Dr Neil Bruce who went through my Chapter 2 and

gave me suggestions and Dr Manikandan Kadirvel who went through my thesis and

gave me suggestions. I would like to thank also the security staff of the School of

Pharmacy and Pharmaceutical Sciences.

I would like to thank all my friends who have motivated me to complete my

thesis. Finally to my family I would like to express my special gratitude for their

constant understanding and encouragements.

8

I Introduction

Chapter 1: Macrolide antibiotics in the treatment of bacterial and protozoaldiseases1.1 Macrolide antibiotics: discovery and structures1.1.1. 14-Membered macrolide antibiotics: erythromycin A, 6-O-methyl

erythromycin A and their derivativesErythromycin A

Discovery of penicillin in 1928 meant the revolution in the treatment of

respiratory diseases and many infections. However, it was later proven that allergic

responses may occur in some patients and hence, the demand for discovery of novel

effective substitutes increased. One of the most successful and the cheapest

substitutes of penicillin was erythromycin A.

In 1952, McGuire et al. from Eli Lilly & Company discovered a compound they

named erythromycin A, which shows anti-bacterial activity both against Gram-positive

and Gram-negative organisms. Erythromycin A is the product of fermentative

processes in soil containing sources of carbohydrates, inorganic salts and nitrogen.

The bacterial strain M5-12559, originally known as Streptomyces erythreus, was the

producer of erythromycin A and belongs to the actinomycete class.1

Erythromycin A exists in both protonated (>96%) and neutral (<4%) forms at

physiological pH because of the pKa of the dimethylamino group (pKa=8.8). The

structure of erythromycin A is represented in Figure 1 as follows:

O

O

OHHO OH

O

O

O

O

N

RO

OOH

O

12

34

16

175

67

8

18

19910

20

11

122113

14

15

1'

2' 3'4'

5'

1" 2" 3"

4"5"

R=H erythromycin AR=

C

O

CH2CH2 C

O

O CH2CH3

erythromycin A 2'-ethyl succinate

Figure 1: Structure of erythromycin A

9

Erythromycin A possesses good activity against Gram-positive bacteria, and

was initially prescribed for treating staphylococcal infections in patients allergic to

penicillin. The classic indication of erythromycin A use has been in the treatment of

lower and upper respiratory tract infections, as well as skin and soft tissue infections. It

is relatively effective, but shows negative side effects exemplified by low bioavailability

and gastrointestinal symptoms. It is unstable in acidic environments and therefore

poorly absorbed across the gastrointestinal barrier. Degradation products of

erythromycin A do not show antimicrobial activity but stimulate intestinal peristalsis.

Clarithromycin

Clarithromycin, 6-O-methyl erythromycin A, was firstly synthesized by the

Japanese pharmaceutical company, Taisho Pharmaceutical. Starting with

erythromycin A, the synthesis of clarithromycin consists of four steps:

1) Protection of 9-oxo group with an oxime

2) Protection of 2’ and 4”-hydroxyl groups

3) Methylation of 6-hydroxyl group

4) Deprotection of 2’, 4” and 9 positions.2

The structure of clarithromycin is represented in Figure 2.

Figure 2: Structure of clarithromycin

The medicine was approved by the Food and Drug administration for the

treatment of upper and lower respiratory tract infections caused by S. pneumoniae, H.

influenzae, M. catarhallis, M. pneumoniae, S. pyogenes and S. aureus. It has good

pharmacokinetic properties with excellent oral bioavailability. It is active against most

respiratory pathogens with an MIC90≤0.25 µg ml-1. Interestingly, it shows activity

against bacteria that are not present in the spectrum of activity of other macrolides (H.

pylori (MIC90 0.03 µg ml-1), Borrelia burgdorferii (MIC90 20.06 µg ml-1)). It gives

gastrointestinal tolerance because of its inability to be cyclised into the 9,12-

10

hemiacetal. In August 1994, clarithromycin was approved by Federal Food

Administration for paediatric use. A very high clinical success rate was shown clinically

against H. influenzae, S. pneumoniae, and M. catarrhalis (90%, 100% and 100%,

respectively). Clarithromycin is safe and effective in non-AIDS patients of M. avium, M.

senopi, M. chelonei, and M. marinum infections.

Metabolism of erythromycin A and clarithromycin

Both erythromycin A and clarithromycin are metabolised through CYP450 3A4.3

However, there are differences in their abilities to bind to and inhibit the cytochrome P-

450 isoform CYP 3A4. On the basis of these differences, macrolides (in general) are

classified into three groups on the basis of data provided by in vitro experiments:

1) Group 1 include erythromycin A and troleandomycin. Both drugs bind strongly

to and markedly inhibit CYP 3A4.

2) Clarithromycin belongs to Group 2 agents. This drug exhibits a lower affinity for

CYP 3A4 compared to erythromycin A, and form complexes to a lesser extent.

3) Group 3 include azithromycin and dirithromycin. These compounds have been

shown to interact poorly with the cytochrome P-450 system in vitro.4

However, results obtained from some clinical studies showed that clarithromycin is

similar to erythromycin A in some drug interactions (e.g. with psychotropic agents).

1.1.2. 15-Membered macrolide antibiotics: 9a-methyl-9-deoxo-9-dihydro-9a-aza-9a-homoerythromycin A, 6-O-methyl homoerythromycins and theirderivatives

Azithromycin

9a-Methyl-9-deoxo-9-dihydro-9a-aza-9a-homoerythromycin A (azithromycin) is

shown in Figure 3. Azithromycin is a weak base and probably because of this property

easily penetrates the cell membrane and stays within the cell, mainly in lysosomes.

Figure 3: Structure of azithromycin

11

Azithromycin was synthesized by a team of researchers led by Slobodan

Djokic in 1981. The synthesis involved the Beckman rearrangement of erythromycin-

9(E)-oxime followed by the reduction of the imino ether intermediate and subsequent

N-methylation (Scheme 1).5,6

O

N

OH

OMe

O

OO

HOOH

O

N

HO

O

OMeOH

TsClpyridineether

O

OMe

O

OO

HO

O

N

HO

O

OMeOH

N

O

O

OMe

O

OO

HO

O

N

HO

O

OMeOH

N

OH

1. H2/PtO22. HCOOH, CHCl3

Scheme 1: The synthesis of azithromycin

Azithromycin dehydrate shows properties superior to the parent drug

(particularly against Gram-negative bacteria). It displays a prolonged and higher tissue

concentration, a short dosing period and a relatively low chance of gastrointestinal

side effects. The principal applications of azithromycin are in the treatment of

respiratory tract infections, skin and soft tissue infections, gastric and duodenal

infections caused by Helicobacter pylori and sexually transmitted diseases.7

Differences and similarities between the 14-membered macrolides

Erythromycin A and clarithromycin, 14-membered macrolide antibiotics are

mostly active against Gram-positive organisms. On the contrary, azithromycin and

derivatives made from azithromycin, azalides, show mostly activity against Gram-

negative organisms. Erythromycin A is acid-labile whilst, azithromycin and

clarithromycin are stable in acidic environments. In addition, azithromycin shows

12

unexpected activity against Plasmodium falciparum. Both 14- and 15-membered

macrolide antibiotics, have similar metabolic process through CYP450 3A4, although

they show differences in binding to CYP 3A4.

Other 15-membered derivatives

The hydroxyl group in position 4” of azithromycin can be oxidized in order to

give the 4-oxo compound, which can be further modified to 4-amino, or via the epoxide

to 4”-methyl amino compounds. The cladinose sugar has been replaced in several

cases in order to get 3-acyl or 3-oxo compounds, which were shown to be less active

than 14-membered acylides or ketolides.8

1.1.3. 16-Membered macrolide antibiotics: tylosin A, josamycin and theirderivatives

There are two sub-families of 16-membered macrolides: the tylosin sub-family

and the leucomycin-spiramycin sub-family. Both groups contain a substituted

disaccharide moiety (4’-O-(α-L-mycarosyl)-β-D-mycaminosyl) attached to the 5-

hydroxyl group of the aglycone.

Tylosin A

Tylosin A, introduced in 1961, has been reserved for use in veterinary

medicine. It contains a second neutral sugar (β-D-mycinose) attached to its 23-

hydroxyl group (Figure 4).9

Figure 4: Structure of tylosin A

Tylosin A shows potent activity against Gram-positive bacteria and is widely

used in veterinary medicine. It was initially isolated from Gram-positive strains of

Streptomyces fradiae that protect themselves from their own product by differential

expression of four resistance determinants, tlrA, tlrB, tlrC and tlrD. Tylosin A has good

absorption from the GI tract, and no enteric coating is required to maintain the stability

13

of the compound in the stomach. Surprisingly, it is distributed to the same tissues as

erythromycin A and metabolized by the liver and excreted via the bile and faeces.10

Tylosin A resistance can be explained by the synergy between G748 and

A2058 methylations. This synergistic mechanism of resistance is specific for the

macrolides tylosin A and mycinamycin and is not observed for macrolides such as

carbomycin, spiramycin and erythromycin A.

Josamycin

Josamycin (identical to leucomycin A3 and introduced in 1954)11 is the most

widely used 16-membered antibiotic (Figure 5). It is a macrolide antibiotic used

increasingly for treating upper and lower respiratory tract infections. Compared to

erythromycin A, it has greater tolerability, safety, and stability to gastric pH.12

Figure 5: Structure of josamycin

Other 16-membered derivatives

Derivatives obtained by modifications of the aldehyde in tylosin A and

desmycosin (tylosin B) (Figure 6) have improved ratio of oral/parenteral efficacy and

bioavailability.

14

Figure 6: Structure of tylosin B

Several modifications of desmicosin were introduced: 1) replacement and

modification of mycinose (labelled red); 2) modification of the dienone moiety (labelled

pink); 3) modification of mycaminose (labelled blue); 4) modification of the C-3

hydroxyl group (labelled green). These modifications were combined with the aldehyde

group modifications in order to improve pharmacokinetic properties (Figure 7).13

Figure 7: Representation of places of modifications on desmicosin

One of the very successful derivatives obtained by modifications of the

aldehyde in tylosin A is tilmicosin (Figure 8), initially used to treat bovine respiratory

disease.

Figure 8: Structure of tilmicosin

15

Tilmicosin, 20-deoxo-20-(3,5-dimethylpiperidin-1-yl) desmycosin, is produced

by the removal of mycarose of tylosin A in an organic solvent such as butyl acetate.14

The solubility of tilmicosin in aqueous solution is dependent on temperature and pH

conditions.15 It has a slow absorption profile, 22% bioavailability, a half-life in plasma of

four hours and shows extensive penetration in milk.16 It has an antibacterial spectrum

similar to tylosin A with enhanced activity against Pasteurella multocida and

Pasteurella haemolytica. The CVMP (Committee for Medicinal Products for Veterinary

Use) evaluated tilmicosin for its use in chicken, cattle, pigs and sheep and two trials

were conducted to test the ability of tilmicosin as a feed additive to prevent

transmission of progressive atrophic rhinitis (AR), including P. multocida, from

endemically affected pigs to AR-free pigs (three weeks old at the start of the trial).17

Spiramycin I (Figure 9) differs from the leucomycins by the presence of the

second amino sugar substituent (D-forosamine) attached to the 9-hydroxyl group via a

β-glycosidic bond.

Figure 9: Structure of spiramycin I

1.2. Macrolide antibiotics as antibacterial agents1.2.1. The mode of action of 14-membered macrolide antibiotics (erythromycin

A, clarithromycin) and effect of substitution on their mode of actionThere is only general knowledge about the interaction of macrolides with

ribosomes. Erythromycin class antibiotics do not block peptidyl transferase activity.18

Erythromycin-type macrolides bind to the peptidyl transferase ring and block the tunnel

that channels the nascent peptides away from the peptidyl transferase centre. 19 It was

found that erythromycin A, clarithromycin and roxithromycin bind to the same site in

the 50S subunit of D. radiodurans, at the entrance of the tunnel (Figure 10).

The desosamine sugar and lactone ring are taking part in the interaction of the

drug with the peptidyl transferase cavity.20 The OH group of the desosamine sugar

forms hydrogen bonds with three positions: N6 and N1 of A2041Dr (A2058Ec) and N6

of A2042Dr (A2059Ec). Hydrogen bonding interactions of hydroxyl group of

16

desosamine with the above mentioned nitrogens in bases explains resistance

mechanisms.21 The following hydroxyl groups have an interaction with 23S rRNA in the

lactone ring: the 6-OH (hydrogen bonding to N6 of A2045Dr (A2062Ec))22, 11-OH and

the 12-OH groups (both form hydrogen bonds with O4 of U2588Dr (U2609Ec)).23

Interestingly, cladinose sugar is not engaged in the interactions with 23S rRNA.

Erythromycin A, clarithromycin and roxithromycin show differences in structure, but a

high degree of similarity in their interactions with ribosomes. For example, etheroxime

in roxithromycin does not participate in the interaction of this drug with 23S rRNA and

ribosomal proteins. It appears that the two ribosomal proteins are involved in drug

resistance. However, the distances of the drug to 23S rRNA and ribosomal proteins

are too far, which suggest that some other interactions, except hydrogen bonding may

be responsible for the action of the drug (van der Waals interactions, hydrophobic

interactions and crowded environments of ribosomal proteins and rRNA around the

drug)24.

Figure 10: Interaction of macrolides with the peptidyl transferase cavity. a) Chemical

structure diagram of macrolides (erythromycin A, clarithromycin and roxithromycin)

showing the interactions of the reactive groups of macrolides with the nucleotides of

the peptidyl transferase cavity. b) Secondary structure of 23S rRNA with marked

contacts of the medicine with the RNA moiety. c) Stereoview showing the erythromycin

17

binding site at the peptidyl transferase cavity of D. radiodurans. In the scheme,

antibiotic is represented with green, part of ribosomal protein with yellow, part of

ribosomal protein L22 with light green. Nucleotide numbering is according to the E. coli

sequence. 24

A common pattern to the interaction and action of 14-membered macrolides to

bacterial ribosomes is apparent: the interaction between N6 of A2041Dr (A2058Ec),

N6 of A2042Dr (A2059Ec) and the non-bridging phosphate oxygen of U2484Dr

(U2505Ec). In order to prevent binding to peptidyl-tRNA, a nucleotide is situated at the

G2484Dr (G2505Ec) position.25

The macrolide binding site can be found at the entrance of the exit tunnel (Figure

11).

Figure 11: Top view of the D. radiodurans 50S subunit showing erythromycin (red)

bound to the entrance of the exit tunnel. Ribosomal proteins are represented by yellow

colour, 23S rRNA by grey and 5S rRNA by dark grey. 25

The nature created the exit tunnel in a way that the 6-8 peptide bonds are getting

created before the nascent protein chain reaches the bound macrolide. In the bound

state, the drug is reducing the exit tunnel diameter from approximately 18 Å to nearly

10 Å. Also, the exit tunnel is occupied by the hydrated Mg2+ ion (Mg33). None of them

show direct interactions with ribosomal proteins.

The first 14-membered macrolide, erythromycin A, possesses good activity

against Gram-positive bacteria, and was initially used for treating staphylococcal

infections in patients allergic to penicillin.

Clarithromycin (6-O-methylerythromycin) is more acid stable than the parent

drug, erythromycin A and shows better activity against Gram positive bacteria.

18

Unsuccessful derivatives of erythromycin A were synthesised and it was

observed that erythromycin A anti-bacterial activity can be reduced by the following

modifications: 1) C-3’ dimethylamino modification; 2) cladinose removal; 3) C-9

carbonyl modification.26 In order to yield the compound with good in vitro activity,

Taisho scientists tried to modify erythromycin A by attaching an additional cladinose

sugar to the C-11 position.27 The synthesis of C-12-O-methyl erythromycin A derivative

was reported by Abbott scientists28, but this compound was shown to be less active

than erythromycin A or clarithromycin.

1.2.2. Mode of action of 15-membered macrolide antibiotics (azithromycin) andthe effect of substitution on their mode of action

Azithromycin differs from other macrolide antibiotics having the secondary

binding site (Figure 12, Figure 13).

Figure 12: The regions of domains II, IV and V of 23S rRNA that contribute to binding.

The contact sites of ABT-773 and azithromycin are indicated by large, coloured letters

in the diagram. 29

19

Figure 13: Interactions of azithromycin with ribosomal proteins L4 and L22 and 23S

rRNA29

The primary site is located in domain V of 23S rRNA forming network of

interactions and with the orientation similar to erythromycin A. For the second

molecule it was found that the hydrogen bond forms between the O4 of U2588Dr with

a direct contact observed with the primary binding site. It is believed that the second

20

molecule of azithromycin interacts with the L4 ribosomal protein. Additional contacts

with domains II and IV lead to another region of 23S rRNA giving the tighter binding.

Hydrophobic interactions of bases A2045Dr (A2062Ec) and U2588Dr (U2609Ec)

contribute to the binding of the first azithromycin molecule. Interestingly, a putative Mg

ion probably in co-ordination with water molecules interacts with the cladinose sugar

and lactone ring. Surprisingly, the nitrogen atom inserted into the lactone ring of

azithromycin does not contribute directly to binding to ribosomes. It is worth

mentioning that the conformation of the lactone ring must be altered in order to involve

new contacts of the putative Mg2+ ion with C2565Dr. The second azithromycin

molecule makes a direct contact with the first azithromycin molecule forming a

hydrogen bond between the desosamine sugar and O1 in the lactone ring of the first

azithromycin molecule (potentially this can also cause an enhancement in activity).

Some contacts with erythromycin A are the same: A2041Dr, A2042Dr (A2059Ec),

and A2045Dr (A2062Ec). The following contact is not present - U2590Dr (C2611Ec),

which can be explained as a consequence of the conformational shift of the bulged

bases U2592Dr (U2613Ec) and A2593Dr (A2614Ec) of helix 73. Also, there is an

additional hydrogen bond between the cladinose of the first azithromycin molecule and

C2565Dr (U2586Ec). This hydrogen bond explains the different orientation of the

cladinose for azithromycin compared to erythromycin A. 29

Azithromycin is active against staphylococci and enterococci and shows lower

activity than erythromycin A against Gram positive bacteria. Gram-negative bacteria

are more susceptible to azithromycin than erythromycin A. Against Enterobacteriaceae

genera and other aerobic Gram-negative bacteria, azithromycin is more potent than

erythromycin A with the MIC90 range of 2-64 µg ml-1. In summary, it can be said that

azithromycin is active against 1) rapidly grown pyogenic bacteria susceptible to beta-

lactam drugs (especially benzyl penicillin), 2) bacteria resistant to benzyl penicillin and

erythromycin A (enterobacteria and Pseudomonas, H. influenzae). Like other

macrolides (roxithromycin, erythromycin A), it accumulates in phagocyte cells. In vitro,

unfractioned blood leukocytes and neutrophils accumulate azithromycin up to 160-fold

greater than the extracellular concentration within one hour at 37°C.

The most promising and successful 15-membered derivatives of azithromycin

were generated when a benzyl amine moiety was attached at the C-4" hydroxyl group

via a carbamate linker. The optimal combination was shown to be the compound with

the diamine linker, 2-methoxybenzyl amine and the erythromycylamine template. 30

Compound CP-544,372 (Figure 14) against macrolide resistant S. pneumoniae and H.

influenzae is similar in activity to telithromycin.31

21

Figure 14: Structure of CP-544 372

4” Modified azalides in which an amino group is attached by opening the epoxide

at this position are especially active against Gram-negative organisms (Figure 15).32

Figure 15: 4”-modified azalide

The C-3, C-6-ketals of azalides have been shown to be active in vitro against

Pasteurella multocida, E. coli and S. aureus. However, because C-3, C-6 diols are

much less active, conversion to a ketal enhances the potency of the descladinosyl

azalides.33

1.2.3. Modifications of erythromycin A and their anti-tuberculosis activity

Recently, Zhu et al.34 explored how modifications of different parts of the

macrolide affect the activity of the drug against Mycobacterium tuberculosis.

22

Table 1: Modifications on erythromycin A and their anti-tuberculosis activities

Position of the

modificationType of derivative Activity

3

cladinose-containing

compounds

Cladinose-containing compounds are

more potent

ketolides

3-hydroxy derivatives

3-carbamoyloxy

derivatives

66-O-substituted

derivativesStrongly depends upon the substituent R6

9 9-oxime

Nature of the substituent is essential for

activity (activity and toxicity are increasing

with increasing lipophilicity of substituent)

11 11,12-carbamate

11,12-carbazateAlso inhibitors of CYPA4

12

16

derivatives obtained by

the substitution of Me

with F

More activity, less toxic

2’ 2’-acyl derivativesAcylation at position 2’ varies depending

on other modifications in the molecule

1.2.4. Spiramycin, tylosin A and tylosin B derivatives and their mode of action

It was found that mycarose removal from spiramycin can cause loss of the

inhibition in poly-U directed polyphenylalanine synthesis35 .Tylosin A is one of the few

macrolides able to inhibit peptidyl transferase. It has been shown that the mycarose

moiety in tylosin A protects A2506Ec and the nucleotide in the neighbourhood

G2484Dr (G2505Ec) is protected against chemical modification.

A group of researchers from Denmark showed that the ability of macrolides to

inhibit the peptidyl transferase reaction strongly depends on the presence of a

23

disaccharide at position 5. The second main finding was that ribosomal-binding sites of

drugs were defined by chemical probing of nucleotides in the rRNA, and interestingly

chemical probe of nucleotides in the rRNA correlates with the presence of a

disaccharide at position 5 in macrolides.

Inhibitory activity of the macrolides was tested by the “puromycin reaction”. A

peptide-bond was formed in this reaction between puromycin and N-acetyl-[14C]Phe-

tRNAPhe which binds in the donor substrate site on the ribosome. The following figure

(Figure 16) presents carbomycin (myc.), spiramycin (myc.) and tylosin (myc.) that

inhibit the puromycin reaction. Erythromycin does not inhibit this reaction, whilst

carbomycin (myc.) and spiramycin inhibit this reaction (i.e. 100% and more than 85%

inhibition, respectively).

Figure 16: Peptidyl-transferase activity by the puromycin reaction as a function of

antibiotic concentration. Abbreviations used: Tyl refers to tylosin A, Des to

desmycosin, Ery to erythromycin A, Cha for chalcomycin and Hyg A to hygromycin. 36

It was found that tylosin A and desmycosin protect A752 in domain II of the 23S

rRNA from DMS modification. Beside convenient 14-membered macrolide antibiotics,

also 16-membered macrolide antibiotics (carbomycin, spiramycin, tylosin A,

desmycosin) protect A2062 and U2506. Moreover, macrolide antibiotics except

chalcomycin protect A2572 and U2609 from modification with DMS and CMCT (Figure

17).36

24

Figure 17: Secondary structural models of (a) The central loop region of domain V of

23S rRNA and (b) the hairpin 35 region of domain II37. Chemical footprints are

indicated by circules obtained with the different macrolides and with hygromycin A

Whilst carbomycin and tylosin A protect U260938, it was found that spiramycin

and desmycosin protect A2058, A2059, A2062 and U2609.

Spiramycin and tylosin A, 16-membered macrolide antibiotics, do not inactivate

cytochrome P450 and are unable to modify the pharmacokinetics of other

compounds.39 Peptide derivatives of tylosin A and desmycosin (tylosin B) were

synthesized by their modification at the C6 position (aminooxyacetyl-L-alanine-L-

alanine methyl ester) in order to explore the position of peptides in the exit tunnel of

the ribosome. Interestingly, Sumbatyan et al. did not report any biological or

crystallographic data on these compounds, but provided evidence that these

compounds inhibit cell transcription-translation.40 Antibacterial activity in 16-membered

macrolides against resistant pathogens was enhanced by the introduction of the

aromatic ring with an adjusted length of an alkyl spacer into the lactone ring.41

25

1.3. Malaria and other protozoal diseases1.3.1. Introduction to protozoa

The protozoa are a phylum of unicellular organisms found worldwide in marine

and freshwater environments where some are free-living and some parasitic. They

move using cilia or pseudopodia and contain a nucleus, cytoplasm and cell wall; some

of them contain chlorophyll. Reproduction is by fission or encystment. Their length is

generally in the range 1-300 µm. This phylum is divided into four classes: Flagellata,

Cnidospora, Ciliophora, and Sporozoa.42

Protozoa can cause serious infections and may infect any human tissue. The

parasites may exist inside or outside the host cell. Some of them produce cysts to

survive outside the body, others use insects as vectors, and some spread during

human sexual contact.

Among the diseases caused by protozoa are toxoplasmosis, amoebic meningitis,

malaria, trypanosomiasis, leishmaniasis (kala-azar) and amoebic dysentery as well as

diarrhoea caused by Cryptosporidium spp. or Giardia intestinalis (lamblia).

1.3.1.1. Toxoplasmosis

Toxoplasmosis is caused by a single-celled parasite called Toxoplasma gondii

(Figure 18).

Figure 18: Toxoplasma gondii (Image taken from the website:

www.hei.org/research/aemi/toxo.htm)

Individuals with compromised immune systems and pregnant women are

particularly vulnerable to this disease. The infection pathways include:

1) Accidentally swallowing cat faeces from an infected cat

2) Consuming contaminated food with knives, utensils, cutting boards and

other foods that may have come into contact with raw meat

26

3) Using contaminated drinking water

Some people infected with toxoplasmosis have flu-like symptoms including swollen

lymph glands or muscle aches and pains that last for a month or more. Severe

toxoplasmosis can cause damage to the brain, eyes, or other organs. Most children

who are infected when in the womb have no symptoms at birth, but may develop

symptoms later in their lives.43

1.3.1.2. Amoebic meningoencephalitisAmoebic meningoencephalitis is an uncommon infection of the central nervous

system (CNS). The first human case of amoebic meningitis was identified in 1965 and

the cause was Naegleria fowleri (Figure 19).44 Acanthamoeba sp. has also been

identified as a cause of this disease.45 Naegleria fowleri is a thermophilic organism, but

it is susceptible to dry conditions and high or low pH and loses viability in seawater. It

is often found in tropical areas, artificially heated water, fresh-water and man-made

lakes, hot springs, warm polluted rivers and streams. This organism is the cause of a

primary meningoencephalitis, infecting the brain directly. The incubation can last from

two or three days up to as much as seven to 15 days, depending on the virulence of

the strain and the number of organisms inoculated. The usual treatment includes using

of amphotericin B. After infection the on-set of death can be rapid, after three to seven

days of onset of symptoms, but hospitalized patients may survive up to two weeks.46

Figure 19: Naegleria fowleri (Image taken from the

website:http://scienceblogs.com/observations/2009/05)

A relatively small number of the Acanthamoebal species are pathogenic. The

following species can cause infections in humans: Acanthamoeba castellanii (Figure

20), Acanthamoeba culbertsoni, Acanthamoeba polyphaga and Acanthamoeba

astronyxis. There are several medicines that show in vitro activity against these

organisms. Unfortunately, most patients do not respond to treatment.47-49

27

Figure 20: S.E.M. of Acanthamoeba castellanii (Image was taken from the website:

www.bms.ed.ac.uk/research/others/smaciver/acanthamoeba.htm)

1.3.1.3. Human African trypanosomiasisHuman African trypanosomiasis or sleeping sickness is a widespread tropical

disease that can be fatal if untreated. It is spread by the bite of an infected tsetse fly

(Glossina Genus). Within a few weeks the person can experience fever, swollen lymph

glands, aching muscles and joints, headaches and irritability. The disease attacks the

central nervous system, causing changes in personality, alteration of the biological

clock, confusion, slurred speech, seizures, and difficulty in talking and walking. The

control of this disease is based on the reduction of the reservoirs of infection by early

diagnosis and control of tsetse flies.50

1.3.1.4. Cutaneous leishmaniasisCutaneous leishmaniasis can be traced back many hundreds of years, but the

first written record appeared in 1756 by Alexander Russell following examination of a

Turkish patient.51 The native inhabitants of the Andes, can become infected as

children, and the effects generally appear in the face. Non-native patients can express

symptoms months after their arrival in an endemic area. Geographic distribution of

cutaneous leishmaniases is represented in Figure 21 (2002).

Figure 21: Geographic distribution of cutaneous leishmaniases (Image was taken

from the website: http://course1.winona.edu/kbates/Parasitology/flagella.htm)

Skin lesions and facial deformities were represented on pre-Inca pottery from

Ecuador and Peru dating back to the first century AD. The leishmaniases are caused

by 20 species pathogenic to humans that belong to the Leishmania genus, a

protozoan transmitted by the bite of a tiny two to three millimetre-long insect vector.

The disease name was derived from the man who identified it in 1901. It is

28

characterised by general debility, irregular and repetitive bouts of fever, severe

anaemia, muscular atrophy and excessive swelling of the spleen.52

1.3.1.5. MalariaMalaria is a parasitic disease resulting from infection with one of four species of

Plasmodium protozoa in humans (P. falciparum, P. vivax, P. malariae, P. ovale). It is

transmitted by the Anopheles mosquito and is characterised by sudden fever and

enlargement of the spleen. Attacks of fever, chills, and sweating recur as new

generations of parasites develop in the blood.53 The life cycle of the malarial parasite

is represented in Figure 22.

Figure 22: The life cycle of the malarial parasite (Image was taken from the website:

www.dpd.cdc.gov/dpdx/html/malaria.htm)

There are two hosts in the malaria parasite life cycle. In the first stage, a

malaria-infected female Anopheles mosquito inoculates sporozoites into the human

host (1). Liver cells are infected by sporozoites (2) and undergo maturation into

schizonts (3), which further rapture and release merozoites (4). P. vivax and P.ovale

have a dormant stage (hypnozoites) which can persist in the liver and cause relapses

by invading the bloodstream weeks, or in some cases years later. After this initial

replication in the liver (exo-erythrocytic schizogony A), the parasites undergo asexual

multiplication in the erythrocytes (erythrocytic schizogony B). Red blood cells get

infected by merozoites. The ring stage trophozoites mature into schizonts, which result

in the release of merozoites (6). It is noteworthy that some parasites differentiate into

29

sexual erythrocytic stages (gametocytes) (7). Clinical manifestations of the disease are

observed on the basis of blood stage parasites. The gametocytes, microgametocytes

(male) and macrogametocytes (female), get ingested by an Anopheles mosquito

during a blood meal (8). The sporogonic cycle (C) is characterized by the multiplication

of the parasite in the mosquito. The microgametes penetrate macrogametes

generating zygotes while they are situated in the mosquito stomach (9). The zygotes

become motile and elongated (ookinetes) (10) which invade the midgut wall of the

mosquito where they develop into oocysts (11). Furthermore, the oocysts grow,

rupture, and release sporozoites (12), which make their way to the mosquito’s salivary

glands. Inoculation of the sporozoites into a new human host perpetuates the malaria

life cycle (1).54

The first antimalarial drug, quinine, was a pathway to synthetics such as

chloroquine. There are five medicines used to prevent malaria (Figure 23):

Atovaquone plus proguanil (also known as Malarone);

Chloroquine (also known as Avloclor tablets and Nivaquine syrup);

Doxycycline (also known as Vibramycin-D);

Mefloquine (also known as Lariam);

Progunail (also known as Paludrine).55

30

Figure 23: Medicines that are in use as antimalarial drugs

Current antimalarial drugs can be divided on the basis of price into two groups:

1) “Commodity generic” antimalarials – drugs that are traded in large

quantities and whose patents have expired, e.g. chloroquine.

2) Relatively new antimalarials – available from only one or a few

sources; they are still under patent and usually available only under brand names.

Malaria does not cause dramatic disfigurement or disability. Therefore, the

ability to raise significant donor funds is limited. However, the travel market and local

elites can afford to pay a high price for prophylactic and curative treatment of malaria,

and this segment of the market is large enough to fuel the development and marketing

of novel, expensive antimalarials.56

Malaria is one of the most widespread diseases (Figure 24), taking each year

more than half of a million lives.53

Figure 24: World map of malaria endemic areas (Image was taken from the website:

www.michellehenry.fr/MalariaMap.gif)

1.3.1.5.1. Anti-malarial medicines and drug resistanceUnfortunately, drug resistance is widespread, and no new chemical class of

antimalarials have been introduced into clinical practice since 1996. There has been a

recent rise in parasite strains with reduced sensitivity to the newest drugs. Some

companies have embarked upon a program of new antimalarial medicine research.

GlaxoSmithKline has screened nearly two million compounds in search for inhibitors of

Plasmodium falciparum and confirmed that 13533 compounds do inhibit parasite

growth by at least 80% at a 2 µM concentration. Activity against multidrug resistant

31

strains was shown in cases of 8000 compounds. Several novel mechanisms of

antimalarial action were found, e.g. inhibition of protein kinases and host-pathogen

interaction related targets. It was found that Plasmodium falciparum causes the most

mortality, mainly in children below the age of five.57 The last effective class of drugs,

the artemisinins, is being compromised by the rise of Plasmodium falciparum strains

with reduced clinical response to artemisinin-containing drug combinations.58,59

It is worth mentioning that the Bill and Melinda Gates Foundation support

investigations into finding anti-malaria vaccines.60

1.4. Macrolide antibiotics used in the treatment of malaria1.4.1. Azithromycin alone and in the combination with existing anti-malarial

medicinesThe first paper about the use of a macrolide as a prophylaxis for liver infection

or both liver and subsequent blood infection with Plasmodium falciparum malaria

appeared in 1995. The performed study showed that azithromycin has the potential to

be at least as effective as mefloquine and doxycycline as a prophylactic agent for

malaria. Compared to mefloquine (pregnancy class C: teratogenic in laboratory

animals but no evidence of teratogenicity in humans) and doxycycline (pregnancy

class D: positive evidence of human foetal risk), azithromycin is in pregnancy class B

(no evidence of teratogenicity in laboratory animals but no well-controlled clinical

studies) and may be used during pregnancy if clearly needed.61

In 2002, it was found that azithromycin cannot be an efficient single agent, but

with quinine, tafenoquine or primaquine was shown to have additive, synergistic

qualities while dihydroartemisinin was additive with a trend towards antagonism.62

Malaria in pregnancy is one of the most important causes of maternal, perinatal

and neonatal morbidity in sub-Saharan Africa. Preventive treatment of malaria in

pregnancy includes use of sulphadoxine-pyrimethamine which reduces the incidence

of low birth-weight, pre-term delivery, intrauterine growth retardation and maternal

anaemia. A potential alternative to sulphadoxine-pyrimethamine treatment is the

combination of azithromycin and chloroquine. 63

The research about the azithromycin combination therapy for malaria was

conducted under the sponsorship of the National Institute of Allergy and Infectious

Diseases (NIAID) in 2006. The survey was performed on 120 people, aged 20-65 in

Thailand (Mahidol University Hospital for Tropical Diseases, Bangkok) using the

intervention: azithromycin, artesunate and quinine. It was shown that azithromycin as

an anti-malarial agent works best in combination with other drugs.64

32

1.5. References

1 J. M. McGuire, R. L. Bunch, R. C. Anderson, H. E. Boaz, E. H. Flynn, H. M. Powelland J. W. Smith: ”Ilotycin”, a new antibiotic, Antibiot. Chemother., 2, 281, 1952.2 J-H. Liu, G. A. Foster Jr. And S. H. Montgomery: Process for 6-O-alkylation oferythromycin derivatives. WO 97/19096, 1997.3 Drug Information Service. 63, 2002.4 N. A. Von Rosenstiel and D. Adam: Macrolide antibacterials. Drug interactions ofclinical significance, Drug safety, 13, 105, 1995.5 G. Kobrehel and S. Djokic: 11-Methyl-11-aza-4-O-cladinosyl-6-O-desosamynyl-15-ethyl-7,13,14-trihydroxy-3,5,7,9,12,14-hexamethyloxacyclopentadecane-2-one andderivatives thereof, U. S. patent 4,517,359, May 14,1985; filed September 22, 1981.6 G. Kobrehel, G. Radobolja, Z. Tamburasev and S. Djokic: 11-aza-10-deozo-10-dihydroerythromycin A and derivatives thereof as well as a process for theirpreparation, U. S. patent 4,328,334, May 4, 1982, filed March 28, 1980.7 http://www.sumamed.com.hr/success.htm8 S. Mutak: Azalides from azithromycin to new azalide derivatives, J. Antibiot., 60, 85,2007.9 W. Schonfeld and H. A. Kirst: Introduction to the macrolide antibiotics (p.1).Macrolide antibiotics, 323 ps, Birkhauser, 2002.10 M. Liu and S. Douthwaite: Resistance to the macrolide antibiotic tylosin is conferredby single methylations at 23S rRNA nucleotides G748 and A2058 acting in synergy,PNAS, 99, 14658, 2002.11 S. Omura, Y. Hironaka and T. Hato: Chemistry of Leukomycin A3 with josamycin, J.Antibiot., 23, 511, 1970.12 X. Bosch, E. Pedrol, X. Casado and A. Urbano-Marquez: Josamycin-induced pedaloedema, BMJ., 307, 26, 1993.13 H. A. Kirst, G. M. Wild, R. H. Baltz, R. L. Hamill, J. L. Ott, F. T. Counter and E. E.Ose: Structure-activity studies among 16-membered macrolide antibiotics related totylosin, J. Antibiot (Tokyo)., 35, 1675, 1982.14 M. Debono, K. E. Willard, H. A. Kirst, J. A. Wind, G. D. Crouse, E. V. Tao, J. T.Vicenzi, F. T. Counter, J. L. Ott, E. E. Ose and S. Omura: Synthesis and antimicrobialevaluation of 20-deoxo-20-(3,5-Dimethylpiperidin-1-yl)desmycosin (Tilmicosin, EL-870)and related cyclic amino derivatives, J. Antibiot., 42, 1253, 1989.15 FAO, 1996.16 G. Ziv, M. Shem-Tov, A. Glickman, M. Winkler and A. Saran: Tilmicosin antibacterialactivity and pharmacokinetics in cows, J. Vet. Pharmacol. Therap., 18, 340, 1995.17 L. Backstrom, J. McDonald, M. T. Collins, W-B. Chung, T. R. Shryock and E. E.Ose: Efficacy of tilmicosin, and a combination of tylosin and sulfamethazine for controlof swine atrophic rhinitis involving infection with toxigenic Pasteurella multocida typeD, Swine Health Prod., 2, 11, 1994.18 D. Vazquez: In Antibiotics. III. Mechanisms of action of antimicrobial and antitumoragents (J. W. Concoran & F. E. Hahn. Eds), pp. 459-479, Springer, Berlin, Heidelberg,New York, 1975.19 P. Nissen, J. Hansen, N. Ban, P. B. Moore and T. A. Steitz: The structural basis ofribosome activity in peptide bond synthesis, Science, 289, 920, 2000.20 W. E. Steinmetz, R. Bersch, J. Towson and D. Pesiri: The conformation of 6-methoxyerythromycin A in water determined by proton NMR spectroscopy, J. Med.Chem., 35, 4842, 1992.21 B. Weisblum: Erythromycin resistance by ribosome modification, Antimicrob. Agents.Chemother., 39, 577, 1995.22 L. H. Hansen, P. Mauvais and S. Douthwaite: The macrolide-ketolide antibioticbinding site is formed by structures in domains II and V of 23S ribosomal RNA, Mol.Microbiol., 31, 623, 1999.23 J. C.-H. Mao and M. Puttermann: The intermolecular complex of erythromycin andribosome, J. Mol. Biol., 44, 347, 1969.

33

24 F. Schlunzen, R. Zarivach, J. Harms, A. Bashan, A. Tocilj, R. Albrecht, A. Yonathand F. Franceschi: Structural basis for the interaction of antibiotics with the peptidyltransferase centre in eubacteria, Nature, 413, 814, 2001.25 U. Saarma, C. M. T. Spahn, K. H. Nierhaus and J. Remme: Mutational analysis ofthe donor substrate binding site of the ribosomal peptidyltransferase center, RNA, 4,189, 1998.26 A. Mereu, E. Moriggi, M. Napoletano, C. Regazonni, S. Manfredini, T. P. Mercurioand F. Pellacini: Design, synthesis and in vivo activity of 9-(S)-dihydro-erythromycinderivatives as potent anti-inflammatory agents, Bioorg. Med. Chem. Lett., 16, 5801,2006.27 T. T. Asaka, Y. T. Misawa, M. T. Kashimura, S. T. Morimoto and K. T. Hatayama: 5-O-desosaminylerythronolide derivative, EP 0619 320, 1994.28 Y. Y. Ku, D. Riley, T. Grieme and A. Nilius: Synthesis and antibacterial activities ofnovel 12-O-methyl erythromycin A derivatives, J. Antibiot. (Tokyo), 52, 908, 1999.29 F. Schlunzen, J. M. Harms, F. Franceschi, H. A. S. Hansen, H. Bartels, R. Zarivachand A. Yonath: Structural basis for the antibiotic activity of ketolides and azalides,Structure, 11, 329, 2003.30 T. Furunchi, K. Kurihara, K. Ajito, T. Yoshida and H. Fushimi: 3-Substitutedleucomycins and pharmaceutical compositions containing them, JP 2004217562.31 A. Bryskier: New research in macrolides and ketolides since 1997, Expert Opin.Invest. Drugs, 8, 1171, 1999.32 B. S. Bronk, M. A. Letavic, C. D. Bertsche, D. M. George, S. F. Hayashi, B. J.Kamicker, N. L. Kolosko, L. J. Norcia, M. A. Rushing, S. L. Santoro and B. V. Yang:Synthesis, stereochemical assignment and biological activity of a novel series of C-4”modified aza-macrolides, Bioorg. Med. Chem. Lett., 13, 1955, 2003.33 S. M. Sakya, P. Bertinato, B. Prat, M. Suarez-Contreras, K. M. Lundy, M. L. Minich,H. Cheng, C. B. Ziegler, B. J. Kamicker, S. F. Haxashi, S. L. Santoro, D. M. Georgeand C. D.Bertsche: Azalide 3,6-ketals: antibacterial activity and structure activityrelationships of aryl and hetero aryl substituted analogues, Bioorg. Med. Chem. Lett.,13, 1373, 2003.34 Z. J. Zhu, O. Krasnykh, D. Pan, V. Petukhova, G. Yu, Y. Liu, H. Liu, S. Hong, Y.Wang, B. Wan, W. Liang and S. G. Franzblau: Structure-activity relationships ofmacrolides against Mycobacterium tuberculosis, Tuberculosis (Edinb)., 88: S49.35 D. Vazquez: Binding to ribosomes and inhibitory effect on protein synthesis of thespiramycin antibiotics, Life Sci., 6, 845, 1967.36 S. M. Poulsen, C. Kofoed and B. Vester: Inhibition of the ribosomal peptidyltransferase reaction by the mycarose moiety of the antibiotics carbomycin, spiramycinand tylosin, J. Mol. Biol., 304, 471, 2000.37 R. R. Gutell, N. Larsen and C. R. Woese: Lessons from an evolving rRNA: 16S and23S rRNA structures from a comparative perspective, Microbiol. Rev., 58, 10, 1994.38 C. Rodriguez-Fonseca, R. Amils and R. A. Garrett: Fine structure of the peptidyltransferase centre on 23S-like rRNAs deduced from chemical probing of antibiotic-ribosome complexes, J. Mol. Biol., 247, 224, 1995.39 A. Anadon and L. Reeve-Johnson: Macrolide antibiotics, drug interactions andmicrosomal enzymes: implications for veterinary medicine, Res. Vet. Sci., 66, 197,1999.40 N. V. Sumbatyan, G. A. Korshunova and A. A. Bogdanov: Peptide derivatives ofantibiotics tylosin and desmycosin, protein synthesis inhibitors, Biochemistry (Mosc).,68, 1156, 2003.41 J. L. Kellenberger, S. R. Shapiro, S. Mathews, P. Guerry and P. J. N. Barbier: Newmacrolides, EP 1846430B1, 2006.42 http://www.mansfield.ohio-state.edu43 http://www.medic8.com/healthguide/articles/toxoplasmosis.html44 M. Fowler and R. F. Carter: Acute pyogenic meningitis probably due toAcanthamoeba sp.; a preliminary report, Br. Med. J., 5464, 740, 1965.

34

45 B. V. Jager and W. P. Stamm: Brain abscesses caused by free-living amoebaprobably of the genus Hartmannella in a patient with Hodgkin’s disease, Lancet. 2,1343, 1972.46 W. Royal III MD: Amebic meningoencephalitis, Curr. Treat. Options Neurol., 8, 202,2006.47 F. L. Schuster and G. S. Visvesvara: Free-living amoebae as opportunistic and non-opportunistic pathogens of humans and animals, Int. J. Parasitol., 34, 1001, 2004.48 F. Marciano-Cabral and G. Cabral: Acanthamoeba spp. as agents of disease inhumans, Clin. Microbiol. Rev., 16, 273, 2003.49 G. S. Visveswara, S. S. Mirra and F. H. Brandt: Isolation of two strains ofAcanthamoeba castellanii from human tissue and their pathogenicity and isoenzymeprofiles, J. Clin. Microbiol., 18, 1405, 1983.50 http://www.who.int/topics/trypanosomiasis_african/en/51 A. Ul Bari: Chronology of cutaneous leishmaniasis: An overview of the history of thedisease, JPAD, 16, 24, 2006.52 http://www.who.int/leishmaniasis/en/53 “Malaria” World Encyclopedia. Philip’s, 2008. Oxford Reference Online. OxfordUniversity Press. University of Manchester. 18 June 2010http://www.oxfordreference.com/views/ENTRY.html? subview=Main & Entry =t142.e713054 www.dpd.cdc.gov/dpdx/html/malaria.htm55 http://www.nhs.uk/Conditions/Malaria/Pages/Treatment.aspx56 S. D. Foster: Pricing, distribution, and use of antimalarial drugs, Bull. W. H. O., 69,349, 1991.57 F-J. Gamo, L. M. Sanz, J. Vidal, C. De Cozar, E. Alvarez, J-L. Lavandera, D. E.Vanderwall, D. V. S. Green, V. Kumar, S. Hasan, J. R. Brown, C. E. Peishoff, L. R.Cordon and J. F. Garcia-Bustos: Antimalarial leads, Nature, 465, 305, 2010.58 V. Andriantscanirina, A. Ratsimbasoa, C. Bouchier, M. Jahevitra, S.Rabearimanana, R. Radrianjafy, V. Andrianaranjaka, T. Randriantsoa, M. A. Rason,M. Tichit, L.P. Rabarijaona, O. Mercereau-Puijalon, R. Durand and D. Menard:Plasmodium falciparum, drug resistance in Madagascar: facing the spread of unusualpfdhfr and pfmdr-1 haplotypes and the decrease of dihydroartemisinin susceptibility,Antimicrob. Agents Chemother., 53, 4588, 2009.59 M. Bonnet, I. Van den Broek, M. Van Herp, P. P. P. Urrutia, C. Van Overmeir, J.Kyomuhendo, C. N. Ndosimao, E. Ashley and J-P. Guthman: Varying efficacy ofartesunate+amodiaquine and artesunate + sulphadoxine-pyrimethamine for thetreatment of uncomplicated falciparum malaria in the Democratic Republic of Congo: areport of two in vivo studies, Malar. J., 8, 192, 2009.60 http://www.gatesfoundation.org/topics/Pages/malaria.aspx61 S. L. Anderson, J. Berman, R. Kuschner, D. Wesche, A. Magill, B. Wellde, I.Schneider, M. Dunne and B. Schuster: Prophylaxis of Plasmodium falciparum malariawith azithromycin administered to volunteers, Ann. Intern. Med., 123, 771, 1995.62 C. Ohrt, G. D. Willingmyre, P. Lee, C. Knirsch and W. Milhous: Assessment ofazithromycin in combination with other antimalarial drugs against Plasmodiumfalciparum in vitro, Antimicrob. Agents Chemother., 46, 2518, 2002.63 R. M. Chico, R. Pittrof, B. Greenwood and D. Chandramohan: Azithromycin-chloroquine and the intermittent preventive treatment of malaria in pregnancy, Malar.J., 7, 255, 2008.64 http://clinicaltrials.gov/ct2/show/NCT00299208

35

Chapter 2: Computational and other supporting techniques in drug discoveryand evaluation2.1. Computational techniques in drug discovery2.1.1. Molecular mechanics and dynamics

The computational investigation of the liquid compounds began in the 1950s

when Metropolis et al.1 and Alder and Wainwright2 performed the first Monte Carlo

(MC) and molecular dynamics (MD) simulations.

The most widely-used theoretical methods in molecular modelling are 1) ab

initio methods3; 2) the Valence Shell Electron Pair Repulsion theory4; 3) the molecular

mechanical (MM) methods5; and 4) semi-empirical molecular orbital methods6.

Molecular mechanics can be defined as an approach to model the behaviour of

matter. The main assumption in molecular mechanics is that the substance consists of

“atoms” with the potential energy of a collection of atoms for every set of positions.

Together, it is treated as a mechanical system moving with potential energy and is

subject to Newton’s or Schrodinger’s equations. 7

Molecular mechanics calculations consider moving particles in an artificial way,

that only average positions of the particles are meaningful (only equilibrium properties

can be calculated). Molecular dynamics was developed in order to study relaxation

phenomena.8

2.1.1.1. Fundamental concepts of molecular dynamics and mechanics

The chemical bond

Molecular mechanics can be used to calculate molecular properties. Results of the

calculations can indicate whether it was used the proper force field or the wrong

parameters were used.

Bond lengths of covalent bonds can be predicted using the Schrodinger equation.

Proper parameterisation is necessary in order to obtain good and reliable results.

MM1 force field does not take into account the electronegativity of groups. The

correction (l0) regarding this electronic effect causing the shortening of the bond was

added in MM29 and MM310 force fields:11

l0 (new)=l0 (old) + lA + (0.62)lB + (0.62)2lC + (0.62)3lD + ...

lA,lB, lC, lD represent the changes in bond lengths and l0 are bond lengths.

Now, the equation has the form12:

l0 = V0 + (1 + cos) + (1 − cos 2),where V0, V1 and V2 are constants.

The separation between electronegativity effect and trans lone pair effect was

achieved via the equation above.

36

The calculation in molecular dynamics can be represented using the following

scheme:

Scheme 1: Scheme of molecular dynamics simulation13

Molecular movement is governed by the following energy equation:13

However, there are several limitations in molecular dynamics calculations: 1) the

force field requires further refinement; 2) high computational demands require further

developments.13

Binding affinity prediction software is based largely on molecular dynamics such as

thermodynamic integration, single-step perturbation and free-energy perturbation.13

Steric energyThe change in the bond length causes change in the energy (termed the strain

energy). This “stress” is caused by steric interactions. The energy connected with the

energy change is termed as steric energy. 14

The steric energy of the molecule was considered by Westheimer to arise from

various strain-producing mechanisms defined with the structural parameters (bond

angles , torsional angles , bond lengths r, and non-bonded distances d)15:

Esteric = E() + E (r) + E () + E (d)

Molecular mechanics is an approach to represent the energy surface for bond

and non-bonded interactions with simple and analytical energy function. The potential

energy (V) is represented by:

37

Energy minimisationThere are several energy minimisation methods:

1) The Fletcher-Powell method

2) The Conjugate gradient method.

Fletcher-Powell method uses the second derivative method.

The conjugate gradient method starts along the steepest descent until the

minimum in this direction is obtained. It continues in the direction perpendicular, or

conjugate, to this current direction.

The force fieldForce field can be regarded as a mathematical function of the conformation of the

system.

There are several principles of designing the force field:

1) Electrons and nuclei are regarded as atom like particles

2) The shape of atom-like particles are spherical and their charges can be

obtained from the theory

3) Interactions are based on springs and classical potentials

4) Interactions must be specified for particular sets of atoms

5) Interactions are determinations of atom like particles energies.

The parameters in the force field are obtained usually either from ab initio or semi-

empirical quantum mechanical calculations, or by fitting experimental data: neutron, X-

ray and electron diffraction, NMR, infrared, Raman and neutron spectroscopy.

Type of force fieldThe energy equation mentioned previously is valuable for most force fields,

including CHARMM16,17, AMBER18, GROMOS19 and OPLS20.

1. AMBER

One of the molecular mechanics software packages that is widely used is

AMBER. AMBER was developed by American scientists using Fortran 77, and it can

be used in different environments. The force field of the package is based on harmonic

bonded interactions with 6-12 non-bonded interactions. The molecular dynamics

simulations are implemented with constant temperature and pressure with the

possibility of using different constraints and restraints. Before AMBER, there were

several available software packages, but unfortunately they were made only for a

particular class of compounds.21

38

The principal use of AMBER force fields is in the area of proteins and nucleic

acids. Because of the limitations, it has limited use in drug design.22

The effect of the polarisability of molecules is incorporated in AMBER

calculation of the total energy, where the induced moments are in the interaction with

fixed charges and not with each other (bond, angle, dihedral, van der Waals,

electrostatic, polarization components in total energy calculation):

Etotal = ∑ (r-req)2 + ∑ (-eq)2 + ∑ [1 + cos( − )] + ∑ (- +

) + ∑ (μ · ), where:

µi = iEi, Ei = + ∑ μ ,

= ∑ ,

= [3 − 1].Internal restraints can be regarded as “time-averaged”, which can be

calculated using the following equation:

= (1/C)(∫ ( )/ ( ) ′)-1/i

where is the time-averaged value of the internal coordinate, is the

exponential decay constant, r(t’)is the value at time t’, i is the average over internals to

the inverse of i, C is the normalization integral.

The free energy can be calculated according to the equation:23

Gi = G ((i+1)) + G ((i)) = (V/(i+1)) - V/(i) ((i+1) - (i))/2.

2. Allinger’s MM force fields

MM1 force field was launched in 1973 and it was suitable for calculations on

hydrocarbons. Later, the next generation force field appeared, named MM2 with the

improvements of using V2 torsional term and smaller and softer hydrogens.

Gibbs’ energy of a molecule is given with the equation:

G~Vrij + K + PV –TS, where:

Vrij is the equilibrium value of the potential energy, K is the kinetic energy (here,

it is included zero-point vibrational energy and thermal energies), PV is the product of

the pressure and volume, S is the entropy, and T is the absolute temperature.24

3. Merck Molecular Mechanics Force Field (MMFF) force fields

The initial published MMFF version was called "MMFF94." MMFF94 makes

significant approximations in its treatment of important physical interactions. The

39

primary use of this force field is in molecular dynamics simulations more than in

energy-optimisations. In form, it is similar to MM2 and MM3 force fields.

The core portion of MMFF94 was derived primarily from ab initio data. The data

were employed for molecular dipole moments, relative energies, and Cartesian first

and second derivatives obtained from HF/6-31G* calculations to characterise the

quantum mechanical energy surface used to derive the QMFF.

The energy in MMFF94 force field has the following expression:

Energy of the bond stretching is represented with the following equation:

EBij = 143.9325 ∆ (1 + ∆ + ∆ )For the angle bending, the equation is given as:

EAijk = 0.043844 ∆ (1 + ∆ )Stretch-bend interactions are assumed using the equation:

EBAijk = 2.51210 (kbaIJKrij + kbaKJIrkj)vijk

The correction in energy is useful for nitrogen containing compounds:

EOOPijk;l = 0.043844 ; ;Torsion interactions are given by:

ETijkl = 0.5 (V1(1 + cos ) + V2 (1 - cos 2) + V3 (1 + cos 3))

Van der Waals interactions are given using “buffered 14-7 form”:

EvdWij = IJ (. ∗. ∗ )7( ∗. ∗ − 2)

Buffered coulombic buffer is used for electrostatic interactions:25

EQij = 332.0716qiqj/(D(Rij + )n)

Choice of potential energy functionsFree energy of solvation was an unexplored field until 1985. The precision of

calculation in that time was at least ±0.2 kcal/mol by the use of statistical perturbation

theory. 26

Selecting appropriate charges can represent the problem, so choosing of

appropriate faster approaches, semiempirical QM charges with bond corrections can

be, in some cases, a suitable choice.

Water models can be problematic. The most used water models TIP3P and TIP4P

were introduced in 1983. The TIP3P model follows the minimalist TIPS form in having

three interaction sites centred on the nuclei (each site has a partial charge for

40

computing the intermolecular Coulomb energy). The popularity of the TIP3P, TIP4P

and SPC models lies in their functional form which is in compatibility with the common

non-polarisable force fields.27

Force field parameters1. Parameter fitting

Force field parameters can be determined using parameter fitting. Forces from

ab initio quantum calculations of small molecules with the combination of data

obtained from experiments can be useful in determining resonant frequencies of

vibration of certain bonds. The approach is particularly useful in computational studies

of biological molecules. The limitation of this approach is the existence of so many

parameters, which makes determination difficult. These force fields are at the

beginning of appearing. 28

2. Parameters based on experimental structures

Values of parameters in the force field can be obtained from experimentally-

determined structures. Extraction of the structures from experimental structures and

fitting them to a two-body form can be used for the calculation (fold the molecules into

their observed structures). It was found that this method is also working correctly at

finite temperatures.

This method uses two approaches. The first is to start with the wrong

parameters and perform a simulation that gives the parameters close to true (iterative

scheme at finite temperature). The iteration continues until good convergence is

achieved.

The idea of clamping is introduced here. It was introduced as a constraint to

prevent calculations straying from the experimental structures.29

3. Parameters based on theoretical structures

Force fields with the parameters based on theoretical structures can be used

when no experimental structural data are available, or these data are not of

satisfactory quality. There are two possible methods: a) rigid PES (neglecting of cross-

terms constrains); b) relaxed PES (parameters correlation). In order to extract

parameters, the first and second derivatives of the energy can be used. Unfortunately,

the disadvantage of the method can be inability to detect the multiple minima.30

Charges

The charge redistribution is the main reason for the better results achieved

using quantum mechanics instead of classical mechanics. Regarding charges, there

are several obstacles to achieve the right answer:

1) There is no experimental technique for the determination of charges of

atoms inside the molecule;

2) Molecular mechanics are assuming nuclei as separated entities.

41

In order to obtain atomic partial charges using electrostatic potential fitting

methods, the electron density of the molecule must be computed using either

semiempirical, Hartree-Fock, density functional theory or post Hartree-Fock methods.31

Review of Hartree-Fock formalismThe total energy of the system is described by the following equations31:

The final energy expression for the correction terms is given by:

Eelec-corr = - ∑ ∗ − 2 ∑ , a a -∑ , . . − ∑ ∗ − 2∑ . .,, . . a a

EMM-corr = ∑ ( - )2 + ∑ ( - )2 + ∑ , (1 + ) +

, (1 − 2 ) + , (1 + 3 )qB denotes the bond charge and a the frozen orbital. The first used term Eelec-corr

contains electrostatic corrections between QM and MM atoms near the frozen orbitals.

This correction is introduced in order to eliminate short-range electrostatic interactions.

DFT FormalismDFT formalism is an extension of the Hartree-Fock equations using standard DFT

quantum energy expression in place of Hartree-Fock:

EQM-DFT = ∑ ∗ + ∑ [2 - ] + ∑ + ∑

With Kc the DFT exchange correlation matrix:32

= ∫ [ ] ( )( )Local MP2 FormalismCorrelation energy within LMP2 is identical to a normal LMP2 calculation with the

caveat that the frozen occupied orbital space cannot be exited from.

42

Published force fields and their evaluationThe evaluation of the published force fields can be done using:

1. Molecular dynamics simulations starting from the experimental crystal

structure

In order to get the correct force field, it is a prerequisite to present accurate

input for parameterisation. The perturbation method and the supermolecular approach

can be useful for the evaluation of intermolecular interactions.

2. Potential to distinguish between correct and incorrect structures

Particular models can be used to predict trends in solvation energies for

majority of the 20 natural amino acids giving the idea that the models are physically

reasonable, not only over-fitted.33

Optimisation methodsIn relation to optimisation methods, the conjugate gradient and steepest descent

options are actually the first derivative methods and they yield quick results when the

molecule's structure is too far from an energy minimum conformation. The Block-

diagonal Neulton Raphson is a second derivative method so it is more computationally

intensive and takes more time to run. The conjugate gradient method is the best

choice for general application.34

Optimisation methods generally in use are:

*Second derivative methods

*Quasi-Newton methods

*Conjugate gradient methods

*Discrete quasi-Newton methods

*Conjugate direction methods

*Direct search methods

Possibilities and problems with the optimization methods are related to:

1. Structure-scanning of the energy surface

Structure-scanning of the energy surface can be achieved using:

Grid search routines

Molecular dynamics (particularly replica-exchange molecular

dynamics)

Multiple minima problem in energy minimisation studies can be avoided using

simulations based on non-Boltzmann probability weight factors. This enables random

walk in energy space which means that it is allowable to pass any energy barrier and

to sample a much wider space. The method based on these principles is

multicanonical algorithm.35 In multicanonical ensembles, there are performances in a

free 1-dimensional random walk in energy space, in the meantime, in simulated

tempering, there is a free random walk in temperature space.36

43

Molecular dynamics method based on a new generalised-ensemble algorithm

is the replica-exchange methods37 (replica Monte Carlo method, multiple Markov chain

method and parallel tempering). In this method, the weighting factor is already known.

The process of replica exchange molecular dynamics consists of the following

parts:

Preparation

Minimisation

Equilibration

Production

Stochastic (Monte Carlo) methods

Monte Carlo is used for integration and simulation and is easy to use, but the

drawback of this method is that it is very slow. An alternative method for speeding up

the process is using the change of the choice of sequences. Monte Carlo methods

require less sophisticated mathematics than other numerical methods. The difference

is also in stability, which is not present in Monte Carlo methods. The answer provided

by Monte Carlo methods is qualitatively correct, but the work on error of this method

must be present in order to reduce it. Quasi-random sequences are better regarding

errors than standard Monte Carlo quadrature. The effectiveness of quasi Monte Carlo

methods are always lost in problems involving discontinuity (e.g. acceptance-rejection

method).38

Markov chain sampling is the basis of optimisation techniques in simulated

annealing. Improving the accuracy of Monte Carlo simulation is possible using different

reduction methods as I previously stated. Control variable techniques can be

particularly efficient using the LR estimators .∇ ( ). 39

2. Strain energies

Strain energies have a meaning if entropic terms are also included.

Sometimes, the omission of entropy does not cause serious problems. Obtained strain

energies are relative energies. The most serious restriction is that molecular

mechanics may only be used for interpolations. Conformational equilibriums

sometimes cannot be predicted using only force fields based on structural data.

However, comparison of the experimental data with the modelling data proves that

agreement can be really good.40

3. The environment

In molecular mechanics, solvents can be imagined as an extended medium

consisting of the astronomical number of low energy states. There are two main types

of models in sorting out the solvent problem: molecular and continuum models.

44

Solvent effects Continuum models

Continuum models have several advantages over the molecular models. There

is a slow convergence of precise answers because of the large numbers of involved

states and particles. Also, computationally molecular solvent models need several

ranges more CPU time than gas phase calculations of the same solute.

In continuum model, the solvent is treated as a continuous medium having the

average properties of the real solvents and surrounding solvent close to the beginning

or close to Van der Waals surface.

There are several developed continuum models. The most efficient is GB/SA

model giving solvation free energies (Gsol) based on the generalised born (GB)

treatment of Gpol and surface areas (SA) for the cavity approximation and van der

Waals contribution to the solvation. In this model, the total free solvation energy (Gsol)

is given by the following equation:

Gsol = Gcav + GvdW + Gpol,

Where Gcav is solvent-solvent cavity term, GvDW solute-solvent van der Waals

term and Gpol solute-solvent electrostatic polarisation term.

In the case of the saturated hydrocarbons, this is simplified by the calculation

of Gcav + GvdW together by evaluation of solvent-accessible surface areas.

Gcav + GvdW = ∑In this equation, is solvent-accessible surface area of all atoms of type k

and is an empirically determined atomic solvation parameter.

For Gpol, there is an equation which can also be used for solutes of the irregular

shapes:

Gpol = -166.0 (1- )∑ ∑ ( ) .where ij =(ij)0.5 and Dij=rij

2/(2ij)2 and the double sum is going over all pairs

of atoms (i and j); Dij is the squared ratio of the i, jth atom pair separation to their

average born diameters. i-Born radius in simple systems can be van der Waals

radius of the solute, but in complex systems, the Born radius of the ith atom depends

upon the positions and volumes of all other atoms in the solute.41

Explicit solvent models

It was found that for the more complex solvation systems, it is very useful to

compare the results of free energy perturbation simulations with explicit solvent as the

experimental data. In that case, explicit/implicit solvent approaches can be analyzed

and compared.

Using the explicit solvent model, the energy difference between the two related

states are equal to:

45

G=-kTln(exp(-[H2-H1]/kT))1 = -kTln(exp(-[V2-V1]/kT))1

In this equation H1 and H2 are the Hamiltonians, V1 and V2 are the total

potential energy of the two states.42

2.1.2. Molecular modelling of proteinsMolecular modelling on proteins represents an extensive field of research in

molecular modelling. Molecular dynamics and mechanics calculations perform on a

wide variety of protein models, and several new force fields were generated in order to

help solving the questions on conformations and structures of proteins with higher

accuracy.

Sequence analysisProblems in sequence analysis are mostly of a statistical nature. Some of the

probabilistic models like HMM (hidden Markov models) are in wide use in sequence

analysis of proteins.

I was impressed when I read in one book that God (nature) is not a thinker, that

it is actually an inventor. Because of this universal principle, sequence analysis was

possible using the homology. The homology principle consists of similarity between

two sequences.

The plausible alignment of two sequences based on homology modelling is

possible using insertions, deletions and substitutions in the sequences.

Pair-wise alignment has statistics as a basis. Gap penalty treating is one of the

main concerns and it is given by linear score:

(g) = -gd

or affine score: (g) = -d-(g-1)e. In these two previous equations, d is open-gap

penalty and e is gap-extension penalty.

The local version of sequence alignment using dynamic programming

sequence alignment algorithm was developed in the 1980s and was called Smith and

Waterman, according to their inventors.

Comparison of the models gives us the opportunity to evaluate the correctness

of them. The Bayesian approach is used for this purpose. 43

Molecular modelling softwares and resourcesNowadays, there are plenty of software and internet resources for sequence

alignment. The most important are:

1. GenBank http://www.ncbi.nlm.nih.gov/Genbank/

GenBank is a database that contains publicly available nucleotide sequences

for approximately 400,000 organisms, with the names given at the genus level or lower

46

obtained by the submissions through the individual laboratories and batch submissions

from individual laboratories. The majority of submissions was made using the web-

based BankIt or standalone Sequin programmes. Worldwide coverage was ensured by

the exchange with the European Nucleotide Archive (ENA) and the DNA Data Bank of

Japan (DDBJ). Sequence similarity searches of GenBank was provided using

BLAST.44

2. Protein Data Bank http://www.rcsb.org

The Protein Data Bank history began in the 1970s. By the beginning of this

century, the number of structures increased, which enables actual understanding of

biology and medicine at the molecular level.

The entry in the PDB contains xyz coordinates, but also chemistry of the

macromolecule, the small-molecule ligands, some details of the data collection and

structure refinement and structural descriptors as well. Typical PDB entry consists of

about 400 unique items of data. The PDB file format was devised in 1976. It is simple

to read and can be used with numerous computer applications.

There are some shortcomings in the temporary database. Improvements must

be made for better representation of disordered structures, X-ray structures refinement

with multiple models and very large macromolecular complexes. 45

3. ExPASy http://ca.expasy.org/

The Expert Protein Analysis System (ExPASy) World Wide Web server is a

service constructed by a multidisciplinary team at the Swiss Institute of Bioinformatics

(SIB) to provide to the life science community a variety of databases and analytical

tools dedicated to proteomics and proteins. ExPASy databases consist of TrEMBL,

SWISS-2DPAGE (proteins sequences identified by the two-dimensional

polyacrylamide gel electrophoresis), PROSITE (database of proteins domains and

families), ENZYME (database of the nomenclature of enzymes) and the SWISS-

MODEL repository (database of the automatically determined protein models).

Available tools include specific tasks related to proteomics, similarity searches, pattern

and profile searches, post-translational modification prediction, topology prediction,

primary, secondary and tertiary structure analysis and sequence alignment. ExPASY

began in 1993 as the first www server in the field of the life sciences. In addition to the

main site in Switzerland, there are seven mirror sites in different continents.

SWISS-PROT and TrEMBL are maintained in collaboration of SIB and

European Bioinformatics Institute (EBI).46

4. NCBI http://www.ncbi.nlm.nih.gov

NCBI is a database server providing 1) literature databases; 2) Entrez

databases; 3) nucleotide databases; 4) genome-specific resources; 5) tools for data

mining; 6) tools for sequence analysis; 7) tools for 3-D structure display and similarity

47

searching; 8) maps; 9) collaborative cancer research; 10) FTP download sites; 11)

resource statistics.

5. EMBL-EBI http://www.ebi.ac.uk/

A comprehensive database of DNA and RNA sequences submitted by

researchers and genome sequencing groups, and collected from the scientific

literature and patent applications, makes up the EMBL Nucleotide Sequence

Database. This database collaborates with DDBJ and GenBank. Dataflow between

these institutions are represented in the following figure:47

Figure 1: Dataflow between international databases DDBJ/EMBL/GenBank.

6. InterPro http://www.ebi.ac.uk/interpro/

The InterPro database contains predictive tools or signatures which represent

protein domains, families and functional sites from multiple, diverse source databases:

Gene3D, PANTHER, Pfam, PIRSF, PRINTS, ProDom, PROSITE, SMART,

SUPERFAMILY and TIGRFAMs. Manual integration is present and around half of the

total - 58 000 - signatures are available in the source databases belonging to an

InterPro entry. Moreover, they provide non-signature data, like structural data, in new

XML files on their FTP site.48

7. UniProt http://www.uniprot.org/

UniProt (The Universal Protein Resource) is a stable, freely accessible

resource for protein sequences and functional annotation. This consortium is a

collaboration between the EBI, the Protein Information Resource (PIR) and the SIB.

Main activities are connected to manual curation of protein sequences, sequence

archiving, and development of a user-friendly UniProt website. The UniProt system is

consisted of the UniProt Knowledgebase, the UniProt Reference Clusters, and the

UniProt Archive and the UniProt Metagenomic and Environmental sequences

database. 49

48

8. SMART http://smart.embl-heidelberg.de/Simple Modular Architecture Research Tool (SMART) is a web device for the

identification and annotation of protein domains and gives the basis for the

comparative study of complex domain architectures in genes and proteins. Currently,

the developments in SMART are based on the integration data from completed

metazoan genomes. Visualisation tools were developed to allow analysis of gene

intron-exon structures within the context of protein-domain structure.

There are several new improvements into SMART:

1) Batch access: It means that the job can be submitted directly to the server

2) Intron positions are shown in the schematic protein figures

3) Extra information are available in the main results page.50

9. Pfam http://pfam.sanger.ac.uk/

Pfam is a collection of protein domains and families, represented as multiple

sequence alignments and using as a profile HMM. It is based currently on the

UniprotKB sequence database. It has also basis in NCBI GenPept and sequences

from the selected metagenomics projects. This database is available from consortium

members using a new, consistent and improved web design in the UK, USA and

Sweden. Mirror sites are also available in France and South Korea. 51

10. PRINTShttp://www.bioinf.manchester.ac.uk/dbbrowser/PRINTS/index.php

The PRINTS database is a collection of protein fingerprints. This can be used

in order to assign uncharacterised sequences to known families. An automatic

supplement, preprints was designed in order to increase the coverage of the resource

and reduce some of the manual burdens.

There are major and minor versions of PRINTS:

1) Minor reflects updates

2) Major releases denote the addition of new material to the resource.

The tools for searching PRINTS include:

1) BLAST server for searches

2) The FingerPRINTScan suite searches against fingerprints.

The following figure illustrates the automated preprints pipeline:

49

Figure 2: Illustration of the automated preprints pipeline.52

11. BLAST http://blast.ncbi.nlm.nih.gov/Blast.cgi

Basic local alignment tool (BLAST) represents the new approach to rapid

sequence comparison which gives direct approximation of alignments in order to

optimise a measure of local similarity. The basic algorithm of BLAST is very simple; it

can be implemented and applied in a number of ways including DNA and protein

sequence database searches, motif searches, analysis of multiple regions of similarity

in long DNA sequences. Compared with other comparison tools, BLAST is faster. 53

12. FASTA http://www.ebi.ac.uk/Tools/fasta/index.html

FASTA represents the algorithm for searching the similarities between the

sequences of newly-discovered amino acids and sequences available in databases.

The method shows an efficiency in identification of regions of similar sequence

giving the score of the aligned identical and differing residues in those regions as an

amino acid replaceability matrix.

Programmers wrote FASTP in the ”C” programming language.54

13. PSI-BLAST http://www.ebi.ac.uk/Tools/psiblast/

The position-specific iterated BLAST (PSI-BLAST) programme runs at

approximately the same speed per iteration as gapped BLAST. This programme was

used to uncover several new and interesting members of the BRCT superfamily.

It was found that PSI-BLAST output depends critically on avoiding the

inadequate inclusion of sequences in the multiple alignment constructes. During

allignment, the new gap generate local alignment of better quality and more

satisfactory. This can lead to more sensitive position-specific score matrices.

PSI-BLAST has several different operations:

1) It makes the construction of multiple alignment from BLAST output data

2) It gives the alignment into a position-specific score matrix

3) It uses the matrix mentioned above in order to search the database.

PSI BLAST imposes the strict scaling rules on the matrices and generates

50

more accurate alignments. It has an ability to report accurate statistics and it can be

used both iteratively and fully automatically. 55

14. HMMER http://hmmer.janelia.org/

The software package which provides the possibility for protein sequence

similarity searches using probabilistic methods is HMMER. Methods on the server are

available for searching either a single protein sequence, multiple protein sequence

alignment or HMM against a target sequence database. This software package was

used by protein family databases such as Pfam and InterPro. Currently, HMM is

improved regarding the speed, but also there is an improvement in sensitivity.

The software package includes four database search programmes for protein

sequence analysis: phmmer (analogues to BLASTP, with BLOSUM62 as a scoring

matrix), hmmscan, hmmsearch and jackhammer. 56

15. ClustalW http://www.ebi.ac.uk/Tools/clustalw2/index.html

For multiple alignment of nucleic acid and protein sequences and for the

preparation of phylogenetic trees the Clustal series of programmes can be used. New

characteristics of the Clustal series include NEXUS and FASTA format output giving

the print of range numbers and faster tree calculations. The first idea for the Clustal

was for it to be run on local computers, but later numerous web servers were used for

that purpose, the most notable at the EBI.

ClustalW represents the third generation of the Clustal series, which includes

number of improvements to the alignment algorithm, sequence weighting, position-

specific gap penalties and the automatic choice of a suitable residue comparison

matrix at the each stage of the multiple alignments.

The web server with Clustal uses the uploaded file from the local computer

where sequences can be used in several formats (GSG, FASTA, EMBL, GenBank,

PIR, NBRF, Phylip or SWISS-PROT).57

16. TMHMM http://www.cbs.dtu.dk/services/TMHMM/

TMHMM represents a membrane protein topology prediction based on HMM. It

was shown that it predicts correctly in 97-98% of the trans-membrane helices. Also,

using TMHMM, it is possible to make difference between soluble and membrane

proteins with specificity and sensitivity higher than 99%. It shows superior behaviour

compared with TOPPRED and ALOM.58

Secondary structure predictionThe sequence and nature of amino acids determine the secondary structure of

the protein. In case the X-ray image of the crystal structure of the protein is not

available, the software for the prediction of the secondary structures must be used.

Jpred3 and PSIPRED represent the secondary structure software prediction.

51

1. Jpred3 http://www.compbio.dundee.ac.uk/www-jpred/

Jpred is a second structure prediction server using the Jnet algorithm. The

current Jnet algoritm gives a three-state prediction (-helix, β-strand and coil) of the

secondary structure with the accuracy of 81.5%. Using a single protein sequence or a

multiple sequence alignment, Jpred derives alignment profiles. The output is available

as coloured HTML, but it is also available in PDF format.59

2. PSIPRED http://bioinf.cs.ucl.ac.uk/psipred/

For the prediction of the secondary structure a two-stage neural network can

be used using the positioning of the specific scoring matrices made by PSI-BLAST.

The used approach is simple and convenient, and results are better compared with the

popular PHD method.

This prediction method is base on three stages: 1) generation of a sequence

profile; 2) prediction of initial secondary structure; 3) filtering of the predicted structure.

Figure 3: The outline of the PSIPRED method showing the processing of PSI-BLAST

score matrices60

Tertiary structure predictionTertiary structure prediction of the protein usually uses the information from the

secondary structure of the protein.

Tertiary structure of the protein can be obtained using:

1. homology or comparative modelling

2. threading

52

3. ab initio or de novo modelling.

2.1.2.1. Template-based modelling (homology modelling)Template-based modelling can be used when there is a similarity between the

template protein within the known structure and the target protein.

The structures of the homologous proteins can be divided into regions in which

the general fold of the polypeptide chain is very similar and those where it is quite

different.

There are many web servers available for homology modelling. The description

and use of them I have reported previously. CASP experiments are providing

information about the quality of the servers that can be used for homology modelling.

The threshold for safe homology modelling can be lower than 25%, particularly

for long sequences. Unfortunately, sequences with 25% sequence identity gives poor

structural models with pretty high uncertainty in side chain positioning. When the chain

identity is greater than 75%, protein homology modelling can be easily achieved using

minor manual intervention. If the sequence identity is between 50-75%, more time is

needed to fine-tune the details of the model and in that case the correct alignment is

necessary. In case of 25-50% identity, the best possible alignment is a concern of the

process of homology modelling. When the sequence identity is lower than 25%, no

template structure can be detected with a simple BLAST techniques.

The homology modelling consists of the following steps:

1) Template recognition and initial alignment

2) Alignment correction

3) Backbone generation

4) Loop modelling

5) Side chain modelling

6) Model optimisation

7) Model validation

8) Iteration. 61

2.1.2.2. Free modelling (ab initio or de novo modelling)Ab initio or de novo modelling is based on quantum mechanics and provides

the most accurate and consistent predictions for chemical systems. The main problem

with ab initio calculations is that they are extremely computer-intensive.

Calculations which require the time can only be used with basic theory (i.e.

Hartree Fock approximation) with minimal basis sets.

Ab initio calculations for many-body systems involves some level of

approximation and some level of empirical parameterisation.

53

Ab initio methods for molecular calculations must satisfy a set of stringent

criteria:

1) Solutions must be well defined

2) The potential energy of the molecule must vary

3) The model must contain no bias

4) The model must be “size-consistent”

5) The model must be variational.62

There are various servers and software programmes for the prediction of the

tertiary structure. In my opinion, the best are:

1) SSPro; 2) MODELLER; 3) Rosetta; 4) I-TASSER.

1) SSPro http://scratch.proteomics.ics.uci.edu/

SCRATCH is a server that can be used for the prediction of the protein tertiary

structure and structural characteristics. It includes predictors for secondary structure,

relevant solvent accessibility, disordered regions, domains, disulfide bridges, single

mutation stability, residue contacts, individual residue contacts and tertiary structure.

The use of the server is very simple, the user only needs to provide an amino acid

sequence and selects the desired predictions and submit information to the server.

The SCRATCH package contains the following main modules:

1) Sspro: three class secondary structure

2) Sspro8: eight class secondary structure

3) ACCpro: relative solvent accessibility

4) CONpro: contacts with other residues compared to average

5) DOMpro: domain boundaries

6) DISpro: disordered regions

7) Mupro: effect of single amino acid mutation on stability

8) Dipro: disulfide bridges

9) CMAPpro: residue-residue contact maps

10) 3Dpro: tertiary structure.

54

Figure 4: Flow diagram for SCRATCH server (in this representation DISpro, DOMPro,

CONpro and Dipro are grouped together because of the same input and output)63

2) MODELLER http://salilab.org/modeller/

Predictions of the structure of proteins is a difficult job, caused mainly by

possible errors in alignment and loop modelling.

There are three main components in modelling the loop: 1) the representation

of a protein; 2) the restraints defining ”energy” function; and 3) the method for the

optimisation of the energy function. The modelling method is completely automated in

MODELLER-5. Hydrogens are removed from the structure used for the manipulation in

MODELLER. Explicit solvent molecules are not present and ligands are not included;

degrees of freedom are the Cartesian coordinates of the loop atoms.

The best energy function is given as:

55

In this equation, b is the bond length, is a covalent bond angle, is a dihedral

angle, is an improper dihedral angle, R is a residue type. The non-bonded atom pairs

with the distance more than 4 Å are ignored.64

3) Rosetta http://robetta.bakerlab.org/

The Rosetta programme is designed to resolve many molecular modelling

problems (e.g. fibril structure prediction to RNA folding and design of new protein

interfaces). The methodology presented in Rosetta enables the creation of novel

molecules with useful functions and gives a promise for faster experimental structural

inference.

In Rosetta, atom-atom interactions can be computed using a Lennard-Jones

potential to describe packing, an implicit solvation model to describe the hydrophobic

effect and the electrostatic desolvation cost connected to the burial of polar atoms, an

explicit hydrogen-bonding potential describing the hydrogen bonding.

Figure 5: Scheme of Rosetta molecular modelling describing de novo structure

prediction: (a) low-resolution fragment; (b) final low-energy conformation produced by

fragment assembly; (c) all-atom model produced after high-resolution refinement65

If we examine the meaning of a molecule being “folded”, it means that it has a

very high probability of being in a single native state. In this case, when this probability

is more than 99.9%, the free-energy gap between the native state and the group of

non-native state must be at least G=kBTlog (0.999/0.001) = 4 kcal/mol, using the

Boltzman equation. In reality, the free-energy gap is usually in the range 3-10 kcal/mol.

The conformational sampling in Rosetta consists from two stages:

1) Firstly, the global minimum must be searched, which involves a large number

56

of local minima. It is necessary to explore as many as possible local minima because

of the coarse-grain smoothing, which can introduce large errors owing to the missing

critical contribution of interatomic packing to the true free energy. Low-resolution

structure prediction for protein and RNA molecules is based on a picture of folding.

2) The second stage in the search for the lowest free-energy minimum starts from

each of the alternative minima identified in the initial low-resolution search. In the case

of proteins, it is necessary to perform firstly the simulated annealing through the

combination of amino acids rotamers. Then, multistep Monte Carlo minimisation

procedure must be applied composed on numerous torsion angle perturbations (each

of them is followed by rotamer optimisation) and gradient-base minimisation of side-

chain and backbone torsion angles.65

4) I-TASSER http://zhang.bioinformatics.ku.edu/I-TASSER/

I-TASSER was constructed in order to be used for protein structure and

function predictions. The process of the construction of the protein consists of several

steps:

1) When the sequence of amino acids is submitted, the server tries to

retrieve the protein template of similar folds (or super-secondary structures) from the

PDB library by LOMETS.

2) The continuous fragments taken from PDB templates are reassembled

into full-length models by replica-exchange Monte Carlo simulations with the threading

unaligned regions built by ab initio modelling. If no appropriate template is identified, I-

TASSER will build the complete structure using ab initio modelling. Low free-energy

states are identified by SPICKER.

3) The fragment assembly simulation is performed again starting from the

SPICKER cluster centroids (spatial restraints are collected from both the LOMETS

templates and the PDB structures by TM-align (guide for simulations)). The second

iteration has as a purpose to remove the steric clash and to refine the global topology

of the cluster centroids). The final models are getting obtained by REMO building the

atomic details from the selected I-TASSER decoys through the optimisation of the

hydrogen-bonding network.66

57

Figure 6: I-TASSER scheme for protein structure and prediction of function66

58

In order to check the quality of the modelled proteins, different software packages

and servers can be used:

1) PROCHECK; 2) WHAT IF; 3) ProSA; 4) Verify 3D.

1) PROCHECK http://www.biochem.ucl.ac.uk/_roman/procheck/ procheck.html

PROCHECK suites of programmes give a detailed check of the

stereochemistry of a protein structure. They are useful for assessing the quality not

only of protein structures, but also of existing structures.

The PROCHECK package comprises five programmes (four written in

Fortran77 and one in C): CLEAN.F (cleaned-up version of the initial structure),

SECSTR.F (secondary-structure assignments), NB.C (all non-bonded interactions

between different pairs of residues), ANGLEN.F (measuring all main chain bond

lengths and angles) and PPLOT.F (all plots and residue by residue listing).67

2) WHAT IF http://swift.cmbi.kun.nl/whatif/

WHAT IF is a computer programme written in Fortran 77 for use in

macromolecular modelling and drug design. The software gives an opportunity to

display, manipulate and analyse small molecules, proteins, nucleic acids and their

interactions. There is a protein database in the software and it is also suitable for

crystallographic work.68

Figure 7: Organisation of WHAT IF software68

59

3) ProSA https://prosa.services.came.sbg.ac.at/prosa.php

The ProSA programme (protein structure analysis) has an application in the

refinement and validation of experimental protein structures and in structure prediction

and modelling. The quality scores of a protein are displayed in the context of all known

protein structures and problematic part of the structure using a 3D molecule viewer.

This is particularly useful for the validation of structures obtained from X-ray analysis,

NMR spectroscopy and theoretical calculations.69

4) Verify3D http://nihserver.mbi.ucla.edu/Verify_3D/

For the testing models, 3D profiles should be used. In structure verification, R-

factor in X-ray analysis depends on comparison of observed properties. In Verify3D,

3D profile for model evaluation relies only on a comparison of the model to its own

amino acid sequence.70

2.1.3. Molecular modelling of RNAsIn molecular modelling of RNAs, the primary problem is how to derive the

proper folding using the RNA sequence.

Typical RNA secondary structuresHydrogen bonds in RNA structures are possible between C and G, and A and U

pairs, but also between less stable G and U pairs. In most cases, isolated base pairs

are usually unstable, but there are seldom more than 10 pairs in an unbroken helix.

Various types of single-stranded regions occur between helices (hairpin loops are

connecting two sides of a single helix, bulges and internal loops are connecting two

helices, multibranched loops-connecting the two sides of a single helix, multibranched

loops are connecting three or more helices).

Typical RNA secondary structures are presented in Figure 8.

Figure 8: Typical RNA secondary structures

60

Secondary StructureThe function of RNA can be predicted by knowing the secondary and tertiary

structure of RNA. It is very well known that RNAs possess catalytic activity but, in

order to predict it, the knowledge about the secondary structure is not sufficient.

The modelling of the secondary structure of RNA is only the initial step and very

significant in modelling the tertiary structure of RNA. 71

In case of a small number of known sequences of RNA, free energy minimisation

can be used to predict the secondary structural models.72

Homology principleIt is known that for the homology modelling of proteins, the 20-30% of sequence

identity is at the lower limit; however, this limit has not been detected yet for RNAs.

The same software and servers can be used in order to perform alignments and

predictions of structures of RNAs (PSI-BLAST, FASTA, CLUSTALW, MUSCLE 73, T-

Coffee 74).

There are several methods developed in order to perform the secondary

structures alignments. They rely on: 1) matching defined RNA sequence/structural

patterns; 2) HMM or Stochastic Content Free Grammar (SCFG) methods; 3) classical

sequence alignment combined with the maximal pairing algorithms.75

2.1.3.1. Molecular modelling of the secondary structure of 23S rRNAExpanded sequence dependence of thermodynamic parameters improves

prediction of RNA

Molecular modelling of the secondary structure of RNA depends on the prediction

by free energy minimization. Algorithmic improvements which enhance the prediction

of the secondary structure include reduced search time and storage for multibranch

loop free energies and improved imposition of folding constraints.

The improved algorithm is available in three forms: 1) RNAstructure 3 for Windows

platform (written in C++); 2) mfold package for the prediction of folding and dot plots (a

collection of FORTRAN and C programmes); 3) on-line server.76

Dynalign: An algorithm for finding the secondary structure common to two RNA

sequences

On average, RNA secondary structure prediction for a single sequence is 73%

accurate. This is accurate enough to be a good starting point for the determination of a

secondary structure either by comparative sequence analysis or by the interpretation

of experimental studies. The improvement in structure prediction was achieved by the

combination of free energy minimisation and comparative sequence analysis to find a

61

low free energy structure common to two sequences without requiring any sequence

identity.

It is available in two forms: 1) ANSI C++ code suitable for compilation on any

platform; 2) user-friendly RNAstructure package for the secondary structure

prediction.77

1) RNA structure

RNA Structure is a software package for the prediction of the secondary structure

and analysis. It is based on thermodynamics and uses the more recent set of nearest

neighbour parameters. The methods for secondary structure prediction (several

algorithms are in use) are included; prediction of base pair probabilities, bimolecular

structure prediction and prediction of structure common to two sequences.78

2) Sfold

Sfold is a nucleic acid folding software package. It is based on a new statistical

sampling paradigm for the prediction of the RNA secondary structure. Major objective

of this software is rational design of RNA-targeting nucleotides (small interfering

RNAs, antisense oligonucleotides and trans-cleaving ribozymes.79

3) Mfold

Mfold represents a number of closely-related software applications available on

internet for the prediction of the secondary structure of single stranded nucleic acids.80

4) RNAfold

In order to compute and compare RNA secondary structures, the Vienna RNA

package was developed based on dynamic programming algorithms and aim to

predict structures with minimum free energies with the calculations on the equilibrium

partition functions and base-pairing probabilities.

All computer codes are written in ANSI C including implementations of modified

algorithms on parallel computers.81

5) ViennaRNA

The Vienna RNA secondary structure provides an interface for the analysis of RNA

secondary structures. It gives the opportunity of the prediction of secondary structure

from a single sequence, prediction of the consensus secondary structure for a set of

aligned sequences and the design of the sequences that will fold into a predefined

structure.82

2.1.3.2. Molecular modelling of the 3D structure of 23S rRNAMolecular modelling of RNAs can be performed using available software:

1) MANIP; 2) YAMPP; 3) NAB; 4) ERNA-3D; 5) S2S; 6) MC-SYM; 7) RNA_2D3D.

62

1) MANIP

RNA represents a big assemble. Because of that, it must be divided into small

pieces (helices, hairpin loops, other recurrent motifs, etc.). MANIP allows the rapid

assembly of separate motifs (each with the specified sequence) into a three-

dimensional structure. Modules previously determined are presented in a database

from where it can be extracted with the appropriate sequence. One of the

performances of MANIP allows the recognition and display, which provides the

opportunity of representation of allowed and non-allowed hydrogen bonds between

residues. It also contains rapid and automatic online refinement tools of partial or full

assemblies; NUCLIN-NUCLSQ. This refinement protocol contains canonical and non-

canonical pairing constraints with restraints imposed by covalent geometry,

stereochemistry, and Van der Waals contacts.83

Figure 9: Main steps of the modelling process used with MANIP83

2) S2S

S2S (Sequence to Structure) represents the framework in which it is easy to

display, manipulate and interconnect heterogeneous RNA data (sequence alignments,

secondary and tertiary structures). It was implemented using the Java language.84

3) RNA_2D3D

RNA2D3D automatically and rapidly produces a first-order approximation of a 3-

dimensional conformation consistent with the provided information. It represents an

efficient tool applicable to structures of arbitrary branching complexity and psedoknot

content. Also, it is an efficient interactive graphical editor for the removal of any

overlaps introduced by the initial generating procedure and for making conformational

changes. In this application, it is possible to conduct fast exploration of alternative 3D

63

conformations. In that case, it is possible to add or delete base pairs interactively,

neighbouring atoms can be coaxially stacked or unstacked, and single strands can be

shaped to accommodate special constraints. Compaction and stacking within stems is

optimally extended into connecting single strands and it is available of strategically

making the structures more compact and revealing folding motifs.85

2.1.4. Molecular dockingMolecular docking can be defined as a computational scheme that aims to find the

best match between two molecules: a ligand and a receptor.

The first practical suggestions came from Crick, who suggested that

complementarity in helical coiled coils could be modelled as knobs fitting the holes.

The first pioneering and widely-used docking programme was GOLD.

There are three main components in docking: 1) representation of the system; 2)

search of the conformational space; 3) ranking of the potential solutions.

The surface of the protein can be described by mathematical models using

geometrical shape descriptors or a grid. 86

1. Basic components

The basic components of molecular docking are:

a) The search algorithm;

b) The scoring function.87

A large number of functions exists today but, in principal, there are two main

groups: 1) scoring schemes based on physical interaction terms (van der Waals

interactions or hydrogen bonding); 2) first principle based scoring functions, terms are

derived from physico-chemical theory and they are not fitted to experimental data

(examples are PMF and DrugScore scoring function).

64

2. Choosing the correct docking software

Figure 10: The algorithm which can be implemented in order to find the appropriate method for modelling protein-ligand interactions (QSAR,

quantitative structure-activity relationship; SBDD, structure-based drug design)87

65

In some algorithms, the ligand is built up incrementally, starting from a docked base

fragment: Hammerhead, DOCK, FlexX. In others, AutoDock, Gold, ICM-Dock and

QXP the ligand is treated in its entirety.88

1. Preparing the molecules

Macromolecule

Preparation of the molecule for docking depends on the software used for docking.

Treating macromolecules can be done using the software, Fitted. The manipulation

inside the software is represented by Figures 11 and 12:

Figure 11: Generation of the population within Fitted 2.689

66

Figure 12: Schematic of the evolution cycle of FITTED 2.6.89

Small molecule ligands

In dealing with small molecules, the attention must be placed on docking

algorithms and de novo design methods. Examples of de novo design tools are

BUILDER, CONCEPTS, CONCERTS, DLD/MCSS, Genstar, Group-Build, Grow,

HOOK, Legend, LUDI, MCDNLG, SMOG and SPROUT. The most useful docking

method will be easy to use and parameterise, and is sufficiently adaptable such that

different functionality may be selected.

It was shown that the use of the rigid receptor/flexible ligand approximation is well

established, and the most successful programmes achieved a success rate of 70-

80%.90

67

Protein ligands

Ligand-binding sites of proteins can be predicted using protein-ligand docking

computation. Using some methods, random libraries can be docked onto the whole

protein surface. The affinity of the true-ligand binding site would be correlated to the

docking scores of the compounds in the random library. This method can be called the

molecular-docking binding-site finding (MolSite) method.91

2. Iterative docking and analysis

In iterative docking algorithm, after performing of the initial FF Dock, QM/MM

calculation is applied to obtain a new set of charge values. With the new charge set,

another docking is performing. These two basic steps can be repeated in order to

achieve improved results.92

3. Post-analysis

Post-analysis can be performed using software such as Discovery Studio or Pymol

or AutoDock Tools.

4. Virtual screening

Virtual screening emerged as a reliable, inexpensive method for identifying the

leading compounds. It has a positive impact on the discovery process.

A complement to experimental screening (HTS), coupled with structural biology,

virtual screening can increase the success. Various methods can be used, such as

similarity and substructure searching, pharmacophore matching and 3D shape

matching.

68

Figure 13: A description of the workflow of a virtual screening run against a specific

target 93

2.2. Supporting techniques used for the determination of macrolideantibiotics in chemical and biological media

2.2.1. Nuclear magnetic resonance techniques

Nuclear magnetic resonance (NMR) spectroscopy was discovered in the 1940s,

and it has from that time experienced a spectacular growth and acceptance in

chemistry. Sensitivity was the main factor affecting its application. In the first period,

the main issue was application on high abundance magnetically active nuclei such as1H, 19F and 31P. The main progress in instrumentation has centred on:

1) increasing the applied magnetic field to increase spectral dispersion and

simplify resonance patterns

2) improving the homogenity of the applied magnetic field for greater resolution of

spectral lines

3) stabilising frequency field ratios for drift-free operation

4) sensitivity improvement

5) multiple-resonance capability (decoupling)

6) and “other nuclei” observation.94

69

Generating NMR phenomenon has been described in many publications and

internet pages such as:

www2.chemistry.msu.edu/faculty/reusch/VirtTxtJml/spectrpy/nmr/nmr1.htm.

In NMR spectroscopy, strong magnetic fields are used. The international unit

for magnetic flux is Tesla (T). Small energy difference (Δt) is usually given as a

frequency in units of MHz (106 Hz), ranging from 20 to 900 MHz, depending on the

magnetic field strength and specific studied nucleus.

Protons in different chemical environments have different chemical shifts.

One factor contributing to differences in chemical shifts are inductive effects.

Elements that are more electronegative than carbon reduce the electron density. The

de-shielding effect of electron withdrawing groups is roughly proportional to their

electronegativity. The strength of the signal is displayed along the vertical axis of a

spectrum, and is proportional to molar concentration of the sample.

Intramolecular hydrogen bonds, especially those defining a six-membered ring,

generally display a very low-field proton resonance.

-Electrons are more polarisable than are σ-bond electrons. In benzene systems,

the external field induced -electron movement giving strong secondary fields that

perturb nearby nuclei. This kind of spatial variation in benzene systems is called

anisotropy and it is usual to find in non-spherical distribution of electrons.

Natural abundance of the stable magnetically active nuclide of carbon (13C, 1.1%

natural abundance, spin=1/2) gives as a consequence chances of existence of two 13C

in one molecule very low, without further complications of homonuclear spin-spin (J)

coupling.95

Decoupling in 13CBecause of the spin number of proton (1/2), proton can couple to 13C and

provides a coupled spectrum. If an intense 1H radio frequency decoupling field is used

to irradiate the protons of the methyl group (which can appear in coupled spectrum as

quartet with intensities 1:3:3:1), it will display one narrow line centred at the chemical

shift of methyl carbon. This experiment can be termed as “coherent single-frequency”

decoupling since only one type of proton in the proton spectrum was irradiated. In

case of complicated molecule, the approach is not working and only broadband

decoupling in which the coherent proton radio frequency is modulated with “white”

noise. 96

Time averagingWeak nuclei (e.g. 13C) require time-averaging (multiscanning) for extraction of

the signals. Long periods of time were necessary in the CW (continuous wave) mode

70

(~500 s). The main advantage of the pulsed Fourier transform method is that each

scan requires only about 1s.96

Pulsed Fourier Transform NMRThe Fourier transform (FT) pair is formed by the steady state and impulse

response in a nuclear spin system at equilibrium. The powerful (kW) field is used

which is pulsed on for times as short as a few microseconds. The response of the

entire spin system is picked up in the normal manner, amplified, and detected in the

spectrometer system. The obtained signal is digitized in an analog-to-digital converter

(ADC) and stored in a computer. For signal-to-noise improvement, subsequent

“transients” are taken in a similar manner and are added to the previously stored data.

Attainable resolution is determined by the length of time recording the response of the

spin system to the pulse (the free induction decay or FID), i.e., resolution ~1/AT. The

limiting resolution attainable will be that corresponding to the natural linewidth (T2)-1

or magnet homogeneity-dictated linewidth.96

Characteristics of 13C NMR spectrumThe 13C chemical shift scale follows that of the proton, including commonly

accepted internal reference tetramethylsilane (TMS).

Figure 14: Structure of tetramehylsilane

Zero of the scale belongs to TMS, resonances to higher frequency (at constant

field) are considered to have positive chemical shifts in ppm-the same as the proton

“” scale. Aliphatic carbons are at lower frequencies (higher shielding) and

consequently have lower chemical shifts. The resonances appeared at higher

chemical shifts with increasing substitution, and particularly heteroatom substitution.

Aromatic and olefinic carbons occur at 90-170 ppm; in contrast, carbonyl carbons have

chemical shifts at 160-220 ppm. Prediction of 13C chemical shifts can be done on the

basis of substituent parameters. This approach usually relies on model compounds

and minimal number of substituted compounds. Grant et al.97 and later Lindeman et

al.98 developed regression analyses for linear and branched hydrocarbons which give

nice comparisons between calculated and experimental shifts. The chemical shift of

the paraffinic carbon k can be expressed as:

71

k = A+ i

iki nB

A-appropriate reference compound, Bi-the ith substituent effect, nik-the number of

substituents of type i relative to carbon k

For branching situations, further parameters are necessary. Parameter sets

have been developed for cyclic hydrocarbons.99 Large effects are present when there

is replacement of a substituent carbon with a polar substituent.100

Coherent Single-Frequency Off-Resonance DecouplingWhen the 1H decoupling power is insufficient and/or the 1H frequency is not exactly set

at the resonance frequency of the proton(s) coupled to the carbon of interest, some

residual coupling will occur, shown as a line broadening or actual breaking up of the

singlet into the multiplet characteristic for the number of protons on the carbon. The

advantage of this approach is the characterisation of the carbon in terms of the degree

of protonation, retention of the full nuclear Overhauser effect, and reduced splittings,

making interpretation convenient.96

Gated decouplingBy turning off the proton decoupler, a coupled spectrum can be obtained. The

Overhauser enhancement can be taken back by turning on the decoupler for a period

just prior to the monitoring pulse and acquisition. The coupled spectrum and

Overhauser enhancement may be obtained by automatically turning off the decoupler

before the pulse and on again after completion of acquisition. Instantaneously the

coupled spectrum is obtained. The reverse “gating” sequence accomplishes the

“suppressed Overhauser spectrum”.96

Suppressed Overhauser SpectrumThe Suppressed Overhauser spectrum can be obtained if the decoupler is

turned on just prior to the pulse and off after acquisition. Overhauser enhancement

build up during the acquisition does not affect the signal being recorded since it only

contributes to the magnetisation along the magnetic field (z) axis, while the receiver

coil only picks up signals along one of the other orthogonal axes. This feature can be

avoided by making the interval between the end of acquisition and the next pulse at

least 3-5 times as long as the acquisition time. If the delay is ≥3-5 T1max, each carbon

will have returned completely to its equilibrium magnetisation. The spectral

integrations are reflective of carbon concentrations and useful for quantitative

analysis.96

72

Selective proton decouplingSpecific decoupling (single-frequency on-resonance) will result in a sharp

singlet for any carbon bonded directly with protons resonating at that frequency.

Complications can occur in cases with overlapping 13C-1H satellite spectra or when the

number of signals is too large for practicability.96

13C, 1H Chemical Shift Cross-CorrelationSelective decoupling is time-consuming. More informative and faster

techniques correlate individual carbons and protons. The operation includes all

protonated carbons at the same time.96

DeuterationDeuteration is a standard procedure in chemistry with special impact

particularly in 13C NMR spectrum. Spin number of deuterium is 1 and hence, it is not

affected by the proton decoupling, and produces a 13C-2H multiplet in the 13C

spectrum. The absence of a resonance indicates the deuterated carbon.96

Specific LabellingSpecific labelling can be done with either magnetically active or inactive nuclei.

Nucleus 15N gives 13C-15N splittings of resonances corresponding to carbons in

proximity to nitrogen. 96

Spin-lattice relaxation (T1)Experimentally, it is possible to determine spin-lattice relaxation time using a

two-pulse sequence. A sequence of spectra can be obtained in which a sequence of

spectra is obtained in which the peaks start out negative and systematically go

positive. There is a very simple equation for calculating relaxation time in small and

medium molecules (200<MW<10000)96:

Ha

C

CH

Hdipolarobs

rN

TT 116

11

11

NH is the number of protons, rCH distance between the carbon and those NH protons, C

is the rotational correlation time for reorientation of the C-H vector

A multinuclear pulse sequence can be used for generating 13C spectra

containing only CH3 and CH signals or just CH2 and quaternary carbon signals. This

sequence enables us to cancel unwanted resonances. Spectral editing of this type

was performed using INEPT sequence.101 However, the main advantages of DEPT

(distortionless excitation by polarization transfer) sequence over the INEPT sequence

73

is in more exact cancelation of residual carbon signals. Single-bond CH coupling

varies somehow, so the average J value must be used in setting the free precession

periods in both DEPT and INEPT sequence. Residual carbon signals appear as a

consequence of divergence from the above mentioned averaged J value.101

NMR spectrometers in use have possibility to distinguish between positive and

negative frequencies along the acquisition axis when 2D experiments are recorded. A

series of free induction decay with the acquisition time, t2, is collected while

incrementing the second time interval, t1, in the range from minimum to maximum

value. Two frequencies axis, f1 and f2, can be obtained using double Fourier

transformation. Distinguishing between the positive and negative frequency signals

can be achieved using phase cycling. The phase cycling can also enable coherence

pathway selection and artefact suppression. As a result, the mixture of absorption and

dispersion phase can be obtained, so it is impossible to get only 2D spectrum with

absorption peaks. Spectra can be represented in absolute or magnitude mode. Peaks

are plotted and calculated as (µ2+2)1/2 where µ is dispersion mode signal and -

absorption mode signal. Although they are positive, they do not show Lorenzian line

shape. Two different methods were invented which can be used for getting Lorenzian

line shape-hypercomplex method (STATES) and time proportional phase increment

(TPPI). In STATES, two data sets are collected, differing in that the phasing of one of

the pulses is changed by 900 between the two data sets. However, in TPPI, only one

data set is collected, but the phase of one of the pulses is incremented by 900 for each

new value of t1, with t1 varying two times faster than in the case of STATES. The

results obtained by these two different methods vary negligibly. Instead of phase

cycling, it can be used Z-axis gradient. The main advantages are that:

1) there is no use of phase cycling

2) there is time saving, particularly in case of homonuclear 2D sequences

3) recycle time for an experiment can be reduced without generating artifacts.102

Correlation spectroscopy (COSY) is a 2D, widely-used NMR method. This

technique was invented by Aue et al. in 1976 in order to solve the structure of

complicated molecules, but limited to weak coupled systems. There are three phases

in 2D experiments ranked according to their significance:

1) t<0, preparation period

2) 0<t<t1, evolution period

3) t1<t, detection period.

Two different peak types can be distinguished:

1) cross peaks (“off-diagonal peaks” in 2D spectrum)

2) dia peaks (they are on the main diagonal in 2D spectrum).

Both peak types give the following information:

74

1) information about the connectivity of transitions in the energy level diagram

2) information on transverse relaxation processes

3) information on initial state of spin system.103

Improvements of this method are very popular. Pulsed gradient technology

can be used to obtain COSY spectra with suppression of axial peaks and without the

need for phase cycling. This method provides an elegant solution to reduce strong

solvent signals.104

Self-focusing soft pulses are applied simultaneously on two coupled spins, I

and S (TSETSE). This technique enables easier determination of coupling

constants.105

In multiple quantum filtered spectra, it is possible to perform elucidation of

NMR coupling networks. Coherence transfer between evolution and detection periods

proceeds through a certain order of multiple quantum coherence (MQC) gives, as a

consequence, elimination of undesired features focusing on relevant peaks. Multiple

quantum filtering can be achieved by the sequence 900()-t1-900 ( + )-900()-

acquisition. Multiple quantum filters can eliminate singlets, e.g. peaks of many

common solvents. Transfer through MQC makes diagonal peak multiplets of

alternating sign such that incomplete resolution reduces them to the same extent as

cross-peak multiplets. The appearance or absence of peaks in a 2-D correlation

spectrum with multiple quantum filtering is governed by the selection rules:

1) The appearance of a diagonal peak in a p-quantum-filtered

spectrum implies that the spin in question has resolved couplings to

p-1 equivalent or non-equivalent spins.

2) The appearance of a cross-peak between two spins in a p-quantum

filtered spectrum indicates that, in addition to their direct coupling,

there are nonvanishing couplings to a common set of at least p-2

additional spins. 106

2D HSQC (Heteronuclear Single-Quantum Correlation) experiment was

introduced by Bodenhausen and Ruben in 1980. It enables us to obtain a 2D

heteronuclear chemical shift correlation map between directly-bonded 1H and X-

heteronuclei (usually they are 13C and 15N). Magnetisation of a proton is detected

during t2-detection time while the low-gamma nuclei evolve during the evolution time-

t1. Sensitivity enhancement (“OverBodenhausen Effect”) is much greater than

enhancement obtainable by simple NOE (Nuclear Overhauser Effect). It is quite

interesting that both HETCOR and HSQC give 2D spectrum which has 13C chemical

shifts on one axis and 1H chemical shifts on the other axis. In HSQC spectrum, 13C

chemical shifts are on the indirect (F1) axis while in the HETCOR spectrum 13C

chemical shifts are on the directly-detected (F2) dimension. There are several

75

differences evident in comparing the HETCOR and HSQC experiment. The

crosspeaks appear in pairs in HSQC and they are separated in F2 (horizontal)

dimension by the large one-bond CH coupling (~150 Hz). This CH coupling can be

eliminated by turning on a 13C decoupler during the acquisition of the FID. 13C Range

of chemical shifts is large (~200 ppm), so 13C decoupler should be put quite a bit of

power and this limits the acquisition time to a maximum of about 220 ms (0.22 s).

Longer relaxation delay (D1) of 1.5 s gives the opportunity for the sample to relax

before the acquisition of the next FID.107-109

2D HMBC (Heteronuclear Multiple Bond Correlation) detects long range

coupling between proton and carbon (two or three bonds away) with great sensitivity.

In order to detect relatively large coupling constants (4-10 Hz), τ delay should be

adjusted to τ=0.06 s or smaller couplings (2-7 Hz) to τ=0.1 s. The technique proved

very useful for the detection of indirectly quaternary carbons to protons - especially

useful if direct carbon-13 is impossible to obtain due to the low amount of available

material.110

Heteronuclear single-quantum coherence (HSQC) sequence107 is used less

frequently (almost four times) than heteronuclear multiple-quantum coherence

(HMQC) sequence,111 although the first technique shows a number of advantages

regard to the second.112

A variation of HSQC method has been made; one which is very applicable and

very useful was suggested by Saka (E-HSQC). According to this method, it is possible

to edit NMR signals (13C/15NH, 13C/15NH2, 13C/15NH3, 13C/15ND, 13C/15ND2, 13C/15ND3)

into subsignals.113 The gradient-enhanced HMQC technique (use of complex linear

back-prediction in order to generate the FIDs at the t1=0) shows increased sensitivity

compare to field-gradient HMQC technique.114

NOESY NMR technique is developed in order to estimate the distance

between the two protons close in space. In regular NOESY experiment, good data are

obtained using relatively long mixing times (about 200 ms) where the water

magnetization was driven significantly toward equilibrium by radiation dumping.115 The

disappearance of HOHAHA distortions is more rapid for large molecules than smaller.

Combination of theory and numerical simulations provide the quantitative explanations

for it. In the rotating frame, the maximum transient NOE increases from 38.5% for

ω0τc<<1 to 67.5% for ω0τc>>1. Consequences arise that 1) the NOE is positive and

does not disappear for any value of ω0τc; and 2) multispin effects will be minor.

Variants of the same method were developed. One provides information about

cross-relaxation rates between pairs of spins. In the same method, migration of

magnetisation through several consecutive steps (spin diffusion via neighbouring

76

spins) is largely suppressed. All these advantages were achieved by inserting a

doubly-selective inversion pulse in a conventional NOESY sequence.116

Instead of NOESY, ROESY experiments are applicable to some molecules.

One modification of ROESY experiment is developed where saturation of the water

magnetisation is avoided without compromising suppression of the water signal during

acquisition. The precise control of the water magnetisation throughout the experiment

was achieved using field gradient and selective RF pulses. Besides, it was avoided

signal losses due to radiation damping and molecular diffusion effects.117

In flip-back ROESY experiment, the time required for efficient radiation

damping is minimized by phase shifting the excitation pulse by 450, so that following a

second pulse the water is never tilted more than 1350 with respect to the z-axis.117

2.2.2. Mass spectrometryThe term “mass spectrometry” and not “mass spectroscopy” originates from the

historical development of this instrumentation. The device employed by Thompson,

and in general all first mass-separating experiments, was a type of spectroscope

displaying blurred signals on a fluorescent screen. It was later followed by Dempster,

who first constructed an instrument with a deflecting magnetic field and an angle of

1800. A mass spectrometer should be equipped with a photographic plate (mass

spectrograph) in order to detect different masses.118

The scheme of the mass spectrometer is represented in Figure 15.

Figure 15: The scheme of the mass spectrometer (image taken from

http://atcinc.net/helium-leak-detectors.asp)

77

Mass spectrometry has an application in:

1) Accurate molecular weight measurements

2) Reaction monitoring

3) Amino acid sequencing

4) Protein structure.

Ionisation methods include:

Atmospheric pressure chemical ionisation (APCI)

Chemical ionisation (CI)

Electron impact (EI)

Electrospray ionisation (ESI)

Fast atom bombardment (FAB)

Field desorption/Field ionisation (FD/FI)

Matrix-assisted laser desorption ionisation (MALDI)

Thermospray ionisation (TSP).119

The usual way of ionisation in mass spectrometry is chemical (CI) or electro-

ionisation (EI).

Regarding chemical ionisation, the initial step is the same as in EI. In CI, the

gas pressure in the ion source is usually increased to 10-3mbar (sometimes even to

atmospheric pressure) when the gas is injected. Chemical ionisation spectra are much

simpler; in most cases, having only protonated molecular ion peaks. Abundant [M-H]-

or [M+X]- can be obtained using negative reagent gases.

Under EI conditions, some compounds fragment so readily that often the

molecular ion does not survive. Consequently, no information about the molecular

mass is available, or is uncertain. Fragmentation during CI occurs very little, so

depending on reagent gas, ions [M+X]+ (X=H, NH4, NO, etc.) or [M+H]+ or [M-H]- or

[M+X]- (X=F, Cl, OH, O, etc.) are the most abundant. It is not surprising that under EI

conditions, information about structure can be obtained, which is not available under

CI conditions. Chemical ionisation conditions can be used to obtain information about

molecular mass. Usually, both CI and EI are used in order to get full information about

the molecule. They are used frequently in techniques such as gas

chromatography/mass spectrometry (GC/MS).

By analogy with the formation of protonated ions, negative ions can be formed

by deprotonation. Strongly basic ions such as OH- or O.- (reagent gas ions) are

capable of abstracting a proton from molecules M giving [M-H]- ions, one mass unit

less than the true molecular mass. It was proven that negative-ion CI is a useful

sensitive technique for substances having a high electron affinity (e.g. halo

compounds and polycyclic aromatic hydrocarbons).

78

Kinetic energy of the electron is gained through the acceleration in the electric

field. For example, after acceleration through 70V, the electron has energy of 70eV.

These electrons in contact with unchanged, neutral molecule, eject an electron from

the compound e - + M M.+ + 2e-

Two electrons exit the reaction zone, leaving a positively-charged species (M.+)

called an ion (it is molecular ion in this case). Strictly speaking, it is a radical-cation. In

the electric field with strength of 70eV with high vacuum, the interaction between

electrons and molecules leaves some ions (M.+) with so much extra energy that they

break up to give ions of smaller mass (A+, B+, etc.). The fragmentation e- + M M.+ +

2e- + A+, B+, etc. is characteristic of the given compound and is similar to a fingerprint

(mass spectrum). For the limited range of substances, negative radical anions (M.-)

can be formed more readily than positive ions.

The EI mass spectrum can be described as a chart on two axis relating the

mass of an ion (m, or, strictly, m/z) to its abundance. During the initial steps, ionisation

and fragmentation in the ion source (place where electron/molecule reaction occurs),

different numbers of ions (M+, A+, B+, etc) are produced and measured using the mass

analyser. A mass spectrum records mass (or m/z) on the x-axis, and corresponding

ion abundance is called the base peak, which, in some cases, corresponds with the

molecular ion or any of the fragment ions.

For FAB methods, an inert gas is usually used because it does not produce

unwanted secondary species in the primary beam, and avoids contaminating the gun

and mass spectrometer. The most commonly used inert gases are helium, argon and

xenon, but the higher mass atoms are preferred for the maximum yield of secondary

ions. A primary ion-beam gun can be used instead of the fast-atom beam. In general,

the ion gun emits a stream of cesium ions (Cs+), which is cheaper for use than xenon

and still have large atom mass (atomic masses: Cs, 139; Xe, 131). Ion guns do not

produce fragment ions in the primary beam, but can contaminate the mass

spectrometer by deposition if used continually.

Because of existing of high vacuum conditions, it is impossible to use samples

dissolved in common solvents used in chemistry (water, ethanol, chloroform, etc.)

because these solvents evaporate very quickly upon introduction to the ion source.

Therefore, it is necessary to use a high-boiling liquid as solvent (matrix) (Table 1).

79

Table 1. Some commonly-used solvents for FAB or LSIMS

Solvent Protonated molecular (m/z) ions

Glycerol 93

Thioglycerol 109

3-NOBA 154

NOP 252

Triethanolamine 150

Diethanolamine 106

Polyethylene glycol (mixtures) Wide mass range which depends on the glycol used

FAB and LSIMS will give excellent molecular mass information in the range

(approximately) of m/z 100-2000.120

2.2.3. Microbiological techniques for the determination of drug activityIn vitro tests of varying complexity are in use: agar diffusion sensitivity tests,

determination of minimum inhibitory concentrations (MIC) and minimum bactericidal

concentration (MBC). 121

MIC can be defined as the lowest concentration of antimicrobial that can

visually inhibit the growth of the microorganism after overnight incubation in most

cases. In contrast, MBC can be defined as the lowest concentration that can prevent

the growth of the organism after subculture on to antibiotic-free media. Mainly, MIC

information can be used to confirm the resistance, but it is also useful for the

determination of in vitro activity of new antimicrobials (“gold standard”).121 MIC wild-

type distribution databases are available for relevant species – drug combinations

(http://www.eucast.org). The highest MIC of the wild-type population is defined as the

“epidemiological cut-off value” or wild-type (WT) cut-off value.122 MBC determinations

are taken less frequently, and their major use has been reserved for isolates taken

from the blood of patients with endocarditis.121

Disk diffusion tests were introduced for the reason that they are less difficult

technically when large number of organisms are tested against many antibiotics.123 A

build-up of bacterial growth is sometimes seen at the edge of inhibition zone. There

are several factors which do not consistently affect the presence or absence of build-

up zones around antibiotic discs: 1) type of antibiotic; 2) bacteriostatic or bactericidal

action; 3) potency of discs; 4) sensitivity or resistance of organisms; 5) method of

inoculation (flooding, spreading); 6) pasteurisation of inoculums; 7) length of pre-

diffusion of antibiotic before incubation (plates upright or inverted, size of stack,

position of plates in stack, plates in canister or free standing); 10) antibiotic as a

solution in wells.124

80

2.2.3.1. Minimum inhibitory concentration (MIC) techniqueOn the basis of used volume, the MIC method can be termed as macrodilution

(total volume 2 ml) or microdilution (used volume is ≤500 μl per well).121

According to the definition of determination of MIC, a population of N bacteria

exposed in vitro to a concentration c of an antibiotic stops growing when c is equal or

higher than the MIC. When concentration c drops below MIC, the population grows

with a rate R0, where R0 is now the fraction of the population that enters the cell

division phase per unit time.

As soon as the antibiotic concentration exceeds the MBC, the bacteria

population decays. The part of bacteria population that dies per unit time is equal to

the fraction of that population that would cell divide in the absence of the cell-wall

antibiotic:

R[c>MIC] = -R[c<MIC] = -R0

R[c<MIC] = growth rate kill rate = growth rate

-R[c>MIC] = kill rate R[c] = R0 – (1+H[c])R0

Where H[c] = -1 for c< MIC

0 for MIC≤c<MBC

1 for c≥MBC

The presented mathematical model is based on the following assumptions:

1) Bacteria population dynamics is given by a non-linear differential equation;

2) The bacteria population reproduces or dies with a rate R[c,t] for which an

empirical form is chosen. R[c,t] depends on two parameters ER[c] and

tmax[c].

3) For the parameters ER[c] and tmax[c] empirical Hill expressions are

chosen.125

Range of antibiotic concentrations should be in doubling dilution steps up and

down from 1mg/L as required.121

Considering the inoculums, it should be adjusted that 104 cfu/spot are applied

to the plates (comparing to 0.5 McFarland standard or adjustment photometrically).

The following procedure can be used in order to prepare the 0.5 McFarland standard.

Preparation of the McFarland standard

It should be added 0.5 ml of 0.048 M BaCl2 (1.17% w/v BaCl22 H2O) to 99.5 ml

of 0.18 M H2SO4 (1% v/v) with constant stirring. This should be followed by distribution

of the standard into screw cap tubes of the same size and with the same volume as

those used for growing broth cultures. Tubes should be sealed tightly in order to

prevent loss by evaporation. It should be stored protected from sunlight at room

81

temperature. The prepared standard can be kept for up to six months. Alternatively,

the McFarland standard can be bought from a supplier (e.g. bioMerieux, Basingstoke,

UK).

Inoculum should be used within 30 min of preparation.121

Table 2. Appropriate media for different organisms 121

Organism Medium to be used

Enterobacteriaceae ISA

Pseudomonas spp. ISA

Enterococci ISA

Streptococcus pneumonia ISA + 5% defibrinated horse blood

-Haemolytic streptococci ISA + 5% defibrinated horse blood

Moraxella catarrhalis ISA + 5% defibrinated horse blood

Haemophilus spp.ISA + 5% defibrinated horse blood + 20

mg/L NAD

Neisseria meningitides ISA + 5% defibrinated horse blood

Neisseria gonorrhoeae ISA + 5% defibrinated horse blood

Staphylococci ISA

AnaerobesWilkins & Chalgren agar + 5%

defibrinated horse blood

Determination of MIC should be conducted according to regulations proposed

by several organisations in different countries worldwide (France, Germany, Norway,

Sweden, UK, USA):

Comite l’Antibiogramme de la Societe Francaise de Microbiologie.

Technical recommendations for in vitro susceptibility testing. Clin.

Microbiol.Infect.1996; 2 (Suppl.1):S11-25.

Deutsches Institut fur Normung. Methods for the determination of

susceptibility of pathogens (except mycobacteria) to antimicrobial

agents. MIC breakpoints of antimicrobial agents. Berlin:DIN,

1998;Suppl 1:58940-4.

Norwegian Working Group on Antibiotics. Susceptibility testing of

bacteria and fungi. Scand. J. Infect. Dis. 1997;103 (Suppl):1-36.

Olsson-Liljequist B, Larsson P, Walder M, Miorner H. Antimicrobial

susceptibility testing in Sweden III. Methodology for susceptibility

testing. Scand. J. Infect. Dis. 1997;105 (Suppl):13-23.

82

BSAC Working Party. A guide to sensitivity testing. J.

Antimicrob.Chemother. 1991;27 (Suppl. D): 1-50.

National Committee for Clinical laboratory Standards. Methods for

dilution antimicrobial susceptibility tests for bacteria that grow

aerobically, 5th edn. Approved Standard M7-A5. Wayne, PA:NCCLS,

2000.126

A clear choice of reference medium depends on researchers. Mueller-Hinton

(MH) agar does not show any principal advantages compared to others, but it is the

most widely-used media in the world. Probably, the reason is the availability of USA

National Committee for clinical laboratory standards (NCCLS) document, which

provides procedures for evaluating MH agar.

Supplements should not be used unless they are necessary for the growth of

the organism. For organisms such as Streptococci and Moraxella catarrhalis, it is

recommended to use 5% defibrinated blood. In case of Haemophilus spp. and

Neisseria spp. organisms, several supplements are recommended (5% defibrinated

blood with 20mg/L NAD, 5% chocolated blood and supplemented GC agar).

Sterilisation of solutions is not usually necessary. However, if it is necessary, it

should be performed using membrane filtration, and the sample should be compared

before and after sterilisation to ensure that no adsorption occurred. Storing of stock

solutions is usually performed by freezing aliquots at -200C or below.127

2.3. References

1 N. Metropolis, A.W. Rosenbluth, M. N. Rosenbluth, A. H. Teller, and E. Teller:Equation of state calculations by fast computing machines. J. Chem. Phys., 2, 1087,1953.2 B. J. Alder and T. E. Wainwright: Phase transition for a hard sphere system. J.Chem. Phys., 27, 1208, 1957.3 (a) R. Boese, T. Haumann and P. Stellberg: In Advances in Molecular StructureResearch M. Hargittai and I. Hargittai (Eds.), Vol. 1, JAI Press, London, pp. 202–226,1995. (b) W. J. Hehre, L. Radom, P. V. R. Schleyer, J. A. Pople: Ab initio MolecularOrbital Theory, Wiley, New York, 1986 and references cited within;4 (a) R. J. Gillespie and I. Hargittai: The VSEPR Model of Molecular Geometry, Allynand Bacon, Boston, 1991 and references cited therein. (b) R. Gillespie: The VSEPRModel Revisited, J. Chem. Soc. Rev., 59, 1992, and references cited therein;5 (a) P. A. Kollman: Advances and continuing challenges in achieving realistic andpredictive simulations of the properties of organic and biological molecules. Acc.Chem. Res., 29, 461, 1996, (b) N. L. Allinger: Molecular Mechanics in AccurateMolecular Structures, A. Domenicano, I. Hargittai. Eds.; Oxford University, New York,1992; and references cited therein. (c) K. Rasmussen: AIP Conf. Proc. (E.C.C.C. 1Computational Chemistry) 330, 269, 1995, (d) C. Jaime: Trends Org. Chem. 4, 469,1993.6 M. J. S. Dewar, C. Jie and J. Yu: SAM1; The first of a new series of general purposequantum mechanical molecular models, Tetrahedron., 49, 5003, 1993, and referencescited therein.

83

7 E. R. Davidson: Molecular Mechanics and Modeling: Overview, Chem. Rev., 93, 7,1993.8 B. J. Alder and T. E. Wainwright: Studies in Molecular Dynamics. I. General Method,J. Chem. Phys. 31, 459, 1959.9 MM2 (77), Quantum Chemistry Program Exchange, Bloomington, IN47405, Program395.10 MM3, Technical Utilization Corporation, Incorporated, 235 Glen Village Court,Powell, OH 4306511 N. L. Allinger, M. R. Imam, M. R. Frierson, Y. H. Yuh, L. Schafer: In mathematicsand computational concepts in chemistry; N. Trinajstic, Ed., E. Horwood, Ltd.: London,1986.12 H. D. Thomas, K. Chen, N. L. Allinger: Toward a better understanding of covalentbonds: the molecular mechanics calculation of C-H bond lengths and stretchingfrequencies, J. Am. Chem. Soc., 116, 5887, 1994.13 J. D. Durrant, J. A. McCammon: Molecular dynamics simulations and drugdiscovery, BMC Biol., 9, 71, 2011.14 A. K. Rappe, C. J. Casewit: Molecular mechanics across chemistry, Universityscience books, Sausalito, 1997.15 E. M. Engler, J. D. Andose and P. von R. Schleyer: Critical evaluation of molecularmechanics, J. Am. Chem. Soc., 95, 1973.16 K. Vanommeslaeghe, E. Hatcher, C. Acharya, S. Kundu, S. Zhong, J. Shim, E.Darian, O. Guvench, P. Lopes, I. Vorobyov, A. D. Mackerell Jr.: CHARMM generalforce field: A force field for drug-like molecules compatible with the CHARMM all-atomadditive biological force fields, J. Comput. Chem., 31, 671, 2010.17 B. R. Brooks, C. L. Brooks 3rd, A. D. Mackerel Jr., L. Nilsson, R. J. Petrella, B. Roux,Y. Won, G. Archontis, C. Bartels, S. Boresch, A. Caflisch, L. Caves, Q. Cui, A. R.Dinner, M. Feig, S. Fischer, J. Gao, M. Hodoscek, W. Im, K. Kuczera, T. Lazaridis, J.Ma, V. Ovchinnikov, E. Paci, R. W. Pastor, C. B. Post, J. Z. Pu, M. Schaefer, B. Tidor,R. M. Venable, H. L. Woodcock, X. Wu, W. Yang, D. M. York, M. Karplus: CHARMM:the biomolecular simulation program, J. Comput. Chem., 30, 1545, 2009.18 W. D. Cornell, P. Cieplak, C. I. Bayly, I. R. Gould, K. M. Merz, Jr., D. M. Ferguson,D. C. Spellmeyer, T. Fox, J. W. Caldwell, P. A. Kollman: A second generation forcefield for the simulation of proteins, nucleic acids, and organic molecules, J. Am. Chem.Soc., 117, 5179, 1995.19 N. Schmid, A. P. Eichenberger, A. Choutko, S. Riniker, M. Winger, A. E. Mark, W. F.Van Gunsteren: Definition and testing of the GROMOS force-field versions 54A7 and54B7, Eur. Biophys. J., 40, 843, 2011.20 W. L. Jorgensen, J. Tirado-Rives: The OPLS potential functions for proteins. Energyminimizations for crystals of cyclic peptides and cramblin, J. Am. Chem. Soc., 110,1657, 1988.21 P. K. Weiner and P. A. Kollman: AMBER: Assisted model building with energyrefinement. A general program for modeling molecules and their Interactions, J.Comput. Chem., 2, 287, 1981.22 J. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman, D. A. Case: Development andtesting of a general AMBER force field, J. Comput. Chem., 25, 1157, 2004.23 D. A. Pearlman, D. A. Case, J. W. Caldwell, W. S. Ross, T. E. Cheatham II, S.DeBolt, D. Ferguson, G. Seibel and P. Kollman: AMBER, a package of computerprograms for applying molecular mechanics, normal mode analysis, moleculardynamics and free energy calculations to simulate the structural and energeticproperties of molecules, Comput. Phys. Commun., 91, 1, 1995.24 J-H. Li, N. L. Allinger: Molecular mechanics. The MM3 force field for hydrocarbons.3. The van der Waals’ potentials and crystal data for aliphatic and aromatichydrocarbons, J. Am. Chem. Soc. 111, 8576, 1989.25 T. A. Halgren: Merck molecular force field. I. Basis, form, scope, parameterization,and performance of MMFF94*, J. Comput. Chem., 17, 490, 1996.26 W. L. Jorgensen and C. Ravimohan: Monte Carlo simulation of differences in freeenergies of hydration. J. Chem. Phys., 83, 3050, 1985.

84

27 W. L. Jorgensen and J. Tirado-Rives: Potential energy functions for atomic-levelsimulations of water and organic and biomolecular systems, PNAS, 102, 6665, 2005.28 V. Daggett, M. Levitt, Annu. Rev. Biophys. Biomol. Struct., 22, 353, 1993.29 J. M. Deutsch and T. Kurosky: Design of force fields from data at finite temperature,Phys. Rev. E: Stat., Nonlinear, Soft Matter Phys., 56, 1997.30 P. Comba and R. Remenyi: Inorganic and bioinorganic molecular mechanicsmodelling-the problem of the force field parametrization, Coord. Chem. Rev., 9, 238,2003.31 D. L. Mobley, E. Dumont, J. D. Chodera, K. A. Dill: Comparison of charge models forfixed-charge force fields: small-molecule hydration free energies in explicit solvent, J.Phys. Chem. B, 111, 2242-2254, 2007.32 B. G. Johnson, P. M. W. Gill and J. A. Pople: The performance of a family of densityfunctional methods, J. Chem. Phys. 98, 5613, 1993.33 D. S. Cerutti, P. L. Freddolino, R. E. Duke Jr. and D. A. Case: Simulations of aprotein crystal with a high resolution X-ray structure: Evaluation of force fields andwater models, J. Phys. Chem. B, 114, 12811, 2010.34 http://www.chem.ox.ac.uk/course/cache/mechanics.html35 B. A. Berg and T. Neuhaus: Multicanonical algorithms for first order phasetransitions, Phys. Lett. B267, 249, 1991.36 Y. Sugita and Y. Okamoto: Replica-exchange molecular dynamics method forprotein folding, Chem. Phys. Lett., 314, 141, 1999.37 K. Hukushima and K. Nemoto: Exchange Monte Carlo method and application tospin glass simulations, J. Phys. Soc. Jpn. 65, 1604, 1996.38 R. E. Caflisch: Monte Carlo and quasi-Monte Carlo methods, Acta Numer., 1, 1998.39 http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.35.401040 J. C. A. Boeyens and P. Comba: Molecular mechanics: theoretical basis, rules,scope and limits, Coord. Chem. Rev., 212, 3, 2001.41 D. Qiu, P. S. Shenkin, F. P. Hollinger and W. C. Still: The GB/SA continuum modelfor solvation. A fast analytical method for the calculation of approximate Born radii, J.Phys. Chem. A, 101, 3005, 1997.42 L. Y. Zheng, E. Gallicchio, R. A. Friesner and R. M. Levy: Solvent models forprotein-ligand binding: comparison of implicit solvent Poisson and surface generalizedBorn models with explicit solvent simulations, J. Comput. Chem., 22, 591, 2001.43 R. Durbin, S. Eddy, A. Krogh and G. Mitchison: Biological sequence analysis,probabilistic models of proteins and nucleic acids, Cambridge University Press,Cambridge, United Kingdom, 1998.44 D. A. Benson, I. Karsch-Mizrachi, D. J. Lipman, J. Ostell and E. W. Sayers:GenBank, Nucleic Acids Res., 39, D32, 2011.45 H. M. Berman: The protein data bank: a historical perspective, Acta Cryst., A64, 88,2008.46 E. Gasteiger, A. Gattiker, C. Hoogland, I. Ivanyi, R. D. Appel and A. Bairoch:ExPASy: the proteomics server for in-depth protein knowledge and analysis, NucleicAcids Res., 31, 3784, 2003.47 G. Stoesser, P. Sterk, M. A. Tuli, P. J. Stohr and G. N. Cameron: The EMBLnucleotide sequence database, Nucleic Acids Res., 25, 7, 1997.48 S. Hunter, R. Apweiler, T. K. Attwood, A. Bairoch, A. Bateman, SD. Binns, P. Bork,U. Das, L. Daugherty, L. Duquenne, R. D. Finn, J. Gough, D. Haft, N. Hulo, D. kahn, E.kelly, A. Laugraud, I. Letunic, D. Lonsdale, R. Lopez, M. Madera, J. Maslen, C.McAnulla, J. McDowall, J. Mistry, A. Mitchell, N. Mulder, D. Natale, C. Orengo, A. F.Quinn, J. D. Selengut, C. J. A. Sigrist, M. Thimma, P. D. Thomas, F. Valentin, D.Wilson, C. H. Wu and C. Yeats: InterPro: the integrative protein signature database,Nucleic Acids Res., 37, D211, 2009.49 The UniProt Consortium: The Universal Protein Resource (UniProt), Nucleic AcidsRes., 36, D190, 2008.50 I. Letunic, R. R. Copley, S. Schmidt, F. D. Ciccarelli, T. Doerks, J. Schultz, C. P.Ponting and P. Bork: Smart 4.0: towards genomic data integration, Nucleic Acids Res.,32, D142, 2004.

85

51 R. D. Finn, J. Tate, J. Mistry, P. C. Coggill, S. J. Sammut, H-R. Hotz, G. Ceric, K.Forslund, S. R. Eddy, E. L.L. Sonnhammer and A. Bateman: The Pfam protein familiesdatabase, Nucleic Acids Res., 36, D281, 2007.52 T. K. Attwood, P. Bradley, D. R. Flower, A. Gaulton, N. Maudling, A. Mitchell, G.Moulton, A. Nordle, K. Paine, P. Taylor, A. Uddin and C. Zygouri: PRINTS and itsautomatic supplement, prePRINTS, Nucleic Acids Res., 31, 400, 2003.53 S. F. Altschul, W. Gish, W. Miller, E. W. Myers and D. J. Lipman: Basic localalignment search tool, J. Mol. Biol. 215, 403, 1990.54 D. J. Lipman and W. R. Pearson: Rapid and sensitive protein similarity searches,Science, 227, 1435, 1985.55 S. F. Altschul, T. L. Madden, A. A. Schaffer, J. Zhang, Z. Zhang and D. J. Lipman:Gapped BLAST and PSI-BLAST: a new generation of protein database searchprograms, Nucleic Acids Res., 25, 3389, 1997.56 R. D. Finn, J. Clements and S. R. Eddy: HMMER web server: interactive sequencesimilarity searching, Nucleic Acids Res., 39, W29, 2011.57 R. Chenna, H. Sugawara, T. Koike, R. Lopez, T. J. Gibson, D. G. Higgins and J. D.Thompson: Multiple sequence alignment with the Clustal series of programs, NucleicAcids Res., 31, 3497, 2003.58 A. Krogh, B. Larsson, G. Von Heijne and E. L. L. Sonnhammer: Predictingtransmembrane protein topology with a Hidden Markov model: application to completegenomes, J. Mol. Biol., 305, 567, 2001.59 C. Cole and J. D. Barber: The Jpred 3 secondary structure prediction server, NucleicAcids Res., 36, W197, 2008.60 D. T. Jones: Protein secondary structure prediction based on position-specificscoring matrices, J. Mol. Biol., 292, 195, 1999.61 http://swift.cmbi.ru.nl/teach/HOMMOD/seminars/biobook.pdf62 http://www.dsto.defence.gov.au/publications/2271/DSTO-GD-0253.pdf63 J. Cheng, A. Z. Randall, M. J. Sweredoski and P. Baldi: SCRATCH: a proteinstructure and structural feature prediction server, Nucleic acids research, 33, W72,2005.64 A. Fiser, R. K. G. Do and A. Sali: Modelling of loopes in protein structures, ProteinSci., 9, 1753, 2000.65 R. Das and D. Baker: Macromolecular modelling with Rosetta, Annu. Rev. Biochem.,77, 363, 2008.66 Y. Zhang: I-TASSER server for protein 3D structure prediction. BMC Bioinformatics,9, 40, 2008.67 R. A. Laskowski, M. W. MacCarthur, D. S. Moss and J. M. Thornton: PROCHECK: aprogram to check the stereochemical quality of protein structures, J. Appl. Cryst., 26,283, 1993.68 G. Vriend: WHAT IF: A molecular modeling and drug design program, J. Mol.Graphics, 8, 52, 1990.69 M. Wiederstein and M. J. Sippl: ProSA-web: interactive web service for therecognition of errors in three-dimensional structures of proteins, Nucleic Acids Res.,35, W407, 2007.70 D. Eisenberg, R. Luthy and J. U. Bowie: VERIFY3D: Assessment of protein modelswith three-dimensional profiles, Methods Enzymol., 277, 396, 1997.71 M. Zuker, D. H. Mathews and D. H. Turner: Algorithms and thermodynamics forRNA secondary structure prediction: A practical guide, RNA Biochemistry andbiotechnology, Proceedings of the NATO advanced research workshop on RNAbiochemistry and biotechnology, Poznan, Poland, 10-17 October 1998.72 D. H. Mathews, J. Sabina, M. Zuker and D. H. Turner: Expended sequencedependence of thermodynamic parameters improves prediction of RNA secondarystructure, J. Mol. Biol., 288, 911, 1999.73 R. C. Edgar: MUSCLE: multiple sequence alignment with high accuracy and highthroughput, Nucleic Acids Res., 32, 1792, 2004.74 P. Di Tommaso, S. Moretti, I. Xenarios, M. Orobitg, A. Montanyola, J-M. Chang, J-F.Taly and C. Notredame: T-Coffee: a web server for the multiple sequence alignment of

86

protein and RNA sequences using structural information and homology extension,Nucleic Acids Res., 39, W13, 2011.75 E. Capriotti and M. A. Marti-Renom: Computational RNA structure prediction, Curr.Bioinform., 3, 2008.76 D. H. Mathews, J. Sabina, M. Zuker and D. H. Turner: Expended sequencedependence of thermodynamic parameters improves prediction of RNA secondarystructure, J. Mol. Biol., 288, 911, 1999.77 D. H. Mathews and D. H. Turner: Dynalign: An algorithm for finding the secondarystructure common to two RNA sequences, J. Mol. Biol., 317, 191, 2002.78 J. S. Reuter and D. H. Mathews: RNAstructure: software for RNA secondarystructure prediction and analysis, BMC Bioinformatics, 11, 129, 2010.79 Y. Ding, C. Y. Chan and C. E. Lawrence: Sfold web server for statistical folding andrational design of nucleic acids, Nucleic acids research, 32, W135, 2004.80 M. Zuker: Mfold web server for nucleic acid folding and hybridization prediction,Nucleic Acids Res., 31, 3406, 2003.81 I. L. Hofacker, W. Fontana, P. F. Stadler, I. S. Bonhoeffer, M. Tacker and P.Schuster: Fast folding and comparison of RNA secondary structures, Monatsh. Chem.,125, 167, 1994.82 I. L. Hofacker: Vienna RNA secondary structure server, Nucleic Acids Res., 31,3429, 2003.83 C. Massire and E. Westhof: MANIP: An interactive tool for modelling RNA, J. Mol.Graphics Modell., 16, 197, 1998.84 F. Jossinet and E. Westhof: Sequence to structure (S2S): display, manipulate andinterconnect RNA data from sequence to structure, Bioinformatics, 21, 3320, 2005.85 H. M. Martinez, J. V. Maizel, Jr and B. A. Shapiro: RNA2D3D: A program forgenerating, viewing, and comparing 3-dimensional models of RNA, Journal ofmolecular structure and dynamics, 25, 669, 2008.86 I. Halperin, B. Ma, H. Wolfson and R. Nussinov: Principles of docking: an overviewof search algorithms and a guide to scoring functions, Proteins: Struct., Funct., Genet.,47, 409, 2002.87 S. Kaur, H. Sadawarti and M. Kumar: Human protein function prediction-a dockingbased approach, 1, 95, 2011.88 R. T. Kroemer: Molecular modelling probes: docking and scoring, Biochem. Soc.Trans., 31, 980, 2003.89 C. R. Corbeil and N. Moiessier: Docking ligands into flexible and solvatedmacromolecules. 3. Impact of input ligand conformation, protein flexibility, and watermolecules on the accuracy of docking programs, J. Chem. Inf. Model., 49, 997, 2009.90 R. D. Taylor, P. J. Jewsbury and J. W. Essex: A review of protein-small moleculedocking methods, J. Comput.-Aided Mol. Des., 16, 151, 2002.91 Y. Fukunishi and H. Nakamura: Prediction of ligand-binding sites of proteins bymolecular docking calculation for a random ligand library, Protein Sci., 20, 95, 2011.92 A. E. Cho, V. Guallar, B. J. Berne and R. Friesner: Importance of accurate chargesin molecular docking: quantum mechanical/molecular mechanical (QM/MM) approach,J. Comput. Chem., 26, 915, 2005.93 P. D. Lyne: Structure-based virtual screening: an overview, DDT, 7, 1047, 2002.94 G. A. Gray: Carbon nuclear magnetic resonance spectroscopy, Anal. Chem., 47,546A, 1975.95 R. Freeman and H. D. W. Hill: High-resolution study of NMR spin echoes: “Jspectra”, J. Chem. Phys., 54, 301, 1971.96 U. Sequin and A. I. Scott: Carbon-13 as a label in biosynthetic studies, Science,186, 101, 1974.97 D. M. Grant and E. G. Paul: Carbon-13 magnetic resonance. II. Chemical shift datafor the alkanes, 86, 2984, 1964.98 L. P. Lindeman and J. A. Adams: Carbon-13 nuclear magnetic resonancespectrometry chemical shifts for the paraffins through C, Anal. Chem., 43, 1245,1971.

87

99 D. K. Dalling and D. M. Grant: Carbon-13 magnetic resonance. IX. Themethylcyclohexanes, J. Am. Chem. Soc., 89, 6612, 1967.100 G. C. Levy: On segmental motion in short aliphatic chains, J. Am. Chem. Soc., 95,6117, 1973.101 D. T. Pegg, D. M. Doddrell and M. R. Bendall: Correspondence between INEPTand DEPT pulse sequences for coupled spin-half nuclei, J. Magn. Reson., 51, 264,1983.102 G. S. Rule and T. K. Hitchens: Fundamentals of Protein NMR spectroscopy, 12.Quadrature detection in multi-dimensional NMR spectroscopy, 239-250, volume 5,Springer, 2006103 Y. Vasavi, N. Parthiban, D. S. Kumar, D. Banji, N. Srisutherson, S. Ghosh and M.V. K. Chakravarthy: Heteronuclear multiple bond correlation spectroscopy-anoverview, International journal of PharmTech Research, 3, 1410, 2011.104 W. F. Reynolds and R. G. Enriquez: Choosing the best pulse sequences,acquisition parameters, postacquisition processing strategies, and probes for naturalproduct structure elucidation by NMR spectroscopy, J. Nat. Prod., 65, 221, 2002.105 J.-M. Nuzillard and R. Freeman: Correlation spectroscopy with two simultaneoussoft pulses (TSETSE), J. Magn. Reson., 112, 72, 1995.106 U. Eliav and G. Navon: Quadrupole-echo techniques in multiple-quantum-filteredNMR spectroscopy of heterogeneous systems, J. Magn. Reson., Series A, 115, 241,1995.107 http://www.chem.queensu.ca/facilities/nmr/nmr/webcourse/hmqc.htm108 http://www.chem.arizona.edu/rss/nmr/manuals/hsqc.pdf109 http://www.nmr/sinica.edu.tu/Cours/Course20030929-30/Lecture_1_NMR_Methods.pdf110 http://www.chem.uky.edu/research/miller/gHMBC.pdf111 http://triton.iqfr.csic.es/guide/eNMR/eNMR2Dinv/hmqc2d.html112

http://web4.uwindsor.ca/units/biochem/nmr/main.nsf/SubCategoryFlyOut/841799DAF6DC7DB98525753D0071E029113 I. Saka: Multiplicity editing by 2D E-HSQC NMR spectroscopy: a product operatortheory study, Phys. Scr., 80, 1, 2009.114 J. H. Davis: Refocusing revisited: an optimized, gradient-enhanced refocusedHSQC and its applications in 2D and 3D NMR and in deuterium exchangeexperiments, J. Biomol. NMR, 5, 433, 1995.115 N. Juranic, Z. Zolnai and S. Macura: Two-dimensional exchange spectroscopyrevisited: accounting for the number of participating spins, J. Serb. Chem. Soc., 65,285, 2000.116 J. Hennig and H. H. Limbach: magnetization transfer in the rotating frame: A newsimple kinetic tool for the determination of rate constants in the slow chemicalexchange range, J. Magn. Reson., 49, 322, 1982.117 D. B. Fulton and F. Ni: ROESY with water flip back for high-field NMR ofbiomolecules, J. Magn. Reson., 129, 93, 1997.118 http://masspec.scripps.edu/mshistory/perspectives/cjudson.php119 http://www.research.uky.edu/ukmsf/jeolioniz.pdf120 http://www.prenhall.com/settle/chapters/ch31.pdf121 J. M. Andrews: Determination of minimum inhibitory concentrations, J. Antimicrob.Chemother., 48, 5, 2001.122 G. Kahlmeter, D. F. J. Brown, F. W. Goldstein, A. P. MacGowan, J. W. Mouton, A.Osterlund, A. Rodloff, M. Steinbakk, P. Urbaskova and A. Vatopoulos: Europeanharmonization of MIC breakpoints for antimicrobial susceptibility testing of bacteria, J.Antimicrob. Chemother., 52, 145, 2003.123 F. D. Schoenknecht, J. A. Washington II, T. L. Gavan and C. Thornsberry: Rapiddetermination of minimum inhibitory concentrations of antimicrobial agents by theAutobac method: A collaborative study, Antimicrob. Agents Chemother., 17, 824,1980.

88

124 M. J. Marshall and C. A. Field: The interpretation of antibiotic disc sensitivities, J.Antimicrob. Chemother., 1 (Suppl.), 13, 1975.

89

Aims

The overall aim of my research presented in the thesis is to model the binding

of macrolides to the apicoplast ribosome of P. falciparum in order to understand both

the drugs and the ribosome.

The first part of the project represents an investigation of the conformational

properties of drugs with anti-malarial activity.

The second part is an attempt to construct an apicoplast ribosome of P.

falciparum in silico.

90

II Results and Discussion

Chapter 3: Theoretical and experimental investigation on clarithromycin,erythromycin A and azithromycin and descladinosyl derivatives ofclarithromycin and azithromycin with 3-O-substitution as anti-bacterial agents

3.1. DECLARATION

This chapter consists of a piece of work in preparation for publication in

Organic and Biomolecular Chemistry - B. Arsic, A. Awan, R. J. Brennan, R. Ledder, A.

McBain, A. C. Regan, G. A. Morris, R. Bryce and J. Barber: Theoretical and

experimental investigation on clarithromycin, erythromycin A and azithromycin and

descladinosyl derivatives of clarithromycin and azithromycin with 3-O substitution as

anti-bacterial agents.

A. Awan prepared the macrolide antibiotics with bacterial ribosomes and did

TRNOESY experiments. I carried out molecular dynamics and molecular mechanics

study in collaboration with Dr Regan and Dr Bryce, microbiological studies in

collaboration with Dr McBain and Dr Ledder, synthesis of descladinosyl derivatives of

clarithromycin and azithromycin, NMR assignment of the synthesized compounds. I

prepared the article for publication along with Supplementary material. It has been

edited by Professor Morris, Dr Regan, Dr Brennan, Dr Bryce and Dr Barber.

CREATED USING THE RSC ARTICLE TEMPLATE (VER. 3.1) - SEE WWW.RSC.ORG/ELECTRONICFILES FOR DETAILS

ARTICLE TYPE www.rsc.org/xxxxxx | XXXXXXXX

This journal is © The Royal Society of Chemistry [year] Journal Name, [year], [vol], 00–00 | 1

Theoretical and experimental investigation on clarithromycin,erythromycin A and azithromycin and descladinosyl derivatives ofclarithromycin and azithromycin with 3-O substitution as anti-bacterialagents

Biljana Arsica, Abida Awana, Richard J. Brennana, Ruth Leddera, Andrew McBaina, Andrew C. Reganb,5

Gareth A. Morrisb, Richard Brycea and Jill Barbera*

Received (in XXX, XXX) Xth XXXXXXXXX 200X, Accepted Xth XXXXXXXXX 200XFirst published on the web Xth XXXXXXXXX 200XDOI: 10.1039/b000000x

Erythromycin A, clarithromycin and azithromycin are the three most important macrolide10

antibiotics and are all very widely used in the clinic. They act by inhibiting the bacterial ribosome,and the molecular basis of their antibacterial activity remains an area of interest. We haveperformed both NMR (transferred NOESY) and modelling studies in order to determine theconformations of these antibiotics when they are bound to ribosomes. All the drugs exhibit the“folded-out” conformation in the bound state, but the dominant conformation of azithromycin is15

not completely superimposable on the clarithromycin and erythromycin A 9-ketone structures.Modelling suggests that clarithromycin and its 3-O-substituted descladinosyl derivatives areconformationally rigid molecules; these are compounds which generally exhibit more activityagainst Gram-positive bacteria. Azithromycin and its 3-O-descladinosyl derivatives are flexible insilico and show more activity against Gram-negative bacteria.20

Introduction

The first macrolide antibiotic, erythromycin A, was isolatedin 1952 from a culture of Saccharopolyspora erythraea 1 andstructurally described in 1957 2,3. In aqueous solution, the 9-ketone (1a) equilibrates with the 12,9-hemiacetal (1b), which25

appears to have no antibacterial activity.4 The secondgeneration macrolides, clarithromycin and azithromycin aresemi-synthetic derivatives of erythromycin A, and showincreased acid stability and broader spectra of antibacterialactivity. In general, clarithromycin is more active against30

Gram-positive bacteria but azithromycin has Gram-negativeactivity, which clarithromycin does not share.5 In combination

with existing antimalarial drugs (artesunate and quinine),azithromycin also shows mild antimalarial activity.6

The macrolide antibiotics inhibit bacterial protein synthesis35

by binding reversibly to the 50S subunit of the bacterialribosome. They block the exit tunnel of the ribosome near thepeptidyl transferase centre, and thus inhibit the elongation ofthe peptide chain. 7

The three-dimensional structures of erythromycin A and its40

derivatives have been analyzed in three distinct situations: inthe free drug (both in solution8 and in the crystalline state9), inthe bound state (by X-ray crystallography of drug-ribosomecomplexes)10-13, and in an intermediate weakly-bound state,detected by NMR spectroscopy, of drug-ribosome mixtures.14-

45

16 Everett et al. (1990) described two conformational statesavailable to the macrolide antibiotics: the folded-inconformation in which H3 approaches H11 and H4approaches Me-18, and the folded-out conformation, in whichH4 approaches H11 and H5 approaches Me-18 (Table 1).17

50

Erythromycin 9-ketone, clarithromycin and azithromycin haveall been shown to adopt folded-out conformations, both in thefree state and when bound to ribosomes (see Fig. 1).

A BFig. 1: Crystal structures of A: erythromycin A 9 (folded-outconformation) and B: dirithromycin 18 (folded-inconformation)

2 | Journal Name, [year], [vol], 00–00 This journal is © The Royal Society of Chemistry [year]

The conformations adopted by these drugs in theintermediate weakly bound state have been the subject ofmuch investigation, but also some confusion. The transferredNOESY experiment is prone to artefacts (especially from spindiffusion). In addition, Everett et al 17 showed that close5

approach of H11 to either H4 or H3 was diagnostic of afolded-out or folded-in conformation respectively for 14-membered macrolides, providing a convenient short-cut to theconformational analysis of 14-membered macrolides by NMRspectroscopy. However, this relation does not necessarily10

hold for 15-membered macrolides, which retain sufficientflexibility to be able to assume a folded-out conformationeven when H3 approaches H11. Several other contacts,especially 4-18 and 5-18, are required to distinguish betweenfolded-out and folded-in conformers.15

Internuclear distances (Å)Contact Erythromycin A DirithromycinH4-H11 2.51 3.33H15-H16 3.28 3.81H3-H8 5.34 2.30H3-H11 3.48 2.46H4-H18 4.02 2.71H5-H18 2.66 3.33H16-H17 3.84 2.51

Table 1: Key internuclear distances in crystal structures of folded-out(erythromycin A) and folded-in (dirithromycin) macrolides

Neither NMR nor X-ray crystallographic studies have thepredictive potential of molecular modelling, and it would beuseful to be able to investigate the conformation of a20

macrolide without having to synthesize it. An attempt todevelop a systematic protocol for describing theconformations of macrolides was described in 2006, usingoleandomycin and some of its derivatives as examples.Unconstrained molecular dynamics suggested that the drug25

forms a mixture of folded-out and folded-in conformations inboth aqueous and chloroform solutions, and this conclusionwas supported by 2D ROESY NMR spectroscopy.14 Othermolecular modelling approaches have generally dependedupon NMR results and were not designed to be used30

independently.15-16

The main purpose of the work described here was to developmolecular modelling protocols for conformationalcharacterisation that would apply to 15-membered as well as14-membered macrolides. We have also described35

systematically the intermediate weakly bound state forerythromycin, clarithromycin and azithromycin, andcorrelated these findings with in vitro anti-bacterial activity.

Results40

The conformations of erythromycin A, clarithromycin andazithromycin bound weakly to bacterial ribosomes

Information about the conformation of a bound ligand infast exchange with free ligand can be obtained using thetransferred NOESY (TRNOESY) NMR experiment. This45

experiment relies on the fact that NOE build-up is fast in largecomplexes and slow (compared with the dissociation rate ofthe complex) in small ligands. NOE information is therefore

transferred from bound to an excess of free drug, so thatexperiments show the chemical shifts of the free drug, but50

Overhauser effects characteristic of the conformation(s) of thebound drug.

For macrolide antibiotics associated with ribosomes,ribosomal core particles (in which the outer proteins areremoved) are used instead of whole ribosomes in control55

experiments (control experiments for non-specific bindingusing large inactive complexes). Because of theirintermediate size, 14- and 15-membered macrolides havesolution NOEs close to zero (at 500 - 600 MHz and shortmixing time).60

ClarithromycinIn the presence of bacterial ribosomes, clarithromycin gives

the TRNOESY spectrum shown as Fig. S1A in theSupplementary Data. The spectrum contains 75 pairs of65

signals, consistent with a folded-out conformation ofclarithromycin (Table S1 in the Supplementary Data). Thirty-three of these signals are not seen in the ROESY spectrum ofthe same drug but are completely consistent with the folded-out structure. Only the weak long-range signal (7'8'-8") gives70

additional information, constraining the relative positions ofthe two sugars. The crystal structure of clarithromycin isclosely similar, although the H4-H11 (3.04 Å) and the H7'H8'-H8" (7.71 Å) distances obtained by molecular modelling(molecular mechanics) are rather long.75

Erythromycin AFigure S1B in the Supplementary Data shows the

TRNOESY spectrum of erythromycin A. There are 83 pairs ofsignals belonging to erythromycin A 9-ketone (Table S1,80

Supplementary Data). Up to four small signals (depending onprocessing) due to the hemiacetal, and due to chemicalexchange between the two isomers, are also seen in thisspectrum, but it is clear that the 9-ketone binds to ribosomesand the 9,12-acetal does not.85

Most of the ketone signals are equivalent to those found inthe clarithromycin TRNOESY spectrum suggesting that thedominant conformation of erythromycin is the same folded-

Fig. 2 Erythromycin A 12,9-hemiacetal: global minimumfrom unconstrained Monte Carlo search


out conformation as adopted by clarithromycin. Thedifferences between the TRNOESY spectra of clarithromycin

and erythromycin are concentrated in two areas of themolecule. Clarithromycin shows more intra-sugar cross-peaksfor the desosamine sugar, but the extra peaks are all very5

small and consistent with spin diffusion. The erythromycinspectrum shows an H8-H11 cross-peak, and H7proS-H318 andH7proS-H319 connectivities are also present. This is consistentwith erythromycin retaining some flexibility in the top portionof the molecule and thus the ability to form an 8-endo folded-10

out conformer, even when bound to ribosomes.TRNOESY data were previously published by Bertho et

al.16 Differences between our results and theirs are indicatedin Table 1 of the Supplementary Data. The additional peaksthat they report can be attributed to spin diffusion, because of15

the use of non-deuteriated ribosomes and a relatively longmixing time.

AzithromycinThe TRNOESY spectrum of azithromycin in the presence

of bacterial ribosomes is presented in Figure S1C of the20

Supplementary Data, together with tabulated data (Table 2 ofthe Supplementary Data). The vast majority of cross-peaks areidentical with those seen for clarithromycin, suggesting thatazithromycin, from a range of available conformations, adoptsthe clarithromycin-like folded-out conformation when bound25

to ribosomes.Azithromycin enjoys some flexibility in the H2-H5 region

so that H3 can approach H11 without significant effects on therest of the molecule. The large H2-H4 cross-peak is a result ofthe same rotation. This signal is small in both the30

erythromycin and clarithromycin spectra. Superimposedfolded-out structures of clarithromycin and azithromycin areshown in Figure S2 in the Supplementary Data. There is clearevidence of rotation about the C14-C15 bond of azithromycin,such that H315 can approach both H316 and H321. This can35

be seen in the crystal structure of azithromycin bound to D.radiodurans ribosomes; in this structure two molecules ofazithromycin bind, one with C15 approaching C21 and theother with C15 (more conventionally) approaching C16.

Comparison of the binding modes of clarithromycin,40

erythromycin and azithromycin

The dominant bound conformations of clarithromycin,erythromycin A and azithromycin are very similar. However,clarithromycin shows almost no flexibility in the bound statewhereas erythromycin shows some limited flexibility and45

azithromycin rather more. For all three drugs, the methylgroup H316 gives a broad 1H NMR signal. Whenclarithromycin is bound to ribosomes, the TRNOESY H16-H2cross-peak almost disappears, whereas in the TRNOESYspectra of the more flexible azithromycin and erythromycin A,50

this signal is clearly visible. Figure 3 shows this region of theTRNOESY spectra of the three drugs.

Unconstrained molecular mechanics and moleculardynamics searches on clarithromycin, erythromycin A,55

azithromycinWe now compared results of unconstrained modelling using

two conformational analysis approaches: a Monte CarloMulitple Minima (MCMM) search and a molecular dynamics-based simulated annealing (MDSA) method. Unconstrained60

MCMM calculations were carried out with Macromodelversion 8 using the AMBER* force field19 and water as asolvent (GBSA solvent model)20. Molecular dynamicscalculations were performed in Sybyl 7.3 (Tripos, St. Louis, )using simulated annealing with energy minimization using65

Tripos force field and water solvation model21. The startingstructures for both modelling protocols were crystal structuresof the macrolides alone (clarithromycin 22, erythromycin A9

and azithromycin23) available in the CambridgeCrystallographic Database.70

ClarithromycinThe molecular mechanics was developed in order to

describe molecular structures and properties in as practicalmanner as possible. However, the goal of molecular dynamics75

Fig.3 Comparison of the TRNOESY NMR spectra of A:erythromycin A, B: clarithromycin and C: azithromycin, showingthe size of the 16,2 crosspeak.


is to simulate the actual changes in a molecule as a function oftime after an energy input is added to a molecule atequilibrium. Molecular mechanics using the AMBER forcefield gave a folded-out global minimum (found 19 times) asshown in Fig. 4A. All the NMR constraints are satisfied in5

this single conformer. The global minimum found bymolecular dynamics (simulated annealing) experiments wasalmost identical with this structure, as shown in Fig. 4B. Keyinternuclear distances for these structures are shown in theSupplementary Data Table S3.10

Fig. 4 A: Global minimum of unconstrained Monte Carlo search onclarithromycin in water; B Superposition of this global minimum (pink)and the global minimum (shown in green) obtained by simulatedannealing. Both structures satisfy all the TRNOESY NMR constraints.

Erythromycin

Fig. 5. Families of conformers of erythromycin A within 1 kcal mol-1

of the global minimum found by Monte Carlo searching: A folded-out; B and C variants of the 8-endo-folded-out conformation; D 7-endo-folded-in conformer. E Superposition of this global minimum(pink) and the global minimum (shown in black) obtained bysimulated annealing

Unconstrained molecular mechanics calculations suggest

that the 6-O-Me group of clarithromycin has a profound effecton its conformational behaviour. Erythromycin A shows four15

different conformers within 1 kcal mol-1 of the globalminimum: Figure 5 shows a folded-out conformer, an 8-endo-folded-out conformer (in which H8 approaches H11) 4 in twoversions differing in their H4-H11 distances, and a 7-endo-folded-in conformer in which H11 approaches both H3 and20

H7 (but not H8). On the basis of the populations of structureswithin 1 kcal mol-1 of the global minimum, we would expectthe folded-out conformer to predominate (about 75% of theminima found belong to this family). On the same basis, the7-endo-folded-in conformer would be present at less than 5%.25

Abundance of conformers was calculated using their energies.Simulated annealing gave very similar results, with a folded-out structure again appearing at lowest energy and 8-endo-folded-out and 7-endo-folded-in structures appearing within 1kcal mol-1. The orientation of the cladinose sugar is slightly30

different in the global minimum obtained by simulatedannealing from that obtained by molecular mechanics (seeTable S4 in the Supplementary Data for details), so that thesuperposition of the two sets of modelling results (Fig 5E) isless good than for clarithromycin (Fig. 4B).35

The TRNOESY NMR spectrum for erythromycin A showsH2"s-H4" and H8"-H2"r peaks which may be assumed to bedue to spin diffusion. Otherwise, all the cross-peaks aresatisfied by the 8-endo-folded-out conformer and, with theexception of H8-H11, by the global minimum folded-out40

conformer.

Azithromycin

Fig. 6. Molecular modelling of azithromycin A: Global minimum ofunconstrained Monte Carlo search in water B: Global minimum ofMonte Carlo search with H3-H11 and H4-H11 constrained. C:Unconstrained molecular dynamics global minimum. D:Unconstrained molecular dynamics structure 3.8 kcal mol-1 above theglobal minimum.

An unconstrained molecular mechanics calculation gaverise to a global minimum structure (Fig. 6A) with an energy of30.6 kcal mol-1. This structure does not resemble any known45

conformation for macrolide antibiotics, neither does it satisfythe NMR constraints (See Table S5 in the SupplementaryData). Interestingly, it closely resembles the 7-endo-folded-inerythromycin, which was found in silico (see Fig. 5). The


conformation 7-endo-folded-in does not resemble any knownmacrolide conformation. It has some characteristics of folded-in and folded-out conformations. The conformation is twistedinside, decreasing the distances of H27 to both H4 and H3.

There were no familiar conformers within 2.4 kcal mol-1 of5

the global minimum; the H3-H11 and H4-H11 distances wereall too large. In order to achieve a folded-out structure thatwould satisfy the NMR data, it was necessary to constrain theMonte Carlo search, setting H3-H11 to 3.5±0.3 Å and H4-H11to 2.5±0.3 Å. This yielded a conformer of energy 35.3 kcal10

mol-1 that satisfied all the NMR constraints (Fig. 6B).Molecular dynamics calculations were superficially more

immediately successful, and the global minimum in theunconstrained calculation was a folded-out structure (Fig. 6C)in which the NMR constraints were all satisfied, except that15

the inter-sugar distance H1'-H5" of 2.77 Å is long – below 2.5Å is expected. The structure, however, is profoundly differentfrom the crystal structure; the orientation of the sugar rings isquite unexpected.

A more conventional folded-out structure with H1'-H5" of20

2.35 Å, was found at 3.8 kcal mol-1 above the global minimum(Fig. 6D).

The conformational flexibility of azithromycin: 1H NMRstudies and molecular dynamics simulation25

The failure of both molecular dynamics and molecularmechanics calculations to identify the global minimum ofazithromycin found experimentally was disappointing. Weconsidered the (highly remote) possibility that this relativelylarge molecule could adopt a metastable conformation in30

aqueous solution, above the global minimum. In order toexplore this possibility, we heated an aqueous solution ofazithromycin in an NMR tube over the temperature range 30°C – 90 °C and allowed the solution to cool, measuring the 1HNMR spectrum at intervals. Figure S3 of the Supplementary35

Data shows the (uninteresting) 1H NMR spectra obtained inthis experiment. There was no change in the spectra over thetemperature range other than the expected slight temperature-dependent shifts. The final spectrum was identical with thestarting spectrum.40

The same experiment was carried out in silico (moleculardynamics calculations) starting from the folded-out crystalstructure of azithromycin. It showed the same result as 1HNMR experiment. In aqueous solution the conventionalfolded-out conformer of azithromycin is stable.45

The properties of descladinosylclarithromycin anddescladinosylazithromycin

As shown above, the second generation macrolides,although often assumed to be equivalent to one another, differhugely in their conformational flexibility. The effect of the50

removal of the cladinose sugar was now explored. 5-Desosaminylerythronolide A (descladinosylerythromycin A)has little or no antibacterial activity24, and unconstrainedmolecular modelling shows a folded-out structure in which thedesosamine sugar fails to superimpose upon that of55

erythromycin A (data not shown).Descladinosylclarithromycin (4) anddescladinosylazithromycin (5) were constructed from the

corresponding drugs 18, 23 in silico, and were subjected tounconstrained Monte Carlo searches.60

Figure 7 shows how the global minimum structure ofdescladinosylclarithromycin is almost superimposable on thatof clarithromycin. As expected, the more flexibleazithromycin derivative did not superimpose on a folded-outmacrolide.65

The synthesis of compounds 4 and 5 by acid-catalyseddegradation of the parent drugs is described in theSupplementary Data, together with their 1H and 13C NMRdata. Measurements of minimum inhibitory concentrations of4 and 5 against an azithromycin-susceptible Escherichia coli70

K12 strain and against macrolide-susceptible Bacillus cereusand Staphylococcus aureus strains showed essentially noactivity (see Supplementary Data Table S7). This is theexpected result for descladinoseazithromycin, but illustratesthat, although descladinoseclarithromycin (4) adopts the same75

conformation as its antibacterial parent, this is insufficient foractivity.

Fig. 7 Global minima of clarithromycin (yellow) anddescladinosylclarithromycin (4, blue) in water, obtained fromunconstrained molecular mechanics, superimposed.


Unconstrained molecular mechanics on 3-O-substituteddescladinosyl clarithromycin and 3-O-substituteddescladinosyl azithromycin derivatives

Derivatives of 4 and 5, substituted at the 3-oxygen have beenreported and some have antibacterial activity 25,26.5

Compounds 6 - 9, together with the unreported 10 - 11 werenow subjected to unconstrained Monte Carlo searches. Thecompounds, their activities and structural characteristicsobtained using unconstrained molecular modelling arerepresented in Table 2. Compounds 6 and 9 have significant10

antibacterial activity, but (as in the case of the parent drugs)the global minima found for them by this method are quitedifferent compared to each other. Derivatives ofdescladinosylazithromycin showed folded-out (active)conformers about 4.8 kcal mol-1 above the global minimum.15

Only when the azithromycin skeleton was substituted with alarge steroid side-chain (compound 12) did it give rise to afolded-out structure in the global minimum of a molecularmechanics calculation. The latter compound was, however,developed not as an antibacterial but as an anti-inflammatory20

agent!

Cmpd MIC(µg ml-1)

Key distances (Å) in the conformation representingthe global minimum

S. a H. i. 3-11 4-11 3-8 8-11 4-18 5-18 15-16 16-176 0.78 12.5 4.0 2.5 6.1 4.3 4.0 2.7 2.8 4.29 3.13 1.56 2.4 4.5 3.9 4.8 2.7 3.5 5.3 2.3

12 NA NA 4.0 2.3 5.9 4.3 4.0 2.6 5.1 4.4Table 2: Key internuclear distances in the global minima of compounds6, 9 and 12. The global minima of 7 and 8 are essentially identical with 6and those of 10 and 11 are essentially identical with 9. (S.a.=Staphylococcus aureus, H.i. = Haemophilus influenzae)

DiscussionMolecular modelling calculations, both molecular25

mechanics and molecular dynamics are successful in findingglobal minima for the 14-membered macrolides. In the caseof clarithromycin, the folded-out structure (Fig 4) is the onlystructure to be found by modelling and the only structureindicated by NMR. The trNOESY data for clarithromycin30

bound to bacterial ribosomes indicate that this drug has verylimited conformational flexibility, hence the almost completedisappearance of the H2-H316 cross-peak. Erythromycin A ismore flexible than clarithromycin, and is able to adopt an 8-endo-folded-out as well as a conventional 8-exo-folded-out35

conformation. Molecular modelling predicts a minor 7- endo-folded-in conformer, but this has not been distinguishedexperimentally.

Azithromycin, the 15-membered macrolide, is, however,quite different. Molecular modelling indicates that a wide40

range of conformations is available to this drug. Although thepreferred conformation (as indicated by NMR) is essentiallythe same folded-out structure as is seen for erythromycin Aand clarithromycin, this drug is conformationally flexible,because of the increased size of the macrolide ring.45

In three distinct situations (free solution 8, weakly-bound tobacterial ribosomes in the present work, and tightly-bound toribosomes detected by X-ray crystallography 13)clarithromycin, erythromycin A 9-ketone and azithromycin

adopt essentially the same conformation, as shown in Fig. 1.50

(The distorted conformation of erythromycin reported in earlyX-ray studies of drug-ribosome complexes 11 has beensuperseded 27; it probably derived from the relatively lowresolution of these studies and from their focus on theribosome rather than the drug.) Molecular dynamics or55

molecular mechanics calculations can be used to predictwhether a macrolide has this conformation available, but inthe case of azithromycin, we have been unable to find asimple modelling protocol using commercially availablesoftware that successfully finds the lowest energy60

conformation seen experimentally. Choosing an appropriatesolvent and solvent model is one of the key stages in MonteCarlo calculations. Implicit solvation represents a compromisesolution in which explicitly represented water molecules arereplaced by a continuum model that reflects the average65

behaviour of water molecules. Our choice of the GBSAsolvent model (Generalized Born model augmented with thehydrophpobic solvent accessible surface area SA term) andAMBER* force field gives good results for conformationallyrigid molecules like clarithromycin, its 3-O-substituted70

descladinosyl derivatives and erythromycin A. However,imperfections in water model and force field causes partialinability of predicting the biologically active conformation ofconformationally flexible molecules like azithromycin and its3-O-substituted descladinosyl derivatives. Where there is75

more flexibility, there are more opportunities for intra-soluteinteractions. These intrasolute interactions could beoveremphasised using implicit solvent rather than explicitwater molecules – the latter could more adequately screen andprovide bridging hydrogen bond networks. Implicit solvent80

has shown inadequate shielding for salt bridges.28,29

At the present time, we can model clarithromycinderivatives and predict that only a molecule that adopts afolded-out conformation (Fig. 1) in the global minimum islikely to be active, whereas for azithromycin derivatives we85

are required to search for the active conformer among therange of structures indicated by modelling.

Macrolides are among the most efficacious and safe drugsin the clinic. A frustration is that they are all based on naturalproducts containing many chiral centres. This means that90

systematic investigations of structure-activity relationships bydirected modification of the skeleton are simply not feasible.It is conventional to think of the semi-synthetic antibioticsazithromycin and clarithromycin as broadly equivalent; indeedhospital antibiotic policies often permit one but not the other.95

Both are acid-stable, hydrophobic derivatives of erythromycinA and both have excellent pharmacokinetic properties.However, it has long been known that these remarkablecompounds have quite different spectra of activity.Azithromycin has the rare property of excellent anti-Gram-100

negative activity (and even anti-malarial activity), whereasclarithromycin has excellent activity against Gram positiveorganisms.

Here we postulate that the chemical basis of the differencesbetween these drugs may lie in their very different105

flexibilities. Azithromycin is conformationally labile,confounding the barriers that cells throw up to bar the entry of


drugs. What clarithromycin lacks in subtlety it makes up forin focus: sensitive Gram positive infections are very sensitiveindeed.

Experimental

Molecular modelling5

Erythromycin A, clarithromycin and its 3-O-substituteddescladinosyl derivatives, azithromycin and its 3-O-substituted descladinosyl derivatives were constructed usingMacromodel 8 software 30 from the appropriate crystalstructures. Structures were minimized using the Truncated10

Newton Conjugate Gradient (TNCG) method in order toobtain local minima. The Monte Carlo Multiple Minimum(MCMM) conformational search was used to find the globalminima.31 The GB/SA model was used for water solvation.The search was set to 10000 structures to be minimized and15

all structures within 6 kcal mol-1 energy range were stored.Molecular dynamics calculations were performed using

Sybyl 7.3 32 Minimization was performed using Tripos forcefield in water as a solvent (=80.4). The Gasteiger-Huckelcharges was used through the calculation. Global minima of20

clarithromycin, erythromycin A and azithromycin wereobtained from their crystal structures minimized by Triposforce field followed by simulated annealing [(clarithromycin;heated to 600 K for 1500 fs and then cooled to 0 K for 10000fs), (erythromycin A; heated to 1000 K for 750 fs and then25

cooled to 0 K for 10000 fs) and (azithromycin; heated to 600K for 2000 fs and then cooled to 0 K for 10000 fs)]. Obtainedstructures were minimized again using Tripos force field. Thenumber of iterations for each minimization was 10000 and thejob was terminated gradiently when the difference in energies30

between two conformations was no more than 0.005 kcal mol-

1.

NMR Analysis

One and two dimensional NMR spectra were obtained bystandard procedures, described in the Supplementary Data.35

Minimum Inhibitory Concentrations of clarithromycin,azithromycin and their derivatives against Escherichiacoli, Bacillus cereus, Staphylococcus aureus, Pseudomonasaeruginosa, Staphylococcus epidermidis, Serratia40

marcescens and Corynebacterium xerosis

These were determined by standard methods 33 and aredescribed in the Supplementary Data.

Conclusions45

Molecular dynamics and molecular mechanics calculationsusing commercially available software can be used to predictwhether macrolide derivatives can adopt the active folded-outconformer required for ribosome binding. In addition, anindication of the likely flexibility of a given molecule can be50

obtained. For antibacterial 14-membered macrolides, thisfolded-out conformer is likely to be found at lowest energy inunconstrained searches. 15-membered macrolides adopt thesame conformation in the active state, as shown by NMR and

X-ray crystallographic measurements, but unconstrained55

molecular modelling cannot be relied on to find thisconformer at lowest energy. Both NMR and modellingstudies indicate that azithromycin is an intrinsically flexiblemolecule and can adopt numerous conformers. The flexibilitycorrelates with the ability of the drug to target Gram negative60

and even malarial cells.

Notes and referencesaSchool of Pharmacy and Pharmaceutical Sciences, University ofManchester, Manchester UK, M13 9PTbSchool of Chemistry, University of Manchester, Manchester UK, M1365

9PL† Electronic Supplementary Information (ESI) available

1 L. Katz, G. Ashley, Chem. Rev., 2005, 105, 499.2 J. C. Craft, D. Stamler, Infectious Disease and Therapy, 1995, 1870

(New Macrolides, Azalides, and Streptogramins in Clinical Practice),359.

3 M. Zief, R. Woodside, E. Huber, Antibiotics and Chemotherapy(Washington, D. C.), 1957, 7, 604.

4 J. Barber, J. I. Gyi, L. Lian, G. A. Morris, D. A. Pye, J. K.75

Sutherland, J. Chem. Soc., Perkin Trans., 1991, 2, 1489.5 M. D. Kitris, F. W. Gold-stein, M. Miegi, J. F. Acar, J. Antimicrob.

Chemother., 1990, 25, Suppl. A, 15.6 H. Noedl, S. Krudsood, K. Chalermratana, U. Silachamroon, W.

Leowattana, N. Tangpukdee, S. Looareesuwan, R. S. Miller, M.80

Fukuda, K. Jongsakul, S. Sriwichai, J. Rowan, H. Bhattacharyya, C.Ohrt, C. Knirsch, Clinical Infectious Diseases, 2006, 43, 1264.

7 J. Poehlsgaard, S. Douthwaite, Curr. Opin. Invest. Drugs, 2003, 4,140.

8 A. Awan, R. J. Brennan, A. C. Regan, J. Barber, J. Chem. Soc.,85

Perkin Trans. 2, 2000, 1645.9 G. A. Stephenson, J. G. Stowell, P. H. Toma, R. R. Pfeiffer, S. R.

Byrn, J. Pharm. Sci., 1997, 86, 1239.10 F. Schlunzen, R. Zarivach, J. Harms, A. Bashan, A. Tocilj, R.

Albrecht, A.Yonath, F.Franceschi, Nature, 2001, 413, 814.90

11 F. Schlunzen, J. M. Harms, F. Franceschi, H. A. Hansen, H. Bartels,R. Zarivach, A. Yonath, Structure, 2003, 11, 329.

12 D. Tu, G. Blaha, P. B. Moore, T. Steitz, Cell, 2005, 121, 257.13 J. L. Hansen, J. A. Ippolito, N. Ban, P. Nissen, P. B. Moore, T. A.

Steitz, Mol. Cell, 2002, 10, 117.95

14 P. Novak, I. Tatic, P. Tepes, S. Kostrun, J. Barber, J. Phys. Chem. A,2006, 110, 572.

15 J. I. Gyi, R. J. Brennan, D. A. Pye, J. Barber, J. Chem. Soc., Chem.Commun., 1991, 1471.

16 G. Bertho, J. Gharbi-Benarous, M. Delaforge, J.-P. Girault, Bioorg.100

Med. Chem., 1998, 6, 209.17 J. R. Everett and J. W. Tyler, J. Chem. Soc., Perkin Trans. 2, 1987,

1659.18 G. A. Stephenson, J. G. Stowell, P. H. Toma, D. E. Dorman, J. R.

Greene, S. R. Byrn, J. Am. Chem. Soc. , 1994, 116, 5766.105

19 J. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman, D. A. Case, J.Comput. Chem., 2004, 25, 1157.

20 W. C. Still, A. Tempczyk, R. C. Hawley, T. Hendrickson, J. Am.Chem. Soc., 1990, 112, 6127.

21 M. Clark, R. D. Cramer III, N. Van Opdenbosch, , J. Comput. Chem.,110

1989, 10 (8), 982.22 J.-H. Liang, G.-W. Yao, J. Chem. Cryst., 2008, 38, 61.23 S. Djokic, G. Kobrehel, N. Lopotar, B. Kamena, A. Nagl, D. Mrvos,

J. Chem. Res., 1988, 152, 1239.24 R. A. LeMahieu, M. Carson, R. W. Kierstead, L. M. Fern, E.115

Grunberg, J. Med. Chem., 1974, 17, 953.25 T. Tanikawa, T. Asaka, M. Kashimura, K. Suzuki, H. Sugiyama, M.

Sato, K. Kameo, S. Morimoto, A. Nishida, J. Med. Chem., 2003, 46,2706.

26 S. Mutak, J. Antibiot., 2007, 60, 85.120


drugs. What clarithromycin lacks in subtlety it makes up forin focus: sensitive Gram positive infections are very sensitive5

indeed.

Experimental5

Molecular modelling

Erythromycin A, clarithromycin and its 3-O-substituteddescladinosyl derivatives, azithromycin and its 3-O-substituted descladinosyl derivatives were constructed usingMacromodel 8 software 30 from the appropriate crystal20

structures. Structures were minimized using the TruncatedNewton Conjugate Gradient (TNCG) method in order toobtain local minima. The Monte Carlo Multiple Minimum(MCMM) conformational search was used to find the globalminima.31 The GB/SA model was used for water solvation.25

The search was set to 10000 structures to be minimized andall structures within 6 kcal mol-1 energy range were stored.

Molecular dynamics calculations were performed usingSybyl 7.3 32 Minimization was performed using Tripos forcefield in water as a solvent (=80.4). The Gasteiger-Huckel35

charges was used through the calculation. Global minima ofclarithromycin, erythromycin A and azithromycin wereobtained from their crystal structures minimized by Triposforce field followed by simulated annealing [(clarithromycin;heated to 600 K for 1500 fs and then cooled to 0 K for 1000040

fs), (erythromycin A; heated to 1000 K for 750 fs and thencooled to 0 K for 10000 fs) and (azithromycin; heated to 600K for 2000 fs and then cooled to 0 K for 10000 fs)]. Obtainedstructures were minimized again using Tripos force field. Thenumber of iterations for each minimization was 10000 and the45

job was terminated gradiently when the difference in energiesbetween two conformations was no more than 0.005 kcal mol-

1.

NMR Analysis

One and two dimensional NMR spectra were obtained bystandard procedures, described in the Supplementary Data.



These were determined by standard methods 33 and are45

described in the Supplementary Data.

Conclusions

Molecular dynamics and molecular mechanics calculations55

using commercially available software can be used to predictwhether macrolide derivatives can adopt the active folded-outconformer required for ribosome binding. In addition, anindication of the likely flexibility of a given molecule can beobtained. For antibacterial 14-membered macrolides, this60

folded-out conformer is likely to be found at lowest energy inunconstrained searches. 15-membered macrolides adopt thesame conformation in the active state, as shown by NMR and

X-ray crystallographic measurements, but unconstrainedmolecular modelling cannot be relied on to find thisconformer at lowest energy. Both NMR and modellingstudies indicate that azithromycin is an intrinsically flexible65

molecule and can adopt numerous conformers. The flexibilitycorrelates with the ability of the drug to target Gram negativeand even malarial cells.

Notes and referencesaSchool of Pharmacy and Pharmaceutical Sciences, University of65

Manchester, Manchester UK, M13 9PTbSchool of Chemistry, University of Manchester, Manchester UK, M139PL† Electronic Supplementary Information (ESI) available

1 L. Katz, G. Ashley, Chem. Rev., 2005, 105, 499.70

2 J. C. Craft, D. Stamler, Infectious Disease and Therapy, 1995, 18(New Macrolides, Azalides, and Streptogramins in Clinical Practice),359.75

3 M. Zief, R. Woodside, E. Huber, Antibiotics and Chemotherapy75

(Washington, D. C.), 1957, 7, 604.4 J. Barber, J. I. Gyi, L. Lian, G. A. Morris, D. A. Pye, J. K.

Sutherland, J. Chem. Soc., Perkin Trans., 1991, 2, 1489.5 M. D. Kitris, F. W. Gold-stein, M. Miegi, J. F. Acar, J. Antimicrob.

Chemother., 1990, 25, Suppl. A, 15.80

6 H. Noedl, S. Krudsood, K. Chalermratana, U. Silachamroon, W.Leowattana, N. Tangpukdee, S. Looareesuwan, R. S. Miller, M.Fukuda, K. Jongsakul, S. Sriwichai, J. Rowan, H. Bhattacharyya, C.85

Ohrt, C. Knirsch, Clinical Infectious Diseases, 2006, 43, 1264.7 J. Poehlsgaard, S. Douthwaite, Curr. Opin. Invest. Drugs, 2003, 4,85

140.8 A. Awan, R. J. Brennan, A. C. Regan, J. Barber, J. Chem. Soc.,

Perkin Trans. 2, 2000, 1645.9 G. A. Stephenson, J. G. Stowell, P. H. Toma, R. R. Pfeiffer, S. R.

Byrn, J. Pharm. Sci., 1997, 86, 1239.90

10 F. Schlunzen, R. Zarivach, J. Harms, A. Bashan, A. Tocilj, R.Albrecht, A.Yonath, F.Franceschi, Nature, 2001, 413, 814.

11 F. Schlunzen, J. M. Harms, F. Franceschi, H. A. Hansen, H. Bartels,R. Zarivach, A. Yonath, Structure, 2003, 11, 329.

12 D. Tu, G. Blaha, P. B. Moore, T. Steitz, Cell, 2005, 121, 257.13 J. L. Hansen, J. A. Ippolito, N. Ban, P. Nissen, P. B. Moore, T. A.

Steitz, Mol. Cell, 2002, 10, 117.14 P. Novak, I. Tatic, P. Tepes, S. Kostrun, J. Barber, J. Phys. Chem. A,

2006, 110, 572.15 J. I. Gyi, R. J. Brennan, D. A. Pye, J. Barber, J. Chem. Soc., Chem.100

Commun., 1991, 1471.16 G. Bertho, J. Gharbi-Benarous, M. Delaforge, J.-P. Girault, Bioorg.

Med. Chem., 1998, 6, 209.17 J. R. Everett and J. W. Tyler, J. Chem. Soc., Perkin Trans. 2, 1987,

1659.105

18 G. A. Stephenson, J. G. Stowell, P. H. Toma, D. E. Dorman, J. R.Greene, S. R. Byrn, J. Am. Chem. Soc. , 1994, 116, 5766.

19 J. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman, D. A. Case, J.Comput. Chem., 2004, 25, 1157.

20 W. C. Still, A. Tempczyk, R. C. Hawley, T. Hendrickson, J. Am.110

Chem. Soc., 1990, 112, 6127.21 M. Clark, R. D. Cramer III, N. Van Opdenbosch, , J. Comput. Chem.,

1989, 10 (8), 982.22 J.-H. Liang, G.-W. Yao, J. Chem. Cryst., 2008, 38, 61.23 S. Djokic, G. Kobrehel, N. Lopotar, B. Kamena, A. Nagl, D. Mrvos,

J. Chem. Res., 1988, 152, 1239.115

24 R. A. LeMahieu, M. Carson, R. W. Kierstead, L. M. Fern, E.Grunberg, J. Med. Chem., 1974, 17, 953.

25 T. Tanikawa, T. Asaka, M. Kashimura, K. Suzuki, H. Sugiyama, M.120

Sato, K. Kameo, S. Morimoto, A. Nishida, J. Med. Chem., 2003, 46,2706.

26 S. Mutak, J. Antibiot., 2007, 60, 85.


drugs. What clarithromycin lacks in subtlety it makes up forin focus: sensitive Gram positive infections are very sensitiveindeed.

Experimental

Molecular modelling

Erythromycin A, clarithromycin and its 3-O-substituteddescladinosyl derivatives, azithromycin and its 3-O-substituted descladinosyl derivatives were constructed using30

Macromodel 8 software 30 from the appropriate crystalstructures. Structures were minimized using the TruncatedNewton Conjugate Gradient (TNCG) method in order toobtain local minima. The Monte Carlo Multiple Minimum(MCMM) conformational search was used to find the global35

minima.31 The GB/SA model was used for water solvation.The search was set to 10000 structures to be minimized andall structures within 6 kcal mol-1 energy range were stored.

Molecular dynamics calculations were performed usingSybyl 7.3 32 Minimization was performed using Tripos force50

field in water as a solvent (=80.4). The Gasteiger-Huckelcharges was used through the calculation. Global minima ofclarithromycin, erythromycin A and azithromycin wereobtained from their crystal structures minimized by Triposforce field followed by simulated annealing [(clarithromycin;55

heated to 600 K for 1500 fs and then cooled to 0 K for 10000fs), (erythromycin A; heated to 1000 K for 750 fs and thencooled to 0 K for 10000 fs) and (azithromycin; heated to 600K for 2000 fs and then cooled to 0 K for 10000 fs)]. Obtainedstructures were minimized again using Tripos force field. The60

number of iterations for each minimization was 10000 and thejob was terminated gradiently when the difference in energiesbetween two conformations was no more than 0.005 kcal mol-

1.

NMR Analysis35

One and two dimensional NMR spectra were obtained bystandard procedures, described in the Supplementary Data.



These were determined by standard methods 33 and aredescribed in the Supplementary Data.

Conclusions

Molecular dynamics and molecular mechanics calculationsusing commercially available software can be used to predict65

whether macrolide derivatives can adopt the active folded-outconformer required for ribosome binding. In addition, anindication of the likely flexibility of a given molecule can beobtained. For antibacterial 14-membered macrolides, thisfolded-out conformer is likely to be found at lowest energy in70

unconstrained searches. 15-membered macrolides adopt thesame conformation in the active state, as shown by NMR and

X-ray crystallographic measurements, but unconstrainedmolecular modelling cannot be relied on to find this70

conformer at lowest energy. Both NMR and modellingstudies indicate that azithromycin is an intrinsically flexiblemolecule and can adopt numerous conformers. The flexibilitycorrelates with the ability of the drug to target Gram negativeand even malarial cells.75

Notes and referencesaSchool of Pharmacy and Pharmaceutical Sciences, University ofManchester, Manchester UK, M13 9PTbSchool of Chemistry, University of Manchester, Manchester UK, M139PL70

† Electronic Supplementary Information (ESI) available70

1 L. Katz, G. Ashley, Chem. Rev., 2005, 105, 499.2 J. C. Craft, D. Stamler, Infectious Disease and Therapy, 1995, 18

(New Macrolides, Azalides, and Streptogramins in Clinical Practice),359.

3 M. Zief, R. Woodside, E. Huber, Antibiotics and Chemotherapy(Washington, D. C.), 1957, 7, 604.

4 J. Barber, J. I. Gyi, L. Lian, G. A. Morris, D. A. Pye, J. K.Sutherland, J. Chem. Soc., Perkin Trans., 1991, 2, 1489.80

5 M. D. Kitris, F. W. Gold-stein, M. Miegi, J. F. Acar, J. Antimicrob.Chemother., 1990, 25, Suppl. A, 15.

6 H. Noedl, S. Krudsood, K. Chalermratana, U. Silachamroon, W.Leowattana, N. Tangpukdee, S. Looareesuwan, R. S. Miller, M.Fukuda, K. Jongsakul, S. Sriwichai, J. Rowan, H. Bhattacharyya, C.Ohrt, C. Knirsch, Clinical Infectious Diseases, 2006, 43, 1264.90

7 J. Poehlsgaard, S. Douthwaite, Curr. Opin. Invest. Drugs, 2003, 4,140.

8 A. Awan, R. J. Brennan, A. C. Regan, J. Barber, J. Chem. Soc.,Perkin Trans. 2, 2000, 1645.90

9 G. A. Stephenson, J. G. Stowell, P. H. Toma, R. R. Pfeiffer, S. R.Byrn, J. Pharm. Sci., 1997, 86, 1239.

10 F. Schlunzen, R. Zarivach, J. Harms, A. Bashan, A. Tocilj, R.Albrecht, A.Yonath, F.Franceschi, Nature, 2001, 413, 814.

11 F. Schlunzen, J. M. Harms, F. Franceschi, H. A. Hansen, H. Bartels,95

R. Zarivach, A. Yonath, Structure, 2003, 11, 329.12 D. Tu, G. Blaha, P. B. Moore, T. Steitz, Cell, 2005, 121, 257.95

13 J. L. Hansen, J. A. Ippolito, N. Ban, P. Nissen, P. B. Moore, T. A.Steitz, Mol. Cell, 2002, 10, 117.

14 P. Novak, I. Tatic, P. Tepes, S. Kostrun, J. Barber, J. Phys. Chem. A,100

2006, 110, 572.15 J. I. Gyi, R. J. Brennan, D. A. Pye, J. Barber, J. Chem. Soc., Chem.

Commun., 1991, 1471.16 G. Bertho, J. Gharbi-Benarous, M. Delaforge, J.-P. Girault, Bioorg.

Med. Chem., 1998, 6, 209.105

17 J. R. Everett and J. W. Tyler, J. Chem. Soc., Perkin Trans. 2, 1987,1659.

18 G. A. Stephenson, J. G. Stowell, P. H. Toma, D. E. Dorman, J. R.Greene, S. R. Byrn, J. Am. Chem. Soc. , 1994, 116, 5766.

19 J. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman, D. A. Case, J.110

Comput. Chem., 2004, 25, 1157.20 W. C. Still, A. Tempczyk, R. C. Hawley, T. Hendrickson, J. Am.

Chem. Soc., 1990, 112, 6127.21 M. Clark, R. D. Cramer III, N. Van Opdenbosch, , J. Comput. Chem.,

1989, 10 (8), 982.115

22 J.-H. Liang, G.-W. Yao, J. Chem. Cryst., 2008, 38, 61.23 S. Djokic, G. Kobrehel, N. Lopotar, B. Kamena, A. Nagl, D. Mrvos,115

J. Chem. Res., 1988, 152, 1239.24 R. A. LeMahieu, M. Carson, R. W. Kierstead, L. M. Fern, E.

Grunberg, J. Med. Chem., 1974, 17, 953.25 T. Tanikawa, T. Asaka, M. Kashimura, K. Suzuki, H. Sugiyama, M.

Sato, K. Kameo, S. Morimoto, A. Nishida, J. Med. Chem., 2003, 46,2706.125

26 S. Mutak, J. Antibiot., 2007, 60, 85.


27 J. A. Dunkle, L. Xiong, A. S. Mankin, J. H. Cate, Proc Natl Acad SciU S A., 2010, 107, 17152.

28 C. Simmerling, B. Strockbine, , A. E. Roitberg. J. Am. Chem.Soc.2002, 124, 11258.

29 R. Geney, M. Layten, R.. Gomperts, V. Hornak, C.Simmerling, J.5

Chem. Theory Comput. 2005, 2, 115.30 F. Mohamadi, N. G. J. Richards, W. C. Guida, R. Liskamp, M.

Lipton, C. Caufield, G. Chang, T. Hendrickson, W. C. Still, J.Comput. Chem., 1990, 11, 440.

31 M. Saunders, K. N. Houk, Z. D. Wu, W. C. Still, M. Lipton, G.10

Chang, W. C. Guida, J. Am. Chem. Soc., 1990, 112, 1419.32 SYBYL 7.3 (2006), Tripos Associates, 1699 South Hanley Road,

Suite 303, St. Louis, MO, 63144-2917.33 A. J. McBain, R. G. Ledder, P. Sreenivasan, P. Gilbert, J.

Antimicrob. Chemother., 2004, 53(5), 772.15

20

1

Supplementary material

Biljana Arsic1, Abida Awan1, Richard J. Brennan1, Andrew C. Regan2,Ruth Ledder1, Andrew McBain1, Gareth A. Morris2, Richard Bryce1 and JillBarber1*

1School of Pharmacy and Pharmaceutical Sciences, University of Manchester,

Manchester UK, M13 9PT2School of Chemistry, University of Manchester, Manchester UK, M13 9PL

2

Fig S1A: 600 MHz TRNOESY NMR spectra of clarithromycin (4 mmol dm-3) in the presence of 0.8µmol dm-3 ribosomes in 50 mmol dm-3 sodium phosphate buffer in D2O apparent pH 7.6 (mixingtime 75 ms)

3

Fig S1B: 600 MHz TRNOESY NMR spectra of erythromycin A (4 mmol dm-3) in the presence of0.8 µmol dm-3 ribosomes in 50 mmol dm-3 sodium phosphate buffer in D2O apparent pH 7.6(mixing time 75 ms). Small signals belonging to the hemiacetal form of the drug are labelled inblue.

4

Fig S1C: 600 MHz TRNOESY NMR spectra of azithromycin (4 mmol dm-3) in the presence of 0.8µmol dm-3 ribosomes in 50 mmol dm-3 sodium phosphate buffer in D2O apparent pH 7.6 (mixingtime 75 ms)

5

2 3 4 5 7s 7r 8 10 11 13 14r 14s 15 16 17 18 19 20 21 22 1' 2' 3' 4' r 4' s 5' 6' 7' 8' 1"PredragNovak"

<[email protected]>

2”r 2”s 4"PredragNovak"

<[email protected]>

5"PredragNovak"

<[email protected]>

6"PredragNovak"

<[email protected]>

7"PredragNovak"

<[email protected]>

8"PredragNovak"

<[email protected]>

2 ● s s s l3 s ● m * s * m m4 s m ● l s m5 m # ● s * m m * l7s # ● l s m * * * s7r * l ● s * s s8 # * s ● * l s10 s ● m m11 m s s s m ● m m m m13 # m ● m s m *14r s ● l m *14s # * l ● s s15 # # l m m ● s16 m s s # m ● l17 m # m s ● m18 s # m s s m # ● l19 s l m ● m20 l s ●21 l s s # s ●22 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● s m1' s m ● ● s M m s m s2' ● s ● s l3' ● s ● s s s * m *4' r ● * S ● l s m4' s ● M l ● s s m5' ● s S s ● m s *6' m ● s # S m m l ● m7' 8' ● # l L m m l ● S1" m l ● ● m m2”r ● s ● l m m M2”s ● s l ● * l *4"PredragNovak"

<[email protected]>

● m s ● l5"PredragNovak"

<[email protected]>

m # # ● m s ● m6"PredragNovak"

<[email protected]>

● # l ●7"PredragNovak"

<[email protected]>

● m m l ● L8"PredragNovak"

<[email protected]>

m ● s s s s s m s l ●2 3 4 5 7s 7r 8 10 11 13 14r 14s 15 16 17 18 19 20 21 22 1' 2' 3' 4' r 4' s 5' 6' 7' 8' 1"

PredragNovak"

<[email protected]>

2”r 2”s 4"PredragNovak"

<[email protected]>

5"PredragNovak"

<[email protected]>

6"PredragNovak"

<[email protected]>

7"PredragNovak"

<[email protected]>

8"PredragNovak"

<[email protected]>

Table S1: 600 MHz TRNOESY connectivities for clarithromycin (4mM, top right) and erythromycin9-ketone (bottom right) in the presence of 0.8 µM ribosomes in 50 mM sodium phosphate buffer inD2O apparent pH 7.6 (mixing time 75 ms). Red asterisks denote signals absent in one spectrum

but present in the other. Green hashes represent additional signals seen by Bertho et al. 1

1 G. Bertho, J. Gharbi-Benarous, M. Delaforge, J.-P. Girault, Bioorg. Med. Chem., 1998, 6, 209.

6

2 3 4 5 7s 7r 8 9s 9r 10 11 13 14r 14s 15 16 17 18 19 20 21 22 1' 2' 3' 4' r 4' s 5' 6' 7' 8' 1"PredragNovak"<[email protected]>

2”r 2”s 4"PredragNovak"<[email protected]>

5"PredragNovak"<[email protected]>




2 ● s s ● ● * s l3 s ● m * ● ● * s * * m m4 l m ● * ● ● l s m5 m ● s ● ● * m m * l7s m ● l s ● ● m * * * s7r * ● s ● ● * s s8 * * ● ● ● * l s9s s ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●9r m l ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●10 s m m ● m * m11 s m m * ● m * m m m13 m ● m s m *14r m ● l m *14s s l ● s s15 s l m m ● s16 s s * m ● l17 m s m m ● m18 s m s s m s ● l19 m m m m ● m20 s m m ●21 m s m s ●22 m ● s m1' m m ● s m m s m s2' s ● s l3' s ● s s s m4' r s ● l s m4' s m l ● s s m5' m s s ● m s6' * s s m ● m7' 8' m m m s ● s1" m m ● m m2”r m ● l m m m2”s s l ● * l *4"PredragNovak"

<[email protected]>

m ● l5"PredragNovak"

<[email protected]>

m m s s ● m6"PredragNovak"

<[email protected]>

m m ●7"PredragNovak"

<[email protected]>

s m m ● l8"PredragNovak"

<[email protected]>

m s s s m m ●2 3 4 5 7s 7r 8 9s 9r 10 11 13 14r 14s 15 16 17 18 19 20 21 22 1' 2' 3' 4' r 4' s 5' 6' 7' 8' 1"

PredragNovak"<[email protected]>

2”r 2”s 4"PredragNovak"<[email protected]>





Table S2: 600 MHz TRNOESY connectivities for clarithromycin (4mM, top right) andazithromycin (bottom right) in the presence of 0.8 µM ribosomes in 50 mM sodium phosphatebuffer in D2O apparent pH 7.6 (mixing time 75 ms). Red asterisks denote signals absent inone spectrum, present in the other. Note that Me22 for the two drugs are not analogous.

7

Figure S2. Superimposed folded-out structuresof clarithromycin (red) and azithromycin (violet)

Internuclear distance (Å)Contact Global minimum (Macromodel) Structure 1 (Sybyl7.3)H4-H11 2.7 2.3

H15-H16 2.7 2.7H3-H8 5.8 6.3

H3-H11 3.8 4.0H4-H18 4.1 4.1H5-H18 2.6 2.7

H16-H17 4.0 4.3Table S3: Key internuclear distances for unconstrained modelling ofclarithromycin: the global minimum of a Monte Carlo search using

Macromodel and the global minimum obtained by simulated annealing.

Internuclear distance (Å)Contact Global minimum (Macromodel) Structure 1 (Sybyl7.3)

5-5” 2.3 2.316-1” 2.6 3.017-8” 4.1 4.11’-5” 2.3 4.35’-5” 2.7 4.31”-3 2.4 2.6

Table S4: Key internuclear distances for unconstrained modelling oferythromycin 9-ketone: the global minimum of a Monte Carlo search using

Macromodel and the global minimum obtained by simulated annealing.

8

Internuclear distance (Å)Contact Global minimum (Macromodel) NMR constraint

4-11 4.0 2.5 ± 18-18 3.9 2.3 ± 1

9proR-18 4.7 2.5 ± 111-13 3.6 2.3 ± 0.515-16 4.9 2.5 ± 218-22 7.5 3.5 ± 2.5

9proR-21 4.0 2.3 ± 12”proR-8” 4.0 2.3 ± 1

Table S5: Global minimum for azithromycin found by unconstrainedconformational search: TRNOESY NMR data not satisfied by this conformer

Figure S3. Part of the 500 MHz 1H NMRspectra of azithromycin dissolved inbuffered D2O (apparent pH=7) in thetemperature range 300C-900C A 1HNMR spectrum of azithromycin at roomtemperature; B 1H NMR spectrum ofazithromycin on 900C ; C 1H NMRspectrum of azithromycin heated on900C, and then cooled to 300C.

9

Assignment of the 1H and 13C NMR spectra of descladinosylclarithromycin anddescladinosylazithromycin.

1D (1H and 13C DEPT45, DEPT90, DEPT135) and 2D (COSY, gHMBC andHSQC) NMR spectra of descladinosyl clarithromycin and descladinosyl azithromycinwere recorded on a Bruker Avance 300 and 500 spectrometer operating at 300 MHzand 500 MHz for 1H, respectively. The spectra were run in CDCl3 (1H spectra wererecorded also with D2O shake). The concentration of descladinosyl clarithromycin anddescladinosyl azithromycin for recording 1H was 4.3 mg/ml and for all other spectra57.1 mg/ml. The 1H spectra were processed using Gaussian window function withparameters optimized empirically. The 13C, DEPT45, DEPT90 and DEPT135 wereprocessed using EM window function with LB value equals to 1.0 Hz. Two dimensionalNMR spectra were processed using QSINE window function with other parametersoptimized empirically (e.g. SSB was equal to 2).

The strategy for assignment of the 1H and 13C NMR spectra ofdescladinosylclarithromycin was similar to that used for other macrolide derivatives. 2

The quaternary carbons C9 and C1 were assigned by comparison of the highfrequency region of the one-dimensional 13C spectrum with that of the DEPT45spectrum. The three methylene carbons, C7, C4' and C14 were assigned non-specifically using the DEPT135 spectrum. H322 gives rise to a characteristic 3H singletand H67'/8' to a broad 6H singlet at 2.95 and 2.37 respectively. H315 gives rise to acharacteristic high field triplet in the 1D 1H spectrum.

The wealth of information in the gHMBC spectrum (see Table S6) could nowbe used to walk around the molecule, assigning each carbon and hydrogen signal.Signals connecting C5 to the sugar hydrogens and H5 to the sugar carbons were usedto connect the macrolide and sugar rings. Stereochemistry at C7, C14 and C4’ wasassigned by analogy with the spectrum of clarithromycin (chemical shifts and J-coupling information).

The strategy for assigning the NMR spectra of descladinosylazithromycin wasexactly analogous.

2 P. Tyson, A. Hassanzadeh, M. N. Mordi, D. G. Allison, V. Marquez, J. Barber,MedChemComm, in press.

10

Position Multiplicity Descladinosylclarithromycin Descladinosylazithromycin HMBC connectivities (1H→13C)1H (ppm) 13C (ppm) 1H (ppm) 13C (ppm)

1 - - 174.0 - 176.70 -2 m 2.65 43.5 2.65 43.60 1,2,163 d 3.57 77.9 3.95 78.58 1,2,4,5,16,174 dq 2.12 34.8 2.00 39.22 3,5,175 s 3.69 87.2 3.66 94.26 4,6,7,17,18,22,1’6 - - 77.0 - 72.15 -

7 ddd p obs

1.561.94 37.8

1.301.60

41.14 8,9,18,196,8,9

8 m 2.56 44.5 2.00 26.90 9,199 - - 219.7 -9 m 2.10 69.99

10 dq p obs 3.01 36.5 2.70 61.58 9,2011 s 3.86 68.8 3.77 73.22 9,20,1312 - - 73.2 - 74.48 -13 dd 5.18 76.3 4.76 78.58 1,11,12,15,21

14 mm

1.961.50 20.4 1.85 1.45 24.86 15

13,1515 t 0.84 9.4 0.88 1.01 13,1416 d 1.25 14.2 1.30 20.16 1,2,317 d 1.12 7.3 1.07 9.86 3,4,518 s 1.37 17.8 1.30 28.68 5,6,719 d 1.12 16.8 0.92 26.90 8,920 d 1.13 11.6 1.07 6.82 9,10,1121 s 1.19 15.2 1.12 20.22 11,12,1322 s 2.98 48.6 622 s 2.37 34.961 d 4.39 105.0 4.47 105.91 5,3’2 dd 3.24 75.6* 3.27 69.53 1’,3’3 m 2.56 64.8 3.95 64.68

4dm 1.68

1.28 27.8 1.50 1.26 28.683’

m obs 3’5 m 3.51 69.6 3.50 69.096 d 1.26 20.2 1.26 25.63 5’,4’

7/8 s 2.27 39.2 2.25 36.09 3’Table S6: NMR spectra of descladinosylclarithromycin and descladinosylazithromycinin CDCl3 (1H spectra were recorded with D2O shake also). Signals unique todescladinosylclarithromycin are in green, those unique to descladinosylazithromycinare in red. Black signals are common to both compounds. Note that for clarity thenumbering of carbons 20-22 used here reflects the numbering used for erythromycinA.(*The position of the signal due to C2’ is very dependent on pH, moving between δ 70and 75 ppm)

11

OrganismMIC( µg ml-1)

Clarithromycin Azithromycin Descladinosylclarithromycin

Descladinosylazithromycin

Escherichia coli 12 2 >1000 500Bacillus cereus 0.2 4 1000 1000Staphylococcus

aureus 0.2 8 1000 1000

Pseudomonasaeruginosa 62.5 15.6 >1000 >1000

Staphylococcusepidermidis 250 31.2 >1000 >1000

Serratia marcescens >1000 >1000 >1000 >1000Corynebacterium

xerosis >1000 250 >1000 >1000

Table S7: Minimum inhibitory concentrations of clarithromycin, azithromycin and theirdescladinosyl derivatives against selected Gram positive and Gram negative bacteria

Experimental ProceduresSynthesis of 3-O-descladinosyl-6-O-methyl-erythromycin A

Clarithromycin (0.1 g) was dissolved in acetone (10 ml) and treated with 0.25 MHCl (2 ml) overnight. The reaction mixture was then washed with dichloromethane (3 x15 ml) to remove free cladinose. The aqueous layer was adjusted to pH 8.5-9 byaddition of saturated sodium bicarbonate solution and extracted using ethyl acetate (4x 15 ml). The combined ethyl acetate layers were washed once with water (50 ml) anddried over anhydrous potassium carbonate. After removal of the drying agent, thesolvent was evaporated to dryness to give colourless crystals. The crude product waspurified using silica column chromatography (chloroform: methanol, 5:1). Yield: 31.6mg (40%). Mp. 115 0C. C30H56NO10 requires: 590.3904; found: 590.3889. The structurewas confirmed using one-dimensional 1H and 13C and two-dimensional COSY andHMBC NMR analysis. The proton and carbon chemical shifts, multiplicities and HMBCconnectivities for the 3-O-descladinosyl-6-O-methyl-erythromycin A are summarized inTable S6.

Synthesis of 3-O-descladinosyl-9-deoxo-9-dihydro-9a-aza-9a-homoerythromycinA.

Azithromycin (200 mg) was treated with 0.25 M HCl (2 ml) overnight. Thereaction mixture was washed using dichloromethane (3x15 ml). The pH of aqueouslayer was adjusted to 9 using saturated solution of sodium bicarbonate and extractedusing ethyl acetate (4x15 ml). The combined organic layers were washed with water(50 ml) and dried over anhydrous potassium carbonate. After removal of the dryingagent, the solvent was evaporated to dryness to give colourless crystals. The crudeproduct was purified using silica column chromatography (chloroform: methanol, 5:1).Yield 91.4 mg, 58%. Mp. 106-1080C. C30H59N2O9 requires: 591.4221; found: 591.4213.The structure was confirmed using one-dimensional 1H and 13C and two-dimensionalCOSY and HMBC NMR analysis. The proton and carbon chemical shifts, multiplicitiesand HMBC connectivities for 3-O-descladinosyl-9-deoxo-9-dihydro-9a-aza-9a-homoerythromycin A are summarized in Table S7.

NMR analysis1D (1H and 13C DEPT45, DEPT90, DEPT135) and 2D (COSY, gHMBC and

HSQC) NMR spectra of descladinosyl clarithromycin and descladinosyl azithromycinwere recorded on a Bruker Avance 300 spectrometer operating at 300 MHz for 1H.The 1H spectra were processed using Gaussian window function with parametersoptimized empirically. The 13C, DEPT45, DEPT90 and DEPT135 were processed

12

using EM window function with LB value equals to 1.0 Hz. Two dimensional NMRspectra were processed using QSINE window function with other parametersoptimized empirically (e.g. SSB was equal to 2).

We used Bruker Avance 500 spectometer operating at 500 MHz for 1H for VTexperiments of azithromycin.

Deuteriated E. coli ribosomes were prepared as previously described 3

(deuteriation reduces the possibility of spin diffusion in transferred NOEexperiments). 600 MHz Spectra were run of each drug alone, in the presence of0.8 pmol dm-3 ribosomes (a concentration known to give a twofold increase inlinewidth of erythromycin A). The buffer in each case was 50 mmol dm -3 sodiumphosphate in D2O,apparent pH 7.6. The spectra were acquired at 600 MHz using adata matrix of 2048 x 1024 points and processed with zero filling in Fl. The mixingtimes were optimised empirically and were 75 ms.

Determination of minimum inhibitory concentrationsInocula for broth dilution end point determination of bacterial antimicrobialsusceptibility were prepared as follows: single colonies of anaerobic test bacteriafrom uncontaminated agar plates were inoculated into sterile, nutrient broth (10mL) contained in 25 mL sterile plastic universals and incubated in a standardaerobic incubator at 37 °C for 24 h. Cultures were then diluted to concentration of105 cfu/mL using MacFarland standard. Stock samples of clarithromycin,descladinosyl clarithromycin, azithromycin and descladinosyl azithromycin wereprepared in distilled water. Testing was performed in 96-well microtitre plates(Becton Dickinson, Franklin Lakes, NJ, USA). Diluted overnight culture (100 μL)was delivered to each test well except the first column of the plate containing 100μl of the antibiotic solution in distilled water diluted by 100 μl of double strengthbroth and the last two columns of the plate which contain 100 μl of single strengthbroth and 100 μl of destilled water, respectively. Doubling dilutions were thencarried out across the plate using a multi-channel pipette, changing the tips at eachdilution step. The plates were then incubated for 24 h in standard incubator at37°C. The MICs were determined as the lowest concentration of antimicrobial atwhich growth did not occur. Growth was detected as turbidity, relative to anuninoculated well using a microtitre plate reader (Anthos HTII; Anthos-LabtecInstruments, Salzburg, Austria). Each MIC determination was carried out intriplicate (in the same 96-well plate).

3 J. I. Gyi, R. J. Brennan, D. A. Pye and J. Barber, J. Chem. Soc.,Chem. Commun., 1991, 1471.

91

Chapter 4: Free and bound state structures of 6-O-methyl homoerythromycinsand epitope mapping of their interactions with ribosomes

4.1. DECLARATION

This chapter consists of a published article:

P. Novak, J. Barber, A. Cikos, B. Arsic, J. Plavec, G. Lazarevski, P. Tepes, N. Kosutic-

Hulita: Free and bound state structures of 6-O-methyl homoerythromycins and epitope

mapping of their interactions with ribosomes, Bioorganic and Medicinal Chemistry, 17,

5857-5867, 2009. It is reproduced unchanged with the Supplementary Material.

I carried out the molecular modelling study of the presented 6-O-methyl

homoerythromycins.

Free and bound state structures of 6-O-methyl homoerythromycinsand epitope mapping of their interactions with ribosomes

Predrag Novak a,*, Jill Barber b, Ana Cikoš c, Biljana Arsic b, Janez Plavec d, Gorjana Lazarevski c,Predrag Tepeš e, Nada Košutic-Hulita f

a Department of Chemistry, Faculty of Natural Sciences, University of Zagreb, Horvatovac 102a, HR-10000 Zagreb, Croatiab School of Pharmacy and Pharmaceutical Sciences, University of Manchester, Stopford Building, Manchester M13 9PL, UKc GSK Research Centre Zagreb Ltd, Prilaz baruna Filipovica 29, HR-10000 Zagreb, Croatiad National Institute of Chemistry, Hajdrihova 19, SI-1115 Ljubljana, Sloveniae Laboratory for Environmental Geochemistry, Faculty of Geotechnical Engineering, University of Zagreb, Hallerova aleja 7, HR-42000 Varazdin, Croatiaf PLIVA—Research and Development, Prilaz baruna Filipovica 29, HR-10000 Zagreb, Croatia

a r t i c l e i n f o

Article history:Received 24 November 2008Revised 26 June 2009Accepted 5 July 2009Available online 9 July 2009

Keywords:HomoerythromycinsFree and ribosome-bound structuresBinding epitopesNMR and molecular modeling

a b s t r a c t

The solution and solid state conformations of several 6-O-methyl homoerythromycins 1–4 were studiedusing a combination of X-ray crystallography, NMR spectroscopy and molecular modelling calculations.In the solid state 1 was found to exist as the two independent molecules with similar structures termed3-endo-folded-out. In solution a significant conformational flexibility was noticed especially in the C2 toC5 region. The compounds 1 and 2 unlike 14-membered macrolides adopted the 3-endo-folded-out con-formation while 3 and 4 existed in the classical folded-out conformation. TrNOESY and STD experimentsshowed that 1 and 2 bound to the Escherichia coli ribosome while 3 and 4, lacking the cladinose sugar, didnot exhibit binding activities, this being in accordance with biochemical data. The bound conformationswere found to be very similar to the free ones, some small differences were observed and discussed. TheSTD experiments provided evidence on binding epitopes. The structural parts of 1 and 2 in close contactwith ribosome were similar, however the degree of saturation transfer was higher for 2. The differencesbetween tr-NOE data and STD enhancements in 1 and 2 arouse as a consequence of structural changesupon binding and a closer proximity of 2 to the ribosome surface. An understanding of the molecularmechanisms involved in the interaction of macrolides with ribosomes can help in developing strategiesaiming at design of potential inhibitors.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Erythromycin and its derivative azithromycin are well-knownmacrolide antibiotics, effective against Gram-positive and certainGram-negative microorganisms.1 Macrolides exert their activityby interacting with bacterial 50S ribosomal subunit at, or close to,the peptidyl-transferase center and thus inhibit the growth of thenascent peptide chain. However, the increasing resistance to ribo-some-targeting antibiotics has become a global problem and mucheffort is now directed toward new and more potent classes of drugsto overcome resistance mechanisms. An effective approach to over-coming this problem is to understand the principles of how thesedrugs interact with the ribosome.2 Recently, crystal structures ofsome ribosome–macrolide complexes3–7 have thrown new lighton the binding mechanisms of macrolides to ribosomes and henceprovide a good basis for the rational design of new ligands and inhib-

itors. However, when analysing solid state structures of ribosome–macrolide complexes one should keep in mind the discrepancies be-tween structures obtained for the halophilic archeon Haloarculamarismortui5 and Deinococcus radiodurans.6 The proposed modelsdiffer significantly even though ribosomal 50S subunits of the twobacteria have drug binding sites whose sequences are highly con-served. Furthermore, the crystal structure data obtained so far oncomplexes of macrolides with ribosome isolated from clinicallynon-relevant bacteria do not explain all the effects of macrolideson different pathogenic strains. An example would be good here.In spite of the knowledge gained so far on macrolide antibiotics anunderstanding of the mode of their interactions with ribosome stillremain incomplete with many issues unresolved.8,9 We believe thatsteps taken in the process of drug design should also include eluci-dation of the solution state structures of free and bound ligand mol-ecules since the structural features of the complex may not beexactly the same in solution and in the solid state.7,10

In our recent papers11,12 we have shown that a systematic ap-proach which combined NMR and molecular modelling calculation

0968-0896/$ - see front matter � 2009 Elsevier Ltd. All rights reserved.doi:10.1016/j.bmc.2009.07.013

* Corresponding author. Tel.: +385 14606184; fax: +385 14606181.E-mail address: [email protected] (P. Novak).

Bioorganic & Medicinal Chemistry 17 (2009) 5857–5867

Contents lists available at ScienceDirect

Bioorganic & Medicinal Chemistry

journal homepage: www.elsevier .com/locate /bmc

http://dx.doi.org/10.1016/j.bmc.2009.07.013

mailto:[email protected]

http://www.sciencedirect.com/science/journal/09680896

http://www.elsevier.com/locate/bmc

could generally be applicable to conformational studies of macro-lides free and bound to ribosomes. Our results,11–13 as well as thoseof others,14–18 demonstrated that macrolides adopt two major con-formational families; folded-out and folded-in, referring to the out-ward and inward folding of the ring fragment C2–C5. The vicinalcoupling constants 3JH2H3 and NOE proton–proton contacts such asH3–H11 and H4–H11 are good indicators of the aglycone folding.12

Furthermore, the 3JCH coupling constants over the glycosidicbonds12,18 might provide information about the position and mobil-ity of the two sugars, (desosamine and cladinose), with respect to thelactone ring. Longitudinal relaxation of methyl protons could be use-ful to probe motions of methyl groups which reflect the aglyconering folding.14,16 Additionally, by applying transferred NOESY (tr-NOESY) and saturation transfer difference (STD) NMR experimentsit is possible to characterize binding of macrolides to ribosomes.11

In this paper we report on conformational analyses of several15-membered azalides (Fig. 1), for example, 6-O-methyl-9a-aza-9a homoerythromycin A (1), 6-O-methyl-8a-aza-8a homoerythro-mycin A (2) and their decladinosyl derivatives (3) and (4). Com-pounds 1 and 2 are particularly interesting to see whetherdifference in the position of the lactam group (Fig. 1) has an impacton the overall conformation of the molecules and how this affectsthe affinity for the ribosome. Binding studies of decladinosyl deriv-atives 3 and 4 are expected to clarify the observed low activity of15-membered azalides without the cladinose sugar attached at po-sition 3.

We now report the use of a combination of NMR measurements(coupling constants, NOE and relaxation data) and molecular mod-elling calculations to determine the solution state structures of 1–4in polar (water) and non-polar (chloroform) media. Variable tem-perature measurements were also performed to check the confor-

mational stability of the studied molecules. The obtained solutionstate conformations of 1 were compared to the solid state confor-mation as determined by X-ray crystallography. Transferred-NOESY and STD (Saturation Transfer Difference) experiments werenext employed to asses bound conformations and to characterizeinteractions of 1–4 with Escherichia coli ribosomes. Tr-NOESYexperiments provide data on the bound conformations of the stud-ied compunds.19 STD is a promising tool for identification of thebinding epitopes of drug–receptor interactions,10,20,21 informationwhich can further be exploited in the design of molecules with bio-activity. The obtained results have been compared with thosefound for the related macrolide–ribosome complexes, both in solu-tion11 and in the solid state.3–7

2. Results and discussion

2.1. NMR assignments

The syntheses and chemical shifts of 1–4 in CDCl3 were de-scribed elsewhere.22 The complete 1H and 13C atom assignmentsin D2O (see Supplementary data) were made by the combineduse of the standard one-dimensional (1H and APT sequences) andtwo-dimensional homo- and heteronuclear (COSY, HSQC andHMBC sequences) experiments. The homo-nuclear coupling con-stants were determined from the high-resolution proton and DQF-COSY spectra (Table 1). Vicinal hetero-nuclear coupling constantsover the glycosidic bond (Table 2) were extracted from the mul-ti-site 13C selective experiments followed by polarisation transferto protons, employing the Hadamard formalism.23,24 NOE andROE data were obtained from the two-dimensional NOESY andROESY spectra (Table 3).

N

O O O

OO

N

O

O

O OO

O

O

O

H

N

O O O

O

N

O

O

O OOO

H

ON

O O O

OO

N

O

O OO

O

O

O

H

ON

O O O

O

N

O

O OOO

H

12

3

45

67

8

9

10

11

12

1314

15

9a

1''

2"3"

4"

5"

1'

2'3'

4'

5'

1

12

3

45

67

8

9

10

11

12

1314

15

8a

1''

2"3"

4"

5"

1'

2'3'

4'

5'

12

3

45

67

8

9

10

11

12

1314

15

9a

1'

2'3'

4'

5'

2

3

12

3

45

67

8

9

10

11

12

1314

15

8a

1'

2'3'

4'

5'

4

Figure 1. Compounds studied and the atom numbering.

5858 P. Novak et al. / Bioorg. Med. Chem. 17 (2009) 5857–5867

2.2. Crystal structure

X-ray structure analysis has shown that in the solid state 1 ex-ists as the two crystallographically independent molecules, 1a and1b with very similar structures (Fig. 2). Each molecule crystallizeswith acetone and seven water molecules in an asymmetrical unit.A complex network of hydrogen-bonds involving water moleculesand 1a and 1b is found, while the acetone molecule is placed in thecavities formed.

Table 4 gives key internuclear distances and torsion angles for1a and 1b, compared with the crystal structure of azithromy-cin.25,26 Like compound 1, azithromycin also forms two very simi-lar structures in the crystalline state.

In both 1a and 1b structures the macrocyclic ring of 1 adopts aconformation intermediate between the classical folded-out andfolded-in structures, and similar to the azithromycin solutionstructure. While the Western portion of the ring is folded-out, asreflected by the H8–H11 and H4–H6Me internuclear distances(these are both <4 Å in folded-in conformers), the H3–H11 internu-clear distance and the H3–H4 torsion angle are more characteristicof folded-in structures. As we have shown previously,17 there isconsiderable torsional freedom in the H2–H5 part of 15-memberedmacrocyclic rings. On balance therefore, these are folded-out struc-tures, but the term 3-endo-folded-out could be used to define thestructures more precisely.

The 6OMe group is oriented toward atom C3 in both 1a and 1b.The oxygen atom in the lactam group is oriented toward the mac-rocyclic ring due to the rotation around the C8–C9 bond, with thetorsion angle C7–C8–C9–N9 of 135.0� and 134.5� for 1a and 1b,respectively. This orientation prevents the formation of the intra-molecular hydrogen-bond N9–H9� � �O6. A slight difference be-tween the two molecules is noticed in the region C13–C2 of themacrocyclic ring with the torsion angles O–C1–C2–C3 and C1–O–C13–C12 of 149.2� and 106.7� for 1a and 132.9� and 122.5� for 1b.

The sugar rings adopt chair conformations and their orientationwith respect to the macrocycle is defined with the torsion anglesC2–C3–O3–C100, C3–O3–C100–O600, C4–C5–O5–C10 and C5–O5–C10–O60 of �90.4, �76.4, �102.3 and �78.3 in 1a and �85.3, �76.0,�99.6 and �77.3 in 1b, respectively. The cladinose is almost paral-lel while desosamine is perpendicular toward the macrocyclic ring.A slight difference in the orientation of the desosamine sugar in 1aand 1b is also indicated (Fig. 2).

2.3. Solution state structure

2.3.1. Macrocyclic ringTo determine the conformational characteristics of the 15-

membered macrocyclic ring and two sugar units in solution ofcompounds 1–4, we measured vicinal proton–proton 3JH,H couplingconstants in polar and non-polar solvents (Table 1) as well as car-bon–proton 3JC,H coupling constants around the glycosidic bonds in1 and 2 (Table 2). As stated previously11–13 only the 3JH2,H3 couplingconstant changed significantly with changing the solvent polarityand raising the temperature.

Table 1 shows vicinal coupling constants obtained for 1–4. Thevalues of 3JH2,H3 are immediately striking. Compounds 3 and 4 ex-hibit large values, suggesting torsion angles of close to 180�; thesevalues are consistent with the classical folded-out conformation.For compounds 1 and 2, however, 3JH2,H3 values in both waterand CDCl3 are much smaller, suggesting average torsion angles ofonly just above 90�. These values are consistent with that obtainedin the crystal structure of 1 as shown in Table 4, suggesting that 1and 2 may form 3-endo-folded-out structures, as described forazithromycin.

Further conformational information about the macrocyclic ringcan be obtained by NOE or ROE experiments, and these were nowused to gain a more complete understanding of the conformationalequilibria shown by compounds 1–4. As already observed for therelated macrolides,11,12,17 NOE’s are close to zero or change signin D2O, and therefore we recorded ROESY spectra of 1–4 in buf-fered solution (as displayed in Fig. 3a for 2) and NOESY spectrain CDCl3. To suppress the unwanted TOCSY–COSY cross-peaks theT-ROESY sequence28 was also employed. The observed ROE con-tacts of 1–4 are displayed in Table 3. The H3–H11 and H4–H11ROE contacts were both observed in buffered solutions of 1 and 2(Table 3), consistent with the 3-endo-folded-out conformation,but also, of course, consistent with folded-in conformers. The ab-sence of H3–H8 and H8–H11 contacts and the presence of H5–H6Me contacts strongly indicate that 1 and 2 adopt 3-endo-

Table 1Vicinal proton–proton coupling constants 3JHH/Hz of compounds 1–4 in CDCl3 and D2O

Protons 1 2 3 4

CDCl3 D2O CDCl3 D2O CDCl3 D2O CDCl3 D2O

H2,H3 5.8 3.6 5.3 4.3 10.1 10.7 10.3 10.7H3,H4 1.6 1.6 1.2 1.5 Unres. 0.9 Unres. Unres.H4,H5 6.8 7.3 7.9 7.3 Unres. 1.5 1.1 1.2H7a,H8 8.2 10.8 9.2 9.3 5.4 7.0 8.1 8.9H7b,H8 Unres. Unres. 2.1 Unres. 2.1 1.5 Unres. Unres.H10,H11 Unres. Unres. 1.4 1.5 1.7 1.8 1.5 Unres.H13,H14a 1.7 2.2 2.6 2.4 1.6 1.8 2.2 1.1H13,H14b 10.7 10.6 9.9 10.4 11.0 11.0 10.6 11.0H10 ,H20 7.2 7.9 7.3 7.6 7.3 7.8 7.3 6.1H20 ,H30 10.1 10.7 10.7 10.7 10.3 10.5 10.3 10.4H30 ,H40a 3.8 3.8 Overlap 4.4 3.9 4.9 3.8 4.6H30 ,H40b 12.4 12.6 12.4 12.2 12.3 12.1 12.3 OverlapH40a,H50 1.7 2.0 2.1 2.0 1.9 1.8 1.8 1.5H40b,H50 10.7 11.0 10.8 11.3 10.8 Overlap 11 OverlapH100 ,H200a 0.7 Unres. Unres. Unres. — — —H100 ,H200b 4.4 4.9 4.5 4.6 — — —H400 ,H500 9.8 9.8 9.3 9.7 — — —

Table 2Comparison of the vicinal carbon–proton coupling constants 3JC,H/Hz over theglycosidic bond obtained from the NMR experiment for compounds 1, 2 andazithromycin11 with those calculated from the X-ray for 1

Atoms 1 2 Azithromycin 1

3JC,H measured U 3JC,H calcd

H3,C100 4.6 4.8 2.2 31.2 5.9C3,H100 4.6 4.9 4.1 41.2 4.5H5,C10 6.6 5.8 6.7 14.4 7.7C5,H10 3.7 3.3 3.5 41.8 4.4

P. Novak et al. / Bioorg. Med. Chem. 17 (2009) 5857–5867 5859

folded-out conformations, however in 1 the H3–H11 cross-peakwas stronger than the 4H–11H cross-peak while the opposite

was found in 2, where H4–H11 cross-peak was stronger thanH3–H11 (Fig. 3a). Compound 1 also exhibits a weak H4–H6Mecross-peak. This signal is characteristic of the folded-in conforma-tion and suggests that 1 has a greater folded-in character than 2.Compounds 3 and 4 both exhibited strong NOEs between protons4 and 11, while no H3–H11 contact was observed in the ROESYspectra. This, together with coupling constant data, strongly sup-ports the conclusion that these compounds adopt the classicalfolded-out conformation in both D2O and CDCl3. The absence ofthe cladinose sugar at position 3 enabled the macrocyclic ring tobe more flexible and to adopt the energetically more favouredfolded-out conformation.

2.3.2. Sugar ringsThe large diaxial coupling constant in cladinose and desosamine

(Table 1) reflected the chair conformations14 of both sugar rings in1 and 2 and desosamine in 3 and 4. The observed intra-sugar ROE’s(Table 3) are in agreement with this.

Table 3ROE contacts of compounds 1–4 in D2O

Atom ROE

1 2 3 4

2 4Me,4,3,11 4Me,4,3,11 4Me,4 4Me,42Me 4Me,3 4Me,3,100 3 33 4Me,2Me,4,2,6OMe,11,5,100 4Me,2Me,4,2,6OMe,11,5,100 2,2Me,4,5,6OMe 2Me,4,5,6OMe4 7a,2,11,5,3 7a,2,11,5,3 3,5,7a,11 2,3,5,7a,114Me 2Me,2,300OMe,5,10 2Me,2,300OMe 20 ,5,10 2 2,5,7a,115 4Me,6Me,500Me,4,6OMe,3,500 ,10 4Me,6Me,500Me,4,6OMe,3,500 ,10 3,4,6Me,6OMe 3,4,6Me,6OMe6Me 4,8,6OMe,5 8,6OMe,5 5,6OMe,8 6OMe,7b,86OMe 6Me,8,11,5,3,500Me, 6Me,8,11,5,3,500Me, 3,5,6Me 3,4,5,6Me7a 8Me,7b,4 8Me, 7b, 11,4 4,4Me,8,8Me 4,4Me,57b 7a,8,8Me, 7a,8,8Me 6OMe, 8Me 6Me,7a,88 7b,6Me,6OMe,8Me 7b,6Me,6OMe,8Me 6Me,6OMe,7a,7b 6Me,7a,7b,8Me8Me 8,7b,7a 8,7b,7a 7b 6OMe,7a,7b,810 10Me,12Me,11 10Me,12Me,11 10Me,11,12Me 11,12Me10Me 11,10 11,10 10,11 —11 10Me,12Me,4,2,3,10, 6OMe,13 10Me,12Me,4,2,3,10, 6OMe,13 4,4Me,10,10Me,12 Me,13 4,10,12Me,1312Me 14b,14a,11,10 14b,14a,11,10,2 11,14a,14b 10,11,14a,14b13 15,14a,14b,12Me,11 15,14a,12Me,11 — 11,12Me,14a,14b,1514a 12Me,13,14b 12Me,13,14b 13 13,14b14b 12Me,14a,13 12Me,14a,13 12Me,15 12Me,14a15 14b, 13,2Me 13,14b 13,14b10 4Me,30 ,300OMe,20 ,5,50 ,500 4Me,30 ,300OMe,20 ,5,50 ,500 — 4Me,5,6Me,20 ,30 ,50

20 40b,30NMe2 40b,30NMe2,50 40b,30NMe2,30 40b,30NMe2

30 30NMe2,50 ,10 ,40a 30NMe2,50 ,1,40a 20 ,30NMe2,50 ,40a 30NMe2,50 ,40a,40b30NMe2 40b,40a,30 ,20 40b,40a,30 ,20 40b,40a,30 ,20 40b,40a,30 ,20

40a 40b,30NMe2,50 ,50Me,30 40b,30NMe2,50 ,50Me,30 40b,30NMe2,50 ,30 40b,50 ,30

40b 40a,30NMe2,20 40a,30NMe2,2’ 40a,30NMe2,20 ,50Me 40a,20

50 5’Me,40b,40a,30 ,500 ,10 5’Me,40a,30 ,500 ,10 5’Me,40a,30 40a,30

50Me 40a,50 40a,40b,50 40a,50 40a,50

10 0 300Me,200b,200a,400 ,3 300Me,200b,200a,400 ,3,30 0OMe — —20 0a 300Me,200b,300OMe 300Me,200b,300OMe — —2‘‘b 300Me,200a,400 ,100 300Me,200a,400 ,100 — —300 Me 200a,200b,400 ,300OMe,100 200a,200b,400 ,500 , 300OMe,100 — —300OMe 200a,300Me,4Me,500 200a,300Me,4Me,500 — —400 300Me,500Me,200b,500 300Me,500Me,200b,500 — —500 500Me,400 ,5,50 ,10 500Me,400 ,300OMe,5,50 ,10 — —500Me 400 ,500 ,6OMe 400 ,500 ,6OMe, — —

Table 4Key torsion angles and internuclear distances for two crystallographically independent molecules 1a and 1b compared with the crystal structure of azithromycin (Azi a and Azi b;data taken from Refs .25,26 is very similar)

Torsion angles (�) Internuclear distances (Å)

H2,H3 H3,H4 H4,H5 H7s,H8 H3,H8 H3,H11 H4,H11 H4,H6Me H5,H6Me H8,H11

1a 112.2 �73.9 144.4 161.4 5.0 3.4 3.5 4.3 2.4 4.31b 113.8 �70.5 141.3 160.1 5.1 3.5 3.4 4.3 2.4 4.2Azi a 111.6 �72.5 141.3 155.6 5.5 3.2 3.5 4.0 2.6 4.8Azi b 112.3 �67.2 139.5 154.0 5.5 3.1 3.2 4.0 2.6 4.8

Figure 2. The superposition of the conformations of the two crystallographicallyindependent molecules 1a and 1b.


The vicinal carbon–proton coupling constant over the glycosidicbond could be useful to determine the relative position of the sug-ars with respect to the macrocyclic ring.12,18 Hence, these couplingconstants for 1 and 2 were extracted from the multi-site 13C selec-tive one-dimensional experiments using Hadamard formalism.23,24

The four simultaneous experiments were used to excite C3, C5, C10

and C100 carbon atoms which afforded the determination of 3JCH val-ues for 1 and 2 listed in Table 3. 3JCH coupling constant over the gly-cosidic bond are similar to those found for azithromycin and relatedmacrolides,12 indicating a similar orientation of the sugars.

The H10–H5 and H10–H4Me ROE contacts were observed for allcompounds, which corresponded to the perpendicular orientationof the desosamine sugar with respect to the macrocycle, being con-

sistent with the X-ray structure of 1 and the data observed for therelated macrolides.12,14–18 On the other hand, a close proximity ofH10 and H500 and H5–H500 atoms indicated approximately parallelorientation of the cladinose in compounds 1 and 2 (Table 3) as alsoobserved in the solid state. Inter-sugar contacts H10–H50 and H10–H300OMe suggest the up–up orientation of the alpha-faces of thetwo sugars in 1 and 2.

2.4. The motional properties of methyl groups

Longitudinal relaxation time measurements (T1) of methyl pro-tons can provide further information about mobility of methylgroups and aglycone ring folding.14,16 T1 values were measured in

Figure 3. Comparison of (a) the ROESY spectrum and (b) tr-NOESY spectrum of compound 2 after addition of E. coli ribosome, both recorded in TRIS buffer at 25 �C (importantNOE peaks are indicated in red).


CDCl3 solution and are given in Table 5. In the 14-membered mac-rolides, the folded-out and folded-in conformations differ by theposition of 4Me and 2Me groups. The folded-out is characterizedby a high energy barrier to the rotation of 2Me resulting in a shortproton longitudinal relaxation times.14,16 In the folded-in confor-mation 2Me group has more motional freedom due to lesseningof the restriction to rotation, while the mobility of 4Me is reducedsince 4Me is positioned in the more congested environment.

For azithromycin, shorter T1 values for 4Me than 2Me were ob-served16, and this is also true of 1 and 2, see Table 5. This would beconsistent with a folded-in conformation. However, as noted previ-ously, 15-membered macrolides show conformational flexibilityaround H3–C3–C4–H4 and this will result in increased freedomfor 4Me. This result therefore is consistent with a 3-endo-folded-out conformation. In the decladinosyl compounds 3 and 4 both2Me and 4Me rotate freely since there is no steric hindrance im-posed by the cladinose, which in turn results in a longer T1 values(Table 5).

2.5. Molecular modelling

Detailed conformational structures may be constructed fromNMR studies combined with molecular modelling in two ways.Either we may carry out unconstrained modelling and then searchfor low energy conformations that satisfy the NMR data, or we mayconstrain the modelling calculations to the NMR data. The dangerwith the first method is that small errors in the assumptions madeby the modelling package may lead to incorrect conformationsbeing proposed. The danger with the second method is that aver-aging of multiple conformations is difficult to recognise.

We carried out unconstrained Monte Carlo searches of both 1and 2 using both AMBER and MM2 force fields. For 1 the AMBERforce field 10 lowest energy structures were characterized by anorientation of the Western half of the macrolides ring such that11H is exo to the ring, giving rise to large distances between 3Hand 11H and also 8H and 11H (see Supplementary data). The samepattern was seen using MM2 as the force field, the global minimumbeing almost identical. Good convergence was seen and no struc-ture within 20 kJ mol�1 of the global minimum showed a conven-tional 11-endo conformation.

A similar picture was seen for compound 2, except that conven-tional folded-out structures were now found 18 kJ mol�1 above theglobal minimum when the AMBER force-field was used. Lower en-ergy structures were all variations of the 11-exo-folded-out confor-mation. The MM2 force-field also found an 11-exo structure withexcellent convergence, but it was not identical with the globalminimum found using the AMBER force-field. Representative 11-exo-folded-out structures are shown in Figure 4, alongside the con-ventional 11-endo conformations.

We investigated the inconsistency between modelling and solu-tion structures in several ways. Firstly, we repeated the modellingstudies, this time for compounds 1–4, using molecular dynamicssimulations. Initially a range of force-fields and gradient functions

were explored using compound 1, and the Tripos force-field andPowell method were found to give the best (lowest energy) results.The four structures were then heated and cooled in silico, varyingthe rates of heating and cooling and the highest and lowest tem-peratures. The energies of the resulting structures were then com-pared. The results are shown in the Supplementary data.

The lowest energy conformer found for compound 2 was onceagain an 11-exo conformation, as was the second conformer. Thethird conformation in energy was, however, an 11-endo conforma-tion and was only about 4 kcal mol�1 higher in energy. Compounds3 and 4 behaved almost identically with an 11-endo conformationenergetically higher by about 4 kcal mol�1 than the 11-exo confor-mation. For compound 1 an 11-endo-folded-in conformation wasfound at lowest energy.

Although it is well known that it is very difficult to representaqueous solution accurately in silico, the persistence of the 11-exo conformers led us to wonder if they had any basis in reality.Analysis of the 11-exo conformers in silico, shows that these con-formers are expected to give rise to strong H11–H12Me, H11–H10Me and H7–H20 cross-peaks. For none of the compounds 1–4do we observe a H7–H20 cross-peak, although this could be weakbecause of the multiplicity of the H20 signal. H11–H10Me is a weaksignal in both 1 and 2. Only H11–H12Me is observed clearly, andthis is expected also in 11-endo conformers. Therefore, there isno clear evidence of 11-exo conformers in 1 or 2.

Finally, constrained Monte Carlo searches were carried out onboth 1 and 2. Key distances dictated by the trNOESY and ROESYspectra were constrained in order to maintain 11-endo-conforma-tions with ‘up–up’ arrangements of the sugars. Using the crystalstructure of 1 as the starting point, and deriving a structure for 2from 1, the distances H11–H4, H11–H3, H3–H100, H5–H500, H10–H500, H10–H17, H10–H800, H50–H500 were constrained to within0.5 Å of the starting structure for distances below 3 Å, or to within1 Å for distances in the range 3–4 Å. The results are summarized inthe Supplementary data.

Even after constraining the Monte Carlo search there is pooragreement between the results obtained for different force fields,and the agreement between experiment (NMR) and modelling isnot complete. While compound 2 adopts a 3-endo-folded-outstructure in silico, the ROESY spectrum of this compound showsweak H4–H18 and H3–H11 signals, suggesting trace amounts ofthe folded-in conformer are also present. For compound 1, a 3-endo-folded-out structure is predicted by the AMBER force-field,but the MM2 force-field prefers a folded-in structure (albeit onein which 8H, constrained by the amide function does not makeclose contact with 3H or 11H). Again, the NMR data support thedominance of the 3-endo-folded-out conformer, but with somefolded-in character also present.

Molecular modelling cannot therefore define the structures ofthese compounds. Rather, it reveals the extreme conformationalflexibility of 1 and 2. Both folded-out and folded-in conformationsare available to them, although they prefer 3H to be endo to themacrocycle and 8H to be exo, resulting in a predominance of the3-endo-folded-out conformation. 11-exo-conformers are also ex-pected, although they are not definitely detectable by NMR sug-gesting that they are present at very low concentrations.

2.6. Binding studies to ribosomes

The compounds 1–4 were further subjected to comparativeNOESY and STD experiments in the presence of E. coli ribosomesto study bound conformations and to determine structural partsin intimate contact with ribosome. As already mentioned, NOE’sin D2O solution of 1–4 were close to zero at 600 MHz, whileROE’s were all positive. Upon the addition of E. coli ribosomescompounds 1 and 2 displayed transferred NOE’s as indicated by

Table 5Longitudinal relaxation times, T1/s for aglycone methyl protons in 1–4

Compound 1 2 3 4Methyl

2 0.422 0.478 0.571 0.4784 0.373 0.377 0.675 0.5476 0.309 0.366 0.434 0.3216OMe 0.590 0.658 0.798 0.6638 0.395 0.380 0.510 Overlap10 0.410 0.590 0.483 0.49012 0.475 0.430 0.576 0.43015 0.680 0.944 0.789 0.663


a negative sign of the cross-peaks (Fig. 3b), corroborating bindingto ribosome. Both compounds gave STD spectra as well (Fig. 5)and thus confirmed the binding activity indicated by trNOESYand allowed for the binding epitopes to be determined. On theother hand compounds 3 and 4 exhibited no trNOE signals anddisplayed very weak STD intensities. This result is consistentwith biochemical data which showed that 3 and 4 were inactiveagainst E. coli, while 1 and 2 exhibited significant activity (seeTable 6).

The features of the trNOESY spectrum of 1 indicate that thebound conformation is very similar to the free one. Namely, allthe cross-peaks found in the ROESY spectrum were also found inthe trNOESY except for the very weak contact 100–500Me. Comparedto the free state the compound 2 exhibited several new but alsovery weak NOE contacts in the bound conformation such as thosebetween protons H7b and H4 in the lactone ring, H50Me andH30NMe in the desosamine and between H200b and H300OMe inthe cladinose. There were also several contacts observed in the

Figure 4. Structures of 1 and 2 found by Monte Carlo conformational searches. Global minima of constrained searches using the AMBER force-field, showing 3-endo-folded-out conformers of (a) compound 1 and (b) compound 2. Global minima of constrained searches using the MM2 force-field, showing (c) folded-in conformation of compound 1and (d) 3-endo-folded-out conformer of compound 2. Global minima of unconstrained searches using the AMBER force-field showing (e) 11-exo-folded-out conformer ofcompound 1 and (f) 11-exo-folded-in conformer of compound 2.


ROESY spectrum which were absent in the trNOESY, such as H13–H2, H100–H500 and H100–H400. These differences point toward thelessening of the conformational flexibility of the ethyl side-chainand sugars as a consequence of motional restrictions imposed bythe close proximity of this structural moieties of 2 to the ribosomesurface. Slight changes in the Eastern part of the macrocyclic re-gion, indicated by the presence of H7b–H4 contact is in agreementwith this. Hence, trNOESY experiments strongly suggest that freeand bound conformations are almost identical.

Subsequently, we performed STD experiments to identify theregions of the studied compounds in close contact to ribosome.As already mentioned, only compounds 1 and 2 exhibited signifi-cant STD enhancements being stronger in the latter than in the for-mer. The STD enhancements are shown in Figure 6. A saturationtransfer in 1 was the most efficient for methyl protons (15Me) inthe ethyl side-chain. Appreciable STD enhancements were also ob-served for dimethylamino protons, and 300Me and 300OMe protonsof the cladinose. The 2Me and 6OMe, protons were saturated toobut to a smaller extent. Similarly, the compound 2 exhibited stron-gest enhancements for methyl protons in the ethyl side-chain too.300Me, 300OMe, 8Me and 2Me groups were moderately saturated.The methyl protons of the dimethylamino group were also en-hanced by the saturation transfer. A comparison between STD ef-fects in 1 and 2 reveals common regions of the two moleculesthat are in intimate contact with ribosome. These include methylgroup at position 15, 300OMe and 300Me groups of cladinose anddimethylamino group of desosamine. Therefore compounds 1

and 2 have the same structural parts that exhibited the largest sat-uration transfer indicating similar binding modes. The only differ-ence is the noticeably higher degree of saturation of 10Me group in2 suggesting that this part is buried deeper into the ribosome pos-sibly due to H-bond interactions involving the oxygen atom of thelactam group. In azithromycin where no oxygen is present at thisposition the STD was found to be much smaller (Fig. 8). This findingis in agreement with tr-NOE data.

In order to differentiate between specific and non-specific inter-actions we have performed competition STD experiments withazithromycin. Namely, the ribosome is very large particle whichmay offer many non-specific binding sites for molecules 1 and 2.Azithromycin is known macrolide antibiotics whose binding hasbeen well characterized by X-ray crystallography,3–6 NMR spec-troscopy11 and by kinetic studies.27

To obtain the best experimental conditions pulse shapes, satura-tion times, on and off-resonance frequencies and other NMR param-eters were varied and tested. After the acquisition of the STD-NMRspectrum of compound 2, we acquired the STD-NMR spectrum ofthe mixture of azithromycin and compound 2 under the same exper-imental conditions (see Section 4). Figure 7 displays the STD-NMRsignal of the methyl group 15 in compound 2 (green) overlappedwith the corresponding signal in azithromycin (red), as well as bothsignals in the mixture (blue). It is clearly seen that the intensities ofthe STD signals in the mixture are much smaller compared to theintensities of STD signals of the individual compounds. This couldonly be possible if the two compounds compete for the same bindingsite on the ribosome. The fact that both signals are observed alsopoints toward similar binding constants, both in nano molar region.

Figure 8 shows STD enhancements observed for azithromycinand distances between azithromycin and ribosome as determinedfor X-ray co-crystallized structure of azithromycin and H. maris-mortui.5 which are in a good correlation. It is seen that those partsof azithromycin found closest to the ribosome in the co-crystal-lized structure exhibit the highest magnetization transfer in thecomplex in solution. A comparison between STD effects in azithro-mycin and 1 and 2 reveals three common regions in closer proxim-ity to the ribosomal surface, desosamine and cladinose sugars(300Me and 300OMe) and methyl group at position 15. This is inagreement with the crystal structure of azithromycin–ribosomecomplex.5 Hence, it is likely that macrolides 1 and 2 interact withribosome in a similar fashion as azithromycin.

The binding epitopes of 1 and 2, as determined by STD, weremapped onto the calculated lowest energy 3-endo-folded-out con-formations and are displayed in Figure 9.

3. Conclusion

These results provide further evidence that the 15-memberedlactam macrolides have more conformational flexibility than 14-membered macrolides, especially in C2–C5 portion of the mole-cule. When cladinose is absent they adopt folded-out conforma-tions, but in the presence of cladinose they prefer anintermediate conformation (termed 3-endo-folded-out) unavail-able to the 14-membered macrolides. The presence of two formsof this structure even in the crystal of 1 reflects the conformationalflexibility of these compounds.

The bound conformations of 1 and 2 are very similar to those ob-served in the free state which is in agreement with the results ob-tained by the crystallography.3–7 Three common regions closest tothe ribosome were found, desosamine, cladinose and methyl groupat position 15. The absence of cladinose sugar in 3 and 4 is the maincause of their inability to bind to the ribosome, as indicated bytrNOESY and STD experiments. Azithromycin has some unexpectedtherapeutic properties, for example, it is a promising anti-malarial

Figure 5. (a) STD-NMR spectrum of 2 prior to addition of the E. coli ribosome, (b)STD-NMR spectrum of 2 after addition of the E. coli ribosome and (c) protonspectrum of compound 2 after addition of the E. coli ribosome, (recorded in TRISbuffer at 25 �C).

Table 6IC50/lM values for compounds 1–4

Compound IC50 (lM)

1 1.32 0.13 >1004 >100Azithromycin 0.3


lead compound,29 and conformational flexibility, also enjoyed bycompounds 1 and 2, is possibly a contributor to these properties.

4. Experimental

4.1. Sample preparation

Fully deuteriated and non-deuteriated ribosomes were pur-chased from Dr. Kalju Vanatalu, CDN Ltd, Leete Str. 13, 11313 Tal-lin, Estonia. Ribosomes were isolated from E. coli MRE 600 strain bythe zonal ultracentrifuge technique. Deuteriated ribosomes wereused for trNOESY experiments, while non-deuteriated were usedfor STD measurements.

Compounds 1–4 were dissolved in buffered D2O (20 mM TRIS-d11, 60 mM KCl) at pD = 7.4 so that their concentration remainsthe same for all four compounds (0.003 M). The pH meter readingwas adjusted using 0.1 M DCl and 0.1 M NaOD at 0.4 units less thanthe required value in order to take into account the change in theglass electrode potential because of D2O.

Ribosomes were added into the solution of 1–4 until the finalconcentration of 8.3 � 10�6 M of ribosome was achieved. The STD

and trNOESY experiments were acquired with the total volume of0.6 ml and the ribosome to macrolide ratio of approximately 1:3500.

4.2. NMR spectroscopy

One and two-dimensional NMR spectra were recorded at 25 �Con Bruker Avance DRX500 and Bruker Avance III 600 spectrome-ters equipped with 5 mm diameter inverse detection probe withz-gradient accessory. In 1H NMR experiments the spectral widthwas 10,000 Hz, the number of data points 65 K and the numberof scans 8–64. TMS was used as the internal standard. The sampleconcentration was 10 mg mL�1 in CDCl3, acetone-d6 and DMSO-d6

solutions and 2 mg mL�1 in D2O and TRIS buffer. The excitationsculpting scheme was used for the water signal suppression. Thedigital resolution was 0.1 Hz per point.

Figure 6. STD enhancements for compounds 1 and 2 (* peak overlap).

Figure 8. Comparison of STD signal enhancements and distances from ribosometaken from X-ray structure of co-crystallized azithromycin with H. Marismortui(* peak overlap).

Figure 7. Comparison of STD signal enhancements of atom H15: (a) azithromycin,(b) compound 2 and (c) mixture of azithromycin and compound 2 (all three STD-NMR spectra recorded in presence of the E. coli ribosome in TRIS buffer at 25 �C.


Two-dimensional gCOSY, ROESY and NOESYspectra were re-corded under the following conditions: spectral width was6000 Hz in both dimensions, 2 K data points were applied in timedomain and 512 increments were collected for each data set withlinear prediction to 1 K and zero filling to 2 K. The number of scansper increment varied between 4 and 32. A relaxation delay was1.5 s. States-TPPI ROESY spectra were obtained with the mixingtime of 250 ms (400 ms for NOESY) and processed with sinesquared function shifted by p/2 in both domains, while gsCOSYspectra were processed with unshifted sine function. The digitalresolution was 2.7 and 10.7 Hz per point in f2 and f1 domains,respectively.

The HSQC and HMBC spectra were recorded with a relaxationdelay of 1.5 s and 32 scans per increment. The spectral widthwas 31,000 Hz in acquisition domain f2 and 7500 Hz in time do-main f1. Data were collected into a 2048 � 512 acquisition matrixand processed using a 2 K � 1 K transformed matrix with zero fill-ing in f1 domain. Sine multiplication was performed before Fouriertransformations. In HMBC spectra the delay for long-range cou-plings was set to 60 ms.

Measurements of long-range 13C–1H couplings were performedusing the multiple 13C site selective excitation experiment on aVarian Unity Inova 600 spectrometer operating at 600.07 MHzfor protons with an inverse detection gradient probe. Half-Gauss-ian shaped pulse truncated at the leading edge at 5% maximumintensity was used for selective 13C excitation with a duration of25–50 ms. The relative sign of each selective pulse was set accord-ing to the Hadamard matrix. Spectral widths of 4000–5000 Hzwere sampled with 32768 data points using 5000 scans for eachHadamard excitation. A BIRD module was used for a better selec-tion of long-range couplings. The two gradients were applied be-fore and after the last pair of 90� pulses for coherence selectionand suppression of artefacts.

1D STD-NMR spectra were collected with a 32k data points anda spectral width of 16 ppm by using a sequence provided by Bru-ker. Selective saturation of the E. coli ribosomes was performedusing several shaped pulses with varying duration and saturationtimes to reach the optimum experimental conditions. The Gauss

cascade pulse gave the highest sensitivity and selectivity. The sat-uration time was 3 s. The saturated and reference spectra were ac-quired and processed simultaneously by creating a pseudo 2Dexperiment. The saturation frequency was switched from on-reso-nance (8 ppm) to off-resonance (50 ppm) after each scan. The exci-tation sculpting scheme was used to suppress the HDO signal.

The trNOESY spectra were acquired using the pulse sequencefor phase sensitive 2D homonuclear correlation (noesygptp19) pro-vided by Bruker software. Data matrix of 2 K � 512 complex pointsand 8 scans with sweep width of 5000 Hz was used. Mixing timewas 150 ms. Data were zero filled in f1 to 1 K and transformedusing sine squared function, giving the digital resolution of2.44 Hz in f2 and 4.88 Hz in f1 dimension. The excitation sculptingand WATERGATE schemes were used to suppress the HDO signal.

T1 proton relaxation time measurements were performed on aBruker Avance DRX500 spectrometer at 500 MHz using the inver-sion recovery sequence and analyzed by the three parameter fitprocedure. 10 s values in the range 15 ms–4 s were used and thedelay was set to five times the longest T1 value for each compound.The solvent was CDCl3.

4.3. X-ray crystallography

The details of crystal data collection and refinement parametersfor 1 are listed in Table 8. Single crystals of the 1 were grown byevaporation from acetone/water solution (1:1, v/v). A colorlessneedle with dimension 0.6 � 0.2 � 0.2 mm was used for X-raymeasurements. Data were collected at 170 K with a Kappa-CCDNonius diffractometer using graphite monochromated Mo Ka radi-ation. Data reduction was carried out using DENZO and SCALE-PACK30 software.

The structure was solved by direct methods implemented inSHELXS9731 and refined by a full-matrix least-squares method basedon F2 using SHELXL97.31 Hydrogen atoms were positioned geometri-cally at calculated positions and allowed to ride on their parentatoms. Hydrogen atoms belonging to water molecules are foundin different Fourier map and refined isotropically. The refinementconverged to R = 4.25%, and wR = 9.01%. The maximum and mini-

Figure 9. STD binding epitopes of 1 and 2. (a) STD binding epitope of 1 determined by molecular modelling using AMBER force field and NMR, (b) STD binding epitope of 2determined by molecular modelling using AMBER force field and NMR and (c) STD binding epitope of 2 determined by molecular modelling using MM2 force field and NMR.


mum residual electron densities were 0.02 and �0.01 eÅ�3,respectively.

4.4. Molecular modelling

Compounds 1, 2, 3 and 4 were constructed from the crystalstructure of compound 1 using Macromodel software.32 The struc-tures were minimized using the Truncated Newton Conjugate Gra-dient (TNCG) method33 in order to obtain local minima.

The Monte Carlo Multiple Minimum (MCMM) conformationalsearch was used to find the global minima.34 The TNCG methodwas used as the minimization procedure. A water solvation optionusing the GB/SA model35 was used for calculations in water. Thesearch was set to 10,000 structures to be minimized and all struc-tures within 20 kJ mol�1 energy range were stored.

Molecular dynamics simulations for compounds 1–4 were per-formed in SYBYL 7.336 Initial structures obtained by change of X-raycrystal structure of dirithromycin were minimized using Triposforce field.37 Molecular dynamics analysis was completed usingSYBYL 7.3 Starting structures were unconstrained and energy mini-mized using the Tripos force field. The applied atomic charges wereGasteiger-Huckel. The number of iterations was 10,000 and thecalculation was terminated when the difference in energy betweentwo conformations was no more than 0.005 kcal/mol.

Acknowledgements

This study was supported by the Ministry of Science, Educationand Sports of the Republic of Croatia (Project No. 119-1191342-1083). We are indebted to Dr. S. Koštrun and I. Tatic for helpful dis-cussions and M. Banjanac for performing biochemicalmeasurements.

Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.bmc.2009.07.013.

References and notes

1. Schönfeld, W.; Mutak, S. In Macrolide Antibiotics; Schönfeld, W., Kirst, H. A.,Eds.; Birkhäuser Verlag: Basel, 2002; pp 73–95.

2. Gaynor, M.; Mankin, A. S. Front. Med. Chem. 2005, 2, 21.

3. Ban, N.; Nissen, P.; Hansen, J.; Moore, P. B.; Steiz, T. A. Science 2000, 289, 905.4. Schlünzen, F.; Zarivach, R.; Harms, J.; Bashan, A.; Tocilj, A.; Albrecht, R.; Yonath,

A.; Franceschi, F. Nature 2001, 413, 814.5. Hansen, J. L.; Ippolito, J. A.; Ban, N.; Nissen, P.; Moore, P. B.; Steitz, T. A. Mol. Cell

2002, 10, 117.6. Schlünzen, F.; Harms, J. M.; Franceschi, F.; Hansen, H. A. S.; Bartels, H.; Zarivach,

R.; Yonath, A. Structure 2003, 11, 329.7. Tu, D.; Blaha, G.; Moore, P. B.; Steitz, T. A. Cell 2005, 121, 257.8. Mankin, A. S. Curr. Opin. Microbiol. 2008, 11, 414.9. Poehlsgaard, J.; Douthwaite, S. Nat. Rev. Microbiol. 2005, 3, 870.

10. Johnson, M. A.; Pinto, B. M. Bioorg. Med. Chem. 2004, 12, 295.11. Novak, P.; Tatic, I.; Tepeš, P.; Koštrun, S.; Barber, J. J. Phys. Chem. A 2006, 110,

572.12. Novak, P.; Banic Tomišic, Z.; Tepeš, P.; Lazarevski, G.; Plavec, J.; Turkalj, G. Org.

Biomol. Chem. 2005, 3, 39.13. Košutic-Hulita, N.; Matak-Vinkovic, D.; Vinkovic, M.; Novak, P.; Kobrehel, G.;

Lazarevski, G. Croat. Chem. Acta 2001, 74, 327.14. Everett, J. R.; Tyler, J. W. J. Chem. Soc., Perkin Trans. 2 1987, 1659.15. Steinmetz, W. E.; Shapiro, B. L.; Roberts, J. J. J. Med. Chem. 2002, 45, 4899.16. Lazarevski, G.; Vinkovic, M.; Kobrehel, G.; Ðokic, S.; Metelko, B.; Vikic-Topic, D.

Tetrahedron 1993, 49, 72.17. Awan, A.; Brennan, R. J.; Regan, A. C.; Barber, J. J. Chem. Soc., Perkin Trans. 2

2000, 1645.18. Bertho, G.; Gharbi-Benarous, J.; Delaforge, M.; Lang, C.; Parent, A.; Girault, J. P. J.

Med. Chem. 1998, 41, 3373.19. Gyi, J. I.; Pye, D. A.; Barber, J. J. Chem. Soc., Chem. Commun. 1991, 1471.20. Mayer, M.; Meyer, B. J. Am. Chem. Soc. 2001, 123, 6108.21. Vogtherr, M.; Peters, T. J. Am. Chem. Soc. 2000, 122, 6093.22. Alihodzic, S.; Fajdetic, A.; Kobrehel, G.; Lazarevski, G.; Mutak, S.; Pavlovic, D.;

Štimac, V.; Cipcic, H.; Dominis Kramaric, M.; Erakovic, V.; Hesenhorl, A.;Maršic, N.; Schoenfeld, W. J. Antibiot. 2006, 59, 753.

23. Kupce, E.; Freeman, R. J. Magn. Reson., Ser. A 1993, 105, 310.24. Blechta, V.; Rio-Portillas, F.; Freeman, R. Magn. Reson. Chem. 1994, 32, 134.25. Montejo-Bernardo, J. M.; Garcia-Granda, S.; Bayod-Jasanada, M. S.; Llavota-

Diaz, L.; Llorente, I. Zeitschrift für Kristallographie 2003, 218, 703.26. Montejo-Bernardo, J. M.; Garcia-Granda, S.; Bayod-Jasanada, M. S.; Llavota-

Diaz, L.; Llorente, I. Zeitschrift für Kristallographie 2005, 220, 66.27. Petropoulos, A. D.; Kouvela, E. C.; Starosta, A. L.; Wilson, D. N.; Dinos, G. P.;

Kalpaxis, D. J. Mol. Biol. 2009, 385, 1179.28. Hwang, T.-L.; Shaka, A. K. J. Am. Chem. Soc. 1992, 114, 3157.29. Sadiq, S. T.; Glasgow, K. V.; Drakeley, C. J.; Muller, O.; Greenwood, B. M.; Mabey,

D. C. W.; Bailey, R. L. Lancet 1995, 346, 881.30. Otwinowski, Z.; Minor, W.. In Macromolecular Crystallography; Carter, C. W., Jr.,

Sweet, R. M., Eds.; Academic Press: New York, 1997; Vol. 276, pp 307–326.Part A.

31. Sheldrick, G. M. SHELXS97 and SHELXL97; University of Göttingen: Germany, 1997.32. Mohamadi, F.; Richards, N. G. J.; Guida, W. C.; Liskamp, R.; Lipton, M.; Caufield,

C.; Chang, G.; Hendrickson, T.; Still, W. C. J. Comput. Chem. 1990, 11, 440.33. Ponder, J. W.; Richards, F. M. J. Comput. Chem. 1987, 8, 1016.34. Saunders, M.; Houk, K. N.; Wu, Y. D.; Still, W. C.; Lipton, M.; Chang, G.; Guida, N.

C. J. Am. Chem. Soc. 1990, 112, 1419.35. Still, W. C.; Tempczyk, A.; Hawley, R. C.; Hendrickson, T. J. Am. Chem. Soc. 1990,

112, 6127.36. SYBYL 7.3, Tripos Associates, 1699 S. Hanley Road, Suite 303, St. Louis, MO, 2006,

63144-2917.37. Clark, M.; Cramer, R. D., III; Van Opdenbosch, N. J. Comput. Chem. 1989, 10, 982.


http://dx.doi.org/10.1016/j.bmc.2009.07.013

Supplementary Material Crystallographic Data for 1 Chemical formula 2(C38 H70 N2 O13) x C3H6O x 7H2O Molecular weight 855.06 Crystal system Triclinic Space group P1 (No. 1) a(Å) 11.9202(6) b(Å) 12.7544(4) c(Å) 17.3664(7) �(º) 94.950(2) �(º) 109.155(2) �(º) 106.429(2) Z 1 Dcalc(g cm-3) 1.211 µ(mm-1) 0.09 F(000) 934.0 Crystal size 0.6 x 0.2 x 0.2 mm � Range (°) 1.3, 23.4 Total number of collected data 11581 Number of unique data 6659 R(int) 0.032 Number of observed data 5883 Threshold I > 2.0�(I) Number of parameters 1152 R(observed) 0.0425 wR2(obs) 0.0901 S 1.00 Largest difference peak and 0.02, -0.01 hole (eÅ-3) Proton and carbon chemical shifts, δ/ppm of compounds 1 -4 in D2O

1 2 3 4 atom 1H 13C 1H 13C 1H 13C 1H 13C

1 - 180.0 - 179.7 - 179.2 - 178.6 2 2.79 45.0 2.87 45.0 2.84 43.0 2.81 43.9

2 Me 1.18 13.9 1.19 13.9 1.25 15.2 1.21 15.8 3 3.88 75.8 3.88 76.3 3.62 76.1 3.60 77.0 4 1.85 42.2 1.93 42.1 1.98 35.7 2.02 36.3

4 Me 0.94 8.3 0.95 8.2 0.92 7.0 0.95 7.8 5 3.61 79.8 3.64 79.3 3.97 80.7 3.90 82.5 6 - 77.7 - 78.3 - 78.0 - 77.7

6 Me 1.27 20.3 1.31 20.0 1.3 18.8 1.26 50.4

6 OMe 3.11 51.0 3.04 51.2 3.18 49.6 3.00 20.1 7a, 7b 1.18;2.04 39.3 1.36;1.67 40.4 1.78;1.41 39.8 1.79;1.45 41.7

8 2.36 35.7 3.94 40.9 2.53 33.4 3.86 41.7 8 Me 0.97 18.8 1.01 22.3 1.02 17.0 1.09 22.3

9 - 180.0 - 175.2 - 179.8 - 175.9 10 4.08 45.3 2.57 40.9 4.13 45.3 2.61 41.7

10 Me 1.06 11.5 1.03 7.9 1.14 13.6 1.06 8.9 11 3.23 73.4 3.43 72.0 3.17 75.2 3.41 73.2 12 - 74.5 - 75.0 - 74.1 - 75.6

12Me 1.19 16.0 1.11 16.0 1.16 15.7 1.14 16.8 13 4.89 76.7 4.89 76.5 4.96 78.3 5.08 77.8 14 1.50;1.74 20.2 1.49;1.78 20.7 1.83;1.64 20.0 1.81;1.56 21.3 15 0.80 9.9 0.79 10.0 0.89 9.6 0.84 10.3 1' 4.37 102.5 4.38 102.5 4.66 101.2 4.63 102.1 2' 3.31 70.4 3.29 70.6 3.35 69.9 3.33 70.5 3' 2.77 63.0 2.70 62.8 2.79 63.1 2.81 63.8

3'NMe2 2.31 39.1 2.25 39.2 2.31 38.9 2.32 39.4 4'a, 4'b 1.27;1.82 30.8 1.24;1.80 31.0 1.86;1.33 30.1 1.86;1.31 30.6

5' 3.74 68.2 3.73 68.3 3.61 69.1 3.59 69.7 5'Me 1.16 20.1 1.16 20.0 1.23 19.3 1.20 19.9

1'' 4.93 94.7 4.92 95.0 - - - 2''a, 2"b 1.55;2.40 33.8 1.58;2.43 38.8 - - -

3'' - 72.8 - 72.7 - - - 3'' Me 1.16 20.1 1.17 20.1 - - - 3''Ome 3.27 48.8 3.27 48.8 - - -

4'' 3.13 77.2 3.13 77.1 - - - 5'' 4.12 65.0 4.12 65.1 - - -

5''Me 1.23 17.2 1.24 17.3 - - -

Key torsion angles and internuclear distances for compounds 1 and 2 (global minima) modelled using a Monte Carlo conformational search AMBER and MM2 force-fields, and for compound 2 structure 19 (18 kJ mol -1 above the global minimum)

Torsion angles/° Internuclear distances/Å H2,H3 H3,H4 H4,H5 H7s,H8 H3,H8 H3,H11 H4,H11 H4,

1 AMBER 179.7 -83.5 167.4 -164.4 5.3 6.9 4.6 1 MM2 179.6 -83.3 165.9 -164.0 5.3 6.9 4.6

2 AMBER 100.8 -70.8 69.8 172.8 4.0 4.7 5.3 Structure 19 108.8 -70.2 137.9 166.4 4.3 3.5 3.0

2 MM2 88.9 160.1 98.8 170.7 4.6 6.6 7.1 Azithromycin 109 -71 134.5 151 5.6 3.2 2.7

Key torsion angles and internuclear distances for compounds 1 and 2 (global minima) modelled using a Monte Carlo conformational search AMBER and MM2 force-fields, constrained using NMR nOE data to 11-endo-conformations

Torsion angles/° Internuclear distances/Å H2,H3 H3,H4 H4,H5 H7s,H8 H3,H8 H3,H11 H4,H11 H4,H6Me

1 AMBER 108.6 -72.1 136.5 165.5 5.2 3.6 3.1 4.1 1 MM2 105.6 -61.5 113 107.2 3.4 2.4 3.8 2.7

2 AMBER 108.7 -70.0 138.1 165.9 5.3 3.4 3.0 4.5 2 MM2 112.0 -72.4 130.0 170.1 5.5 3.5 2.8 4.6

The energies of the calculated structures for compounds 2, 3 and 4.

Conditions

Heating Cooling

Distances in Å Compound

Temperature in K Time in fs Temperature in

K Time in fs

Energies in kcalmol-1

3-11 4-11

2000 2000 0 10000 46.767 2.76 5.58 1000 2000 0 10000 50.341 4.42 2.96 800 2000 0 10000 54.113 5.66 8.21 700 2000 0 10000 58.233 2.83 5.04 600 2000 0 10000 49.218 5.64 5.10 500 2000 0 10000 45.930 4.68 2.58 1000 1500 0 10000 46.878 4.13 3.47 1000 1000 0 10000 49.275 7.30 5.22 1000 750 0 10000 49.503 4.69 2.58 1000 500 0 10000 39.229 3.90 2.62 1000 500 50 10000 38.219 4.21 4.60 1000 500 100 10000 43.474 6.29 7.43 1000 500 150 10000 35.080 7.53 5.63 1000 500 200 10000 49.091 6.33 7.86 1000 500 250 10000 51.148 5.46 5.38 500 1500 0 10000 44.097 3.40 4.42

2

500 1000 0 10000 41.630 2.59 3.84 2000 2000 0 10000 32.174 7.36 7.10 1000 2000 0 10000 35.784 3.04 3.60 800 2000 0 10000 30.557 5.55 6.43 700 2000 0 10000 37.422 5.04 4.30 600 2000 0 10000 30.083 4.38 4.91 500 2000 0 10000 40.613 4.92 4.36 700 1500 0 10000 38.453 3.11 2.65 700 1000 0 10000 30.479 6.57 4.60 700 750 0 10000 33.664 6.27 5.29 700 500 0 10000 47.187 2.61 5.32 700 1500 50 10000 36.542 2.74 3.60 700 1500 100 10000 35.276 4.68 2.72 700 1500 150 10000 33.030 3.21 6.10 700 1500 200 10000 31.531 3.35 3.75 700 1500 250 10000 31.272 5.26 5.34 700 1500 300 10000 39.282 2.67 3.45 700 1500 100 5000 32.725 3.87 2.70 700 1500 100 2500 43.490 4.43 5.79 700 1500 100 2000 44.782 4.87 3.88 700 1500 100 1500 32.390 3.54 3.19 700 1500 100 1000 34.612 4.55 2.61

3

700 1500 100 500 44.299 3.70 4.66

700 1500 100 10000 37.259a 2.58 3.83 700 1500 100 10000 33.859b 4.55 2.75 700 1500 100 2000 33.237a 2.56 2.64 700 1500 100 2000 31.778b 3.44 2.65 700 1000 100 1000 37.551a 4.22 4.86

700 1000 100 1000 46.681b 4.52 4.95 2000 2000 0 10000 33.203 7.53 8.83 1000 2000 0 10000 29.690 7.60 6.86 800 2000 0 10000 34.769 6.37 7.46 700 2000 0 10000 37.032 3.89 4.23 600 2000 0 10000 35.813 2.75 3.18 500 1500 0 10000 34.797 5.32 4.50 500 1000 0 10000 40.566 2.87 2.79 500 750 0 10000 36.767 4.67 2.74 500 500 0 10000 39.831 2.64 2.79 500 1000 50 10000 34.017 4.71 2.73 500 1000 100 10000 33.273 6.11 6.85

4

500 1000 150 10000 34.813 2.54 3.89

a Linear descent annealing function b Steepest descent annealing function

92

Chapter 5: Structures and antibacterial activity of tylosin A and tylosin B

5.1. DECLARATION

This chapter consists of a piece of work in preparation for publication in

Medicinal Chemistry Communications - B. Arsic, J. Aguilar Malavia, G. Morris, A.

McBain, R. Bryce, J. Barber: Structures and antibacterial activity of tylosin A and

tylosin B.

I synthesized tylosin B and did full NMR assignments of tylosin A and tylosin B.

In NMR recording and processing I had a help of Dr Malavia. Microbiological work was

done by me with the collaboration with Dr McBain. I performed the molecular

modelling with the collaboration of Dr Bryce. I prepared the article for publication along

with Supplementary material. It has been edited by Dr Barber.

Journal Name

Cite this: DOI: 10.1039/c0xx00000x

www.rsc.org/xxxxxx

Dynamic Article Links ►

ARTICLE TYPE

This journal is © The Royal Society of Chemistry [year] [journal], [year], [vol], 00–00 | 1

Structures and antibacterial activity of tylosin A and tylosin B

B. Arsica, J. Aguilar Malaviab, G. Morrisb, A. McBaina, R. Brycea, J. Barbera*

Received (in XXX, XXX) Xth XXXXXXXXX 20XX, Accepted Xth XXXXXXXXX 20XXDOI: 10.1039/b000000x

Full 1H and 13C NMR assignments of tylosin A in water and5

tylosin B in chloroform were performed. Molecular modellingon tylosin A gave the conformation which satisfies almost all2D ROESY NMR constraints. Tylosin A occupies the space oftwo azithromycin molecules, which was shown bysuperposition of structures.10

Introduction

The macrolide antibiotic tylosin A was first isolated from a strainof Streptomyces fradiae in 1961, 1 and its chemical structuredetermined by several groups over the next decade. 2-6 Althoughthe 14-membered ring macrolide was assigned to human use,15

tylosin A has been used in domestic animals, principally as ananti-inflammatory agent. There are few adverse side-effects, butthe foul taste means that the drug needs to be formulated incapsules.

Tylosin A acts by binding to the 50S ribosomal subunit of20

susceptible bacteria and resistance is mediated by methylation ofN1 of the nucleotide G748 within hairpin 35 of 23S rRNA,catalysed by the methyltransferase Rlm AII (formerly known asTlrB). 7 This is analogous to the methylation of rRNA by the ermgene product giving rise to erythromycin resistance.25

Remarkably, the first crystal structure of intact tylosin A has thedrug bound to 50S ribosomal subunits of Haloarculamarismortui.There are reports of NMR assignment of tylosin A in aproticsolvents (CDCl3, CD3CN).9, 10 Simova et al. 10 published that30

spectral assignment for tylosin in protic solvents (CD3OD, D2O)is very difficult due to the existence of several different tylosin Aforms in CD3OD (carbonyl, hemiacetal 1, hemiacetal 2) and D2O(ketone, hydrate). They have found that, in D2O at 300 K, ratiofor ketone:hydrate is 65:35. It is known that macrolides often35

exist in several forms.11-14 Tylosin A shows the existence ofdifferent forms, depending on pH.15-17

Tylosin B is a degradation product of tylosin A as a result of acidhydrolysis.18 However, in solutions for injection containingtylosin A, an alkaline degradation product was isolated-tylosin40

aldol (TAD).19 Tylosin B is important as a compound mainlybecause it can be used as a starting material for the synthesis ofmore active derivatives, particularly in position 20. The mostsuccessful derivative of tylosin B until now is tilmicosin, 20-deoxo-20-(3,5-dimethylpiperidin-1-yl) desmycosin.20

45

There are published modelling works on conformations oftylosin A21,22 in aprotic solvents using different force fields.

There are no published results on tylosin A conformation inwater. Also, there is no published work on full assignments oftylosin A in water and tylosin B in chloroform. The antibacterial50

activity of tylosin A and tylosin B was determined and theirproperties have never been compared with the rest of themacrolide class of antibiotics.

In order to understand the conformation of tylosin A wetherefore carried out both molecular modelling and NMR55

analyses. Tylosin B was synthesized using tylosin A as a startingmaterial.

Results and Discussion

Determination of ratio carbonyl:hydrate of tylosin A inwater, using NMR spectroscopy60

Simova et al. 10 reported that at 300 K tylosin A exists as amixture of carbonyl and hydrate forms in the ratio of 65:35. OurNMR investigations of tylosin A sample in D2O in thetemperature range of 298-338 K shows that increasingtemperature causes equilibrium to move to aldehyde form (the65

percentage is ~80% on 338 K).

Unconstrained conformational search on tylosin A

An unconstrained conformational search on tylosin A in waterwas performed using molecular mechanics (Macromodel). MM2force field was used.70

The starting point for modelling of tylosin A was a structure oftylosin A extracted by Molecular Viewer 2 from the pdb file ofco-crystal structure of tylosin A bound to the 50S ribosomalsubunit of Haloarcula marismortui.23 It is noteworthy that thecrystal structure of tylosin A is not available in the Cambridge75

crystallographic database, although it was isolated for the firsttime in the mid-1960s. It is obvious, from our preliminaryinvestigations and published data, that tylosin A exists as anequilibrium mixture of different tautomeric compounds.The global minimum of tylosin A (Figure 1) with energy E=280.980

kJmol-1 using Macromodel was found 17 times. The globalminimum satisfies all NMR constraints from 2D ROESY exceptH11-H13, H13-H11, H320-H4′, H4′-H320 and H36″-H2″b(ROESY connectivity table is available in Supplementary Data asTable 1). Contacts H320-H4′ and H4′-H320 are negligible small. It85

seems that H36″-H2″s belongs to spin-diffusion peaks.


Figure 1. The global minimum of tylosin A obtained using Macromodel

Two molecules of azithromycin were extracted from their co-crystal structure with the ribosome of D. radiodurans and tylosinA from its co-crystal structure with the ribosome of H.5

marismortui using MOE, and then they were subjected tosuperposition using Macromodel. Azithromycin alone possessesmild antimalarial activity, but more in combination with otherantimalarial drugs. We have shown that tylosin A actuallyoccupies the space of two azithromycin molecules, as10

demonstrated in Figure 2.

Figure 2. Superimposed structures of two azithromycin molecules withone molecule of tylosin A

Baram et al.24 suggest the existence of two binding sites for15

azithromycin, which can explain azithromycin properties. It isquite interesting that involving of nitrogen in a macrolide ring,replacing at the same time carbonyl group changes so muchproperty and conformation of the drug compared witherythromycin A and clarithromycin. This small part of modelling20

can lead us to the hypothesis that tylosin A can exhibit differentproperties and uses, similar to azithromycin.

Full 1H and 13C assignments of tylosin A in phosphatebuffered solution (0.047 mol dm-3) (apparently pH=7)

The numbering and tylosin A structure are reported in Figure 3.25

Figure 3. Structure of tylosin A

The proton and carbon assignments of tylosin A were based on1D spectra (1H, 13C) and 2 D spectra (COSY, HMBC andHSQC).30

The results of proton and carbon assignments are represented

in Table 2S in Supplementary material.1H and 13C assignments of tylosin B (desmycarosyl tylosin) inCDCl3 are shown in Table 3S in Supplementary material.

Determination of minimum inhibitory concentrations of35

tylosin A and tylosin B

We were interested to see and compare the minimum inhibitoryconcentrations of tylosin A and tylosin B. It was shown that theyshow negligible activity against both Gram-positives and Gram-negatives compared to clarithromycin and azithromycin. In40

general, it can be concluded that tylosin B shows better activityagainst both Gram-positive and Gram-negative bacteria comparedto tylosin B. Minimum inhibitory concentrations of tylosin A andtylosin B are represented in Table 3.

Table 3. Minimum inhibitory concentrations (MICs) expressed in μg ml-145

for tylosin A and tylosin B compared to clarithromycin and azithromycin

Experimental

Synthesis of tylosin B (desmycarosyl tylosin):

The mixture of tylosin A tartrate (1.0000 g) and 0.2 M HCl (2050

ml) was stirred for 4h. After the reaction was finished, extractionwith chloroform (4x15 ml) was performed in order to removemycarose. The acidity of the solution was adjusted to pH=8.0using saturated solution of sodium bicarbonate and universalindicator and extracted with chloroform (4x15 ml). The extract55

was dried over anhydrous magnesium sulphate, solid removed byfiltration and filtrate evaporated to dryness. White crystallinetylosin B was obtained (0.5101 g, yield 70.45 %). Mp. 113-1160C. Compound was identified using combination of 1D (1Hand 13C) and 2 D NMR techniques (DQF COSY and HMBC) and60

low and high resolution mass spectrometry. m/z (ES+): 772.6;HRMS (EI): m/z calculated for C39H66O14N1: 772.4478. Found772.4467.

Molecular modelling

Tylosin A was constructed from the crystal structure of tylosin A65

using Macromodel software.25 The structure was minimizedusing the Truncated Newton Conjugate Gradient (TNCG) methodin order to obtain local minima. The Monte Carlo MultipleMinimum (MCMM) conformational search was used to find theglobal minima.26 The TNCG method was used as the70

minimization procedure. A water solvation option using the

OrganismMIC( µg ml-1)

ClarithromycinAzithromycinTylosin

ATylosin

BEscherichia coli 12 2 125 31.25Bacillus cereus 0.2 4 0.98 0.98

Staphylococcus aureus 0.2 8 31.25 >500Pseudomonas

aeruginosa62.5 15.6 250 62.5

Staphylococcusepidermidis

250 31.2 250 62.5

Serratia marcescens >500 >500 500 250Corynebacterium

xerosis>500 250 500 250


GB/SA model27 was used for calculations in water. The searchwas set to 10000 structures to be minimized and all structureswithin 20 kJmol-1 energy range were stored.

NMR analysis

Tylosin A tartarate was obtained from Aldrich and used without5

further purification. All measurements were made using a 0.047mol dm-3 phosphate buffered tylosin A solution with apparent pHof 7. TSP was used as an internal standard.1H and 2 D (COSY, HMBC, HSQC and ROESY) NMR spectraof tylosin A were recorded on Bruker spectrometer operating at10

400 MHz for 1H. 13C spectra of tylosin A were recorded onBruker spectrometer operating at 500 MHz.Previously synthesized tylosin B was used in NMR experiments.The solution was obtained by dissolving 40 mg of tylosin B in700 µl of CDCl3.

1H, 13C and 2 D (COSY, HMBC, HSQC) NMR15

spectra of tylosin B were recorded on Bruker spectrometeroperating at 500 MHz for 1H.

Minimum inhibitory concentrations

These were determined by standard methods28.Inocula for broth dilution end point determination of bacterial20

antimicrobial susceptibility were prepared as follows: singlecolonies of anaerobic test bacteria from uncontaminated agarplates were inoculated into sterile, nutrient broth (10 ml)contained in 25 ml sterile plastic universals and incubated in astandard aerobic incubator at 37 °C for 24 hours. Cultures were25

then diluted to concentration of 105 cfu ml-1 using theMacFarland standard. Stock samples of clarithromycin,descladinosyl clarithromycin, azithromycin and descladinosylazithromycin were prepared in distilled water. Testing wasperformed in 96-well microtitre plates (Becton Dickinson,30

Franklin Lakes, NJ, USA). A diluted overnight culture (100 μl)was delivered to each test well, except the first column of theplate containing 100 μl of the antibiotic solution in distilled waterdiluted by 100 μl of double strength broth, and the last twocolumns of the plate which contain 100 μl of single strength broth35

and 100 μl of distilled water, respectively. Doubling dilutions wasthen carried out across the plate using a multi-channel pipette,changing the tips at each dilution step. The plates were thenincubated for 24 hours in a standard incubator at 37°C. The MICswere determined as the lowest concentration of antimicrobial at40

which growth did not occur. Growth was detected as turbidity,relative to an uninoculated well using a microtitre plate reader(Anthos HTII; Anthos-Labtec Instruments, Salzburg, Austria).Each MIC determination was carried out in triplicate (in the same96-well plate).45

Conclusions

We performed full 1H and 13C NMR assignments on tylosin A inwater, and tylosin B in chloroform. NMR investigations on theequilibrium of carbonyl: hydrate ratio were shown that thepercentage of aldehyde form increases with the augmentation of50

experimental temperatures. The molecular mechanics calculationin Macromodel gave, as a global minimum, conformation oftylosin A which satisfies almost all 2D ROESY NMR constraints.Two azithromycin molecules are almost completelysuperimposable on one tylosin A structure, which can provide a55

clue that tylosin A can, potentially, be used in antimalarialmedicine. Tylosin B shows better antibacterial activity thantylosin A.

Notes and referencesa School of Pharmacy and Pharmaceutical Sciences, The University of60

Manchester, Oxford road, M13 9PT, Manchester, United Kingdomb School of Chemistry, The University of Manchester, Oxford road, M139PL, Manchester, United Kingdom† Electronic Supplementary Information (ESI) available

65

1 R. L. Hamill, M. E. Haney Jr, M. Stamper and P. F. Wiley, Antibiot.Chemother., 1961, 11, 328.

2 H. Achenbach, W. Regel, W. Karl, Chemische Berichte, 1975, 108 (7),759.

3 S. Omura, A. Nakagawa, M. Machida, H. Imai, Tetrahedron Letters,70

1977, 12 (8), 1045.4 S. Omura, A. Nakagawa, A. Neszmelyi, S. D. Gero, A. M. Sepulchre, F.

Piriou, G. Lukacs, Journal of the American Chemical Society, 1975,97 (14), 4001.

5 A. Neszmelyi, S. Omura, G. Lukacs, Journal of the Chemical Society,75

Chemical Communications, 1976, 3, 97.6 R. B. Morin, M. Gorman, Tetrahedron Letters, 1964, 33-34, 2339.7 I. Lebars, C. Husson, S. Yoshizawa, S. Douthwaite, D. Fourmy, Journal

of Molecular Biology, 2007, 372 (2), 525.8 S. Omura, H. Matsubara, A. Nakagawa, A. Furusaki, T. Matsumoto,80

Journal of Antibiotics, 1980, 33 (8), 915.9 C. Rossi, A. Donati, M. P. Picchi, M. R. Sansoni, G. Corbini, P. Corti,

Magnetic Resonance in Chemistry, 1992, 30 (10), 954.10 S. Simova, G. Ivanova, Magnetic Resonance in Chemistry, 1996, 34

(4), 255.85

11 J. Barber, J. I. Gyi, L. Lian, G. A. Morris, D. A. Pye, J. K. Sutherland,J. Chem. Soc., Perkin Trans. 2, 1991, 10, 1489.

12 J. R. Everett, E. Hunt, J. W. Tyler, J. Chem. Soc., Perkin Trans. 2,1991, 10, 1481.

13 P. Tyson, H. Barjat, W. Mann, J. Barber, J. Chem. Res., Synop., 1998,90

727 (12), 3146.14 Jpn. Pat., 1985, 60038394.15 J. Paesen, W. Cypers, K. Pauwels, E. Roets, J. Hoogmartens, Journal

of Pharmaceutical and Biomedical Analysis, 1995, 13 (9), 1153.16 I. A. Aksenova, E. M. Ter-Sarkisyan, Biotekhnologiya, 1988, 4 (2),95

217.17 I. A. Aksenova, E. M. Ter-Sarkisyan, R. D. Soifer, G. Ya. Glorova, L.

S. Yustratova, Antibiotiki (Moscow), 1984, 29 (3), 179.18 J. Paesen, W. Cypers, K. Pauwels, E. Rots, J. Hoogmartens, Journal of

Pharmaceutical and Biomedical Analysis, 1995, 13, 1153.100

19 H. A. Kirst, J. E. Toth, M. Debono, K. E. Willard, B. A. Truedell, J. L.Ott, F. T. Counter, A. M. Felty-Duckworth, R. S. Pekarek, J. Med.Chem., 1988, 31, 1631.

20 M. Debono, K. E. Willard, H. A. Kirst, J. A. Wind, G. D. Crouse, E.V. Tao, J. T. Vicenzi, F. T. Counter, J. L. Ott, E. E. Ose, S. Omura,105

The journal of antibiotics, 42 (8), 1253.21 P. M. Ivanov, J. Mol. Struct., 1998, 440, 121.22 P. M. Ivanov. J. Mol. Struct., 2002, 606, 217.23 J. L. Hansen, J. A. Ippolito, N. Ban, P. Nissen, P. B. Moore, T. A.

Steitz, Mol. Cell, 2002, 10, 117.110

24 D. Baram, A. Yonath, FEBS Letters, 2005, 579, 948.25 F. Mohamadi, N. G. J. Richards, W. C. Guida, R. Liskamp, M. Lipton,

C. Caufield, G. Chang, T. Hendrickson, W. C. Still, J. Comput.Chem., 1990, 11 (4), 440.

26 M. Saunders, K. N. Houk, Z. D. Wu, W. C. Still, M. Lipton, G. Chang,115

W. C. Guida, J. Am. Chem. Soc., 1990, 112, 1419.27 W. C. Still, A. Tempczyk, R. C. Hawley, T. Hendrickson, J. Am.

Chem. Soc., 1990, 112, 6127.28 A. J. McBain, R. G. Ledder, P. Sreenivasan, P. Gilbert, J. Antimicrob.

Chemother., 2004, 53 (5), 772.120

1

Supplementary material

B. Arsica, J. Aguilar Malaviab, G. Morrisb, A. McBaina, R. Brycea, J. Barbera*

a School of Pharmacy and Pharmaceutical Sciences, The University of Manchester,

Oxfor d road, M13 9PT, Manchester, United Kingdomb School of Chemistry, The University of Manchester, Oxford road, M13 9PL,

Manchester, United Kingdom

2

2s 2r 3 4 5 6 7s 7r 8 10 11 13 14 15 16s 16r 17 18 19s 19r 20 21 22 23s 23r 1' 2' 3' 4' 5' 6' 7' 8' 1" 2”s 2”r 4" 5" 6" 7" 1‴ 2‴ 3‴ 4‴ 5‴ 6‴

2s ● L V M S

2r L ● V V M

3 V V ● S

4 S ●5 S M ● S S

6 M S ● V S V

7s S ● L S

7r S L ● S

8 S V ● V S

10 ● S S M

11 S ● M

13 M ● V S V

14 V ● V V V M

15 V ● V V

16s V ● L M V

16r V S L ● S M

17 S S M ● V

18 S ● S

19s V ● L

19r L ●20 V ● V

21 S S S ●22 S M V ●23s M V V ● S

23r V V M V S ●1' S ● S S S V

2' V S ●3' V ● S M

4' V S ● S M S

5' S S ●6' S ● V M

7' 8' M M V ● M

1" S ● S S S

2”s M ● S S

2”r M ● V V S

4" V ● M S S

5" S M ● M

6" S S M M ●7" M ●1‴ ● M S

2‴ M ● S

3‴ S S ● S

4‴ V M ● M S

5‴ S M ● M

6‴ M M ●2s 2r 3 4 5 6 7s 7r 8 10 11 13 14 15 16s 16r 17 18 19s 19r 20 21 22 23s 23r 1' 2' 3' 4' 5' 6' 7' 8' 1" 2”s 2”r 4" 5" 6" 7" 1‴ 2‴ 3‴ 4‴ 5‴ 6‴

Table 1S: 400 MHz ROESY connectivities for tylosin A (0.047 mol dm-3 phosphate buffered tylosin A solution with apparent

pH of 7). TSP was used as an internal standard.

3

Table 2S. The full 1H and 13C NMR assignments of tylosin A in phosphate buffered D2O (apparent pH=7)

Position Multiplicity 1H (ppm) JHH (Hz) 13C (ppm) HMBC connectivities (13C1H)1 - - - 178.3 H2a, H2b

2mm

2.182.67

--

39.5 H318

3 m 3.83 - 68.3 H3184 m 1.67 - 40.6 -5 m 3.65 - 80.6 H3186 m 1.94 - 31.0 H321

7mm

1.501.90

--

31.6 H321

8 m 2.55 - 45.0 H3219 - - - 209.2 H10, H11, H321

10 d 6.59 15.5 118.6 -11 d 7.32 15.5 149.5 H32212 - - - 136.1 H11, H32213 dd 5.96 10.4 143.8 H10, H11, H32214 m 3.14 - 44.6 -15 dt 4.89 10.4; 2.6 75.8 H317

16mm

1.851.60

--

24.6 H317

17 t 0.94 7.3 8.8 -18 d 0.98 3.9 8.3 -

19mm

2.932.60

--

43.6 H321

20 s 9.67 - 207.8 -21 d 1.24 6.5 16.9 H722 s 1.87 - 12.2 H11

23ddm

4.003.71

10.2; 3.4-

69.3 H14

1′ d 4.45 7.2 101.8 H2′2′ m 3.73 - 74.2 H36′3′ m 3.79 - 43.6 H67′/8′

7′/8′ s 3.06 - 40.9 H67′/8′4′ m 3.69 - 72.5 H36′5′ m 3.42 - 73.7 H36′6′ d 1.36 8.6 17.7 H2′1″ d 5.19 2.5 96.3 H36″, H37″

2″ mdd

2.051.99

-4.1

39.9 H37″

3″ - - - 69.7 H1″, H2″a, H2″b, H36″, H3 7″4″ d 3.23 9.1 74.9 H36″, H37″5″ m 3.86 - 66.8 H4″, H36″6″ d 1.33 2.7 17.1 H4″7″ s 1.28 - 25.0 H4″1″′ d 4.62 8.1 100.5 H2″′, H3″′2″′ dd 3.16 8.4;2.8 80.1 H3″′, HOCH32″′

OCH32″′ s 3.47 - 57.9 H2″′, H4″′, H5″′3″′ t 3.95 2.7 78.9 HOCH33″′, H2″′, H4″’

OCH33″′ s 3.57 - 61.5 H3″′, H4″′, H5″′4″′ dd 3.34 9.8;2.7 69.73 H3″′, H6″′5″′ m 3.75 - 69.2 H1″′, H4″′, H6″′6″′ d 1.24 6.5 16.6 -

From the structure of tylosin A, it is obvious that the highest chemical shifts in the carbon spectrum will havecarbonyl carbons. These include carbons C1, C9 and C20. The expectation is that C1 has the lowest chemical shift ofall of the above mentioned signals. Among C9 and C20, it is very easy to determine the exact chemical shifts for C9and C20 compared with 13C and DEPT90 spectra. Chemical shift for C1 is 178.3 ppm, for C9 is 209.2 ppm and forC20 is 207.8 ppm. Aldehyde proton is easily visible in the proton spectrum. It can be found at chemical shift of 9.67ppm. Using the COSY spectrum, we can find exact chemical shifts of H219, which are 2.93 ppm and 2.60 ppm. Withthe exception of this signal at low field, we can easily notice three other singlets with chemical shifts in the range 3-4ppm (OCH32″′, OCH33″′ and N(CH 3)2). Because of different magnetic environments, singlets of OCH32″′ andOCH33″′ must have negligibly different chemical shifts. Dimet hylamino protons are magnetically the same, so we willsee one signal as singlet of six protons in 1H spectrum (3.06 ppm). From HSQC spectrum, we can determine the

4

carbon chemical shift for C7′/8′ which is 40.9 ppm. The HMBC spectrum shows further that H 67′/8′ has connectionwith C7′/8′ and C3′ (strong crosspeak), which enables us to find the exact chemical shift (43.6 ppm). The chemicalshift of C3′ can help us to determine the chemical shift for H3′ using HSQC spectrum (3.79 ppm). The H3′ can affordclose contacts with H2′ and H4′. If we take a look of the structure of the compound, we can see that H2′ has slightlymore electronegative environment than H4′. Therefore, the chemical shift for H2′ is 3.73 ppm and 3.69 ppm for H4′.From HSQC spectrum, we find carbon chemical shifts: (C2′) is equal to 74.2 ppm and (C4′) is equal to 72.5 ppm.Carbon C2′ from the HMBC spectrum provides contact with methyl protons of low chemical shift (H36′)=1.36 ppm.The contact with the same proton is achievable for C4′. From HSQC spectrum, we can find the chemical shift for C6′which is equal to 17.7 ppm. The chemical shift for carbon C12 can be found on the basis of difference between 13Cand DEPT90. Looking at the chemical environment of C12 and C3″, we can expect much higher chemical shift of C12compared with C3″. Therefore, chemical shift for C12 is 136.1 ppm and 69.7 ppm for C3″.From HMBC, it is visible that C12 has two contacts: one with the proton at high chemical shift and another one withthe proton on low chemical shift. The proton at high chemical shift can be H11 (value is 7.32 ppm). Proton(s) whichcan have HMBC contact with C12 on high field are H322 which chemical shift is 1.87 ppm. From HSQC spectrum, wecan find the chemical shift for C11 and C22. The values are 149.5 ppm for C11 and 12.2 ppm for C22. Looking at theHMBC spectrum, we can see that H11 has crosspeak with carbon of a high chemical shift. Presumably, values shouldbe for C13 or C10. The higher chemical shift should be expected for C13 due to the presence of more electronegativeenvironment. Therefore, chemical shift for C13 is 143.8 ppm and for C10 is 118.6 ppm. From HSQC, we can findchemical shifts for H13 and H10. They are 5.96 ppm for H13 and 6.59 ppm for H10. The proton with the highchemical shift (expectation) is also H315 due to electronegative magnetic environment. Its chemical shift is 4.89 ppm.From HSQC spectrum, we can find the value for C15 which is 75.8 ppm. HMBC spectrum shows a connectivity ofC15 with H317 which enables us to find the exact chemical shift for H317 which is 0.94 ppm. The chemical shiftfound from the HSQC spectrum for C17 is 8.8 ppm. If we compare DEPT90 and DEPT135, we can see that there areseveral methylene carbons: C2, C7, C16, C19, C23 and C2″. The chemical shifts are 39.5 ppm, 31.6 ppm, 24.6 ppm,43.6 ppm, 69.3 ppm and 39.9 ppm. It is not easy to determine which chemical shift belongs to particular carbon. FromHMBC crosspeak of H317, we can see the connection with C16. The exact chemical shift of C16 is, therefore, 24.6ppm. The HSQC spectrum gives us information about the chemical shift of protons H216, which chemical shifts are1.85 ppm and 1.60 ppm. Carbons which should have high chemical shifts are C1′, C1″ and C1″′. However, if weconsider the magnetic environment, we can expect due to electronegative surrounding that C1′ has the highestchemical shift, then C1″′ and then C1″. We can assign their chemical shifts as 101.8 ppm for C1′, then 100.5 ppm forC1″′ and at the end 96.3 ppm for C1″. From HSQC spectrum, we can see that respective proton chemical shifts are4.45 ppm (H1′), 4.62 ppm (H1″′) and then 5.19 ppm (H1″). The carbon C1″′ has two very close contacts -with H3″′ andH2″′. Carbon C1″ have two crosspeaks with low chemical shift, which should be H7″ and H6″. Carbon C9 has HMBCconnectivities with H11, H10 and proton(s) with low chemical shift, which should be H321. The chemical shift forH321 is then 1.24 ppm. From HSQC spectrum, it is visible that the carbon chemical shift is 16.9 ppm. Because H321has HMBC connectivity with one of protons H27, which chemical shift is 1.50 ppm. From the HSQC spectrum, it canbe seen that carbon chemical shift of C7 is 31.6 ppm. At the same time, this value for carbon chemical shift enables usto find the chemical shift of the second H27 proton, which is 1.90 ppm. Protons H321 can afford HMBC contacts withH6 and H8 as well. Because of the proximity of the carbonyl group to H8, H8 can have higher chemical shift than H6.Therefore, chemical shift of H8 is 2.55 ppm and 1.94 ppm for H6. From the HSQC spectrum, we can see that carbonchemical shift C6 is 31.0 ppm and 45.0 ppm for C8. From the HMBC spectrum, besides the above mentionedconnectivities, protons H36′ can also have contact with C5′. The chemical shift of carbon C5′ is 73.7 ppm. Using theHSQC spectrum, we can obtain the chemical shift of proton H5′, which is 3.42 ppm. Besides the above mentionedconnectivities for H321 with carbons, it is also possible for them to have contacts with C19. The chemical shift ofcarbon C19 is 43.6 ppm. Using the HSQC spectrum, we have found that chemical shifts of protons H219 are 2.93 ppmand 2.60 ppm. Carbon C3″ has connectivity with H1″. Also, it can afford con tacts with both H22″ protons, then H37″and H36″. Because H 37″ and H36″ are methyl protons, we can expect that they have low chemical shift. From thisdifference, we can find the proton chemical shifts of protons H22″ (2.05 ppm and 1.99 ppm). Using the HSQCspectrum, we have found that carbon chemical shift of C2″ is 39.9 ppm. The carbon C2″ has HMBC crosspeak withproton(s) of low chemical shift, which appears in the proton spectrum as a singlet. It must be only H37”, which exactchemical shift is 1.28 ppm. Using HSQC spectrum, we have found the value of carbon chemical shift C7″ of 25.0ppm. Carbon C3″ has a lot of HMBC connectivities: H1″, H 22″, H37″ and also with protons of low chemical shift,which can be only H36″. The chemical shift of H36″ is 1.33 ppm. Using HSQC spectrum, we can see that carbonchemical shift C6″ is equal to 17.1 ppm. From the HMBC spectrum, we can see that both C7″ and C6″ have HMBCconnectivity with the proton which can be only H4″. The exact chemical shift is 3.23 ppm of proton H4″. UsingHSQC spectrum, we can see that carbon chemical shift of C-4″ is equal to 74.9 ppm. Proton H4″ has HMBC contactswith C6″, C7″ but also with carbon of higher chemical shift than those previously described. It can be only C5″, whichhas chemical shift of 66.8 ppm. Using the HSQC spectrum, we can see that the proton chemical shift of H5″ is 3.86ppm. We have already found that C1″′ can have HMBC contacts with H3″′ and H2″′. From the COSY spectrum, wecan see that H1″′ can afford only one contact, which presumably is H2″′, which chemical shift is equal to 3.16 ppm.

5

Using the HSQC spectrum, we can see that the chemical shift of C2″′ is equal to 80.1 ppm. The proton chemical shiftof H3″′ is then 3.95 ppm. The HSQC spectrum provides information that the c arbon chemical shift of C3″′ is 78.9ppm. Carbon C2″′ can have HMBC contact with H3″′ and a singlet of chemical shift in the range 3.0 -3.5 ppm, whichcan be either H(OCH3)2″′ or H(OCH3)3″′. Carbon C3″′ have HMBC contacts with H2″′, then proton which chemicalshift is ~3.30 ppm and a singlet which can be either H(OCH3)3″′ or H(OCH3)2″′. The COSY spectrum shows that H3″′has HMBC contacts with H2″′ and a proton with the chemical shift of ~3.30 ppm which is obviously H4″′. We can seefrom 1H spectrum that the exact chemical shift is 3.34 ppm. It is visible from the HSQC spectrum that the carbonchemical shift is 69.7 ppm. From the HMBC spectrum, it can be seen that H4″′ has contacts with H3″′ and proton(s) oflow chemical shift. It can be only H36″′, whose chemical shift seen from 1H spectrum is 1.24 ppm. Using the HSQCspectrum, we can find that carbon chemical shift is equal to 16.6 ppm. Sizes of crosspeaks with OCH33″′ and OCH32″′are large which tell us that seen connections must be H(OCH3)3″′, C3″′ and H(OCH3)2″′, C2″′. Using these relations,we have found that chemical shifts for protons OCH32″′ is 3.47 ppm, and that one for OCH 33″′ is equal to 3.57 ppm.Using the HSQC spectrum, we can find carbon chemical shifts of COCH32″′ which is 57.9 ppm and that one o fCOCH33″′ which is 61.5 ppm. Carbon COCH 32″′ shows HMBC connectivity with H4″′, H2″′ and a proton of chemicalshift of ~3.7 ppm. On the other hand, carbon COCH33″′ shows HMBC connectivities to H3″′, H4″′ and a proton withchemical shift of ~3.7 ppm. It is not difficult to find that the proton in question is H5″′ which exact chemical shift is3.75 ppm. Using the HSQC spectrum, we have found that the carbon chemical shift is equal to 69.2 ppm. We havealready found that C1 has chemical shift of 178.3 ppm and that it has HMBC connectivities with two protons ofchemical shift around 2.2 and 2.70 ppm. If we take a look of these signals and their multiplicities than we can see thatthey must correspond to two H22 protons. Their exact chemical shifts are 2.18 ppm and 2.67 ppm. From the HSQCspectrum, we can see that the carbon chemical shift must be 39.5 ppm. The carbon C2 has HMBC contact with protonsof low chemical shift which are not H317. Therefore, we can conclude that the protons must be H318, which chemicalshift is 0.98 ppm. From HSQC, we can see that the exact carbon chemical shift is 8.3 ppm. Protons H318 have HMBCcontacts with C2, but also with two other carbons with chemical shift around 80 ppm and 70 ppm. On the basis oftheir chemical shifts, we can see that they must be surrounded with electronegative groups. The only choices are C5and C3 where C5 must have higher chemical shift than C3. Therefore, chemical shift for C5 is 80.6 ppm and for C3 is68.3 ppm. Using the HSQC spectrum, we have found that (H3) is equal to 3.83 ppm. Found carbon chemical shift forC5 enables us to find the proton chemical shift of H5 which is 3.65 ppm. From DEPT135, we have found twomethylene protons which still need to be assigned (C14 and C23). The chemical shift of C23 is higher compare toC14, so the carbon chemical shift of C23 is 69.3 ppm. The carbon chemical shift, C14, is 44.6 ppm. The exactchemical shift of H14 was found from the HSQC spectrum to be 3.14 ppm. Proton H14 has HMBC connectivity withC23, which proves the above found value. The proton chemical shift of H223, found from the HSQC spectrum is 4.00ppm and 3.71 ppm. Using the COSY spectrum, we have found that strong crosspeaks, which H5 and H3 show are withH4, which chemical shift is equal to 1.67 ppm. The HSQC spectrum gives us the exact chemical shift of carbon C4which is 40.6 ppm.

6

Table 3S. Multiplicities, proton and carbon chemical shifts and HMBC connectivities (13C→1H) for tylosin B (desmycosine tylosin) inCDCl3

Position Multiplicity 1H (ppm) 13C (ppm) HMBC connectivities (13C→1H)1 - - 173.9 H2a, H2b, H3

2md

2.451.94

39.4 H318

3 d 3.84 70.1 H2a4 m 1.56 75.1 -5 m 3.71 81.9 H318, H1′,H2′6 m 2.27 30-35* H321

7mm

1.471.64

30-35* H321

8 m 2.57 44.7 H3219 - - 203.2 H10, H11

10 d 6.25 118.4* -11 d 7.30 148.1 H13, H32212 - - 134.9 H10, H11, H3 2213 d 5.91 142.3 H11, H322, H23a, H23b14 m 2.92 45.0 H23a, H23b15 t 4.98 75.1 H16b, H317, H322, H23a

16mm

1.861.53

25.5 H317

17 t 0.94 9.7 H16a18 d 1.1 9.0 -

19mm

2.932.36

43.7 H320

20 s 9.70 203.0 H19a, H32121 d 1.21 17.4 H7a, H7b22 s 1.79 13.0 H11, H13, H16b

23ddm

4.013.57

69.0 H2″′

1′ d 4.25 104.0* H5, H3′, H4′2′ m 3.60 71.0 H3′, H4′, H6 7′/8′3′ m 2.47 70.6 H4’, H5’, H6 7’/8’4′ m 3.52 75.1 H2′, H 36′5′ m 3.30 73.3 -6′ d 1.26 17.8 H2′

7′/8′ s 2.50 41.7 H3′, H 36′, H 6 7′/8′1″′ d 4.56 101.1 H15, H23a, H23b, H2″′, H3″′2″′ m 3.02 81.9 H3″′, H4″′

OCH3-2″′ s 3.49 59.7 H2″′, H3″′, H4″′3″′ m 3.75 79.8 H1″′

OCH3-3″′ s 3.62 61.8 H3(OCH3-2″′), H3″′4″′ m 3.27 72.6 H2″′, H5″′5″′ - 3.75 70.8 H1″′, H3″′, H36″′6″′ d 1.26 17.8 H2″′

* Signals are very broad due to chemical exchange. In the case of C1′ and C10, it was possible to see the apex of the peak, whose

values are presented in the table. Regarding C6 and C7, however, peaks are merged and broad, so it is not possible to determine the

value of the apex

The chemical shift of aldehyde proton was easy to find (=9.70 ppm). Chemical shifts for protons within methylgroups attached to oxygens and nitrogens were also distinguishable. Protons close to nitrogen have lower chemicalshift than protons attached to oxygen. Therefore, the chemical shift for H67′/8′ is 2.50 ppm. Methyl protons H 3(OCH3-3″′) and H 3(OCH3-2″′) are not recognisable at first sight. Methyl protons H322, due to very electronegativesurrounding, should have a relatively high chemical shift for methyl group protons not directly attached toelectronegative atom. The chemical shift for protons H322 is 1.79 ppm. Using HMBC spectrum, we can see that theremust be crosspeak corresponding to interaction of H320 with C19 (=43.7 ppm). Protons H67′8′ shows always in allmacrolide antibiotics strong interaction with carbons nearby i.e. with C2′ and C3′. Due to more electronegativeenvironment, C3′ shoul d have higher chemical shift than C2′ (c3′ = 70.6 ppm and c2′ =71.0 ppm). Protons H322 havea number of interactions, including C11, C13, C12 and C15. Due to different magnetic environment, they havedifferent chemical shifts (c11 = 148.1 ppm, c13 = 142.3 ppm, c12 = 134.9 ppm, c15 = 75.1 ppm). Carbonyl carbonsbelong to C9 and C20 must have very high chemical shifts (over 200 ppm). Due to a possibility for resonance of C9and most electronegative environment, it has higher chemical shift (203.2 ppm) than C20 (203.0 ppm). In the HMBCspectrum, it is visible connectivity of C9 with H11 and H10, keeping in mind that chemical shift of H11 must behigher than H10 (H11 = 7.30 ppm, H10 = 6.25 ppm). Connectivities of H11 with carbons can be seen in HMBCspectrum (C13, C12 and C22). Chemical shifts of carbons C13 and C12 were previously determined, so the chemicalshift of C22 in high magnetic field is 13.0 ppm. Carbon C20 in HMBC has connectivities with proton H19 and H321

7

(chemical shift of H19 is higher than H321). Therefore, we have H19 = 2.93 ppm and H21=1.21 ppm. Except withH322, carbon C11 has HMBC connectivity with H13 which chemical shift is 5.91 ppm. Protons H322 have HMBCconnectivity with C13, which also gives connection with both protons H223 which chemical shifts are 3.57 ppm and4.01 ppm. Proton Ha23 has HMBC contact with C14 and C15, which have different chemical shifts due to differentmagnetic environment. Their exact chemical shifts are 45.0 ppm and 75.1 ppm for C14 and C15, respectively. P rotonH15 has connectivities in the HMBC spectrum, except with Ha23 and H322, close contact with one of H216 protonsand H317 protons. Protons H317 are situated in high magnetic field with chemical shift of 0.94 ppm. The H317 protonsalso have HMBC connectivity with C16. Its chemical shift in high magnetic field is 25.5 ppm. Proton Ha23 has strongHMBC connectivity with C1″′, too. The chemical shift of C1″′ is 101.1 ppm. Carbon C1″′ possesses a number ofcontacts including Ha23, but also crosspeak with proton(s) in low magnetic field and two protons in proton range 3-4ppm. Protons in low magnetic field can be H15 due to very electronegative environment. Its chemical shift is 4.98ppm. Using the COSY spectrum, it can be seen two connectivities which can be ascribed to H14 and Hb16, whichchemical shifts are 2.92 ppm and 1.53 ppm, respectively.Lactone carbonyl group due to inner group resonance should have lower chemical shift compare to C9 and C20.Therefore, its chemical shift is 173.9 ppm. It should have HMBC connectivities with H3 and H22. The strength ofconnectivity is reflected on the size of crosspeaks. Corresponding chemical shifts are 1.94 ppm and 2.45 ppm for H22and 3.84 ppm for H3. Carbons C1’ and C1″′ should have close chemical shifts. However, because of proximity ofmethylene group and its positive inductive effect, carbon C1″′ should have negligible low chemical shift than C1′,which chemical shift is 104.0 ppm. In HMBC, it is visible very strong like in case with other macrolides the crosspeakcorresponds to C7′8′ and H67′8′. The chemical shift of C7′8′ is 41.7 ppm. The strong crosspeak, which C7′8′ giveswith H3′, can help us find the chemical shift of H3′, which is 2.47 ppm. Proton chemical shifts for H1″′ and H1′ mustbe very similar. In this case, proton chemical shift H1″′ is higher than H1′. The chemical shifts are equal 4.56 ppm forH1″′ and 4.25 ppm for H1′. We can expect HMBC contact of H1″′ with C3″′ and C5″′. In this case, C3″′ should havehigher chemical shift due to more electronegative environment. Therefore, the chemical shift for C3″′ is 79.8 ppm and70.8 ppm for C5″′. Proton H1″′ shows strong correlation with H2″′ in the COSY spectrum, which enables us todetermine the chemical shift of H2″′ (3.02 ppm). The chemical shift for proton H2′ can be seen from the COSYspectrum as well due to its large crosspeak with H1′. The chemical shift of H2′ is 3.60 ppm. Carbon C1″′ has a strongHMBC connection with H3″′ and H 3(OCH3-2″′). Their chemical shifts are 3.03 ppm (H3″′) and 3.49 ppm (H3(OCH3-2″′)). Proton H3″′ shows HMBC connections with C2″′ and C(OCH 3-3″′). Due to different magnetic environments,their chemical shifts are 81.9 ppm (C2″′) and 61.8 ppm (C(OCH 3-3″′)). Carbon C2″′ shows HMBC connectivity withH4″′ (3.27 ppm). The crosspeak in the HMBC spectrum of H4″′ with C(OCH3-2″′) enables us to find the chemicalshift of C(OCH3-2″′) which is 59.7 ppm. After finding the chemical shift for H 3(OCH3-2″′), it is easy to find thechemical shift for H3(OCH3-3″′) (3.62 ppm). The chemical shift for H2″′, and using the HMBC spectrum, gives us anopportunity to determine the chemical shifts of C4″′, C23 and C6″′. Regarding the magnetic environment, they allhave different environments, so their chemical shifts are 72.6 ppm (C4″′), 69.0 ppm (C23) and 17.8 ppm (C6″′).Proton H1′ has only one aspect of HMBC connectivity with C5 (81.9 ppm). In contrast, carbon C5 has several HMBCconnectivities, including H2′ and H318. The chemical shift of H318 is 1.10 ppm. Protons H321 have severalconnections in the HMBC spectrum. They are C8, C7 and C6. Their chemical shifts are 44.7 ppm (C8), but exactchemical shifts for C6 and C7 are impossible to find due to chemical exchange. Their carbon chemical shift lies in therange 30-35 ppm. Carbon C5″′ shows large crosspeak in the HMBC spectrum with proton(s) with low chemical shift.It can be H36″′, whose chemical shift is 1.26 ppm. Using the COSY spectrum, we can find the chemical shift for H4′,which is 3.52 ppm. Chemical shift of H2′ helps us by using the HMBC spectrum to find chemical shifts of C4′ (75.1ppm) and C6′ (17.8 ppm). Carbon chemical shift C4′ gives us an opportunity to find chemical shift H 36

′ (1.26 ppm).Determination of H36′ helps us to determine C5′ (73.3 ppm). Due to more electronegative environment C21 shouldhave higher chemical shift than C22. Its chemical shift is 17.4 ppm. Using the COSY spectrum and knowing thechemical shift of one H216 proton, we can find chemical shift of the second H216 (1.86 ppm). One of protons H216has connection with carbon in high magnetic field, which is C17 (9.7 ppm). The chemical shift of carbon C18 is verylow and is equal to 9.0 ppm. Chemical shift of another proton H219 can be seen using the COSY spectrum and knownchemical shift of one of protons (2.93 ppm). Its chemical shift is 2.36 ppm. Using the chemical shift of C21, we candetermine from the HMBC spectrum a chemical shift of H27 protons (1.47 ppm and 1.64 ppm). Using the COSYspectrum and the proton chemical shift of one H27, we can find the chemical shift of H8 (2.57 ppm). Due to chemicalexchange, carbon signal for C10 is very week and broad (118.4 ppm). Using carbon chemical shift, we were able todetermine the chemical shift H5 (3.71 ppm). Chemical shift of proton H5 using COSY spectrum helps us to determinethe chemical shift of H4 (1.56 ppm). The chemical shift of protons H318 gives rise to determination of C2 (39.4 ppm).Using H5 chemical shift, we determined chemical shift of C4 (75.1 ppm) and C3 (70.1 ppm). The COSY Spectrumand proton chemical shift of one H219 (2.36 ppm) enables us to find the chemical shift H6 (2.27 ppm).

93

Chapter 6: Anti-malarial activity of macrolide antibiotics: in silico study on theapicoplast ribosomal exit tunnel of Plasmodium falciparum

6.1. IntroductionMalaria is a disease caused by several strains of protozoa from the Plasmodium

genus. Several attempts have been made in order to control the disease that has

included vaccination, vector control and parasiticidal drugs. 1 Currently, parasiticidals

constitute the most routinely applied approach in combating malaria. However, there is

widespread resistance to existing anti-malarial drugs. Also, many studies have been

conducted into the development of prophylactic treatment (i.e. vaccine). The most

promising vaccine to date described as RTS, S/AS01, is currently in phase 3 of clinical

trials, the outcome of which will be known by 2014. Early results indicate that there is

an increased risk in children administered with this medicine of meningitis infection,

which is one of the drawbacks of this vaccine.2

Apicoplast, a plastid inside the protozoa, contains its own house-keeping

processes which may be exploited as a drug target.

Plastid is a generic term to describe organelles of all colours. Apicoplast was

originated from the same endosymbiosis as other plastids.3 The apicoplast can be

regarded as a mini-bacterium living inside the malaria parasite. At the time of

discovery, Mereschkowsky poetically describe plastids in 1905 as “little green slaves”

who are working for their hosts to produce food from the sunshine.4 All familiar

processes happening inside this cell within a cell are bacterial in nature (DNA

replication, transcription, translation, post-translational modification, catabolism and

anabolism), and can be potential drug targets.

There are several families of drugs that are in use or under investigation as

potential medicines against the malarial parasite. Some of them act on metabolic

targets such as DNA replication (ciprofloxacin)5, RNA transcription (rifampicin)6,

protein translation (clindamycin, erythromycin, azithromycin, spiramycin7, thiostrepton8,

micrococcin9, chloramphenicol, doxycycline, tetracycline, amythiamicin10)11, amino acid

biosynthesis (glyphosate)12, IPP biosynthesis (fosmidomycin)13 and fatty acid

biosynthesis (thiolactomycin, clodinafop, quizalofop, haloxyfop, triclosan)14. However,

only few of them show significant IC50 values for activity against P. falciparum

(rifampicin, azithromycin, thiostrepton, tetracycline, amythiamicin and fosmidomycin). It

has previously been reported that rifampicin has a putative target plastid RNA

polymerase β-subunit, but its mode of action yet remains unclear. Azithromycin has

activity on plastid 23S rRNA, but there is no substantial evidence of its mode of action

and strong indirect evidence for the activity is provided for thiostrepton (putative target

plastid 23S rRNA). Tetracyclines are known to be active on 16S rRNA, but can also

94

target the mitochondrial protein synthesis. Elongation factor Tu is the putative target

for amythiamicin and was found by the observation of polysome formations.

Fosmidomycin has a target DOXP reductoisomerase, which was confirmed by the

inhibition of recombinant Pf enzyme.

In order to explain the activity of azithromycin in P. falciparum and make further

predictions on potential anti-malarial medicines, the apicoplast ribosomal 23S rRNA

exit tunnel had to be constructed. This was necessary because no crystallographic

information on apicoplast ribosomes is available. It is not possible to isolate the

apicoplast from the mitochondria in P. falciparum (both of them contain ribosomes).

The construction of the ribosomal exit tunnel of P. falciparum using D. radiodurans

ribosome crystal structure was the approach adopted in this study.

Azithromycin in D. radiodurans possesses two unexpected binding sites one of

which is located and has an orientation similar to erythromycin and direct contact is

existent between the second molecule and first molecule. The second molecule also

exhibits interaction with L4. It was found that the first azithromycin molecule has

interactions mostly with domains IV and V of 23S rRNA. On the contrary, the second

azithromycin molecule interacts with the ribosomal proteins L4 (hydrogen bonds to

Thr64 and Gly60) and L22 (hydrophobic interactions between Arg111 to both the

cladinose and desosamine of the second azithromycin molecule) and domain II of 23S

rRNA. Regarding sugars, the second azithromycin molecule makes a direct interaction

with the first azithromycin molecule through its desosamine sugar and O1 in the

lactone ring of the first azithromycin molecule.15

95

Figure 1: Interactions of azithromycin with ribosomal proteins L4 and L22 and 23S

rRNA15

96

Fidock et al. modelled the L4 segment from P. falciparum (Lys57 to Pro97) using

MODELLER and E. coli and D. radiodurans as templates. However, they did not

provide any data about the quality of the model.16

Azithromycin exhibits both mild anti-malarial activity and anti-bacterial activity

against Gram-negative bacteria. We wanted to investigate whether other macrolide

antibiotics possess anti-malarial activity. However, the problem was the crystal

structure of the exit tunnel of the apicoplast ribosomal exit tunnel from P. falciparum

(inability to separate apicoplast from mitochondria). The bacterial crystal structure of

D. radiodurans was used in order to do superpositions of the modelled P. falciparum

exit tunnel and D. radiodurans.

6.2. Results and Discussion6.2.1. Modelling on apicoplast-encoded P. falciparum L4 and nuclear genome-encoded P. falciparum L22 ribosomal proteins6.2.1.1. Construction of a segment of L4 ribosomal protein

1) Homology modelling

Ribosomal protein L4 from P. falciparum shows very little similarity with proteins

that have reported crystal structures, even in the protein region responsible for

macrolide activity against bacteria.

Here are the sequences of D. radiodurans and P. falciparum L4 ribosomal

proteins:

>D. radiodurans L4

AQINVIGQNGGRTIELPLPEVNSGVLHEVVTWQLASRRRGTASTRTRAQVS

TGRKMYGQKGTGNARHGDRSVPTFVGGGVAFGPKPRSYDYTLPRQVRQLGLAMAI

ASRQEGGKLVAVDGFDIADAKTKNFISWAKQNGLDGTEKVLLVTDDENTRRAARNVS

WVSVLPVAGVNVYDILRHDRLVIDAAALEIVEEEAGEEQQ

>P. falciparum L4

IIILNNNTLNNIIFKYKYNFFIKLYFNNYIKICKLIIYIIKYLFTYIYNIYMYKHTKNKS

VYFSNKKIRVQKGLGKARLKNFKSPVCKQGACNFGPFYKENKIFTISKINYRLIFVYLLI

NKRSNIIIIKLENIINLLNIFYKNKNYCIFKLLYLKGIINNKYIFTLINLNNKLFNKNIFINIIMY

NYLIFLI

There is very little similarity. For example, 5 Thr residues and 18 Tyr amino

acids in P. falciparum sequence may be observed, but 13 Thr and 3 Tyr amino acids in

D. radiodurans sequence can be seen. However, the same sequence of three amino-

acids in both residues (61-63 in D. radiodurans and 16-18 in P. falciparum) may be

observed:

97

Tyr-Gly-Gln-Lys-Gly-Thr D. radiodurans

59 60 61 62 63 64

Arg-Val-Gln-Lys-Gly-Leu P. falciparum

14 15 16 17 18 19

This was very important in the alignment of these two sequences.

Fidock et al. attempted to construct the L4 fragment (Lys57 to Pro97) from P.

falciparum. However, his group did not provide any data about the quality of the

modelled protein segment. He used homology modelling based on E. coli and D.

radiodurans. Using SWISS Model17 and E. coli18 and D. radiodurans19 as templates,

the whole L4 protein from P. falciparum was constructed. Both models show similar

structural characteristics. The model obtained using D. radiodurans as a template

analyzed by YASARA20 yielded 29.9% Helix, 11.4% Sheet, 9.6% Turn, 41.2% Coil and

0.0% 3-10 Helix. Whilst, the model obtained using E. coli as a template analyzed by

YASARA gave 24.2% Helix, 5.2% Sheet, 13.4% Turn, 50.5% Coil and 6.7% 3-10

Helix. The Ramachandran Z-score is a simple method to judge the quality of the model

based on the distribution of backbone dihedral angles. Based on the experience of the

developer of this method, if the Z-score is -4.0 or lower, the proposed structure is non-

satisfactory.21 The Ramachandran Z-scores obtained from these two models were low

(model obtained using D. radiodurans as a template has the value of -7.192 and model

using E. coli as a template has the value of -5.982) showing that they are not

satisfactory models (Figure 2).

A B

Figure 2: A An apicoplast ribosomal L4 protein of Plasmodium falciparum

obtained using SWISS model and D. radiodurans as a template; B An

apicoplast ribosomal L4 protein of Plasmodium falciparum obtained using

SWISS model and E. coli as a template

2) Ab initio

Ab initio calculations were performed using an I-TASSER server22 by

submitting the sequence of the protein to be modelled in FASTA format and using

the option for modelling without the template (ab initio). The whole sequence of L4

98

ribosomal protein from P. falciparum was used. The obtained model (Figure 3)

shows c-score=-2.26 and Ramachandran Z-score=-4.081. Again, the model was

not satisfactory because the Z-score was on the border of acceptability. Analysis of

the model by YASARA gave 28.9% Helix, 2.1% Sheet, 6.2% Turn, 60.8% Coil and

2.1% 3-10 Helix.

Figure 3: L4 protein from P. falciparum obtained using ab initio method on I-TASSER

In order to improve the modelling, ab initio modelling was used on the I-TASSER

server, an internet service for protein structure and function predictions, to design the

L4 ribosomal protein segment (Lys57 to Pro97) also previously modelled by Fidock’s

group. A model was obtained with a corresponding Ramachandran Z-score of -3.404,

which is satisfactory (Figure 4). However, it does not resemble the L4 ribosomal

protein from D. radiodurans.

Figure 4: Segment of L4 ribosomal protein from P. falciparum obtained using ab initio

method on I-TASSER

99

6.2.1.2. Construction of the L22 ribosomal protein1) Homology modelling

Nuclear genome-encoded P. falciparum ribosomal protein L22 from

Plasmodium falciparum shows more sequence similarities with the known crystal

structures, particularly in the region involved in macrolide binding to ribosomes (Figure

5).

Score = 33.1 bits (74), Expect = 1e-06, Method: Compositionalmatrix adjust.Identities = 26/104 (25%), Positives = 49/104 (47%), Gaps =1/104 (0%)

Query 9ATAKYIKMSFMKTRKILWKIRYMPIIKAFAFLYYYGTNKYTVNIYKCIKSCLHNAINKYG68

A AKY++MS K R ++ IR + A L + + + + K + SNA++Sbjct 26 AVAKYVRMSPRKVRLVVDVIRGKSVQDAEDLLRFIPRS-ASEPVAKVLNSAKANALHNDE 84

Query 69 RNNIKPVFHTLQANMGGYTKKINIRARGKTDIIREPHTHIRVVL 112+ + G K++ RARG +II++ +HI +++

Sbjct 85 MLEDRLFVKEAYVDAGPTLKRLIPRARGSANIIKKRTSHITIIV 128

Figure 5: Results from sequence alignment of segment of L22 protein from D.

radiodurans and segment L22 protein from P. falciparum using BLASTP 2.2.22+23 and

compositional matrix adjust24

The Swiss Model homology modelling was performed using L22 ribosomal

proteins from D. radiodurans and E. coli. The model obtained using D. radiodurans

gave as a result model (Figure 6) with 37.7% Helix, 11.4% Sheet, 9.6% Turn, 41.2%

Coil and 0.0% 3-10 Helix. The Ramachandran Z-score was -4.859. However,

homology modelling with the E. coli template did not give the protein of the whole

sequence.

Figure 6: L22 protein obtained using SWISS model and D. radiodurans as a template

100

2) Ab initio

Ab initio modelling on L22 ribosomal protein from P. falciparum was performed

using I-TASSER server. The obtained model (Figure 7) contained 39.5% Helix, 36.0%

Sheet, 0.0% Turn, 24.6% Coil and 0.0% Helix (YASARA analysis). It yielded a c-

score=0.88 and Ramachandran Z-score=1.684. On the basis of these parameters, it

was concluded that indeed a reliable model for the nuclear genome-encoded P.

falciparum ribosomal protein L22 was utilised (i.e. Z-score > -4.0).

Figure 7: L22 protein from P. falciparum obtained using ab initio method on I-TASSER

6.2.2. Modelling on apicoplast 23S rRNA from P. falciparumThere are previous modelling studies on mammalian mitochondrial ribosome

by Mears et al.25 They combined molecular modelling techniques with cryo-electronic

microscopic data. However, nobody has performed an in silico study on the

construction of the apicoplast ribosomal exit tunnel. To the best of our knowledge,

apicoplast has never been isolated separately from mitochondria in P. falciparum, so it

is impossible to get an X-ray image of its crystal structure. Comparison of the

sequence and secondary structure of P. falciparum and organisms with known crystal

structures yielded a significant similarity in domain V, but very small similarity in

sequence and base pairs of domains II and IV of 23S rRNA. The entire 23S rRNA was

modelled using the software RNA2D3D26. However, the obtained big structure could

not be refined using the software, so the sequence of 23S rRNA from P. falciparum

was divided into two separate files. The obtained modelling was performed using the

input files containing sequence and information about base pairs obtained on the basis

of the secondary structure of P. falciparum. The final models (Figure 8) appeared to be

101

satisfactory, but there were concerns about the folding because of the mutual

influence of one residue on other.

A B

Figure 8: A Model of large subunit ribosomal RNA-3’ half of P. falciparum; BModel of large subunit ribosomal RNA-5’ half of P. falciparum

6.2.3. Construction of the exit tunnel of the apicoplast ribosomal exit tunnel ofPlasmodium falciparum

The model of the apicoplast ribosomal exit tunnel was constructed using the

modelled fragment of L4 protein, whole L22 protein, parts of domains II, IV and V of

23S rRNA. Superposition was performed using D. radiodurans as a template and

Pymol software25 for the superposition of RNA residues and MOE software27 for the

superposition of proteins. The obtained model (Figure 9) of the apicoplast ribosomal

exit tunnel shows different folding of RNA compared to D. radiodurans. One of the

reasons for the different folding can be explained by the large number of differences in

domains II and IV in 23S rRNA of P. falciparum compared to D. radiodurans or

imperfection of the built RNA model.

Figure 9: Model of the apicoplast ribosomal exit tunnel of P. falciparum

Obtained model of the apicoplast ribosomal exit tunnel from P. falciparum

shows only binding of one molecule of azithromycin (Figure 10).

102

Figure 10: Model of the apicoplast ribosomal exit tunnel from P. falciparum binding one molecule of azithromycin

103

6.3. DiscussionThe model is difficult to construct and harder to test. Prior to this study, there

has been no reported attempt at the construction of the exit tunnel of the apicoplast

ribosome from Plasmodium falciparum. There is a published paper on the construction

of the mammalian ribosome, but using a combination of experimental and modelling

techniques.

In order to test the quality of the protein models (apicoplast-encoded P.

falciparum L4 and nuclear genome-encoded P. falciparum L22 apicoplast ribosomal

proteins) the Ramachandran Z-score was utilised. The results demonstrated that the

homology modelling on the ribosomal proteins from P. falciparum had less similarities

to the templates (D. radiodurans and E. coli) and all known and available crystal

structures of L4 and L22 ribosomal proteins was less successful than ab initio

modelling of the same structures. The modelled segment of apicoplast-encoded P.

falciparum L4 ribosomal protein (Lys57 to Pro97 in D. radiodurans) and nuclear

genome-encoded P. falciparum L22 ribosomal protein using ab initio method represent

reliable models on the basis of their Ramachandran Z-scores which could be useful for

the construction of the exit tunnel of the apicoplast ribosome from P. falciparum.

Modelling of the 23S rRNA presented an additional challenge. In order to

model the sequence, due to its length the sequence had to be divided into two halves.

Modelling was subsequently performed and refinement and energy minimization of the

two halves was performed separately. Consequently, a problem arose as to whether

the folding of the two halves was indeed appropriate or not. Unfortunately, there is no

method similar to protein methods for the checking of the quality of generated

structures.

To the best of our knowledge, there is no available quality control approach

for the generated model of the exit tunnel of the apicoplast ribosome from P.

falciparum. The following good features were observed: 1) It is the first model of the

ribosomal exit tunnel of P. falciparum and 2) L22 Ribosomal protein is predicted

correctly, L4 represents the best modelled L4 ribosomal protein from P. falciparum, to

the best of our knowledge. Due to a lack of crystal data and appropriate programs for

checking the model, we cannot validate and ascertain the relevance and accuracy of

the model.

Clearly the model would bind clarithromycin which has no anti-malarial activity.

The flexibility in the azithromycin molecule that enables gram negative activity is

probably also important here.

104

6.4. ExperimentalHomology modelling using the SWISS Model server17

Homology modelling using the SWISS Model server (Project mode under the

Modelling section of the SWISS model page) was performed by submitting the

sequence of the target sequences of proteins (apicoplast-encoded P. falciparum L4

and nuclear genome-encoded P. falciparum L22) and template pdb files of L22 and L4

from D. radiodurans and E. coli extracted using Molecular Viewer.

Ab initio molecular modelling using I-TASSER server22

Ab initio molecular modelling using I-TASSER server of L4 and L22 ribosomal

proteins L4 and L22 from P. falciparum was performed by submitting the sequences of

these two proteins and using the option of modelling without the template.

Modelling of apicoplast 23S rRNA using RNA_2D3D26

The molecular modelling of apicoplast 23S rRNA from P. falciparum was

performed making the two separate input files containing the RNA sequences of 5’ and

3’-half of 23S rRNA and information of the base pairs obtained from the secondary

structure of apicoplast 23S rRNA from P. falciparum. The obtained crude models were

refined according to the manual given by the creators of the software RNA_2D3D24.

Energy refinement was performed using the TINKER software available in

RNA_2D3D. Initially, it was generated one input file containing the whole sequence of

23S rRNA and information about the base pairs from the whole sequence. However,

the software could not refine and perform energy minimization of such a big structure.

Energy and structure refinement were tried using MOE, but without any significant

success. So, the structure was divided and two separate models generated.

PyMOL27 superposition of 23S rRNA of D. radiodurans domains and constructed

domains of apicoplast 23S rRNA from P. falciparum

As a template for the alignment, the D. radiodurans moiety containing L4, L22

ribosomal proteins, 23S rRNA and two molecules of azithromycin were examined.

Firstly, separate alignments were performed using Pymol software of modelled

domains II, IV and V. The smallest RMS was shown with the alignment of domain IV

(RMS=14.528) using the align command and specifying the nucleotides to be aligned

(It was chosen on the same base according to A. Yonath et al.15 are in contact with

azithromycin and responsible for the binding of azithromycin) of domains IV with the

corresponding template of D. radiodurans. Using MOE, it was deleted from the file

containing the domain IV from P. falciparum and the template of D. radiodurans the

105

nucleotides corresponding to domain IV in the D. radiodurans template. This file was

used after where reduced template of D. radiodurans was further aligned to the

modelled domain II from P. falciparum as previously described before also adding

some nucleotides as restrictions for alignment (D. radiodurans: U716, A720, U727,

A729, A731, G742; P. falciparum: U657, A661, U668, A670, A672, G683)

(RMS=11.475). Following alignment from the reduced template of D. radiodurans,

nucleotides belonging to domain II from the file were deleted using MOE. The file with

reduced template of D. radiodurans was used for the alignment with the modelled

domain V from P. falciparum (RMS=36.293). From the generated alignment file,

domain V from D. radiodurans was deleted and the file once more saved for the further

superpositions of ribosomal proteins L4 and L22 from P. falciparum with L4 and L22

from D. radiodurans using MOE.

MOE28 superposition of segment of L4 and the whole L22 from P. falciparum and L4

and L22 proteins from D. radiodurans

MOE Superposition was performed using the file containing domains II, IV and

V from P. falciparum and template of D. radiodurans containing only L4 and L22

ribosomal proteins and two molecules of azithromycin. The first alignment of L22 from

the mentioned file with L22 ribosomal protein from P. falciparum was used with the

pre-selected group options by Chain Tag and Synchronize Selection (RMSD= 2.347

Å). The second alignment with the same components was performed using above-

mentioned Optimize Gap penalties for Superposition. The RMSD was smaller in the

second case (RMSD= 2.031 Å). However, there was no alignment of the conserved

regions in D. radiodurans and P. falciparum in the second case. From the alignment

files, it was removed the used chain Q for the alignment from D. radiodurans. Both

files containing no more chain Q were submitted for superposition with the fragment L4

from P. falciparum with two different options as in the above-mentioned case (2.879 Å

and 2.384 Å, respectively). From the obtained four files, the protein L4 was removed

from D. radiodurans and obtained files contained only fragment L4 and L22 ribosomal

proteins, domains II, IV and V from P. falciparum and two molecules of azithromycin

from the D. radiodurans template. Removal of two azithromycin molecules in MOE

gave the exit tunnel of the apicoplast ribosome.

6.5. ConclusionThe model of the exit tunnel of the apicoplast exit tunnel from P. falciparum

was created using different computational approaches and softwares in order to create

it in silico. The L22 ribosomal protein and the segment of L4 ribosomal protein were

modelled using both homology modelling and ab initio. Ab initio modelling gave better

106

models with acceptable Ramachandran Z-scores. The challenging task entailing

modelling of the long sequence of RNA was performed. In the presented model there

is uncertainty surrounding the folding of the RNA and its appropriateness due to a non-

existence of methods for quality checking of the model. The model can predict only

one binding site for the azithromycin. The next aim will involve docking of different

compounds into the created apicoplast ribosomal exit tunnel of P. falciparum. If some

medicines can be anti-malarial drugs, we can prove by testing them, at the same time

managing to test the correctness of the proposed model.

6.6. References1 S. A. Ralph, M. C. D’Ombrain and G. I. McFadden: The apicoplast as an antimalarialdrug target, Drug Resist. Update, 4, 145, 2001.2 N. J. White: A vaccine for malaria, N. Engl. J. Med., 365, 1926, 2011.3 M. T. Gleeson: The plastid in Apicomplexa: what use is it? Int. J. Parasitol., 30, 1053,2000.4 C. Mereschkowsky: Uber Natur und Ursprung der Chromatophoren imPflanzenreiche. Biol. Zentralbl., 25, 593, 1905; C. Mereschkowsky, Theorie der zweiPlasmaarten als Grundlage der Symbiogenesis, einer neuen Lehre von der Enstehungder Organismen. Biol. Zentralbl., 30, 278, 1910.5 R. Gozalbes, M. Brun-Pascaud and R. Garcia-Domenech: Anti-Toxoplasma activitiesof 24 quinolones and fluoroquinolones in vitro: Prediction of activity by moleculartopology and virtual computational techniques, Antimicrob. Agents Chemother., 44,2771, 2000.6 S. Pukrittayakamee, C. Viravan and P. Charoenlarp: Antimalarial effects of rifampinin Plasmodium vivax malaria, Antimicrob. Agents. Chemother, 38, 511, 1994.7 E. R. Pfefferkorn and S. E. Borotz: Comparison of mutants of Toxoplasma gondiiselected for resistance to azithromycin, spiramycin, or clindamycin, Antimicrob. AgentsChemotherapy, 338, 31, 1994.8 G. A. McConkey, M. J. Rogers and T. F. McCutchan: Inhibition of Plasmodiumfalciparum protein synthesis: targeting the plastid-like organelle with thiostrepton, J.Biol. Chem., 272, 2046, 1997.9 M. J. Rogers, E. Cundliffe and T. F. McCutchan: The antibiotic micrococcin is apotent inhibitor of growth and protein synthesis in the malaria parasite, Antimicrob.Agents Chemoth., 42, 715, 1998.10 B. Clough, K. Rangachari, M. Strath, P. R. Preiser and R. Wilson: Antibioticinhibitors of organellar protein synthesis in Plasmodium falciparum, Protist, 150, 189,1999.11 K. M. Woods, M. V. Nesterenko and S. J. Upton: Efficacy of 101 antimicrobials andother agents on the development of Cryptosporidium parvum in vitro, Ann. Trop. Med.Parasitol., 90, 603, 1996.12 F. Roberts, C. Roberts and J. Johnson: Evidence for the shikimate pathwayapicomplexan parasites, Nature, 393, 801, 1998.13 H. Jomaa, J. Wiesner and S. Sanderbrand: Inhibitors of the nonmevalonate pathwayof isoprenoid biosynthesis as antimalarial drugs, Science, 285, 1573, 1999.14 N. Surolia and A. Surolia: Triclosan offers protection against blood stages of malariaagainst blood stages of malaria by inhibiting enoyl-ACP reductase of Plasmodiumfalciparum, Nat. Med., 7, 167, 2001.15 F. Schlunzen, J. Harms, F. Franceschi, H. Hansen, H. Bartels, R. Zarivach and A.Yonath: Structural basis for the antibiotic activity of ketolides and azalides, Structure,11, 328, 2003.16 A. B. S. Sidhu, Q. Sun, L.J. Nkrumah, M. W. Dunne, J. C. Sacchettini and D. A.Fidock: In vitro efficacy, resistance selection, and structural modelling studies implicate

107

the malarial parasite apicoplast as the target of azithromycin, J. Biol. Chem., 282,2494, 2007.17 K. Arnold, L. Bordoli, J. Kopp and T. Schwede: The SWISS-MODEL Workspace: Aweb-based environment for protein structure homology modelling, Bioinformatics,22,195, 2006; F. Kiefer, K. Arnold, M. Künzli, L. Bordoli, T. Schwede: The SWISS-MODEL Repository and associated resources, Nucleic Acids Research, 37, D387,2009; M. C. Peitsch: Protein modeling by E-mail, Bio/Technol., 13, 658, 1995.18 K. Mitra, C. Schaffitzel, F. Fabiola, M. S. Chapman, N. Ban and J. Frank: Elongationarrest by SecM via a cascade of ribosomal RNA rearrangements, Mol. Cell, 22, 533,2006.19 J. M. Harms, D. N. Schlunzen, S. R. Connell, T. Stachelhous, Z. Zaborowska, C. M.Spahn and P. Fucini: Translational regulation via L11: molecular switches on theribosome turned on and off by thiostrepton and micrococcin, Mol. Cell, 30, 26, 2008.20 www.yasara.org21 R. W. W. Hooft, C. Sander and G. Vriend: Objectively judging the quality of a proteinstructure from a Ramachandran plot, CABIOS, 13, 425, 1997.22 R. Ambrish, K. Alper and Z. Yang: I-TASSER: a unified platform for automatedprotein structure and function prediction, Nature Protocols, 5, 725, 2010; R. Ambrish,X. Dong, P. Jonathan and Z. Yang: A Protocol for Computer-Based Protein Structureand Function Prediction, Journal of Visualized Experiments, 57, 3259, 2011; Z. Yang:I-TASSER server for protein 3D structure prediction, BMC Bioinformatics, 9, 40, 2008.23 S. F. Altschul, J. C. Wootton, E. M. Gertz, R. Agarwala, A. Morgulis, A. A. Schafferand Y-K. Yu: Protein database searches using compositionally adjusted substitutionmatrices, FEBS J., 272, 5101, 2005.24 S. F. Altschul, T. L. Madden, A. A. Schaffer, J. Zhang, Z. Zhang, W. Miller and D. J.Lipman: Gapped BLAST and PSI-BLAST: a new generation of protein databasesearch programs, Nucleic Acids Res., 25, 3389, 1997.25 J. A. Mears, M. R. Sharma, R.R. Gutell, A. S. McCook, P. E. Richardson, T. R.Caulfield, R. K. Agrawal and S. C. Harvey: A structural model for the large subunit ofthe mammalian mitochondrial ribosome, J. Mol. Biol., 358, 193, 2006.26 http://www-lmmb.ncifcrf.gov/~bshapiro/software.html27 http://www.pymol.org/28 http://www.chemcomp.com/software.htm

108

III Conclusion

Macrolide antibiotics show activity against bacteria. It was observed that 14-

membered macrolide antibiotics have activity against Gram-positive bacteria. More flexibility

is present in erythromycin A than its 6-O-methyl derivative, clarithromycin. Azithromycin, 15-

membered macrolide antibiotic, is flexible and active against Gram-negative bacteria. We

have found that it is possible to predict anti-bacterial activity of 14-, 15- and 16-membered

macrolide antibiotics using molecular modelling and construct new macrolide antibiotics with

desired properties. Flexibility was found to be responsible for the activity of macrolide

antibiotics against Gram-negative organisms.

Assignments of 16-memebered macrolide antibiotics, tylosin A and tylosin B, was

shown to be extremely difficult due to close chemical shifts and existence of different forms

of tylosin A in aqueous solutions. To the best of our knowledge, nobody before tried to

assign tylosin A in D2O. Accurate assignments of tylosin A and tylosin B are essential for the

further exploration of their properties. Their anti-bacterial activity was shown to be weaker

compared to 14- and 15-membered macrolide antibiotics, clarithromycin and azithromycin.

Azithromycin, flexible 15-membered macrolide antibiotic and tylosin A, rigid 16-membered

macrolide antibiotic have anti-malarial activity. The alignment of tylosin A and two molecules

of azithromycin gave that tylosin A occupy the space of two molecules of azithromycin and

possible these argument together with similar binding behaviour can be used for the

explanation of its anti-malarial activity.

Even today, it is not possible to isolate separately apicoplast from mitochondria in

Plasmodium falciparum and both of them contain ribosomes. In order to construct the exit

tunnel of the apicoplast ribosome from P. falciparum, we had to construct apicoplast-

encoded P. falciparum L4 and nuclear genome-encoded P. falciparum L22 ribosomal

proteins and apicoplast ribosomal 23S rRNA in silico. Apicoplast ribosomal L4 was

constructed using both homology modelling and ab initio server. Ramachandran Z- score

was used for the judgement about the quality of the model. Models obtained using ab initio

server show more reliability than those obtained using homology modelling. Because of the

unsatisfactory value for Ramachandran Z-score, we constructed and used in further

construction of the exit tunnel one segment of apicoplast ribosomal L4 protein, important for

azithromycin binding in bacteria. Apicoplast ribosomal 23S rRNA was made using the

109

software RNA_2D3D (input file contained both sequences and the information about base

pairing obtained from its secondary structure) where it was performed the refinement. The

model of the exit tunnel of Plasmodium falciparum was gotten using constructed building

blocks and the relevant X-ray crystal structure of D. radiodurans with azithromycin as

template. Superposition procedure performed using Pymol and MOE gave the model of the

exit tunnel of the apicoplast ribosome from P. falciparum which can bind one azithromycin

molecule. This is the first model of the exit tunnel from apicoplast ribosome from

Plasmodium falciparum. The model is extremely difficult to make and better model in this

moment can not be made. The performed work can be extended by docking of different

macrolide antibiotics into the constructed exit tunnel.

110

IV Appendix

Appendix 1

Appendix 1 consists of published papers or papers in preparation made as a

collaboration with other researchers from different research groups from The School of

Pharmacy and Pharmaceutical Sciences or different countries (India, Republic of

Serbia) (supporting_file.pdf).

The content of this section is available as a CD attached to the back cover of

the thesis.

111

Appendix 2

1D and 2D NMR spectra of descladinosyl clarithromycin

Figure A1: 1D 1H spectrum of descladinosyl clarithromycin in CDCl3 at 300 MHz

112

Figure A2: 1D 13C spectrum of descladinosyl clarithromycin in CDCl3 at 300 MHz

113

Figure A3: DEPT 45 spectrum of descladinosyl clarithromycin in CDCl3 at 300 MHz

114

Figure A4: DEPT 90 spectrum of descladinosyl clarithromycin in CDCl3 at 300 MHz

115Figure A5: DEPT 135 spectrum of descladinosyl clarithromycin in CDCl3 at 300 MHz

116

Figure A6: 2D COSY spectrum of descladinosyl clarithromycin spectrum in CDCl3 at 300 MHz

117

Figure A7: 2D HMBC spectrum of descladinosyl clarithromycin in CDCl3 at 300 MHz

118

Figure A8: 2D HMQC spectrum of descladinosyl clarithromycin in CDCl3 at 300 MHz

119

Figure A9: 2D HSQC spectrum of descladinosyl clarithromycin in CDCl3 at 400 MHz

120

1D and 2D NMR spectra of descladinosyl azithromycin(a)

121

(b)

Figure A10: (a) and (b) 1D 1H spectrum of descladinosyl azithromycin in CDCl3 at 300 MHz

122

Figure A11: 1D 13C spectrum of descladinosyl azithromycin in CDCl3 at 300 MHz

123

Figure A12: DEPT 45 spectrum of descladinosyl azithromycin in CDCl3 at 300 MHz

124


125


126

Figure A15: 2D DQF COSY spectrum of descladinosyl azithromycin in CDCl3 at 500 MHz

127Figure A16: 2D HMBC spectrum of descladinosyl azithromycin in CDCl3 at 300 MHz

128Figure A17: 2D HSQC spectrum of descladinosyl azithromycin in CDCl3 at 500 MHz

129

Figure A18: 2D ROESY spectrum of descladinosyl azithromycin in CDCl3 at 500 MHz

130

1D and 2D NMR spectra of tylosin A in phosphate buffered D2O, apparent pH=7

Figure A19: 1D 1H spectrum of tylosin A in phosphate buffered D2O (apparent pH=7) at 500 MHz

131

Figure A20: 1D 13C spectrum of tylosin A in phosphate buffered D2O, apparent pH=7 at 500 MHz

132

Figure A21: DEPT 90 spectrum of tylosin A in phosphate buffered D2O at 500 MHz

133

Figure A22: DEPT 135 spectrum of tylosin A in phosphate buffered D2O at 500 MHz

134

Figure A23: 2D COSY spectrum of tylosin A in phosphate buffered D2O at 400 MHz

135

Figure A24: 2D HMBC spectrum of tylosin A in phosphate buffered D2O at 400 MHz

136

Figure A25: 2D HSQC spectrum of tylosin A in phosphate buffered D2O (apparent pH=7) at 400 MHz

137

Figure A26: 2D TOCSY spectrum of tylosin A in phosphate buffered D2O (apparent pH=7) at 400 MHz

138

Figure A27: 2D ROESY spectrum of tylosin A in phosphate buffered D2O (apparent pH=7) at 400 MHz

139

1D and 2D spectra of tylosin A in CDCl3

Figure A28: 1D 1H spectrum of tylosin A in CDCl3 at 500 MHz

140

Figure A29: 1D 13C spectrum of tylosin A in CDCl3 at 500 MHz

141

Figure A30: DEPT 90 spectrum of tylosin A in CDCl3 at 500 MHz

142

Figure A31: DEPT 135 spectrum of tylosin A in CDCl3 at 500 MHz

143

Figure A32: 2D COSY spectrum of tylosin A in CDCl3 at 500 MHz

144

Figure A33: 2D HMBC spectrum of tylosin A in CDCl3 at 500 MHz

145

Figure A34: 2D HMQC spectrum of tylosin A in CDCl3 at 500 MHz

146

Figure A35: 2D ROESY spectrum of tylosin A in CDCl3 at 500 MHz

147

1D and 2D spectra of tylosin B in phosphate buffered D2O (apparent pH=7)

Figure A36: 1D 1H spectrum of tylosin B in phosphate buffered D2O (apparent pH=7) at 500 MHz

148

Figure A37: 1D 13C spectrum of tylosin B in phosphate buffered D2O (apparent pH=7) at 500 MHz

149

Figure A38: DEPT 90 spectrum of tylosin B in phosphate buffered D2O (apparent pH=7) at 500 MHz

150

Figure A39: DEPT 135 spectrum of tylosin B in phosphate buffered D2O (apparent pH=7) at 500 MHz

151Figure A40: 2D HMBC spectrum of tylosin B in phosphate buffered D2O (apparent pH=7) at 500 MHz

152

Figure A41: 2D HSQC spectrum of tylosin B in phosphate buffered D2O (apparent pH=7) at 500 MHz

153

1D and 2D spectra of tylosin B in CDCl3

(a)

154

(b)

Figure A42: 1D 1H spectrum of tylosin B in CDCl3 at 300 MHz

155

Figure A43: 1D 13C spectrum of tylosin B in CDCl3 at 300 MHz

156

Figure A44: DEPT 90 spectrum of tylosin B in CDCl3 at 300 MHz

157

Figure A45: DEPT 135 spectrum of tylosin B in CDCl3 at 300 MHz

158

Figure A46: 2D HMBC spectrum of tylosin B in CDCl3 at 300 MHz

159

Figure A47: 2D HMQC spectrum of tylosin B in CDCl3 at 300 MHz

160

Acquisition and processing NMR parameters1) tylosin A in phosphate buffered D2O (apparent pH=7)

Current Data ParametersNAME B_tyl_1xi07EXPNO 10PROCNO 1

F2 - Acquisition ParametersDate_ 20071101Time 9.20INSTRUM spectPROBHD 5 mm BBI 1H/D-PULPROG zg30TD 65536SOLVENT D2ONS 16DS 2SWH 8223.685 HzFIDRES 0.125483 HzAQ 3.9846387 secRG 71.8DW 60.800 usecDE 6.00 usecTE 303.7 KD1 2.40000010 secTD0 1

======== CHANNEL f1 ========NUC1 1HP1 7.79 usecPL1 0.40 dBSFO1 400.1318761 MHz

F2 - Processing parametersSI 32768SF 400.1300000 MHzWDW EMSSB 0LB 0.30 HzGB 0PC 1.00


F2 - Acquisition ParametersDate_ 20071101Time 17.38INSTRUM spectPROBHD 5 mm BBI 1H/D-PULPROG hmbcgplpndqfTD 4096SOLVENT D2ONS 24DS 16SWH 3378.378 HzFIDRES 0.824799 HzAQ 0.6062580 secRG 16400DW 148.000 usecDE 6.50 usecTE 303.0 KCNST13 10.0000000CNST2 145.0000000d0 0.00000300 secD1 1.28700805 secD16 0.00020000 secd2 0.00344828 secd6 0.05000000 secIN0 0.00002240 sec

======== CHANNEL f1 ========NUC1 1HP1 7.79 usecp2 15.58 usecPL1 0.40 dBSFO1 400.1314726 MHz

======== CHANNEL f2 ========NUC2 13CP3 13.90 usecPL2 -4.00 dBSFO2 100.6228138 MHz

====== GRADIENT CHANNEL =====GPNAM1 SINE.100GPNAM2 SINE.100GPNAM3 SINE.100GPZ1 50.00 %GPZ2 30.00 %GPZ3 40.10 %P16 1000.00 usec

F1 - Acquisition parametersND0 2TD 128SFO1 100.6228 MHzFIDRES 174.386154 HzSW 221.833 ppmFnMODE QF

F2 - Processing parametersSI 8192SF 400.1300000 MHzWDW SINESSB 1LB 0.00 HzGB 0PC 1.40

F1 - Processing parametersSI 512MC2 QFSF 100.6127690 MHzWDW SINESSB 1LB 0.00 HzGB 0

161


F2 - Acquisition ParametersDate_ 20071101Time 9.26INSTRUM spectPROBHD 5 mm BBI 1H/D-PULPROG cosygpmfprgpqf.ptgTD 2048SOLVENT D2ONS 8DS 8SWH 4157.428 HzFIDRES 2.029994 HzAQ 0.2463561 secRG 912DW 120.267 usecDE 6.00 usecTE 303.1 KD0 0.00000300 secD1 2.00000000 secd13 0.00000400 secD16 0.00020000 secIN0 0.00024030 sec

======== CHANNEL f1 ========NUC1 1HP1 7.79 usecPL1 0.40 dBPL9 56.55 dBSFO1 400.1318770 MHz

====== GRADIENT CHANNEL =====GPNAM1 SINE.100GPNAM2 SINE.100GPNAM3 SINE.100GPNAM4 SINE.100GPZ1 16.00 %GPZ2 12.00 %GPZ3 40.00 %GPZ4 34.17 %P16 1000.00 usec


F2 - Processing parametersSI 4096SF 400.1300000 MHzWDW QSINESSB 1LB 0.00 HzGB 0PC 1.40

F1 - Processing parametersSI 2048MC2 QFSF 400.1300000 MHzWDW QSINESSB 1LB 0.00 HzGB 0


F2 - Acquisition ParametersDate_ 20071101Time 14.52INSTRUM spectPROBHD 5 mm BBI 1H/D-PULPROG hsqcetgpsi2TD 1024SOLVENT D2ONS 24DS 16SWH 3378.378 HzFIDRES 3.299198 HzAQ 0.1516020 secRG 18400DW 148.000 usecDE 6.50 usecTE 303.2 KCNST2 145.0000000d0 0.00000300 secD1 1.44449902 secd11 0.03000000 secd13 0.00000400 secD16 0.00020000 secD24 0.00086207 secd4 0.00172414 secDELTA 0.00127158 secDELTA1 0.00120800 secDELTA2 0.00006207 secDELTA3 0.00052414 secIN0 0.00003000 secST1CNT 128ZGOPTNS

======== CHANNEL f1 ========NUC1 1HP1 7.79 usecp2 15.58 usecP28 0.10 usecPL1 0.40 dBSFO1 400.1314726 MHz

======== CHANNEL f2 ========CPDPRG2 garpNUC2 13CP3 13.90 usecp4 27.80 usecPCPD2 70.00 usecPL12 10.04 dBPL2 -4.00 dBSFO2 100.6203124 MHz

====== GRADIENT CHANNEL =====GPNAM1 SINE.100GPNAM2 SINE.100GPNAM3 SINE.100GPNAM4 SINE.100GPZ1 80.00 %GPZ2 20.10 %GPZ3 11.00 %GPZ4 -5.00 %P16 1000.00 usecP19 600.00 usec

F1 - Acquisition parametersND0 2TD 256SFO1 100.6203 MHzFIDRES 65.104164 HzSW 165.639 ppmFnMODE Echo-Antiecho


F1 - Processing parametersSI 1024MC2 echo-antiechoSF 100.6069187 MHzWDW QSINESSB 2LB 0.00 HzGB 0

162


F2 - Acquisition ParametersDate_ 20071101Time 19.21INSTRUM spectPROBHD 5 mm BBI 1H/D-PULPROG roesyphpr.2TD 2048SOLVENT D2ONS 48DS 8SWH 4401.409 HzFIDRES 2.149125 HzAQ 0.2327028 secRG 181DW 113.600 usecDE 6.00 usecTE 303.1 Kd0 0.00010464 secD1 2.00000000 secd11 0.03000000 secd12 0.00002000 secd13 0.00000400 secIN0 0.00022720 secl4 400P15 200000.00 usecST1CNT 64

======== CHANNEL f1 ========NUC1 1HP1 7.79 usecP25 250.00 usecPL1 0.40 dBPL27 24.51 dBPL9 56.55 dBSFO1 400.1318770 MHz

F1 - Acquisition parametersND0 1TD 512SFO1 400.1319 MHzFIDRES 8.596501 HzSW 11.000 ppmFnMODE States-TPPI


F1 - Processing parametersSI 1024MC2 States-TPPISF 400.1300000 MHzWDW QSINESSB 2LB 0.00 HzGB 0

Current Data ParametersNAME 2012-07-02-evb-3EXPNO 11PROCNO 1

F2 - Acquisition ParametersDate_ 20120702Time 13.31INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG zgpg30TD 65536SOLVENT H2O+D2ONS 14336DS 2SWH 29761.904 HzFIDRES 0.454131 HzAQ 1.1010548 secRG 512DW 16.800 usecDE 32.78 usecTE 293.0 KD1 2.00000000 secD11 0.03000000 secTD0 1

======== CHANNEL f1 ========NUC1 13CP1 8.80 usecPL1 -1.00 dBPL1W 104.35516357 WSFO1 125.7703643 MHz

======== CHANNEL f2 ========CPDPRG2 waltz16NUC2 1HPCPD2 80.00 usecPL2 0.00 dBPL12 17.23 dBPL13 120.00 dBPL2W 25.62124252 WPL12W 0.48484197 WPL13W 0.00000000 WSFO2 500.1320005 MHz


163


F2 - Acquisition ParametersDate_ 20120703Time 7.57INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG dept90_mTD 65536SOLVENT H2O+D2ONS 4914DS 4SWH 29761.904 HzFIDRES 0.454131 HzAQ 1.1010548 secRG 512DW 16.800 usecDE 6.00 usecTE 293.0 KCNST2 145.0000000D1 2.00000000 secD2 0.00344828 secD12 0.00002000 secTD0 1

======== CHANNEL f1 ========NUC1 13CP1 8.80 usecP2 17.60 usecPL1 -1.00 dBPL1W 104.35516357 WSFO1 125.7703643 MHz

======== CHANNEL f2 ========CPDPRG2 waltz16NUC2 1HP3 11.00 usecP4 22.00 usecPCPD2 80.00 usecPL2 0.00 dBPL12 17.23 dBPL2W 25.62124252 WPL12W 0.48484197 WSFO2 500.1320005 MHz







164

2) descladinosyl clarithromycin in CDCl3

Current Data ParametersNAME sp2EXPNO 11PROCNO 1

F2 - Acquisition ParametersDate_ 20090523Time 1.12INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG zgpg30TD 65536SOLVENT CDCl3NS 5000DS 4SWH 18832.393 HzFIDRES 0.287360 HzAQ 1.7400308 secRG 16384DW 26.550 usecDE 6.50 usecTE 297.3 KD1 2.00000000 secd11 0.03000000 secDELTA 1.89999998 secTD0 1

======== CHANNEL f1 ========NUC1 13CP1 7.80 usecPL1 8.00 dBSFO1 75.4764278 MHz

======== CHANNEL f2 ========CPDPRG2 waltz16NUC2 1HPCPD2 100.00 usecPL12 21.94 dBPL13 120.00 dBPL2 0.00 dBSFO2 300.1312005 MHz



F2 - Acquisition ParametersDate_ 20090523Time 3.26INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG dept45TD 65536SOLVENT CDCl3NS 2048DS 4SWH 17985.611 HzFIDRES 0.274439 HzAQ 1.8219508 secRG 16384DW 27.800 usecDE 6.50 usecTE 297.4 KCNST2 145.0000000D1 2.00000000 secd12 0.00002000 secd2 0.00344828 secDELTA 0.00000993 secTD0 1

======== CHANNEL f1 ========NUC1 13CP1 7.80 usecp2 15.60 usecPL1 8.00 dBSFO1 75.4752953 MHz

======== CHANNEL f2 ========CPDPRG2 waltz16NUC2 1HP3 8.00 usecp4 16.00 usecPCPD2 100.00 usecPL12 21.94 dBPL2 0.00 dBSFO2 300.1312005 MHz


165











166


F2 - Acquisition ParametersDate_ 20090523Time 8.02INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG cosydfetgpprf2.2TD 2048SOLVENT CDCl3NS 48DS 8SWH 2620.545 HzFIDRES 1.279563 HzAQ 0.3908084 secRG 9195.2DW 190.800 usecDE 6.00 usecTE 297.2 Kd0 0.00000300 secD1 1.97706294 secd11 0.03000000 secD13 0.00000400 secD16 0.00020000 secDELTA 0.00120300 secDELTA1 0.00120800 secIN0 0.00038160 secST1CNT 64


======== CHANNEL f2 ========NUC2 1HPL2 0.00 dBPL21 55.92 dBSFO2 300.1307124 MHz






F2 - Acquisition ParametersDate_ 20090524Time 12.10INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG hmbcgpndqfTD 4096SOLVENT CDCl3NS 72DS 16SWH 2688.172 HzFIDRES 0.656292 HzAQ 0.7619060 secRG 23170.5DW 186.000 usecDE 6.50 usecTE 297.4 KCNST13 8.0000000d0 0.00000300 secD1 1.24932504 secD16 0.00020000 secd6 0.06250000 secIN0 0.00002980 sec







167


F2 - Acquisition ParametersDate_ 20090525Time 18.30INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG roesyphprf2.2TD 2048SOLVENT CDCl3NS 96DS 8SWH 2670.940 HzFIDRES 1.304170 HzAQ 0.3834356 secRG 322.5DW 187.200 usecDE 6.00 usecTE 297.9 Kd0 0.00017792 secD1 1.98525500 secd11 0.03000000 secd12 0.00002000 secd13 0.00000400 secIN0 0.00037440 secl4 400P15 200000.00 usecST1CNT 64

======== CHANNEL f1 ========NUC1 1HP1 8.30 usecP25 250.00 usecPL1 0.00 dBPL27 23.56 dBSFO1 300.1310821 MHz

======== CHANNEL f2 ========NUC2 1HPL2 0.00 dBPL21 55.92 dBSFO2 300.1307116 MHz





F2 - Acquisition ParametersDate_ 20090526Time 3.29INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG hmqcgpqfTD 1024SOLVENT CDCl3NS 96DS 16SWH 2796.421 HzFIDRES 2.730880 HzAQ 0.1831412 secRG 26008DW 178.800 usecDE 6.50 usecTE 297.7 KCNST2 145.0000000d0 0.00000300 secD1 1.48504901 secd12 0.00002000 secd13 0.00000400 secD16 0.00020000 secd2 0.00344828 secDELTA1 0.00222428 secIN0 0.00004000 sec


======== CHANNEL f2 ========CPDPRG2 garpNUC2 13CP3 8.30 usecPCPD2 65.00 usecPL12 28.00 dBPL2 8.00 dBSFO2 75.4734083 MHz





168

Current Data ParametersNAME dc_pr_chEXPNO 13PROCNO 1

F2 - Acquisition ParametersDate_ 20091216Time 1.22INSTRUM spectPROBHD 5 mm BBI 1H/D-PULPROG hsqcetgpsi2TD 1024SOLVENT CDCl3NS 72DS 16SWH 2688.172 HzFIDRES 2.625168 HzAQ 0.1905140 secRG 2050DW 186.000 usecDE 6.50 usecTE 298.4 KCNST2 145.0000000d0 0.00000300 secD1 1.40558696 secd11 0.03000000 secd13 0.00000400 secD16 0.00020000 secD24 0.00086207 secd4 0.00172414 secDELTA 0.00127158 secDELTA1 0.00120800 secDELTA2 0.00006207 secDELTA3 0.00052414 secIN0 0.00003000 secST1CNT 128ZGOPTNS

======== CHANNEL f1 ========NUC1 1HP1 7.79 usecp2 15.58 usecP28 0.10 usecPL1 0.40 dBSFO1 400.1309202 MHz

======== CHANNEL f2 ========CPDPRG2 garpNUC2 13CP3 13.90 usecp4 27.80 usecPCPD2 70.00 usecPL12 10.04 dBPL2 -4.00 dBSFO2 100.6203124 MHz





169

3) descladinosyl azithromycin in CDCl3

Current Data ParametersNAME da_ba_chEXPNO 10PROCNO 1

F2 - Acquisition ParametersDate_ 20100201Time 16.21INSTRUM spectPROBHD 5 mm BBI 1H/D-PULPROG zg30TD 65536SOLVENT CDCl3NS 16DS 2SWH 8223.685 HzFIDRES 0.125483 HzAQ 3.9846387 secRG 144DW 60.800 usecDE 6.00 usecTE 292.1 KD1 2.40000010 secTD0 1



Current Data ParametersNAME A_1EXPNO 10PROCNO 1





170

Current Data ParametersNAME samp_21EXPNO 12PROCNO 1










171







F2 - Acquisition ParametersDate_ 20120703Time 11.26INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG cosygpmfphTD 4096SOLVENT CDCl3NS 48DS 4SWH 4132.231 HzFIDRES 1.008846 HzAQ 0.4956660 secRG 2050DW 121.000 usecDE 6.00 usecTE 293.0 KD0 0.00010827 secD1 1.91398394 secD13 0.00000400 secD16 0.00020000 secIN0 0.00024200 sec

======== CHANNEL f1 ========NUC1 1HP1 10.00 usecP2 20.00 usecPL1 -0.10 dBPL1W 26.21803665 WSFO1 500.1321688 MHz

====== GRADIENT CHANNEL =====GPNAM1 SINE.100GPNAM2 SINE.100GPZ1 10.00 %GPZ2 20.00 %P16 1000.00 usec




172


F2 - Acquisition ParametersDate_ 20120703Time 20.00INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG hsqcetgpTD 2048SOLVENT CDCl3NS 48DS 16SWH 4132.231 HzFIDRES 2.017691 HzAQ 0.2478580 secRG 2050DW 121.000 usecDE 6.00 usecTE 293.2 KCNST2 145.0000000D0 0.00000300 secD1 1.40538204 secD4 0.00172414 secD11 0.03000000 secD13 0.00000400 secD16 0.00020000 secIN0 0.00001590 secZGOPTNS

======== CHANNEL f1 ========NUC1 1HP1 10.00 usecP2 20.00 usecP28 1000.00 usecPL1 -0.10 dBPL1W 26.21803665 WSFO1 500.1321688 MHz

======== CHANNEL f2 ========CPDPRG2 garpNUC2 13CP3 9.00 usecP4 18.00 usecPCPD2 65.00 usecPL2 -1.00 dBPL12 16.17 dBPL2W 104.35516357 WPL12W 2.00222993 WSFO2 125.7702390 MHz






F2 - Acquisition ParametersDate_ 20120704Time 1.43INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG roesyetgp.2TD 2048SOLVENT CDCl3NS 48DS 4SWH 4132.231 HzFIDRES 2.017691 HzAQ 0.2478580 secRG 9DW 121.000 usecDE 6.00 usecTE 293.0 KD0 0.00000300 secD1 1.95289600 secD11 0.03000000 secD16 0.00020000 secIN0 0.00024200 secL4 400P15 200000.00 usec

======== CHANNEL f1 ========NUC1 1HP1 10.00 usecP2 20.00 usecP25 250.00 usecPL1 -0.10 dBPL27 21.84 dBPL1W 26.21803665 WPL27W 0.16772591 WSFO1 500.1321688 MHz





173

4) tylosin A in CDCl3


F2 - Acquisition ParametersDate_ 20120622Time 12.03INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG zgpg30TD 65536SOLVENT CDCl3NS 15360DS 2SWH 29761.904 HzFIDRES 0.454131 HzAQ 1.1010548 secRG 512DW 16.800 usecDE 32.78 usecTE 293.1 KD1 2.00000000 secD11 0.03000000 secTD0 1




174

Current Data ParametersNAME 2012-06-22-evb-2_2EXPNO 13PROCNO 1

F2 - Acquisition ParametersDate_ 20120623Time 12.19INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG dept90_mTD 65536SOLVENT CDCl3NS 6144DS 4SWH 29761.904 HzFIDRES 0.454131 HzAQ 1.1010548 secRG 512DW 16.800 usecDE 6.00 usecTE 293.0 KCNST2 145.0000000D1 2.00000000 secD2 0.00344828 secD12 0.00002000 secTD0 1





F2 - Acquisition ParametersDate_ 20120623Time 6.56INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG dept135_mTD 65536SOLVENT CDCl3NS 6144DS 4SWH 29761.904 HzFIDRES 0.454131 HzAQ 1.1010548 secRG 2050DW 16.800 usecDE 6.00 usecTE 293.0 KCNST2 145.0000000D1 2.00000000 secD2 0.00344828 secD12 0.00002000 secTD0 1




175


F2 - Acquisition ParametersDate_ 20120624Time 12.41INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG cosygpmfphTD 4096SOLVENT CDCl3NS 32DS 4SWH 6250.000 HzFIDRES 1.525879 HzAQ 0.3277300 secRG 2050DW 80.000 usecDE 6.00 usecTE 293.0 KD0 0.00006727 secD1 2.08191991 secD13 0.00000400 secD16 0.00020000 secIN0 0.00016000 sec







F2 - Acquisition ParametersDate_ 20120623Time 12.26INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG hmbcgplpndqfTD 4096SOLVENT CDCl3NS 96DS 16SWH 6250.000 HzFIDRES 1.525879 HzAQ 0.3277300 secRG 2050DW 80.000 usecDE 6.00 usecTE 293.0 KCNST2 145.0000000CNST13 10.0000000D0 0.00000300 secD1 1.48689401 secD2 0.00344828 secD6 0.05000000 secD16 0.00020000 secIN0 0.00001655 sec







176


F2 - Acquisition ParametersDate_ 20120624Time 1.17INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG hsqcetgpTD 2048SOLVENT CDCl3NS 96DS 16SWH 6250.000 HzFIDRES 3.051758 HzAQ 0.1638900 secRG 2050DW 80.000 usecDE 6.00 usecTE 293.0 KCNST2 145.0000000D0 0.00000300 secD1 1.48934996 secD4 0.00172414 secD11 0.03000000 secD13 0.00000400 secD16 0.00020000 secIN0 0.00001655 secZGOPTNS

======== CHANNEL f1 ========NUC1 1HP1 10.00 usecP2 20.00 usecP28 1000.00 usecPL1 -0.10 dBPL1W 26.21803665 WSFO1 500.1321627 MHz

======== CHANNEL f2 ========CPDPRG2 garpNUC2 13CP3 9.00 usecP4 18.00 usecPCPD2 65.00 usecPL2 -1.00 dBPL12 16.17 dBPL2W 104.35516357 WPL12W 2.00222993 WSFO2 125.7702390 MHz






F2 - Acquisition ParametersDate_ 20120624Time 18.14INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG roesyetgp.2TD 2048SOLVENT CDCl3NS 64DS 4SWH 6250.000 HzFIDRES 3.051758 HzAQ 0.1638900 secRG 18DW 80.000 usecDE 6.00 usecTE 293.0 KD0 0.00000300 secD1 2.00655293 secD11 0.03000000 secD16 0.00020000 secIN0 0.00016000 secL4 400P15 200000.00 usec






177

5) Tylosin B in phosphate buffered D2O (apparent pH=7)

Current Data ParametersNAME BiljanaEXPNO 10PROCNO 1

F2 - Acquisition ParametersDate_ 20100528Time 11.45INSTRUM spectPROBHD 5 mm TXI 1H/D-PULPROG zg30TD 65536SOLVENT D2ONS 64DS 2SWH 10288.065 HzFIDRES 0.156983 HzAQ 3.1850996 secRG 32DW 48.600 usecDE 6.00 usecTE 294.1 KD1 2.40000010 secTD0 1

======== CHANNEL f1 ========NUC1 1HP1 8.00 usecPL1 2.50 dBPL1W 14.40788269 WSFO1 500.1323450 MHz



F2 - Acquisition ParametersDate_ 20100528Time 11.46INSTRUM spectPROBHD 5 mm TXI 1H/D-PULPROG hmbcgplpndqfTD 4096SOLVENT D2ONS 24DS 8SWH 7500.000 HzFIDRES 1.831055 HzAQ 0.2731167 secRG 2050DW 66.667 usecDE 6.00 usecTE 294.2 KCNST2 145.0000000CNST13 10.0000000D0 0.00000300 secD1 1.53604496 secD2 0.00344828 secD6 0.05000000 secD16 0.00010000 secIN0 0.00001590 sec







178


F2 - Acquisition ParametersDate_ 20100529Time 0.36INSTRUM spectPROBHD 5 mm TXI 1H/D-PULPROG hsqcetgpprsisp2.2TD 1024SOLVENT D2ONS 48DS 8SWH 7500.000 HzFIDRES 7.324219 HzAQ 0.0683167 secRG 2050DW 66.667 usecDE 6.00 usecTE 294.2 KCNST2 145.0000000CNST17 -0.5000000D0 0.00000300 secD1 1.50000000 secD4 0.00172414 secD11 0.03000000 secD12 0.00002000 secD16 0.00010000 secD24 0.00086207 secIN0 0.00001590 sec


======== CHANNEL f2 ========CPDPRG2 garpNUC2 13CP3 9.10 usecP14 500.00 usecP24 2000.00 usecPCPD2 70.00 usecPL0 120.00 dBPL2 1.50 dBPL12 19.22 dBPL0W 0.00000000 WPL2W 58.68321991 WPL12W 0.99200535 WSFO2 125.7671682 MHzSP3 10.48 dBSP7 10.48 dBSPNAM3 Crp60,0.5,20.1SPNAM7 Crp60comp.4SPOAL3 0.500SPOAL7 0.500SPOFFS3 0.00 HzSPOFFS7 0.00 Hz






F2 - Acquisition ParametersDate_ 20100529Time 11.43INSTRUM spectPROBHD 5 mm TXI 1H/D-PULPROG cosydfetgppr.2TD 2048SOLVENT D2ONS 48DS 4SWH 7500.000 HzFIDRES 3.662109 HzAQ 0.1365833 secRG 2050DW 66.667 usecDE 6.00 usecTE 294.3 KD0 0.00000300 secD1 2.40000010 secD11 0.03000000 secD13 0.00000400 secD16 0.00020000 secIN0 0.00013330 sec

======== CHANNEL f1 ========NUC1 1HP1 8.00 usecP2 16.00 usecPL1 2.50 dBPL9 52.48 dBPL1W 14.40788269 WPL9W 0.00014474 WSFO1 500.1323450 MHz





179


F2 - Acquisition ParametersDate_ 20100604Time 10.59INSTRUM spectPROBHD 5 mm TXI 1H/D-PULPROG roesyetgp.2TD 2048SOLVENT D2ONS 48DS 6SWH 7500.000 HzFIDRES 3.662109 HzAQ 0.1365833 secRG 456DW 66.667 usecDE 6.00 usecTE 294.4 KD0 0.00000300 secD1 2.00000000 secD11 0.03000000 secD16 0.00010000 secIN0 0.00013330 secL4 600P15 300000.00 usec







F2 - Acquisition ParametersDate_ 20120702Time 13.31INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG zgpg30TD 65536SOLVENT H2O+D2ONS 14336DS 2SWH 29761.904 HzFIDRES 0.454131 HzAQ 1.1010548 secRG 512DW 16.800 usecDE 32.78 usecTE 293.0 KD1 2.00000000 secD11 0.03000000 secTD0 1




180






181

6) Tylosin B in CDCl3

Current Data ParametersNAME TB145_cdcl3EXPNO 10PROCNO 1

F2 - Acquisition ParametersDate_ 20081122Time 13.00INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG zg30TD 65536SOLVENT CDCl3NS 16DS 2SWH 6172.839 HzFIDRES 0.094190 HzAQ 5.3084660 secRG 128DW 81.000 usecDE 6.00 usecTE 294.4 KD1 2.40000010 secTD0 1








182


F2 - Acquisition ParametersDate_ 20081123Time 7.37INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG hmqcgpqfTD 1024SOLVENT CDCl3NS 96DS 16SWH 3360.215 HzFIDRES 3.281460 HzAQ 0.1524212 secRG 26008DW 148.800 usecDE 6.50 usecTE 293.8 KCNST2 145.0000000d0 0.00000300 secD1 1.51576900 secd12 0.00002000 secd13 0.00000400 secD16 0.00020000 secd2 0.00344828 secDELTA1 0.00222428 secIN0 0.00004000 sec


======== CHANNEL f2 ========CPDPRG2 garpNUC2 13CP3 8.30 usecPCPD2 65.00 usecPL12 28.00 dBPL2 8.00 dBSFO2 75.4734083 MHz






F2 - Acquisition ParametersDate_ 20081122Time 23.45INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG cosygpmfphTD 2048SOLVENT CDCl3NS 96DS 4SWH 3360.215 HzFIDRES 1.640730 HzAQ 0.3047924 secRG 20642.5DW 148.800 usecDE 6.50 usecTE 294.5 Kd0 0.00013823 secD1 1.97173703 secd13 0.00000400 secD16 0.00020000 secd20 0.00120400 secIN0 0.00029760 secST1CNT 0






183

Current Data ParametersNAME TB123EXPNO 12PROCNO 1










184


F2 - Acquisition ParametersDate_ 20081108Time 16.52INSTRUM spectPROBHD 5 mm PABBO BB-PULPROG hmbcgplpndqfTD 4096SOLVENT CDCl3NS 256DS 16SWH 3205.128 HzFIDRES 0.782502 HzAQ 0.6390260 secRG 26008DW 156.000 usecDE 6.50 usecTE 296.2 KCNST13 10.0000000CNST2 145.0000000d0 0.00000300 secD1 1.38531196 secD16 0.00020000 secd2 0.00344828 secd6 0.05000000 secIN0 0.00002980 sec







Date post:	24-Feb-2023
Category:	Documents
Upload:	khangminh22
View:	0 times
Download:	0 times

Macrolide antibiotics as anti-bacterial and potential anti ...

Documents