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

Article abstract of DOI:10.1039/c4md00220b

Erythromycin A, clarithromycin and azithromycin are the three most important macrolide antibiotics and are all very widely used in the clinic. They act on the bacterial ribosome inhibiting protein synthesis. We have performed both NMR (transferred NOESY) and modelling studies in order to determine the conformations of these antibiotics when they are bound to ribosomes. All three drugs exhibit the folded-out conformation in the bound state, but the dominant conformation of azithromycin is not completely superimposable on the clarithromycin and erythromycin A 9-ketone structures. Modelling suggests that clarithromycin (based on a 14-membered ring) and its 3-O-substituted descladinosyl derivatives are conformationally rigid molecules; these are compounds which generally exhibit more activity against Gram-positive bacteria. Azithromycin (based on a 15-membered ring) and its 3-O-descladinosyl derivatives are flexible in silico and show more activity against Gram-negative bacteria in culture.

Full text of DOI:10.1039/c4md00220b

MedChemComm
View Article Online
View Journal | View Issue
CONCISE ARTICLE
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†
Cite this: Med. Chem. Commun., 2014,
5, 1347
Biljana Arsic,a Abida Awan,a Richard J. Brennan,a Juan A. Aguilar,b Ruth Ledder,a
a
b
a
*
Andrew J. McBain, Andrew C. Regan and Jill Barber
Erythromycin A, clarithromycin and azithromycin are the three most important macrolide antibiotics and
are all very widely used in the clinic. They act on the bacterial ribosome inhibiting protein synthesis. We
have performed both NMR (transferred NOESY) and modelling studies in order to determine the
conformations of these antibiotics when they are bound to ribosomes. All three drugs exhibit the
“folded-out” conformation in the bound state, but the dominant conformation of azithromycin is not
completely superimposable on the clarithromycin and erythromycin A 9-ketone structures. Modelling
suggests that clarithromycin (based on a 14-membered ring) and its 3-O-substituted descladinosyl
derivatives are conformationally rigid molecules; these are compounds which generally exhibit more
activity against Gram-positive bacteria. Azithromycin (based on a 15-membered ring) and its 3-O-
descladinosyl derivatives are flexible in silico and show more activity against Gram-negative bacteria in
culture.
Received 20th May 2014
Accepted 11th June 2014
DOI: 10.1039/c4md00220b
www.rsc.org/medchemcomm
The three-dimensional structures of erythromycin A and its
derivatives have been analyzed in three distinct situations: in
the free drug (both in solution9 and in the crystalline state10), in
the bound state (by X-ray crystallography of drug-ribosome
complexes),11–14 and in an intermediate weakly-bound state,
detected by NMR spectroscopy, of drug-ribosome mixtures.15–17
Introduction
The rst macrolide antibiotic, erythromycin A, was isolated in
1952 from a culture of Saccharopolyspora erythraea1 and struc-
turally described in 1957.2,3 In aqueous solution, the 9-ketone
(1a) equilibrates with the 12,9-hemiacetal (1b), which appears to
have no antibacterial activity.4,5 The second generation macro-
lides, clarithromycin and azithromycin are semi-synthetic
derivatives of erythromycin A, and show increased acid stability
and broader spectra of antibacterial activity. In general, clari-
thromycin is more active against Gram-positive bacteria but
azithromycin has Gram-negative activity, which clarithromycin
does not share.6 In combination with existing antimalarial
drugs (artesunate and quinine), azithromycin also shows mild
antimalarial activity.7
The macrolide antibiotics inhibit bacterial protein synthesis
by binding reversibly to the 50S subunit of the bacterial ribo-
some. They block the exit tunnel of the ribosome near the
peptidyl transferase centre, and thus inhibit the elongation of
the peptide chain.8
aManchester Pharmacy School, University of Manchester, Manchester, M13 9PT, UK.
E-mail: jill.barber@manchester.ac.uk
bSchool of Chemistry, University of Manchester, Manchester, M13 9PL, UK
† Electronic supplementary information (ESI) available. See DOI:
10.1039/c4md00220b
This journal is © The Royal Society of Chemistry 2014
Med. Chem. Commun., 2014, 5, 1347–1354 | 1347

View Article Online
MedChemComm
Concise Article
Table 1 Key internuclear distances in crystal structures of folded-out
(erythromycin A) and folded-in (dirithromycin) macrolides
folded-in conformations in both aqueous and chloroform
solutions, and this conclusion was supported by 2D ROESY
NMR spectroscopy.15 Other molecular modelling approaches
have generally depended upon NMR results and were not
designed to be used independently.16,17
˚
Internuclear distances (A)
Contact
Erythromycin A
Dirithromycin
The main purpose of the work described here was to develop
molecular modelling protocols for conformational character-
ization that would apply to 15-membered as well as 14-
membered macrolides. We have also described systematically
the intermediate weakly bound state for erythromycin, clari-
thromycin and azithromycin, and correlated these ndings with
in vitro anti-bacterial activity.
H4–H11
H15–H16
H3–H8
H3–H11
H4–H18
H5–H18
H16–H17
2.51
3.28
5.34
3.48
4.02
2.66
3.84
3.33
3.81
2.30
2.46
2.71
3.33
2.51
Results
Everett et al. (1990) described two conformational states avail-
able to the macrolide antibiotics: the folded-in conformation in
which H3 approaches H11 and H4 approaches Me-18, and the
folded-out conformation, in which H4 approaches H11 and H5
approaches Me-18 (Table 1).18 Erythromycin 9-ketone, clari-
thromycin and azithromycin have all been shown to adopt fol-
ded-out conformations, both in the free state and when bound
to ribosomes (see Fig. 1).
The conformations of erythromycin A, clarithromycin and
azithromycin bound weakly to bacterial ribosomes
Information about the conformation of a bound ligand in fast
exchange with free ligand can be obtained using the transferred
NOESY (TRNOESY) NMR experiment. This experiment relies on
the fact that NOE build-up is fast in large complexes and slow
(compared with the dissociation rate of the complex) in small
ligands. NOE information is therefore transferred from bound
to an excess of free drug, so that experiments show the chemical
shis of the free drug, but Overhauser effects characteristic of
the conformation(s) of the bound drug. For macrolide antibi-
otics associated with ribosomes, ribosomal core particles
replace whole ribosomes in control experiments (controlling for
non-specic binding). Because of their intermediate size, 14-
and 15-membered macrolides have solution NOEs close to zero
(at 500–600 MHz and short mixing time).
The conformations adopted by these drugs in the interme-
diate weakly bound state have been the subject of much inves-
tigation, but also some confusion. The transferred NOESY
experiment is prone to artefacts (especially from spin diffusion).
In addition, Everett et al.18 showed that close approach of H11 to
either H4 or H3 was diagnostic of a folded-out or folded-in
conformation respectively for 14-membered macrolides,
providing a convenient short-cut to the conformational analysis
of 14-membered macrolides by NMR spectroscopy. However,
this relation does not necessarily hold for 15-membered mac-
rolides, which retain sufficient exibility to be able to assume a
folded-out conformation even when H3 approaches H11.
Several other contacts, especially 4–18 and 5–18, are required to
distinguish between folded-out and folded-in conformers.
Neither NMR nor X-ray crystallographic studies have the
predictive potential of molecular modelling, and it would be
useful to be able to investigate the conformation of a macrolide
without having to synthesize it. An attempt to develop a
systematic protocol for describing the conformations of mac-
rolides was described in 2006, using oleandomycin and some of
its derivatives as examples. Unconstrained molecular dynamics
suggested that the drug forms a mixture of folded-out and
Clarithromycin
1
The H NMR spectrum of clarithromycin in aqueous solution
has been described,19 and in the presence of bacterial ribo-
somes, clarithromycin gives the TRNOESY spectrum shown as
Fig. S1A in the ESI.† The spectrum contains 75 pairs of signals,
consistent with a folded-out conformation of clarithromycin
(Table S1 in the ESI†). Thirty-three of these signals are not seen
in the ROESY spectrum of the same drug but are completely
consistent with the folded-out structure. Only the weak long-
range signal (7080–800) gives additional information, constrain-
ing the relative positions of the two sugars. The crystal structure
of clarithromycin is closely similar, although the H4–H11(3.04
0
0
00
˚
˚
A) and the H7 H8 –H8 (7.71 A) distances obtained by molecular
modelling (molecular mechanics) are rather long.
Erythromycin A
Fig. S1B in the ESI† shows the TRNOESY spectrum of erythro-
mycin A. There are 83 pairs of signals belonging to erythromycin
A 9-ketone (Table S1, ESI†). Up to four small signals (depending
on processing) due to the 9,12-hemiacetal, and due to chemical
exchange between the two isomers, are also seen in this spec-
trum, but it is clear that the 9-ketone binds to ribosomes and
the 9,12-hemiacetal does not. Fig. 2 shows the global minimum
Fig. 1 Crystal structures of (A) erythromycin A10 (folded-out confor-
mation) and (B) dirithromycin18 (folded-in conformation).
1348 | Med. Chem. Commun., 2014, 5, 1347–1354
This journal is © The Royal Society of Chemistry 2014

View Article Online
Concise Article
MedChemComm
conformational search in Macromodel) and azithromycin
(global minimum from the constrained search in Macromodel)
are shown in Fig. S2 in the ESI.† Another small difference
between the two drugs is that there is clear evidence of rotation
about the C14–C15 bond of azithromycin, such that H315 can
approach both H316 and H321. This can also be seen in the
crystal structure of azithromycin bound to Deinococcus radio-
durans ribosomes; in this structure two molecules of azi-
thromycin bind, one with C15 approaching C21 and the other
with C15 (more conventionally) approaching C16.
Comparison of the binding modes of clarithromycin,
erythromycin and azithromycin
The dominant bound conformations of clarithromycin, eryth-
romycin A and azithromycin are very similar. However, clari-
thromycin shows almost no exibility in the bound state
whereas erythromycin shows some limited exibility and azi-
thromycin rather more. For all three drugs, the methyl group
H316 gives a broad 1H NMR signal. When clarithromycin is
Fig.
2 Erythromycin A 12,9-hemiacetal: global minimum from
unconstrained Monte Carlo search.
of an unconstrained Monte Carlo search on erythromycin A
9,12-hemiacetal, showing close approach of H2 to H317.
Most of the ketone signals are equivalent to those found in
the clarithromycin TRNOESY spectrum suggesting that the
dominant conformation of erythromycin is the same folded-out
conformation as adopted by clarithromycin. The differences
between the TRNOESY spectra of clarithromycin and erythro-
mycin are concentrated in two areas of the molecule. Clari-
thromycin shows more intra-sugar cross-peaks for the
desosamine sugar, but the extra peaks are all very small and
consistent with spin diffusion. The erythromycin spectrum
shows an H8–H11 cross-peak, and H7proS–H318 and H7proS
–
H319 connectivities are also present. This is consistent with
erythromycin retaining some exibility in the top portion of the
molecule and thus the ability to form an 8-endo folded-out
conformer, even when bound to ribosomes.
TRNOESY data were previously published by Bertho et al.17
Differences between our results and theirs are indicated in
Table 1 of the ESI.† The additional peaks that they report can be
attributed to spin diffusion, because of the use of non-deuter-
ated ribosomes and a relatively long mixing time.
Azithromycin
The TRNOESY spectrum of azithromycin in the presence of
bacterial ribosomes (analysed with the aid of published
assignments20) is presented in Fig. S1C of the ESI,† together
with tabulated data (Table 2 of the ESI†). The vast majority of
cross-peaks are identical with those seen for clarithromycin,
suggesting that azithromycin, from a range of available
conformations, adopts the clarithromycin-like folded-out
conformation when bound to ribosomes.
Azithromycin enjoys some exibility in the H2–H5 region so
that H3 can approach H11 without signicant effects on the rest
of the molecule. The large H2–H4 cross-peak is a result of the
same rotation. This signal is small in both the erythromycin and
clarithromycin spectra. Superimposed folded-out structures of
Fig. 3 Comparison of the TRNOESY NMR spectra of (A) erythromycin
A, (B) clarithromycin and (C): azithromycin, showing the size of the 16,2
clarithromycin (global minimum from the unconstrained crosspeak.
This journal is © The Royal Society of Chemistry 2014
Med. Chem. Commun., 2014, 5, 1347–1354 | 1349

View Article Online
MedChemComm
Concise Article
bound to ribosomes, the TRNOESY H16–H2 cross-peak almost
disappears, whereas in the TRNOESY spectra of the more ex-
ible azithromycin and erythromycin A, this signal is clearly
visible. Fig. 3 shows this region of the TRNOESY spectra of the
three drugs.
Erythromycin
Unconstrained molecular mechanics calculations suggest that
the 6-O-Me group of clarithromycin has a profound effect on its
conformational behaviour. Erythromycin A, which carries a
hydroxyl group at C-6, shows four different conformers within 1
kcal molꢀ1 of the global minimum: Fig. 5 shows a folded-out
conformer, an 8-endo-folded-out conformer (in which H8
approaches H11)9 in two versions differing in their H4–H11
distances, and a 7-endo- folded-in conformer in which H11
approaches both H3 and H7 (but not H8). On the basis of the
populations of structures within 1 kcal molꢀ1 of the global
minimum, we would expect the folded-out conformer to
predominate (about 75% of the minima found belong to this
family). On the same basis, the 7-endo-folded-in conformer
would be present at less than 5%.
Simulated annealing gave very similar results, with a folded-
out structure again appearing at lowest energy and 8-endo-fol-
ded-out and 7-endo-folded-in structures appearing within 1 kcal
molꢀ1. The orientation of the cladinose sugar is slightly
different in the global minimum obtained by simulated
annealing from that obtained by molecular mechanics (see
Table S4 in the ESI† for details), so that the superposition of the
two sets of modelling results (Fig. 5E) is less good than for
clarithromycin (Fig. 4B).
Unconstrained molecular mechanics and molecular dynamics
searches on clarithromycin, erythromycin A, azithromycin
We now compared results of unconstrained modelling using
two conformational analysis approaches:
a Monte Carlo
Multiple Minima (MCMM) search and a molecular dynamics-
based simulated annealing (MDSA) method. Unconstrained
MCMM calculations were carried out with Macromodel version
8 using the AMBER* force eld21 and water as a solvent (GBSA
solvent model).22 Molecular dynamics calculations were per-
formed in Sybyl 7.3 (Tripos, St. Louis) using simulated anneal-
ing with energy minimization using Tripos force eld and water
solvation model.23 The starting structures for both modelling
protocols were crystal structures of the macrolides alone (clar-
ithromycin,24 erythromycin A10 and azithromycin25) available in
the Cambridge Crystallographic Database.
Molecular mechanics calculations are generally preferred for
small molecules: the exploration of conformational space is
oen comprehensive and very accurate predictions of lowest
energy structures can be obtained. Molecular dynamics calcu-
lations are generally preferred for large molecules. The
exhaustive exploration of conformational space is simply not
practical for large, exible molecules. Macrolide antibiotics do
not t comfortably into the description “small molecules” but
neither are they macromolecules. For this reason we used both
molecular dynamics and molecular mechanics approaches.
Clarithromycin
In the case of clarithromycin, molecular mechanics calculations
appeared to be satisfactory and a calculation using the AMBER
force eld gave a folded-out global minimum which was found
19 times and is shown in Fig. 4A. All the NMR constraints are
satised in this single conformer. The global minimum found
by molecular dynamics (simulated annealing) experiments was
almost identical with this structure, as shown in Fig. 4B. Key
internuclear distances for these structures are shown in the ESI
(Table S3†).
Fig. 5 Families of conformers of erythromycin A within 1 kcal molꢀ1 of
Fig. 4 (A) Global minimum of unconstrained Monte Carlo search on the global minimum found by Monte Carlo searching: (A) folded-out;
clarithromycin in water; (B) superposition of this global minimum (pink) (B) and (C) variants of the 8-endo-folded-out conformation; (D) 7-
and the lowest energy conformation (shown in green) obtained by endo-folded-in conformer. (E) Superposition of this global minimum
simulated annealing. Both structures satisfy all the TRNOESY NMR (pink) and the lowest energy conformation (shown in black) obtained
constraints.
by simulated annealing.
1350 | Med. Chem. Commun., 2014, 5, 1347–1354
This journal is © The Royal Society of Chemistry 2014

View Article Online
Concise Article
MedChemComm
1
The TRNOESY NMR spectrum for erythromycin A shows
H200s–H400 and H800–H200r peaks which may be assumed to be
due to spin diffusion. Otherwise, all the cross-peaks are satis-
ed by the 8-endo-folded-out conformer and, with the exception
of H8–H11, by the global minimum folded-out conformer.
The conformational exibility of azithromycin: H NMR
studies and molecular dynamics simulation
The failure of both molecular dynamics and molecular
mechanics calculations to identify the global minimum of azi-
thromycin found experimentally was disappointing. We
considered the (highly remote) possibility that this relatively
large molecule could adopt a metastable conformation in
aqueous solution, above the global minimum. In order
to explore this possibility, we heated an aqueous solution of
azithromycin in an NMR tube over the temperature range 30 ꢂC-
90 ꢂC and allowed the solution to cool, measuring the 1H NMR
spectrum at intervals. Fig. S3 of the ESI.† shows the (uninter-
esting) 1H NMR spectra obtained in this experiment. There was
no change in the spectra over the temperature range other than
the expected slight temperature-dependent shis. The nal
spectrum was identical with the starting spectrum.
Azithromycin
An unconstrained molecular mechanics calculation gave rise to
a global minimum structure (Fig. 6A) with an energy of 30.6 kcal
molꢀ1. This structure does not resemble any known confor-
mation for macrolide antibiotics, neither does it satisfy the
NMR constraints (see Table S5 in the ESI†). Interestingly, it
closely resembles the 7-endo-folded-in erythromycin, which was
found in silico (see Fig. 5). As in the conventional folded-in
conformation, H11 approaches H3, but whereas a conventional
folded-in conformation shows a close approach of H8 to H3,
this conformation has H7 approaching H3 and H4.
The same experiment was carried out in silico (molecular
dynamics calculations) starting from the folded-out crystal
structure of azithromycin. It showed the same result as 1H NMR
experiment. In aqueous solution the conventional folded-out
conformer of azithromycin is stable.
There were no familiar conformers within 2.4 kcal molꢀ1 of
the global minimum; the H3–H11 and H4–H11 distances were
all too large. In order to achieve a folded-out structure that
would satisfy the NMR data, it was necessary to constrain the
˚
Monte Carlo search, setting H3–H11 to 3.5 ꢁ 0.3 A and H4–H11
˚
to 2.5 ꢁ 0.3 A. This yielded a conformer of energy 35.3 kcal
The properties of descladinosylclarithromycin and
descladinosylazithromycin
molꢀ1 that satised all the NMR constraints (Fig. 6B).
Molecular dynamics calculations were supercially more
immediately successful, and the global minimum in the
unconstrained calculation was a folded-out structure (Fig. 6C)
in which the NMR constraints were all satised, except that the
As shown above, the second generation macrolides, although
oen assumed to be equivalent to one another, differ hugely in
their conformational exibility. The effect of the removal of the
cladinose sugar was now explored. 5-Desosaminylerythronolide
A (descladinosylerythromycin A) has little or no antibacterial
activity,26 and unconstrained molecular modelling shows a
folded-out structure in which the desosamine sugar fails to
superimpose upon that of erythromycin A (data not shown).
Descladinosylclarithromycin (4) and descladinosylazithromycin
(5) were constructed from the corresponding drugs24,25 in silico,
and were subjected to unconstrained Monte Carlo searches.
Fig. 7 shows how the global minimum structure of descla-
dinosylclarithromycin is almost superimposable on that of
clarithromycin. As expected, the more exible azithromycin
derivative did not superimpose on a folded-out macrolide.
The synthesis of compounds 4 and 5 by acid-catalysed
degradation of the parent drugs is described in the ESI,†
together with their 1H and 13C NMR data. Measurements of
minimum inhibitory concentrations of 4 and 5 against an azi-
thromycin-susceptible Escherichia coli K12 strain and against
inter-sugar distance H10–H500 of 2.77 A is long-below 2.5 A is
expected. The structure, however, is profoundly different from
the crystal structure; the orientation of the sugar rings is quite
unexpected.
˚
˚
A more conventional folded-out structure with H10–H500 of
2.35 A, was found at 3.8 kcal molꢀ1 above the global minimum
˚
(Fig. 6D).
Fig. 6 Molecular modelling of azithromycin (A) Global minimum of
unconstrained Monte Carlo search in water (B) global minimum of
Monte Carlo search with H3–H11 and H4–H11 constrained. (C)
Unconstrained molecular dynamics lowest energy structure. (D) Fig. 7 Global minima of clarithromycin (yellow) and descladinosyl-
Unconstrained molecular dynamics structure 3.8 kcal molꢀ1 above the clarithromycin (4, blue) in water, obtained from unconstrained
lowest energy structure.
molecular mechanics, superimposed.
This journal is © The Royal Society of Chemistry 2014
Med. Chem. Commun., 2014, 5, 1347–1354 | 1351

View Article Online
MedChemComm
Concise Article
Table 2 Key internuclear distances in the global minima of compounds 6, 9 and 12. The global minima of 7 and 8 are essentially identical with 6
and those of 10 and 11 are essentially identical with 9. (S.a. = Staphylococcus aureus, H.i. = Haemophilus influenzae)
MIC (mg mlꢀ1
)
Key distances (A) in the conformation representing the global minimum
˚
Cmpd
S.a.
H.i.
3–11
4–11
3–8
8–11
4–18
5–18
15–16
16–17
6
9
12
0.78
3.13
NA
12.5
1.56
NA
4.0
2.4
4.0
2.5
4.5
2.3
6.1
3.9
5.9
4.3
4.8
4.3
4.0
2.7
4.0
2.7
3.5
2.6
2.8
5.3
5.1
4.2
2.3
4.4
macrolide-susceptible Bacillus cereus and Staphylococcus aureus is a proper folded-out structure. Derivatives of descladinosyla-
strains showed essentially no activity (see ESI Table S7†). This is zithromycin showed folded-out (active) conformers about 4.8
the expected result for descladinoseazithromycin, but illus- kcal molꢀ1 above the global minimum. Only when the azi-
trates that, although descladinoseclarithromycin (4) adopts the thromycin skeleton was substituted with a large steroid side-
same conformation as its antibacterial parent, this is insuffi- chain (compound 12) did it give rise to a folded-out structure in
cient for activity.
the global minimum of a molecular mechanics calculation. The
latter compound was, however, developed not as an antibacte-
rial but as an anti-inammatory agent!
Unconstrained molecular mechanics on 3-O-substituted
descladinosyl clarithromycin and 3-O-substituted
descladinosyl azithromycin derivatives
Discussion
Derivatives of 4 and 5, substituted at the 3-oxygen have been
reported and some have antibacterial activity.27,28 Compounds Molecular modelling calculations, both molecular mechanics
6–9, together with the unreported 10–11 were now subjected to and molecular dynamics are successful in nding global
unconstrained Monte Carlo searches. The compounds, their minima for the 14-membered macrolides. In the case of clari-
activities and structural characteristics obtained using uncon- thromycin, the folded-out structure (Fig. 4) is the only structure
strained molecular modelling are represented in Table 2. to be found by modelling and the only structure indicated by
Compounds 6 (based on clarithromycin) and 9 (based on azi- NMR. The TRNOESY data for clarithromycin bound to bacterial
thromycin) have signicant antibacterial activity, but (as in the ribosomes indicate that this drug has very limited conforma-
case of the parent drugs) the global minima found for these tional exibility, hence the almost complete disappearance of
compounds are quite different from one another. Compound 6 the H2–H316 cross-peak. Erythromycin A is more exible than
1352 | Med. Chem. Commun., 2014, 5, 1347–1354
This journal is © The Royal Society of Chemistry 2014

View Article Online
Concise Article
MedChemComm
clarithromycin, and is able to adopt an 8-endo-folded-out as well azithromycin and clarithromycin as broadly equivalent; indeed
as a conventional 8-exo-folded-out conformation. Molecular hospital antibiotic policies oen permit one but not the other.
modelling predicts a minor 7-endo-folded-in conformer, but Both are acid-stable, hydrophobic derivatives of erythromycin A,
this has not been distinguished experimentally.
and both have excellent pharmacokinetic properties. However,
Azithromycin, the 15-membered macrolide, is, however, it has long been known that these remarkable compounds have
quite different. Molecular modelling indicates that a wide range quite different spectra of activity. Azithromycin has the rare
of conformations is available to this drug. Although the property of excellent anti-Gram-negative activity (and even anti-
preferred conformation (as indicated by NMR) is essentially the malarial activity), whereas clarithromycin generally has excel-
same folded-out structure as is seen for erythromycin A and lent activity against Gram positive organisms.
clarithromycin, this drug is conformationally exible, because
of the increased size of the macrolide ring.
Molecular modelling, even with commercial packages, has
advanced our understanding of the properties of these molecules.
In three distinct situations (free solution,9 weakly-bound to Hence we are able to postulate that the chemical basis of the
bacterial ribosomes in the present work, and tightly-bound to clinical differences between these drugs may lie in their very
ribosomes detected by X-ray crystallography14) clarithromycin, different exibilities. Azithromycin is conformationally labile,
erythromycin A 9-ketone and azithromycin adopt essentially the confounding the barriers that cells throw up to bar the entry of
same conformation, as shown in Fig. 1. (The distorted confor- drugs. What clarithromycin lacks in subtlety it makes up for in
mation of erythromycin reported in early X-ray studies of drug- focus: sensitive Gram positive infections are very sensitive indeed.
ribosome complexes12 has been superseded29-it probably derived
from the relatively low resolution of these studies and from their
focus on the ribosome rather than the drug.). Molecular
dynamics or molecular mechanics calculations can be used to
Experimental
Molecular modelling
predict whether a macrolide has this conformation available, but
in the case of azithromycin, we have been unable to nd a simple
modelling protocol using commercially available soware that
successfully nds the lowest energy conformation seen experi-
mentally. Choosing an appropriate solvent and solvent model is
one of the key stages in Monte Carlo calculations. Implicit
solvation represents a compromise solution in which explicitly
represented water molecules are replaced by a continuum model
that reects the average behaviour of water molecules. Our choice
of the GBSA solvent model (Generalized Born model augmented
with the hydrophobic solvent accessible surface area (SA) term)
and AMBER* force eld gives good results for conformationally
rigid molecules like clarithromycin, its 3-O-substituted descladi-
nosyl derivatives and erythromycin A. However, approximations
in the water model and force eld have the potential to impair the
prediction of biologically active conformations of conforma-
tionally exible molecules like azithromycin and its 3-O-
substituted descladinosyl derivatives. Where there is more exi-
bility, there are more opportunities for intra-solute interactions.
These intrasolute interactions could be over emphasised using
implicit solvent rather than explicit water molecules – the latter
could more adequately screen and provide bridging hydrogen
bond networks. Implicit solvent has shown inadequate shielding
for salt bridges.30,31
Erythromycin A, clarithromycin and its 3-O-substituted descla-
dinosyl derivatives, azithromycin and its 3-O-substituted
descladinosyl derivatives were constructed using Macromodel 8
soware32 from the appropriate crystal structures. Structures
were minimized using the Truncated Newton Conjugate
Gradient (TNCG) method in order to obtain local minima. The
Monte Carlo Multiple Minimum (MCMM) conformational
search was used to nd the global minima.33 The GBSA model
was used for water solvation. The search was set to 10 000
structures to be minimized and all structures within 6 kcal
molꢀ1 energy range were stored.
Molecular dynamics calculations were performed using
Sybyl 7.3.34 Minimization was performed using the Tripos force
eld in water as a solvent (3¼80.4). The Gasteiger–Huckel
charges were used throughout the calculation. Global minima
of clarithromycin, erythromycin A and azithromycin were
obtained from their crystal structures minimized by Tripos
force eld followed by simulated annealing [(clarithromycin;
heated to 600 K for 1500 fs and then cooled to 0 K for 10 000 fs),
(erythromycin A; heated to 1000 K for 750 fs and then cooled to
0 K for 10 000 fs) and (azithromycin; heated to 600 K for 2000 fs
and then cooled to 0 K for 10 000 fs)]. The structures obtained
were minimized again using the Tripos force eld. The number
of iterations for each minimization was 10 000 and the mini-
mization simulation was terminated gradiently when the
difference in energies between two conformations was no more
At the present time, we can model clarithromycin derivatives
and predict that only a molecule that adopts a folded-out
conformation (Fig. 1) in the global minimum is likely to be
active, whereas for azithromycin derivatives we are required to
search for the active conformer among the range of structures
indicated by modelling.
than 0.005 kcal molꢀ1
.
The abundance of different conformers (in case of erythro-
mycin A) was calculated according the simple formula:
Nðconformation typeÞ
Macrolides are among the most efficacious and safe drugs in
the clinic. A frustration is that they are all based on natural
u% ðconformation typeÞ ¼
ꢃ100;
NðtotalÞ
products containing many chiral centres. This means that where N (conformation type) is the number of structures of the
systematic investigations of structure–activity relationships by same type (e.g. folded-out) within 1 kcal molꢀ1 of the global
directed modication of the skeleton are simply not feasible. It minimum and N (total) is the number of structures of all types
is conventional to think of the semi-synthetic antibiotics within 1 kcal molꢀ1 of the global minimum.
This journal is © The Royal Society of Chemistry 2014
Med. Chem. Commun., 2014, 5, 1347–1354 | 1353

View Article Online
MedChemComm
Concise Article
9 A. Awan, R. J. Brennan, A. C. Regan and J. Barber, J. Chem.
Soc., Perkin Trans. 2, 2000, 1645.
10 G. A. Stephenson, J. G. Stowell, P. H. Toma, R. R. Pfeiffer and
S. R. Byrn, J. Pharm. Sci., 1997, 86, 1239.
11 F. Schlunzen, R. Zarivach, J. Harms, A. Bashan, A. Tocilj,
R. Albrecht, A. Yonath and F. Franceschi, Nature, 2001,
413, 814.
12 F. Schlunzen, J. M. Harms, F. Franceschi, H. A. Hansen,
H. Bartels, R. Zarivach and A. Yonath, Structure, 2003, 11,
329.
NMR analysis
One and two dimensional NMR spectra were obtained by
standard procedures, described in the ESI.†
Minimum inhibitory concentrations of clarithromycin, azi-
thromycin and their derivatives against Escherichia coli, Bacillus
cereus, Staphylococcus aureus, Pseudomonas aeruginosa, Staphy-
lococcus epidermidis, Serratia marcescens and Corynebacterium
xerosis
These were determined by standard methods35 and are
described in the ESI.†
13 D. Tu, G. Blaha, P. B. Moore and T. Steitz, Cell, 2005, 121,
257.
14 J. L. Hansen, J. A. Ippolito, N. Ban, P. Nissen, P. B. Moore and
T. A. Steitz, Mol. Cell, 2002, 10, 117.
Conclusions
Molecular dynamics and molecular mechanics calculations
using commercially available soware can be used to predict
whether macrolide derivatives can adopt the active folded-out
conformer required for ribosome binding. In addition, an
indication of the likely exibility of a given molecule can be
obtained. For antibacterial 14-membered macrolides, this fol-
ded-out conformer is likely to be found at lowest energy in
unconstrained searches. 15-membered macrolides adopt the
same conformation in the active state, as shown by NMR and X-
ray crystallographic measurements, but unconstrained molec-
ular modelling cannot be relied on to nd this conformer at
lowest energy. Both NMR and modelling studies indicate that
azithromycin is an intrinsically exible molecule and can adopt
numerous conformers. The exibility correlates with the ability
of the drug to target Gram negative and even malarial cells.
15 P. Novak, I. Tatic, P. Tepes, S. Kostrun and J. Barber, J. Phys.
Chem. A, 2006, 110, 572.
16 J. I. Gyi, R. J. Brennan, D. A. Pye and J. Barber, J. Chem. Soc.,
Chem. Commun., 1991, 1471.
17 G. Bertho, J. Gharbi-Benarous, M. Delaforge and J.-P. Girault,
Bioorg. Med. Chem., 1998, 6, 209.
18 J. R. Everett and J. W. Tyler, J. Chem. Soc., Perkin Trans. 2,
1987, 1659.
19 A. Awan, J. Barber, R. J. Brennan and J. A. Parkinson, Magn.
Reson. Chem., 1992, 30, 1241.
20 J. Barber, Magn. Reson. Chem., 1991, 29, 740.
21 J. Wang, R. M. Wolf, J. W. Caldwell, P. A. Kollman and
D. A. Case, J. Comput. Chem., 2004, 25, 1157.
22 W. C. Still, A. Tempczyk, R. C. Hawley and T. Hendrickson,
J. Am. Chem. Soc., 1990, 112, 6127.
23 M. Clark, R. D. Cramer III and N. Van Opdenbosch,
J. Comput. Chem., 1989, 10(8), 982.
Acknowledgements
24 T. Adachi, S. Morimoto, Y. Watanabe, N. Kamiya and
H. Iwasaki, J. Antibiot., 1989, 42, 1012.
25 S. Djokic, G. Kobrehel, N. Lopotar, B. Kamena, A. Nagl and
D. Mrvos, J. Chem. Res., 1988, 152, 1239.
We are thankful to Mr Roger Speak for his valuable help during
recording of NMR spectra. We also thank Professor Gareth
Morris and Dr Richard Bryce for fruitful discussion.
26 R. A. LeMahieu, M. Carson, R. W. Kierstead, L. M. Fern and
E. Grunberg, J. Med. Chem., 1974, 17, 953.
Notes and references
27 T. Tanikawa, T. Asaka, M. Kashimura, K. Suzuki,
H. Sugiyama, M. Sato, K. Kameo, S. Morimoto and
A. Nishida, J. Med. Chem., 2003, 46, 2706.
1 L. Katz and G. Ashley, Chem. Rev., 2005, 105, 499.
2 J. C. Cra and D. Stamler, Infect. Dis. Ther., 1995, 18, 359,
New Macrolides, Azalides, and Streptogramins in Clinical
Practice.
28 S. Mutak, J. Antibiot., 2007, 60, 85.
29 J. A. Dunkle, L. Xiong, A. S. Mankin and J. H. Cate, Proc. Natl.
Acad. Sci. U. S. A., 2010, 107, 17152.
3 M. Zief, R. Woodside and E. Huber, Antibiot. Chemother.,
1957, 7, 604.
30 C. Simmerling, B. Strockbine and A. E. Roitberg, J. Am. Chem.
Soc., 2002, 124, 11258.
4 D. A. Pye, J. I. Gyi and J. Barber, J. Chem. Soc., Chem.
Commun., 1990, 1143.
31 R. Geney, M. Layten, R. Gomperts, V. Hornak and
C. Simmerling, J. Chem. Theory Comput., 2005, 2, 115.30F.
32 F. Mohamadi, N. G. J. Richards, W. C. Guida, R. Liskamp,
M. Lipton, C. Caueld, G. Chang, T. Hendrickson and
W. C. Still, J. Comput. Chem., 1990, 11, 440.
33 M. Saunders, K. N. Houk, Z. D. Wu, W. C. Still, M. Lipton,
G. Chang and W. C. Guida, J. Am. Chem. Soc., 1990, 112, 1419.
34 SYBYL 7.3, Tripos Associates, 1699 South Hanley Road, Suite
303, St. Louis, MO, 2006, 63144–2917.
5 J. Barber, J. I. Gyi and D. A. Pye, J. Chem. Soc., Chem.
Commun., 1991, 1249.
6 M. D. Kitris, F. W. Goldstein, M. Miegi and J. F. Acar,
J. Antimicrob. Chemother., 1990, 25(suppl. A), 15.
7 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 and C. Knirsch, Clin.
Infect. Dis., 2006, 43, 1264.
35 A. J. McBain, R. G. Ledder, P. Sreenivasan and P. Gilbert,
J. Antimicrob. Chemother., 2004, 53, 772.
8 J. Poehlsgaard and S. Douthwaite, Curr. Opin. Invest. Drugs,
2003, 4, 140.
1354 | Med. Chem. Commun., 2014, 5, 1347–1354
This journal is © The Royal Society of Chemistry 2014