Exploring eukaryotic: Versus prokaryotic ribosomal RNA recognition with aminoglycoside derivatives

Article abstract of DOI:10.1039/c8md00001h

New derivatives of aminoglycosides containing 6′-carboxylic acid or 6′-amide on their ring I were designed, synthesized and their ability to readthrough nonsense mutations was examined in vitro, along with the protein translation inhibition in prokaryotic and eukaryotic systems. The observed structure-activity relationships, along with the comparative molecular dynamics simulations within the eukaryotic rRNA decoding site, showed high sensitivity of 6′-position to substitution, indicating that the rational design of potent stop-codon read-through inducers requires consideration of not only the structure and energetics of the drug-RNA interaction but also the dynamics associated with that interaction.

Full text of DOI:10.1039/c8md00001h

MedChemComm
View Article Online
View Journal
RESEARCH ARTICLE
Exploring eukaryotic versus prokaryotic ribosomal
RNA recognition with aminoglycoside derivatives†
Cite this: DOI: 10.1039/
c8md00001h
a
Narayana Murthy Sabbavarapu,
*
Tomasz Pieńko,bc Bat-Hen Zalman,a
a
b and Timor Baasov
*
Joanna Trylska
New derivatives of aminoglycosides containing 6′-carboxylic acid or 6′-amide on their ring I were
designed, synthesized and their ability to readthrough nonsense mutations was examined in vitro, along
with the protein translation inhibition in prokaryotic and eukaryotic systems. The observed structure–activity
relationships, along with the comparative molecular dynamics simulations within the eukaryotic rRNA
decoding site, showed high sensitivity of 6′-position to substitution, indicating that the rational design of
potent stop-codon read-through inducers requires consideration of not only the structure and energetics
of the drug–RNA interaction but also the dynamics associated with that interaction.
Received 1st January 2018,
Accepted 31st January 2018
DOI: 10.1039/c8md00001h
rsc.li/medchemcomm
lar data on the interaction of AGs within their eukaryotic
target site, all those lead structures were developed based on
Introduction
Aminoglycosides (AGs) are well known antibiotics that specifi-
cally bind to the decoding-site RNA within the small ribo-
somal subunit, and thereby interfere with translation fidelity,
leading to bacterial cell death.1 In the last several years, it was
shown that some AG antibiotics can also impact the eukary-
otic ribosome and promote read-through of premature termi-
nation codons (PTCs) in mRNA. This unique phenomenon of
AGs, also termed “PTC suppression” or “translational
readthrough”, has been demonstrated in several in vitro and
in vivo experiments including clinical trials.2 AGs are there-
fore considered as potential therapeutics for PTC underlined
genetic disorders. However, the lack of potency of standard
AGs and their recognized nephrotoxicity and ototoxicity at
high doses encouraged further research towards the develop-
ment of new derivatives.3
the reported structural data of conventional AGs within the
prokaryotic target site, and their subsequent biochemical and
toxicity data. Recent structural and mechanistic insights on
the interaction of AGs with eukaryotic ribosome, provided
reasonable basis for structure-based rational design of novel
read-through inducers.6
The crystal structure of the AG geneticin (G418) bound
to 80S yeast ribosome,7 along with the structures of G418
(ref. 8) and paromomycin9 bound to their putative binding
site of Leishmania ribosomes, are among recently reported
structures deciphering the interaction of AGs with the eukary-
otic ribosomal target. Based on these structural data, we very
recently reported on the design, synthesis and evaluation of
the new lead structure 2 that differs from its parent structure
1 by the presence of an additional hydroxyl on the side-chain
of the glucosamine ring (ring I) (compounds 1 and 2 are also
named NB124 and NB157, respectively, Fig. 1).10 Preliminary
modeling study suggested that the extension of the side-
chain methyl group in compound 1 so as to generate 6′,7′-
diol in 2, could allow additional H-bond interaction and sub-
sequently increased specificity toward eukaryotic ribosome.
Indeed, compound 2 showed substantially enhanced specific-
ity to the eukaryotic cytoplasmic ribosome and induced
higher PTC read-through in comparison to that of 1.10
Encouraged, we reasoned that converting the primary 6′-OH
group in the previously reported 4 and 5 (ref. 4) to the corre-
sponding 6′-carboxylic acid or 6′-amide would improve the
binding of new derivatives into the eukaryotic ribosome, and
subsequently increase the PTC read-through activity. To test
this hypothesis, herein we report on the design, synthesis
and evaluation of a series of pseudo-disaccharides (6 and 7)
During last several years, we systematically developed a
series of lead compounds,4 which was found to improve
readthrough activity and reduce toxicity as demonstrated in
various models of a variety of different diseases underlying
by PTC mutations.5,6 Because of the lack of detailed molecu-
a The Edith and Joseph Fischer Enzyme Inhibitors Laboratory, Schulich Faculty of
Chemistry, Technion – Israel Institute of Technology, Haifa 32000, Israel.
E-mail: chtimor@tx.technion.ac.il; Tel: +972 4 829 2590
b Centre of New Technologies, University of Warsaw, Banacha 2c, 02-097 Warsaw,
Poland. E-mail: joanna@cent.uw.edu.pl; Tel: +48 22 554 3683
c Department of Drug Chemistry, Faculty of Pharmacy with the Laboratory
Medicine Division, Medical University of Warsaw, Banacha 1a, 02-097 Warsaw,
Poland
† Electronic supplementary information (ESI) available: Simulation methods,
biochemical assays, synthetic procedures, and copies of NMR spectra (PDF). See
DOI: 10.1039/c8md00001h
This journal is © The Royal Society of Chemistry 2018
Med. Chem. Commun.

View Article Online
Research Article
MedChemComm
bonds from the Watson–Crick site of the G1408 N1 and N2
amino groups, thereby significantly improve the affinity to-
wards the eukaryotic versus the prokaryotic ribosome (Fig. 2).
Similarly, the planar 6′-amide enhances the efficacy of the
ligands to improve the PTC suppression activity. We expected
that the proposed structural manipulations, 6′-acid and 6′-
amide in 6–11, would preserve all the existing crucial interac-
tions of the parent AG (with 6′-OH) to the ribosomal RNA,
while the additional interactions generated due to added 6′-
acid and/or 6′-amide would increase binding affinity and sub-
sequent PTC read-through activity of the resulting structure.
The synthesis of 6 and 7 are illustrated in Scheme 1. The
known14 compound 12, was selectively protected at the
primary hydroxyl, followed by protection of all the secondary
hydroxyls with benzoate esters to afford the completely
protected 13. Treatment with HF/Py gave the corresponding
6′-alcohol, which was subsequently oxidized using TEMPO/
BIAB15 to afford the acid 14.
Base hydrolysis was followed by the Staudinger reaction to
produce 6 with 6′-CO2H functionality. The corresponding 6′-
CONH2 was synthesized from the intermediate acid 14 by fol-
lowing the procedure reported by Hermann and co-workers16
with some modifications. To a precooled mixture of DMF
and oxalyl chloride at −30 °C was added the intermediate
6′-acid 14 to generate the corresponding acid chloride. The
reaction mixture was cooled to −78 °C and the ammonia gas
was condensed into the vessel. In situ generated acid chloride
reacted with ammonia to afford the amide 15. Once again,
base hydrolysis followed by the Staudinger reaction produced
the fully deprotected target 6′-amide 7. The assembly of the
pseudo-trisaccharides 8–11 is illustrated in Scheme 2.
The intermediate 6′-silyl ether (Scheme 1) was subjected
to a regioselective benzoylation (BzCl) to yield the corre-
sponding C5 acceptor 16. Glycosylation with the trichloro-
acetimidate donors 17 (ref. 4) and 18 (ref. 4) in the presence
of catalytic BF3·OEt2 afforded the protected pseudo-
trisaccharides 19 and 20 exclusively as β-anomers in excellent
Fig. 1 Chemical structures of synthetic AGs that were investigated in
this study.
and pseudo-trisaccharides (8–11), exhibiting either carboxylic
acid or amide function at 6′-position of the glucosamine ring
(ring I) (Fig. 1).
Results and discussion
In selecting the 6′-acid and 6′-amide in 6–11, we have taken
into consideration the following points. Firstly, from the
available data on AGs as readthrough inducers, it has been
well documented that AGs containing a 6′-OH on their ring I
(such as G-418 and paromomycin) are more effective than
those with the amine functionality at the same position.6,11
These observations are usually rationalized by one of the
main differences between their putative binding sites in
eukaryotic vs. prokaryotic ribosomes: the 1408 position (E.
coli numbering, A1408 in bacteria and G1408 in eukary-
otes).12 It has been suggested that while both 6′-OH and
6′-NH2 can form H-bond interactions with A1408 in bacterial
ribosomes, the 6′-NH2 derivatives are prevented from such
interactions with G1408 in eukaryotic ribosomes due to the
electrostatic repulsion between the positively charged nitro-
gen atom of the guanine residue and the 6′-ammonium of
AGs.13 Second, the observed differential selectivity in regards
to the substitution at the 6′ position, is further supported by
the data showing that 6′-NH2 AGs are less effective against
the growth of eukaryotic parasite Leishmania8 and against AG
resistant bacterial strains containing A1408G mutation.13
Finally, we have demonstrated that the increased specificity
toward eukaryotic cytoplasmic ribosome correlates with in-
creased PTC suppression activity.4 These observations, along
with our own modeling studies, suggested that a planar 6′-
carboxylate on ring I might serve as an acceptor of hydrogen
Fig. 2 Three-dimensional structure of ring I of compound 8 docked
into the 80S ribosome from S. cerevisiae (PDB code 4U4O)7
highlighting the interactions of only its ring
I (green) into the
decoding-site rRNA (E. coli numbering). Atom coloring: C – cyan, O –
red, N – blue, P – brown. H-Bonds (dashed black lines) and C–H⋯π
stacking (dotted red line) are shown. The inset shows chemical struc-
ture of ring I of 8.
Med. Chem. Commun.
This journal is © The Royal Society of Chemistry 2018

View Article Online
MedChemComm
Research Article
were confirmed by a combination of various 1D and 2D NMR
1
techniques, including 1D TOCSY, 2D COSY, 2D H–13C HMQC
and HMBC, along with mass spectrometry.
Initially, we tested 6 and 7 for in vitro PTC suppression ef-
ficacy (Fig. 3A). The observed data show that oxidation of 6′-
OH in 3 to yield the 6′-acid in 6, is accompanied with a sig-
nificant reduction in readthrough activity. This reduction in
activity is even more significant in the 6′-amide 7. The ob-
served reduction in the PTC suppression activity of 6 and 7
was further supported by their significantly reduced inhibi-
tion of the eukaryotic protein translation process (Table 1).
The measured half maximal inhibitory concentration values,
Euk
50
IC , for 6 and 7 are 5.4-fold and 4.0-fold higher compared
to that of 3. When the same comparative translation inhibi-
tion tests were performed in the prokaryotic system, the neg-
ative impacts of both 6 and 7 versus that of 3 were more sub-
stantial. For example, the 6′-acid
inhibitor than the paromamine 3 (IC
6
is 293-fold poorer
= 0.01 mM and 2.93
Pro
50
mM, for 6 and 3, respectively). The comparative PTC suppres-
sion and translation inhibition data with the pseudo-
disaccharides 3, 6 and 7 (Fig. 3 and Table 1) show that the
conversion of 6′-OH to corresponding 6′-CO2H or 6′-CONH2
significantly reduces the compounds' ability to inhibit pro-
tein translation process and that this effect is much stronger
in the prokaryotic than in the eukaryotic system. Importantly,
our comparative data on the pseudo-disaccharides 3, 6 and 7
in regards to the bacterial ribosome, are in agreement with
that previously reported by Hermann and co-workers.16 This
group synthesized 6 and 7 and tested them for their impact
only on the inhibition of prokaryotic protein synthesis. Both
Scheme 1 Reagents and conditions: (i) (a) TIPSCl, DMF, 4-DMAP, 0–25
°C, 83%; (b) BzCl, 4-DMAP, Py, 80 °C, 80%; (ii) (a) HF/Py, Py, 0–4 °C,
82%; (b) TEMPO, BIAB, CH2Cl2:H2O, 5 °C, 100%; (iii) (a) NaOMe, MeOH,
0–60 °C, 88%; (b) PMe3, THF, NaOH (0.1 M), 67%; (iv) (a) oxalyl chloride
(COCl)2, DMF, CH2Cl2, −30–0 °C, then NH3IJgas), −78–0 °C, 62%; (v)
NaOMe, MeOH, 0–60 °C, 77%; (b) PMe3, THF, NaOH (0.1 M), 57%.
6 and 7 were about 2-order of magnitude poorer inhibitors
Pro
50
than the parent 3 (IC
values of 3.9, 620 and >100 μM for
Scheme 2 Reagents and conditions: (i) BzCl, 4-DMAP, Py, −5 °C, 44%;
(ii) BF3–OEt2, CH2Cl2, −30 °C; (iii) (a) HF/Py, Py, 0–4 °C; (b) TEMPO,
BIAB, CH2Cl2:H2O, 5 °C, 92% of 21, 94% of 22; (iv) (a) NaOMe, MeOH,
0–60 °C, 89% (RH), ∼100% (RCH3); (b) PMe3, THF, NaOH (0.1 M),
58% of 8, 72% of 10; (v) oxalyl chloride (COCl)2, DMF, CH2Cl2, −30–
0 °C, then NH3 (gas), −78–0 °C, 69% of 23, 71% of 24; (vi) (a) NaOMe,
MeOH, ∼100% (RH), ∼100% (RCH3); (b) PMe3, THF, NaOH (0.1 M),
47% of 9, 71% of 11.
isolated yields. Treatment of the silyl ether products with HF/
Py furnished the corresponding 6′-alcohols, which after oxida-
tion provided the corresponding 6′-acids 21 and 22. Hydrolysis
of the benzoate esters was followed by the Staudinger reaction
to afford the target 6′-acids, 8 and 10. For the synthesis of the
pseudo-trisaccharides with 6′-amide, the intermediate 6′-acids
21 and 22 were first converted to the corresponding 6′-amides
23 and 24; two subsequent deprotection steps provided the tar-
get 6′-amides 9 and 11. The structures of all new compounds
Fig. 3 Comparative in vitro stop codon suppression levels induced by
pseudo-disaccharides (A) paromamine 3, 6′-acid 6 and 6′-amide 7, and
by pseudo-trisaccharides (B and D) compounds 4–5, 8–11 in two differ-
ent nonsense constructs representing genetic diseases (shown in pa-
renthesis): (A–C) R3X (USH1) and (D) G542X (CF). The assays were
performed as previously described by us.4 The results are averages of
at least three independent experiments.
This journal is © The Royal Society of Chemistry 2018
Med. Chem. Commun.

View Article Online
Research Article
MedChemComm
Table 1 Comparative translation inhibition and antibacterial activity testsa
Translation inhibitionb
Antibacterial activity MICc (μg mL−1
)
Euk
50
Pro
Comp.
IC
(mM)
IC (mM)
E. coli R477/100
B. subtilis ATCC6633
50
3
6
7
4
8
9
5
10
11
0.80 0.08
4.32 0.69
3.18 0.42
0.03 0.00
0.70 0.07
0.50 0.04
0.02 0.00
0.60 0.05
0.52 0.04
0.01 0.00
2.93 0.57
2.41 0.29
0.0005 0.0000
1.30 0.10
0.42 0.04
0.002 0.000
1.30 0.08
2.31 0.32
80
>192
>192
>192
790
>192
>192
2659
>192
>192
>192
>192
>192
100
>192
>192
83
1.5
1.3
60
0.5
1.2
10
0.5
0.2
>192
>192
a
All tested compounds, in all biological tests, were in their sulfate salt forms and the concentrations reported refer to that of the free amine
form of each AG. Prokaryotic (IC ) and eukaryotic (IC ) translation inhibition values were quantified as previously described by us.4 c The
b
Pro
Euk
50
50
MIC values were determined by using the double-microdilution method.
3, 6 and 7, respectively), indicating that the chemical func-
tion at the 6′-position of AGs is pivotal in bacterial target site
recognition.
observed inhibition data in prokaryotic translation, is corrob-
orated by the significant reduction in antibacterial activity of
8–11 against both Gram-negative (E. coli) and Gram-positive
(B. subtilis) bacteria, by measuring the minimal inhibitory
concentration (MIC) values. Thus, the comparative biological
data in Fig. 3 and Table 1 clearly show that the negative im-
pact of the 6′-acid and 6′-amide in the pseudo-disaccharides
6 and 7 is largely retained in the subsequent pseudo-
trisaccharides 8–11.
In attempts to provide an explanation on the molecular
level for the observed negative impacts of the 6′-acid and 6′-
amide in the pseudo-trisaccharides 8 and 9, in comparison
with the parent compound 4, we performed full-atom molec-
ular dynamics (MD) followed by random acceleration molecu-
lar dynamics (RAMD)17 simulation studies using NAMD18
package (for details and system set-up see section S1 in
ESI†). Simulations were performed for a 100 000 atom frag-
ment of the 80S ribosome containing the decoding A site and
docked compounds (Fig. S1†). The starting structures for
RAMD simulations were taken from 500 ns long stable MD
trajectories (Fig. S2†). For broader evaluation of the struc-
ture–activity relationship, our recently reported lead com-
pound 2 (ref. 10) was also included. RAMD simulations were
used to compare the kinetic stability of 2, 4, 8 and 9 by
enforcing their dissociation from the A site. Total MD and
RAMD simulation time was 10 μs.
Fig. 4 compares the probability of dissociation of AGs 2, 4,
8 and 9 from the A site, as obtained from RAMD simulations.
In the 2.5 ns time frame, dissociation probability of 4 equals
99% and that of 2 is 91%, suggesting that 4 dissociates faster
than 2. This indicates that the enhanced read-through activ-
ity of 2 (as compared to that of 4)10 may be due to longer
residence time of compound 2 in the A site. The probability
of 9 is about 67%, meaning that in one-third of RAMD
simulations compound 9 did not leave the A site up to 2.5
ns. This result indicates that 9 may have a higher affinity
to A site than compounds 4 and 2. Interestingly, the most sta-
ble complex in decoding A site was created by the 6′-acid 8.
In fact, these previous data of 3, 6 and 7 with the prokary-
otic system was the main trigger for our current work. Thus,
while our observed data with the prokaryotic system is simi-
lar to that of Hermann and co-workers, the scenario in the
eukaryotic system is significantly different (Table 1). While
the compound 3 is 80-fold more selective towards prokaryotic
Euk
Pro
versus eukaryotic ribosome (IC /IC = 80) this selectivity is
50
50
Euk
Pro
dramatically reduced for the compounds 6 and 7, IC /IC
50
50
values of 1.5 and 1.3, respectively. The observed differential
selectivity of the pseudo-disaccharides 6 and 7 versus that of
the parent 3 was very intriguing and we were interested to
know whether this trend would be retained in the corre-
sponding pseudo-trisaccharides.
To test this issue, the pseudo-trisaccharides with the 6′-
acid, compounds 8 and 10, and with the 6′-amide, com-
pounds 9 and 11, were separately evaluated against their par-
ent structures 4 and 5, respectively (Fig. 3). For the PTC sup-
pression tests we used the nonsense reporter plasmids R3X
for Usher syndrome (USH) and G542X for cystic fibrosis (CF).
The observed data in Fig. 3 show that none of the pseudo-
trisaccharides (8–11) exhibit superior activity to the corre-
sponding parents (4 and 5). The activities of 8 and 9 were sig-
nificantly lower than that of the parent 4 in both mutations
(R3X and G452X, Fig. 3B and D, respectively) tested. Introduc-
tion of the exocyclic chiral methyl group as in 10 and 11, did
not restore the readthrough activity (Fig. 3C). Furthermore,
the observed decrease in comparative PTC suppression activ-
ity of 8–11 is supported by their significantly reduced inhibi-
tion of the eukaryotic protein translation (Table 1). The com-
Euk
50
parison of IC
values reveal that 8 and 9 are 23-fold and
16.7-fold poorer inhibitors than their parent 4, and 10 and 11
are 30-fold and 26-fold poorer inhibitors than their parent 5.
Similar to 6 and 7, the comparative translation inhibition
tests of 8–11 in prokaryotic system (Table 1) reveals that in
this system the reduction in activity is more dramatic. The
Med. Chem. Commun.
This journal is © The Royal Society of Chemistry 2018

View Article Online
MedChemComm
Research Article
N7 and its phosphate, and with U1495:O4. These interac-
tions anchor ring II at A site. RAMD trajectories show that
compound 8 interacts with G1408 and C1407 more often
than other AGs (Fig. 5A and B). The carboxyl group of
8 (O6′, ring I) is close to G1408:N1 and N2 (Fig. 5C). From
three 500 ns long MD simulations, the average occurrence
of hydrogen bonds of any O6′ with G1408:N1 is 57% and
with G1408:N2 is about 61%. Another interaction involves
8:N5″ (ring III) that approaches G1408:O6. Also, the ring III
hydroxyl (3″-OH) hydrogen bonds with C1407:N4. Apart
from interactions with G1408, another factor contributing to
the stability of the 8-rRNA complex is the intra-molecular
interaction in 8 formed between ring I (O6′ and O′) and N5″
of ring III (Fig. 5C, green dashed lines). Interestingly, we
also observed similar intra-molecular interaction pattern in
MD simulations of free 8 in solution. Therefore, it seems
that upon binding of 8 to the eukaryotic A site, its intra-
molecular contacts and supposedly its conformation, are
preserved.
Fig. 4 Probability of dissociation of AG from the decoding A site in
time from RAMD simulations. Time from
0 to 2.5 ns considers
simulations in which AGs left the A site by at least 10 Å (section S1†). If
compounds did not dissociate up to 2.5 ns, probabilities are depicted
by Gaussian distributions spanning from 2.5 ns to infinity.
Compound 8 did not leave the A site even once under the
same RAMD force (of 35 kcal mol−1 Å−1) as applied for the
other AGs. Actually, the dissociation of 8 has not occurred
(on a 50 ns time scale) until the RAMD force was increased
up to 65 kcal mol−1 Å−1. The simulation data indicate that
Overall, compound 8 interacts preferentially with G1408,
and this particularly strong interaction diminishes the num-
ber of contacts formed with the opposite backbone U1489–
U1495 including the important A1492/93. Thus the enhanced
stability of the 8-rRNA complex is achieved by its strong inter-
actions (especially of ring I carboxyl group) with the G1408
base and the fact that its intra-molecular bonds in solu-
tion are preserved into the A site. However, due to shifting of
interactions toward the opposite backbone of helix 44, the
mobility of A1492/A1493 residues, related to the efficacy of
the compound 8 in translational fidelity,9 might not be so
much affected. This conclusion is supported by the data of
Pilch and co-workers on the prokaryotic ribosome, demon-
strating that the AG-induced reduction in the mobility of the
A1492 residue in the rRNA A-site is a more important determi-
nant of antibacterial activity than drug affinity for the A-site.20
Compound 9 has binding pattern that involves not only
G1408 but also C1409 (Fig. 5D). The amide nitrogen of 9
(N6′) hydrogen bonds with C1409:N3, while the oxygen of the
amide (O6′) interacts with C1409:N4. The carbonyl O6′ in 9
may also hydrogen bond with G1408:N1 and N2 amines but
less strongly than in compound 8 (Fig. S3†). Similar as for 8,
extra stabilization of 9 in A site is achieved by its intra-
molecular contacts even though these contacts are different
in the A site and free compound 9 (Fig. 5D and S4†). In the A
site, ring I (O6′ and O′) interacts with ring III (N5″); in solu-
tion, however, there is an interaction between N2′ (of ring I)
and O″ (of ring III).
8 and 9 have significantly enhanced stability in the A site, in
Euk
comparison with 2 and 4, even though the IC
values of
50
8 and 9 are one order of magnitude higher than that of com-
pound 4 (Table 1). In addition, since the interactions be-
tween the AGs and rRNA are mainly electrostatic,19 it was sur-
prising that the compound 8-rRNA complex is so stable
despite lower positive total charge of 8 (+3e) in comparison
with +4e of compounds 2, 4, 9. In attempts to explain the ob-
served unique properties of the 6′-acid 8 and the 6′-amide 9,
we analyzed the short-range contacts of 2, 4, 8 and 9 within
the A-site nucleotides as observed in simulations (Fig. 5).
All compounds have a similar interaction pattern with
G1494 and U1495; ring II (N1 and N3) interacts with G1494:
The difference between compounds 2, 4, 8, and 9 lies also
in their interactions with the decoding site adenines: A1491,
A1492, and A1493. Both the 6′-acid 8 and the 6′-amide 9 con-
tact A1491 less frequently than 4 and 2 (Fig. 5A and B). Com-
pound 8 preferentially interacts with A1493 (phosphate oxy-
Fig. 5 (A) Occurrence of short-range interactions (within 3.2 Å) be-
tween AGs and eukaryotic A-site nucleotides derived from RAMD sim-
ulations. Normalization was carried out independently for each AG
complex. (B) Positions of AGs (compound 4) with respect to eukaryotic
A site but with numbering of nucleotides as in E. coli ribosomes.
(C and D) Examples of binding modes of compounds 8 and 9. Inter-
and intra-molecular contacts are marked as black and green dashed
lines, respectively. The distances (in Å) represent averages from trajec-
tory frames of the most occupied cluster. For clarity hydrogen atoms
are not shown.
gens) than with A1492. In contrast, compound
9
preferentially binds to A1492 than to A1493. Interactions of 2
and 4 with A1492 and A1493 are more balanced and they
interact with both adenine phosphates.
This journal is © The Royal Society of Chemistry 2018
Med. Chem. Commun.

View Article Online
Research Article
MedChemComm
Our MD and RAMD simulations suggest that the binding References
mode of the AG at the eukaryotic decoding site might influ-
ence the efficacy of AG-induced stop codon read-through.
The computed binding modes of the 6′-acid 8 and the 6′-
amide 9 are clearly shifted towards the nucleotide stretch
G1405–C1409 of helix 44, in comparison with the binding
modes of the previously reported lead structures 2 and 4.
This shift is more pronounced for the 6′-acid 8 than for the
6′-amide 9, making the stability of 8 highest among all the
AG derivatives examined. In addition, the compounds differ
in their interactions with A1492 and A1493. Compounds 2
and 4 interact with both A1492 and A1493 phosphates but
compound 8 prefers A1493 and compound 9 prefers A1492.
Such imbalance affects the dynamics of this adenine switch,
specifically its backbone, which could be critical for transla-
tional fidelity.
1 S. Magnet and J. S. Blanchard, Chem. Rev., 2005, 105, 477–498.
2 K. M. Keeling and D. M. Bedwell, Curr. Pharmacogenomics,
2005, 3, 259–269.
3 L. V. Zingman, S. Park, T. M. Olson, A. E. Alekseev and A.
Terzic, Clin. Pharmacol. Ther., 2007, 81, 99–103.
4 J. Kandasamy, D. Atia-Glikin, E. Shulman, K. Shapira, M.
Shavit, V. Belakhov and T. Baasov, J. Med. Chem., 2012, 55,
10630–10643.
5 E. Shulman, V. Belakhov, G. Wei, A. Kendall, E. G. Meyron-
Holtz, D. Ben-Shachar, J. Schacht and T. Baasov, J. Biol.
Chem., 2014, 289, 2318–2330.
6 M. Shalev and T. Baasov, Med. Chem. Commun., 2014, 5,
1092–1105.
7 N. G. de Loubresse, I. Prokhorova, W. Holtkamp, M. V.
Rodnina, G. Yusupova and M. Yusupov, Nature, 2014, 513,
517–522.
8 M. Shalev, J. Kondo, D. Kopelyanskiy, C. L. C. L. Jaffe, N.
Adir and T. Baasov, Proc. Natl. Acad. Sci. U. S. A., 2013, 110,
13333–13338.
9 M. Shalev, H. Rozenberg, B. Smolkin, A. Nasereddin, D.
Kopelyanskiy, V. Belakhov, T. Schrepfer, J. Schacht, C. L.
Jaffe, N. Adir and T. Baasov, Nucleic Acids Res., 2015, 43,
8601–8613.
10 N. M. Sabbavarapu, M. Shavit, Y. Degani, B. Smolkin, V.
Belakhov and T. Baasov, ACS Med. Chem. Lett., 2016, 7,
418–423.
11 M. Manuvakhova, K. Keeling and D. M. Bedwell, RNA,
2000, 6, 1044–1055.
12 P. Pfister, S. Hobbie, Q. Vicens, E. C. Böttger and E.
Westhof, ChemBioChem, 2003, 4, 1078–1088.
13 J. Kondo, Angew. Chem., Int. Ed., 2011, 51, 465–468.
14 P. T. Nyffeler, C.-H. Liang, K. M. Koeller and C.-H. Wong,
J. Am. Chem. Soc., 2002, 124, 10773–10778.
15 L. J. van den Bos, J. D. Codee, J. C. van der Toorn, T. J.
Boltje, J. H. Van Boom, H. S. Overkleeft and G. A. van der
Marel, Org. Lett., 2004, 6, 2165–2168.
16 K. B. Simonsen, B. K. Ayida, D. Vourloumis, M. Takahashi,
G. C. Winters, S. Barluenga, S. Qamar, S. Shandrick, Q. Zhao
and T. Hermann, ChemBioChem, 2002, 3, 1223–1228.
17 S. K. Lüdemann, V. Lounnas and R. C. Wade, J. Mol. Biol.,
2000, 303, 797–811.
Conclusions
In summary, the observed structure–activity relationships of
6–11, along with the comparative MD and RAMD simula-
tions, indicate that the rational design of potent PTC read-
through inducers is a complex process and requires consider-
ation of not only the structure and energetics of the drug–
RNA interaction but also the dynamics associated with that
interaction, especially since the eukaryotic A site bulge was
found more dynamically variable than the prokaryotic one.21
This conclusion is supported by a very recently published
crystal structures of the 80S ribosome in complex with
paromomycin, G418, gentamicin and TC007.22 Interestingly,
AGs containing a 6′-NH2 in ring I, including gentamicin and
TC007, do not bind helix 44 at the decoding A site in a
canonical fashion, rather they exhibit multiple binding sites
within the large and small subunits. AGs with a 6′-OH sub-
stituent in ring I, including paromomycin and G418, how-
ever, bind helix 44 in a canonical fashion. It was suggested
that the chemical composition at 6′ position of AGs and their
distinct modes of interaction with 80S ribosome lead to inhi-
bition of intersubunit movement within the eukaryotic ribo-
some that influence PTC read-through efficiency.
Conflicts of interest
The authors declare the following competing financial
interestIJs): T. B. declares that the compounds 8–11 discussed
in this publication are subject to license agreement granted
to a commercial third party.
18 J. C. Phillips, R. Braun, W. Wang, J. Gumbart, E.
Tajkhorshid, E. Villa, C. Chipot, R. D. Skeel, L. Kalé and K.
Schulten, J. Comput. Chem., 2005, 26, 1781–1802.
19 M. Kulik, A. M. Goral, M. Jasiński, P. M. Dominiak and J.
Trylska, Biophys. J., 2015, 108, 655–665.
20 M. Kaul, C. M. Barbieri and D. S. Pilch, J. Am. Chem. Soc.,
2006, 128, 1261–1271.
Acknowledgements
This work was supported by Eloxx Pharmaceutical LTD Re-
search Fund (Grant No. 2019230 for TB) and by the Interdisci-
plinary Centre for Mathematical and Computational Model-
ling University of Warsaw (grants G31-4 and GA65-16, for JT
and TP). N. M. S. thanks to Fine Postdoctoral Fellowship at
the Technion.
21 J. Panecka, J. Šponer and J. Trylska, Biochimie, 2015, 112,
96–110.
22 I. Prokhorova, R. B. Altman, M. Djumagulov, J. P. Shrestha,
A. Urzhumtsev, A. Ferguson, C. T. Chang, M. Yusupov, S. C.
Blanchard and G. Yusupova, Proc. Natl. Acad. Sci. U. S. A.,
2017, 114, E10899–E10908.
Med. Chem. Commun.
This journal is © The Royal Society of Chemistry 2018