Discovery of aminoglycoside mimetics by NMR-based screening of Escherichia coli A-site RNA

Article abstract of DOI:10.1021/ja021354o

A method is described for the NMR-based screening for the discovery of aminoglycoside mimetics that bind to Escherichia coli A-site RNA. Although aminoglycosides are clinically useful, they exhibit high nephrotoxicity and ototoxicity, and their overuse ha

Full text of DOI:10.1021/ja021354o

Published on Web 03/21/2003
Discovery of Aminoglycoside Mimetics by NMR-Based
Screening of Escherichia coli A-site RNA
Liping Yu,* Thorsten K. Oost, Jeffrey M. Schkeryantz,† Jianguo Yang,
Dave Janowick,‡ and Stephen W. Fesik
Contribution from the Pharmaceutical DiscoVery DiVision, GPRD, Abbott Laboratories,
Abbott Park, Illinois 60064-6098
Received November 15, 2002; E-mail: liping.yu@abbott.com
Abstract: A method is described for the NMR-based screening for the discovery of aminoglycoside mimetics
that bind to Escherichia coli A-site RNA. Although aminoglycosides are clinically useful, they exhibit high
nephrotoxicity and ototoxicity, and their overuse has led to the development of resistance to important
microbial pathogens. To identify a new series of aminoglycoside mimetics that could potentially overcome
the problems associated with toxicities and resistance development observed with the aminoglycosides,
we have prepared large quantities of E. coli 16 S A-site RNA and conducted an NMR-based screening of
our compound library in search for small-molecule RNA binders against this RNA target. From these studies,
several classes of compounds were identified as initial hits with binding affinities in the range of 70 µM to
3 mM. Lead optimization through synthetic modifications of these initial hits led to the discovery of several
small-molecule aminoglycoside mimetics that are structurally very different from the known aminoglycosides.
Structural models of the A-site RNA/ligand complexes were prepared and compared to the three-dimensional
structures of the RNA/aminoglycoside complexes.
and viral RNA motifs14 open up numerous opportunities for
the discovery of small molecules that modulate RNA functions.
Introduction
The increasing realization of the essential roles of RNA in
many biological processes and in the progression of diseases
makes RNA an attractive target in drug discovery. The search
for small organic molecules that interact with RNA is therefore
drawing growing interest for drug discovery.1-4 There are many
potential RNA targets, including RNA that is involved in cellular
protein interactions such as transcription, splicing, and transla-
tion, and RNA that is involved in viral infection such as human
immunodeficiency virus (HIV) Rev response element (RRE),5
the trans-activation responsive element (Tar),6 and the hepatitis
C-virus internal ribosome entry site (IRES) RNA.7 The recently
solved crystal structures of the ribosome8-13 and other cellular
One of the most promising RNA targets is the bacterial
ribosome 16 S decoding region aminoacyl-tRNA site (A-site)
RNA. This RNA target is well validated, since aminoglycoside
antibiotics bind to this A-site RNA15 and cause misreading of
the bacterial genetic code, inhibition of translocation, and
bacterial cell death.16,17 Aminoglycosides are effective drugs
and widely used in the therapy against bacterial infections.18
However, aminoglycoside antibiotics display high nephrotoxicity
and ototoxicity in the clinic, resulting in kidney failure, hearing
loss, and deafness.19,20 Furthermore, as is the case with virtually
all classes of antibiotics, bacterial resistance has been widely
developed against this class of antibiotics.19,21,22
† Current address: Eli Lilly and Co., Division of Chemistry Research,
Lilly Corporate Center, Indianapolis, IN 46285.
The structures of the E. coli A-site RNA when free and
complexed with the antibiotics paromomycin and gentamicin
have been determined by NMR and crystallography.11,23-26 The
‡ Current address: Pfizer Inc., La Jolla Laboratories, 3550 General
Atomics Court, San Diego, CA 92121.
(1) Michael, K.; Tor, Y. Chem. Eur. J. 1998, 4, 2091.
(2) Afshar, M.; Prescott, C. D.; Varani, G. Curr. Opin. Biotechnol. 1999, 10,
59.
(3) Gallego, J.; Varani, G. Acc. Chem. Res. 2001, 34, 836.
(4) Swayze, E. E.; Griffey, R. H. Expert Opin. Ther. Pat. 2002, 12, 1367.
(5) Zapp, M. L.; Stern, S.; Green, M. R. Cell 1993, 74, 969.
(6) Wang, S.; Huber, P. W.; Cui, M.; Czarnik, A. W.; Mei, H.-Y. Biochemistry
1998, 37, 5549.
(13) Yusupov, M. M.; Yusupova, G. Z.; Baucom, A.; Lieberman, K.; Earnest,
T. N.; Cate, J. H. D.; Noller, H. F. Science 2001, 292, 883.
(14) Lukavsky, P. J.; Otto, G. A.; Lancaster, A. M.; Sarnow, P.; Puglisi, J. D.
Nature Struct. Biol. 2000, 7, 1105.
(15) Fourmy, D.; Recht, M. I.; Puglisi, J. D. J. Mol. Biol. 1998, 277, 347.
(16) Noller, H. F. Annu. ReV. Biochem. 1991, 60, 191.
(17) McGowan, J. P. Cancer InVest. 1998, 16, 528.
(18) Begg, E. J.; Barclay, M. L. Br. J. Clin. Pharmacol. 1995, 39, 597.
(19) Spelman, D. W.; McDonald, M.; Spicer, W. J. Med. J. Aust. 1989, 151,
346.
(7) Le, S.-Y.; Siddiqui, A.; Maizel, J. V., Jr. Virus Genes 1996, 12, 135.
(8) Ban, N.; Nissen, P.; Hansen, J.; Moore P. B.; Steitz, T. A. Science 2000,
289, 905.
(9) Brodersen, D. E.; Clemons, W. M., Jr.; Carter, A. P.; Morgan-Warren, R.
J.; Wimberly, B. T.; Ramakrishnan, V. Cell 2000, 103, 1143.
(10) Carter, A. P.; Clemons, W. M.; Brodersen, D. E.; Morgan-Warren, R. J.;
Wimberly, B. T.; Ramakrishnan, V. Nature 2000, 407, 340.
(11) Wimberly, B. T.; Brodersen, D. E.; Clemons, W. M., Jr.; Morgan-Warren,
R. J.; Carter, A. P.; Vonrhein, C.; Hartsch, T.; Ramakrishnan, V. Nature
2000, 407, 327.
(20) Slaughter, R. L.; Cappelletty, D. M. Pharmacoeconomics 1998, 14, 385.
(21) Zembower, T. R.; Noskin, G. A.; Postelnick, M. J.; Nguyen, C.; Peterson,
L. R. Int. J. Antimicrob. Agents 1998, 10, 95.
(22) Haddad, J.; Vakulenko, S.; Mobashery, S. J. Am. Chem. Soc. 1999, 121,
11922.
(23) Fourmy, D.; Recht, M. I.; Blanchard, S. C.; Puglisi, J. D. Science 1996,
274, 1367.
(24) Fourmy, D.; Yoshizawa, S.; Puglisi, J. D. J. Mol. Biol. 1998, 277, 333.
(25) Yoshizawa, S.; Fourmy, D.; Puglisi, J. D. EMBO J. 1998, 17, 6437.
(12) Ogle, J. M.; Brodersen, D. E.; Clemons, W. M., Jr.; Tarry, M. J.; Carter,
A. P.; Ramakrishnan, V. Science 2001, 292, 897.
9
4444
J. AM. CHEM. SOC. 2003, 125, 4444-4450
10.1021/ja021354o CCC: $25.00 © 2003 American Chemical Society

Aminoglycoside Mimetics of E. coli A-site RNA
A R T I C L E S
aminoglycosides bind to the pocket formed by the residues
located in the internal loop. The complex formation is stabilized
by a combination of intermolecular hydrogen bonds and
electrostatic and hydrophobic interactions. Previous studies have
primarily centered on chemical modifications of known amino-
glycosides for the improvement of their antibiotic properties.27-31
The goal of this study is to discover small molecules that bind
to the A-site RNA that could serve as aminoglycoside mimetics.
These compounds may exhibit less toxicity and may be effective
against bacteria that are resistant to aminoglycosides and other
classes of antibiotics. There are several methods that could be
used to identify ligands for the A-site RNA. One approach is
the mass-spectrometry-based screening of RNA as recently
described for the 23 S rRNA target.32,33 Another approach
utilizes a fluorescently labeled aminoglycoside as a probe to
discover competitive binders.34 Alternatively, surface plasmon
resonance (SPR) could be employed to detect ligand binding
to RNA by monitoring changes in the optical properties of the
surface when ligands bind to immobilized RNA molecules.28,35
Here, we describe an NMR-based screening approach to
identify ligands that bind to RNA. This approach has the
advantages of identifying not only ligands that bind to RNA
but also the RNA residues that are perturbed by the binding of
the ligands simultaneously. However, this approach requires
relatively large amounts of RNA. To prepare large quantities
of E. coli A-site RNA for the NMR-based screening, we used
a ribozyme approach that generates a high yield of transcrip-
tional RNA and easy separation of the targeted RNA. The
purified E. coli A-site RNA was then used in the NMR-based
screening of our compound library that contains ∼10 000
compounds. The screening hit rate is about 3%. This approach
led to the discovery of several novel series of RNA-binding
ligands that were structurally very different from the known
aminoglycosides.
Figure 1. RNA sequence of the A-site oligonucleotide used in the current
studies. The sequence inside the box corresponds to the decoding region of
E. coli 16 S A-site rRNA. The sequence numbers of the E. coli 16 S rRNA
are indicated in parentheses.
Figure 2. Representative gels of the optimized transcription reaction
products are shown in lanes 1 and 2. The gels were stained with ethidium
bromide. Lanes 1 and 2 were loaded with 2 and 1 µL of the transcription
products. The bands on the gel correspond to the cleaved target E. coli
A-site RNA (29-mer), 5′-hammerhead ribozyme 1 (58-mer), and 3′-
hammerhead ribozyme 2 (77-mer), as well as the uncleaved transcription
product (164-mer). Some other bands around 77-mer correspond to the
partially cleaved products. Even for the overloaded lane 1, the target 29-
mer RNA is still well resolved from the other RNA molecules. Thus, it
can be easily separated.
Results and Discussion
Preparation of Large Quantities of RNA. To apply NMR-
based screening to the E. coli A-site RNA, a key requisition
was to prepare large quantities of RNA. The RNA sequence
used in this study is shown in Figure 1. The E. coli A-site RNA
corresponds to the sequence inside the box. The additional three
base pairs at the top were introduced to stabilize the top base
stem, while a four-base loop at the bottom was constructed to
connect the two RNA strands.
We have adopted the method that uses cis-acting hammerhead
ribozyme flanking at both the 5′ and 3′ ends of the target RNA
sequence for large-scale RNA production.36 There are two
advantages of this method over the traditional transcriptional
runoff reaction. First, higher yields of RNA can be obtained,
since the optimal sequence can be used at the 5′ end of the
hammerhead ribozyme. Second, a clean cut can be obtained at
both the 5′ and 3′ ends of the target sequence. In contrast, when
using the traditional transcriptional runoff reaction,37 the 3′
end of the RNA sequence usually is heterogeneous (with n and
n ( 1 lengths). The separation of the n and n ( 1 bands is
very difficult, especially when purifying large quantities of
target RNA. However, with this approach, it is now much
easier to separate the cleaved target RNA from the hammer-
head ribozymes, which have very different sizes compared to
the 29-mer target RNA (Figure 2). A similar approach has been
reported by using different ribozymes for cleavage.38 Small-
(26) Lynch, S.; Puglisi, J. D. J. Mol. Biol. 2001, 306, 1037.
(27) (a) Wang, H.; Tor, Y. Bioorg. Med. Chem. Lett. 1997, 7, 1951. (b)
Simonsen, K. B.; Ayida, B. K.; Vourloumis, D.; Takahashi, M.; Winters,
G. C.; Barluenga, S.; Qamar, S.; Shandrick, S.; Zhao, Q.; Hermann, T.
ChemBioChem 2002, 3, 1223. (c) Vourloumis, D.; Takahashi, M.; Winters,
G. C.; Simonsen, K. B.; Ayida, B. K.; Barluenga, S.; Qamar, S.; Shandrick,
S.; Zhao, Q.; Hermann, T. Bioorg. Med. Chem. Lett. 2002, 12, 3367.
(28) Wong, C.-H.; Hendrix, M.; Manning, D. D.; Rosenbohm, C.; Greenberg,
W. A. J. Am. Chem. Soc. 1998, 120, 8319.
(29) Alper, P. B.; Hendrix, M.; Sears, P.; Wong, C.-H. J. Am. Chem. Soc. 1998,
120, 1965.
(30) Sucheck, S. J.; Wong, A. L.; Koeller, K. M.; Boehr, D. D.; Draker, K.-A.;
Sears, P.; Wright, G. D.; Wong, C.-H. J. Am. Chem. Soc. 2000, 122, 5230.
(31) Sucheck, S. J.; Greenberg, W. A.; Tolbert, T. J.; Wong, C.-H. Angew.
Chem., Int. Ed. 2000, 39, 1080.
(32) Swayze, E. E.; Jefferson, E. A.; Sannes-Lowery, K. A.; Blyn, L. B.; Risen,
L. M.; Arakawa, S.; Osgood, S. A.; Hofstadler, S. A.; Griffey, R. H. J.
Med. Chem. 2002, 45, 3816.
(33) Griffey, R. H.; Hofstadler, S. A.; Sannes-Lowery, K. A.; Ecker, D. J. Proc.
Natl. Acad. Sci. U.S.A. 1999, 96, 10129.
(34) Tok, J. B.-H.; Rando, R. R. J. Am. Chem. Soc. 1998, 120, 8279.
(35) Hendrix, M.; Priestley, E. S.; Joyce, G. F.; Wong, C.-H. J. Am. Chem.
Soc. 1997, 119, 3641.
(36) Price, S. R.; Ito, N.; Oubridge, C.; Avis, J. M.; Nagai, K. J. Mol. Biol.
1995, 249, 398.
(37) Milligan, J. F.; Groebe, D. R.; Witherell, G. W.; Uhlenbeck, O. C. Nucleic
Acids Res. 1987, 15, 8783.
9
J. AM. CHEM. SOC. VOL. 125, NO. 15, 2003 4445

A R T I C L E S
Yu et al.
Table 1. Dissociation Constants for Members of the
2-Aminobenzimidazole Series that Bind to the
Aminoglycoside-Binding Pocket of the E. coli A-site RNA
1
Figure 3. 1D H spectra of the imino region of the RNA molecule at 25
°C: (A) free RNA; (B) RNA in the presence of paromomycin (1:1 ratio);
(C) typical 1H spectrum of the RNA in the presence of library compounds
that bind to this piece of RNA (compound 1 in this case at 1.0 mM). The
assignments for the imino protons of the RNA are indicated for the free
RNA.
a Dissociation constants were derived from an analysis of the changes
in the imino proton chemical shifts as a function of compound concentration.
The errors in the dissociation constants are ∼16% as estimated by using a
Monte Carlo simulation as previously described.41
scale transcription reactions were usually carried out before a
large batch preparation was undertaken in order to optimize the
buffer components, the NTP concentrations, and the amount of
DNA template and T7 RNA polymerase in the transcription
reaction. We typically obtained 0.8 mg of purified A-site RNA
per mL transcription reaction. The RNA products from a typical
transcription reaction are shown in Figure 2.
The RNA sequence used in this study is identical to that
reported previously39 except that one additional base pair AU
was inserted into the top base stem which is not part of the E.
coli A-site RNA sequence. This insertion was designed such
that the cis-acting 3′ hammerhead ribozyme could cleave the
target RNA at the 3′ end of the target sequence. The purified
RNA binds tightly to paromomycin, as evidenced by the
dramatic shift of the imino protons of U19 and G20 (Figure
3B), which is in good agreement with the results previously
described.39
Identification of Ligands That Bind to the 29-mer RNA.
The imino proton region of the 29-mer RNA is well resolved
and easily assigned from 2D NOESY experiments (Figure 3A).
Therefore, the imino proton region of one-dimensional NMR
spectra was used to monitor the spectral changes upon the
addition of compounds for the identification of ligands that bind
to this RNA target. The screening protocol that was employed
in this study is similar to that published previously for screening
protein targets.40,41 First, mixtures of 10 compounds (at 0.5 mM
each) were added to a 50 µM solution of 29-mer RNA in 20
mM sodium phosphate with a final DMSO concentration of
0.5%. The 1H NMR spectrum of the imino proton region of the
RNA in the presence of compounds was collected and compared
with the corresponding spectrum of the free RNA. If changes
in chemical shifts were observed, the compounds in this mixture
were tested individually in order to identify the compound
responsible for causing the changes in chemical shifts. We
screened about 10 000 compounds from our compound library.
A representative 1H NMR spectrum of the RNA in the presence
of representative screening hit 1 (Table 1) is shown in Figure
3C, where the imino protons of G20, G23, and U19 are clearly
shifted. The RNA binding affinities (Kd) were determined by
fitting the chemical shift data as a function of ligand concentra-
tion as described previously.41 Representative fits are shown in
Figure 4. Since the imino protons of this RNA are well resolved,
usually very selective changes in the NMR spectra were
observed for the RNA binders. The resolved G20 imino proton
oftentimes was monitored for the Kd measurements, since it was
one of the most perturbed peaks upon ligand binding and had
a reasonably narrow line width. We typically collected 6 data
points in the titration for the Kd measurements. Ligand binding
to RNA could also be detected by a saturation transfer difference
(STD) experiment as recently described.42
From NMR-based screening, several classes of compounds
were identified that bind to this RNA molecule (Tables 1 and
2). Benzimidazole derivatives in Table 1 were found to bind in
(41) Hajduk, P. J.; Sheppard, G.; Nettesheim, D. G.; Olejniczak, E. T.; Shuker,
S. B.; Meadows, R. P.; Steinman, D. H.; Carrera, G. M., Jr.; Marcotte, P.
A.; Severin, J.; Walter, K.; Smith, H.; Gubbins, E.; Simmer, R.; Holzman,
T. F.; Morgan, D. W.; Davidsen, S. K.; Summers, J. B.; Fesik, S. W. J.
Am. Chem. Soc. 1997, 119, 5818.
(38) Ferre-D’Amare, A. R.; Doudna, J. A. Nucleic Acids Res. 1996, 24, 977.
(39) Recht, M. I.; Fourmy, D.; Blanchard, S. C.; Dahlquist, K. D.; Puglisi, J.
D. J. Mol. Biol. 1996, 262, 421.
(40) Shuker, S. B.; Hajduk, P. J.; Meadows, R. P.; Fesik, S. W. Science 1996,
274, 1531.
(42) Mayer, M.; James, T. L. J. Am. Chem. Soc. 2002, 124, 13376.
9
4446 J. AM. CHEM. SOC. VOL. 125, NO. 15, 2003

Aminoglycoside Mimetics of E. coli A-site RNA
A R T I C L E S
Table 3. Dissociation Constants for Analogues of the
2-Aminoquinoline Lead that Bind to the Aminoglycoside-Binding
Pocket of the E. coli A-site RNA
Figure 4. Representative fits of the chemical shift of the imino proton
of the ligand-binding pocket nucleotide G20 as a function of ligand concen-
tration. The chemical shift perturbation is the chemical shift difference
(δbound - δfree) of the RNA. The fitted RNA-binding affinities (Kd) are
60 µM for compound 11 (filled square) and 220 µM for compound 4
(filled circle).
a Dissociation constants were derived from an analysis of the changes
in the imino proton chemical shifts as a function of compound concentration.
b Standard deviation was obtained from duplicate measurements.
Table 2. Dissociation Constants for Members of the
2-Aminoquinoline Series that Bind to the Aminoglycoside-Binding
Pocket of the E. coli A-site RNA
6-10 were found to bind the RNA with affinities ranging from
∼90 µM to 3.2 mM (Table 2). In this series, the 2-methylamine
analogue 8 was found to bind more tightly than the 2-amino
and 2-dimethylamino analogues (6, 7). 4-Methyl substitution
(2-aminolepidine series) led to a large increase in binding
affinity (compounds 6 vs 9, and 7 vs 10). These observations
prompted us to synthesize 2-(methylamino)lepidine (11). By
combining the methyl substitutions at the 2-amino and 4
positions, which individually exhibited improved affinity, the
highest affinity analogue 11 in this series was produced (Kd of
60 µM).
The 1D screening method employed in the current study is
simple and fast and does not require 15N- or 13C-labeling of
RNA. This method is applicable to any RNA targets in which
the imino proton region can be resolved. For larger RNA targets
in which the imino protons are severely overlapped, the
screening could be carried out by using 15N- (and/or 13C)-labeled
RNA and by collecting 2D HSQC spectra, which would
significantly reduce the spectral overlap.
Further Optimization of the 2-Aminoquinoline Lead. The
2-aminoquinoline lead was chosen for further optimization. On
the basis of structural information derived from NMR data as
well as from ligand docking using molecular models of the
RNA/ligand complexes, we envisioned that the 4- and 7-position
of the 2-aminoquinoline scaffold could provide vectors toward
additional RNA pockets that are occupied by paromomycin.
According to the previously published synthetic protocols,43,44
7-amino-2-dimethylaminolepidine (12) and 4-amino-2-dimethyl-
aminoquinoline (14) were prepared. Both compounds can readily
be derivatized via their amino moiety. We found that the parent
amines (12 and 14) have increased binding affinities when
compared to the desamino analogues (10 and 7) (Tables 2 and
3). In a parallel synthesis approach, numerous amide, urea, and
carbamate analogues were prepared starting from the core
intermediates 12 and 14. Amides 13 and 15 represent active
a Dissociation constants were derived from an analysis of the changes
in the imino proton chemical shifts as a function of compound concentration.
Based on the pKa calculation with the ACD software, the methyl substitution
at the position 4 increased the pKa by 0.57, while the mono- and dimethyl
substitution at the 2-amino position increased pKa by 0.33 and 0.28,
respectively. The increase in pKa values is correlated with the increase in
RNA binding affinity.
the submillimolar range. In this series, 2-amino substitution led
to a 4-fold increase in binding affinity (compound 1 vs 4).
Methyl substituents on the phenyl portion had little effect on
the binding affinity (compound 2 vs 3-5). 2-Aminoquinolines
(43) Nasr, M.; Drach, J. C.; Smith, S. H.; Shipman, C.; Burckhalter, J. H. J.
Med. Chem. 1988, 31, 1347.
(44) Eiden, F.; Berndl, K. Arch. Pharm. (Weinheim, Ger.) 1986, 319, 347.
9
J. AM. CHEM. SOC. VOL. 125, NO. 15, 2003 4447

A R T I C L E S
Yu et al.
Table 4. Dissociation Constants for Members of the
2-Aminopyridine Series that Bind to the Aminoglycoside-Binding
Pocket of the E. coli A-site RNA
1
Figure 5. 1D H spectra of the imino region of the RNA molecule at 20
°C: (A) 80 µM free RNA (sample A); (B) sample A plus 110 µM neamine
(sample B); (C) sample B plus 110 µM compound 18; (D) sample B plus
220 µM compound 18; (E) sample A plus 80 µM paromomycin (sample
E); (F) sample E plus 160 µM compound 18. The assignments for the imino
protons of the RNA are indicated for the free RNA.
a Dissociation constants were derived from an analysis of the changes
in the imino proton chemical shifts as a function of compound concentration.
b Standard deviations were obtained from duplicate measurements.
paromomycin complex or in a flipped up orientation occupying
the space above ring II near G5 nucleotide.15 The binding of
neamine to the RNA caused a shift of imino proton peaks of
U19, G5, G20, and U6/U24 (Figure 5B), which is consistent
with the reported structural data.15 Addition of compound 18
to the RNA/neamine complex caused a shift reversal of the
imino protons of U19, G5, and G20, indicating that compound
18 competes with the binding of neamine (Figure 5C,D).
Similarly, the binding of paromomycin caused major chemical
shifts of the imino protons of U19 and G20 (Figure 5E). This
shift pattern is identical to that previously reported.15,39 Addition
of compound 18 to the RNA/paromomycin complex partially
displaced paromomycin, as indicated by the presence of a new
set of peaks of U19 and G20 (Figure 5F) when compared to
Figure 5E. Further addition of compound 18 caused line-
broadening of the peaks. These results provide additional
evidence that compound 18 binds to the paromomycin-binding
site.
It is interesting to note that the U13/U14 and U6/U24 imino
peaks are severely broadened in the presence of compound 18
and neamine or paromomycin. These uridine imino peaks are
generally broader than the rest of the imino protons (see Figure
3A), and their line widths are sensitive to sample temperature.
When compound 18 alone was titrated into the RNA, the U6/
U24 and G20 imino protons had very large chemical shift
perturbations, while the U13/U14 and G16 imino protons
experienced no chemical shift changes. Therefore, we are
convinced that compound 18 primarily binds to the paromo-
mycin-binding pocket where these perturbed residues are
located. Chemical exchange could have caused the broadening
of these uridine imino peaks in the presence of multiple ligands
at relatively high concentrations.
library members. Compared to the parent amines, they exhibit
improved RNA binding affinities with Kd values of 18 and 9
µM, respectively (Table 3).
Optimization of the 2-Aminopyridine Lead. During the
initial NMR-based screening, 2-aminopyridine 16 was found
to bind to the E. coli A-site RNA with an affinity of 68 µM
(Table 4). This screening hit was attractive, since it is even
smaller than the previously discovered 2-aminoquinolines and
yet binds to the RNA with good affinity. Therefore, a series of
analogues was prepared to test whether the RNA binding affinity
of this hit can be further improved. A systematic “methyl scan”
of the pyridine ring revealed an interesting structure-activity
relationship. The 3- and 6-methyl substitutions (17 and 20) led
to decreased binding, whereas, the 4- and 5-substitutions (18
and 19) resulted in an improved affinity. The 4-methyl analogue
18 was the tightest binder in this series, with a Kd of 3 µM,
which represents a 22-fold improvement over the parent
compound 16. Interestingly, we had observed a similar increase
in binding affinity earlier when we switched from the 2-amino-
quinoline to the 2-aminolepidine series (Table 2, e.g., compound
8 vs 11). The homologous analogue 21, with an (aminopropyl)-
amino side chain, was also found to bind to the RNA with
good affinity. Analogues with an electron-withdrawing 5-nitro
substituent (22) were devoid of binding affinity, which may be
attributed to the significantly reduced basicity of the 2-amino-
pyridine moiety.
Competition Experiments. The antibiotic neamine is known
to bind the E. coli A-site RNA15 with a Kd of ∼8 µM based on
our NMR titration. Neamine consists of rings I and II of
paromomycin39 and binds to the RNA in two conformations
with the ring I either occupying the same position as in the
Molecular Modeling of the 29-mer RNA/Ligand 18
Complex. To provide information on how compound 18 binds
9
4448 J. AM. CHEM. SOC. VOL. 125, NO. 15, 2003

Aminoglycoside Mimetics of E. coli A-site RNA
A R T I C L E S
Figure 6. Structure of the E. coli A-site RNA in complex with paromomycin (A). Structural model of the lead compound 18 when bound to the aminoglycoside-
binding pocket of the E. coli A-site RNA (B). The four subunits (I-IV) of paromomycin are labeled in A. Nucleotides G20 and A22 are colored in yellow.
Nucleotides U6 and U24 are colored in red and magenta, respectively.
Table 5. Intermolecular NOEs Observed between 29-mer RNA
and Compound 18 at a Mixing Time of 150 ms
of compound 17 in which the 3-methyl substitution could cause
steric hindrance and prevent the formation of these hydrogen
bonds.
Conclusions. To address the problems associated with
Protons on
Ligand 18
Protons on
29-mer RNA
Relative NOE
Intensity
4-methyl
4-methyl
H5
H5
H3
A22 H8
A21 H8
A22 H8
A21 H8
A22 H2
U6 H5
medium
medium
weak
weak
weak
toxicity and development of resistance against aminoglycoside
antibiotics, we chose the E. coli A-site RNA as a target and
screened for small molecules that bind to this site and that could
potentially be developed into novel aminoglycoside mimetics.
Large quantities of E. coli A-site RNA were generated for the
NMR-based screening and NMR-based structural studies using
a bis-ribozyme construct. High-throughput NMR-based screen-
ing of our compound library delivered several series of initial
hits that were further optimized. This led to the rapid discovery
of small-molecule RNA binders that bind to the target RNA in
the single digit micromolar range.
Aminoglycosides are known to bind to many different RNA
targets. For example, the aminoglycoside neomycin binds to
the prokaryotic 16 S A-site RNA, group I intron RNA, HIV
TAR element, and hammerhead ribozyme.5,6,15,45 Therefore, the
binding of the aminoglycosides to RNAs is not very specific.
The compounds discovered from the current study likely also
bind to other RNA targets, although they may exhibit different
binding affinities. With the rapidly expanding knowledge of the
three-dimensional structures of RNA/ligand complexes and the
better understanding of RNA dynamics, it may be possible to
design specific RNA inhibitors through the use of maximized
shape and electrostatic complementarity and hydrogen-bond
formation. The leads from this study could serve as novel
templates, which can be further developed into possible replace-
ments for the aminoglycoside antibiotics, which suffer from
toxicity and resistance development.
H9a
H9b
strong
strong
U6 H5
to the A-site RNA, NOE experiments were performed. A total
of seven intermolecular NOEs were obtained for the 29-mer
RNA/compound 18 complex at a mixing time of 150 ms (Table
5). Although this small number of intermolecular NOEs was
insufficient to calculate well-resolved structures of this complex,
they did allow a model of the complex to be prepared (Figure
6B).
The model was generated by docking 18 into the free E. coli
A-site RNA structure previously determined by Puglisi and co-
workers,24 and the docking was guided by the observed
intermolecular NOE restraints (Table 5). The NOEs observed
between the compound pyridine ring 4-methyl and H5 protons
(see compound 18 atom numbering in Table 4) and the RNA
A21 and A22 H8 protons place this pyridine ring near a location
that is occupied by rings I and II of paromomycin (Figure 6A).
The compound pyridine ring stacks above the base G20 and
against the base of A21. The observed NOEs between A22 H2
and U24 H5 indicate that the base of A22 has rotated toward
the inside of the ligand-binding pocket and stacks on the top of
the compound pyridine ring. This ligand and RNA base stacking
is consistent with the observed intermolecular NOEs between
A22 H2 and compound pyridine H3 and between A22 H8 and
pyridine 4-methyl and H5. The side-chain off the 2-amino
pyridine ring of compound 18 points toward the base of U6.
This is evidenced by the strong NOEs observed between the
ligand H9 methylene protons and RNA U6 H5, but not between
the ligand H8 methylene protons and RNA U6 H5. On the basis
of the model, the positively charged terminal NH3 group of the
ligand is positioned to form an electrostatic interaction with the
negatively charged phosphate of G5. The ligand NH (H7) likely
forms hydrogen bonds with G23 O6 and/or U6 O4. This binding
mode is also consistent with the significantly reduced binding
Experimental Section
Preparation of E. coli A-Site Oligonucleotide RNA. The DNA
sequences of the T7 promoter (ATAATACGACTCACTATA), 58-mer
5′-hammerhead ribozyme 1 (GGGAGAGTGTGACGCTCCTGAT-
GAGTCCGTGAGGACGAAACGGTATCTAGATACCGTC), 29-mer
target RNA (GAGCGTCACACCTTCGGGTGAAGTCGCTC), and
77-mer 3′-hammerhead ribozyme 2 (GACGGAGTCTAGACTCCGTC-
CTGATGAGTCCGTGAGGACGAAAGCGACTAAATAAACCAA-
CTGCCGGCATGCAAGCT) were cloned into pUC18 vector with an
EcoR I cleavage site at the 5′ end and Hind III cleavage site at the 3′
(45) (a) von Ahsen, U.; Davies, J.; Schroeder, R. J. Mol. Biol. 1992, 226, 935.
(b) Hermann, T.; Westhof, E. J. Mol. Biol. 1998, 276, 903.
9
J. AM. CHEM. SOC. VOL. 125, NO. 15, 2003 4449

A R T I C L E S
Yu et al.
end of the DNA sequence. The vector was then transformed into a
DH5R cell, and its inserted DNA sequence was confirmed from
sequencing gels. After growing a large batch of the transformed cells,
the plasmid DNA was isolated with Maxi or Mega kits from Qiagen.
Subsequently, the purified plasmid DNA was linearized through cutting
with Hind III restriction enzyme and further purified with phenol/
CH3Cl extraction. The linearized DNA was then used as a template in
the in vitro transcription reaction.
The in vitro transcription reaction was carried out in a transcription
buffer typically containing 40 mM Tris-HCl (pH 7.9), 2 mM spermidine,
40 mM MgCl2, 10 mM NaCl, 10 mM DTT, 21.2 mM NTP (4.0 mM
UTP, 5.3 mM CTP, 5.5 mM ATP, and 6.4 mM GTP, which are
optimized based upon the nucleotide composition of the transcribed
sequence), plasmid DNA template 28 nM, 0.05% Triton-X100, and
T7 RNA polymerase 2000 units/mL reaction. The amount of MgCl2
needed in the transcription buffer is very sensitive to the amount of
NTP present as well as how the NTPs were prepared. For instance, the
in-house-prepared 13C- and 15N-labeled NTPs may already contain a
significant amount of Mg2+. Thus, much less MgCl2 was added in the
transcription buffer. The T7 RNA polymerase was expressed and
purified from a N-terminal His-tagged version of this protein. The
products from the in vitro transcription reaction were separated from
20% (w/v) polyacrylamide (29:1 (w/w) acrylamide:bis)/7 M urea gels.
The unlabeled RNA was obtained by using unlabeled NTPs purchased
from Sigma, and [15N,13C]-doubly labeled RNA was synthesized in the
in vitro transcription reaction by employing [15N,13C]-doubly labeled
NTPs prepared as previously described.46,47
Detection of Ligand Binding. Ligand binding was detected by
acquiring 1D spectra with the jump-return method48 using a 500 µL
sample of 50 µM E. coli A-site RNA in 20 mM sodium phosphate
(pH 6.5) in the presence and absence of added compound. Compounds
were added as solutions in perdeuerated DMSO or in aqueous buffer
of 20 mM sodium phosphate (pH 6.5). We screened about 10 000
compounds from our compound library. Compounds were initially
tested at 0.5 mM each, and binding was determined by monitoring
changes in the imino proton region of the 1D 1H spectra. Dissociation
constants were obtained by fitting the chemical shift as a function of
ligand concentration through a least-squares grid search by varying the
values of dissociation constant Kd and the chemical shift of the fully
saturated RNA.41 For several selected compounds, duplicate experiments
were carried out by using 40 µM RNA samples, and their average Kd
values and standard deviations were reported.
otherwise specified temperature. Pulsed field gradients were applied
where appropriate as described to afford suppression of solvent signal
and spectral artifacts. 1D spectra for the samples in water were acquired
with the jump-return method.48 RNA/ligand complexes were assigned
from COSY, TOCSY, and NOESY spectra.49 The data were processed
and analyzed on Silicon Graphics computers using in-house written
software.
Molecular Modeling. To provide information on how compound
18 binds to the A-site RNA, compound 18 was titrated into the RNA
with increasing compound concentrations, and a series of TOCSY and
NOESY spectra with different mixing times were collected at 35 °C at
each ligand concentration. This allowed us to trace the cross-peaks that
were shifted upon the addition of the compound. On the basis of the
published assignments of the free A-site RNA by Puglisi and co-
workers,24 these perturbed peaks/residues were shown to be located in
the paromomycin-binding pocket. Seven intermolecular NOEs were
observed at mixing time 150 ms (Table 5). Additional intermolecular
NOEs between ligand H8 and H9 methylene protons and A22 H2 and
between ligand H6 and A22 H8 and A21 H8 were also observed at
mixing times of 200 and 300 ms.
Ligand 18 was docked into the free A-site E. coli RNA structure
that was previously determined by Puglisi and co-workers.24 The
docking was guided by the NMR-derived NOE restraints with the
InsightII program. No solvent and sodium counterions were added, and
no electrostatic potential was used in the docking. On the basis of the
observed NOE connectivity between A22 H2 and U24 H5, the
nucleotide 22 was rotated along the phosphate backbone toward the
inside of the binding pocket from the initial structure and the RNA
structure was then energy-minimized. As a result of this rotation, the
base of A22 stacks against the ligand pyridine ring, which is consistent
with the observed intermolecular NOEs. During the NOE assignments,
it is observed that, at a low ratio of compound to RNA, A22 H2 has
NOEs to both U24 H5 and G23 H1′, indicating that the base of A22
has a mixture of free and ligand-bound conformations. Either using
NMR samples that contain higher ligand to RNA ratio or lowering
experimental temperature drives the RNA to the complete bound
conformation.
Supporting Information Available: Experimental procedures
and spectral characterization for the compounds described in
the text. This material is available free of charge via the Internet
at http://pubs.acs.org.
NMR Spectroscopy. All NMR spectra were collected on a Bruker
DRX500, DRX600, or DRX800 NMR spectrometer at 25 °C or
(46) Nikonowicz, E. P.; Sirr, A.; Legault, P.; Jucker, F. M.; Baer, L. M.; Pardi,
A. Nucleic Acids Res. 1992, 20, 4507.
JA021354O
(47) Batey, R. T.; Battiste, J. L.; Williamson, J. R. Methods Enzymol. 1995,
261, 300.
(49) Wu¨thrich, K. NMR of Proteins and Nucleic Acids; John Wiley & Sons:
(48) Plateau, P.; Gueron, M. J. Am. Chem. Soc. 1982, 104, 7310.
New York, 1986.
9
4450 J. AM. CHEM. SOC. VOL. 125, NO. 15, 2003