Two-dimensional combinatorial screening identifies specific aminoglycoside-RNA internal loop partners

Article abstract of DOI:10.1021/ja803234t

Herein is described the identification of RNA internal loops that bind to derivatives of neomycin B, neamine, tobramycin, and kanamycin A. RNA loop-ligand partners were identified by a two-dimensional combinatorial screening (2DCS) platform that probes RNA and chemical spaces simultaneously. In 2DCS, an aminoglycoside library immobilized onto an agarose microarray was probed for binding to a 3 x 3 nucleotide RNA internal loop library (81 920 interactions probed in duplicate in a single experiment). RNAs that bound aminoglycosides were harvested from the array via gel excision. RNA internal loop preferences for three aminoglycosides were identified from statistical analysis of selected structures. This provides consensus RNA internal loops that bind these structures and include: loops with potential GA pairs for the neomycin derivative, loops with potential GG pairs for the tobramycin derivative, and pyrimidine-rich loops for the kanamycin A derivative. Results with the neamine derivative show that it binds a variety of loops, including loops that contain potential GA pairs that also recognize the neomycin B derivative. All studied selected internal loops are specific for the aminoglycoside that they were selected to bind. Specificity was quantified for 16 selected internal loops by studying their binding to each of the arrayed aminoglycosides. Specificities ranged from 2- to 80-fold with an average specificity of 20-fold. These studies show that 2DCS is a unique platform to probe RNA and chemical space simultaneously to identify specific RNA motif-ligand interactions.

Full text of DOI:10.1021/ja803234t

Published on Web 07/25/2008
Two-Dimensional Combinatorial Screening Identifies Specific
Aminoglycoside-RNA Internal Loop Partners
Matthew D. Disney,*,† Lucas P. Labuda,† Dustin J. Paul,† Shane G. Poplawski,‡
Alexei Pushechnikov,† Tuan Tran,† Sai P. Velagapudi,† Meilan Wu,† and
Jessica L. Childs-Disney‡
Department of Chemistry, UniVersity at Buffalo, The State UniVersity of New York, and the
New York State Center of Excellence in Bioinformatics and Life Sciences, 657 Natural Sciences
Complex, Buffalo, New York 14260, and Department of Chemistry & Biochemistry,
Canisius College, 2001 Main Street, Buffalo, New York 14208
Received May 1, 2008; E-mail: mddisney@buffalo.edu
Abstract: Herein is described the identification of RNA internal loops that bind to derivatives of neomycin
B, neamine, tobramycin, and kanamycin A. RNA loop-ligand partners were identified by a two-dimensional
combinatorial screening (2DCS) platform that probes RNA and chemical spaces simultaneously. In 2DCS,
an aminoglycoside library immobilized onto an agarose microarray was probed for binding to a 3 × 3
nucleotide RNA internal loop library (81 920 interactions probed in duplicate in a single experiment). RNAs
that bound aminoglycosides were harvested from the array via gel excision. RNA internal loop preferences
for three aminoglycosides were identified from statistical analysis of selected structures. This provides
consensus RNA internal loops that bind these structures and include: loops with potential GA pairs for the
neomycin derivative, loops with potential GG pairs for the tobramycin derivative, and pyrimidine-rich loops
for the kanamycin A derivative. Results with the neamine derivative show that it binds a variety of loops,
including loops that contain potential GA pairs that also recognize the neomycin B derivative. All studied
selected internal loops are specific for the aminoglycoside that they were selected to bind. Specificity was
quantified for 16 selected internal loops by studying their binding to each of the arrayed aminoglycosides.
Specificities ranged from 2- to 80-fold with an average specificity of 20-fold. These studies show that 2DCS
is a unique platform to probe RNA and chemical space simultaneously to identify specific RNA motif-ligand
interactions.
expression12,13 and diseases such as cancer.14,15 Small molecules
can also target mRNA by binding to riboswitches in untranslated
Introduction
RNA is an important target for therapeutic intervention due
to its essential biological functions.1–8 These functions are often
predicated on both the RNA sequence and structure. The most
commonly exploited RNA drug target for small-molecule
intervention is bacterial rRNA.
regions5,6,16–18 or in the coding regions themselves. Part of the
reason for the underexploitation of RNA drug targets is the
limited information available on the RNA secondary structure
motifs that bind small molecules and the features in small
molecules that are important for binding RNA structures. If such
information were known, then it would facilitate the develop-
ment of improved methods for the design of probes or drugs
targeting RNA.
9,10
Other less explored RNA
drug targets are pre-microRNAs that are processed into
microRNAs,1,11 which play important roles in regulating gene
† University at Buffalo, The State University of New York, and the New
York State Center of Excellence in Bioinformatics and Life Sciences.
‡ Canisius College.
There are two main methods currently used to identify
RNA-ligand interactions: selective evolution of ligands by
exponential enrichment (SELEX) and high-throughput screening
of small molecules. SELEX finds an RNA that binds a ligand
(1) Lau, N. C.; Lim, L. P.; Weinstein, E. G.; Bartel, D. P. Science 2001,
294, 858–62.
(2) Zaug, A. J.; Cech, T. R. Science 1986, 231, 470–5.
(3) Guerrier-Takada, C.; Gardiner, K.; Marsh, T.; Pace, N.; Altman, S.
Cell 1983, 35, 849–57.
(4) Vonahsen, U.; Noller, H. F. Science 1993, 260, 1500–3.
(5) Winkler, W. C.; Cohen-Chalamish, S.; Breaker, R. R. Proc. Natl. Acad.
Sci. U.S.A. 2002, 99, 15908–13.
(12) He, Y. D.; Friedman, A.; Perrimon, N.; Medina-Martinez, O.; Jamrich,
M.; Zhao, Y.; Srivastava, D.; Miller, L. D.; Liu, E. T. Cancer Biomark.
2006, 2, 103–33.
(6) Winkler, W.; Nahvi, A.; Breaker, R. R. Nature 2002, 419, 952–6.
(7) Gallego, J.; Varani, G. Acc. Chem. Res. 2001, 34, 836–43.
(8) Disney, M. D.; Childs, J. L.; Turner, D. H. Biopolymers 2004, 73,
151–61.
(13) Zhao, Y.; Srivastava, D. Trends Biochem. Sci. 2007, 32, 189–97.
(14) Calin, G. A.; Croce, C. M. Oncogene 2006, 25, 6202–10.
(15) Yeung, M. L.; Bennasser, Y.; Jeang, K. T. Curr. Med. Chem. 2007,
14, 191–7.
(9) Tenson, T.; Mankin, A. Mol. Microbiol. 2006, 59, 1664–77.
(10) Poehlsgaard, J.; Douthwaite, S. Nat. ReV. Microbiol. 2005, 3, 870–
81.
(16) Nahvi, A.; Sudarsan, N.; Ebert, M. S.; Zou, X.; Brown, K. L.; Breaker,
R. R. Chem. Biol. 2002, 9, 1043–9.
(17) Blount, K. F.; Breaker, R. R. Nat. Biotechnol. 2006, 24, 1558–64.
(18) Blount, K. F.; Wang, J. X.; Lim, J.; Sudarsan, N.; Breaker, R. R. Nat.
Chem. Biol. 2007, 3, 44–9.
(11) Lagos-Quintana, M.; Rauhut, R.; Lendeckel, W.; Tuschl, T. Science
2001, 294, 853–8.
9
10.1021/ja803234t CCC: $40.75
2008 American Chemical Society
J. AM. CHEM. SOC. 2008, 130, 11185–11194 11185

A R T I C L E S
Disney et al.
by multiple rounds of selection.19 These RNA-ligand com-
plexes can be used as sensors20 or as biochemical tools.21 The
aptamer-ligand complex that is the output of SELEX, however,
is difficult to apply to RNA targeting endeavors because the
selected RNAs are large (derived from a g20-nucleotide random
region)19,22 and unlikely to be found in a biologically relevant
RNA. Despite this impediment, there have been cases where
parts of an RNA aptamer identified via SELEX have been found
in a biological RNA.23,24 In high-throughput screening of small
molecules, the more common approach in RNA drug discovery,
a library of chemical ligands is probed for binding a validated
RNA drug target such as the bacterial rRNA aminoacyl-tRNA
site (A-site),25–30 HIV trans-activating responsive element (TAR)
and Rev responsive element (RRE) RNAs,31,32 or the hepatitis
C internal ribosomal entry site.33,34 When applied toward
“druglike ligands”, however, high-throughput screening has been
hampered by low hit rates, much lower rates than are typically
found by screening protein targets. Part of the reason for this is
little is known about what types of RNA structures like to bind
organic ligands and what types of organic ligands like to bind
RNA structures.
By merging chemical (high-throughput screening of drugs)
and RNA (selection) screening, we envision that features in both
the RNA and ligand that facilitate molecular recognition can
be identified.35,36 This would be especially useful if the RNAs
identified that bound ligands were small RNA motifs (hairpins,
loops, or bulges) that are typically found in RNA targets. These
results could help to establish a database that could serve as a
rational design tool for the modular construction of RNA-
targetingligands.Wepreviouslyreportedachemicalmicroarray35–45
platform to screen RNA libraries for binding small molecules.35
In that report, a selection was completed by use of an RNA
internal loop library and 6′-N-derivatized kanamycin A. It was
determined that the kanamycin derivative preferred loops with
potential AC pairs. The platform was also used to demonstrate
that an array with four related aminoglycosides could be probed
simultaneously for binding to an RNA library. The RNAs that
bound to each aminoglycoside, however, were not sequenced
nor were their binding affinities studied. Herein, we describe
the results from screening chemical and RNA spaces simulta-
neously (named two-dimensional combinatorial screening,
2DCS) to find the small RNA internal loops that bind to
derivatives of four aminoglycosides (kanamycin A, tobramycin,
neamine, and neomycin B). Results show that the RNA
motif-ligand partners identified by 2DCS are specific for the
aminoglycosides for which they were selected. Sequencing and
structure modeling of the selected internal loops show that
neomycin prefers loops with potential GA pairs, tobramycin
prefers RNA internal loops with potential GG pairs, and
kanamycin A prefers internal loops with potential pyrimidine-
pyrimidine pairs. Neamine binds to a variety of internal loops,
including internal loops with potential GA pairs.
Materials and Methods
Synthesis. All azido-aminoglycosides were synthesized from the
corresponding free base forms of the parent aminoglycosides
according to previously published procedures.35–37 The azido-
aminoglycosides were fluorescently labeled with 5-fluorescein
isothiocyanate (5-FITC, Toronto Research Chemicals) by a two-
step procedure in which the Boc-protected azido-aminoglycosides
were reacted with propargylamine to install a free amine that was
then conjugated to 5-FITC.46
Construction of Alkyne-Displaying Microarrays. Microarrays
were constructed as described.35,36 Briefly, ∼2 mL of a 1% agarose
solution was applied to a Silane-Prep slide (Sigma-Aldrich Co.,
St. Louis, MO), and the agarose was allowed to dry to a thin film
at room temperature. The agarose was oxidized by submerging the
slides in 20 mM NaIO4 for 30 min.47 The oxidized agarose slides
were then washed in water (3 × 30 min) with frequent water
changes. Residual NaIO4 was removed by incubating the slides in
10% aqueous ethylene glycol for 1.5 h at room temperature, which
was followed by washing with water as described above. The slides
were then incubated with 20 mM propargylamine in 0.1 M NaHCO3
overnight and reduced with NaCNBH3 the following morning (100
mg in 40 mL of 1× phosphate-buffered saline + 10 mL of ethanol;
3 min at room temperature). Slides were washed with water and
dried before use.
(19) Osborne, S. E.; Ellington, A. D. Chem. ReV. 1997, 97, 349–70.
(20) Ellington, A. D.; Szostak, J. W. Nature 1990, 346, 818–22.
(21) Werstuck, G.; Green, M. R. Science 1998, 282, 296–8.
(22) Joyce, G. F. Curr. Opin. Struct. Biol. 1994, 4, 331–6.
(23) Tao, J.; Frankel, A. D. Biochemistry 1996, 35, 2229–38.
(24) Carlson, C. B.; Vuyisich, M.; Gooch, B. D.; Beal, P. A. Chem. Biol.
2003, 10, 663–72.
Construction of Azido-aminoglycoside Microarrays. Azido-
aminoglycosides were spotted onto alkyne-functionalized slides in
10 mM Tris-HCl (pH 8.5), 100 µM tris(benzyltriazolylmethyl)amine
(TBTA)48 (dissolved in 4:1 butanol/dimethyl sulfoxide, DMSO),
1 mM CuSO4, 1 mM ascorbic acid, and 10% glycerol. The slides
were placed into a humidity chamber for 2 h and then washed with
water. The slides were dried at room temperature.
(25) Wong, C. H.; Hendrix, M.; Manning, D. D.; Rosenbohm, C.;
Greenberg, W. A. J. Am. Chem. Soc. 1998, 120, 8319–27.
(26) Disney, M. D.; Seeberger, P. H. Chemistry 2004, 10, 3308–14.
(27) 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, 3816–9.
(35) Childs-Disney, J. L.; Wu, M.; Pushechnikov, A.; Aminova, O.; Disney,
M. D. ACS Chem. Biol. 2007, 2, 745–54.
(36) Disney, M. D.; Childs-Disney, J. L. ChemBioChem 2007, 8, 649–56.
(37) Disney, M. D.; Barrett, O. J. Biochemistry 2007, 46, 11223–30.
(38) Ratner, D. M.; Adams, E. W.; Disney, M. D.; Seeberger, P. H.
ChemBioChem 2004, 5, 1375–83.
(28) Jefferson, E. A.; Arakawa, S.; Blyn, L. B.; Miyaji, A.; Osgood, S. A.;
Ranken, R.; Risen, L. M.; Swayze, E. E. J. Med. Chem. 2002, 45,
3430–9.
(29) Yu, L.; Oost, T. K.; Schkeryantz, J. M.; Yang, J.; Janowick, D.; Fesik,
S. W. J. Am. Chem. Soc. 2003, 125, 4444–50.
(39) Hergenrother, P. J.; Depew, K. M.; Schreiber, S. L. J. Am. Chem.
Soc. 2000, 122, 7849–50.
(30) Disney, M. D.; Magnet, S.; Blanchard, J. S.; Seeberger, P. H. Angew.
Chem., Int. Ed. 2004, 43, 1591–4.
(40) MacBeath, G.; Koehler, A. N.; Schreiber, S. L. J. Am. Chem. Soc.
1999, 121, 7967–8.
(31) Mei, H. Y.; Mack, D. P.; Galan, A. A.; Halim, N. S.; Heldsinger, A.;
Loo, J. A.; Moreland, D. W.; Sannes-Lowery, K. A.; Sharmeen, L.;
Truong, H. N.; Czarnik, A. W. Bioorg. Med. Chem. 1997, 5, 1173–
84.
(41) Barnes-Seeman, D.; Park, S. B.; Koehler, A. N.; Schreiber, S. L.
Angew. Chem., Int. Ed. 2003, 42, 2376–9.
(42) Koehler, A. N.; Shamji, A. F.; Schreiber, S. L.; Barnes-Seeman, D.;
Park, S. B. J. Am. Chem. Soc. 2003, 125, 8420–1.
(43) Fukui, S.; Feizi, T.; Galustian, C.; Lawson, A. M.; Chai, W. Nat.
Biotechnol. 2002, 20, 1011–7.
(32) Peled-Zehavi, H.; Horiya, S.; Das, C.; Harada, K.; Frankel, A. D. RNA
2003, 9, 252–61.
(33) Seth, P. P.; Miyaji, A.; Jefferson, E. A.; Sannes-Lowery, K. A.;
Osgood, S. A.; Propp, S. S.; Ranken, R.; Massire, C.; Sampath, R.;
Ecker, D. J.; Swayze, E. E.; Griffey, R. H. J. Med. Chem. 2005, 48,
7099–102.
(44) Blixt, O.; et al. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 17033–8.
(45) Lee, M. R.; Shin, I. Org. Lett. 2005, 7, 4269–72.
(46) Details of chemical synthesis are provided in Supporting Information.
(47) Afanassiev, V.; Hanemann, V.; Wolfl, S. Nucleic Acids Res. 2000,
28, E66.
(48) Chan, T. R.; Hilgraf, R.; Sharpless, K. B.; Fokin, V. V. Org. Lett.
2004, 6, 2853–5.
(34) Jefferson, E. A.; Seth, P. P.; Robinson, D. E.; Winter, D. K.; Miyaji,
A.; Osgood, S. A.; Swayze, E. E.; Risen, L. M. Bioorg. Med. Chem.
Lett. 2004, 14, 5139–43.
9
11186 J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008

Specific Aminoglycoside-RNA Internal Loop Partners
A R T I C L E S
Figure 1. Secondary structures of the internal loop library (1, 4096 members) and competitor oligonucleotides 2-5.35,36 The competitor oligonucleotides
were chosen to ensure RNA-ligand interactions occur with the random region in 1 and are RNA-specific. 6 is the empty cassette into which the random
region was inserted. It serves as a control in fluorescence-based binding assays. The dashed line in 6 indicates where the random region was inserted to
create 1.
General Nucleic Acids. All DNA oligonucleotides were pur-
chased from Integrated DNA Technologies Inc. (IDT, Coralville,
IA) and used without purification unless noted otherwise. The RNA
competitor oligonucleotides were purchased from Dharmacon
(Lafayette, CO) and deprotected according to the manufacturer’s
standard procedure. All aqueous solutions were made with nanopure
water.
protocol and 4 µL of [R-32P]ATP (3000 Ci/mmol; PerkinElmer,
Waltham, MA) in the transcription reaction.
RNA Selection. Internally labeled internal loop library (1, 12
pmol) and competitor oligonucleotides (2-5, 20 nmol each) (Figure
1) were annealed separately in 1× hybridization buffer [HB1; 20
mM N-(2-hydroxyethyl)piperazine-N′-2-ethanesulfonic acid (Hepes),
pH 7.5, 140 mM NaCl, and 5 mM KCl] at 60 °C for 5 min and
allowed to slow-cool on the benchtop. MgCl2 was then added to a
final concentration of 1 mM. The annealed RNAs were mixed
together for a total volume of 400 µL. Azido-aminoglycoside
microarrays were pre-equilibrated with 1× hybridization buffer
supplemented with 1 mM MgCl2 and 40 µg/mL bovine serum
albumin (BSA) (HB2) for 5 min at room temperature to prevent
nonspecific binding. After the slides were pre-equilibrated, the
annealed RNAs were pipetted onto the slide and evenly distributed
across the slide surface with a custom-cut piece of Parafilm. The
slides were hybridized at room temperature for 30 min, at which
point the RNA solution was removed. The slides were washed by
submersion in 30 mL of HB2 for 30 min with gentle agitation.
This step was repeated. Excess buffer was removed from the slides,
and the slides were dried.
RNA Library and Competitor Oligonucleotides. The RNA
library (1) was designed to display a 3 × 3 nucleotide internal loop
pattern35,36 and was embedded in a hairpin cassette (6, Figure 1).49
Library 1 was synthesized by in vitro transcription from the
corresponding DNA template, which was custom-mixed at the
randomized positions to ensure equivalent representation of all four
nucleotides after deprotection. Competitor oligonucleotides were
used to ensure that RNA-ligand interactions were confined to
the random region. The two oligonucleotides that form 2 mimic
the stem; the sequence was changed so that it does not bind to
reverse transcription-polymerase chain reaction (RT-PCR) primers
but maintains the nearest neighbors (Figure 1). Competitor 3 is a
mimic of the hairpin loop displayed in 1, while DNA oligonucle-
otides 4 and 5 ensure that interactions are RNA-specific.
RNA Transcription and Purification. RNA oligonucleotides
were transcribed by use of an RNAMaxx transcription kit (Strat-
agene) with 5 µL of the amplified DNA from the PCR reaction
described below or 1 pmol of DNA template purchased from IDT
according to the manufacturer’s protocol. After transcription, 0.5
unit of RQ DNase I (Promega) was added, and the sample was
incubated for an additional 30 min at 37 °C.
The arrays were exposed to a phosphorimager screen and imaged
using a Bio-Rad FX phosphorimager. The image was used as a
template to mechanically remove selected RNAs from the surface.
A 200 nL aliquot of HB1 was added to the spot to be removed.
After 30 s, excess buffer (the majority of buffer was reabsorbed
by the surface) was pipetted from the surface and the gel at that
position on the surface was excised.
The transcribed RNAs were then purified by gel electrophoresis
on a denaturing 15% or 17% polyacrylamide gel. The RNAs were
visualized by UV shadowing and extracted into 300 mM NaCl by
tumbling overnight at 4 °C. The resulting solution was concentrated
with 2-butanol and ethanol-precipitated. The pellet was resuspended
in 200 µL of diethyl pyrocarbonate- (DEPC-) treated water, and
the concentrations were determined by their absorbances at 260
nm and the corresponding extinction coefficients. Extinction
coefficients were determined by use of HyTher version 1.0 (Nicolas
Peyret and John SantaLucia, Jr., Wayne State University).50,51 These
parameters were calculated from information on the extinction
coefficients of nearest neighbors in RNA.52
Reverse Transcription-Polymerase Chain Reaction Amplifica-
tion. The agarose containing bound RNAs that was removed from
the slide surface was placed into a thin-walled PCR tube with 18
µL of H2O, 2 µL of 10× RQ DNase I buffer, and 2 units of RQ
DNase I (Promega, Madison, WI). The tube was vortexed and
centrifuged for 4 min at 8000g. The tube was then incubated at 37
°C for 2 h. The reaction was quenched by addition of 2 µL of 10×
DNase stop solution (Promega, Madison, WI), and the sample was
incubated at 65 °C for 10 min to completely inactivate the DNase.
Aliquots of this solution were used for RT-PCR amplification.
Reverse transcription reactions were completed in 1× RT buffer
(supplied by the manufacturer), 1 mM dNTPs, 5 µM RT primer
(5′-CCTTGCGGATCCAAT), 200 µg/mL BSA, 3.5 units of reverse
transcriptase (Life Sciences, Inc., St. Petersburg, FL), and 20 µL
of the selected RNAs treated with DNase I and incubated at 60 °C
for 1 h. To 20 µL of the RT reaction were added 6 µL of 10×
PCR buffer (1× PCR buffer is 10 mM Tris, pH 9, 50 mM KCl,
and 0.1% Triton X-100), 4 µL of 100 µM PCR primer (5′-
G G C C G G A T C C T A A T A C G A C T C A C T A T A G G -
GAGAGGGTTTAAT), 2 µL of 100 µM RT primer, 0.6 µL of 250
Internally labeled internal loop library was synthesized by
including half the amount of cold ATP per the manufacturer’s
(49) Bevilacqua, J. M.; Bevilacqua, P. C. Biochemistry 1998, 37, 15877–
84.
(50) Peyret, N.; Seneviratne, P. A.; Allawi, H. T.; SantaLucia, J., Jr
Biochemistry 1999, 38, 3468–77.
(51) SantaLucia, J., Jr Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 1460–5.
(52) Puglisi, J. D.; Tinoco, I., Jr Methods Enzymol. 1989, 180, 304–25.
9
J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008 11187

A R T I C L E S
Disney et al.
Figure 2. (A, top), Chemical structures of the azido-aminoglycosides used in this study: 7 (kanamycin A derivative), 8 (tobramycin derivative), 9 (neamine
derivative), and 10 (neomycin B derivative). (A, bottom) Immobilization of 7-1035,37 onto alkyne-displaying agarose microarrays for 2DCS or conjugation
to fluorescein (green ball) via a Huisgen dipolar cycloaddition reaction to study binding affinities; AmG refers to aminoglycoside. (B) Image of a microarray
composed of 7-10 after hybridization with the oligonucleotides shown in Figure 1 (top) and after removal of bound RNA (bottom). Circles indicate positions
where RNA was harvested. (C) Plot of data for binding of 32P-internally labeled 1 by array-immobilized 7-10 in the presence of competitor oligonucleotides
2-5. The signals in the plot were normalized relative to the highest signal obtained for binding 10.35 Panel B has been modified from ref 35.
mM MgCl2, and 0.1 µL of Taq DNA polymerase. PCR cycling
conditions (2-steps) were 95 °C for 1 min and 72 °C for 1 min.49
Aliquots of the RT-PCR were checked every five cycles starting at
cycle 10 on a denaturing 17% polyacrylamide gel stained with
ethidium bromide or SYBR Gold (Invitrogen, Carlsbad, CA).
Cloning and Sequencing. The RT-PCR products were cloned
into pGEM T Vector (Promega, Madison, WI) according to the
manufacturer’s protocol with 2 µL of the RT-PCR product. The
ligation mixture was incubated overnight at room temperature, and
then a 5 µL aliquot was transformed into DH5-R Escherichia coli.
White colonies were used to inoculate 1 mL of TB medium
containing 50 mg/L ampicillin in a well of a square, deep-well 96-
well plate. The plate was shaken for 24-48 h at 37 °C until the
cultures reached OD600 > 4. The cultures were pelleted and sent to
Functional Biosciences Inc. (Madison, WI) for sequencing.
PCR Amplification of DNA Templates Encoding Selected
RNAs. The DNA templates encoding the selected RNAs were PCR-
amplified from the harvested plasmid DNA (Eppendorf Fast Plasmid
Mini kit) in 50 µL of 1× PCR buffer, 4.25 mM MgCl2, 0.33 mM
dNTPs, 2 µM each primer (RT and PCR primers), and 0.1 µL of
Taq DNA polymerase. The DNA was amplified by 25 cycles of
95 °C for 30 s, 50 °C for 30 s, and 72 °C for 1 min. All PCR
reactions were checked by gel electrophoresis on a 5% agarose gel
stained with ethidium bromide.
labeled azido-aminoglycoside. The solutions were incubated for 30
min at room temperature and then transferred to a well of a black
96-well plate. Fluorescence intensity was measured on a Bio-Tek
FLX-800 plate reader. The change in fluorescence intensity as a
function of RNA concentration was fit to55
I ) I0 + 0.5∆ε([FL]0 + [RNA]0 + Kt) -
{([FL]0 + [RNA] + Kt)2 - 4[FL]0[RNA]0)0.5
}
where I is the observed fluorescence intensity, I0 is the fluorescence
intensity in the absence of RNA, ∆ε is the difference between the
fluorescence intensity in the absence of RNA and in the presence
of infinite RNA concentration, [FL]0 is the concentration of
the fluorescently labeled azido-aminoglycoside, [RNA]0 is the
concentration of the selected internal loop or control RNA, and Kt
is the dissociation constant. Control experiments were also com-
pleted in the same manner for FITC-triazole,35 which contains the
dye and triazole linkage but no aminoglycoside.56 No change in
fluorescence is observed up to 5 µM 1 (entire internal loop library)
or 3 µM Neo IL15, which was tested as a representative individual
of the selected loops (Figures 1 and 5, respectively). These results
show that the change in fluorescence is due to binding of the
aminoglycoside to the oligonucleotides and not the dye itself.
RNA Secondary Structure Prediction and Identification of
Trends in Selected RNAs. The secondary structures of all selected
RNAs were predicted by free energy minimization using the
RNAstructure program.53,54 The structures were then inspected for
commonalities. To determine if the commonalities and trends were
statistically significant, the percentage of RNAs displaying the trend
in the selected RNAs was compared to the percentage of the RNAs
that display that trend in the entire library. A Z-test was applied
and the corresponding p-value was calculated. Please see Supporting
Information for all calculations.
Results and Discussion
Our library of azido-aminoglycosides was chosen for these
studies because their parent aminoglycosides have been shown
to bind bacterial rRNA A-sites.57–63 The azido-aminoglycoside
library (7-10, Figure 2A) was site-specifically immobilized onto
alkyne-functionalized agarose arrays via a Huisgen 1,3 dipolar
cycloaddition reaction to create chemical microarrays (Figure
Fluorescence Binding Assays. Dissociation constants were
determined by an in-solution, fluorescence-based assay. A selected
RNA or RNA mixture was annealed in HB1 and 40 µg/mL BSA
at 60 °C for 5 min and was allowed to slow-cool to room
temperature. Then MgCl2 and fluorescently labeled aminoglycoside
(7-Fl, 8-Fl, 9-Fl, or 10-Fl, Figure 2A) were added to final
concentrations of 1 mM and 10 nM, respectively. Serial dilutions
(1:2) were then completed in 1× HB2 + 10 nM fluorescently
(55) Wang, Y.; Rando, R. R. Chem. Biol. 1995, 2, 281–90.
(56) For details see Supporting Information.
(57) Moazed, D.; Noller, H. F. Nature 1987, 327, 389–94.
(58) Fourmy, D.; Recht, M. I.; Blanchard, S. C.; Puglisi, J. D. Science
1996, 274, 1367–71.
(59) Recht, M. I.; Fourmy, D.; Blanchard, S. C.; Dahlquist, K. D.; Puglisi,
J. D. J. Mol. Biol. 1996, 262, 421–36.
(60) Lynch, S. R.; Gonzalez, R. L.; Puglisi, J. D. Structure (Cambridge,
MA, U.S.) 2003, 11, 43–53.
(61) Wong, C. H.; Hendrix, M.; Priestley, E. S.; Greenberg, W. A. Chem.
Biol. 1998, 5, 397–406.
(53) Mathews, D. H.; Disney, M. D.; Childs, J. L.; Schroeder, S. J.; Zuker,
M.; Turner, D. H. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 7287–92.
(54) Mathews, D. H.; Sabina, J.; Zuker, M.; Turner, D. H. J. Mol. Biol.
1999, 288, 911–40.
(62) Ryu, D. H.; Rando, R. R. Bioorg. Med. Chem. 2001, 9, 2601–8.
(63) Griffey, R. H.; Hofstadler, S. A.; Sannes-Lowery, K. A.; Ecker, D. J.;
Crooke, S. T. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 10129–33.
9
11188 J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008

Specific Aminoglycoside-RNA Internal Loop Partners
A R T I C L E S
Figure 3. Secondary structures of the RNA internal loops that were selected to bind 7 (A, kanamycin A derivative) and 8 (B, tobramycin derivative) and
their corresponding dissociation constants (nanomolar) (Figure 1). The nucleotides shown are derived from the boxed region in 1. Analysis of these data
shows that 7 prefers internal loops that contain potential pyrimidine-pyrimidine pairs (in green) and that 8 prefers internal loops that contain potential
guanine-guanine pairs (in green). Tob IL3 may have a structure in equilibrium in which there is a 3 × 3 internal loop instead of the two bulges. Secondary
structure studies with RNAstructure53 predict that the 3 × 3 nucleotide loop is 3.2 kcal/mol less stable than the structure containing two 1-nucleotide bulges.
2A).35–37,48,64 The position where the azido group was installed
in each library member was governed by the ease of function-
alizing a primary hydroxyl group.35,37,65,66 In addition, crystal
structures of kanamycin A (7-like) and tobramycin (8-like)
complexed with an oligonucleotide mimic of the bacterial rRNA
A-site show that no contacts are formed between the A-site and
either aminoglycoside’s 6′-OH group.67,68 Therefore, 7 and 8
would be immobilized in a manner that mimics their biological
presentations, at least for their binding to the bacterial rRNA
A-site. Structures 9 (neamine-like) and 10 (neomycin B-like)
were functionalized at the chosen positions (5- and 5′′-OH,
respectively) because both present rings I and II in similar
manners, potentially allowing for RNA motifs that interact with
these rings to be identified via analysis of the selection data
from the two compounds. In contrast to the sites of function-
alization with 7 and 8, interactions are observed between the
5- and the 5′′-OH groups in neamine and neomycin B,
respectively, and an oligonucleotide mimic of the bacterial rRNA
A-site.67
loops can form two 1-nucleotide bulges; see Tob IL3 in Figure
3), and 216 unique base-paired regions. One potential problem
with using 1 to identify RNA loop-ligand interactions is that
interactions can occur outside of the random region. This would
cause selection of all library members. To deal with this issue,
competitor oligonucleotides (Figure 1) were used that mimic
the stem (2) and the hairpin (3) in 1 to compete off these
interactions. DNA oligonucleotides 4 and 5 were also added as
competitors to ensure that RNA-specific interactions were
selected. Array hybridization (Figure 2B) was completed with
20 nmol of each competitor oligonucleotide and 12 pmol of 1;
the large excess of competitors was used to ensure that it is in
excess of both the loading of 7-10 onto the surface and 1.
Hybridization of the arrays under the conditions above
showed a clear dose response and the aminoglycosides bound
different amounts of 1 (Figure 2B,C). Plots of the normalized
signal for binding of 1 showed, not surprisingly, that the signal
scaled with the number of amino groups present on the
aminoglycosides (Figure 2C). Bound RNAs were harvested from
the arrays via excision of the agarose at the positions indicated
in Figure 2B. Excision of the bound RNAs is possible because
of the unique features of the agarose microarray: ligands can
be site-specifically immobilized onto the surface for screening,
and rehydration of the surface allows for the positions where
RNA is bound to ligand to be harvested via simple gel
extraction.35 These bound RNAs can be amplified by RT-PCR,
cloned, and sequenced to identify the binders. The lowest signal
(or ligand loading) spots that were able to be amplified over
background were harvested because RNAs that bind at lower
ligand loading are higher affinity.35 The mixtures of selected
RNA motifs were then amplified via RT-PCR, transcribed into
RNA, and studied for binding each aminoglycoside via a
Aminoglycoside arrays were probed for binding to a 32P-
internally labeled RNA library that displays a 3 × 3 nucleotide
internal loop pattern (1, Figures 1 and 2B).35,36 The random
region was embedded in oligonucleotide 6, and the RNA motifs
displayed by 1 include 1080 1 × 1 internal loops, 1200 2 × 2
internal loops, 1600 3 × 3 internal loops (some 3 × 3 internal
(64) Kolb, H. C.; Finn, M. G.; Sharpless, K. B. Angew. Chem., Int. Ed.
2001, 40, 2004–2021.
(65) Michael, K.; Wang, H.; Tor, Y. Bioorg. Med. Chem. 1999, 7, 1361–
71.
(66) Wang, H.; Tor, Y. Angew. Chem., Int. Ed. 1998, 37, 109–11.
(67) Francois, B.; Russell, R. J.; Murray, J. B.; Aboul-ela, F.; Masquida,
B.; Vicens, Q.; Westhof, E. Nucleic Acids Res. 2005, 33, 5677–90.
(68) Vicens, Q.; Westhof, E. Chem. Biol. 2002, 9, 747–55.
9
J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008 11189

A R T I C L E S
Disney et al.
Table 1. Binding of 1, 6, and Libraries of Selected RNAs to Different Aminoglycosidesa
aminoglycoside from which bound RNAs were harvestedb
other RNAs
aminoglycoside
7
8
9
10
1
6
A-sitec
7
8
9
10
75 ( 15; -
850 ( 105; 11
900 ( 03; 12
625 ( 82; 8
>1500; >30
50 ( 12; -
>1500; >30
700 ( 84; 14
>1500; >7
1100 ( 113; 6
800 ( 95; 5
1100 ( 127; 6
160 ( 20; -
>1200
>5000
18000
1500
7800
19
>1000; >11
200 ( 33; -
500 ( 53; 2.5
>1200
∼2500
>2500
1200 ( 112
>5000
1600 ( 212
a All dissociation constants are reported in nanomolar. Selectivies (given after the semicolon) were calculated by dividing the Kd for the other
aminoglycoside by the Kd for the aminoglycoside that the loop was selected to bind. b These are the pools of RNA structures that were selected to bind
each aminoglycoside. c Binding affinities for an oligonucleotide mimic of the bacterial rRNA A-site were measured using surface plasmon resonance
(SPR) with the RNA immobilized onto the SPR chip.61
fluorescence-based emission assay with fluorescein-labeled
7-10, denoted 7-Fl, 8-Fl, 9-Fl, and 10-Fl (Figure 2A).35
Specificity in the RNA Motif-Aminoglycoside Partners
Identified by 2DCS. Table 1 summarizes the dissociation
constants determined for all selected RNA mixtures, the entire
RNA library (1, Figure 1), and the empty hairpin cassette (6,
Figure 1) for all aminoglycoside derivatives. The mixtures of
selected RNAs bind to their respective ligand with Kd values
from 75 to 200 nM. In comparison, binding to the whole library,
1, was g1200 nM and binding to the empty cassette, 6, was
g1600 nM for all aminoglycosides. These experiments were
completed at pH 7.5 in a buffer containing 1 mM MgCl2, while
those completed in our previous report were completed at pH
7.0 without MgCl2.35 Higher pH, the presence of divalent metal
ions, and an increase in ionic strength all decrease aminogly-
coside binding affinity and help explain the differences in the
dissociation constants between the two studies.35,69–71 The new
conditions were used because they more closely mimic physi-
ological conditions than the previous buffer. Not surprisingly,
the neomycin B mimetic (10) with the largest number of amines,
exhibits the highest affinities toward 1 and 6, suggesting it has
a significant nonspecific binding mode. This has also been
observed in studies of aminoglycosides binding to mimics of
the bacterial rRNA A-site.61 Additionally, competition dialysis
studies with a series of nucleic acids showed neomycin binds
helical A-form nucleic acids that are present in the nonvariable
region of 1.72
The binding of the selected RNA mixtures to the other arrayed
aminoglycosides was studied to determine whether the selected
structures are specific for the aminoglycoside that they were
selected to bind. Interestingly, the RNAs selected to bind 7 and
8 are >8-fold and g14-fold specific, respectively, for their
corresponding aminoglycosides. Lower specificities are observed
between structures selected to bind 9 (neamine-like) and 10
(neomycin B-like); in the worst case, there was only a 2.5-fold
specificity between the RNAs selected to bind 9 for binding
10. This is not surprising since they have rings I and II in
common (Figure 2A), making it likely that these two structures
bind overlapping RNA motif space.
Figure 4. Secondary structures of the RNA internal loops that were selected
to bind 9 (neamine derivative) and their corresponding dissociation constants
(nanomolar). The nucleotides shown are derived from the boxed region in
1 (Figure 1). Analysis of these data shows preferences for internal loops
that contain potential guanine-adenine pairs (in green).
computed for groups of the selected structures to find RNA
structure trends in the data. Trends in RNA structure space
were observed for aminoglycosides 7 (Figure 3A), 8 (Figure
3B), and 10 (Figure 5). Compound 7 prefers pyrimidine-rich
internal loops (Figure 3A), in particular internal loops
displaying potential UU pairs (two-tailed p-value ) 0.0373),
while 8 prefers internal loops that display potential GG pairs
(Figure 3B, two-tailed p-value ) 0.0109). Compound 10
prefers internal loops that display potential GA pairs (Figure
5, two-tailed p value ) 0.0019), which is also the most highly
represented type of internal loop for 9 (Figure 4). This could
explain, at least in part, the smaller specificity between
structures selected to bind 9 and 10 (Table 1). Further
inspection of the GA-containing internal loops that bind 9
show that 66% of the other nucleotides in the internal loop
(i.e., non-GA) that are un- or noncanonically paired are
pyrimidines; this trend is not observed for the internal loops
that bind 10.
Dissociation constants were determined for each statistically
significant RNA motif-ligand pair. Figures 3–5 show the
secondary structures of selected internal loops and their dis-
sociation constants. The ranges of dissociation constants are
consistent with the dissociation constants for the RNA mixtures
(Table 1).
In general 1 × 1, 2 × 2, and 3 × 3 internal loops are well
represented for each of the ligands (Figures 3–5). Further
analysis of the global structures shows that some general features
Studies on the Specific RNA Motifs Selected To Bind
7-10. To deconvolute the specific RNA structures that were
selected, the mixtures of RNAs selected to bind 7-10 were
cloned and sequenced and their secondary structures were
modeled with the RNAstructure program (Figures 3–5).53,54
The data were then statistically analyzed and p-values were
(69) Thomas, J. R.; Liu, X.; Hergenrother, P. J. Biochemistry 2006, 45,
10928–38.
(70) Tor, Y.; Hermann, T.; Westhof, E. Chem. Biol. 1998, 5, R277–83.
(71) Hermann, T.; Westhof, E. J. Mol. Biol. 1998, 276, 903–12.
(72) Arya, D. P.; Xue, L.; Willis, B. J. Am. Chem. Soc. 2003, 125, 10148–
9.
9
11190 J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008

Specific Aminoglycoside-RNA Internal Loop Partners
A R T I C L E S
Figure 5. Secondary structures of the RNA internal loops that were selected to bind 10 (neomycin B derivative) and their corresponding dissociation
constants (nanomolar). The nucleotides shown are derived from the boxed region in 1 (Figure 1). Analysis of these data shows preferences for internal loops
that contain potential guanine-adenine pairs (in green).
can be found in the 1 × 1 internal loops (Figures 3–5). For all
of the selected 1 × 1 internal loops, there is an underrepresen-
tation of internal loops with GC closing pairs; of the eight 1 ×
1 internal loops selected, only one loop has a single closing
GC pair (Neo IL1, binds 10) and none are closed with two GC
pairs. In addition, all of the internal loops but one (Nea IL2,
binds 9) have at least one GU closing pair. Similar results have
been found with internal loops selected to bind 6′-N-5-hexynoate
kanamycin A.35 These collective results suggest the thermody-
namic stability of the 1 × 1 loops is an important determinant
in molecular recognition because there is a 1.2 kcal/mol penalty
for closing a loop with an AU or GU pair instead of a GC pair.73
Perhaps in the presence of ligand these pairs partially form or
do not form at all. This may indicate an induced-fit type of
binding; conformational changes in RNA often occur upon
binding to ligand.74–77 Structural studies must be completed to
test this hypothesis, however.
Comparison of the groups of RNA motifs selected to bind
each aminoglycoside show that there are similar features in each
group that go beyond the presence of a potential loop non-
canonical pair. Many loops share common closing pairs or share
some other loop feature. For example, with compound 7 (Figure
3A), Kan A IL1 and Kan A IL8 each have a 1 × 1 CC internal
loop closed by a GU and an AU pair but are different in their
orientations. Loop Kan A IL2 appeared twice in the sequencing
data and is also an all-pyrimidine (UU) 1 × 1 internal loop
that is also closed by a GU and an AU pair. For compound 8
(Figure 3B), five of the eight loops that have potential GG pairs
were 2 × 2 internal loops (Tob IL4-8). Both Tob IL4 and Tob
IL7 have a 5′-CG-3′/3′-CG-5′ internal loop but with different
closing pairs. The Tob IL5 and Tob IL8 internal loops are
similar, 5′-GA-3′/3′-GC-5′ and 5′-AG-3′/3′-CG-5′, respectively,
and both are closed with GC and AU pairs that are oriented in
the same manner relative to the loop nucleotides. For compound
9, two 1 × 1 GA internal loops (Nea IL1 and Nea IL2) were
identified that are closed by AU pairs in different orientations.
The four 2 × 2 internal loops (Nea IL3, IL5, IL6, and IL7)
each have either a pyrimidine across from a pyrimidine (Nea
IL3 and IL6) or a purine across from a purine (Nea IL5 and
IL7) adjacent to the potential GA pair. Four of the 3 × 3 internal
loops identified for binding 9 have an adenine across from
cytosine in the internal loop (Nea IL9-12). Of these four, three
(Nea IL9, Nea IL10, and Nea IL12) have the adenine across
from the cytosine in a position that is adjacent to the internal
loop-closing AU pair, and there are also nucleotides that can
form pairs (GU or AU) in the internal loop’s central position
(Figure 4). These results may indicate the central position is
paired (GU or AU) in these internal loops. For loops that bind
10 (Figure 5), each of the 1 × 1 GA internal loops has at least
one closing GU pair (Neo IL1, Neo IL2, and Neo IL13). For
the 2 × 2 internal loops, Neo IL3 and IL6 have the same loop
composition with an adenine across from a cytosine in addition
to an adenine across from a guanine, but the orientation of the
internal loop and the loop closing pairs are different. For the
eight 3 × 3 internal loops selected to bind 10, four have
nucleotides that can form canonical pairs in the loop’s central
position (Neo IL4, Neo IL11, Neo IL12, and Neo IL16), which
is similar to observations with 9.
Once the specific RNA motifs interacting with ligands were
identified, studies on the specificity of a selected RNA for the
corresponding aminoglycoside were revisited. Dissociation
constants were measured for 16 selected RNAs for binding to
each of 7-10-Fl (Table 2). Specificities range from 2- to 80-
fold depending on the aminoglycoside and the internal loop
studied. The average specificity for these subsets of internal
(73) Davis, A. R.; Znosko, B. M. Biochemistry 2007, 46, 13425–36.
(74) Hermann, T.; Patel, D. J. Science 2000, 287, 820–5.
(75) Gosser, Y.; Hermann, T.; Majumdar, A.; Hu, W.; Frederick, R.; Jiang,
F.; Xu, W.; Patel, D. J. Nat. Struct. Biol. 2001, 8, 146–50.
(76) Puglisi, J. D.; Tan, R.; Calnan, B. J.; Frankel, A. D.; Williamson,
J. R. Science 1992, 257, 76–80.
(77) Fourmy, D.; Yoshizawa, S.; Puglisi, J. D. J. Mol. Biol. 1998, 277,
333–45.
9
J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008 11191

A R T I C L E S
Disney et al.
Table 2. Binding and Selectivities of Single Internal Loops
Selected from 7-10
One question that arises is “Why does 2DCS allow for the
selection of specific aminoglycoside-RNA internal loop inter-
actions?” A likely possibility is the nature of microarray-based
screening that sets up a competition experiment on the surface
of an array between the four arrayed ligands and members of
1. Such a competition is possible because 1 is being exposed
to each of the four aminoglycosides simultaneously during
hybridization and the amount of aminoglycoside delivered
to an array is in excess (4.6 nmol)37,78 over the amount of 1
hybridized (12 pmol), 380-fold excess total aminoglycoside
over 1.
dissociation constants (nM); selectivity for binding 7-10-Fla
internal loop
7-Fl
8-Fl
9-Fl
10-Fl
Selected from 7
778 ( 60; 7
b
b
>1200 ; 12c >1500 ; 12c
Kan A IL2 100 ( 15
Kan A IL4 110 ( 20
Kan A IL6 150 ( 20
Kan A IL8 64 ( 7
d
>2000 ; 18c
780 ( 48; 7
300 ( 26; 2
1100 ( 192; 10
d
>2250 ; 15c
580 ( 50; 4
685 ( 84; 11
b
d
>2250 ; 35c
>2250 ; 35c
Selected from 8
42 ( 12
37 ( 9
b
>1200 ; 29c
Tob IL1
655 ( 18; 16 342 ( 20; 8
d
>2250 ; 61c
Tob IL2
Tob IL7
Tob IL8
690 ( 22; 19
1600 ( 28; 43
d
>2250 ; 45c
1400 ( 280; 28 50 ( 4
930 ( 42; 19
660 ( 63; 9
Results for Inserting Internal Loops into Other Cassettes:
Are Cassette Nucleotides Important for Binding? To determine
if specific RNA internal loop-ligand partners were selected,
we changed the cassette in which the internal loops were
displayed (11, Figure 6) by changing the AU and GU pairs in
6 to GC pairs. This was done because previous experiments
have shown that that 2-deoxystreptamine (2-DOS) binds weakly
to GU pairs and GU steps in RNA,79 which are present in 6
(Figure 1). The simplest statistically significant internal loops
(1 × 1 nucleotide) and their nearest neighbors were inserted
into 11. The nearest neighbors were included because previous
experiments have shown that internal loop nearest neighbors
can affect internal loop structure80,81 and because of the
thermodynamic arguments for loop closing pairs raised above.
Results show that dissociation constants for the four internal
loops tested are similar when they are displayed in either 6 or
11 (Figure 6). As also demonstrated for a selection to identify
RNA internal loop preferences for a 6′-N-derivatized kanamycin
A,35 these results show that cassette nucleotides do not
contribute significantly to aminoglycoside affinity and that the
selected RNA motif-ligand pairs are portable.
Comparison to Other Studies on RNAs That Bind Aminogly-
cosides. We have probed chemical and RNA spaces simulta-
neously in order to determine the RNA sequence or structure
that is preferred by four related aminoglycoside derivatives
(Figure 7). The kanamycin A derivative, 7, shows a preference
for pyrimidines, in particular loops containing potential UU
pairs. Interestingly, 6′-N-5-hexynoate kanamycin was previously
found to prefer RNA internal loops that can form potential AC35
and UU pairs.36 This suggests that these two related structures
(7 and 6′-N-5-hexynoate kanamycin A) can bind to similar RNA
motifs. The bacterial rRNA A-site also contains a UU pair, and
a crystal structure of the bacterial rRNA A-site complexed with
several aminoglycosides has been reported.67 Direct contacts
are made between kanamycin and the UU pair in the A-site,
though this contact also occurs with tobramycin. Comparison
of the contacts between kanamycin A and tobramycin and the
bacterial rRNA A-site show that 21 and 20 contacts are formed,
respectively.67,68 Of these contacts, only eight are present in
both complexes. This indicates that the two related aminogly-
cosides have somewhat different structures in their bound state
with the A-site, suggesting that they may also bind different
RNA space. Indeed we have found that tobramycin-like 8 binds
d
>2250 ; 32c
560 ( 62; 8
70 ( 5
Selected from 9
960 ( 70; 4
840 ( 28; 12
640 ( 20; 2
Nea IL2
Nea IL3
Nea IL6
490 ( 15; 2
230 ( 40
70 ( 10
850 ( 77; 2
780 ( 58; 11
542 ( 35; 8
d
>2250 ; 7c
307 ( 31
1435 ( 15; 5
1000 ( 144; 12
Nea IL10 860 ( 4; 10
1050 ( 120; 13 84 ( 8
Selected from 10
d
d
>2250 ; 80c
860 ( 50; 11
920 ( 85; 16
>2250 ; 80c
Neo IL6
Neo IL9
Neo IL13 460 ( 47; 8
620 ( 75; 22
28 ( 3
76 ( 14
58 ( 10
49 ( 9
d
>2250 ; 30c
580 ( 4; 8
>2250 ; 39c
d
d
d
>2250 ; 46c
>2250 ; 46c
Neo IL15 1150 ( 130; 23
a Selectivities (given after the semicolon) were calculated by dividing
the Kd for the other aminoglycoside by the Kd for the aminoglycoside
that the loop was selected to bind. b These values are a lower limit of
the dissociation constants because the titration curves did not reach
saturation at these concentrations. c These values represent a lower limit
of specificity. d In these binding experiments there was no change in
fluorescence observed.
loops is 14-, 26-, 7-, and 34-fold for 7, 8, 9, and 10, respectively.
The highest specificity was measured for Neo IL6 that was
selected for binding to 10, which binds 22-, 80-, and 80-fold
more weakly to 7-Fl, 8-Fl, and 9-Fl, respectively. Other high
specificity RNA motif-ligand pairs are 7 and Kan A IL8 with
an average specificity of 27-fold, 8 and Tob IL2 with an average
specificity of 41-fold, and 9 and Nea IL10 with an average
specificity of 12-fold. The lower specificity for 9 is also mirrored
in the specificity results observed for binding to the entire
mixture of selected structures (Table 1). Interestingly, the 1 ×
1 GA internal loop selected to bind 9 (Nea IL2) has the lowest
selectivity for binding the different aminoglycosides (average
specificity of 2.6-fold) of all internal loops tested for specificity.
This loop and Nea IL1 are both closed with AU pairs. In
contrast, the single GA internal loops (Neo IL1, Neo IL2, and
Neo IL13) selected to bind 10 have at least one loop closing
GU pair. One of these loops (Neo IL13) was tested for
aminoglycoside specificity and binds 39-fold more weakly to
9-Fl than 10-Fl. Evidently, the identity of the loop closing pairs
can affect aminoglycoside selectivity with 1 × 1 internal loops.
A particular challenge in finding ligands that bind RNA has
been to identify specific aminoglycoside-RNA partners. The
observation that mixtures of selected RNA structures and
individual selected RNAs exhibit specificities for the different
aminoglycosides is encouraging in this regard. This is especially
true given the fact that RNAs selected for 7 and 8 bind at least
8-fold more tightly to their respective ligands than to 10, which
has more NH2 groups and binds to the bacterial rRNA A-site
950- and 78-fold more tightly than 7 and 8, respectively (Table
1).61 These results suggest that specific interactions between
the aminoglycosides and RNA motifs identified in these
selections are dependent on the aminoglycoside size and shape
and the features in the RNA, not simply charge-charge
interactions.
(78) Using a fluorescamine-based assay described in ref 37, it was
determined for the highest concentration of aminoglycoside spotting
solution that 48%, 62%, 79%, and 36% of the aminoglycoside
delivered to the array surface is immobilized for 7, 8, 9, and 10,
respectively.
(79) Yoshizawa, S.; Fourmy, D.; Eason, R. G.; Puglisi, J. D. Biochemistry
2002, 41, 6263–70.
(80) Wu, M.; Turner, D. H. Biochemistry 1996, 35, 9677–89.
(81) SantaLucia, J., Jr.; Turner, D. H. Biochemistry 1993, 32, 12612–23.
9
11192 J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008

Specific Aminoglycoside-RNA Internal Loop Partners
A R T I C L E S
Figure 6. Secondary structures of the simplest loops selected to bind each aminoglycoside derivative and the alternative cassette (11) in which they were
embedded. All GU and AU pairs were changed to GC pairs in 11 to determine whether cassette nucleotides contribute to binding affinity. Dissociation
constants for the loops are below the secondary structures and are reported in nanomolar. Dissociation constants for the loops in the original cassette (6,
Figure 1) are in parentheses. The GC cassette alone (11) binds weakly to 7-10-Fl with dissociation constants g2 µM. In each case the binding constants
of the ligands are similar when the internal loops are displayed in the different cassettes, showing that interactions to cassette nucleotides are not important
for binding.
aptamer since it was selected by immobilization through the
6′-amine. Our tobramycin derivative was selectively immobi-
lized through the 6′′-amine. In the NMR structure of the
tobramycin aptamer, the 6′′-amine is buried within the RNA87
and it is therefore unlikely that 8 could bind in a similar fashion.
Compound 10 prefers loops with potential GA pairs. A
previously reported neomycin aptamer also contains a GA pair,
90
which sits on top of neomycin. This aptamer was selected
by random attachment to a resin via either the 6′- or 6′′-amine.
On the basis of the orientation of neomycin in the solution
structure, the authors postulated that the NMR structure is
representative of the RNAs that bind when neomycin is
immobilized through the 6′′-amine since it is not buried within
the RNA, while the 6′-amine is.90 Our 10 neomcyin derivative
was selectively immobilized through ring III, which is also not
buried within the aptamer.
Compound 9 prefers loops with potential GA pairs over other
types of mismatches (40% of the structures contain potential
GA pairs), though it is not statistically significant. Because
neamine is smaller than most aminoglycosides and has the
largest cationic charge density, it likely can accommodate many
more RNA folds. We are currently employing multiple rounds
of selection in order to better define the RNA motif preferences
for neamine.
Figure 7. Schematic representation of the 2DCS data reported in this study.
By probing both RNA and chemical spaces simultaneously, two aminogly-
coside derivatives (tobramycin and kanamycin A) prefer to bind different
RNA spaces while two other aminoglycoside derivatives (neamine and
neomycin B) prefer to bind overlapping RNA space, presumably because
they share the top two rings in their structures.
(82) Blount, K. F.; Zhao, F.; Hermann, T.; Tor, Y. J. Am. Chem. Soc. 2005,
127, 9818–29.
to different RNA structures than kanamycin A-like 7, in support
of this hypothesis.
(83) Zhao, F.; Zhao, Q.; Blount, K. F.; Han, Q.; Tor, Y.; Hermann, T.
Angew. Chem., Int. Ed. 2005, 44, 5329–34.
One method to introduce or affect the specificity of ami-
noglycosides may therefore be to restrict the conformations they
can sample or alter the conformation altogether. Indeed, several
groups have synthesized and studied aminoglycosides that have
been rigidified by introducing additional ring systems.82–86 It
would be interesting to see how constraining the three-
dimensional space that these rigidified aminoglycosides can
sample affects the RNA space the ligands recognize.
Aminoglycoside 8 prefers loops that display potential GG
pairs. Solution structures of tobramycin aptamers show that
tracts of purines are important for selective binding; a guanine
forms a cap for the aminoglycoside.87–89 It is difficult to make
direct comparisons to the previously reported tobramycin
(84) Kling, D.; Hesek, D.; Shi, Q.; Mobashery, S. J. Org. Chem. 2007, 72,
5450–3.
(85) Asensio, J. L.; Hidalgo, A.; Bastida, A.; Torrado, M.; Corzana, F.;
Chiara, J. L.; Garcia-Junceda, E.; Canada, J.; Jimenez-Barbero, J.
J. Am. Chem. Soc. 2005, 127, 8278–9.
(86) Bastida, A.; Hidalgo, A.; Chiara, J. L.; Torrado, M.; Corzana, F.; Perez-
Canadillas, J. M.; Groves, P.; Garcia-Junceda, E.; Gonzalez, C.;
Jimenez-Barbero, J.; Asensio, J. L. J. Am. Chem. Soc. 2006, 128, 100–
16.
(87) Jiang, L.; Patel, D. J. Nat. Struct. Biol. 1998, 5, 769–74.
(88) Cho, J.; Hamasaki, K.; Rando, R. R. Biochemistry 1998, 37, 4985–
92.
(89) Wang, Y.; Killian, J.; Hamasaki, K.; Rando, R. R. Biochemistry 1996,
35, 12338–46.
(90) Jiang, L.; Majumdar, A.; Hu, W.; Jaishree, T. J.; Xu, W.; Patel, D. J.
Structure 1999, 7, 817–27.
9
J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008 11193

A R T I C L E S
Disney et al.
Interestingly, neamine derivatives91 and neomycin92,93 have
been shown to inhibit interactions between the protein Rev and
the RRE RNA in HIV. The binding site of Rev and neomycin
B in RRE have been mapped and show that these ligands interact
at a site that contains a 3 × 2 RNA internal loop that contains
GA and GG pairs.92,93 Although the aminoglycoside binding
site in RRE is an asymmetric internal loop, the similarities
between the symmetric internal loops that we selected are
interesting.
It should be noted that, in contrast to SELEX, RNA sequence
preferences for kanamycin, tobramycin, and neomycin deriva-
tives were identified using only a single round of selection. This
is likely due to reduced size of 1 (4096 members) versus typical
SELEX libraries (1013 to 1015 members)19 and the use of
competitor oligonucleotides (2-5) to select against binding to
common elements in 6 and 1, which also introduced stringency
into the selection. An additional factor may also be the
competition that is occurring for RNA-ligand partners on the
array surface that is mentioned above. This may allow 2DCS
to be of general use for the selection of selective RNA-ligand
interactions.
data described herein with searching tools may elucidate new
RNA targets for aminoglycosides.
It should also be noted that aminoglycoside antibiotics display
less than favorable toxicity profiles.95 One challenge that 2DCS
has not been tested for is identifying RNA motif-ligand partners
with more druglike ligands, especially ones that have an
unknown RNA binding capacity. The ability to probe both RNA
and chemical spaces simultaneously, however, may allow 2DCS
to identify such partners. These interactions could go undetected
if only a single or a few biologically important RNAs were
probed for binding to a chemical library. Although the challenge
of enabling the rational and modular design of small molecules
targeting RNA by use of a database of RNA motif-ligand pairs
is daunting given the diversity of RNA structures present in
nature, we hope that 2DCS will accelerate these developments
by serving as an enabling technology.
Acknowledgment. We thank Professor Yitzak Tor (University
of California, San Diego) for helpful suggestions and Olivia Barrett
for completing the fluorescamine assay to determine aminoglycoside
loading on the microarrays. This work was funded by the University
at Buffalo, The State University of New York, The New York State
Center of Excellence in Bioinformatics and Life Sciences, A
NYSTAR J. D. Watson Award (M.D.D.), The Camille and Henry
Dreyfus Young Investigator Award (M.D.D.), a Cottrell Scholar
Award from the Research Corporation (M.D.D.), and the Kapoor
endowment fund.
Implications. Our results developed a set of consensus RNA
internal loops that prefer derivatives of four aminoglycosides
(Figure 7). In terms of understanding general features in RNA
motifs that bind ligands, it will be interesting to see how the
data described here for a library of symmetric loops correlate
with 2DCS of asymmetric RNA internal loops or hairpin loops.
Bioinformatics tools are already available to find RNA motifs
in genomic sequences (RNAMotif).94 Therefore, coupling the
Supporting Information Available: Details on the synthesis
of the fluorescently labeled azido-aminoglycosides 7-10-Fl,
results from statistical analysis of the selected structures, and
representative fluorescence binding assays. This material is
available free of charge via the Internet at http://pubs.acs.org.
(91) Park, W. K. C.; Auer, M.; Jaksche, H.; Wong, C. H. J. Am. Chem.
Soc. 1996, 118, 10150–5.
(92) Werstuck, G.; Zapp, M. L.; Green, M. R. Chem. Biol. 1996, 3, 129–
37.
JA803234T
(93) Zapp, M. L.; Stern, S.; Green, M. R. Cell 1993, 74, 969–78.
(94) Macke, T. J.; Ecker, D. J.; Gutell, R. R.; Gautheret, D.; Case, D. A.;
Sampath, R. Nucleic Acids Res. 2001, 29, 4724–35.
(95) Silva, J. G.; Carvalho, I. Curr. Med. Chem. 2007, 14, 1101–19.
9
11194 J. AM. CHEM. SOC. VOL. 130, NO. 33, 2008