Conformational constraint as a means for understanding RNA-aminoglycoside specificity

Article abstract of DOI:10.1021/ja050918w

The lack of high RNA target selectivity displayed by aminoglycoside antibiotics results from both their electrostatically driven binding mode and their conformational adaptability. The inherent flexibility around their glycosidic bonds allows them to easi

Full text of DOI:10.1021/ja050918w

Published on Web 06/18/2005
Conformational Constraint as a Means for Understanding
RNA-Aminoglycoside Specificity
Kenneth F. Blount,†,§ Fang Zhao,† Thomas Hermann,‡ and Yitzhak Tor*,†
Contribution from the Department of Chemistry and Biochemistry, UniVersity of California,
San Diego, La Jolla, California 92093-0358 and Anadys Pharmaceuticals, Inc.,
3115 Merryfield Row, San Diego, California 92121
Received February 12, 2005; E-mail: ytor@chem.ucsd.edu
Abstract: The lack of high RNA target selectivity displayed by aminoglycoside antibiotics results from both
their electrostatically driven binding mode and their conformational adaptability. The inherent flexibility around
their glycosidic bonds allows them to easily assume a variety of conformations, permitting them to structurally
adapt to diverse RNA targets. This structural promiscuity results in the formation of aminoglycoside
complexes with diverse RNA targets in which the antibiotics assume distinct conformations. Such differences
suggest that covalently linking individual rings in an aminoglycoside could reduce its available conformations,
thereby altering target selectivity. To explore this possibility, conformationally constrained neomycin and
paromomycin analogues designed to mimic the A-site bound aminoglycoside structure have been
synthesized and their affinities to the TAR and A-site, two therapeutically relevant RNA targets, have been
evaluated. As per design, this constraint has minimal deleterious effect on binding to the A-site. Surprisingly,
however, preorganizing these neomycin-class antibiotics into a TAR-disfavored structure has no deleterious
effect on binding to this HIV-1 RNA sequence. We rationalize these observations by suggesting that the
A-site and HIV TAR possess inherently different selectivities toward aminoglycosides. The inherent plasticity
of the TAR RNA, coupled to the remaining flexibility within the conformationally constrained analogues,
makes this RNA site an accommodating target for such polycationic ligands. In contrast, the deeply
encapsulating A-site is a more discriminating RNA target. These observations suggest that future design
of novel target selective RNA-based therapeutics will have to consider the inherent “structural” selectivity
of the RNA target and not only the selectivity patterns displayed by the low molecular weight ligands.
Introduction
with the observation that they can inhibit the activity of several
ribozymes8 or can bind to two important HIV-1 RNAs, the RRE
and the TAR, thereby preventing the necessary binding of their
cognate protein.9
In the mandated search for new approaches to target
pathogens, viral and bacterial RNA sequences have recently
emerged as intriguing targets.1,2 Key events in bacterial and viral
life cycles that are mediated by RNA or RNA-protein interac-
tions present vulnerable points at which RNA-binding small
molecules could capitalize.3-5 Observations linking the interfer-
ence of protein synthesis by aminoglycosides with their direct
binding to the ribosomal decoding site (A-site) were the first
example of this RNA targeted approach.6,7 Recently, the
aminoglycosides have been the focus of more intense study,
The ability of aminoglycosides to bind diverse RNA targets
has an unfortunate caveat: these antibiotics are rather promiscu-
ous RNA binders. This promiscuity is partly a result of their
electrostatically driven RNA binding mode.10 From an electro-
static viewpoint, it is likely to be difficult to distinguish different
RNA folds in solution. This problem is exacerbated by the
ability of the aminoglycosides to conformationally adapt to
diverse RNA targets. For example, neomycin B, 1a, or the
closely related paromomycin, 1b, (Figure 1a) assumes a different
conformation, when bound to different RNA targets11-14 (Figure
† University of California.
‡ Anadys Pharmaceuticals.
§ Current address: Molecular, Cellular, and Development Biology, Yale
University, 266 Whitney Avenue, KBT 516, New Haven, Connecticut
06520.
(8) Stage, T. K.; Hertel, K. J.; Uhlenbeck, O. C. RNA 1995, 1, 95-101. Wang,
H.; Tor, Y. J. Am. Chem. Soc. 1997, 119, 8734-8735.
(1) Tor, Y. Angew. Chem., Int. Ed. 1999, 38, 1579-1582.
(2) For selected review articles, see: Hermann, T.; Westhof, E. Curr. Opin.
Biotechnol. 1998, 9, 66-73. Ritter, T. K.; Wong, C.-H. Angew. Chem.,
Int. Ed. 2001, 40, 3508-3533. Griffey, R. H.; Swayze, E. C. Expert Opin.
Ther. Pat. 2002, 12, 1367-1374.
(9) Zapp, M. L.; Stern, S.; Green, M. R. Cell 1993, 74, 969-978. 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, 1173-1184.
(10) Tor, Y.; Hermann, T.; Westhof, E. Chem. Biol. 1998, 5, R277-283.
(11) Fourmy, D.; Recht, M. I.; Blanchard, S. C.; Puglisi, J. D. Science 1996,
274, 1367-1371.
(3) Tor, Y. ChemBioChem 2003, 4, 998-1007.
(4) For selected review articles, see: Chow, C. S.; Bogdan, F. M. Chem. ReV.
1997, 97, 1489-1513. Michael, K.; Tor, Y. Chem.sEur. J. 1998, 4, 2091-
2098. Afshar, M.; Prescott, C. D.; Varani, G. Curr. Opin. Biotechnol. 1999,
10, 59-63. Wilson, W. D.; Li, K. Curr. Med. Chem. 2000, 7, 73-98.
Sucheck, S. J.; Wong, C.-H. Curr. Opin. Chem. Biol. 2000, 4, 678-686.
(5) Hermann, T. Angew. Chem., Int. Ed. 2000, 39, 1891-1905.
(6) Schroeder, R.; Waldsich, C.; Wank, H. EMBO J. 2000, 19, 1-9.
(7) Von Ahsen, U.; Noller, H. F. Science 1993, 260, 1500-1503.
(12) Faber, C.; Sticht, H.; Schweimer, K.; Rosch, P. J. Biol. Chem. 2000, 275,
20660-20666.
(13) Jiang, L.; Majumdar, W. H.; Jaishree, T. J.; Weijun, X.; Patel, D. J. Structure
1999, 7, 817-827. Varani, L.; Spillantini, M. G.; Goedert, M.; Varani, G.
Nucleic Acids Res. 2000, 28, 710-719.
(14) Vicens, Q.; Westhof, E. Structure 2001, 9, 647-658.
9
9818
J. AM. CHEM. SOC. 2005, 127, 9818-9829
10.1021/ja050918w CCC: $30.25 © 2005 American Chemical Society

RNA-Aminoglycoside Specificity
A R T I C L E S
skeleton in an unfavorable orientation for binding to competing
targets. In this contribution we describe the design and synthesis
of two new restricted aminoglycoside analogues 2a and 2b
(Figure 1a), which are specifically designed to exhibit enhanced
selectivity for the A-site relative to the TAR. A comparison of
the RNA binding affinities of these novel products and their
parent, nonrestricted counterparts suggests that conformationally
restricted aminoglycosides show promise as target-selective
binders, but the A-site and TAR RNAs may exhibit a different
level of structural selectivity.
Results
Design of Conformationally Restricted Aminoglycosides.
Antibiotics of the neomycin class exhibit very different structural
features when bound to either the A-site RNA or the HIV-1
TAR (Figure 1b). All three published structures of paromomycin
bound to the A-site reveal a similar structure, with a very
compact arrangement of the four rings and rings II and III in
spatial proximity.11,15,18 Additionally, the B-factors reported for
these structures indicate that the bound conformations of rings
I and II are fairly static, while rings III and IV are increasingly
dynamic. In contrast, when bound to the TAR, the antibiotics
are more dynamic and assume a more extended conformation.12
The published structure of TAR-bound neomycin includes a
collection of 17 refined NMR structures and reveals two primary
conformations, with a range of glycosidic torsion angles among
the first three rings (Figure 1c; Table 1). In both conformations,
rings II and III are distal to one another. Notably, the torsional
angles between rings I and III in both TAR-bound conformations
are very different from the torsional angles in any of the A-site
bound paromomycin structures.
Given these major structural differences, we reasoned that a
short covalent link between rings II and III should constrain
the aminoglycoside into a structure resembling the A-site bound
form, thereby altering its RNA affinity and target selectivity.
The 2′-nitrogen and 5′′-carbon were chosen for placement of
the linker, identifying compounds 2a and 2b as our target mole-
cules (Figure 1b). The average distance between these two atoms
is 3.67 Å when bound to the A-site,14 whereas the average dis-
tance between these two atoms in the TAR-neomycin structures
is 6.9 ( 0.6 Å (standard deviation), further underscoring the
significant conformational difference between the two interac-
tions. Since the newly formed covalent bond would span a length
of approximately 1.5 Å, these restricted aminoglycosides were
expected to bind more readily to the A-site than to the TAR.
To predict the conformational effect induced by the covalent
linker, we performed molecular dynamics simulations of 1a and
2a in water. Using the AMBER molecular dynamics simulation
program, 100 iterative rounds were performed of phases of
heating to 5000 K followed by cooling. Following energy
Figure 1. (A) Chemical structures of the neomycin class aminoglycosides
studied. Neomycin B (1a) and paromomycin (1b) differ only at the 6′
position. Likewise, restricted neomycin (2a) and restricted paromomycin
(2b) differ only at this position. Ring numbers are shown in red roman
numerals. (B) The structure of neomycin bound to the TAR (blue) is overlaid
with the structure of paromomycin bound to the A-site (red). The two
structures were overlaid using the Pymol Molecular Graphics System based
on a best fit of ring II. The 2′N-to-5′′-C distances are indicated. (C) Tor-
sional angles for the glycosidic linkages defined as ΦI/II (O1′-C1′-
O4-C4), ΨI/II (C1′-O4-C4-C3), ΦI/III (C5-O5-C1′′-O1′′), and ΨI/III
(C4-C5-O5-C1′′).
1b). When bound to the prokaryotic ribosomal A-site,14 the
aminoglycoside is compactly arranged, with rings II and III in
spatial proximity. In contrast, when bound to the HIV-1 TAR,12
the aminoglycoside is in a more extended conformation, with
rings II and III distal to one another. The aminoglycoside
conformation is still different when bound to other RNA targets
(see Supporting Information, Figure S.16). Although the rela-
tive positions of the individual neomycin rings differ among
the complexes, the affinity with which neomycin or paromo-
mycin binds each target is not dramatically different.
In an attempt to circumvent this conformational flexibility
and the resulting RNA target promiscuity, conformationally
constrained aminoglycoside analogues have been designed,
synthesized, and evaluated.15-17 Covalently “freezing” the
aminoglycoside conformation may, under ideal circumstances,
yield the following advantageous features: (a) increased affinity
to the desired target due to limited entropy losses upon binding
and (b) increased selectivity by locking the aminoglycoside
(15) For other approaches for exploring and improving aminoglycosides’ RNA
specificity, see: Alper, P. B.; Hendrix, M.; Sears, P. S.; Wong, C.-H. J.
Am. Chem. Soc. 1998, 120, 1965-1978. Wong, C. H.; Hendrix, M.;
Priestley, E. S.; Greenberg, W. A. Chem. Biol. 1998, 5, 397-406.
(16) Luedtke, N. W.; Liu, Q.; Tor, Y. Biochemistry 2003, 42, 11391-11403.
(17) A related approach has been explored for controlling oligosaccharide
interactions. See: Bundle, D. R.; Alibes, R.; Nilar, S.; Otter, A.; Warwas,
M.; Zhang, P. J. Am. Chem. Soc. 1998, 120, 5317-5318. Bundle, D. R.;
Alibes, R.; Nilar, S.; Otter, A.; Warwas, M.; Zhang, P. J. Am. Chem. Soc.
1998, 120, 5317-5318. Navarre, N.; Amiot, N.; van Oijen, A.; Imberty,
A.; Poveda, A.; Jiminez-Barbero, J.; Cooper, A.; Nutley, M. A.; Boons,
G. Chem.sEur. J. 1999, 5, 2281-2294. Wacowich-Sgarbi, S. A.; Bundle,
D. R. J. Org. Chem. 1999, 64, 9080-9089.
(18) Carter, A. P.; Clemons, W. M.; Brodersen, D. E.; Morgan-Warren, R. J.;
Wimberly, B. T.; Ramakrishnan, V. Nature 2000, 407, 897-902.
9
J. AM. CHEM. SOC. VOL. 127, NO. 27, 2005 9819

A R T I C L E S
Blount et al.
reference
Table 1. Conformational Properties of RNA Bound Aminoglycosides
a
ΦI/
ΨI/
ΦI/
ΨI/
III
2′
N/5′′C distance (Å)
II
II
III
paromomycin-A-site NMR
paromomycin-A-site crystalb
neo-TAR Class 1c
78.1
80.6
111.6
89.2
-145.0
-150.9
-42.0
-94.1
-118.1
-90.5
-23.0
-175.5
116.2
160.2
-51.7
129.8
4.38
3.67
7.24
6.36
11
14
12
12
neo-TAR Class 2
a ΦI/II (O1′-C1′-O4-C4), ΨI/II (C1′-O4-C4-C3), ΦI/III (C5-O5-C1′′-O1′′), ΨI/III (C4-C5-O5-C1′′). b Numbers are an average of two bound
paromomycin molecules in an asymmetric unit. c Torsional angles for TAR-bound neomycin are averages of 11 structures (class 1) or 6 structures (class 2)
of the collected set of 17 best-fit structures.
structures (triangles) are very similar to those predicted in this
simulation for neomycin and restricted neomycin (open and
filled circles, respectively). The slight differences that are
observed likely represent minor structural adaptations that occur
upon RNA binding.
It is informative to locate the TAR-bound neomycin confor-
mation within this map. The published structure of TAR-bound
neomycin includes a collection of 17 refined models which best
fit the observed NMR data.12 The range of torsional angles
between rings I and III within these models (blue, Figure 2)
represents a conformational family that is distinctive from that
encountered in the solvent simulation for either 1a or 2a.
Consistent with our design concept, this predicts that it would
be energetically disfavored for the restricted 2a to assume a
conformation necessary for binding to the TAR. Thus, both the
conformations obtained from the molecular dynamics simula-
tions and the 2′N/5′′C distance constraint imposed by the linker
predict that 2a and 2b should bind less well than the parent 1a
and 1b to the TAR, thereby altering target selectivity.
Synthesis. A concise synthesis of the restricted aminogly-
coside analogues has been developed (Scheme 1). A fully Boc-
protected neomycin was activated at the 5′′ position with 2,4,6-
triisopropylbenzenesulfonyl chloride to give 3 as previously
reported19 (Scheme 1a). TFA-mediated deprotection of all Boc
groups, followed by dilution and neutralization with Et3N,
facilitated a slow (10 days) intramolecular cyclization to give
the desired 2a. To simplify isolation, this highly polar product
was first protected as the penta-Boc derivative 4 and column
purified. Acidic deprotection of all Boc groups followed by
reversed-phase purification afforded 2a in 12% overall yield
(based on 3).20
The conformationally restrained paromomycin derivative 2b
was prepared in a fashion similar to 2a (Scheme 1b). To ensure
5′′-selective activation of paromomycin 1b that contains two
Figure 2. Torsional angles of aminoglycosides free in solution or bound
to A-site or the TAR RNAs. (A) Torsional angles encountered during
primary alcohols, the 6′-hydroxyl was first protected as the
isopropylidene ketal. Boc protection of all amines and activation
of the 5′′ position proceeded smoothly to yield 5. The ketal
group was then conveniently removed under the same conditions
utilized for Boc deprotection, providing the 5′′-activated paro-
momycin (Scheme 1b). After neutralization and high dilution
cyclization, followed by a similar protection/purification/depro-
tection scheme, the conformationally restrained paromomycin
2b was obtained in 20% overall yield (based on 5).
molecular dynamics simulations in water are shown for neomycin 1a (open
circles) and restricted neomycin 2a (closed circles). Note the tighter
distribution of 2a conformations. Torsional angles of TAR-bound neomycin
are shown in blue. Torsional angles of paromomycin 1b when bound to
the A-site as determined by NMR11 or crystallography14 are shown in green
or red, respectively.
minimization, the torsional angles between each of the first three
rings of the cooled conformation were recorded, and their
distributions were examined (Figure 2). Interestingly, the
conformational space available to the restricted 2a is very similar
to that available to 1a. The primary difference is that the
distribution of different conformations is somewhat more
scattered for the nonrestricted 1a, particularly around the
glycosidic bond connecting rings I and III (Figure 2b). This
difference is consistent with the expected reduction of confor-
mational flexibility enforced by the covalent linkage. Impor-
tantly, the torsional angles in the A-site-bound paromomycin
It is important to note that protection of both conformationally
constrained analogues 2a and 2b has always yielded (n - 1)
protected derivatives 4 and 6, respectively (where n is the total
number of amines), as confirmed by NMR and MS analysis.21
(19) Michael, K.; Wang, H.; Tor, Y. Bioorg. Med. Chem. 1999, 7, 1361-1371.
(20) See Supporting Information for synthetic procedures and characterization.
(21) The NH peaks are well-resolved in acetone/d6-DMSO/d6 (3:1) and full
assignment shows that the 2′-N is lacking a Boc group.
9
9820 J. AM. CHEM. SOC. VOL. 127, NO. 27, 2005

RNA-Aminoglycoside Specificity
A R T I C L E S
Scheme 1. Synthesis of Conformationally Constrained Analogues 2a and 2ba
a Reagents and conditions: (a) TFA, CHCl3; (b) NEt3, DMF; (c) Boc2O, NEt3, methanol; (d) TFA, CHCl3. Ar ) 2,4,6-triisopropylbenezene.
In both cases the least reactive amine is the newly formed
bridging secondary amine. This lower nucleophilicity may be
a result of steric or electronic effects. These reactivity observa-
tions are pertinent to the RNA binding characteristics of the
new derivatives as elaborated below.
anomeric protons of rings II, III, IV (δ 5.0-6.0) and the
methylene protons of ring I (δ ) 2.58 and 1.94) provide a
convenient starting point, and gCOSY spectra assist in assigning
the proton system of each ring (see Supporting Information for
all spectra). Phase sensitive gHSQC spectra provide a definitive
assignment for all methylene groups at positions 2, 6′, 5′′ and
6′′′ (see Figure 1 for numbering). Edited gHSQC spectra
differentiate the carbons connected to oxygen (δ > 55.0) from
the carbon centers connected to nitrogen or other carbon atoms
(δ < 55.0), with one exception: 2′-C at 60.2 ppm.23,24 This
significant downfield shift is consistent with converting the
primary amine at this position to a secondary amine.23 Full
assignment was not possible until 1H-13C gHMBC spectroscopy
unequivocally secured the assignment of ring II and ring IV by
showing correlations across the glycosidic bonds (i.e., 4-H to
1′-C, 4-C to 1′-H, 3′′-H to 1′′′-C, 3′′-C and 1′′′-H).25 It is
important to note that this technique is the method of choice to
confirm the cyclization site by demonstrating a correlation
between the two nuclei across the newly formed secondary
amine. Indeed, a cross-peak between 5′′-Hb and 2′-C observed
for both 2a and 2b unambiguously cements the ring connectivity
(Figure 3).26
Spectroscopic Characterization. The intramolecular cy-
clization imposes a significant conformational constraint on the
aminoglycosidic skeleton. In terms of composition, however,
the cyclized structures 2a and 2b can essentially be viewed as
“dehydrated” antibiotics (i.e., the two structures differ only by
a molecule of water). To unequivocally prove the formation of
the N-C bond between the amine at the 2′ position and the 5′′
carbon on the D-ribose, a series of NMR experiments has been
conducted to assign all resonances and establish the cyclization
site.
1
A qualitative comparison of the H NMR spectra of the
restricted neomycin derivative 2a and its parent natural product
1a reveals intriguing and informative changes.22 The 1H NMR
spectrum of the cyclized derivative, under the same pH and
counteranion conditions, shows a dramatic upfield shift of the
5′′ methylene group, when compared to neomycin’s spectrum.22
This observation is in full agreement with the replacement of
an oxygen by a nitrogen functional group replacement. The
change in chemical shift pulls the 5′′-Hb peak out of a crowded
spectral region and significantly facilitates its assignment.
Interestingly, the rest of the proton signals of the D-ribose system
(ring III) experience a dramatic downfield shift.22 The anomeric
proton (H1′′) shifts further downfield to H1′, a rare observation
for neomycin derivatives. These extraordinary changes in
chemical shifts are likely due to a significant conformational
distortion of the D-ribose ring enforced by the intramolecular
cyclization.
15N NMR Studies. To further cement the proposed structure
and to evaluate the basicity of the new restricted antibiotics,
15N NMR spectra of the TFA salts of compound 2a and
neomycin 1a were recorded. Comparing the spectra at pH >
10.5 shows that while most amines exhibit similar chemical
shifts, the 15N signal of the 2′ amine in 2a is approximately 5
ppm upfield shifted compared to the corresponding amine in
the parent neomycin B 1a (Figure 4). This observation further
(23) Similar chemical shift changes have been predicted and observed for the
corresponding carbon in cyclohexylammonium (51.6 ppm) vs N-methyl-
cyclohexylammonium (59.4 ppm). See: Sarneski, J. E.; Surprenant, H. L.;
Molen, F. K.; Reilley, C. N. Anal. Chem. 1975, 47, 2116-2124.
(24) This secondary carbon is clearly distinguished by its phase.
(25) This spectrum also unequivocally cements the assignment of the pseudo-
symmetrical 2-DOS ring.
(26) The cross-peak between 5′′-Ha and 2′-C was not observed. As with H-H
coupling, the value of this 3-bond coupling constant is dependent on the
dihedral angle. The absence of cross-peaks is not uncommon and can be
understood as the distortion of compounds 2a and 2b from a common
conformation.
The full assignment of compound 2a was accomplished using
1D proton NMR as well as 2D gradient-assisted correlation
spectroscopy (gCOSY), gradient-assisted heteronuclear single
quantum coherence (1H-13C gHSQC) and heteronuclear mul-
tiple bond correlation (1H-13C gHMBC) spectroscopies. The
(22) See Supporting Information for additional NMR spectra.
9
J. AM. CHEM. SOC. VOL. 127, NO. 27, 2005 9821

A R T I C L E S
Blount et al.
Figure 3. 1H-13C gHMBC spectrum of restricted neomycin B 2a (top)
and an expansion (bottom) highlighting the cross-peak from 5′′-Hb to 2′-C.
supports the higher substitution of this particular nitrogen and
is in agreement with the 13C NMR and the reactivity data
discussed above.
Recording the 15N NMR spectra as a function of pH results
in well-defined titration curves that can be fitted to yield the
pKa value of each amine (Figure 4 and Table 2).27,28 The newly
formed bridging secondary amine is found to have significantly
attenuated basicity. With a pKa value of 6.37, it is almost 2 pKa
units below the amine at the same position in neomycin B (pKa
8.14). The lower basicity of this amine may be due to its
crowded environment that hinders suitable solvation of the
corresponding ammonium ion (and is in agreement with the
low nucleophilicity exhibited by this functional group). Interest-
ingly, the lower basicity of this center renders all other amines
in 2a more basic when compared to their corresponding amines
in neomycin 1a (Table 2), as would be expected based on
intramolecular electrostatic considerations. For example, the pKa
of the NH2 at position 1 of 2a is 0.4 units higher than the
analogous amine in neomycin 1a. While these absolute pKa
values are likely to depend on the counterions,29 the experi-
mentally determined values clearly reflect the impact of cy-
clization on the basicity of the 2′-amine.
Figure 4. 15N NMR spectra of restricted neomycin 2a and neomycin 1a
at pH 10.7 show the distinct chemical shift of the 2′-nitrogen. The chemical
shifts of each nitrogen are shown below as a function of pH. Symbols
highlight individual nitrogens: 1 (open triangles), 2 (closed triangles), 2′
(blue filled squares), 6′ (filled circles), 2′′′ (open squares), and 6′′′ (open
circles).
analogue 2-aminopurine (2AP) into the A-site as a means for
detecting and quantifying aminoglycoside binding.30 The 2AP
substitution at A1492 accurately reports the unstacking of this
nucleotide upon binding of the neomycin-class antibiotics.
Notably, the magnitude of the 2AP fluorescence increase upon
binding is relatively constant among different aminoglycosides
and is strongly correlated with the specific structural changes
that occur upon binding.30 Capitalizing on these reports, the
binding of the aminoglycosides discussed here was measured
by titration into a fixed concentration (∼200 nM) of A-site
2AP(1492) (Figure 5a,b). Using the directly measured 1:1 stoi-
chiometry of paromomycin binding to the A-site, the fluores-
Ligand Binding to the A-Site RNA. Two recent contribu-
tions report the incorporation of the fluorescent nucleoside
(29) Kaul, M.; Barbieri, C. M.; Kerrigan, J. E.; Pilch, D. J. Mol. Biol. 2003,
326, 1373-1387.
(30) Kaul, M.; Barbieri, C. M.; Pilch, D. J. Am. Chem. Soc. 2004, 126, 3447-
3453. Shandrick, S.; Zhao, Q.; Han, Q.; Ayida, B.; Takahashi, M.,
Takahashi; Winters, G. C.; Simonsen, K. B.; Vourloumis, D.; Hermann,
T. Angew. Chem., Int. Ed. 2004, 43, 3177-3182.
(27) Assignment of 15N chemical shift is based on: Botto, R. E.; Coxon, B. J.
J. Am. Chem. Soc. 1983, 105, 1021-1028.
(28) The pKa values obtained for neomycin B are in good agreement with the
values recently reported by Pilch. See: Kaul, M.; Barbieri, C. M.; Kerrigan,
J. E.; Pilch, D. J. Mol. Biol. 2003, 326, 1373-1387.
9
9822 J. AM. CHEM. SOC. VOL. 127, NO. 27, 2005

RNA-Aminoglycoside Specificity
A R T I C L E S
Table 2. 15N NMR Determination of pKa Values for All Amine Groups
neomycin (1a)
restricted neomycin (2a)
δNH3+ (ppm)
a
amine
δ
NH2 (ppm)a
δNH3+ (ppm)a
pKa
δNH2 (ppm)
pKa
1
3
2′
6′
2′′′
6′′′
33.5 ( 0.2
35.8 ( 0.1
25.1 ( 0.1
17.2 ( 0.4
18.7 ( 0.1
18.9 ( 0.3
40.5 ( 0.2
42.0 ( 0.3
36.6 ( 0.1
28.7 ( 0.2
32.3 ( 0.1
29.0 ( 0.2
8.50 ( 0.07
6.59 ( 0.08
8.14 ( 0.03
9.36 ( 0.08
8.07 ( 0.02
9.65 ( 0.07
32.9 ( 0.1
35.9 ( 0.1
29.7 ( 0.1
16.0 ( 0.1
18.9 ( 0.1
18.9 ( 0.2
40.2 ( 0.1
41.8 ( 0.1
40.9 ( 0.2
28.4 ( 0.1
32.0 ( 0.1
29.3 ( 0.1
8.89 ( 0.04
6.68 ( 0.04
6.37 ( 0.03
9.45 ( 0.03
8.34 ( 0.03
9.65 ( 0.04
a Tolerances indicate the standard error determined from curve fitting.
Figure 5. Measuring the RNA target binding of aminoglycosides. (A) Secondary structure of the fluorescently labeled A-site RNA used in this study, A-site
2AP (1492) RNA. Position 1492 is highlighted in red. The structure of 2-aminopurine is shown. (B) Sample binding isotherms of paromomycin (1b) and
restricted paromomycin (2b) binding to A-site 2AP(1492) at pH 7.5. Shown below is the dependence of dissociation constants on pH. Error bars indicate
plus or minus one standard deviation of three independent measurements. (C) Secondary structure of the fluorescently labeled TAR RNA used in this study,
TAR 2′p(25) with position 25 highlighted in orange. The structure of the 2′-aminobutyrl-pyrene label is shown. (D) Sample binding isotherm of paromomycin
(1b, right axis) and restricted paromomycin (2b, left axis) binding to TAR 2′p(25) at pH 7.5. The axes show the normalized increase in fluorescence upon
aminoglycoside binding as described.35 Shown in the bottom panel is the pH-dependence of binding. Error bars indicate plus or minus one standard deviation
of three independent measurements.
cence increase upon binding of the aminoglycoside was con-
verted to fractional saturation and plotted against the concentra-
tion of the unbound aminoglycoside. The binding isotherm fits
well to a two-state binding model and yields a reproducible
measure of binding affinity. Using this assay, paromomycin
binds to the A-site at pH 7.5 with a Kd of 170 nM (Table 3),
which compares favorably with literature reports.31-33 Much like
previous studies,31,34 under the conditions used to measure
paromomycin binding, neomycin binding to the A-site does not
adhere to a simple two-state model and does not yield repro-
ducible affinity values from this assay. Therefore, for the pur-
pose of comparison, the approximate dissociation constant for
the first neomycin binding event is compared here to the value
of 19 nM at pH 7.5 measured by surface plasmon resonance.31
Notably, restricted-neomycin 2a does show single-state binding,
much like paromomycin. This may implicate the 2′-amine as a
contributor to neomycin’s nonspecific binding.
(31) Alper, P. B.; Hendrix, M.; Sears, P. S.; Wong, C.-H. J. Am. Chem. Soc.
1998, 120, 1965-1978.
(32) 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, 10129-10133.
(33) Fourmy, D.; Recht, M. I.; Puglisi, J. D. J. Mol. Biol. 1998, 277, 347-362.
(34) Kaul, M.; Pilch, D. Biochemistry 2002, 41, 7695-7706.
9
J. AM. CHEM. SOC. VOL. 127, NO. 27, 2005 9823

A R T I C L E S
Blount et al.
Table 3. Target Binding Affinities of Aminoglycosides
A-site
Kda pH 5.8
M)
K
d pH 6.8
M)
K
d pH 7.5
M)
(
µ
(
µ
(µ
neomycin B 1a
restricted neomycin 2a
paromomycin 1b
restricted
ND
0.07 ( 0.02
0.05 ( 0.02
ND
0.16 ( 0.01
0.06 ( 0.01
0.02b
0.42 ( 0.03
0.17 ( 0.01
paromomycin 2b
0.12 ( 0.03
0.35 ( 0.07
2.4 ( 0.4
TAR
K
d pH 5.8
M)
K
d pH 6.8
M)
K
d pH 7.5
M)
(
µ
(
µ
(µ
neomycin B 1a
restricted neomycin 2a
paromomycin 1b
restricted
1.2 ( 0.2
2.2 ( 0.6
6.2 ( 0.8
1.5 ( 0.2
3.8 ( 0.2
12.6 ( 0.4
2.4 ( 0.3
25.8 ( 0.6
50 ( 3
paromomycin 2b
9.6 ( 0.7
23.5 ( 0.2
118 ( 3
a Tolerances indicate the standard deviation of at least three independent
determinations. b From Alper et al.31
At pH 7.5, the restricted-neomycin 2a and restricted-paro-
momycin 2b bind the A-site with 22- and 14-fold lower affinity,
respectively, compared to neomycin and paromomycin, the
parent natural products. This decreased binding affinity is
likely to be largely due to the lower overall charge of the
cyclized derivatives due to the diminished basicity of the
2′-amine in the restricted molecules. Indeed, upon decreasing
the pH to 5.8 to more fully protonate this amine, the affinities
of restricted-neomycin and restricted-paromomycin are only
3.8- and 2.3-fold lower than the parent aminoglycosides. Thus,
under conditions where the overall protonation states are sim-
ilar, the conformational restriction of these neomycin-class
antibiotics only modestly decreases their binding affinity to the
A-site.36
Figure 6. DMS modification footprint of aminoglycoside binding the A-site
RNA, detected by primer extension. “No dd” lane shows the primer
extension of untreated RNA without any dideoxy-NTPs present. U,G,C,A
lanes are untreated RNA, sequenced using dideoxy-NTPs. “0” lanes show
the modification by DMS and subsequent reduction and strand cleavage in
the absence of aminoglycoside. DMS footprinting reactions in the presence
of aminoglycoside are at concentrations corresponding to Kd, 4 × Kd, 16
× Kd, and 64 × Kd. The similarities in the footprint between aminoglyco-
sides are noted as arrows, corresponding to a decrease in cleavage as the
concentration of aminoglycoside is increased. Note that because aniline
strand cleavage removes the modified nucleotide, primer extension stops
indicated modification at one nucleotide closer to the 5′-end.
A-site 2AP(1492), which is strongly correlated with the
structural changes at A1492 and A1493 that accompany binding,30
is similar among all the aminoglycoside tested, also confirming
that the structural details of the resulting complexes are similar.40
Ligand Binding to the HIV-1 TAR. Binding of all ami-
noglycosides to the TAR was measured using our previously
validated and published assay, in which a 2′-aminobutyrylpyrene
label at U25 accurately reports binding (Figure 5c,d).35 At pH
7.5, restricted-neomycin 2a and restricted-paromomycin 2b bind
the TAR with 11- and 2.4-fold lower affinity, respectively, when
compared to neomycin and paromomycin (Table 3). Based on
the structural design of these molecules, this modest decrease
in affinity is surprising. More surprising, however, is that, upon
decreasing the pH to 5.8 to more fully protonate the less basic
secondary amine, the affinities of restricted-neomycin and
restricted-paromomycin compared to the parent aminoglycosides
are only 1.8- and 1.6-fold lower (Table 3). Thus, under
conditions where the amine protonation states are normalized,
restricting these neomycin-class antibiotics has no deleterious
effect on binding to the HIV-1 TAR.
The surprising lack of deleterious effects of covalent cycliza-
tion on TAR binding prompted a deeper examination of the
structural details of this RNA-ligand interaction. It is conceiv-
able that the restricted aminoglycosides bind to different
locations on the TAR or induce very different RNA structural
changes upon binding. To examine this possibility, the RNase
V1 footprint of each aminoglycoside on the TAR was evaluated
as described.35 RNase V1 cleaves primarily double-stranded or
strongly stacked nucleotides. All natural and cyclized aminogly-
To correctly assess the effect of the conformational restraint
on A-site binding, it is important that 1a and 2a bind at the
same place on the A-site RNA. Previous dimethyl sulfate (DMS)
footprinting studies indicate that aminoglycoside binding only
modestly alters the A-site structure.37,38 Specifically, the primary
nucleotides whose reactivities significantly change in response
to ligand binding are G1494, G1405, G1491, and to a lesser extent
G
1497. Using similar methods as those reported,38 the DMS
reactivity of a 3′-extended A-site RNA39 indicates that binding
of the restricted and nonrestricted paromomycin analogues
reported here produce a similar footprint (Figure 6). As with
previous studies, the primary ligand-dependent changes for both
the nonrestricted and restricted aminoglycosides are at G1494
,
A1493, and G1497, implying that the conformational constraint
does not alter the structural context of the aminoglycoside
interaction with the A-site. Furthermore, the magnitude of the
ligand-dependent increase in 2-aminopurine fluorescence of
(35) Blount, K. F.; Tor, Y. Nucleic Acids Res. 2003, 31, 5490-5500.
(36) As the pH is decreased, the difference in affinity between neomycin and
paromomycin decreases slightly. This suggests the possibility that the cause
for convergence of the affinities of the restricted and nonrestricted
aminoglycosides could include smaller, additional terms besides the pKa
of the 2′-amine.
(37) Miyaguchi, H.; Narita, H.; Sakamoto, K.; Yokoyama, S. Nucleic Acids Res.
1996, 24, 3700-3706.
(38) Recht, M. I.; Fourmy, D.; Blanchard, S. C.; Dahlquist, K. D.; Puglisi, J.
D. J. Mol. Biol. 1996, 262, 421-436.
(39) The 3′ extended version of the A-site oligonucleotide, 5′-GGCGUCA-
CACCUUCGGGUGAAGUCGCCGGUUGGCGUGGCUCGCG-3′, facili-
tates detection of DMS reactivity by primer extension via a 17-nucleotide
DNA primer complimentary to the 3′-extension.
9
9824 J. AM. CHEM. SOC. VOL. 127, NO. 27, 2005

RNA-Aminoglycoside Specificity
A R T I C L E S
Figure 7. Enzymatic footprint (RNAse V1) of aminoglycoside binding to all 2′-hydroxy HIV-1 TAR. “Std” lanes are untreated RNA. “OH-” lanes are an
alkaline hydrolysis ladder of each RNA, and “T1” lanes show cleavage by ribonuclease T1 under denaturing conditions, in the absence of neomycin. “0”
lane shows the cleavage by RNAse V1 in the absence of aminoglycoside. RNAse V1 footprints in the presence of aminoglycoside are at concentrations
corresponding to 0.25 × Kd, Kd, 4 × Kd, 16 × Kd, and 64 × Kd. A strong cleavage indicates stacked or double-stranded regions of the structure. Similarities
in the footprint between aminoglycosides are noted as arrows pointing up or down, corresponding to either increases or decreases in cleavage, respectively,
as the concentration of aminoglycoside is increased. The loop and bulged regions are highlighted in the sequence as red or blue, respectively.
cosides tested herein generate a very similar RNase V1 footprint
on the TAR (Figure 7). Most notable among these similarities
are a ∼2.5-fold increase in cleavage 3′ of C39 and U40 and at
the 5′ side of the loop at U31, and decreases in cleavage near
the bulge at U25, G26, and A27, and near the 5′-end at C19, A20,
and G21. Quantification of these changes as a function of
aminoglycoside concentrations indicates a half-maximal protec-
tion concentration that agrees well with the data measured by
the fluorescence assay (Supporting Information).41 Thus, to the
extent that RNAse V1 can detect, the location of the binding
site and the TAR structural changes associated with binding
are very similar among all the aminoglycosides tested here.
modulate their binding to a desired target, to the exclusion of
competing targets within a cell. At present, the aminoglycosides
are rather promiscuous binders, owing to both their electrostatic
binding mode and their ability to conformationally adapt to
diverse RNA folds. In an attempt to circumvent the latter, we
have synthesized conformationally restricted aminoglycoside
analogues designed to bind to the prokaryotic ribosomal A-site
to the exclusion of the HIV-1 TAR. By measuring the binding
of these novel aminoglycosides to both RNA targets, one can
accurately assess and more thoroughly understand selectivity
determinants.
A-Site-Restricted Aminoglycoside Interaction. As il-
lustrated in Figure 1, the 2′-nitrogen and 5′′-carbon of paro-
momycin are approximately 3.6 Å apart when bound to the
A-site,14 and a hydrogen bond is likely to exist between the
proton of the 2′-nitrogen and the 5′′-oxygen.14 Thus, a newly
formed 2′-N/5′′-C covalent bond should constrain rings II and
III in proximity and preorganize paromomycin into a conforma-
tion resembling the A-site bound form. In theory, this preor-
ganization could be predicted to enhance the A-site binding
affinity, due to a smaller change in the conformational entropy
upon binding.43 In practice, however, this phenomenon is seldom
observed.17 Rather, the decreased conformational entropy is
often compensated for by enthalpic changes, resulting in similar
Discussion
Currently, numerous examples exist in the literature of low
molecular weight ligands which bind to and alter the function
of therapeutically important RNA targets.8,9,42 Although much
work has focused on ways to enhance the affinity of these
interactions, comparatively little research has examined the issue
of target selectivity.16 Ultimately, for these molecules to be
clinically successful, one will need to understand how to
(40) Initial refinement of cocrystal diffraction data (T. Hermann, unpublished
data) shows that 2a binds within the A-site essentially the same as the
parent 1a. Thus, addition of the conformational restraint does not change
the binding mode or grossly alter the ligand-RNA interaction.
(41) The half-maximal footprint of Ribonuclease A also corresponds well with
the measured dissociation constants. See Supporting Information Figure
S.19.
(43) Estimates for loss of conformational entropy upon binding for one bond
are ∼0.6 kcal mol-1. See: Nivotny, J.; Bruccoleri, R. E.; Saul, F. A.
Biochemistry 1989, 28, 4735-4749.
(42) Von Ahsen, U.; Davies, J.; Schroeder, R. Nature 1991, 353, 368-370.
9
J. AM. CHEM. SOC. VOL. 127, NO. 27, 2005 9825

A R T I C L E S
Blount et al.
affinities for constrained versus nonconstrained carbohydrates
toward a specific target.44 Alternatively, some minor structural
adaptations, with an accompanying slight energetic cost, may
still be necessary for 2b to bind the A-site. Indeed, under
conditions where their protonation states are equal, 1b and 2b
bind the A-site with very similar affinities. Likewise, the
neomycin derivatives 1a and 2a bind to the A-site with similar
affinities.40 Thus, despite not increasing the binding affinity,
these conformational constraints can be tolerated isoenergeti-
cally, providing an opportunity for enhancement of selectivity
among targets. This provides an important precedent for the
rational design and development of conformationally restricted
aminoglycosides.
TAR-Restricted Aminoglycoside Interaction. As illus-
trated in Figure 1, the 2′-nitrogen and 5′′-carbon that are
covalently linked in the restricted aminoglycosides 2a and 2b
are ∼6.9 Å apart in neomycin when bound to the TAR. Since
footprinting results indicate that the restricted aminoglycosides
bind at the same site on the TAR as the parent compounds, it
is surprising that both ligands are bound with comparable affinity
to the nonrestricted parent compounds. This result is more
baffling given the molecular dynamics prediction that the
aminoglycoside conformation typically assumed while binding
to the TAR is not easily accessed in solution by the restricted
aminoglycosides. That is, the covalent link enforces an ami-
noglycoside “conformation” that is incompatible with the known
structural data.
Figure 8. Structural implications of docking restricted neomycin 2a
(orange) onto the TAR. Grey traces represent the RNA backbone of the 17
best-fit NMR structures of the TAR with bound neomycin (blue). The
backbone at U23 and A22 is highlighted in green and yellow, respectively.
To qualitatively explore this apparent quandary, we used the
PyMol structural viewing program to attempt to dock restricted
neomycin onto the TAR. Specifically, a model aminoglycoside
was created in which the neomycin glycosidic torsion angles
resembled those encountered in the molecular dynamics simula-
tion (Figure 2). In addition the conformation was altered so that
rings II and III were close enough to form the 2′N/5′′C bond.
Since the neamine conformation is relatively constant among
known aminoglycoside-RNA interactions, only modest changes
were introduced in the torsional angles between rings I and II,
and the position of ring I was not changed. The resultant model
of 2a (Figure 8) has torsional angles typically representative of
the conformational space encountered in molecular dynamics
simulations (Figure 2; Table 1) and is similar to the A-site bound
structures. Furthermore, the NH3+-NH3+ distances within the
aminoglycoside are consistent with the energetically favorable
ranges reported for neomycin in solution.45
The torsional angles of this conformation of 2a are ΦI/II ) 102.5, ΨI/II
-79.3, ΦI,III ) -64.5, ΨI/III ) 100.4.
)
Table 4. Distance (Å) from Each Amine to the Nearest Phosphate
When Bound to Different RNA Targets
A-site
TAR
N1
N3
N2′
N6′
N2′′′
N6′′′
6.4
3.2
4.7
NA
3.2
4.8
3.1
5.6
6.2
7.1
6.3
6.8
is at least 30% electrostatic,5,35 the position of the amines relative
to RNA phosphate groups must be an important binding
determinant. Of the six total amines on neomycin, N1 and N3
on ring I are the closest to neighboring RNA phosphate groups
(Table 4), implying their energetic importance. If the binding
energy provided by these amines is to be maintained, it seems
unlikely that the relative position of this ring would be drastically
altered.
The second possibility for overcoming the steric clash induced
by the conformational restraint would be RNA structural
adaptation. That is, the aminoglycoside binding site on the TAR
may have the requisite plasticity to form significantly different
structural interactions with the restricted or nonrestricted deriva-
tives. From an RNA perspective, accommodation of the
restricted neomycin conformer shown would require movement
of U23 toward the major groove and a shift of the backbone at
A22 by several angstroms. The ESR signature of a nitroxide
label46 at U23 demonstrates that it is very dynamic in the
presence or absence of neomycin and, thus, could adapt without
As Figure 8 clearly indicates, the interaction of the TAR with
restricted neomycin must inherently be different than the
neomycin interaction. As shown here, rings III and IV would
encounter serious steric clash with the TAR backbone at A22
and U23 (yellow and green, respectively, in Figure 8) and the
base of U23. Two primary possibilities exist for how this could
be overcome. The first is that the conformation and relative
position of the neamine core could change more dramatically
than expected, permitting a different orientation within the
binding site with perhaps less steric clash. Energetically, this
seems unlikely, given that the conformation of neamine is mostly
conserved among known RNA-aminoglycoside interactions.
Further, since the binding energy of aminoglycosides to the TAR
(44) Carver, J. P. Pure Appl. Chem. 1993, 65, 763-770. Alibe´s, R.; Bundle, D.
R. J. Org. Chem. 1998, 63, 6288-6301.
(45) Hermann, T.; Westhof, E. J. Mol. Biol. 1998, 276, 903-912.
9
9826 J. AM. CHEM. SOC. VOL. 127, NO. 27, 2005

RNA-Aminoglycoside Specificity
A R T I C L E S
serious energetic penalty.47 The necessarily extensive movement
of the backbone at A22 would further widen the minor groove
and propagate other significant localized structural perturbations.
Since this bulged region is largely single-stranded, the RNase
V1 footprints can neither confirm nor deny these possible
differences. However, quantification of the footprint at the
3′-side of the bulge at G26 suggests slight differences between
the interaction with 2a and 2b or 1a and 1b.48 Although it is
unclear what energetic impact these possible RNA structural
adaptations would have, it is clear that the structural interaction
must be inherently different from the published TAR-neomycin
structure. Thus, from either an aminoglycoside or an RNA
perspective, it is apparent that, despite having a similar binding
affinity, the restricted aminoglycosides must have a different
structural binding interface with the TAR than the nonrestricted
aminoglycosides.
the A-site with 10- to 20-fold higher affinity compared to
tobramycin.32,49 Thus the A-site is considerably more selective
and less tolerant of the aminoglycoside structure compared to
the TAR.
Summary
A straightforward synthesis affords two new and intriguing
conformationally constrained aminoglycosides. Examining the
affinity of these unique analogues to two different RNA targets,
the A-site and the HIV-1 TAR, opens a window to the complex
world of RNA selectivity. While we tend to view aminogly-
cosides as rather nonselective RNA binders, our results reveal
that different RNA sites may exhibit inherently different ligand
selectivity. The deeply encapsulating A-site appears to be a
much more discriminating RNA target when compared to the
“shallow” HIV TAR. For the future design of novel target
selective therapeutics, it will be necessary to more fully consider
the structural selectivity of the RNA target and not only the
inherent selectivity displayed by the small molecules.
TAR vs A-Site Specificity. The binding of aminoglycosides
to RNA targets is highly promiscuous. The common paradigm
is that this lack of selectivity is a consequence of the highly
cationic nature of the aminoglycosides and the inherent flex-
ibility of their glycosidic bonds. This “plasticity” facilitates a
structural adaptation wherein the aminoglycosides can remodel
themselves to match the electrostatic field generated by the RNA
fold. In attempt to circumvent this promiscuous adaptation, we
have introduced a conformational constraint into the aminogly-
cosides. Contrary to our design principles, the TAR RNA can
bind these restricted aminoglycosides equally as well as the
natural antibiotics. This suggests a converse paradigm; different
RNA targets may exhibit varying degrees of selectivity for
natural and modified aminoglycosides. Indeed, closer compari-
son of the interaction of aminoglycosides with either the A-site
or TAR suggests that these therapeutically relevant RNA sites
may possess inherently different selectivity traits. The ami-
noglycosides are more fully enveloped by the RNA fold of the
A-site than in the TAR, as evidenced by solvent accessibility
and the proximity of the amines to nearby phosphates (Table
4). Accordingly, thermodynamic data predict that four of five
possible ion pairs are formed between paromomycin and the
A-site,34 whereas only two of six possible ion pairs are formed
between neomycin and the TAR.35 In addition, the very minimal
chemical and enzymatic footprint of aminoglycoside binding
to the A-site implies that it is a relatively preformed scaffold
for aminoglycoside binding. In contrast, the footprinting data
for aminoglycoside binding to the TAR indicate changes
throughout the RNA, suggesting an inherently adaptable ami-
noglycoside binding scaffold. Finally, as demonstrated here, the
TAR can accommodate a very different aminoglycoside struc-
ture with minimal deleterious effects on binding. Previous
studies have suggested that the A-site is comparatively less
forgiving of structural differences.49 A notable example of the
discriminatory difference of these two RNAs is the relative
affinity to paromomycin and tobramycin. These two similarly
charged aminoglycosides are bound by the TAR with roughly
the same affinity.35 In contrast, however, paromomycin binds
Experimental Section
Synthesis. All synthetic procedures and characterization as well as
NMR spectra and assignment are included in the Supporting Informa-
tion.
Materials. Neomycin and paromomycin were purchased as their
sulfate salts from Sigma-Aldrich. Mes (4-morpholinoethanesulfonic
acid), Mops (4-morpholinepropanesulfonic acid), and Hepes (4-(2-
hydroxyethyl)-1-piperazineethanesulfonic acid) buffers, and all inor-
ganic salts were purchased from Fisher (enzyme grade). Nonidet 40
was purchased from Fluka.
1
NMR Spectroscopy. H, 13C, and 2D NMR spectra were recorded
on a Varian Mercury 400 MHz spectrometer. Dioxane was used as the
internal standard (δ ) 66.60) for 13C chemical shifts unless otherwise
noted. 15N NMR spectra were recorded at 50.7 MHz at 23 °C on a
Varian Unity 500 spectrometer using a 5 mm inverse broad band probe
with an acquisition time of 1 s and a recycle delay of 0.5 s. 15NH4Cl
(1M, 10% enriched) in 85:15 (v:v) H2O/D2O contained in a coaxial
2-mm inner cell (from New Era) was used as an external reference (δ
) 24.00). The 15NH4Cl standard was purchased from Cambridge Isotope
Laboratories, Inc. NMR solutions were prepared by dissolving ami-
noglycosides (TFA salt form, 340 mg each) in 0.5 mL 85:15 (v:v)
H2O/D2O to yield the final concentration of 0.5-0.6 M. The solution
pH was adjusted by 31% HCl and 40% KOH in 85:15 (v:v) H2O/D2O.
All pH measurements were acquired with a Corning 320 pH meter
interfaced with a microstem glass/calomel combination electrode (from
Aldrich). The 15N resonances were assigned according to Botto and
Coxon.27 The pKa values were determined by fitting the data to the
equation shown below using Kaleidagraph:
(δNH - δNH )(10pH-pK
)
a
2
3
δ )
+ δNH
3
1 + 10pH-pK
a
Molecular Dynamics Simulations. For both neomycin and restricted
neomycin, simulations in a box of explicit water (∼580 H2O and 4
Cl-) were run under periodic boundary conditions at a time step of 1
fs. AMBER50 force field parameters were used as described in our
previous aminoglycoside simulations.45 100 cycles each of heating to
5000 K for 10 ps followed by cooling to 10 K over 5 ps were performed.
Structures were extracted at the end of the 10 K cycles and subjected
to 1000 steps of conjugate gradient minimization. Torsion angles were
measured and recorded. There was no filtering for high-energy structures
(46) Edwards, T. E.; Sigurdsson, S. T. Biochemistry 2002, 41, 14843-14847.
(47) The NMR solution structure also reveals significant mobility for this residue.
See ref 12.
(48) Quantification of RNAse V1 footprint indicates that 2a and 2b may induce
more protection at G26 than 1a and 1b.
(49) Simonsen, K. B.; Ayida, B.; Vourloumis, D.; Takahashi, M.; Winters, G.
C.; Barluenga, S.; Qamar, S.; Shandrick, S.; Zhao, M.; Hermann, T.
ChemBioChem 2002, 3, 1223-1228.
(50) Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M. J.;
Ferguson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.; Kollman, P.
A. J. Am. Chem. Soc. 1995, 117, 5179-5197.
9
J. AM. CHEM. SOC. VOL. 127, NO. 27, 2005 9827

A R T I C L E S
Blount et al.
which might be conformationally trapped by the solvent or ions.
However, comparison of plots without the 50% highest energy
conformations showed no significant difference in distribution of the
conformers over the map.
aminoglycoside, or at saturation, respectively. As previously described,51
the concentration of free aminoglycoside, [ag]free, can be calculated from
this fractional saturation and the total aminoglycoside concentration,
[ag]tot:
Preparation of Oligonucleotides. All oligonucleotides were pur-
chased from Dharmacon RNA technologies (Lafayette, CO), depro-
tected as per manufacturer protocol, and lyophilized. The deprotected
oligonucleotides were resuspended in water and purified using 20%
denaturing PAGE. Excised gel slices were eluted overnight into a buffer
containing 50 mM Hepes pH 7.5 and 1 M NaCl, followed by ethanol
precipitation. After a 70% ethanol/water wash, the RNA pellets were
dried and resuspended in water, quantified by UV absorbance, and their
identities were confirmed by MALDI-TOF mass spectrometry. For
binding experiments, the TAR was refolded by heating a 15 µM solution
of the pyrene-labeled TAR in a Tris binding buffer (50 mM Tris pH
7.4, 100 mM NaCl, 1 mM MgCl2) to 95 °C for 5 min, followed by
cooling on ice for 20 min. Correspondingly, the A-site RNA was
refolded by heating a 15 µM solution of A-site 2AP(1492) in the A-site
binding buffer (20 mM Hepes 7.5, 100 mM NaCl, 0.5 mM EDTA) to
65 °C for 5 min, followed by cooling at room temperature for 20 min.
Fluorescence Binding Assays. All fluorescence spectra were
measured on a Perkin-Elmer LS50B fluorimeter, with an excitation
slit width of 10 nm and an emission slit width of 10 nm. Upon excitation
at 345 or 310 nm for TAR 2′p(25) or A-site 2AP(1492), respectively,
the emission spectrum was recorded three times (scan rate of 400
nm/min) and averaged for a composite spectrum. As we have pre-
viously observed, ideal solution conditions for measuring binding to
TAR 2′p(25) include 100 mM NaCl and 1 mM MgCl2. Optimal solution
conditions for measuring binding to A-site 2AP(1492) include 100 mM
NaCl and 0.5 mM EDTA, consistent with literature reports.29,30 To
buffer the pH, the titrations were performed in the sulfonate buffers
Mes, Mops, or Hepes for pH values of 5.8, 6.8, or 7.5, respectively.
The amount of NaOH added to each to adjust to the appropriate pH
was recorded and found to give less than 5 mM difference in total
sodium ion concentration over the this pH range. Notably, at low pH
values the fluorescence of TAR 2′p(25) quenched dramatically upon
saturation with each aminoglycoside. Inclusion of 0.001% nonionic
detergent P40 (Fluka) in all TAR titrations alleviated this problem,
suggesting aggregative behavior in the presence of fully protonated
aminoglycosides.
For a typical binding experiment, the fluorescence spectrum of a
148 µL solution of buffer in the absence of any RNA or aminoglycoside
was recorded. This spectral blank, for which only Raman scatter was
observed, was subtracted from all subsequent spectra within each
binding experiment. Following determination of the buffer blank, 2
µL of a 15 µM solution of refolded RNA was added (final concentration
is 200 nM), the solution was mixed, and the spectrum was again
recorded. Subsequent aliquots of 1 µL of an aqueous aminoglycoside
solution (increasing concentrations) were added, and the fluorescence
spectrum was recorded after each aliquot until the fluorescence signal
saturated. Over the entire range of aminoglycoside concentrations, the
emission maxima of both fluorophores varied less than 1.5 nm. To
correct for the presence of any fluorescent contamination in each
aminoglycoside, the full titration was repeated in the absence of labeled
RNA. Any observed “background” at the emission maximum was then
subtracted from each corresponding point of the labeled RNA titrations,
and the resulting fluorescence intensity was corrected for dilution.
Based on definitive literature reports and direct measurements, a
two-state model for a 1:1 interaction was assumed for the A-site, and
the corrected fluorescence at each titration point, Fli, was converted to
a fractional fluorescence saturation (fsat):
[ag]free ) [ag]total - fsat*[RNA]tot
The subsequent plot of fsat against [ag]free was fit to the equation for a
simple binding isotherm:
([ag]free/Kd)
fsat
)
(1 - [ag]free/Kd)
For the TAR, the corrected fluorescence intensity value at each titration
point (Fli) was then normalized to the initial fluorescence of the pyrene-
derivatized TAR in the absence of aminoglycoside (Fl0) and plotted as
a function of the dilution-corrected aminoglycoside concentration as
shown in Figure 2. From this plot, an accurate binding affinity and
Hill coefficients for the interaction of each aminoglycoside with the
TAR were determined as described.35
DMS Footprint of Aminoglycosides on the A-Site. DMS foot-
printing experiments were performed essentially as described.38 Briefly,
2 µL of a 1:2 dilution of DMS in ethanol was added to a 75 nM solution
of an 3′-extended A-site RNA (includes additional 3′-end nucleotides:
5′-GGUUGGCGUGGCUCGCG-3′) in a total volume of 50 µL of a
buffer containing 80 mM sodium cacodylate pH 7.5, 50 mM NaCl,
0.4 mM EDTA, and the indicated concentration of aminoglycoside.
After 30 min at room temperature, the reaction was stopped by addition
of 12.5 µL of DMS stop buffer, resulting in the final additional
concentrations of 150 mM Tris pH 7.5 and 200 mM â-mercapto-ethanol.
Each reaction was then precipitated with the addition of 30 µg of
Glycoblue (Ambion, Austin, TX), 12 µL of 3 M NaOAc, and 320 µL
of ethanol. Sodium borohydride reduction and subsequent aniline
cleavage were performed as described,38 except that 30 µg of Glycoblue
was added to aid quantitative precipitation at each step.
After resuspending in water, ∼250 fmol of each RNA was combined
with ∼125 fmol of a 5′-[32P]-labeled DNA primer (5′-CGCGAGC-
CACGCCAACC-3′) in a buffer of 50 mM Tris pH 8.3, 60 mM NaCl,
and 10 mM DTT. This mixture was heated to 95 °C for 5 min, followed
by immediately cooling on dry ice/ethanol for 1 min. After briefly
centrifuging, Mg(OAc)2 in the same buffer was added to a final
concentration of 6 mM. While maintaining the same buffer conditions,
deoxy NTPs (625 µM final concentration, Promega) and 1 unit of
AMV-reverse transcriptase (Invitrogen) were added to the RNA/primer
mix, and the solution was incubated at 42 °C for 30 min. These reactions
were quenched by dilution into an equal volume of loading dye (88%
formamide with 0.02% bromophenol blue and xylene cyanol), followed
by immediate separation with 20% denaturing PAGE. To aid interpreta-
tion, sequencing reactions containing one dideoxyNTP were performed
as per manufacturer protocol.
RNase Footprint of Aminoglycosides on the TAR. A trace
concentration (<20 pM) 5′-[32P]-labeled, refolded TAR (95 °C 5 min,
ice 20 min) was incubated with 1.6 milliunits of ribonuclease V1
(Ambion) for 15 min at room temperature in a solution containing 30
mM Tris pH 7.5, 100 mM NaCl, 2 mM MgCl2, 0.1 µg/mL yeast torula
RNA (Ambion), and the indicated concentrations of aminoglycosides
(Figure 6). These reactions were quenched by dilution into an equal
volume of loading dye (88% formamide with 0.02% bromophenol blue
and xylene cyanol), followed by immediate separation with 20%
denaturing PAGE.
Ribonuclease T1 reactions contained a trace concentration (<20 pM)
of 5′-[32P]-labeled, refolded TAR and 1 unit of ribonuclease T1
(Boerrhinger Moenheim) in a solution of 20 mM sodium citrate pH
5.0, 1 mM EDTA, 3.63 M urea, and 0.05 µg/µl yeast torula RNA
(Fli - Fl0)
fsat
)
(Fl∞ - Fl0)
(51) Winzor, D. J.; Sawyer, W. H. QuantitatiVe Characterization of Ligand
Binding; Wiley-Liss, Inc.: New York, 1995.
where Fl0 and Fl∞ are the observed fluorescence in the absence of
9
9828 J. AM. CHEM. SOC. VOL. 127, NO. 27, 2005

RNA-Aminoglycoside Specificity
A R T I C L E S
(Ambion). These reactions were incubated at 55 °C for 20 min, followed
by dilution into an equal volume of loading dye and separation with
20% denaturing PAGE.
Alkaline Hydrolysis Reactions. A trace amount (<20 pM) of
5′-[32P]-labeled TAR was incubated in a buffer of 50 mM NaHCO3
pH 9.0 at 95 °C for 7 min, followed by quenching on ice for 10
min. After dilution with an equal volume of loading dye, the reactions
were separated on 20% PAGE. All polyacrylamide gels were run for
approximately 3 h at 20 W and analyzed using a Molecular Dynamics
Storm Phosphorimager.
experiments. Special thanks to DeLano Scientific (San Carlos,
CA) for use of the open-source PyMol molecular graphics
system. This work was generously supported by the National
Institutes of Health (Grant Number AI 47673 to Y.T., GM
065652 to K.B. and AI51104 to T.H.).
Supporting Information Available: Synthesis procedures,
characterization as well as 1D and 2D NMR spectra of key
intermediates and final compounds. This material is available
free of charge via the Internet at http://pubs.acs.org.
Acknowledgment. Dedicated to Professor Peter B. Dervan
on the occasion of his 60th birthday. We are grateful to Drs.
Qiang Cong and John Wright for their help with the NMR
JA050918W
9
J. AM. CHEM. SOC. VOL. 127, NO. 27, 2005 9829