RNA structure analysis at single nucleotide resolution by Selective 2′-Hydroxyl Acylation and Primer Extension (SHAPE)

Article abstract of DOI:10.1021/ja043822v

The reactivity of an RNA ribose hydroxyl is shown to be exquisitely sensitive to local nucleotide flexibility because a conformationally constrained adjacent 3′-phosphodiester inhibits formation of the deprotonated, nucleophilic oxyanion form of the 2′-hydroxyl group. Reaction with an appropriate electrophile, N-methylisatoic anhydride, to form a 2′-O-adduct thus can be used to monitor local structure at every nucleotide in an RNA. We develop a quantitative approach involving Selective 2′-Hydroxyl Acylation analyzed by Primer Extension (SHAPE) to map the structure of and to distinguish fine differences in structure for tRNA^Asp transcripts at single nucleotide resolution. Modest extensions of the SHAPE approach will allow RNA structure to be monitored comprehensively and at single nucleotide resolution for RNAs of arbitrary sequence and structural complexity and under diverse solution environments.

Full text of DOI:10.1021/ja043822v

Published on Web 03/04/2005
RNA Structure Analysis at Single Nucleotide Resolution by
Selective 2′-Hydroxyl Acylation and Primer Extension
(SHAPE)
Edward J. Merino, Kevin A. Wilkinson, Jennifer L. Coughlan, and Kevin M. Weeks*
Contribution from the Department of Chemistry, UniVersity of North Carolina,
Chapel Hill, North Carolina 27599-3290
Received October 11, 2004; E-mail: weeks@unc.edu
Abstract: The reactivity of an RNA ribose hydroxyl is shown to be exquisitely sensitive to local nucleotide
flexibility because a conformationally constrained adjacent 3′-phosphodiester inhibits formation of the
deprotonated, nucleophilic oxyanion form of the 2′-hydroxyl group. Reaction with an appropriate electrophile,
N-methylisatoic anhydride, to form a 2′-O-adduct thus can be used to monitor local structure at every
nucleotide in an RNA. We develop a quantitative approach involving Selective 2′-Hydroxyl Acylation analyzed
by Primer Extension (SHAPE) to map the structure of and to distinguish fine differences in structure for
tRNAAsp transcripts at single nucleotide resolution. Modest extensions of the SHAPE approach will allow
RNA structure to be monitored comprehensively and at single nucleotide resolution for RNAs of arbitrary
sequence and structural complexity and under diverse solution environments.
Introduction
is idiosyncratic, multiple reagents must be used to interrogate
all positions in an RNA, and it is difficult to ascribe a single,
unifying set of rules for the classes of structures that react versus
those that do not.
Most RNA molecules function correctly in biology only once
they fold back on themselves to form complex and difficult-
to-predict three-dimensional structures.1 In addition, the biology
of many RNAs requires that specific regions have local structural
flexibility or the ability to transition from one conformation to
another. Understanding the function of almost all RNAs thus
requires an accurate and, ideally, comprehensive knowledge of
the base-paired secondary structure and identification of nucleo-
tides involved in tertiary interactions.
Traditional chemical and enzymatic reagents2 and backbone-
based cleavage3 are useful for mapping RNA secondary struc-
ture, especially when multiple reagents are used in concert.
However, traditional approaches for monitoring base pairing and
local nucleotide flexibility suffer from two broad shortcomings.
(1) Reagents are leaky in the sense that the absolute magnitude
of discrimination between single-stranded and base-paired
regions can be small and also because helix termini and some
base pairs (especially G-U pairs) are often reactive.4 (2) The
information obtained by existing approaches is also sparse
because a given reagent typically reacts with only a subset of
the four RNA nucleotides2 or backbone sites.3 Because the
chemistry of each sparsely reactive structure-selective reagent
An alternate approach for mapping RNA structure exploits
the observation that the chemical reactivity of the 2′-ribose
position is strongly modulated by the adjacent 3′-phosphodiester
anion.5,6 Acylation of synthetic 2′-amine-substituted nucleotides
to form the 2′-amide product is strongly gated by the underlying
local nucleotide flexibility.5,7,8 Flexible nucleotides in RNA are
better able to reach a facilitated transition state in which the
3′-phosphodiester becomes appropriately positioned with the 2′-
amine to exert a catalytic effect.6 2′-Amine acylation thus
robustly detects essentially all base-paired RNA secondary
structures and many tertiary interactions when the sites of 2′-
adduct formation are mapped by primer extension.5,8
We have found that this 2′-ribose-based chemistry is signifi-
cantly more straightforward to implement than traditional
chemical or enzymatic approaches for monitoring RNA second-
ary and tertiary structure. However, introduction of an artificial
2′-amine group makes an experiment more complex, prevents
this chemistry from being used in vivo, and may perturb some
tertiary interactions involving the 2′-ribose position. We con-
sidered that the proximity of a 3′-phosphodiester anion might
modulate the reactivity of the 2′-hydroxyl group normally
present in RNA. Since every nucleotide has a 2′-hydroxyl,
structural information is, in principle, obtainable for every
(1) (a) Brion, P.; Westhof, E. Annu. ReV. Biophys. Biomol. Struct. 1997, 26,
113-137. (b) Gesteland, R. F.; Cech, T. R.; Atkins, J. F. The RNA World,
2 ed.; Cold Spring Harbor Laboratory Press: Cold Spring Harbor, 1999.
(c) Tinoco, I.; Bustamante, C. J. Mol. Biol. 1999, 293, 271-281.
(2) Ehresmann, C.; Baudin, F.; Mougel, M.; Romby, P.; Ebel, J.-P.; Ehresmann,
B. Nucleic Acids Res. 1987, 15, 9109-9128.
(3) (a) Soukup, G. A.; Breaker, R. R. RNA 1999, 5, 1308-1325. (b) Helm,
M.; Brule, H.; Degoul, F.; Cepanec, C.; Leroux, J. P.; Giege, R.; Florentz,
C. Nucleic Acids Res. 1998, 26, 1636-1643. (c) Wittberger, D.; Berens,
C.; Hammann, C.; Westhof, E.; Schroeder, R. J. Mol. Biol. 2000, 300,
339-352.
(4) 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, 7287-7292.
(5) Chamberlin, S. I.; Weeks, K. M. J. Am. Chem. Soc. 2000, 122, 216-224.
(6) Chamberlin, S. I.; Merino, E. J.; Weeks, K. M. Proc. Natl. Acad. Sci. U.S.A.
2002, 99, 14688-14693.
(7) (a) John, D. M.; Weeks, K. M. Chem. Biol. 2000, 7, 405-410. (b) John,
D. M.; Weeks, K. M. Biochemistry 2002, 41, 6866-6874. (c) John, D.
M.; Merino, E. J.; Weeks, K. M. J. Mol. Biol. 2004, 611-619.
(8) Chamberlin, S. I.; Weeks, K. M. Biochemistry 2003, 42, 901-909.
9
10.1021/ja043822v CCC: $30.25 © 2005 American Chemical Society
J. AM. CHEM. SOC. 2005, 127, 4223-4231
4223

A R T I C L E S
Merino et al.
Figure 1. Reaction of a ribose 2′-hydroxyl with NMIA via formation of an oxyanion nucleophile. Nucleotides constrained in the RNA C3′-endo and other
conformations are unreactive due to destabilization (red overlap symbol) of the oxyanion by the proximal 3′-phosphodiester.
position in an RNA. Formation of a bulky 2′-O-adduct could
then be detected as an impediment to cDNA synthesis by a
reverse transcriptase enzyme.5,8
In this work, we show that selective acylation of the ribose
2′-hydroxyl position by N-methylisatoic anhydride (NMIA) is
exquisitely sensitive to local nucleotide flexibility. We then use
selective 2′-hydroxyl acylation analyzed by primer extension
(SHAPE) to analyze comprehensively local structure in yeast
tRNAAsp transcripts at nucleotide resolution in a single series
of experiments.
Results
A 2′-Hydroxyl Reactive Reagent. Since prior work showed
that the nucleophilicity of a 2′-ribose substituent is modulated
by the proximal 3′-group,6 we initially monitored reactivity of
ATP and 3′-deoxy-ATP toward diverse reagents9,10 potentially
able to react with hydroxyl groups under mild conditions in
aqueous solution. The most promising reagent was N-methyl-
isatoic anhydride (NMIA), which liberates CO2 to form a
2-methylaminobenzoic acid nucleotide 2′-ester10 (Figure 1). The
2′-O-adduct is readily resolved as a low mobility band by gel
electrophoresis (Figure 2A). ATP reacts 2.5-fold more rapidly
than 3′-deoxy-ATP to form the monoadduct (Figure 2B) and,
once modified, reacts slowly to form a second adduct with
NMIA.
An electrophile capable of reacting with a hydroxyl group in
aqueous solution will face competition from the analogous
hydrolysis reaction (Figure 3A). To measure the rate of NMIA
Figure 2. Structure-selective reaction of NMIA with ribose hydroxyl
groups. (A) Visualizing nucleotide 2′-ester formation with ATP and 3′-
deoxy-ATP. NMIA-modified nucleotides migrate with reduced mobility by
hydrolysis, we monitored formation of the 2-methylaminoben-
zoate product by fluorescence. 2-Methylaminobenzoate was
excited selectively at 375 nm and emission detected at 440 nm
(Figure 3B). NMIA hydrolysis shows good first-order kinetics
with a rate constant of 0.031 min-1 (at 25 °C, Figure 3C). Given
this rate constant for reagent hydrolysis, we are able to calculate
rates for ribose 2′-O-adduct formation explicitly.
polyacrylamide gel electrophoresis. (B) Kinetic analysis of nucleotide adduct
formation; lines represent least-squares fits to eq 1.
of NMIA requires formation of the significantly more nucleo-
philic 2′-oxyanion prior to reaching the ester-forming transition
state (see first step in Figure 1). Formation of the 2′-oxyanion
would be destabilized if the negatively charged 3′-phosphodi-
ester were constrained to be adjacent to the 2′-position (see red
overlap symbols in Figure 1). Because formation of the oxyanion
involves loss of a proton, the hydroxyl pKa is a good indicator
for the relative nucleophilicity of a 2′-hydroxyl. The pKa for
deprotonation of the 2′-hydroxyl of adenosine is 12.2, and this
pKa increases to 13.4 for adenosine-3′-phosphate.11 The relative
nucleophilicity of a 2′-hydroxyl is thus strongly modulated by
NMIA Reactivity with Nucleotides is Modulated by the
3′-Substituent. Attack of a 2′-hydroxyl at the anhydride moiety
(9) (a) de Belder, A. N.; Wik, K. O. Carbohydr. Res. 1975, 44, 251-257. (b)
Stearns, N. A.; Prigent-Richard, S.; Letourneur, D.; Castellot, J. J. Anal.
Biochem. 1997, 247, 348-356. (c) Cardinaud, R.; Baker, B. R. J. Med.
Chem. 1970, 13, 467-470.
(10) (a) Moorman, A. R.; Abeles, R. H. J. Am. Chem. Soc. 1982, 104, 6785-
6786. (b) Hiratsuka, T. Biochim. Biophys. Acta 1983, 742, 496-508.
9
4224 J. AM. CHEM. SOC. VOL. 127, NO. 12, 2005

RNA Structure Analysis at Single Nucleotide Resolution by SHAPE
A R T I C L E S
Figure 3. NMIA hydrolysis. (A) Reaction of NMIA with water to form the fluorescent 2-methylaminobenzoate. (B) Formation of the 2-methylaminobenzoate
product is accompanied by a shift in absorbance maximum from 320 to 335 nm; product fluorescence was thus detected by exciting at 375 nm (heavy line)
to eliminate inner filter effects. (C) Pseudo-first-order hydrolysis of NMIA (25 °C); every tenth data point is shown.
electrophoresis (Figure 2A). Rates were then obtained by fitting
complete reaction profiles (Figure 2B). For the remaining
analogues, product formation was determined as the fraction
2′-O-adduct formation after complete NMIA hydrolysis, using
HPLC (Figure 4; see Experimental Section). Rates were obtained
by both methods for ATP and 2′-deoxy-ATP and agreed within
error.
All analogues contained either a 5′-mono- or triphosphate to
block reaction at the 5′-hydroxyl. The kadduct values for ATP
and AMP are similar at 0.015 and 0.022 min-1, respectively,
indicating that reaction of the 2′-hydroxyl with NMIA is not
significantly modulated by the number of phosphate moieties
at the 5′-position (Figure 5E). When 2′,3′-dideoxy-ATP (Figure
5F) was treated with NMIA, no new species were detected by
HPLC, confirming that adduct formation requires a free ribose
hydroxyl group.
We evaluated the reactivity of two reference analogues, pAp-
ddC and pAp-ethyl (Figure 5A), designed to mimic the
2′-hydroxyl reactivity of unstructured regions in larger RNAs.
Both analogues possess an authentic 3′-phosphodiester group
and a single reactive 2′-hydroxyl. The pAp-ddC and pAp-ethyl
react with kadduct values of 0.0010 and 0.0018 min-1, respec-
tively. The slightly faster reactivity of the pAp-ethyl analogue
Figure 4. End point analysis to determine rate of reaction of ATP and
cAMP with NMIA. Reaction endpoints were quantified by HPLC. Fraction
shows that simple steric effects make, at most, a small
contribution to 2′-hydroxyl reactivity at flexible nucleotides.
ATP or cAMP product was obtained by integration of the indicated peaks.
Assignment of nucleotide starting material, products, and buffer components
is straightforward because all components absorb at 260 nm, whereas only
2-methylaminobenzoate products are detected at 360 nm.
We then evaluated a series of nucleotide analogues designed
to assess the impact of a 3′-phosphodiester on the reactivity of
the adjacent 2′-hydroxyl (Figure 5B). When the 3′-phosphodi-
ester in pAp-ddC or pAp-ethyl is replaced with either a
3′-methoxy (3′-O-methyl-ATP) or a 3′-hydrogen (3′-deoxy-
ATP), kadduct increases by 3- or 6-fold, respectively. Analogously,
constraining the 3′-phosphodiester away from the 2′-hydroxyl
in a 3′,5′-cyclic linkage (cAMP, Figure 5C) yields an analogue
that reacts at a rate of 0.018 min-1, 10-18-fold faster than that
of the RNA-like analogues, pAp-ddC and pAp-ethyl.
neighboring functional groups at the 3′-position. Formation of
the oxyanion also presumably induces secondary changes in
ribose conformation to yield complex relationships between
nucleotide structure and the 2′-hydroxyl pKa. We, therefore,
explored reactivity for an instructive set of nucleotides (Figure
5) whose functional group or conformation at the 3′-position
had the potential to inhibit or facilitate 2′-ester formation with
NMIA (see Figure 1).
Two methods were used to obtain pseudo-first-order rate
constants for adduct formation (kadduct) for reaction of the
nucleotide analogues with NMIA. Most analogues could be
radiolabeled and were resolved from their reaction products by
Finally, the inhibitory effect of a proximal charged phosphate
moiety is obvious from comparing reactivities for 3′- versus
2′-deoxy-ATP (Figure 5B,D). The 5′-triphosphate is closer in
distance to the 3′-hydroxyl than the 2′-hydroxyl, and cor-
respondingly, 2′-deoxy-ATP reacts 12-fold slower than 3′-
deoxy-ATP (kadduct is 0.0005 versus 0.0060 min-1). We infer
(11) Velikyan, I.; Acharya, S.; Trifonova, A.; Fo¨ldesi, A.; Chattopadhyaya, J.
J. Am. Chem. Soc. 2001, 123, 2893-2894.
9
J. AM. CHEM. SOC. VOL. 127, NO. 12, 2005 4225

A R T I C L E S
Merino et al.
Figure 6. 2′-Hydroxyl acylation by NMIA is selective for unpaired
nucleotides. (A) 2′-Adduct formation with the oligonucleotide (5′-UAC-
UAACACC-ddC) and the corresponding duplex visualized by gel electro-
phoresis. Site of [32P]-label is indicated with an asterisk. (B) Quantitative
analysis of 2′-O-adduct formation averaged over the 10 hydroxyl groups
in the labeled oligonucleotide. Rates for the single strand (triangles) and
duplex (squares) were obtained by fitting to eq 1.
was treated with NMIA, a slowly migrating adduct formed,
whose overall synthesis occurs at a rate of 0.0048 min-1 (left
panel in Figure 6A). Since this oligonucleotide contains 10
hydroxyl groups, each reacts, on average, at a rate of (0.0048/
10) or ∼0.0005 min-1 (triangles in Figure 6B), or within 2-fold
of the rate for the pAp-ddC reference dinucleotide (0.0010
min-1, Figure 5A). The slightly slower average rate may reflect
stronger stacking and other constraining interactions in the
longer oligonucleotide.
Figure 5. Structure-reactivity analysis for reaction of nucleotide analogues
with NMIA. For each nucleotide or analogues series (A-F), kadduct is
reported in min-1 at 13 mM NMIA and 25 °C. Rates were obtained from
complete profiles visualized by gel electrophoresis (e) or by single time
point analysis using HPLC (h); a subset of nucleotides was evaluated by
both methods.
When the 11-mer oligonucleotide was hybridized to its
complement, the rate of adduct formation decreased by 10-fold
to 0.0005 min-1 (compare triangles and squares in Figure 6B).
Residues at RNA helix termini experience fraying12 and are
more flexible than internal nucleotides. Thus, the calculated
that the 5′-triphosphate more strongly destabilizes oxyanion
formation at the 3′-position than at the more distant 2′-hydroxyl
(see 2′-deoxy-ATP in Figure 5D).
In sum, these data support the model (Figure 1) that the
reactivity of a ribose hydroxyl group is decreased by the
proximity of any phosphate group. Reactivity can be recovered
either by replacing a phosphate diester with an uncharged
analogue or by increasing the distance between the hydroxyl
and phosphate groups.
Structure-Selective 2′-Hydroxyl Reactivity in an Oligo-
nucleotide. We extended these studies on the relationship
between constrained local structure and 2′-hydroxyl reactivity
by evaluating reactivity in an 11-mer oligonucleotide (Figure
6). This RNA was synthesized with a 5′-monophosphate
radiolabel and a 3′-dideoxycytidine nucleotide so that all 10
remaining hydroxyl groups lie in internal linkages and adjacent
to 3′-phosphodiester groups. When the 5′-labeled oligonucleotide
average rate at a single position (0.0005/10 or 5 × 10-5 min-1
)
must represent an upper limit for reaction of hydroxyl groups
constrained by base pairing.
The differential reactivity for single-stranded versus helical
oligonucleotides (Figure 6) supports two important conclusions.
First, the average reactivity at an individual site in the 11-mer
oligonucleotide is within a factor of 2 for that of the pAp-ddC
reference dinucleotide. The similar rate supports the interpreta-
tion that the analogue studies reproduce key mechanistic features
of 2′-O-adduct-forming reactions at flexible sites in large RNAs.
Second, because ribose 2′-hydroxyl groups project outward from
an RNA duplex, the differential reactivity of single-stranded
(12) (a) Preisler, R. S.; Mandal, C.; Englander, S. W.; Kallenbach, N. R.; Frazier,
J.; Miles, H. T.; Howard, F. B. Biopolymers 1984, 23, 2099-2125. (b)
Snoussi, K.; Leroy, J.-L. Biochemistry 2001, 40, 8898-8904.
9
4226 J. AM. CHEM. SOC. VOL. 127, NO. 12, 2005

RNA Structure Analysis at Single Nucleotide Resolution by SHAPE
A R T I C L E S
versus duplex RNA (Figure 6B) rules out solvent accessibility
as the primary explanation for structure-sensitive 2′-hydroxyl
reactivity.
A Structure Cassette for Analyzing tRNAAsp Conforma-
tion. We apply selective 2′-hydroxyl acylation analyzed by
primer extension (SHAPE) to explore the structure of Saccha-
romyces cereVisiae tRNAAsp transcripts. The well-studied L-
shaped tRNAAsp structure5,13,14 (Figure 7A) encompasses inter-
actions that comprise any folded RNA, including double-
stranded helices, single-stranded loops and connecting regions,
and a network of tertiary interactions. The secondary structure
shown (Figure 7A) emphasizes the three-dimensional fold of
the RNA in which (i) the anticodon and D-stems and (ii) the
acceptor and T-stems stack. These two pairs of coaxially stacked
helices are then held together in the tRNA architecture by
tertiary interactions between the D- and T-loops and involving
the joining regions at positions U8-A9 and G45-U48.
The tRNAAsp transcript was placed in the center of an RNA
“structure cassette” to facilitate analysis of 2′-O-adduct forma-
tion by primer extension (Figure 7B). Each stem-loop structure
in the cassette contains a stable UUCG tetraloop15 to enforce
the designed fold and eliminate alternate foldings with tRNAAsp
.
The hairpins that comprise the flanking 5′ and 3′ cassette
structures have three functions. (1) The 3′-most sequence (green
in Figure 7B) provides the DNA primer binding site required
to initiate reverse transcription. (2) The 3′-linker (blue) between
the primer binding site and the first position in the tRNA allows
the reverse transcriptase enzyme to become fully processive
prior to reaching the region of structural interest. This linker
also prevents nontemplated primer extension products from
masking structural information at the 3′-end of the RNA. (3)
The 5′-portion of the cassette (red in Figure 7B) displaces the
abundant full length extension product that would otherwise
obscure structural information at the 5′-end of the RNA. Overall,
the design of the structure cassette balances the requirements
that it fold autonomously, not interact with the internal RNA,
and, yet, efficiently bind a primer DNA.
SHAPE Footprinting of tRNAAsp. The tRNAAsp construct
was folded under strongly native conditions (100 mM NaCl,
10 mM MgCl2)5,13 at 35 °C. NMIA was added and allowed to
react with the RNA for five hydrolysis half-lives (54 min). Sites
of 2′-esterification (Figure 1) were identified as stops to primer
extension by reverse transcriptase. The presence of a 2′-O-adduct
causes the reverse transcriptase to stop exactly one nucleotide
prior to the modified base.
Primer extension products were resolved in a sequencing gel
and converted to intensity versus position plots (Figure 8A).
When tRNAAsp was incubated under the standard reaction
conditions in the absence of NMIA, primer extension yields a
strong full-length product and only minor bands of intermediate
length (green trace, Figure 8A). When an otherwise identical
SHAPE experiment was performed in the presence of NMIA,
a strong and reproducible pattern of 2′-O-adducts is detected
Figure 7. The tRNAAsp transcript and structure cassette. (A) Secondary
structure of tRNAAsp. Structure is drawn to suggest arrangement of RNA
helices in three-dimensional space. Tertiary interactions are emphasized with
dashed lines. The RNA transcript does not contain post-transcriptional
modifications, and the native tRNAAsp U1-A72 base pair was changed to
a G-C pair to permit comparison with prior work.5 By convention,14
nucleotides in yeast tRNAAsp are numbered 1-76, with number 47 omitted.
(B) The RNA structure cassette can be appended to any RNA of interest
by transcription from a PCR-generated template.
by primer extension (compare black (+NMIA reaction) with
green (-NMIA) traces in Figure 8A). The fraction full-length
extension product is 70% of that for the no reaction control,
indicating that NMIA reacts with single-hit kinetics.16 Sites of
adduct formation were identified by comparison with a dideoxy
sequencing lane (blue trace, Figure 8A).
(13) (a) Perret, V.; Garcia, A.; Puglisi, J.; Grosjean, H.; Ebel, J. P.; Florentz,
C.; Giege, R. Biochimie 1990, 72, 735-743. (b) Romby, P.; Moras, D.;
Dumas, P.; Ebel, J. P.; Giege, R. J. Mol. Biol. 1987, 195, 193-204.
(14) (a) Westhof, E.; Dumas, P.; Moras, D. J. Mol. Biol. 1985, 184, 119-145.
(b) Westhof, E.; Dumas, P.; Moras, D. Acta Crystallogr. A 1988, 44, 112-
123.
(15) (a) Tuerk, C.; Gauss, P.; Thermes, C.; Groebe, D. R.; Gayle, M.; Guild,
N.; Stormo, G.; d’Aubenton-Carafa, Y.; Uhlenbeck, O. C.; Tinoco, I.;
Brody, E. N.; Gold. L. Proc. Natl. Acad. Sci. U.S.A. 1988, 85, 1364-
1368. (b) Cheong, C.; Varani, G.; Tinoco, I. Nature 1990, 346, 680-682.
(16) Brenowitz, M.; Senear, D. F.; Shea, M. A.; Ackers, G. K. Methods Enzymol.
1986, 130, 132-181.
9
J. AM. CHEM. SOC. VOL. 127, NO. 12, 2005 4227

A R T I C L E S
Merino et al.
Figure 8. SHAPE analysis of tRNAAsp. (A) Profile for 2′-O-adduct formation as a function of nucleotide position, visualized by primer extension. Reactions
performed in the presence and absence of NMIA are shown in black and green, respectively. Regions that react selectively with or are excluded from
reaction with NMIA are emphasized with red or green bars, respectively. A dideoxy sequencing marker is shown in blue and is one nucleotide longer than
corresponding positions in the (+) and (-) NMIA lanes. (B) Absolute 2′-hydroxyl reactivities superimposed on a secondary structure for the tRNAAsp
construct. Column heights are proportional to the band intensities shown in part A (black trace) minus background.
Regions that showed the strongest reactivity relative to
background are emphasized with red bars in Figure 8A, and all
reactivities are superimposed on a secondary structure for the
RNA transcript in Figure 8B. Comparing the most reactive
nucleotide, U33 in the anticodon loop, with unreactive residues
in, for example, the T- and acceptor stems indicates that the
dynamic range of the SHAPE experiment spans 30-50-fold.
2′-Hydroxyl reactivity observed in the 5′- and 3′-linkers is
exactly that expected for proper folding of the cassette,
indicating the structure cassette folds as designed (Figure 8B).
in Figure 8B). Tertiary interactions, including interaction of G45
with the G10-U25 pair, the A15-U48 Levitt pair, the A14-
A21-U8 base triple, and the U54-A58 intraloop interaction,
are unreactive.
2′-Ribose-based NMIA chemistry is clearly generic to all
RNA nucleotides. Each nucleotide (A, G, C, and U) shows
strong reactivity in at least one part of the tRNAAsp transcript,
and all four residues show protection at other positions when
constrained by base pairing or a tertiary interaction (Figure 8B,
see Figure 7A for nucleotide assignments).
When superimposed on the secondary structure for tRNAAsp
and the flanking structure cassette, absolute 2′-hydroxyl reac-
tivities correspond closely to whether a given nucleotide is
constrained. Nucleotides that do not participate in either base
pairing or tertiary interaction are most reactive (red and orange
columns in Figure 8B). Conversely, NMIA modification is
significantly reduced in each of the structures that comprise the
four canonical tRNA helices (green, blue, and black columns
Discussion
2′-Hydroxyl Acylation with NMIA Scores Local Nucleo-
tide Flexibility. RNA SHAPE footprinting reproduces the
canonical structure of tRNAAsp. Reactive nucleotides lie almost
exclusively in the conserved T-, D-, and anticodon loops and
at the 3′-CCA terminus (red and orange columns in Figure 8B).
Of the five noncanonical G-U or G-A pairs in tRNAAsp, only
9
4228 J. AM. CHEM. SOC. VOL. 127, NO. 12, 2005

RNA Structure Analysis at Single Nucleotide Resolution by SHAPE
A R T I C L E S
Figure 9. RNA SHAPE reactivity is not correlated with solvent accessibility. Relative 2′-O-adduct formation (gray bars) was calculated from the data
shown in Figure 8A. Solvent accessibilities (heavy lines) were computed19 assuming a probe radius of 1.4 Å. For comparison, results from chemical mapping
studies on tRNAAsp transcripts using DMS and DEPC13a are shown with spheres; reactive and unreactive positions are emphasized with filled and open
spheres, respectively.
one, in the anticodon stem, is reactive. SHAPE footprinting thus
appears to be more sensitive to constraints imposed by nonca-
nonical pairing than traditional chemistries because G-U and
G-A pairs are typically reactive toward DMS and CMCT.4,13
NMIA reactivity also detects tertiary interactions. Essentially,
all nucleotides that form higher-order interactions (dashed lines
in Figure 8B) are unreactive at 35 °C. This is a striking result
given that the unreactive structures are constrained by diverse
local tertiary interactions, including Hoogsteen paring, base
triples, and cross-loop interactions.
solvent accessible but unreactive (Figure 9). 2′-Hydroxyl
reactivity is thus uncorrelated with static solvent accessibility,
strongly supporting models (Figure 1) that emphasize local
conformational changes as the primary determinant of 2′-
hydroxyl reactivity.
We also compare SHAPE with chemical mapping experi-
ments13a employing the traditional base-reactive reagents, di-
methyl sulfate (DMS) and diethylpyrocarbonate (DEPC). The
correlation between nucleotides reactive by SHAPE and with
DMS and DEPC is very strong (compare gray bars and filled
circles in Figure 9). This is a striking conclusion considering
that DMS and DEPC react largely at nucleophilic positions in
purine bases,2 while NMIA reacts at the RNA backbone and,
again, strongly supports the model (Figure 1) that local
nucleotide flexibility is the primary determinant for reactivity
by SHAPE chemistry. The major difference occurs at position
68, where the guanosine residue in a G-U pair is reactive
toward DMS, but not with the more selective SHAPE chemistry
(Figure 9).
The direct comparison between traditional and SHAPE
chemistries also emphasizes important advantages of SHAPE.
First, all nucleotides are interrogated in a single reaction so there
are no or fewer regions lacking structural information. Second,
instead of simply classifying positions as reactive or not, SHAPE
lends itself naturally to quantitative analysis (Figures 8B and
9).
2′-Ribose Chemistry, Local Nucleotide Flexibility, and the
Influence of the 3′-Phosphodiester Anion. Reactions of 2′-
hydroxyl groups (Figure 10A) and of 2′-amine substitutions
(Figure 10B) in RNA to form ester or amide linkages,
respectively, are both strongly sensitive to local nucleotide
flexibility (Figures 5 and 8B and ref 6). We postulate that the
physical basis for this effect in both cases stems predominantly
from the influence of the proximal 3′-phosphodiester anion on
reactivity at the 2′-position. The contribution of the 3′-
phosphodiester potentially includes direct charge-charge elec-
trostatic effects, differential hydrogen bonding, changes to the
pKa of the 2′-group, and generalized electrostatic and dielectric
effects in the local environment.
Finally, the SHAPE approach is also uniquely information-
rich in the sense that the absolute reactivity amplitudes appear
to be meaningful. First, nucleotides in the T- and D-loops, while
generally reactive, are less so than those in the anticodon loops,
consistent with formation of significant tertiary interactions in
the former loops. Within the T-loop, the cross-loop U54-A58
interaction is reported by low reactivities in an otherwise reactive
region. Second, nucleotides in the C-G-rich T- and acceptor
stems show reactivities that are measurably lower than those
of many nucleotides in the D- and anticodon stems (Figure 8B).
Third, inspection of the local patterns of reactivity within each
of the loops in tRNAAsp, at the 3′-CCA region, and in the 3′-
cassette (the +10 - +14 loop) shows a reproducible trend in
which the center-most loop nucleotides are more reactive than
nucleotides that abut constrained positions in the adjacent
duplexes. These data together emphasize that SHAPE mapping
reveals significant fine gradations in local nucleotide flexibility.
NMIA Modification Does Not Score Solvent Accessibility.
We compared the relative reactivities obtained by SHAPE with
calculated molecular surface areas at the ribose 2′-position,
assuming a solvent-sized probe and using the crystallographic
coordinates14 for tRNAAsp (Figure 9). Consistent with the results
obtained with the 11-mer oligonucleotide (Figure 6), 2′-O-adduct
formation in tRNAAsp does not correlate with solvent acces-
sibility. For example, the entire anticodon stem and loop
structures, spanning nucleotides 26-44, have roughly uniform
high solvent accessibility (g20 Å2 per ribose 2′-O atom) but
feature both unreactive and highly reactive positions (compare
gray bars with black line in Figure 9). Analogously, both the
D- and T-loops include residues with low calculated static
solvent accessibility but high 2′-hydroxyl reactivity; conversely,
the acceptor and T-stems (positions 49-53 and 60-72) are
However, the detailed reaction pathways for the 2′-ester-
versus the 2′-amide-forming reactions are quite different.
Reaction of a 2′-hydroxyl to form a 2′-ester proceeds via
9
J. AM. CHEM. SOC. VOL. 127, NO. 12, 2005 4229

A R T I C L E S
Merino et al.
adjacent 3′-phosphodiester anion (Figures 1, 5, and 10). SHAPE
footprinting affords a comprehensive and nucleotide resolution
view of base pairing and tertiary interactions in an RNA that is
generic to all four nucleotides (Figure 8). SHAPE reactivity
amplitudes also measure quantitatively subtle differences in the
degree to which a given nucleotide is constrained. RNA structure
mapping chemistries based on reaction at the 2′-hydroxyl will
likely make possible facile nucleotide resolution analysis of
RNA as a function of ion environment, temperature, ligand
binding, and conformational switches.
Experimental Section
General. Nucleotides (see Figure 5) ATP, AMP, 2′-deoxy-ATP, 3′-
5′-cyclic-adenosine monophosphate (cAMP), and adenosine 3′-mono-
phosphate were obtained from Sigma-Aldrich (St. Louis, MO); 2′,3′-
dideoxy-ATP and 3′-(O-methyl)-ATP were from Trilink Biotech (San
Diego, CA); 3′-(2′,3′-dideoxycytidine)-adenosine 5′-monophosphate and
the 5′-UACUAACACC-ddC oligo and its complement (Figure 6) were
synthesized by Dharmacon (Boulder, CO), and adenosine-3′-(O-ethyl)-
phosphate was the gift of J. Chattopadhyaya (University of Uppsala).
Radiolabeled (R-32P) ATP and 2′- and 3′-deoxy-ATP were from
Amersham. The pAp-ethyl and pAp-ddC were synthesized by 5′-[32P]-
labeling the appropriate precursor using polynucleotide kinase and were
purified by polyacrylamide gel electrophoresis.6
Nucleotide Analogue and Model Duplex Reaction Kinetics.
Nucleotide analogue reactions were performed in 100 mM potassium
phosphate (pH 8.0), 250 mM NaCl, supplemented with 10% (v/v)
NMIA (13 mM final, Molecular Probes) in dimethyl sulfoxide at 25
°C. For [32P]-labeled nucleotides, reaction aliquots (10 000 cpm/µL)
were quenched by addition to 1 vol 250 mM dithiothreitol and 2 vol
formamide. Reaction products were resolved by gel electrophoresis
(30% polyacrylamide; 29:1 acrylamide:bisacrylamide; 0.4 mm × 28.5
cm × 23 cm; 30 W; 1 h) and visualized by phosphorimaging (Molecular
Dynamics). Reaction rates were obtained using an equation that
Figure 10. Distinct contributions of the 3′-phosphodiester anion for reaction
of a flexible (gray arrows) ribose 2′-hydroxyl versus 2′-amine. Reaction
pathways for (A) 2′-hydroxyl and (B) 2′-amine acylation. (C and D) Relative
reactivities of cAMP analogues. Reference dinucleotides were pA2′-OHp-
6
ddC (Figure 5A) and pU2′-NH pU, respectively.
2
formation of an oxyanion in a pre-equilibrium step (Figure 1)
and yields negative charge on the nucleophile in the transition
state (Figure 10A). In contrast, 2′-amide formation involves
direct attack of the unperturbed, 2′-NH2 form6,17 of the nucleo-
phile and a transition state in which there is a partial positive
charge on the amine nucleophile (Figure 10B).
accounts for parallel reaction of NMIA by 2′-O-adduct formation (kadduct
)
and by hydrolysis (khydrolysis):
kadduct
hydrolysis
fraction product ) 1 - exp
(e-k
t - 1)
(1)
{
}
A proximal 3′-phosphodiester anion thus destabilizes forma-
tion of the 2′-oxyanion formed in a pre-equilibrium step (see
interference symbol in Figure 10A) but stabilizes the amine
nucleophile (dashed green line in the transition state in Figure
10B). These contrasting effects are supported directly by
nucleotide analogue studies (Figure 5 and ref 6) and especially
by the cAMP analogues, which constrain the 3′-phosphodiester
anion to be distant from a 2′-hydroxyl or 2′-amine nucleophile.
Because the 3′-phosphodiester group electrostatically destabilizes
the 2′-oxyanion, cAMP reacts (18-fold) faster than a reference
dinucleotide (Figure 10C). In contrast, reaction of the 2′-amine,
whose transition state is stabilized by the proximal anion,
becomes undetectable (at least 30-fold slower) for 2′-amino-
cAMP (Figure 10D). Flexible nucleotides sample a broad
ensemble of conformations (gray arrows in Figure 10), distinct
constellations are able to both minimize the electrostatic
influence of the 3′-phosphodiester and facilitate 2′-oxyanion
formation or, alternately, to align the 3′-group and facilitate the
amide-forming transition state.
khydrolysis
Nonradiolabeled analogues (10 µL, 1 mM in nucleotide) were allowed
to react for five half-lives, lyophilized to dryness, resuspended in
600 µL of gradient buffer [100 mM potassium phosphate (pH 6)], and
resolved by HPLC (Waters Xterra C-18 column, gradient to 50%
(v/v) methanol) over 40 min. Nucleotides typically eluted within the
first 15 min, while 2′-O-adducts eluted at ∼35 min. In the limit of
complete NMIA hydrolysis, eq 1 simplifies to kadduct/khydrolysis
)
-ln(1 - fraction product). Thus, the absolute rate of 2′-O-adduct
formation is uniquely determined by the fraction product at the reaction
endpoint.
Reaction of the 5′-[32P]-labeled 11-mer oligonucleotide (0.55 µM,
in 90 µL) was followed for both the single strand and duplex
(complement concentration ) 14 µM). RNAs were heated to 90 °C
and allowed to cool to ambient temperature over 20 min. Reactions
were initiated by addition of 10 µL NMIA (130 mM in DMSO), and
aliquots were quenched by addition of an equal volume of formamide,
100 mM DTT, and 40 µM unlabeled 11-mer (to liberate the radiolabeled
oligo from the duplex). Reaction aliquots were resolved on a (30%)
denaturing polyacrylamide gel.
Perspective. SHAPE mapping takes advantage of the strong
linkage between reactivity at the 2′-hydroxyl position and
destabilization of nucleophilic chemistry at this site by the
NMIA Hydrolysis. Reactions were initiated by adding 1 µL of 10
mM NMIA to reaction buffer (2 mL) in a fluorescence cuvette. Pseudo-
first-order formation of the hydrolysis product, 2-methylaminobenzoate,
was monitored by fluorescence (excitation 375 nm, emission 440 nm,
15 s time points).
(17) Verheyden, J. P. H.; Wagner, D.; Moffatt, J. G. J. Org. Chem. 1971, 36,
250-254.
9
4230 J. AM. CHEM. SOC. VOL. 127, NO. 12, 2005

RNA Structure Analysis at Single Nucleotide Resolution by SHAPE
A R T I C L E S
Synthesis of tRNAAsp Construct. A DNA template for transcription
of tRNAAsp in the context of the structure cassette was generated by
PCR [1.5 mL; containing 20 mM Tris (pH 8.4), 50 mM KCl, 2.5 mM
MgCl2, 200 µM each dNTP, 500 nM each forward and reverse primer,
5 pM template, and 0.025 units/µL of Taq polymerase; denaturation at
94 °C, 45 s; annealing 55 °C, 30 s; and elongation 72 °C, 90 s; 34
cycles]. Primer and tRNAAsp coding sequences are indicated in Figure
7. The PCR product was recovered by ethanol precipitation and
resuspended in 300 µL of HE [10 mM Hepes (pH 8.0), 1 mM EDTA].
Transcription reactions (1.5 mL, 37 °C, 3 h) contained 40 mM Tris
(pH 8.0), 10 mM MgCl2, 10 mM DTT, 2 mM spermidine, 0.01% (v/v)
Triton X-100, 4% (w/v) poly(ethylene) glycol 8000, 2 mM each NTP,
75 µL of PCR-generated template, and 0.1 mg/mL of T7 RNA
polymerase. The RNA product was purified by denaturing (8%)
polyacrylamide gel electrophoresis, excised from the gel, and recovered
by electroelution and ethanol precipitation. The purified RNA (∼2.5
µmol) was resuspended in 100 µL of HE.
Structure-Sensitive RNA Modification. RNA (20 pmol) in 6 µL
of sterile water was heated at 95 °C for 3 min, quickly cooled on ice,
treated with 3 µL of folding buffer [333 mM NaCl, 333 mM Hepes
(pH 8.0), 33.3 mM MgCl2], and incubated at 37 °C for 20 min. The
RNA solution was treated with NMIA (1 µL, 130 mM in anhydrous
DMSO, 35 °C), allowed to react for 54 min (equal to five NMIA half-
lives), and placed on ice. Control reactions contained 1 µL of DMSO
in place of NMIA. In prequench reactions, the RNA was added after
NMIA was allowed to degrade by hydrolysis. Modified RNAs were
subjected to primer extension without further purification.
buffer [6 µL; 167 mM Tris (pH 8.3), 250 mM KCl, 10 mM MgCl2,
1.67 mM each dNTP, 16.7 mM DTT] was added; the RNA was heated
to 52 °C; Superscript III (Invitrogen, 1 µL, 200 units) was added and
mixed by gentle pipetting, and reactions were incubated at 52 °C for
5 min. Primer extension reactions were quenched by addition of 1 µL
of 4 M NaOH, heating (95 °C for 5 min), subsequent addition of
neutralizing gel loading solution (29 µL; 40 mM Tris/borate, 276 mM
Tris/HCl, 5 mM EDTA, 0.01% (w/v) each xylene cyanol and bro-
mophenol blue, 73% (v/v) formamide), and heating (95 °C, for an
additional 5 min). Dideoxy sequencing markers were generated using
unmodified RNA and by adding 1 µL of 10 mM dideoxy nucleotide
triphosphate after addition of reverse transcription buffer. The cDNA
extension products were separated by denaturing electrophoresis [90
mM Tris/borate, 2 mM EDTA, 7 M urea, 8% (29:1) acrylamide:
bisacrylamide); 100 W, 5 min; 72 W, 2.5 h; 0.75 mm × 31 cm × 38.5
cm] and visualized by phosphorimaging. Two-dimensional gel images
were converted to plots of intensity versus position using ImageQuant.
Individual band intensities for the (+) and (-) NMIA reactions were
integrated using SAFA.18
Acknowledgment. The work was supported by an NSF
Career Award (MCB-9984289) to K.M.W., and we are indebted
to the authors of SAFA for sharing this program prior to
publication.
JA043822V
Primer Extension. Modified RNA (2 µL, 4 pmol) was added to
5′-[32P]-radiolabeled reverse transcription DNA primer (5′-32P-GAA
CCG GAC CGA AGC CCG, ∼0.2 pmol in 11 µL of H2O), heated to
65 °C (6 min), and incubated at 35 °C (20 min). Reverse transcription
(18) Das, R.; Laederach, A.; Pearlman, S.; Herschlag, D.; Altman, R. B.
SAFA: Semi-automated Footprinting Analysis Software for High-
Throughput Quantification of Nucleic Acid Footprinting Experiments. RNA
2005, 11, 344-354.
(19) Lee, B.; Richards, F. M. J. Mol. Biol. 1971, 55, 379-400.
9
J. AM. CHEM. SOC. VOL. 127, NO. 12, 2005 4231