J. Mol. Biol.

(1997) 269, 52±66

The Structure of an RNA ``Kissing'' Hairpin Complex

of the HIV TAR Hairpin Loop and its Complement
Kung-Yao Chang and Ignacio Tinoco Jr*

Department of Chemistry We have used nuclear magnetic resonance (NMR) to obtain the structure
University of California at of an RNA ``kissing'' hairpin complex formed between the HIV-2 TAR
Berkeley and Structural Biology hairpin loop and a hairpin with a complementary loop sequence. Kissing
Division, Lawrence Berkeley hairpins are important in natural antisense reactions; their complex is a
National Laboratory, Berkeley speci®c target for protein binding. The complex has all six nucleotides of
CA, 94720-1460, USA each loop paired to form a bent quasicontinuous helix of three coaxially
stacked helices: two stems plus a loop-loop interaction helix. Experimen-
tal constraints derived from heteronuclear and homonuclear NMR data
on 13C and 15N-labeled RNA led to a structure for the loop-loop helix
with an average root-mean-square deviation of 0.83 (0.10) A Ê for 33 con-
verged structures relative to the average structure. The loop-loop helix of
the kissing complex is distorted compared to A-form RNA. Its major
groove is blocked by the phosphodiester bonds that connect the ®rst loop
residue of each hairpin with its own stem, and it is ¯anked by two nega-
tively charged phosphate clusters. The loop-loop helix has alternating
helical twists between adjacent base-pairs. The base-pairs at the helix
junctions are overwound and three base-pairs near the helix junctions
adopt high propeller twists. All these changes reduce the distance needed
for the bridging phosphodiester bonds connecting each stem and loop to
cross the major groove of the loop-loop helix, and result in a deformed
RNA helix with localized perturbations in the minor groove surface. The
alternating helical twist pattern, plus other distortions in the loop-loop
helix may be important for Rom protein recognition of the kissing hair-
pin complex.
# 1997 Academic Press Limited
Keywords: NMR spectroscopy; RNA-RNA interactions; kissing hairpin
*Corresponding author complex; alternating helical twist; RNA-protein recognition

Introduction plasmid control the initiation of DNA replication

by interacting with RNAII to block the formation
Loop-loop interactions between RNA hairpins of the speci®c structure required. One is an RNA
participate in the antisense regulation of a variety (RNAI) complementary to the 50 end of RNAII and
of cellular processes in prokaryotic cells and in bac- the other is a protein, Rom (RNA One Modulator),
teriophages (Wagner & Simons, 1994). Antisense that recognizes the RNAI-RNAII complex. The
regulation is widespread in eukaryotes as well (for complementary sequences of RNAI and the 50 end
example, see Pilipenko et al., 1996; Cavaille et al., of RNAII each can form three tandem stem-loop
1996). The most extensively studied biological sys- structures with complementary loop sequences.
tem that involves RNA loop-loop interactions is The interaction of RNAI and RNAII is initiated by
the antisense regulation of ColE1 plasmid DNA base-pairing between these complementary hairpin
replication (Tomizawa et al., 1991 and references loops through a kinetically favored intermediate,
therein). An RNA primer is used for the initiation the ``kissing hairpin complex''. The ®nal state is a
of DNA synthesis during replication of the ColE1 duplex formed by hybridization of the complemen-
plasmid in Escherichia coli. The RNA primer is gen- tary RNAs. Rom speci®cally binds to the RNA
erated from an RNA precursor (RNAII) by RNase complex formed by two complementary hairpins
H cleavage. A speci®c structure in RNAII is re- derived from RNAI and RNAII and stabilizes their
quired for it to bind to the DNA near the replica- association (Tomizawa, 1990, 1993). It is the
tion origin, and to form the substrate for RNase H. speci®c loop-loop interaction, the kissing hairpin
Two regulatory factors encoded by the ColE1 complex, rather than the duplex that is recognized

0022±2836/97/210052±15 $25.00/0/mb971021 # 1997 Academic Press Limited

RNA Kissing Hairpins 53

by the Rom protein (Eguchi & Tomizawa, 1990,

1991).
Knowledge of the structures and stabilities of
RNA loop-loop interactions will be valuable for
understanding how RNAs interact with each other,
and how the folding of larger RNAs occur. NMR
and gel electrophoresis studies of the ColE1 kissing
hairpin complex (Marino et al., 1995) showed that
all seven nucleotides of the loop-loop helix formed,
and provided strong evidence that the complex is
bent. However, detailed structural information was
not obtained. Interactions between the kissing
hairpin complex and the Rom protein make this
system particularly attractive for studying RNA-
protein recognition (Predki et al., 1995) and pro-
tein-mediated RNA folding processes. One import-
ant question to ask is what are the structural
features of the kissing hairpin complex that are
recognized by the Rom protein. We had previously
studied the formation of a kissing complex be-
Figure 1. A representation of the secondary structure of
tween the TAR hairpin loop of HIV-2 (part of the Tar-Tar* based on exchangeable and non-exchangeable
transcriptional activation region of the viral RNA) proton NMR data. The intra-strand and cross-strand
and its complement (Chang & Tinoco, 1994). We AH2-H1 NOE network of the ®ve adenine residues of
found that a 16-mer RNA hairpin (Tar*-16) with a Tar-Tar* are indicated by arrows.
loop sequence (UCCCAG) complementary to the
TAR loop sequence (CUGGGA) associates speci®-
cally with the TAR hairpin loop (Tar-16) to form a
stable kissing hairpin complex. This complex can of the complex indicated that all the sugar puckers
be bound by the Rom protein with similar af®nity of the residues were in a 30 -endo conformation, as
to the wild-type kissing hairpin complex from the expected for an A-form RNA duplex. The ®ve ade-
ColE1 plasmid (Chang, 1995). Here, we describe nine H2 protons were distinguished from the H6/
the solution structure of the Tar-Tar* RNA kissing H8 protons by their very different 13C chemical
hairpin complex. The loop-loop interaction helix shifts in a natural abundance 13C HMQC exper-
contains unique structural features that may be a iment (Varani & Tinoco, 1991b). Three adenine H2
general characteristic of kissing hairpin complexes protons were assigned to the A2 of Tar-16 and A*10,
recognized by the Rom protein. Loop-loop inter- A*14 of Tar*-16 from the strong NOE seen between
actions are important in folding of RNAs and the imino proton of the A
U base-pair and the H2
make likely sites for metal ion and protein binding. of the adenine residue.
A 2D 2H2O NOESY spectrum of the complex
was very crowded with few crosspeaks easily
Results identi®ed. However, the spectrum was greatly sim-
Spectral assignment and characterization of pli®ed by isotope-editing of an RNA sample with
secondary structure features one hairpin 13C-labeled and the other hairpin left
unlabeled. We used nucleotides speci®cally 13C-la-
Our previous study (Chang & Tinoco, 1994) had beled to simplify the spectrum and to reduce the
shown that the Tar-Tar* kissing hairpin complex broadening effects introduced by uniformly labeled
involved the formation of at least ®ve base-pairs nucleotides (SantaLucia et al., 1995). Figure 2(a)
between the two hairpin loops as shown in and (b) are the H10 -H6/H8 portion of a 13C-1/2 ®l-
Figure 1. The imino proton resonances of all 11 tered NOESY spectrum (Otting & WuÈthrich, 1990)
possible G
C base-pairs and three of ®ve possible of the complex with all 16 residues of Tar-16
A
U base-pairs were seen in the exchangeable pro- speci®cally 13C-labeled at the C6 (pyrimidines) or
ton NMR spectrum of the Tar-Tar* complex. The C8 position (purines) of the bases. Figure 2(a) is
formation of A*12
U*5 for the Tar*-16 in the complex, the sub-spectrum containing the H10 -H6/H8 NOE
and the loop-loop A11
U*6 base-pair could not be that involves 13C-attached protons and corresponds
established because of the absence of U*5 and U*6 to the Tar-16 part of the complex. There are two
imino protons in the 1D spectra, presumably due H10 -H6/H8 NOE connectivity walks in Figure 2(a)
to fast exchange with solvent. that correspond to the NOE connectivity of two
To obtain a high-resolution structure of the Tar- fragments of stacked RNA oligonucleotides. One is
Tar* kissing complex we used unlabeled samples, a 5-mer with a 50 to 30 base sequence of Pur-Pur-
plus samples in which one of the hairpins in the Pur-Pyr-Pyr and the other is an 11-mer with a base
complex was 13C-labeled and the other unlabeled. sequence of Pyr-Pyr-Pur-Pur-Pur-Pur-Pur-Pur-Pyr-
The absence of H10 -H20 sugar proton crosspeaks in Pyr-Pyr. The six sequential purines in the 11-mer
a DQF-COSY spectrum (Marion & WuÈthrich, 1983) segment indicate that this segment is C6UGG-
54 RNA Kissing Hairpins

GAGGCUC16 and thus helps assign the 5-mer seg-

ment to G1AGCC5. Similarly, in the NOESY sub-
spectrum that originated from 12C-attached H6/H8
protons (Figure 2(b)), two NOE connectivity walks
belonging to a 5-mer and an 11-mer are assigned as
the G*1C*U*G*U*5 and U*6C*C*C*A*G*A*C*A*G*C*16
of the Tar*-16 in the complex.
3D HMQC-NOESY experiments (Fesik &
Zuiderweg, 1990) on the complex with either hair-
pin fully labeled with 13C-enriched C6/C8 NTPs
were used in the assignment of non-anomeric
sugar protons. A short mixing time 2H2O NOESY
experiment on unlabeled complex was used to as-
sign the H20 sugar protons through intra-ribose
H10 -H20 and internucleotide H20 -H6/H8 NOEs.
This information, plus the previously assigned H10
and base protons, help to assign each H20 sugar
proton of the complex. An H20 -aromatic proton
NOE connectivity walk for the loop residues of the
Tar*-16 part of the complex is presented as an as-
sembly of 2D slices taken from the 3D HMQC-
NOESY spectrum (Figure 3). In A-form RNA, the
H20 , H30 sugar protons are less than 4 A Ê away
0
from the H5 base proton of an adjacent 3 pyrimi-
dine (Varani & Tinoco, 1991a). This property was
used to con®rm the H20 proton assignments as
well as the assignment of 15 H30 protons. H10 -H30
proton NOEs were then used to con®rm these as-
signments and to locate the remaining H30 protons.
An H30 -H6/H8 NOE connectivity walk for the
same region of loop residues of Tar*-16 is shown
in Figure 3 and is consistent with the previous 10
and 20 -H6/H8 NOE connectivity walks. The sugar-
aromatic proton NOE connectivity patterns and
the observations of aromatic-aromatic proton
NOEs on adjacent residues (data not shown)
suggest continued stacking within these four frag-
ments of RNA (two 5-mers and two 11-mers). The
identi®cation of two non-interrupted 11-mers indi-
cates that the continuing stacking of loop residues
extends into the 30 side of both hairpin stems.
Two extra crosspeaks (boxed) in Figure 2(a) and
(b) suggest that intra-hairpin NOEs exist between
the C5H10 and C6H6 of Tar-16, as well as between
the U*5H10 and U*6H6 of Tar*-16. However, because
of spectral overlap, alternative interhairpin connec-
tivities are possible. Similar ambiguity occurs in
tracing the H20 /H30 -aromatic proton NOE connec-
tivities at the same stem ±loop junction. Therefore,
13
C double-1/2-®ltered NOESY experiments
(Otting & WuÈthrich, 1990) on a sample containing
Figure 2. H10 -H6/H8 region of 2H2O NOESY spectra of uniformly labeled Tar-16 and unlabeled Tar*-16
Tar-Tar*. (a) Two H10 -H6/H8 NOE connectivity walks were used to distinguish the intrahairpin from in-
in the 13C sub-spectrum from a 300 ms 13C 1/2-®ltered terhairpin NOEs. The non-anomeric sugar proton-
NOESY spectrum of Tar-Tar* with C6/C8 13C-labeled aromatic proton region of a 13C-edited NOESY
Tar-16. They correspond to nucleotides in Tar-16 and spectrum showed that there is no intra-hairpin
lead to a secondary structure that has all six loop resi- U*5H20 -U*6H6 NOE in Tar*-16 in a 12C double-se-
dues stacking on the 30 side of Tar-16 stem. (b) Two
lected NOESY spectrum compared with the same
H10 -H6/H8 NOE connectivity walks in the 12C sub-spec-
trum from a 300 ms 13C 1/2-®ltered NOESY spectrum region of a NOESY spectrum of an unlabeled com-
of Tar-Tar* with C6/C8 13C-labeled Tar-16. They corre- plex. In contrast, an inter-hairpin NOE between the
spond to nucleotides in Tar*-16 and lead to a secondary C5H20 of Tar-16 and the U*6H6 of Tar*-16 as well as
structure that has all six loop residues stacking on the 30 an inter-hairpin NOE between the U*5H10 , H20 of
side of Tar*-16 stem. Tar*-16 and the C6H6 of Tar-16 can be detected in
RNA Kissing Hairpins 55

Figure 4. AH2-H10 NOEs of the ®ve adenine residues of

the complex in the 12C sub-spectrum from a 300 s 13C
1/2-®ltered NOESY spectrum of Tar-Tar* with C6/C8
13
C-labeled Tar-16. Two unusually strong AH2-H10
NOEs (labeled in bold face) are seen. These crosspeaks
have similar intensities to pyrimidine H5-H6 crosspeaks;
they are the only adenine crosspeaks seen at 60 ms mix-
ing times.

base-pairing partner of its 50 nearest neighbor

(Varani & Tinoco, 1991a). The AH2-H10 proton
NOE spectra show that every adenine residue in
the complex has two AH2 to H10 proton NOEs
(Figure 4). This indicates that all the adenine resi-
dues are surrounded by a duplex environment, in-
Figure 3. H20 and H30 -H6/H8 connectivity walk on 2D cluding A*12 of Tar*-16 and A11 of Tar-16 (Figure 1),
slices of 3D HMQC-NOESY of Tar-Tar* with Tar*-16 which also have interhairpin AH2-H10 NOEs with
13
C-labeled at C6/C8 position. It shows the 20 -walk the H10 of C6 in Tar-16 and C*7 in Tar*-16, respect-
(black line) and 30 -walk (gray line) from U*6 to G*15 of ively. Thus, although the imino protons of the po-
Tar*-16 and further supports a secondary structure with tential A11
U*6 and A*12
U*5 are not seen in the
30 -stacking of all the loop residues. The number shown exchangeable proton spectrum, their characteristic
in the right corner within each slice is the 13C chemical AH2-H10 NOE patterns indicate that these adenine
shift of each corresponding C6 or C8 carbon atom.
residues are in a double helix environment. These
data also indicate the existence of a continuous
helix starting from the end of the Tar-16 hairpin
stem and ending at the end of the Tar*-16 hairpin
a 12C-selected, 13C-®ltered double-1/2-®ltered stem with the loop-loop interaction helix in be-
NOESY spectrum (data not shown). These inter- tween.
hairpin NOEs combined with the previous H10 / There are two unusually strong AH2 to H10
H20 /H30 -H6/H8 NOE connectivity walks ®x the intra-strand NOEs in the short mixing time
orientation of the loop-loop helix. They support the NOESY spectrum. They occur between A11H2 and
secondary structure shown in Figure 1, with con- G12H10 of Tar-16, and between A*10 H2 and G*11 H10
tinuous stacking of the 30 -side of each stem into of Tar*-16. These unusual intrastrand AH2-H10
each loop. However, the ®rst base of each loop NOE intensities are very different from those
(nucleotide 6 of each hairpin) stacks on the term- expected from a normal A-form RNA helix and
inal residue of the 50 side of the stem of the other suggest an unusual conformation in the minor
hairpin (nucleotide 5 of each hairpin). It does not groove of the loop-loop helix. Examination of the
stack on its own stem. imino-H10 crosspeak intensities of a 100 ms mixing
In an A-form helix, each adenine H2 proton has time H2O NOESY also reveals the existence of rela-
an intra-strand NOE interaction with the H10 pro- tively strong G8NH-G9H10 and G10NH-A11H10
ton of the ribose on its 30 side as well as a cross- NOEs, which are much weaker in a normal A-form
strand AH2 to H10 NOE to the ribose of the helix (data not shown).
56 RNA Kissing Hairpins

Phosphorus chemical shifts can be used to locate straints, and 39 hydrogen bond constraints were
unusual phosphate backbone torsion angles in nu- used for the structure calculations (Table 1); this
cleic acids (Gorenstein, 1981). Several phosphorus corresponds to 16 experimental constraints per
chemical shifts outside the normal phosphorus residue. Weak A-form constraints were used for
chemical shift region of an A-form RNA were seen the ®rst two base-pairs of Tar* and the ®rst three
in a 31P-1H heteronuclear COSY spectrum (Sklenar base-pairs of Tar. One hundred random structures
et al., 1986) with a span of 2.2 ppm for the phos- were generated using X-PLOR (BruÈnger, 1990) and
phorus chemical shift range (data not shown). Un- were then subjected to a simulated annealing-re-
fortunately, attempts to make a sequence-speci®c strained molecular dynamics protocol followed by
assignment of these phosphorus resonances an energy minimization step. The minimization
through a 2D hetero-TOCSY-NOESY experiment steps were carried out without including electro-
(Kellogg et al., 1992) were not successful, probably static interactions. The total energy and the NOE
because of the large size of the complex. violation energy of each structure were used as cri-
The unassigned non-anomeric sugar protons- teria for identi®cation of converged structures.
H6/H8 NOEs in the 3D HMQC-NOESY spectra
Figure 6 shows the calculated structures arranged
were picked as potential candidates for the intra-
in order of increasing total and NOE violation
nucleotide H40 and H50 , H500 -H6/H8 NOEs. They
energies. There is a small jump in the total energy
are assigned to H40 , H50 or H500 if corresponding
H10 -H40 or H10 -H50 /H500 proton NOEs can also be after the ®rst 39 structures, and a jump in NOE
found. Because the 13C chemical shift of C4 and C5 violation energies after 33 structures. The six struc-
of sugars are very different, they help to dis- tures in between did not form the A*
U* base-pair
tinguish and con®rm the H40 and H50 /H500 proton at the end of the Tar*-16 stem (Figure 1); these
assignments available through 13C HMQC and structures had short interproton distances not seen
constant-time HSQC spectra (Vuister & Bax, 1992). as NOE crosspeaks. We therefore chose the ®rst 33
This led to the assignment of most of the H40 , H50 / structures for further analysis. A second set of
H500 sugar protons of the Tar*-16 and some H40 of structural constraints that included hydrogen bond
Tar-16. constraints for A11
U*6 and A*12
U*5 was used for
similar structure calculations. These constraints led
to a higher convergence ef®ciency with 44 struc-
Structure determination tures with low total energy and NOE-violation en-
ergy generated from 100 random structures.
Experimental constraints and However, the average structure and the precision
structure calculations
of the range of structures for both sets of con-
Figure 5 shows a diagram summarizing the in- straints were very similar. We are convinced that
ternucleotide NOE network observed in the all base-pairs in the complex form, although the
NOESY spectra of the complex. A total of 423 NOE two U imino groups from the junction region are
distance constraints, 60 experimental dihedral con- in fast exchange with solvent.

Figure 5. The inter-nucleotide NOE

network of the Tar-Tar* kissing
hairpin complex. The two hairpins
are arranged in duplex form with
the shaded rectangular bars repre-
senting the presence of the imino
protons in the 1D exchangeable
proton spectrum. Inter-hairpin
NOE bonding constraints were
added only for base-pairs that had
imino protons visible in the spectra.
Inter-hairpin NOEs occur at the
helix junctions (C5-U*6 and U*5-C6),
and in the loop-loop helix; 16
NOEs for exchangeable protons are
not shown.
RNA Kissing Hairpins 57

Table 1. Statistics of experimental constraints used in the structure calculations, and precision
of the lowest-energy structures
A. Distance and dihedral constraints
Intraresidue 197 Ê)
Strong (1.8± 3.0 A 70
Interresidue 226 Medium (2.0 ±4.0 A Ê) 150
Intrahairpin 198 Weak (3.0 ±5.0 AÊ) 164
Interhairpin 28 Very weak (4.0 ±7.0 A Ê) 39
Hydrogen bond 39
Dihedral constraints 60
Constraints per residue 16

B. Precision of lowest-energy structuresa

RMSD (A Ê)
Whole complex (16 bp) 1.71  0.33
One-turn (11 bp; C4-G13, U*3-A*14) 1.28  0.19
Loop-loop interaction helix (6 bp)
C6-A11, U*6-G*11 0.83  0.10
a
The root-mean-square deviation (RMSD) of the 33 lowest-energy structures from the average structure.

The precision of the ®nal calculated structure is the superposition of the central part of the 22
was measured by comparing the root-mean-square lowest energy structures (RMSD of 1.28 (0.19) A Ê
deviation (RMSD) of the 33 converged structures calculated for 33 converged structures). They
from the average structure (Table 1). The RMSD correspond to two and three base-pairs of the
for the stem regions of Tar-16 and Tar*-16 are 0.55 stems of Tar-16 and Tar*-16, respectively, as well
(0.13) and 0.58 (0.11) A Ê , respectively, and the as the loop-loop interaction helix. The remaining
loop-loop interaction helix has a calculated RMSD terminal base-pairs in each stem are not included
of 0.83 (0.10) AÊ . Using all atoms in the complex, because their proton chemical shift values remain
an RMSD of 1.71 (0.33) A Ê relative to the average unchanged upon complex formation. The RMSD
structure was obtained. This variation is mainly value suggests that the local geometry is relatively
caused by different global features among the con- well de®ned within each helical region. The heli-
verged structures. There are 26 bent structures of cal parameters of the 33 converged structures
different bending angle (from 30 to 90 ) and seven generated from the different sets of constraints
linear structures among the 33 converged struc- were analyzed using the program CURVE
tures. No obvious energy differences or NOE viola- (Lavery & Sklenar, 1989). The three parameters
tions distinguish these two different global (twist, roll and tilt) that de®ne the relative orien-
structures. It is well known that accurate global tation of successive base-pairs (Dickerson et al.,
conformations of elongated nucleic acids can be 1989) are listed in Table 2. They are precisely
very dif®cult to obtain from NMR data. Thus, we determined with standard deviations that are
concentrate on the local structural features that can within 10% of the mean value for most of the
be determined well by NMR. Shown in Figure 7(a) values. A representative structure is shown in
Figure 7(b).

Structural features of the Tar-Tar* kissing

hairpin complex
The formation of a new helix of six base-pairs
involving all the residues within the loops and the
requirement to maintain the covalent linkage in the
stem ±loop junction of each hairpin imposes severe
constraints on the 16 base-pair Tar-Tar* kissing
hairpin complex. Shown in Figure 8 is a side-by-
side comparison of the central part (11 bp) of the
complex with one turn of standard A-form RNA
and B-form DNA (Saenger, 1984) of identical se-
quences (with thymine in the DNA). The path of
the phosphate backbone in the junction region
where the three helices of the complex stack is a
very distinct structural feature of this kissing hair-
pin complex. The continuous 30 -stacking of the
Figure 6. Structures plotted as a function of increasing loop residues that extend into the 30 -side of each
NOE-violation energies (®lled circles) and increasing hairpin stem create a gap between U*5 and U*6 as
total energies (®lled squares). well as between C5 and C6 that can be connected
58 RNA Kissing Hairpins

Figure 7. (a) Superposition of the central 11 base-pairs of the Tar-Tar* complex (six from the loop-loop helix, three
from the Tar* stem and two from the Tar stem) for 22 converged structures with lowest total energies. The protons
are omitted to allow a clearer presentation of the backbones. (b) Stereo view of a single low-energy structure.

most easily by crossing the major groove of the Short interphosphorus distances are observed
loop-loop interaction helix. These gaps are about between the 50 phosphate group of C5 and the 50
Ê long for A-form geometry and need at least
10 A phosphate groups of U7 and G8 of Tar-16
two nucleotides to cross them (Haasnoot et al., (5.25(0.77) AÊ and 9.25(1.21) A Ê , respectively) as
1986). In the three-dimensional structure of the well as between U*5, and C*7, C*8 of Tar*-16
Tar-Tar* complex, only one phosphodiester bond is (6.7(0.9) AÊ and 7.2(1.3) A
Ê , respectively). In con-
used to connect each gap (Figure 9(a)). The phos- trast, the corresponding interphosphorus distances
phodiester bonds that cross the major groove to in an A-form RNA helix are 10.4 A Ê and 12.7 AÊ.
bridge the ®fth and sixth residues in each hairpin This results in two sets of spatially close, nega-
are found to be pushed toward each other, thus tively charged phosphate clusters located at the
blocking accessibility of the major groove formed edges of the major groove (Figure 9(a)). One set is
by the loop-loop helix (Figure 9(b)). the 50 phosphate groups of C5, C6, U7 and G8 of
RNA Kissing Hairpins 59

Table 2. Helical parameters between successive base-pairs in the loop-loop interaction helix
and in both hairpin stems from 33 lowest-energy structures
Base-pair step Twist Roll Tilt
Loop-loop helix:
U*6-A11 to C*7-G10 35.0  5.3 21.2  9.0 5.1  8.0
C*7-G10 to C*8-G9 27.8  3.8 13.1  2.1 3.3  6.0
C*8-G9 to C*9-G8 42.2  4.2 29.3  6.6 10.6  12.6
C*9-G8 to A*10-U7 33.2  3.4 ÿ5.2  4.3 ÿ3.0  5.0
A*10-U7 to G*11-C6 40.9  3.5 7.0  4.6 ÿ14.0  5.4
Average: 35.8  2.8 0.3  1.8

Tar-16 stem:
G1-C16 to A2-U15 33.6  3.0 6.1  8.4 8.0  7.1
A2-U15 to G3-C14 26.4  2.1 4.6  5.2 ÿ1.3  9.4
G3-C14 to C4-G13 29.1  2.3 26.1  9.5 ÿ6.8  7.7
C4-G13 to C5-G12 32.1  3.2 22.4  11.5 ÿ0.7  5.0

Tar*-16 stem:
G*1-C*16 to C*2-G*15 33.1  1.6 13.5  3.8 11.5  4.7
C*2-G*15 to U*3-A*14 29.2  1.4 6.8  3.2 ÿ9.7  3.1
U*3-A*14 to G*4-C*13 30.5  2.1 ÿ3.7  5.9 2.3  3.3
G*4-C*13 to U*5-A*12 22.1  2.1 6.6  5.2 2.3  3.3

Tar-16 and the other set is the 50 phosphate et al., 1977). These two phosphate clusters ¯ank-
groups of U*5, U*6, C*7 and C*8 of Tar*-16. The in- ing the major groove of the loop-loop interaction
terphosphorus distances in the Tar* phosphate helix may constitute part of a speci®c metal ion-
cluster are very similar to those between the 50 binding site. The Tar-Tar* complex (Chang &
phosphate groups of U8, A9, C11 and U12 of yeast Tinoco, 1994) and the wild-type ColE1 kissing
tRNAPhe that form a Mg2‡-binding site (Holbrook hairpin complex (Tomizawa, 1990; Gregorian &

Figure 8. Comparison of the central 11 base-pairs of the Tar-Tar* complex with standard B-form DNA and A-form
RNA. The red residues correspond to the stem of Tar*-16 and the violet residues belong to the stem of Tar-16. The
pink and blue residues in the middle are the loop residues of Tar*-16 and Tar-16, respectively.
60 RNA Kissing Hairpins

Figure 10. The high propeller twist of the A*12
U*5

base-pair in the helical junction.

prove the stacking between the purines in the

junction regions and may make important contri-
butions to the stability of the complex.
Table 2 shows that the twist angle between adja-
cent base-pairs along the loop-loop helix varies al-
ternately from the average twist angle of the helix.
An overtwist in one step is followed by an under-
twist in the next step. The two central base-pairs of
the loop-loop helix have an average twist angle of
42(4) for the step from G8
C*9 to G9
C*8. Away
from this central step, there is an undertwist step
and an A-form step (average 28(4) and 33(3) )
that are followed by another two overtwist steps
(average 35(5) and 41(4) ). This leads to an
Figure 9. (a) View into the major groove of the Tar- average twist angle of 36(3) in each step
Tar* kissing hairpin complex with the ¯anking phos- (Table 2).
phate clusters shown as yellow spheres. (b) Comparison
of the loop-loop helix with the same sequence of stan-
The unique structural features ®t well with the
dard A-RNA looking down from the top of the major NMR data, especially the unusual AH2/NH-H10
groove. It shows the blocking of the major groove of the NOE intensities in the short-mixing time 2H2O/
loop-loop helix by the two crossing phosphodiester H2O NOESY spectra (Figure 4). These intensities
bonds. The arrows indicate the phosphate groups of C6 within the loop-loop interaction helix are very im-
of Tar-16 and U6* of Tar*-16. portant characteristics for this distorted helix and
can be used to distinguish it from the normal A-
form RNA helix. These NOEs occur in the base-
pairs adopting the high propeller twist and alter-
Crothers, 1995) are speci®cally stabilized by mag- nating helical twist and are an indication of a
nesium ions. narrow minor groove. There are several unusual
There is base stacking continuing from the A- aromatic proton chemical shifts in residues near
form stems through the entire RNA complex with the two terminal base-pairs of the loop-loop helix.
some distortions within the loop-loop helix and They are the unusually down®eld-shifted H6 of C6
overwinding at the interhairpin stacking steps be- in Tar-16 (8.51 ppm) and H6 of U*6 (8.65 ppm), and
unusually up®eld-shifted H10 (5.01 ppm), H8
tween the bases of C5 and U*6, and between U*5 and
(6.65 ppm) of G*11 in Tar*-16. These unusual chemi-
C6. Speci®cally, three base-pairs located at the helix cal shifts can be qualitatively explained by the
junctions are found to adopt very high propeller stacking patterns found in the structure. The
twist. They are C6
G*11, A11
U*6 and U*5
A*12, and down®eld H6 chemical shift of the two pyrimi-
their propeller twist values are 38(4) , 30(14) dines in the sixth position of both hairpins can be
and 26(7) , respectively, with one of them shown rationalized by the deshielding of both H6 protons
in Figure 10. These high propeller twists can im- caused by the destacking. In contrast, the position
RNA Kissing Hairpins 61

G*11 and A*12 are 7.9(0.4) A Ê and 7.2(0.3) AÊ , re-

spectively. The corresponding distances in an A-
form helix are 8.7 AÊ and 9.3 A
Ê . Figure 11 is a view
of the minor groove surface of the loop-loop inter-
action helix compared with a normal A-form RNA
helix of the same sequence. There is a clear differ-
ence in the water-accessible surfaces, and the phos-
phodiester backbone in the loop-loop interaction
helix is pushed toward the minor groove. The spa-
cings between phosphate groups and the 20 -hy-
droxyl groups create distinct features within the
minor groove surface. This perturbation may be a
general feature of the kissing hairpin complex, and
is consistent with the distinct RNase V1 cleavage
pattern in the ColE1 kissing complex compared
with a duplex of the same sequence (Eguchi &
Tomizawa, 1990).

Discussion
The helix formed by base-pairing between the
two hairpin loops in the kissing complex is signi®-
cantly different from an A-form helix. The main
NMR data that determine its structure are: (1) H10 ,
H20 , H30 -aromatic NOE connectivities that require
continuous stacking of bases throughout the com-
plex. (2) Cross-strand AH2-H10 NOEs indicating
formation of the A
U pairs at the ends of the loop-
loop helix. (3) An unusual pattern of NOE intensi-
ties that are consistent with an alternate overwind-
ing and underwinding in the loop-loop helix,
Figure 11. Minor groove water-accessible surface of the
overwinding at the junctions between the helices,
loop-loop helix of the Tar-Tar* complex compared with and a high propeller twist of the base-pairs at the
the minor groove surface of a standard A-form RNA junctions (Figure 10). All these structural features
duplex of the same sequence. The surfaces are generated are well characterized by the precision of the range
using a water probe of 1.4 AÊ radius on Insight II (Bio- of structures consistent with the NMR data. Large
sym Technologies Inc.) with the colors red, blue, yellow positive roll angles and large (but varied) tilts are
and green representing O1P, O2P, phosphorus and the found for the base-pairs along the loop-loop helix
hydrogen atoms of 20 -hydroxyl groups, respectively. (Table 2). The large standard deviations observed
in the tilt angle of the central step is indeed caused
by the existence of two conformers with very
different tilt angles (0 versus 25 ). The average
structure of the kissing complex is bent. However,
of the purine ring of A*10 can explain the up®eld there was a wide range of bend angles; 26 struc-
G*11 H8 chemical shift. tures had bending angles ranging from 30 to 90 ,
The average structure of the kissing complex is and seven were approximately linear. Bending can
bent towards the major groove (although with a also be interpreted as anisotropic ¯exibility (Crick
large variation in bending angles among the & Klug, 1975) in this center region of the loop-loop
converged structures) as seen in Figure 7(b). The interaction helix.
main contributions to the bend are the large posi- Gel electrophoresis and NMR data showed that
tive roll angles in the loop-loop helix, particularly the ColE1 kissing hairpin complex is bent towards
at the center step (Table 2). Together, the over- the major groove with an estimated bending angle
wound loop-loop helix and the bend in the center of 45 (Marino et al., 1995). All seven bases in the
of the helix combine to reduce the distance for the loop-loop helix were paired and continuous stack-
phosphodiester bonds to cross the major groove of ing of the loop bases on the 30 -side of each stem
the helix. The loop-loop interaction helix also pro- occurred. It is encouraging that the two kissing
duces a narrower minor groove with localized sur- complexes, although containing different sequences
face perturbations compared with a normal A-form and loop lengths, have qualitatively similar struc-
helix. This perturbed minor groove surface can be tures.
characterized by the base-pair tilts and twists Sequence-dependent, local structure variations
(Table 2), and by cross-strand C10 -C10 distances. are well documented in DNA, such as in alter-
For example, the C10 -C10 distances from U7 to nating B-DNA (Klug et al., 1979; Yuan et al.,
62 RNA Kissing Hairpins

1992) and A-tract DNA (Nelson et al., 1987; Yoon context between two adjacent helices. The distor-
et al., 1988), and may be important in a variety tion is presumably induced by the constraints im-
of transcription regulation processes involving posed by the maintenence of 30 -helical stacking of
DNA-protein interactions (Calladine & Drew, the loop residues, plus the A-form geometry of the
1992). For example, sequence-dependent alternat- hairpin stems.
ing helical twist is involved in the met repressor-
operator complex (Somers & Phillips, 1992) and
in the complex between telomeric DNA and the
Rom binding
DNA-binding domain of RAP1 (KoÈnig et al., For the hybridization of RNAI and RNAII to
1996). Also, DNA twisting on the non-contacted proceed, the base-pairs from the helices of the hair-
bases of the bacteriophage 434 operator has been pins need to melt; this is a very slow process
shown to affect the af®nity of the bacteriophage (Turner et al., 1990) without the help of helicases. It
434 operator for bacteriophage 434 repressor was shown that the Rom protein can affect the kin-
(Koudelka et al., 1987, 1988; Koudelka & Carlson, etics of hybridization between RNAI and RNAII
1992). by decreasing the dissociation rate of the kissing
In contrast to DNA, RNA helices are found to be hairpin complex, but there is no evidence that the
more rigid (Egli et al., 1996). We thus think that the Rom protein is acting as a helicase. Because the
structural variations seen in the loop-loop helix are Rom protein can bind to kissing hairpin complexes
caused by the constraints imposed by the presence of very different sequences, it suggests that com-
of the three coaxially stacked helices. For example, mon structural features shared among different
analysis of the phenylalanine tRNA structure did complexes are recognized by Rom. The unique fea-
not reveal sequence-speci®c structure variation tures of the central loop-loop helix of the Tar-Tar*
(Holbrook et al., 1981). Our analysis of available kissing complex include: a compressed major
RNA crystal structures also suggests that the alter- groove, two groups of four closely spaced phos-
nating helical twist pattern observed here is not a phate groups around the major groove, and a dis-
result of a sequence-dependent effect. There are torted minor groove surface. The loop-loop helix
two CUG steps in the complex: one in the stem of distortion is characterized by successive steps of
Tar* and one in the loop-loop helix. The two helical over- and underwinding, and increased propeller
twist angles in the C*U*G* step of the Tar* stem twists for the junction base-pairs. Presumably, all
are found to have similar value of 30 compared these structural characteristics contribute to Rom
with the overtwist, undertwist pattern seen in the recognition. Further studies on other kissing hair-
loop-loop interaction helix of the identical CUG pin complexes of different loop sequences and loop
step (Table 2). This pattern remains unchanged size will be needed to address this problem.
when structures are calculated with the presence of
electrostatic interaction in the ®nal energy mini-
mization stage. The helix distortions seen in the Materials and Methods
Tar-Tar* kissing complex suggest that deformabil-
Preparation of RNA sample
ity may be important for the formation of a stable
complex. The sequence requirement for the for- C6/C8 13C-labeled NMPs were chemically synthesized
mation of a kissing hairpin complex may thus be (SantaLucia et al., 1995) and were gifts from Dr John San-
the exclusion of sequence combinations that lead to taLucia Jr. Uniformly 13C-labeled NMPs were isolated
reduced deformability (Gregorian & Crothers, from Methylophilus methylotrophus bacterial cells supplied
1995). by the Stable Isotope Resource at Los Alamos National
Distortion of RNA helices has been found to be Laboratory following the published procedures (Batey
et al., 1992; Nikonowicz et al., 1992). The 13C-labeled
a major contribution for RNA recognition. It can be
NMPs were converted into their NTP forms by enzy-
achieved by either the insertion of a bulge into an matic phosphorylation (Haynie & Whitesides, 1990).
RNA helix as found in the HIV TAR hairpin NMR quantities of RNA samples were synthesized
(Puglisi et al., 1992; Aboul-ela et al., 1995), or the in vitro using phage T7 RNA polymerase with a T7 pro-
formation of a mismatch within a helix, as seen in moter and speci®c DNA templates. The transcription
the HIV RRE internal loop (Battiste et al., 1996). In yield of isotope-labeled RNA was optimized on an ana-
both cases, the distortion of an RNA helix leads to lytical scale by varying the concentration of Mg2‡ and
the opening of the major groove and thus allows then scaled up to a 20 ml transcription reaction for the
sequence-speci®c RNA recognition. Both the P1 preparation of RNA for NMR studies. The concen-
substrate recognition by the core of Tetrahymena trations of Tar-16 and Tar*-16 RNA were measured at
280 nm because of their high G ‡ C contents.
group I intron (Doudna et al., 1989; Strobel & Cech,
1995) and the acylation of the tRNAAla acceptor
stem by tRNAAla synthetase (Gabriel et al., 1996) NMR spectroscopy
require a distorted RNA helix with a conserved
RNA samples were dialyzed against 50 mM NaCl at
G
U wobble base-pair for recognition. The recog- pH 6.4 (H2O sample) or 180 mM NaCl at pH 7.0 (2H2O
nition of the kissing hairpin complex by the Rom sample) in 0.01 mM EDTA, 10 mM sodium phosphate
protein represents another type of protein recog- buffer. The ®nal sample concentrations were 1 to 1.5 mM
nition of an RNA helix. A fully Watson-Crick- for the 1D NMR studies and 2 to 2.5 mM for 2D or 3D
paired helix is recognized, but it is distorted by its homonuclear and heteronuclear NMR studies. The vo-
RNA Kissing Hairpins 63

lumes of NMR samples were between 0.4 and 0.5 ml for for ef®cient excitation of aromatic resonances
the unlabeled sample and 0.2 to 0.25 ml for the labeled (JCH ˆ 200 Hz). The 3D HMQC-NOESY spectra (Fesik &
samples. All the heteronuclear NMR experiments were Zuiderweg, 1990) were taken on the complex with C6/
done on a Bruker AMX 600 spectrometer except for the C8 13C-labeled Tar-16 or Tar*-16 with a delay 1/2J set to
2D 31P-1H heteronuclear correlation experiment, which 3.3 ms for ef®cient excitation of the sugar resonances
was measured on a Bruker AMX 300 spectrometer. All (JCH ˆ 150 Hz). The carbon carrier was set to the middle
NMR data were acquired using the time-proportional of the carbon sugar spectrum with a spectral width of
phase incrementation (TPPI, Marion & WuÈthrich, 1983) 6250 Hz (o1), 1500 Hz (o2) and 1200 Hz (o3). The total
or States-TPPI (Marion et al., 1989) formats and then number of increments in every dimension was t1 ˆ 32,
transferred to a Silicon Graphics workstation and pro- t2 ˆ 64 resulting in 2048 FIDs and the ®nal real matrix
cessed by the Felix 2.30 program (Biosym Technologies after zero-®lling contained o1 ˆ 512  o2 ˆ 128  o3 ˆ
Inc.). 64 points.
The experimental constraints used in the determi-
nation of structure were all taken from NMR data with- NOE distance constraints
out magnesium present. However, 1D and 2D
exchangeable and non-exchangeable proton NMR data The 2D 2H2O NOESY and 13C-®ltered NOESY exper-
in the presence of 5 mM Mg2‡ or 180 mM Na‡ showed iments of 60, 90, 120 and 200 ms mixing time were
no signi®cant changes in spectra, such as unusual chemi- done on unlabeled and labeled samples. The NOE vo-
cal shifts or changes in AH2 to imino or H10 NOE inten- lumes of all well-isolated crosspeaks from spectra of
sities. different mixing times were estimated and compared
The H2O NOESY spectrum of 2.5 to 3 mM RNA in with crosspeaks of constant distances in order to obtain
50 mM NaCl at pH 6.4 at 3 C were measured with 100 distance constraints for non-exchangeable protons. To
to 300 ms mixing times using a 1-1 NOESY pulse se- alleviate the ambiguity of crosspeak measurement
quence (Sklenar & Bax, 1987) with the carrier frequency caused by crosspeak overlap for residues in some cru-
set on the water resonance. In all, 2K complex points in cial junction regions, 13C-labeled C6/C8 NTPs were
t2 and 300 to 350 t1 increments were acquired with 96 speci®cally incorporated into the complex. For example,
13
scans per t1 increment and spectral width of 12,000 Hz C-labeled C6/C8 UTPs were incorporated into Tar*-
on a Bruker 600 spectrometer. Prior to the Fourier trans- 16 to help distinguish U5, U6 and C7 in the stem± loop
formation, the data sets were processed by digital shift junction. NOE crosspeaks were classi®ed as strong,
subtraction in both dimensions and apodized using a 30 medium, weak and very weak. NOE crosspeaks that
skewed sine bell in the ®rst dimension and 60 in the can be seen at 60 ms mixing times with crosspeak in-
second dimension; the data were zero ®lled in both di- tensities similar to those of pyrimidine H5-H6 cross-
mensions to give a ®nal 2K  2K real data matrix. peaks were classi®ed as strong and were assigned an
The 2D double-quantum-®ltered COSY spectra (DQF- inter-proton distance range of 1.8 to 3.0 A Ê . Nearly all
COSY, Marion & WuÈthrich, 1983) and 2H2O NOESY (90%) of the strong crosspeaks are intranucleotide H10 -
spectra (Macura et al., 1982) at 60 to 300 ms mixing time H20 and internucleotide H20 -H6/H8 crosspeaks. Those
were recorded at 25 C with a low power preirradiation NOEs not found at 60 ms mixing times but that ap-
during a two second relaxation delay to suppress the re- peared at 90 ms mixing times were grouped as med-
sidual H2O peak. The 2K complex points in t2 and 350 to ium and given a distance range of 2.0 to 4.0 A Ê. A
400 (FIDs) in t1 were collected with a spectral width of distance range of 3.0 to 5.0 A Ê was assigned for NOE
4608 Hz (500 MHz spectrometer) or 5000 Hz (600 MHz). crosspeaks classi®ed as weak that were absent in short
A proton-detected heteronuclear 31P-1H COSY was re- mixing time NOESYs but occurred at 120 ms mixing
corded as proposed by Sklenar et al. (1986) with a spec- times. Finally, NOE crosspeaks that are seen only at
tral width of 600 Hz in the 31P dimension and 1400 Hz in 200 ms or longer mixing times were put into the group
the 1H dimension. 118 FIDs of 2K complex points were of very weak with an assigned inter-proton distance of
collected with 176 scans per FID on a Bruker AMX 300 4.0 to 7.0 AÊ . Fixed distances as suggested by Saenger
spectrometer. A natural abundance proton-detected (1984) were also assigned between atoms involved in
13
C-1H HMQC spectrum (Bax et al., 1983) was recorded hydrogen bond formation in Watson-Crick base-pairs
with a spectral width of 7550 Hz in the 13C dimension as indicated by exchangeable proton NMR data. The
having the 13C carrier set at the resonance frequency be- 100 ms H2O NOESY data were used to estimate the
tween that of C5 and C6 of the bases. This leads to the inter-proton distances that involve the exchangeable
out-of-phase folding of the adenine C2 resonances and protons, using imino-AH2 crosspeaks as reference.
all non-anomeric carbon resonances in ribose into an Imino-imino proton distances between base-pairs were
empty space of the spectrum (Schmieder et al., 1991). A set based on the result from 300 ms H2O NOESY con-
GARP1 composite sequence (Shaka et al., 1985) for broad nectivity walks. In total, 423 experimental distance con-
band carbon decoupling was applied during acquisition straints involving non-exchangeable and exchangeable
and a refocusing delay of 2.5 ms was set for optimal exci- protons were obtained from observed proton-proton
tation of aromatic resonances. Exponential multiplication NOEs.
was applied with 3 and 5 Hz linewidth broadening in t2
and t1, respectively, during data processing. 13C-1H Dihedral constraints
HMQC and HSQC (Bax et al., 1990; Vuister & Bax, 1992)
spectra were also taken on the complex with C6/C8 or A 30 -endo conformation was assigned to all the ribose
uniform 13C-labeling of either hairpin with an acquisition units of the complex based on the absence of H10 -H20
time of 2.5 hours. crosspeaks in a DQF-COSY spectrum and was achieved
13
C-editing NOESY experiments were done on C6/C8 by restraining the endocyclic torsion angles u0, u1, u2
or uniformly 13C-labeled samples with a spectral width and u3. The glycosidic torsion angles w of A11, U*5, U*6
of 5000 Hz in both dimensions and 1K complex and A*12 were left randomized due to the absence of U
points using the pulse sequences suggested by Otting & imino exchangeable proton NMR data. The glycosidic
WuÈthrich (1990). The delay 1/4J was tuned to 1.25 ms torsion angle w was set to be anti for all other
64 RNA Kissing Hairpins

nucleotides, because of the lack of strong H10 -H6/H8 ment of Energy (DE-FG05-86ER75281) and from the
NOE. National Science Foundation (DMB 86-09305).
For the other ®ve backbone dihedral angles, a set of
loose A-form helix constraints (Saenger, 1984) were used
to constrain the backbone of residues in the ®rst two
(Tar*) or three (Tar) base-pairs of the stem of each
References
hairpin with a range of 20 for torsion angles b, g, e Aboul-ela, F., Karn, J. & Varani, G. (1995). The structure
and a range of 30 for torsion angles a, z. No non-ex- of the human immunode®ciency virus type-1 TAR
perimental dihedral constraint was used to restrain the RNA reveals principles of RNA recognition by Tat
other residues in each hairpin: C4 to G13 in Tar16 and U*3 protein. J. Mol. Biol. 253, 313± 332.
to A*14 in Tar*16. A set of arti®cial constraints was also Batey, R. T., Inada, M., Kujawinski, E., Puglisi, J. D. &
used to keep the amino groups of the bases in the same Williamson, J. R. (1992). Preparation of isotopically
plane as the aromatic ring. A total of 60 experimental di- labeled ribonucleotides for multi-dimensional NMR
hedral constraints and 72 loose A-form helical constraints spectroscopy of RNA. Nucl. Acids. Res. 20, 4515±
were used. 4523.
Battiste, J. L., Mao, H., Rao, N. S., Tan, R., Muhandiram,
D. R., Kay, L. E., Frankel, A. D. & Williamson, J. R.
Molecular dynamics (1996). a helix-RNA major groove recognition in an
HIV-1 Rev peptide-PRE RNA complex. Science, 273,
Structure calculation on the complex was done by re- 1547± 1551.
strained molecular dynamics followed by energy mini- Bax, A., Griffey, R. H. & Hawkins, B. L. (1983). Corre-
mization using the X-PLOR program (BruÈnger, 1990) lation of proton and nitrogen-15 chemical shift by
with a force-®eld consisting of bond lengths, bond multiple quantum NMR. J. Magn. Reson. 55, 301 ±
angles, improper angles, repulsive van der Waals poten- 315.
tial and the experimental distance and dihedral angle Bax, A., Ikura, M., Kay, L. E., Torchia, D. A. &
constraints from the NMR data. Random structures were Tschudin, R. (1990). Comparison of different modes
created with random backbone torsion angles and used of two-dimensional reverse-correlation NMR for the
as the initial structures for a two-stage structure calcu- study of proteins. J. Magn. Reson. 86, 304± 318.
lation. First, a modi®ed simulated annealing protocol BruÈnger, A. T. (1990). X-PLOR: A System for Crystallogra-
(Wimberly, 1992) was used to search for global fold phy and NMR, Version 2.1 Yale University, New
structures that are consistent with secondary structure Haven, CT.
features determined by NMR using only NOE distance Calladine, C. R. & Drew, H. R. (1992). Understanding
constraints. The global fold structure was then subjected DNA, Academic Press, New York.
to the second stage structure re®nement that included all CavailleÂ, J., Nicoloso, M. & Bachellerie, J.-P. (1996). Tar-
the dihedral angle constraints in addition to the NOE geted ribose methylation of RNA in vivo directed
distance constraints. Dihedral angle constraints were by tailored antisense RNA guides. Nature, 383, 732 ±
added gradually: ®rst b, g, d, e and then a and z. This 735.
structure re®nement procedure did not change the global Chang, K.-Y. (1995). Solution structure of a bent RNA
shapes of calculated structures. The converged structures complex formed by two complementary RNA hair-
were selected based on low NOE-violation energies and pin loops. PhD thesis, University of California, Ber-
total energies of the structures (Figure 6). These con- keley.
verged structures were then subjected to the ®nal stage Chang, K.-Y. & Tinoco, I. (1994). Characterization of a
of energy minimization with attractive Lennard-Jones ``kissing'' hairpin complex from the human immu-
potentials introduced to obtain proper van der Waals node®ciency virus genome. Proc. Natl Acad. Sci.
contacts. USA, 91, 8705± 8709.
The program CURVES (Lavery & Sklenar, 1989) was Crick, F. H. & Klug, A. (1975). Kinky helix. Nature, 255,
used to analyze the helical parameters of the structure. 530.
Dickerson, R. E. (1989). De®nitions and nomenclature of
nucleic acid structure components. Nucl. Acids Res.
17, 1797± 1803.
Doudna, J. A., Cormack, B. P. & Szostak, J. W. (1989).
Acknowledgements RNA structure, not sequence, determines the 50
splice-site speci®city of a group I intron. Proc. Natl
We thank the Stable Isotope Resource at Los Alamos Acad. Sci. USA, 86, 7402± 7406.
National Laboratory for providing cells used in isolation Egli, M., Portmann, S. & Usman, N. (1996). RNA hy-
of the uniformly 13C/15N-labeled NMPs, and Dr John dration: a detailed look. Biochemistry, 35, 8489±
SantaLucia Jr for the synthesis of C6/C8 labeled 13C 8494.
NMP. We thank Professor Lynne Regan and Dr Paul Eguchi, Y. & Tomizawa, J. I. (1990). Complex formation
Predki at Yale University for their kind gift of the plas- by complementary RNA stem-loops and its stabiliz-
mid encoding the Rom protein. It is a pleasure to thank ation by a protein: function of ColE1 Rom protein.
Mr David Koh for synthesizing all the DNA oligonucleo- Cell, 60, 199± 209.
tides used throughout this work, Dr Jeffrey Pelton for Eguchi, Y. & Tomizawa, J. I. (1991). Complexes formed
help with the NMR experiments and Ms Barbara Den- by complementary RNA stem-loops: their for-
gler for managing the laboratory. We appreciate the criti- mations, structures and interaction with ColE1 Rom
cal reading of this manuscript and useful suggestions protein. J. Mol. Biol. 220, 831± 842.
from Dr Gabriele Varani. This research was supported in Fesik, S. W. & Zuiderweg, E. R. P. (1990). Heteronuclear
part by National Institute of Health grant GM 10840, by three-dimensional NMR spectroscopy of isotopically
the Department of Energy grant DE-FG03-86ER60406, labeled biological macromolecules. Quart. Rev. Bio-
and through instrumentation grants from the Depart- phys. 23, 97± 131.
RNA Kissing Hairpins 65

Gabriel, K., Schneider, J. & McClain, W. H. (1996). Func- ling constants in proteins. Biochem. Biophys. Res.
tional evidence for indirect recognition of G
U in Commun. 113, 967 ± 974.
tRNAAla by alanyl-tRNA synthetase. Science, 271, Marion, D., Ikura, M., Tschudin, R. & Bax, A. (1989).
195± 197. Rapid recording of 2D NMR spectra without phase
Gorenstein, D. G. (1981). Nucleotide conformational cycling: application to the study of hydrogen
analysis by 31P nuclear magnetic resonance exchange in proteins. J. Magn. Reson. 85, 393± 399.
spectroscopy. Annu. Rev. Biophys. Bioeng. 10, 355± Nelson, H. C., Finch, J. T., Luisi, B. F. & Klug, A. (1987).
386. The structure of an oligo(dA)
oligo(dT) tract and its
Gregorian, R. S. & Crothers, D. M. (195). Determinants biological implications. Nature, 330, 221 ± 226.
of RNA hairpin loop-loop complex stability. J. Mol. Nikonowicz, E. P., Sirr, A., Legault, P., Fiona, M. J.,
Biol. 248, 968 ± 984. Baer, L. M. & Pardi, A. (1992). Preparation of 13C
Haasnoot, C. A. G., Hilbers, C. W., van der Marel, G. A., and 15N labeled RNA for heteronuclear multi-
van Boom, J. H., Singh, U. C., Pattabiraman, N. & dimensional NMR studies. Nucl. Acids Res. 20,
Kollman, P. A. (1986). On loop folding in nucleic 4507± 4513.
acid hairpin-type structures. J. Biomol. Struct. Otting, G. & WuÈthrich, K. (1990). Heteronuclear ®lters
Dynam. 3, 843±857. in two-dimensional [1H, 1H]-NMR spectroscopy:
Haynie, S. L. & Whitesides, G. M. (1990). Preparation of combined use with isotope labeling for studies of
a mixture of nucleoside triphosphates suitable for macromolecular conformation and intermolecular
use in synthesis of nucleotide phosphate sugars interactions. Quart. Rev. Biophys. 23, 39± 96.
from ribonucleic acid using nuclease P1, a mixture Pilipenko, E. V., Poperechny, K. V., Maslova, S. V.,
of nucleoside monophosphokinases and acetate Melchers, J. G., Bruins, H. J. & Agol, V. I. (1996).
kinase. Appl. Biochem. Biotechnol. 23, 205± 220. Cis-element, oriR, involved in the initiation of (-)
Holbrook, S. R., Sussman, J. L., Warrant, R. W., Church, strand poliovirus RNA: a quasi-globular multi-
G. M. & Kim, S.-H. (1977). RNA-ligand interactions: domain RNA structure maintained by tertiary (`kis-
(1) magnesium binding sites in yeast tRNAphe. Nucl. sing') interaction. EMBO J. 15, 5428±5436.
Acids Res. 4, 2811± 2820. Predki, P. F., Nayak, L. M., Gottlieb, M. B. & Regan, L.
Holbrook, S. R., Sussman, J. L. & Kim, S.-H. (1981). (1995). Dissecting RNA-protein interactions: RNA-
Absence of correlation between base-pair sequence RNA recognition by Rop. Cell, 80, 41±50.
and RNA conformation. Science, 212, 1275± 1277. Puglisi, J. D., Tan, R., Canlan, B. J., Frankel, A. D. &
Kellogg, G. W., Szewczak, A. A. & Moore, P. B. (1992). Williamson, J. R. (1992). Conformation of the TAR-
arginine complex by NMR. Science, 257, 76± 80.
Two-dimensional hetero-TOCSY-NOESY-correlation
Saenger, W. (1984). Principles of Nucleic Acid Structure,
of 31P resonances with anomeric and aromatic 1H
Springer-Verlag, New York.
resonances in RNA. J. Am. Chem. Soc. 114, 2727±
SantaLucia, J. S., Jr, Shen, L. X., Cai, Z., Lewis, H. &
2728.
Tinoco, I., Jr (1995). Synthesis and NMR of RNA
Klug, A., Jack, A., Viswamitra, M. A., Kennard, O.,
with selective isotopic enrichment in the base
Shakked, Z. & Steitz, T. A. (1979). A hypothesis on
moieties. Nucl. Acids Res. 23, 4913± 4921.
a speci®c sequence-dependent conformation of
Schmieder, P., Zimmer, S. & Kessler, H. (1991).
DNA and its relation to binding of the lac-repressor Increased resolution in proton detected heteronuc-
protein. J. Mol. Biol. 131, 669± 680. lear NMR experiments by folding in the hetero-
KoÈnig, P., Giraldo, R., Chapman, L. & Rhodes, D. dimension. Magn. Reson. Chem. 29, 375 ±380.
(1996). The crystal structure of the DNA-binding Shaka, A. J., Barker, P. B. & Freeman, R. (1985). Compu-
domain of yeast RAP1 in complex with telomeric ter-optimized decoupling scheme for wide band
DNA. Cell, 85, 125± 136. applications and low-level operation. J. Magn.
Koudelka, G. B. & Carlson, P. (1992). DNA twisting and Reson. 64, 547 ± 552.
the effects of non-contacted bases on af®nity of 434 Sklenar, V. & Bax, A. (1987). Spin-echo water suppres-
operator for 434 repressor. Nature, 355, 89 ±91. sion for generation of pure-phase two-dimensional
Koudelka, G. B., Harrison, S. C. & Ptashne, M. (1987). NMR spectra. J. Magn. Reson. 74, 469 ± 479.
Effect of non-contacted bases on the af®nity of 434 Sklenar, V., Miyashiro, H., Zon, G. & Bax, A. (1986).
operator for 434 repressor and Cro. Nature, 326, Assignment of the 31P and 1H resonances in oligo-
886 ± 888. nucleotides by two-dimensional NMR spectroscopy.
Koudelka, G. B., Harbury, P., Harrison, S. C. & Ptashne, FEBS Letters, 208, 94± 98.
M. (1988). DNA twisting and the af®nity of bac- Somers, W. S. & Phillips, S. E. (1992). Crystal structure
teriophage 434 operator for bacteriophage 434 of the met repressor-operator complex at 2.8 A Ê res-
repressor. Proc. Natl Acad. Sci. USA, 85, 4633± 4637. olution reveals DNA recognition by b-strands.
Lavery, R. & Sklenar, H. (1989). De®ning the structure Nature, 359, 387± 393.
of irregular nucleic acids: conventions and Strobel, A. A. & Cech, T. R. (1995). Minor groove recog-
principles. J. Biomol. Struct. Dynam. 6, 655 ± 667. nition of the conserved G
U pair at the Tetrahymena
Macura, S., WuÈthrich, K. & Ernst, R. R. (1982). Separ- ribozyme reaction site. Science, 267, 675± 679.
ation and suppression of coherent transfer effects in Tomizawa, J. I. (1990). Control of ColE1 plasmid replica-
two-dimensional NOE and chemical exchange tion: interaction of Rom protein with an unstable
spectroscopy. J. Magn. Reson. 46, 269 ± 282. complex formed by RNAI and RNAII. J. Mol. Biol.
Marino, J. P., Razmic, S. G., Csanlovszki, G., Jr & 212, 695 ± 708.
Crothers, D. M. (1995). Bent helix formation Tomizawa, J. I. (1993). Evolution of functional structures
between RNA hairpins with complementary loops. of RNA. In The RNA World (Gesteland, R. F. &
Science, 268, 1448± 1454. Atkins, J. F., eds), pp. 419± 445, Cold Spring Harbor
Marion, D. & WuÈthrich, K. (1983). Application of phase Laboratory Press, Cold Spring Harbor, NY.
sensitive two-dimensional correlated spectroscopy Tomizawa, J. I., Eguchi, Y. & Itoh, T. (1991). Antisense
(COSY) for measurement of 1H-1H spin-spin coup- RNA. Annu. Rev. Biochem. 60, 631± 652.
66 RNA Kissing Hairpins

Turner, D. H., Sugimoto, N. & Freier, S. M. (1990). Ther- its relationship to A-tract DNA. Proc. Natl Acad. Sci.
modynamics and kinetics of base-paring and of USA, 85, 6332± 6336.
DNA and RNA self-assembly and helix coil Yuan, H., Quintana, J. & Dickerson, R. E. (1992).
transition. In Nucleic Acids, pp. 201 ±227, Springer- Alternative structures for alternating poly(dA-dT)
Verlag, Berlin. tracts: the structure of the B-DNA decamer C-G-A-
Varani, G. & Tinoco, I., Jr (1991a). RNA structure and T-A-T-A-T-C-G. Biochemistry, 31, 8009± 8021.
NMR spectroscopy. Quart. Rev. Biophys. 24, 479 ±
532.
Varani, G. & Tinoco, I., Jr (1991b). Carbon assignments Edited by P. E. Wright
and heteronuclear coupling constants for an RNA
ologonucleotide from natural abundance 13C-1H
correlated experiments. J. Am. Chem. Soc. 113, 9349±
9354. (Received 2 December 1996; received in revised form
Vuister, G. W. & Bax, A. (1992). Resolution enchance- 28 February 1997; accepted 4 March 1997)
ment and spectral editing of uniformly 13C-enriched
13
proteins by homonuclear broadband C
decoupling. J. Magn. Reson. 98, 428± 435.
Wagner, E. G. H. & Simons, R. W. (1994). Antisense
RNA control in bacteria, phage, and plasmids.
Annu. Rev. Microbiol. 48, 713 ±742.
Wimberly, B. (1992). NMR-derived structures of RNA
loops: the conformation of eukaryotic 5 S ribosomal
loop E. PhD thesis, University of California, Berke- http://www.hbuk.co.uk/jmb
ley.
Yoon, C., Prive, G. G., Goodsell, D. S. & Dickerson, R. E. Supplementary material comprising one Table is
(1988). Structure of an alternating-B DNA helix and available from JMB Online