ABSTRACT
An RNA oligonucleotide that contains the binding site for
Escherichia coli
ribosomal protein S8 was prepared with uniform
15
N isotopic enrichment and uniform deuterium enrichment at all non-exchangeable sites using enzymatic methods. The RNA binding site, which
contains 44 nt, forms a hairpin in solution and requires Mg
2+
for proper folding. The longitudinal magnetization recovery rates of the
exchangeable protons were compared for the [
2
H,
15
N]-enriched RNA molecule and for the corresponding fully [
2
H,
15
N]-enriched RNA hairpin. It was found that
1
H-
1
H dipolar relaxation significantly contributes to the recovery of exchangeable
proton longitudinal magnetization. The exchangeable proton resonance line
widths were less affected by deuteration, indicating that chemical exchange
with H
2
O remains the dominant mechanism of transverse magnetization relaxation.
Nevertheless, deuteration of this RNA hairpin was found to enhance the
sensitivity of NOE-based experiments relative to the fully protonated hairpin and to simplify
2D NMR spectra. The increased signal-to-noise ratio facilitated the assignment of the cytidine amino
resonances and several of the purine nucleotide amino resonances and permitted
the identification of NOE crosspeaks that could not be observed in spectra of
the fully protonated RNA hairpin.
Heteronuclear multi-dimensional NMR methods have extended the size and complexity of proteins
and nucleic acids that can be studied in solution by improving spectral
resolution and permitting the correlation of resonances through scalar coupled
pathways. In particular, NOE-based experiments have significantly benefited from the enhanced
resolution that is provided through
13
C and
15
N isotopic enrichment. The
13
C and/or
15
N separated NOESY experiments provide much structural information and are
routinely applied to obtain high resolution structures of proteins up to ~22 kDa and are beginning to be applied to RNA molecules as large as 12-15 kDa (
1
,
2
). However, the sensitivity of NOE-based experiments can be dramatically reduced by line broadening
associated with
1
H dipolar relaxation, thus limiting the amount of structural information that
can be obtained from these experiments.
The ill effects of dipolar relaxation generally become more pronounced as the
effective correlation time of a nucleus increases and as the number of proximal
relaxation partners of a proton increase. These effects are not restricted to
large molecules, however. Even moderately sized proteins and oligonucleotides
that exhibit long correlation times can suffer the problems associated with
dipolar relaxation. Proteins and oligonucleotides that are components of
protein-nucleic acid complexes exhibit correlation times characteristic of the
complexes which can be significantly longer than the correlation times of
either of the individual components free in solution. Mg
2+
, which is often required for the proper folding of RNA molecules (
3
,
4
), may also act to slow the correlation time of RNA oligonucleotides. Free Mg
2+
can promote non-specific intermolecular interactions among RNA molecules which result in
an increase in the molecular correlation time and a decrease in spectral
sensitivity.
Several investigations have shown that deuteration can be used to reduce the
adverse effects associated with
1
H dipolar relaxation (
5
-
7
). Recently, Venters and co-workers (
8
) and Grzesiek
et al.
(
9
) have demonstrated that perdeuteration of the non-exchangeable sites in
15
N-labeled proteins improves the sensitivity and resolution of
15
N separated NOESY experiments, permitting structure analysis of slowly tumbling
proteins. The line broadening effects associated with dipolar relaxation
exhibited by the amide protons are significantly reduced when the number of
relaxation partners surrounding them is decreased. The narrower resonances
ultimately result in improved sensitivity of experiments that employ deuterated
proteins. It has recently been shown that selective deuteration within the
sugar moieties of DNA (
10
) and RNA (
11
) oligonucleotides decreases the transverse relaxation rates of the remaining
non-exchangeable protons. The more favorable relaxation rates and the
elimination of
1
H scalar coupling result in decreased line widths and improved spectral
resolution for the remaining proton resonances (
10
-
1
2
).
The exchangeable imino and amino proton resonances of nucleic acids tend to be
broad. Importantly, many of the
1
H-
1
H NOE interactions that report on the secondary and tertiary structural features
of these molecules involve the exchangeable protons (
13
,
14
). Although chemical exchange processes contribute significantly to the line
widths of the exchangeable resonances (
15
,
16
),
1
H dipolar interactions can also contribute. Deuteration of the non-exchangeable sites offers a method to eliminate several of the dipolar
relaxation partners of the exchangeable protons, thus providing enhanced
sensitivity and improved spectral quality.
In this report, we compare the NMR spectral properties of the exchangeable
protons of a deuterated,
15
N-enriched RNA molecule and the corresponding fully protonated,
15
N-enriched molecule. The oligoribonucleotide used in this study forms a
hairpin in solution and contains the binding site for ribosomal protein S8
(Fig.
1
). The apparent longitudinal relaxation times of several of the imino protons of
the deuterated RNA hairpin were found to be longer than the corresponding
resonances of the protonated RNA hairpin. In addition, NOE-based spectra of the deuterated RNA hairpin exhibited increased
sensitivity and a greater number of crosspeaks between exchangeable protons.
Finally, spectra from the deuterated RNA hairpin were used to identify several
NOE crosspeaks involving exchangeable proton resonances that could not be
obtained from spectra of the fully protonated RNA hairpin.
All enzymes were purchased from Sigma Chemical Co. (St Louis, MO) with the
exception of the T7 RNA polymerase. The T7 RNA polymerase was prepared as
previously described (
17
). Deoxyribonuclease I type II, pyruvate kinase, adenylate kinase and the
nucleotide monophosphate kinase were obtained as powders and were dissolved in
solutions of 15% glycerol, 1 mM dithiothreitol and 10 mM Tris-HCl, pH 7.4, and stored at -20oC. The guanylate kinase and nuclease P
1
were obtained as solutions and stored at -20oC. Phosphoenolpyruvate (potassium salt) was obtained from Bachem.
The labeled RNA hairpin was transcribed
in vitro
with T7 RNA polymerase using a synthetic DNA template as previously described (
18
,
19
) and
2
H/
15
N-labeled 5'-nucleoside triphosphates (5'-NTPs). The 5'-NTPs were prepared by growing
Escherichia coli
on minimal medium containing 88%
2
H
2
O, [
15
N]ammonium sulfate as the sole nitrogen source and [
2
H
3
]sodium acetate as the sole carbon source. Starter cultures were prepared in LB medium
and transferred to 25 ml minimal medium containing ammonium sulfate (5 mM) and
sodium acetate (40 mM). The cultures were passed sequentially through 25 ml
medium containing 0, 50, 75 and 88%
2
H
2
O and 90%
2
H
2
O/100% [
2
H]sodium acetate, each time being allowed to reach an OD A
600
of 0.8 before being transferred. The final culture contained 55 ml medium and
was used to inoculate 11 cultures in 320 ml 88%
2
H
2
O/100% [
2
H]sodium acetate. The 320 ml cultures were harvested when an OD of 2.0 (A
600
) was reached and yielded 18.3 g wet cell paste. rRNA was extracted and the
labeled 5'-NTPs were prepared as previously described (
19
); 7330 A
260
OD units were obtained.
Two 12 ml
in vitro
transcriptions were carried out as described (
18
,
19
) using either
15
N-labeled 5'-NTPs or
2
H/
15
N-labeled 5'-NTPs. The PAGE purified RNA molecules were dissolved in 1.0 M
NaCl, 20 mM potassium phosphate, pH 6.8, and 2.0 mM EDTA and dialyzed
extensively against 10 mM NaCl and 10 mM potassium phosphate, pH 6.8, using a
Centricon-3 concentrator (Amicon Inc.). The sample was diluted with buffer to a
volume of 200 [mu]l and annealed. MgCl
2
was added to a concentration of 30 mM and the samples were dialyzed twice
against a buffer of 10 mM NaCl and 10 mM potassium phosphate, pH 6.8, and 12.5
mM MgCl
2
. The RNA was diluted to a volume of 200 [mu]l using the final dialysis buffer and lyophilized to a powder. The samples
were resuspended in 90% H
2
O/10%
2
H
2
O to a concentration of 75 A
260
OD units in 220 [mu]l (~0.85 mM).
Spectra were acquired on a Bruker AMX-500 spectrometer equipped with a
1
H-{X} reverse detection probe. Broadband decoupling of the imino nitrogen
resonances was achieved using GARP (
20
) ([gamma]B
2
= 1570 Hz). Solvent suppression was achieved using either spin lock pulses (
21
) or binomial 1
1
or 1
3
3
1
read pulses (
22
) with maximum excitation at 12.5 p.p.m. and were acquired at 12oC. Quadrature detection was achieved using the States-TPPI method and acquisition was delayed by a half-dwell in all indirectly detected dimensions.
1
H longitudinal relaxation measurements were performed as 2D HSQC experiments as
reported elsewhere (
23
), except that the final read pulse was of the binomial type and the H
2
O resonance was not presaturated. Nine experiments with longitudinal
magnetization recovery delays ranging from 20 to 800 ms were acquired for each
molecule and each experiment was repeated five times. The spectral widths were [omega]
1
= 1500 Hz and [omega]
2
= 12 205 Hz and the system was allowed 3.5 s to recover between scans. The
spectral widths for the 2D NOESY experiments were [omega]
1
= 5500 Hz and [omega]
2
= 12 205 Hz. The 2D
15
N [omega]
1
half-filtered NOESY experiments (
24
) were collected at a mixing time ([tau]
m
) of 260 ms and the filter delay was set to 10.4 ms, which is equal to 1/J
HN
. For the 3D {
15
N}-HSQC-NOESY experiments, [tau]
m
= 240 ms, the
15
N-
1
H anti-phase magnetization was allowed to develop for 5.0 ms and spectral widths
were [omega]
1
= 1200 Hz, [omega]
2
= 5500 Hz and [omega]
3
= 12 205 Hz.
2
H decoupling was not employed in the 2D and 3D experiments.
Problems associated with low signal-to-noise ratios and resonance overlap limit the analysis of NMR spectra
of large or slowly tumbling molecules. Complete deuteration of the non-exchangeable sites of proteins facilitates the structure determination of large proteins by reducing dipolar relaxation-associated line broadening of the amide proton resonances (
8
,
9
,
25
). An analogous isotopic enrichment scheme has been applied to the RNA binding
site for ribosomal protein S8 (Fig.
1
), leaving only the exchangeable imino and amino protons available for
1
H NMR studies. The exchangeable protons participate in hydrogen bonding networks
that mediate base-base interactions in nucleic acids and are the source of NOEs for
establishing RNA secondary structure. NOEs involving these protons can also
provide more global structural information if tertiary interactions can be
identified, such as those involving coaxial stacked helices (
14
). Thus, NOE-based experiments involving the exchangeable proton resonances of
ribonucleic acids are of particular interest for RNA structure studies.
The procedure for preparation of the perdeuterated RNA is similar to published
procedures for preparation of
13
C- and
15
N-enriched RNAs (
19
,
26
,
27
). Uniformly
15
N/
2
H-labeled 5'-NMPs were prepared from labeled rRNA that was extracted from
E.coli
cultured in medium of 90%
2
H
2
O with [
15
N]ammonium sulfate and [
2
H]sodium acetate as the sole nitrogen and carbon sources. Cells were grown on
11.0 g [
2
H]sodium acetate and yielded 7330 A
260
absorbance units (~6.60 * 10
-4
mol using an average extinction coefficient of [epsilon]
260
= 11 100) of rRNA. This yield is ~65% less per g carbon source than we have obtained from cells cultured
using glucose and H
2
O, however, the reduced yield is offset by the lower per mol cost of [
2
H]sodium acetate relative to [
2
H]glucose. The fraction of
2
H enrichment was estimated from one-dimensional
13
C NMR spectra. All of the ribose carbon positions exhibited a 90-95% level of deuterium enrichment. The C2 (adenine) and C8 (purine)
positions contained 85 and 90% deuterium respectively and the C6 and C5
positions of the pyrimidines contained 90 and 88% deuterium respectively.
The secondary structure of the RNA oligonucleotide used in this study is
depicted in Figure
1
. Stabilization of the RNA tertiary structure and complex formation between the
RNA and ribosomal protein S8 require a Mg
2+
concentration of ~15 mM. The combination of molecular size and divalent cation concentration
extends the correlation time of the free RNA hairpin, resulting in broad imino
proton resonances (~33 Hz) and limiting the sensitivity of the NMR experiments.
Recovery of longitudinal magnetization of the imino protons after inversion was
measured for the deuterated and fully protonated RNA hairpins. Although the
observed rates of recovery contain contributions from both proton dipolar
interactions and from chemical exchange (
15
), the relative rates of recovery between corresponding protons within the
deuterated and protonated hairpins reflect primarily differences originating
from proton dipolar interactions. Figure
2
compares the distribution of imino proton recovery times for the two molecules
at the same concentration and under the same buffer conditions. The recovery
data were fitted to single exponential decays to provide a qualitative
comparison between the two molecules. This method is supported by the high
quality of fits achieved, as determined by calculated values of [chi]
2
(typical values were <0.5). The observed recovery times for protons of the deuterated hairpin ranged
between 20 and 40% longer than those of the fully protonated RNA molecule (33%
longer on average). Only
1
H dipolar interactions contribute significantly to the different relaxation
rates, since deuteration is not expected to influence the chemical exchange processes. Deuteration of the non-exchangeable sites in proteins has similarly been shown to decrease the longitudinal
relaxation rates of the amide protons by almost 35% relative to those of the
fully protonated protein (
25
). Thus, deuteration of the non-exchangeable sites in RNA molecules reduces the number of dipolar
interactions of the imino protons and results in a decrease in the recovery
rate of longitudinal magnetization of the imino protons even in the presence of
chemical exchange.
Several NOE-based experiments were performed using the S8 RNA hairpin to investigate
the effects of complete deuteration on the spectroscopic properties of medium
sized RNA molecules. Figure
3
a-d compares the imino regions of 260 ms mixing time NOESY spectra of the
deuterated,
15
N-enriched and fully protonated,
15
N-enriched RNA hairpins. Both crosspeaks and diagonal peaks in the spectrum
of the deuterated molecule exhibit 2- to 5-fold higher signal-to-noise ratios than the corresponding peaks of the fully
protonated RNA molecule. However, the line widths of the peaks are similar in
the spectra of both RNA molecules, indicating that the increased signal-to-noise ratio exhibited by the deuterated RNA hairpin results from a
reduced number of cross-relaxation pathways, rather than decreased transverse relaxation rates of
the protons in the deuterated molecule. Figure
4
compares one-dimensional vectors along [omega]
1
of the 260 ms NOESY spectra at the imino proton frequency of U591H3. In order
to ensure that the sensitivity differences observed between the two spectra at
260 ms were not significantly influenced by the NOE build-up rate, 240 and 280 ms mixing time NOESY spectra of the fully protonated
RNA hairpin were also collected. The difference in the crosspeak intensities is
greater between the latter spectra of the fully protonated RNA molecule and the
260 ms NOESY spectrum of the deuterated RNA molecule. This suggests that the
sensitivity improvement results from the reduced number of cross-relaxation pathways available to the imino protons in the deuterated
molecule rather than a more rapid decay of the NOEs in the fully protonated
molecule. Thus, deuteration of the non-exchangeable sites limits cross-relaxation to a few pathways and enhances the sensitivity of the
exchangeable proton NOESY spectrum.
The exchangeable protons can provide valuable information for a variety of
structural features, including non-standard base pairs (
28
), base triples (
29
,
30
) and coaxial stacked helices (
14
). However, the crosspeaks between imino and amino protons can be difficult to
interpret because of resonance overlap with non-exchangeable proton resonances. The NOESY spectrum of the deuterated RNA
hairpin contains crosspeaks involving only the
15
N-bound exchangeable protons, which markedly simplifies the imino region of
the spectrum (Fig.
3
a and b). An [omega]
1
-filtered NOESY experiment must be employed to achieve comparable spectral
simplification using the fully protonated RNA hairpin. However, the sensitivity
of this experiment is considerably reduced relative to a standard NOESY
experiment, due to the effects of chemical exchange and relaxation during the
filter delay periods, and results in the loss of most crosspeaks in this
molecule (data not shown).
The RNA binding site for ribosomal protein S8 consists of two helical regions
that flank a core of 9 nt that are evolutionarily conserved (
31
). Core residues A596, G597, C643 and U644 are required for RNA-protein S8 association and have been proposed to form A[middot]U and G[middot]C base pairs (
31
,
32
). Crosspeaks between the imino protons of adjacent base pairs in helices
facilitate the sequence-specific resonance assignment of the imino protons. The NOE between G597H1
and U644H3 is critical for identification of the G597[middot]C643 base pair, but this interaction is too weak to be observed in the
NOESY spectrum of the fully protonated RNA hairpin (Fig.
3
). The G597H1-U644H3 NOE is present in the spectrum of the deuterated RNA hairpin and
helps to confirm that G597 and C643 form a base pair. The signal-to-noise ratio of the NOESY spectrum of the fully protonated hairpin,
and hence the possibility of identifying the G597-U644 NOE, can be improved by increasing the number of scans in the 2D
experiment, but the time required to achieve the necessary improvement is
impractical.
Comparative phylogenetic analysis studies predict that residues A595, A596 and
U644 form an A[middot](A[middot]U) base triple (
31
) and NMR spectral evidence supporting the base triple has recently been
reported (
33
). The NOESY spectrum of the deuterated hairpin contains crosspeaks between the
amino protons of A596 and the imino proton of U644. Importantly, both of the
A596 amino protons resonate in the low field region of the spectrum, indicating
that both protons are involved in base-base interactions. These NOEs cannot be identified in spectra of the
fully protonated RNA hairpin. The NOESY spectrum of the deuterated RNA hairpin
also contains a weak crosspeak between the imino resonance of U644 and the H2
resonance of A596 (Fig.
3
a). This crosspeak is more easily identified in the spectrum of the fully
protonated RNA hairpin. Together, these interactions are critical for
confirming the presence of the A[middot](A[middot]U) base triple in the RNA hairpins containing the complete
binding site for protein S8.
15
N separated 2D and 3D NOE experiments provide a means to resolve crowded regions
of the exchangeable proton spectrum. We have assigned the cytidine amino
resonances of the deuterated RNA hairpin (Table
1
) using a 3D HSQC-NOESY experiment that was optimized for detection of the imino protons. In
this experiment, the imino-amino proton crosspeaks of the NOESY spectrum were labeled in the third
dimension with their corresponding amino nitrogen chemical shifts. Figure
5
a and b shows [omega]
2
-[omega]
3
NOE planes of the 3D spectrum at the amino nitrogen chemical shifts
corresponding to C637 and C651. The intra-base pair imino-amino crosspeaks of all G[middot]C base pairs, except G604[middot]C634 and G598[middot]C643, which exhibit weak imino proton
resonances, are present in the 3D spectrum. Many of the crosspeaks involving
the adenine and guanine amino resonances are not readily observed in this
experiment. However, the 2D version of this experiment permits identification
of several intra-base pair and inter-base pair crosspeaks involving purine amino protons (Fig.
5
c). The crosspeak between A596N6 and U644H3 helps to establish the participation
of A596 in the A[middot](A[middot]U) base triple by providing a means to assign NOEs that are
observed between amino protons and U644H3 in the NOESY spectrum (Fig.
3
a). These and other imino-amino interactions are not readily distinguished in the NOESY spectrum,
but are resolved and can be assigned using the
15
N separated NOESY spectrum. Analogous HSQC-NOESY experiments acquired using the fully protonated RNA hairpin exhibit
poor sensitivity and could not be used to obtain these correlations.
The variety and complexity of RNA systems amenable to high resolution NMR
structural analysis has dramatically increased with the introduction of
13
C and
15
N heteronuclear NMR methods, in part by increasing the spectral dispersion. We
have shown that perdeuteration of the non-exchangeable sites of RNA oligonucleotides can improve the quality of NOE-based spectra involving the exchangeable proton resonances.
Perdeuteration decreases the longitudinal relaxation rates of the imino protons
but has little effect on the exchangeable proton line widths, suggesting that
chemical exchange dominates the apparent transverse relaxation rates.
Nevertheless, deuteration results in an increased signal-to-noise ratio of NOE-based experiments and permits extraction of additional
structural information. The incorporation of deuterium also simplifies the
exchangeable region of the NOESY spectrum without the need to apply
15
N or
13
C filters, which can reduce sensitivity. Further, the residual protonation of
adenine C2 permits the identification of A[middot]U base pairs through the AH2-UH3 crosspeak in the deuterated molecule. Although chemical exchange processes remain active in the
deuterated molecule, the NOE crosspeaks contain fewer contributions from spin
diffusion, which should provide the opportunity to extract more accurate inter-proton distances. We are beginning efforts to explore the utility of RNA
deuteration in the protein S8-RNA complex, where the molecular correlation time is significantly
increased.
We thank Dr R.Zimmermann for helpful discussions. This work was supported by
grants from the Welch Foundation (C-1277) and the National Institutes of Health (GM52115) to E.P.N.
REFERENCES
Return