ABSTRACT
The so-called spine of hydration in the minor groove of AnTn tracts in DNA is
thought to stabilise the structure, and kinetically bound water detected in the
minor groove of such DNA species by NMR has been attributed to a narrow minor groove
[Liepinsh, E., Leupin, W. and Otting, G. (1994)
Nucleic Acids Res.
22
,
2249-2254]. We report here an NMR study of hydration of an RNA dodecamer which
has a wide, shallow minor groove. Complete assignments of exchangeable protons, and a large number of non-exchangeable protons in r(CGCAAAUUUGCG)
2
have been obtained. In addition, ribose C2
'
-OH resonances have been detected, which are probably involved in hydrogen bonds. Hydration at
different sites in the dodecamer has been measured using ROESY and NOESY experiments at 11.75 and 14.1 T. Base protons in both the major and minor grooves are in
contact with water, with effective correlation times for the interaction of
~
0.5 ns, indicating weak hydration, in contrast to the hydration of adenine C2H in the homologous DNA sequence. NOEs to H1'
in the minor groove are consistent with hydration water present that is not
observed in the analogous DNA sequence. Hydration kinetics in nucleic acids may be determined by chemical factors such as hydrogen-bonding more than by simple conformational factors such as groove width.
Nucleic acids are strongly hydrated in both the crystal state and in solution
and the degree of hydration determines the overall conformation. Hydration in
DNA has been measured by a variety of methods including NMR (
1
-
4
), thermodynamics (
5
), and crystallography (
6
,
7
). Thus at low water activity, DNA adopts the A conformation, whereas at high
water activity, it is found in the B conformation. In contrast, RNA is always in the A conformation. Chemically, the important difference between DNA and RNA is the 2'-OH on the sugar in RNA, though the methyl group of thymine affects primarily the thermodynamic stability of duplexes of either RNA or DNA (
8
).
Although there are numerous X-ray structures of DNA in both the B and A forms, there are no high
resolution solution structures of DNA in the A form, and only a few solution
structures of RNA. Many of the crystal structures have been reported on the
hydration of DNA and RNA, and the results have been summarised in detail (
9
). In contrast, there have been only a few reports of hydration of nucleic acids using NMR (
1
-
4
,
10
) and none on RNA in solution.
The concentration of water is very high (~55 M), and therefore statistically can be expected to be close to all
exposed sites on the surface of a solute particle. In the absence of any
interaction between water and solute, the rate constant for dissociation of a
water molecule from a solute surface depends on the diffusion constant, and can be expected to be ~2-3 * 10
9
s
-1
at 10oC (i.e. an average residence time of ~0.3-0.5 ns) (
11
). The diffusion limited association rate constant will be of similar magnitude,
i.e. 10
9
-10
10
M
-1
s
-1
, so that the effective dissociation constant is ~0.2-2 M. The free energy change is comparable with the free energy of mixing, and represents the case where water is not
thermodynamically bound. However, at 55 M water, the occupancy of each exposed site will be >96%. Thus, water can be considered to be bound if
the effective correlation time is significantly longer than ~0.5 ns.
A particularly important structural role for water has been proposed in the
spine of hydration found in the narrowed minor groove of dAnTn sequences (
6
,
7
) which has been proposed to be responsible for the relatively slow exchange of
NMR-visible water molecules from the neighbourhood of the adenine C2H (
1
,
2
). Because an RNA duplex has a very wide, shallow minor groove, the groove-width hypothesis predicts that there should be no slowly exchanging water
molecules in the minor groove. However, it is possible that other factors such
as hydrogen bonding are important. Thus, the minor groove of RNA is lined with
C2'-OH groups, and is therefore relatively more hydrophilic than the
minor groove of DNA. It would therefore be expected that the minor groove of
RNA should show extensive contact with water, particularly around the C2' positions. We have therefore examined the hydration of r(CGCAAAUUUGCG)
2
. The hydration properties of this molecule can be directly compared with the DNA analogue
d(CGCAAATTTGCG)
2
, which has been extensively studied by X-ray diffraction (
12
) and NMR methods (
13
).
r(CGCAAAUUUGCG) was synthesised using phosphoramidite chemistry and purified by
reversed-phase HPLC (
14
). A quantity of 112 A
260
units of the purified and annealed dodecamer
were dissolved in 0.6 ml 90% H
2
O:10% D
2
O containing 0.01 M sodium phosphate, 100 mM KCl, 0.2 mM EDTA and 0.1 mM DSS, pH
7.
1
H NMR spectra were recorded at 14.1 T on a Varian Unity NMR spectrometer and at
11.75 T on a Varian UnityPlus spectrometer. 2D NMR spectra were recorded in the phase-sensitive mode (
15
). Spectra in H
2
O were recorded using the Watergate pulsed gradient method for solvent
suppression (
16
) with acquisition times of 0.4 s in t
2
and 0.05 s in t
1
. NOESY spectra were obtained using mixing times of 25, 50, 100 and 250 ms. A
ROESY spectrum was obtained with a mixing time of 25 ms, and a spin-lock field strength of 4.5 kHz.
NOESY spectra in D
2
O were recorded at 30oC with acquisition times of 0.7 s in t
2
and 0.06 s in t
1
, with mixing times of 50, 100 and 250 ms. A 2-quantum COSY experiment (
17
) was recorded with a mixing pulse of 135o, acquisition times of 0.5 s in t
2
and 0.04 s in t
1
, with a 2-quantum creation time of 40 ms. In this experiment, cross-peaks are disposed either side of the two-quantum diagonal, which is free of peaks. Further, the 135o pulse suppresses the remote connectivities. The spectrum produced in this
experiment was phased to pure absorption along F2, and absolute value along F1. Data
matrices were transformed as 16384 by 2048 complex points, using a Gaussian function for apodisation in both dimensions.
Hydration was assessed by observing both NOEs and ROEs from water to different
protons in cross-sections along F2 at the water frequency. In the absence of spin
diffusion, ROEs are positive, and the cross-peaks have the opposite sign to the diagonal. Chemical exchange peaks in
ROESY spectra are negative, and have the same sign as the diagonal. In NOESY
spectra, exchange cross-peaks are negative, and in general, NOEs are also negative unless the
effective correlation time is short (less than ~0.3 ns). The ratio of the cross-relaxation rate constants in the laboratory and rotating frames, R,
depends only on correlation times and (known) Larmor frequencies. Thus, for a
tightly bound water molecule (i.e. for which the residence time is longer than
the rotational correlation time), the cross relaxation rate constants in the
two experiments are given by:
[sigma]
1
= a[6J(2[omega]) - J(0)]/r
6
1
[sigma]
r
= a[3J([omega]) + 2J(0)]/r
6
2
R = [6J(2[omega]) - J(0)]/[3J([omega]) + 2J(0)]
3
where a is a nuclear constant and J([omega]) are the spectral density functions.
Equation
3
predicts that the ratio R, will approach -0.5 for long correlation times, as in macromolecules. If the bound water
molecule can undergo rapid, large amplitude motions, the spectral density
functions become more complex. For sufficiently rapid internal motions, each spectral density function can be decomposed into two parts, corresponding to overall rotation of the complex, and internal
mobility as (
18
):
J([omega],[tau]) = S
2
J([omega],[tau]
0
) + (1 - S
2
)J([omega],[tau]
e
)
4
where S
2
is the order parameter, [tau]
e
= [tau]
0
[tau]
i
/([tau]
0
+[tau]
i
) and [tau]
0
, [tau]
i
are the correlation times for global rotation and internal motion,
respectively. If the internal motion is of large amplitude, such that S
2
= 0, the form of the spectral density function reduces to that of a rigid rotor,
except that it is determined by [tau]
e
rather than [tau]
0
. In this case, the value of R can vary between unity (very short correlation
times) and -0.5 (long correlation times). This allows NOEs to become positive. An
alternative is that water molecules are not strongly bound, but diffuse in and
out of the molecular surface on a time-scale comparable with the global correlation time. Models of translation
diffusion give quite complex spectral density functions (
19
), but the ratio, R, still depends only on the correlation time and Larmor
frequency (and not on distances) (
19
-
21
). For this model, the correlation time is related to the diffusion coefficient
of the diffusing water molecule, which may be smaller than that of pure water
if there is an interaction between the water molecules and the solute particle.
Hence, these models can provide limits on the effective correlation time for water-solute interactions as probed by NMR.
Experimentally, the ratio R was determined by measuring cross peak areas in cross-sections at the water frequency, and normalising them to the intensity of Cyt or Uri H5-H6 cross peaks. This eliminates the effects of autorelaxation and non-uniform excitation with the Watergate pulse. Corrections were made for off-resonance effects in the ROESY spectra as described (
22
,
23
). Effective correlation times were then determined using equation
3
for both rotational and translational diffusion models. Where absolute volume ratios were used, corrections were also made for the Watergate excitation
profile, which is particularly important for the H1' resonances.
Magnetisation time courses were calculated either analytically for three-spin systems, or by numerical integration of the appropriate Bloch
equations for complete spin-systems, using the program NUCFIT (
24
). Coordinates were obtained for energy-minimised RNA using DISCOVER (Molecular Simulations, San Diego).
Calculations were carried out assuming a correlation time of 5 ns and a recycle
time of 3.5 s. The complete spin-system calculations also provided values for the autorelaxation rate
constants of each proton.
Non-exchangeable protons were assigned using a combination of NOESY and DQF-COSY and 2-quantum COSY in D
2
O solution. Figure
1
shows
a NOESY spectrum in D
2
O recorded at 14.1 T and 30oC. It was possible to trace the sequential base-H1'-base proton connectivities. In addition, the adenine
C2H resonances (and see below) showed three cross-peaks. The weak cross-peak corresponds to the intraresidue interaction with H1', and the two strong peaks correspond to the sequential H2(i)-H1'(i+1) and the cross-strand H2(i)-H1-(i-1) interactions. The latter cross-peak was more
intense than expected for standard A-RNA, where this distance is ~4.5 Å, compared with the sequential H2-H1' distance of ~3.6 Å. The H2' resonances were assigned
using the strong H1'-H2' NOE cross-peaks, and also the sequential pathways H1'(i)-H2'(i)-H8/H6(i+1) pathway. H3' peaks were
assigned from both the NOESY, using the intraresidue H8/6(i)-H'(i) and sequential H3'(i)-H8/H6(i+1) cross-peaks, and the H2'-H3' cross-peaks in the DQF-COSY and DQ-COSY
spectra (not shown). The H4' resonances were initially assigned from traces through H1' of NOESY spectra. The most intense peak in this cross-section is the H2', followed by the H4' and H3' peaks. The H4' resonance was assigned by
elimination, and by H3'-H4' cross-peaks in the DQF-COSY spectrum. The H3' and H4' resonances were confirmed using
the 2-quantum COSY experiment, which revealed scalar interactions between
protons of similar chemical shifts (not shown). The 2Q-COSY and DQF-COSY spectra also revealed numerous cross-peaks that could arise only from inequivalent H5'/H5'', which in many instances had large
differences in chemical shifts. In some instances, these could be independently
verified from NOESY spectra by observing H6-H5/H5'' and H1'-H5'/H5'' (cf. U7 and C11). The
assignments are given in Table
1
.
Figure
3
shows cross-sections at the water frequency from NOESY and ROESY spectra of r(CGCAAAUUUGCG)
2
recorded at 10oC and 14.1 T, using mixing times of 50 and 25 ms, respectively. Imino
protons, which resonate between 12 and 14 p.p.m. (not shown) showed some
exchange with water, in the order: G12N1H>U7N3H[approx]U8N3H>U9N3H[approx]G2N1H>G10N1H([approx]0). The N4H2 (hydrogen bonded amino proton) of C3 and C11 showed essentially no exchange whereas that of C1 exchanged extensively
with the water. The intense peaks between 6.6 and 6.9 p.p.m. arise from chemical exchange primarily of C2'-OH (see above) with water. The relatively strong peak at 7.58
p.p.m. is predominantly a direct NOE between U7H5 ([delta] = 4.97 p.p.m.) and U7H6 (cf. Table
1
). However, it provides a useful internal intensity standard, as the distance between these protons is 2.42 Å.
We have shown that C2'-OH can be observed in RNA under appropriate conditions, and that
they are probably hydrogen bonded to groups within the RNA molecule. This
indicates that additional information is available about RNA conformation and stability. Measurements of the NOEs involving the ribose hydroxyl protons should assist in the calculation of structures.
Both grooves of the RNA are clearly in contact with water, as has been observed
by NMR for DNA. RNA has a deep major groove and a wide, shallow minor groove.
Further, in RNA, the minor groove should be quite hydrophilic owing to the
presence of the C2'-OH groups. However, the presence of water molecules close to H1' occurs despite the wide and shallow minor groove of RNA.
This is in contrast to the DNA analogue, d(CGCAAATTTGCG)
2
, which is the B form and has a narrow minor groove in the A3T3 tract (
12
,
13
). Slowly exchanging water molecules have been detected in the minor groove in
the region of the adenine residues of the related d(CGCGAATTCGCG)
2
, and more rapidly exchanging water elsewhere (
2
,
3
). In DNA there is no nearby hydroxyl group to stabilise a water molecule, and
under conditions similar to those used in this work, no NOEs were observed
between water and H1' (Lane, Jenkins and Frenkiel, unpublished). Because the hydroxyl proton
exchanges with the solvent, it must be both accessible to, and interact with
water, which indicates hydration at this site. Our results indicate that it is
possible for water protons to be sufficiently close to H1' for a direct NOE to be observed, and that hydrogen bonding to the
hydroxyl oxygen may be responsible for the relatively long residence time,
compared with DNA. This would be consistent with recent X-ray studies of RNA duplexes where water molecules interacting directly with the C2'-OH have been observed (
27
-
29
).
The reason for slow exchange in dAnTn tracts may be related more to the high
propeller twist in these sequences, rather than the narrowed minor groove
per se
, which allows bridging of water molecules between neighbouring base pairs. This
interaction is not possible in mixed-sequence DNA or RNA, where as we have shown the AdeC2H is only weakly
hydrated. However, even in the narrowed minor groove of AnTn DNA tracts,
hydration near to H1' is very weak or non-existent, whereas in RNA it is substantial, and can be attributed
to the presence of the OH group. The water close to this group is relatively
long lived, which does not correlate with a narrow minor groove. Although only
a lower limit to the water residence time can be obtained by NMR, we imagine
that even the long-lived molecules have residence times in nanoseconds rather than much
longer as is sometimes observed in proteins (
20
). This is supported by the observed dependence of the ratio R on the magnetic field strength. For correlation times longer than ~2 ns, R becomes independent of the magnetic field strength, whereas we observed a significant magnetic field dependence for
all protons (Table
2
).
This work was supported by the Medical Research Council of the UK, by a Wellcome
Travelling Research Fellowship to MRC and a Royal Society of Edinburgh
Caledonian Research Fellowship to GC. We thank Dr J. Feeney for helpful
discussions.
REFERENCES
Return