ABSTRACT
The pyrimidine base 5-(hydroxymethyl)-2
'
-deoxyuridine (HmU) is a common nucleotide in SPO1 phage DNA. Numerous transcriptional proteins bind HmU-containing DNA preferentially implicating a regulatory function of HmU. We
have investigated the conformation and dynamics of d-(5
'
-CHmUCHmUACACGHmUGHmUAGAG-OH-3
'
)
2
(HmU-DNA). This oligonucleotide mimics the consensus sequence of Transcription
Factor 1 (TF1). The HmU-DNA was compared to the thymine-containing oligonucleotide. NOESY and DQF COSY spectroscopy provided
resonance assignments of nonexchangeable and exchangeable protons,
intranucleotide, internucleotide and intrastrand proton-proton distances, and dihedral angle constraints. Methylene protons of
the hydroxymethyl group are nonequivalent protons and the hydroxymethyl group
is not freely rotating. The hydroxymethyl group adopts a specific orientation
with the OH group oriented on the 3
'
side of the plane of the base. Analysis of imino proton resonances and NOEs
indicates additional end base pair fraying and a temperature-induced transition to a conformation in which the internal HmU-A base pairs are disrupted or have reduced lifetimes. Orientation of the hydroxymethyl group indicates the presence of internucleotide intrastrand hydrogen bonding between the HmU12C5 hydroxyl group and A13. All sugars in both DNAs show a
C2' endo conformation (typical of B-DNA).
Numerous nucleoside analogs including the pyrimidine base 5-(hydroxymethyl)-2'-deoxyuridine (HmU)
demonstrate antibacterial, antiviral or mutagenic activity (
1
-
4
). Biosynthesis of HmU results in G-C to A-HmU transition (
5
,
6
). Excision of HmU is carried out by the activity of HmUra-DNA glycosylase, found to be decreased in Werner's syndrome cells (a
disease marked by early symptoms of accelerated aging) indicating a role of HmU
accumulation in the aging process (
7
-
9
).
While the presence of HmU is cytotoxic, it is present in the DNA of several
Bacillus subtilis
bacteriophages such as SPO1 instead of thymine as a natural nucleotide (
10
). HmU containing DNA has been shown to be a good template-primer for different DNA and RNA polymerases and may provide a selective
advantage over host DNA in transcription or replication (
11
). In fact, numerous bacteriophage encoded proteins such as the Transcription
Factor 1 (TF1) show greater affinity for HmU containing DNA relative to thymine
containing DNA (
12
). These DNA binding proteins are found in only a small subset of bacteriophages
and the influence of HmU on the binding of eucaryotic transcription factors is
as yet unknown. TF1 is a type II DNA binding protein (DBPII) and shows sequence
homology to other DBPII which have been shown to induce DNA bending and
compaction. The crystal structure of the protein HU (a
Bacillus stearothermophilus
encoded DBPII) has been solved and a model of the HU-DNA binding has been described (
13
). DNA binding is proposed to involve two [beta]-ribbon `arms'. TF1 is a DBPII proposed to have the same binding
motif as HU (
14
). TF1, however, shows greater DNA affinity (specifically for HmU containing
DNA) and sequence specificity. This potentially makes TF1 a good candidate for
the structural study of the novel [beta] sheet binding motif proposed for DBPII-DNA complexes.
We are studying the structural and dynamic features of the DNA binding site of TF1 (
15
). High resolution NMR spectroscopy provides a means of studying the structure and dynamics of double-stranded DNA (
16
-
20
). In order to simplify structural analysis a single base deletion was made to
the consensus sequence of the known TF1 binding site resulting in a self
complementary 16 bp DNA containing eight HmU residues. We find that the 16mer
establishes a conformation with sugar puckers in the C2'endo configuration (typical of B-DNA). The average orientation of the hydroxymethyl group
consistently places the oxygen on the 3' side of the plane of the base in all HmU base pairs. Orientation of the
hydroxymethyl group indicates the presence of an intrastrand internucleotide hydrogen bond between the HmU12C5 hydroxyl group and A13. Comparison of several experiments indicates
instability of A-HmU basepairing resulting in additional fraying, and a temperature-induced transition to a conformation in which the internal HmU-A base pairs are disrupted or have reduced lifetimes.
The synthesis of 5-hydroxymethyl-uracil containing oligonucleotides has been previously described (
21
). Thymine containing 16mer 5'-CTCTACACGTGTAGAG-OH-3'
2
(T-DNA) was purchased from Oligo's, Inc. Wilsonville, OR. Samples were dissolved to 2 mM in 100
mM NaCl in 80% H
2
O, 20% D
2
O (purchased from Cambridge Isotope Laboratories, Woburn, MA) and titrated to pH
6.2 with NaCl. Samples were dried under a stream of nitrogen and resuspended in
D
2
O when appropriate.
Melting profiles of the HmU-DNA [d-(5'-CHmUCHmUACACGHmUGHmUAGAG-OH-3')
2
] and T-DNA were compared by optical melting curves. The melting temperature of
the HmU-DNA is ~10o lower than the T-DNA. The denaturation appears to be a simple two-state process in both the HmU-DNA and T-DNA. No significant melting is apparent
prior to 30oC in either oligonucleotide.
Proton NMR spectra were recorded on a Bruker AMX500 FT-NMR spectrometer (499.896 MHz). NOESY spectra for analysis of exchangeable
protons were recorded in 80% H
2
O/20% D
2
O solvent using a jump-return pulse sequence for water suppression over a spectral width of 10
504 Hz (
22
). The time domain consisted of 2048 data points along t2 and 480 data points
along t1 with 96 summed scans for each t1. Mixing times varied from 80 to 200
ms as noted. A relaxation delay of 30 ms was used (total recycle time of 1.31
s). NOESY spectra for analysis of nonexchangeable protons were recorded in a
100% D
2
O solvent over a spectral width of 5000 Hz. The time domain consisted of 2048
data points along t2 and 480 data points along t1 with 96 summed scans for each t1. A relaxation delay of 100 ms was used (total recycle time of 1.38 s). The effect of increasing recycle time
on signal intensity was evaluated by 1D spectra of the HmU-DNA. No significant increase (<5%) in signal intensity occurs after 1.3 s. Spectra were analyzed using
Felix software. In all cases data workup included a zero fill to 2048 along the
t1 axis and shifted sine bell functions along both axes prior to two
dimensional Fourier transformation. Spectra were referenced to H
2
O at 4.7 p.p.m.
Double quantum filtered correlated spectroscopy (DQF COSY) data were collected in D
2
O over 5000 Hz spectral width with 2048 data points and 80 summed scans per t1.
One dimensional data was collected on a Bruker AMX 500 FT NMR spectrometer using
a jump-return pulse with presaturation (
22
). 400 transients of 16 K data points were collected over a spectral width of 10
504 Hz. Data processing included shifted sine bell adipodization. All spectra
were referenced by designating the center of the spectral width as 4.7 p.p.m.
Interstrand, internucleotide, intranucleotide and dihedral angle constraints
were determined from NOESYspectra taken at 80 ms mixing time in H
2
O and D
2
O and DQF COSY spectra for both HmU-DNA and T-DNA.
Distance contraints were estimated using very strong (1.8-2.2 Å), strong (2.2-2.5 Å), medium-strong (2.5-2.9 Å), medium (2.9-3.3 Å), weak-medium (3.3-3.7 Å), weak (3.7-4.1 Å) and very weak (4.1-6.0 Å) (
23
). Cross peak intensities were categorized by measurement of crosspeak volumes
in NOESY spectra acquired at 80 ms mixing time and application of the relationship
r
ij
=
r
ref
(
R
ref
/
R
ij
)
1/6
where (
r
ij
) is the proton-proton distance, (
r
ref
) is the fixed reference distance (cytosine H5-H6 = 2.46 Å) and
R
ref
/
R
ij
is the relative initial cross-relaxation rate. NOE buildup curves were used in order to determine the mixing time appropriate for the
`two spin' approximation used in molecular dynamics.
The pseudorotational angle was determined by a combination of H1'-H4', H1'-H2' and H4'-H2'' measured
distances (
24
). Individual torsional angles for all dihedral restraints within the sugar ring
were determined from the pseudorotational angle (
25
). The angle (5'O-5'C-4'C-3'C) was determined from the [Sigma]
J
H4'
measured by DQF COSY
(
18
). The angle [C4'-C3'-O3'-(n+1)P] was determined from the [Sigma]
J
H3'
(
18
). Glycosidic angle (4'O-1'C- 1N-2C) for pyrimidines and (4'O-1'C-9N-4C) for purines
was determine by H8/H6-H1' distances (
24
). Hydrogen bond restraints were added in order to maintain base pairing.
Starting structures for molecular modeling ranging from B-DNA to A-DNA were generated using the Insight software. Five starting structures having an average RMSD of 3.42 Å (all atom) were generated by building the 16mer duplex of T-DNA (used directly for T-DNA modeling) and modifying the thymine bases to
reflect the 5-hydroxymethyl group (for HmU-DNA modeling). Molecular modeling was performed using Biosym software on a Silicon Graphics
Iris workstation. All calculations were carried out on a SGI Challenge L
compute server. Counterions were not included, however, their effects were
mimicked using distance-dependent dielectric constant with e =
r
ij
(
26
). Molecular dynamics was carried out with a simulated annealing protocol
utilizing the Amber forcefield (
26
). Molecular dynamics included (i) energy minimization until a root mean square
of 0.1 kcal/mol Å
was achieved, (ii) temperature increase to 800 K, (iii) ramping of force
constants at 800 K followed by 14 phases of molecular dynamics with a temperature reduction from 800 to 100 K, and (iv) minimization until a root mean square gradient of 0.03 kcal/mol Å was achieved. Twenty-nine final structures were obtained which converged to an RMSD of
0.50 Å (all atom) for the internal 12 bp.
Table 1
Proton chemical shifts of HmU-DNA and T-DNA (in parentheses) at 30oC. Chemical shifts differing by >0.1 p.p.m. are in bold face
The region of the spectrum associated with the methylene protons of two of the
hydroxymethyl groups of HmU-DNA is shown in Figure
1
. The two methylene protons (HM1 and HM2) are nonequivalent indicating that the
hydroxymethyl group is not freely rotating. These resonances are doublets due
to the spin coupling of the two methylene protons which show a strong NOE to
each other (data not shown). Each proton has two significant NOEs, one
intranucleotide to the H6 and one internucleotide to the neighboring base H8.
Thus for each HmU residue four distances (six when there is cytosine 5' of the HmU) can be determined. The orientation of the hydroxymethyl
group can be seen by inspection of the 80 ms NOESY (Fig.
1
). The two protons show similar NOEs to the neighboring (n-1)H8 proton and very
dissimilar NOEs to the intraresidue H6; one strong and one weak NOE. This
defines the orientation of the hydroxymethyl group in space. Assignment of
these methylene protons was accomplished by trial molecular dynamics, once with
M1 assigned as the upfield resonance and once with M1 assigned as the downfield resonance. Intra- and interresidue distances in the final models were compared with the
experimentally determined distances. In the model calculated with M1 assigned
as the upfield resonance the two interresidue distances involving the methylene
protons are quite different from each other (3.95 and 2.83 Å) while the experimentally measured interresidue distances are similar at
3.39 and 3.28 Å. However, in the model calculated where M1 is assigned as the downfield
resonance the two interresidue distances involving the methylene protons are
quite similar at 3.09 and 3.29 Å and close to the experimentally determined distances. Thus, in this
case, assignment of M1 as the downfield resonance results in an orientation of
the hydroxymethyl group that more closely adheres to the experimental data.
Each HmU residue was evaluated in this manner and in each case calculations
performed with M1 assigned as the downfield resonance resulted in an
orientation of the hydroxymethyl group that best satisfies the experimentally determined distances. Molecular dynamics calculations with M1 assigned as the upfield resonance result in deviations of up to 0.71 Å (average overall [brvbar]deviation[brvbar] = 0.45 Å) from the measured values of the M1 and M2
associated interproton distances described above. Calculations with M1 as the downfield resonance,
however, result in maximum deviation of only 0.45 Å (average overall [brvbar]deviation[brvbar] = 0.25 Å) from the measured distances.
Two dimensional
1
H NOESY NMR spectra of HmU-DNA and T-DNA in 80% H
2
O/20% D
2
O were recorded at 200 ms mixing time at 10oC. Sequence specific assignment of the exchangeable protons has been
carried out and their chemical shifts listed in Table
1
.
One dimensional
1
H NMR and two dimensional
1
H NOESY spectra were acquired of HmU-DNA and T-DNA in 80% H
2
O/20% D
2
O at increasing temperatures. At low field, imino proton resonances associated
with HmU/T and guanine can be observed. In HmU-DNA loss of resonances associated with the internal A-HmU base pair imino protons can be seen as early as 15oC. The melting profile obtained from 1D NMR measurements of HmU-DNA shows the downfield shift typical of imino protons at
higher temperatures (data not shown). At 30oC a distinct difference can be seen between the 1D resonances of the imino
protons of T-DNA and HmU-DNA (Fig.
2
A). In addition, distinct differences are seen in the NOEs associated with A-T/HmU and G-C base pairing between the two DNAs (Fig.
2
B). NOEs associated with the A-HmU base pairing in HmU-DNA are smaller than those in T-DNA indicating a temperature-dependent transition to a conformation in which these
base pairs are disrupted or have reduced lifetimes. In addition, spectra
acquired at 10oC show that while a large NOE is observed for T2H3-A15H2 of T-DNA the corresponding NOE observed for HmU2H3-A15H2 is quite small (Fig.
3
). Thus, while T-DNA demonstrates a typical fraying profile for DNA duplexes at 10oC, HmU-DNA shows excess fraying at 10oC.
NOESY spectra of HmU-DNA were acquired with increasing mixing times and NOE buildup curves were
constructed in order to determine the effect of mixing time on estimated
interproton distances. A mixing time of 80 ms was chosen for the qualitative
estimation of interproton distances for both HmU-DNA and T-DNA. Data were acquired at 20oC in order to reduce the effects of fraying and premelting
phenomena noted in the previous sections. All conformational analysis described
below was restricted to the internal 14 bp as, even at 20oC, considerable fraying is apparent at the last base pair.
The pseudorotational angle of
P
= 160o was determined by combination of H1'-H4' (3.21 Å), H1'-H2' (2.91 Å) and H4'-H2'
(>4.0 Å), distances measured from
1
H-
1
H NOESY data acquired at 80 ms mixing time (
24
). Individual torsional angles for all dihedral restraints within the sugar ring
were determined from the pseudorotational angle based on an average of 12
structures of
P
= 163o +- 10o (
25
). The glycosidic torsion angle (4'O-1'C-1N-2C) for pyrimidines and (4'O-1'C-9N-4C) for purines
was constrained to -125o +- 35o and 125o +- 75o respectively based on the
experimentally determined H6/H8-H1' distances (
24
). The angle (5'O-5'C-4'C-3'C) was established by the [Sigma]
J
H4'
determined from DQF COSY. The average total coupling of the six discernible H4' resonances was 8.1 Hz. This limits the possible range of angle (5'O-5'C-4'C-3'C) to 60o +- 30o or 240o +-
30o (
18
). Comparison of the relative intensities of H6/H8-H5'/H5'', H6/H8-H1', H2'-H5'/H5'' and
H2'-H1' in the NOESY spectrum eliminates the 240o +- 30o possibility. Thus angle (5'O-5'C-4'C-3'C) is determined
to be 60o +- 30o. Distance bounds of 2.28-2.70 Å were set for H4'-H5'/5'' based on this
angle (
18
). Based on the pseudorotational angle and (5'O-5'C-4'C-3'C), additional distance constraints involving the H5'/H5'' backbone
protons were determined (
24
). This was verified by close inspection of NOESY spectra acquired at 80 and 200
ms that show all H1'-H5'/H5'' distances to be >3 Å. Imposing a lower bound of 3 Å on all H1'-H5'/H5''
(interresidue and intraresidue) effectively restricts the torsional angles (3'O-P-5'O-5'C) and (3'C-3'O-P-5'O)
(
18
). The dihedral angle [n4'C-n3'C-n3'O-(n+1)P] was restricted to 140-360o based on the sum of the H3' coupling constants ([Sigma]
J
H3'
). With this angle restriction a distance constraint of 2.8-3.25 Å is imposed for nH3'-(n+1)P (
18
). Similar measurements were made from the T-DNA.
We have studied the HmU-DNA containing TF1 binding site and its corresponding T-DNA. Sequence specific assignment of the nonexchangeable and
exchangeable protons of HmU-DNA and T-DNA has been carried out, models developed, and exchangeable protons analyzed for base pair stability.
Inspection of the imino (exchangeable protons) region of the
1
H NOESY NMR spectra indicate a temperature-dependent transition to a conformation in which these base pairs are
disrupted or have reduced lifetimes. Exchangeable protons associated with the
end base pairs of both the HmU-DNA and T-DNA are not detectable. This is typical of short oligonucleotides and is commonly referred to as `fraying'. A NOE is observed for
T2H3-A15H2 of T-DNA (Fig.
3
A) indicating that the second base pair is substantially annealed. In contrast,
only a small NOE is observed for HmU2H3-A15H2 (Fig.
3
B) suggesting significant fraying through the second base pair in HmU-DNA. It is also interesting to note that additional NOEs between HmU12H3-A13H2 and HmU4H3-A5H2 are apparent in the NOESY spectrum of the HmU-DNA that do not appear in the spectrum of T-DNA (Fig.
3
). These connections are between adjacent base pairs of the same strand
indicating that at 10oC there is a slight tilt of neighboring 5'-HmUA-OH-3' base pairs relative to 5'-TA-OH-3' base
pairs. At 10oC imino proton resonances and NOEs of the internal base pairs are quite
similar between the two oligomers while at 30oC a distinct difference can be seen between the imino proton resonances
associated with A-T and A-HmU base pairs (Fig.
2
). This may account for the lower
T
m
determined for HmU-DNA relative to T-DNA. Lower
T
m
values for DNA containing HmU residues was first reported by Kallen in 1962 (
10
). Previous studies, however, suggest that the presence of HmU in DNA does not
change the torsional flexibility of the whole molecule (
28
). In these studies steady state fluorescence polarization anisotropy (FPA) of intercalated ethidium was used to compare the torsional flexibility of HmU-containing SPO1 DNA and thymine-containing DNA. The data indicate that the rigidity is the same in
both DNAs. Reduction in DNA flexibility is, however, apparent upon binding to
type II DNA binding proteins (
28
). Thus, while potential DNA flexibility may be important in TF1 binding, sequence-dependent charge distribution and local structural fluctuations may also play an important role in the
increased affinity of HmU-DNA to TF1.
We thank Maria V. Silva, Anne Grove and E. Peter Geiduschek for many helpful
discussions. We wish to acknowledge support from the National Institutes of
Health (GM-40635).
Present addresses:
+
Department of Biochemistry, The University Dundee, UK,
[sect]
Dipartimento di Chimica delle Sostanze Naturali, Universita di Napoli Federico
II, via Domenico Mentesano 49, I-80131 Napoli, Italy and
[para]
La Jolla Cancer Research Foundation, La Jolla, CA, USA
Two dimensional
1
H NMR NOESY spectra of nonexchangeable protons of HmU-DNA and T-DNA in D
2
O were recorded at 30oC. Sequence specific assignment has been carried out and the chemical
shifts are listed in Table
1
. The upfield region of the NOESY of HmU-DNA shows the connectivities of the H2'/H2'' nonexchangeable protons. Examination of the H8/H6-H2'/H2'' NOEs and absence of H2''-H4'
NOEs confirms the sugar pucker conformation to be C2'endo, typical of B-DNA (
27
). DQF COSY spectra of T-DNA and HmU-DNA identify through bond connectivities of H1' with H2'/H2''. Characteristic couplings of these
resonances assign the H2', H2'' and H1' protons.
REFERENCES
Return