(Received for publication, October 18, 1996, and in revised form, January 19, 1997)
From the Department of Chemistry, Wesleyan University, Middletown, Connecticut 06459
Oxidative stress, ionizing radiation, and other events can induce the oxidation of the thymine in DNA to thymine glycol. The presence of thymine glycol can have significant biological consequences, and there are specific repair enzymes for thymine glycol in a wide range of organisms. The structure of a duplex DNA containing a single thymine glycol (5,6-dihydroxy-5,6-dihydrothymidine) has been determined by the combined use of NMR and restrained molecular dynamics. The duplex of d(C1G2C3G4A5Tg6A7C8G9C10C11) paired with d(G22C21G20C19T18A17T16G15C14G13G12), with Tg indicating thymine glycol, has been used for these studies. The structure shows that the thymine glycol induces a significant, localized structural change with the thymine glycol largely extrahelical. This structural information is consistent with the biological consequences of thymine glycol in DNA. This structure is compared with that of a DNA duplex with an abasic site in the same sequence context.
Damage to DNA can occur by the spontaneous deamination of cytosine to uracil and through the action of alkylating agents, oxidants, drugs, and toxins, ionizing radiation, and other modes of action (1-3). The exposure of DNA in cells bound to proteins as a solid or free in solution to ionizing radiation or to oxidative stress can lead to the conversion of thymine to thymine glycol (5,6-dihydroxy-5,6-dihydrothymidine). Approximately 10-20% of the damage to DNA induced by ionizing radiation, including that used to treat tumors, is the result of thymine base oxidation and fragmentation. These products can also be produced by oxidative stress. In addition, the importance of damaged DNA as a control on the cell cycle is increasingly becoming a focus of research efforts so that the effects of damaged thymines on the structures, stabilities, dynamics, and interactions of DNA are of interest.
It has been known for several decades that ionizing radiation can stop the reproduction of cells as well as kill cells. These activities form the basis for using radiation in chemotherapy, and the research in this area has been frequently reviewed (1-13). Since the mid-1950s there have been studies on the effects of ionizing radiation on the chemical integrity of DNA (4, 5, 7, 9, 14). Ionizing radiation can induce a number of types of damage to DNA including single and double strand breaks, oxidation of the purine and pyrimidine bases, and cross-linking of DNA with proteins. Radiation-generated oxidants are thought to react with thymine to lead to the formation of thymine glycol, thymine peroxide, thymine hydroperoxide, and other oxidized forms of the base. Some of these oxidized forms of thymine subsequently react further to form urea. Some of the same types of damage also occurs to DNA during oxidative stress (5, 15, 16).
The damage to the thymine bases in DNA is of special interest because the major forms of damaged thymine are known; thymine is the most easily oxidized base, and many of the biological consequences of damaged thymines are known. Thus, the damaged thymines offer the opportunity to correlate the changes in the physical properties of DNA caused by ionizing radiation or oxidative stress with the biological effects.
Thymine glycol in DNA can be excised in vitro by
Escherichia coli endonuclease III or other enzymes, which
liberates the thymine glycol and subsequently carries out a
-elimination reaction to cleave the 3
phosphodiester as was first
shown by Demple and Linn (17, 18). We have shown that
endo1 III cleaves the 3
phosphodiester of
abasic sites via a syn
-elimination reaction (17-20). A
mammalian homolog of endo III has recently been found (21).
X174
containing thymine glycol are inactivated in E. coli hosts
deficient in endo III much more so than in wild type hosts, indicating
that endo III is most likely involved in thymine glycol repair in
vivo (22).
X174 containing thymine glycol are also inactivated
in E. coli hosts deficient in exonuclease III and endo IV
much more so than in wild type hosts, indicating that class II apurinic
endonuclease activity is needed for thymine glycol repair (22). Thymine
glycols can apparently also be repaired through the ultraviolet induced
SOS repair mechanism, which is more commonly associated with thymine
dimer repair. Thus, there may be as many as three routes to thymine
glycol repair: endo III; exonuclease III and endo IV; and SOS (13).
These results indicate that thymine glycol sites can be repaired
in vivo in organisms ranging from E. coli to
mammals.
The presence of thymine glycol in DNA can have profound consequences on DNA replication (9, 13, 23-27). The presence of thymine glycol is a block to replication (23-28). Studies of the effects of thymine glycol on the replication of M13 in vitro using polymerase I, E. coli DNA polymerase I, and T4 polymerases indicate that thymine glycol stops replication either one residue before or at the site of damage. Examination of the replication of M13, prepared with a single thymine glycol, indicated that thymine glycol is a weak mutagen about 0.3% of the time and primarily acts as a replication block (23).
The presence of thymine glycol may also induce structural changes at the initiation and termination points leading to alterations in the relative amounts of the termination bands of DNA synthesis (22). These effects of thymine glycol may be due to the interactions of the double stranded DNA containing the damaged site with the polymerase.
The presence of a pyrimidine 5 to the damaged thymine seems to enhance
the probability of reading through a thymine glycol site in replication
more so than a 5
purine, indicating a sequence-dependent effect (22). The base 3
to a thymine glycol is apparently also important in determining the extent of the block to replication (28).
There is also evidence that dA is the residue placed opposite thymine
glycol by DNA polymerase (9, 13, 23).
Taken together these results indicate that thymine glycol sites can be repaired in vivo as well as in vitro and that if there is no repair, replication can be blocked and thymine glycol is mildly mutagenic. These molecular biology studies also suggest that the sequence context may be important and that the presence of thymine glycol might have structural consequences. Because thymine glycol can not be planar, unlike the normal DNA bases, structural consequences are not unexpected. It is likely that the presence of thymine glycol has effects on transcription, regulation, and DNA packaging as well.
The studies described below on the physical properties of DNA containing thymine glycol are aimed at determining how the changes in DNA structure, dynamics, and stability can be related to the biological consequences of thymine glycol in DNA. These results on thymine glycol will also be compared with those from our ongoing studies of the structures of DNAs containing abasic sites.
DNA containing thymine glycol was prepared by oxidation of the parent ssDNA with 0.1 M KMnO4. The oxidation was carried out in a 300-ml plastic jar containing 20 ml of 0.2 M KHPO4 at pH 8.6 and 50 OD260 of d(C1G2C3G4A5T6A7C8G9C10C11). The mixture was stirred with a magnetic stirrer for 20 min in an ice bath. With the sample at 4 °C, it was treated with 8 ml of 0.1 M KMnO4 (Aldrich) for 5 min. The reaction was quenched by the addition of 0.5 ml of allyl alcohol (Aldrich), which converts the MnO4 into MnO2. The sample was kept at 4 °C for at least 1 h to allow the MnO2 to completely precipitate. The reaction mixture was then centrifuged to remove the MnO2, and the supernatant containing the products was diluted to 400 ml with distilled water and was desalted with the use of a Waters C-18 cartridge. The DNA was eluted from the C-18 column with 5 ml of 60% CH3CN/H2O in three steps (2 × 2 ml and 1 × 1 ml).
Purification of the Oxidized DNAThe ssDNA containing
thymine glycol was purified by HPLC on a semi-preparative
reversed-phase PRP-1 column (Hewlett-Packard) and eluted (elution time
is 18 min) with a 2-25% gradient of 50% acetonitrile, 25 mM phosphate buffer at pH 7.0 mixed with 25 mM phosphate buffer at pH 7.0. A typical chromatogram is shown in Fig. 1.
The column flow rate was 2 ml/min, and detection was at 254 nm. The
other peaks produced by oxidation were not identified and are likely to
contain oxidized guanine, adenosine, or other oxidation products. The
purified single strand containing thymine glycol was collected,
diluted, and desalted using a C-18 column. The isolated yield was
typically about 24%.
Characterization of the Thymine Glycol in the Single Stranded DNA
To confirm that the DNA prepared as described above contains a single thymine glycol and no other modifications, further analysis was carried out. A sample of 1 OD of the ssDNA containing thymine glycol was incubated with 2 units of venom phosphodiesterase, 4.1 units of alkaline phosphatase, 1 M Tris, 1 M MgCl2 at pH 8 and 37 °C for 24 h. The parent strand was also digested using the same conditions except that the incubation time needed for complete digestion was only 1 h. Both DNAs were completely digested to nucleosides by this procedure, and the nucleoside mixtures were analyzed by HPLC.
The HPLC was performed using a PRP-1 column with a 3-18% gradient of 50% acetonitrile, 25 mM phosphate buffer at pH 7.0 mixed with 25 mM phosphate buffer at pH 7.0. The column flow rate was 2 ml/min, and the nucleosides were detected at 254 nm. The order of elution of nucleosides from the parent strand is C, G, T, and A, and the elution times were 11.9, 15.2, 16.8, and 21.0 min, respectively. Typical chromatograms are shown in Fig. 1. The relative ratios of C, G, and A from the thymine glycol containing DNA were the same as in the parent strand. Thymine glycol was eluted at about the same time as T. The thymine glycol was monitored at 220 nm, collected, and used for stereoisomer determination discussed below.
Determination of the Stereoisomer Form of the Thymine GlycolPermanganate oxidation of thymine has been shown to yield only the two cis-isomers (4, 28). The determination of the stereochemistry of the cis-thymine glycol produced by the oxidation of single stranded DNA was determined by HPLC and confirmed by NMR. The thymine glycol nucleoside was prepared from the thymine glycol containing DNA by the digestion procedure described above. The nucleosides were chromatographed on a Whatman ODS-3 semi-preparative column, 10 mm × 50 cm, with a mobile phase of 3% acetonitrile in water at a flow rate of 2 ml/min with a typical chromatogram shown in Fig. 1. The elution time of the (5S,6R)-stereoisomer is 16.5 min, and that of the (5R,5S)-stereoisomer is 17.8 min.
For comparison the oxidation of thymine nucleoside was carried out under the same conditions as those used for the ssDNA. The ratio of the (5S,6R)- to (5R,6S)-stereoisomers obtained in the nucleoside oxidation is about 1 to 3, whereas (5R,6S) is the only detectable stereoisomer from the oxidation of ssDNA. Similar stereochemical selectivity in the oxidation of ssDNA was previously reported by our group (29). The improved sensitivity and resolution of HPLC methods now allows determination that the ratio of (5R,6S) to (5S,6R) is more than 10 to 1. The identification of each of the two stereoisomers has been confirmed by one- and two-dimensional NMR as described previously (29).
An examination of the products obtained by the oxidation of mono- and
dinucleotides by similar methods has indicated that the presence of a
5-phosphate is the determinant of the preference for the
(5R,6S)-stereoisomer (30). It is not obvious to
us how a 5
-phosphate controls the stereochemistry of the permangenate oxidation.
The heteroduplex was formed by mixing equimolar quantities based on the extinction coefficients of the two strands and by monitoring the titration of the single strand containing the thymine glycol with its complementary strand via the one-dimensional NMR spectrum. The duplex was lyophilized several times in 2H2O and dissolved in 0.5 ml of 99.96% 2H2O. The purified duplex was studied at 1-1.5 mM concentration in pH 7.0 buffer containing 10 mM sodium phosphate, 100 mM sodium chloride, and 0.05 mM EDTA in 99.96% 2H2O. For experiments involving the exchangeable imino protons, the duplex was lyophilized and dissolved in 90% H2O, 10% 2H2O.
NMR ProceduresTwo-dimensional NOESY, total correlation spectroscopy, and phased easy correlation spectroscopy spectra were collected at 400 and 750 MHz. The 400-MHz data were obtained using a Varian Unityplus spectrometer at Wesleyan, and the 750-MHz spectra were obtained using a Bruker DMX 750 spectrometer at the University of Wisconsin at Madison. The Varian NMR results were processed using VNMR software, and FELIX 95.0 software was used for the Bruker data.
NOESY experiments in 2H2O were carried out at mixing times of 150 and 250 ms with a 1 s equilibration delay with presaturation of the water resonance, using the Bruker 750 MHz with a spectral width of 11904.8 Hz in each dimension. For each mixing time, 512 t1 increments were acquired with 64 scans for each increment of the evolution time. The F1 dimension was zero-filled to 2,000, and the data were processed with sinebell apodization in each dimension prior to 4,000 × 2,000 Fourier transformation. These data were used for quantification of the NOESY cross-peaks.
The total correlation spectroscopy spectra were obtained at 750 MHz and 25 °C with a mixing time of 60 ms and a delay time of 1 s between acquisitions. In this experiment, 512 t1 increments were collected with 64 scans for each increment. The free induction decay along t1 was zero-filled to 4,000 data points to give a final 4,000 × 1,000 spectrum. The data were processed with sinebell apodization in t1 and t2 and base line corrected along both dimensions.
The phased easy correlation spectroscopy spectra were collected at 400 MHz and 25 °C. In this experiment, 256 acquisitions for each of the 256 t1 values were obtained into 2,000 points. The free induction decay along t1 was zero-filled to 2,000 data points prior to Fourier transformation to give a final 2,000 × 2,000 spectrum. The data were processed with Gaussian apodization and Gaussian shift constant in t1 and t2 and drift corrected in both dimensions.
The 1H spectrum for the nonexchangeable protons was carried out at 400 MHz at 25 °C. The spectrum of the exchangeable protons was obtained at 400 MHz with a spectral width of 10,000 Hz using a jump and return pulse for water suppression. One-dimensional 31P NMR data were obtained at 161.9 MHz with proton decoupling. The spectra width was 2687 Hz with 4288 complex points. A Lorentzian apodization of 3 Hz was applied prior to Fourier transformation.
Quantitation of NOE Cross-peak VolumesThe volumes in the NOE cross-peaks of the data obtained with 150 and 250 ms mixing times were quantified using FELIX 95.0 software as described previously. For each assigned and resolved cross-peak, the volume in a standard area was determined.
Structure Determination ProcedureStructure refinement was
performed as described previously (31) by restrained molecular dynamics
using X-PLOR 3.1 with the following modifications. 387 NOE volume
constraints were used for each of the two mixing times and 31 dihedral
constraints (H1-C1
-C2
-H2
and H1
-C1
-C2
-H2") were used. The
optimum relative weighting of the NOE constraint (60 kcal/mol) was
found to be larger for the thymine glycol site DNA than for undamaged
and abasic site containing DNAs. The other force constraints were the
same as those we have previously used for damaged and undamaged DNA
(31-34).
The starting structure was generated from a canonical B-form DNA in INSIGHT II by modifying the thymine residue in parent strand to thymine glycol and then minimized by DISCOVER 3.0. Because there are four possible thymine glycol structural forms that are the two chair forms and two boat forms, a MNDO minimization using INSIGHT II and molecular dynamics using X-PLOR were performed on thymine glycol. The potential was determined from MNDO, and the molecular dynamics showed that the free thymine glycol underwent rapid interconversion between the boat and chair forms. This indicated that the MNDO-generated potential was a reasonable one for thymine glycol. This potential for thymine glycol was then inserted to the topology file of the duplex DNA.
The energy of the starting structure was minimized in 100 steps of Powell's conjugate gradient minimization using X-PLOR. The relaxation matrix refinements were carried out in vacuum at 300 K. These were further minimized using the force field with all restraints for 100 steps of minimization and then subjected to a 60 ps relaxation matrix simulation followed by 200 steps of conjugate gradient energy minimization. The structures were saved at every 2-ps interval. The root mean square deviation over the 50-60-ps portion of the trajectory is 0.212 Å.
The NOE cross-peak volumes for the structures every 2 ps were back
calculated using an overall correlation time of 5 ns, a leakage rate of
0.33 s1, and a distance cut-off of 5.5 Å.
The NMR R, Q, and root mean square values (35) for the averaged structures are 0.10, 0.05, and 0.12, which are slightly better than the 0.11, 0.05, and 0.12 obtained for the 60 ps structure. The R, Q, and root mean square values all declined about 10-20% from 10 to 50 ps, at which time they stabilized. The structures and experimental data have been submitted to the Brookhaven data bank.
The basic properties of the damaged DNA duplex could be determined
from the one-dimensional spectra shown in Fig. 2. The
number of signals in the one-dimensional spectrum of the
nonexchangeable protons indicates that the duplex adopts one structure
on the NMR time scale. The intensity of the imino region of the
spectrum indicates that there are 10 base pairs present in the duplex. The range of 31P chemical shifts indicates that there are
not any phosphodiesters with highly unusual conformations.
Assignment of the NMR Spectrum of the DNA Duplex Containing Thymine Glycol
The spectra were assigned by the application of sequential
assignment procedures used for B-form DNA (36, 37). A portion of a
750-MHz NOESY spectrum is shown in Fig. 3 with many of
the assignments of the H6/H8-H1 region indicated. The H6/H8-H2
and H2" regions of a 750-MHz NOESY spectrum are shown in Fig.
4, and many of the H6/H8-H2
assignments are indicated.
The assignments of the thymine glycol containing duplex are listed in
Table I.
|
The sequential 3-base to 5
-deoxyribose connectivities associated with
B-form DNA were observed for residues 1-5 and 6-11 of the thymine
glycol containing strand. There is a clear break in the inter-residue
NOE connectivities found between the thymine glycol at position 6 and
the A at position 5. The sequential 3
-base to 5
-deoxyribose
connectivities associated with B-form DNA were observed for all of the
residues of the complementary strand.
The NMR data presented to this point indicate that the thymine glycol containing DNA can be described as being close to an normal B-form DNA structure as monitored by the imino proton, the 31P chemical shifts, and the proton-proton NOEs except at the A5-Tg6 junction. At this base step the NOE connectivities are not consistent with B-form DNA.
The structure of the thymine glycol containing duplex was determined by
the combination of molecular dynamics with experimental NOE and
dihedral angle constraints. The range of structures consistent with the
experimental data are shown in Figs. 5 and
6. These structures are those obtained at 2-ps intervals
in the molecular dynamics trajectory between 50 and 60 ps. These
structures indicate the range of structures consistent with the data
and not the range of structures that the molecule would sample over a
10-ps time period. The experimental NOE data and that predicted from
these structures are shown in Figs. 3 and 4. The comparison shows that the agreement between the predicted and experimental data is very good.
The predicted data are the averages over the structures obtained from
50 to 60 ps. The average over the predicted structures offers better
agreement with the experimental data than any of the individual
structures.
The structure of duplex DNA containing an abasic site in the same
position as this DNA has a thymine glycol has been previously determined. Figs. 7 and 8 show the
structures of the duplex DNAs that have the and
forms of the
abasic site, as well as the structure of the duplex DNA containing the
thymine glycol. The comparison shows that the structures that are found
depend on which damaged site is present. The damaged DNAs containing
the
and
forms of the abasic site are distinct from one another as well as from the thymine glycol containing duplex DNA. The thymine
glycol is not accommodated into the normal DNA stacking.
The "Extrahelicity" of the Thymine Glycol Residue
Extrahelical bases have been previously observed in DNA duplexes by crystallography (38, 39) and by NMR methods (40-45). The thymine glycol residue in this DNA appears to be extrahelical because it is not in the central stack of the DNA. However, there is no generally agreed upon definition of extrahelicity.
To move toward a quantitative measure of extrahelicity, we have examined the surface area of DNA residues as a function of the radius of the probe molecule. It was thought that extrahelical residues would have a larger surface area for probe molecules about the size of a water molecule and that the percentage of the surface area accessible to the probe molecule might provide a means to quantify the extent of extrahelicity. The surface areas of nucleic acid residues in canonical structures as a function of probe molecule radius has been previously examined for other reasons.
The percentage of the surface area of a residue in canonical B-form DNA
accessible to a probe of various radii is shown in Fig.
9. The cytosine found in the DNA covalently complexed to HaellI methyltransferase is the most extrahelical residue we
are aware of (46), and its percentage accessible surface area is about
80% for probe molecules in the range of about 2-5 Å. For a terminal
residue the percentage of the surface area that is accessible is about
70% for probe molecules in the range of about 2.5-5 Å. For a residue
in a normal, nonterminal Watson-Crick base pair in B-form DNA, the
accessible surface area is about 15% over this range. It seems that
over the range of probe radii from 2.5 to 5 Å, the maximal accessible
surface area is about 85%, and the minimum is about 10%. An
extrahelicity scale could use these as the two extremes as indicated in
Fig. 9.
The percentage of the surface area of the thymine glycol residue that is accessible to probe molecules is about midway between the most extrahelical residue and an interior residue as indicated in Fig. 9. The residue A5 also has a larger accessible surface area than a normal, interior residue as indicated in Fig. 9. Thus, the thymine glycol residue may be considered to be approximately half extrahelical.
The details of the surface area that is exposed to different size probe
molecules can also offer insights. Fig. 10 shows the accessible surface area of the thymine glycol containing DNA that is
obtained with 1.5- and 4-Å probe molecules. The surface obtained with
the 1.5-Å probe shows that the thymine glycol can be accessed from
both the major and minor grooves. This is not the case for a residue in
an interior base pair as shown for the case of G2. The
surface area obtained with the 4-Å probe molecule has the thymine
glycol only accessible from the major groove side. The 4-Å surface
indicates that the presence of a thymine glycol could be detected by a
protein, for example, without having to further separate the DNA
strands.
These NMR results on thymine glycol containing DNA indicate that the presence of thymine glycol induces a significant and highly localized alteration in the structure of the DNA. The base of the thymine glycol is largely accessible to the solvent and other molecules. It is likely that this structural perturbation due to the presence of thymine glycol effects the recognition of the DNA by proteins and may also effect the packaging of DNA in the replication complex, in nucleosomes, in viruses, and in other contexts. The distortions that accompany the presence of the thymine glycol may provide a readily recognizable target for repair enzymes as well as the direction interaction with the thymine glycol.
The structural perturbation due to thymine glycol is quite different from that found for DNA, which contains aldehydic abasic sites. The comparison of the structures of duplex DNAs of the same sequence that have abasic site damage or thymine glycol shows that the structure of the damaged DNA is dependent on the nature of the damaged site. Thus, the partial repair of a thymine glycol site to an abasic site will reduce the structural consequences of the damage to DNA.
The 750-MHz spectra were obtained with the assistance of Dr. Frits Abildgaard using the NMR facility at the University of Wisconsin at Madison, which is supported by National Institutes of Health Grant RR02301 with equipment purchased with support from the University of Wisconsin (NSF DMB-8415048 and NIH RR02781) and the United States Department of Agriculture.