The structure of d(TpA)*, the major photoproduct of thymidylyl-(3
'
-5
')-deoxyadenosine
The structure of d(TpA)*, the major photoproduct of thymidylyl-(3 ' -5 ')-deoxyadenosine
Xiaodong
Zhao
,
Sourena
Nadji
,
Jeffrey L.-F.
Kao
1
and
John-Stephen
Taylor*
Department of Chemistry and
1
Chemistry Department High Resolution NMR Facility, Washington University,
St Louis
, MO 63130,
USA
Received October 3, 1995;
Revised and Accepted March 5, 1996
ABSTRACT
Irradiation of the dinucleotide TpdA and TA-containing oligonucleotides and DNA produces the TA* photoproduct which
was proposed to be the [2+2] cycloaddition adduct between the C5-C6 double bonds of the T and the A [Bose,S.N., Kumar,S., Davies,R.J.H.,
Sethi,S.K. and McCloskey,J.A. (1984)
Nucleic Acids Res.
12, 7929-7947]. The proposed structure was based on a variety of spectroscopic and
chemical degradation studies, and the assignment of a trans-syn-I stereochemistry was based on an extensive
1
H-NMR and molecular modeling study of the dinucleotide adduct [Koning,T.M.G., Davies,R.J.H. and Kaptein,R. (1990)
Nucleic Acids Res.
18, 277-284]. However, a number of properties of TA* are not in accord with the originally proposed structure, and prompted a re-evaluation of the structure. To assign the
13
C spectrum and establish the bond connectivities of the TA* photoproduct of TpdA [d(TpA)*],
1
H-
13
C heteronuclear multiple-quantum coherence (HMQC) and heteronuclear multiple bond correlation (HMBC) spectra were obtained. The
13
C shifts and connectivities were found to be inconsistent with the originally proposed cyclobutane ring fusion between the thymine and adenine, but could be explained by a subsequent ring-expansion reaction to give an eight-membered ring valence isomer. The new structure for the d(TpA)* resolves the inconsistencies with the originally proposed structure, and could have a stereochemistry that arises from the
anti
,
anti
glycosyl conformation found in B form DNA.
INTRODUCTION
The TA* photoproduct was first isolated from the irradiation products of
thymidylyl-(3'-5')-deoxyadenosine, d(TpA), and was proposed to be a
[2+2] cycloadduct between the 5,6 double bonds of the T and the A (Fig.
1
; Structure
1a
) based on spectroscopic and chemical degradation studies (
1
). Assignment of the trans-syn-I stereochemistry to the proposed cycloadduct was based on the
analysis of
1
H-NMR coupling constant and NOE data which was then used to constrain
molecular dynamics simulations (
2
). Because of the unique fluorescence properties of the acid hydrolysis product
of TA*, compound
2
, (
1
,
3
) it was possible to demonstrate the formation of TA* in duplex poly(dA-dT) and calf thymus DNA irradiated with 254 nm light (
4
). The TA* photoproduct is of interest because of its implications for gene
expression, which depends on a highly conserved TATA site in the promoter
region. Recently, we have succeeded in constructing a 49mer containing a site-specific TA* photoproduct for biological studies, and in particular,
in vivo
mutagenesis studies (
5
). To understand properly the biological properties of this interesting photoproduct requires knowledge of its precise structure, and though the proposed
structure was well supported by various spectroscopic and chemical studies,
there are a number of inconsistencies. Unlike most [2+2] cycloadducts involving the 5,6-double bond of thymine, the TA* product was reported not to be
photoreversed by 254 nm UV light (
1
). Another surprising aspect of structure
1
is that it would be stable in spite of what must be a sizeable amount of ring
strain, and a number of viable pathways for ring fragmentation.
In addition to d(TpA), direct irradiation of thymidylyl- (3',5')-2'-deoxyinosine, d(TpI), with 254 nm has
also been reported to give the analogous d(TpI)* photoproduct with a similar
quantum yield (
4
). The far-UV irradiation of various other adenine-containing substrates, however, have been reported to lead to other
ring fusions (Fig.
2
). The photoadduct
3
is proposed to form between the 5,6-double bond of thymine and the 4,5-double bond of hypoxanthine when the two bases are linked by a
trimethylene chain which then hydrolyzes to give
4
. Irradiation of 5-alkyldeoxyuridylyl-(3'-5')-deoxyadenosine has been proposed to give
5
, a [2+2] adduct between the 5,6-double bond of the uridine derivative and the 7,8-double bond of adenine. Irradiation of d(ApA) gives two major
products A=A and d(ApA)* which have been assigned structures
7
and
8
(
6
). These products were proposed to arise from fragmentation of an initial [2+2]
cycloadduct between the 5,6- and 7,8-double bonds, compound
6
. Although previous spectroscopic measurements and chemical degradation studies
of d(TpA)* appeared to rule out ring fusions involving anything but the 5,6-double bonds of T and A, direct spectroscopic evidence for the actual
nature of the ring fusion has been lacking. Herein, we report the assignment of
the
13
C-NMR spectrum of d(TpA)*, and the results of
1
H-
13
C heteronuclear multiple-quantum coherence (HMQC) and hetero multiple bond correlated (HMBC) NMR
experiments. The NMR data is inconsistent with the originally proposed
structure
1a
for d(TpA)*, but is consistent with an eight-membered ring valence isomer of
1a
that explains its inability to be photoreversed by 254 nm irradiation.
MATERIALS AND METHODS
Reagents and equipment
Anion-exchange HPLC was carried out on a Nucleogen DEAE 60-7 column. Reverse phase HPLC was carried out on a Dynamax C-18 column (4.6 * 250 mm for analytical work and 21.6 * 250 mm for preparative work). NMR spectra were
obtained on Varian UNITY plus-500 and UNITY-600 spectrometers and the data processed on a SPARC 10 computer with
VNMR software.
Preparation of d(TpA)*
TpdA,
3
, (20.3 mg, 0.036 mmol) in 20 ml double-distilled water was dispersed on two Petri dishes to give an overall exposure area of 127 cm
2
and irradiated with 2.23 mW/cm
2
254 nm light for 12 h on ice. The irradiated solution was concentrated and
subjected to preparative C-18 HPLC on a Dynamax column (8 [mu]m, 21.6 mm * 250 mm) with a 4.0 ml/min gradient of acetonitrile in 75 mM KH
2
PO
4
/K
2
HPO
4
, pH 6.8 that went from 0-10% over 30 min, followed by 20 min at 10%, and then 10-25% over 10 min. The fraction with an approximate retention time
of 49 min was concentrated and then desalted on the same C-18 column by washing with 100 ml water to give 7.2 mg (36%) of d(TpA)*.
The
1
H-NMR spectrum was the same as that reported in the literature (
2
);
31
P-NMR (121.5 MHz, D
2
O, referenced to TMP), [delta] -1.35; MS (FAB), m/z (intensity), M+K, 594.
NMR spectroscopy of d(TpA)*
All 2D spectra were acquired in phase-sensitive mode by employing the Hypercomplex method (
7
). ROESY (
8
) spectra were recorded with a 125 and 250 ms mixing times and a 2 kHz spin-lock rf field strength. A total of 2 * 320 * 2048 data matrix with 16 scans per t1 value were collected. Gaussian line broadening and sine-bell function were used in weighting t2 and t1 dimensions, respectively. After two-dimensional Fourier transformation, the spectra
were obtained as 2k * 2k data matrices, and were phase and baseline corrected in both
dimensions. TOCSY spectra (
9
) were acquired with a 125 ms mixing time and were processed as for the ROESY
spectra. NOESY (
10
,
11
) spectra were obtained from 2 * 320 * 2048 data matrices with 16 scans per t1 value. The spectral data were
acquired at 298 K with 250, 500 and 750 ms mixing times, but the 750 ms mixing
time gave the best spectrum. The time domain data were zero filled to yield 2k * 2k spectral data matrices and were processed in the similar way as the
2D ROESY spectrum. The proton-detected heteronuclear multiple quantum coherence (HMQC) (
12
,
13
) spectrum was recorded on a Unity plus-500 spectrometer using a 0.35 s
1
H-
13
C nulling period (Fig.
3
). The 90o
1
H pulse width was 7.8 [mu]s and the 90o
13
C pulse width was 14 [mu]s. The proton spectral width was set to 4200 Hz and carbon spectral width to 12 500 Hz. A 2 * 200 * 2048 data matrix with 64 scans per t1 value was collected.
Gaussian line broadening was used in weighting both t2 and t1 dimensions. Two-dimensional Fourier transformation resulted in spectra with 512 * 2048 data points, which were phase and baseline corrected in both
dimensions. Heteronuclear multiple bond coherence (HMBC) spectra were recorded
in a similar way to the HMQC experiment except that delays of 35, 55, 75 and 95
ms were employed to observe the multiple-bond correlations. Selective HMBC (
14
) was employed to resolve the long-range heteronuclear correlations of the two very closely spaced
resonances, TC4 and AC6. The spectrum was recorded on a Varian Unity 600
spectrometer using two 90o
13
C Gaussian pulses of 1.5 ms and a
13
C spectral width of 1358 Hz to cover the two resonances in question. A 2 * 256 * 2048 data matrix with 64 scans per t1 value was used and zero
filled to 1024 * 2048 data points.
Chemical shift calculations
13
C chemical shift calculations were carried out with the C-13 NMR module of ChemWindow Version 3.1.3 which is an implementation of
the C13Shift program (
15
). This program utilizes linear additivity rules and an extensive parameter set
(
16
,
17
) to predict
13
C shifts with a mean deviation of -0.29 p.p.m. and a standard deviation of 5.5 p.p.m. for 97% of 160 000
shifts that were analyzed (
15
).
Molecular modeling
Torsion angle constraints were derived from the reported coupling constant data
(
2
) as has been previously described for the (6-4), Dewar and trans-syn-II photoproducts of TpT (
18
,
19
). The newly derived torsion angles were used along with the previously derived
interproton distance ranges as constraints in molecular mechanics minimizations
with SYBYL. The starting geometry for the thymine-adenine ring system of d(TpA)* was obtained by first carrying out the
cycloaddition between the 5,6-double bonds of T and A from a B form conformation in which the T was in
either an
anti
or
syn
conformation. The AC5-AC6 bonds of the resulting trans-syn-I-
1a
and cis-syn-
1a
adducts were then broken and the resulting structures minimized with torsion
angles and distance ranges imposed as constraints. Semi-empirical AM1 molecular orbital calculations were carried out with
Hyperchem Version 2.0.
RESULTS AND DISCUSSION
Assignment of the
13
C-NMR spectrum and
1
H-
13
C connectivities
The assignment of the
13
C-NMR spectrum of d(TpA)* was facilitated by the previously reported proton
NMR assignment based upon COSY and NOESY spectra (
2
) which was confirmed by TOCSY, ROESY and NOESY spectroscopy. To assign the
13
C spectrum and obtain bond connectivities
1
H-
13
C HMQC and HMBC NMR spectra were obtained. The strong HMQC crosspeaks (Fig.
3
) enabled assignment of all carbons directly attached to protons and included
all of the sugar carbons (Table
1
). The only base carbons that could be assigned by HMQC were the two aromatic
carbons bearing non-exchangeable protons of the adenine subunit and the C6 and methyl carbon
signals of the thymine subunit.
Structure and mechanism of formation of d(TpA)*
The presence of only three saturated carbon signals ([delta] 21.8, 49.4 and 58.7) that are assignable to the base portion of d(TpA)*
is inconsistent with the originally proposed structure for d(TpA)* (
2
,
3
) which would have five such signals. This data rules out any [2+2] adducts
between the 5,6-double bond of the thymine ring and any double bond of the adenine ring,
but suggests, that the d(TpA)* structure may be the result of a fragmentation
of an initial [2+2] adduct. Of the three [2+2] pyrimidine-purine adducts typified by structures
1
,
3
and
5
, only
1
appears to be able to fragment to a structure containing only two saturated
carbons that is also consistent with all of the other available data (Fig.
5
, structure
9
). A similar fragmentation has been proposed for the formation of A=A,
7
, from the [2+2] adduct of d(ApA),
6
(
6
) (Fig.
2
). Structure
9a
has the same molecular mass as
1a
, and because the crucial bond connections are preserved, structure
9a
would lead to the same acid hydrolysis product,
2
(Fig.
1
). Structure
9a
would also explain why d(TpA)* is not photoreversed by 254 nm irradiation, as
it is not a [2+2] adduct. The increase in pK of d(TpA)* relative to TpdA (from
3.5 to 5.5) (
1
,
3
) is also consistent with the formation of the more basic amidino functional
group present in
9a
.
Conformational analysis
To obtain models for the d(TpA)* photoproduct, torsion angle and distance range
constrained molecular mechanics energy minimizations were carried out. The
torsion angle constraints were derived from the reported coupling constant data
(
2
), according to methods that we have used to analyze the (6-4), Dewar and trans-syn-II photoproducts of TpT (
18
,
19
,
23
). Our evaluation of the coupling constant data according to unrestricted one
and two state models is in general agreement with that previously reported (
2
) (Table
3
). Unlike the previous study, which concluded that the deoxyribose ring of the T
could be described by only 55% of a C3'-endo-C4'-exo conformation (P = 42, [Phi] = 35), we conclude that this is the major if not
exclusive conformation (P = 30, [Phi] = 35 RMS deviation of 0.55 Hz). In accord with the previous study, we
conclude that the pdA sugar is indeed represented by 50% of a C3'-exo conformation, and in addition, that the remaining conformers
can be represented by a single C2'-exo conformer (RMS deviation of 0.32 Hz for the two state model).
We are also in agreement with the analysis of the backbone angles, except for
the previously reported value of 152o for [epsilon]. The coupling constant of 7.4 Hz corresponds to an HCOP torsion
angle of +-30o, which corresponds to [epsilon] values of 210o (-150o) and 270o (-90o), of which -150o is more
reasonable.
.
Conformational parameters (in degrees) of d(TpA)* used to constrain modeling
calculations
Tp
pdA (50%)
pdA' (50%)
[beta]
-
180
180
[gamma]
none
60
60
[delta]
+85
+158
+96
[epsilon]
-150
-
-
P,[Phi]
M
30, 35
198, 40
342, 40
C3'-endo
C3'-exo
C2'-exo
[tau]0,[tau]1,[tau]2,[tau]3,[tau]4
-8, -15, 30, -35, 26
-1,23,-39,38,-24
23,-39,38,-24,-1
Molecular models of d(TpA)*
Stereochemical assignment
The angle and distance constrained models for the 5
S
,6
R
and 5
R
,6
S
isomers of structure
9a
fit all the available NMR data quite well (Table
4
). The molecular models, however, suggest that the 5
S
,6
R
stereoisomer might be distinguishable from the 5
R
,6
S
isomer by the presence of an NOE between TH6 and TH3'. Such a NOE was not reported in the previous NMR study of d(TpA)* (
2
) or in the present study but an extremely weak crosspeak was observed between
these two protons in the NOESY spectrum of d(GTATTA*TG), but not in
d(GTA*TTATG).
.
Comparison of selected interproton distances in Å corresponding to observed and unobserved NOE's in d(TpA)* for the models
of the 5
S
,6
R
and 5
R
,6
S
stereoisomers of
9a
derived from distance and angle constrained molecular mechanics calculations
1st H
2nd H
Exp
a
5
S
,6
R
-a
5
S
,6
R
-b
Avg
b
5
R
,6
S
-a
5
R
,6
S
-b
Avg
b
TH6
TH1'
<3.5
3.3
3.3
3.3
2.6
2.5
2.5
TH2'
>3.5
4.3
4.3
4.3
3.0
3.0
3.0
TH2''
>3.5
4.9
4.9
4.9
3.5
3.5
3.5
TH3'
>3.5
c
4.0
4.0
4.0
4.7
4.8
4.7
AH8
TH6
>3.5
c
4.7
4.6
4.6
5.4
5.4
5.4
TH3'
<3.5
d
3.1
3.3
3.2
3.4
3.8
3.6
AH1'
>3.5
3.6
3.8
3.7
3.8
3.9
3.8
AH2'
<2.5
2.7
2.5
2.6
2.3
2.8
2.5
AH2''
>3.5
d
3.1
4.0
3.4
3.3
4.3
3.6
AH3'
>3.5
c
4.9
2.6
2.9
4.8
2.6
2.9
a
Distance ranges reported in the previous study (
2
).
b
The average corresponds to an r
-6
weighted average distance.
c
Distances with an asterisk correspond to NOEs that were not specifically described in the previous study and were confirmed to be absent in the present
study.
d
Misassigned in the previous study (see text).
Evidence that the d(TpA)* could have the 5
S
,6
R
-
9a
structure, comes from an explanation for the highly unusual chemical shift of
TH3'. The TH3' signal appears at 2.73 p.p.m. in d(TpA)*, compared with 4.59
p.p.m. in TpdA (
26
), and appears at 2.68 p.p.m. in d(GTATTA*TG) compared with 4.93 p.p.m. in the
parent oligonucleotide (
5
). The authors of the previous NMR study of d(TpA)* (
2
) attributed the large upfield shift of TH3' to shielding by the C2 carbonyl of the thymine in a syn glycosyl
conformation, which could also explain the upfield shift in the 5
R
,6
S
stereoisomer of
9a
. The large upfield shift of TH3' is also consistent with the 5
S
,6
R
stereoisomer of
9a
, as TH3' is situated directly under the imidazole ring and therefore in its
shielding region (Fig.
7
).
Thus, we conclude that the d(TpA)* product could arise from an
anti
,
anti
glycosyl conformation of the two nucleotides, as would be found in B DNA, and
as was originally proposed for the [2+2] adduct (
3
). Only one stereoisomer of the TA* product of d(GTAT
TA
TG) appears to be produced by 254 nm irradiation, and has an
1
H-NMR spectrum that is very similar to that of d(TpA)*, suggesting that it
also has the same stereochemistry (
5
). This fact, coupled with the fact that the TA* product is also produced in
double-strand DNA, in which the
anti
glycosyl conformation predominates, further supports the notion that TA* has
the 5
S
,6
R
stereochemistry. Such a stereochemical assignment would be more consistent with
the fact that the cis-syn dimer, which also arises from an
anti
,
anti
glycosyl conformation, is the major product of TpT, and TT in both single- and double-strand DNA, while the trans-syn-I isomer, which arises from the
syn
,
anti
conformation, is a very minor product (
23
,
27
). Though all the available data for d(TpA)* is consistent with structure
9a
, a crystal structure will be required to unambiguously determine its
stereochemistry and conformation, and attempts to crystallize d(TpA)* are in
progress.
ACKNOWLEDGEMENTS
This work was supported by NIH Grant R37 CA40463. The assistance of the
Washington University High Resolution NMR Facility, funded in part through NIH
Biomedical Research Support Shared Instrument Grants RR-02004, RR-05018 and RR-07155 is also gratefully acknowledged.
REFERENCES
1 Bose,S.N., Davies,R.J.H., Sethi,S.K. and McCloskey,J.A. (1983) Science 220, 723-725.
25 Sakamoto,M., Takahashi,M., Kimura,M., Fujihira,M., Fujita,T., Iida,I., Nishio,T. and Watanabe,S. (1994) J. Org. Chem. 59, 5117-5119.
26 Cheng,D.M. and Sarma,R.H. (1977) J. Am. Chem. Soc. 99, 7333-7348.
27 Patrick,M.H. and Rahn,R.O. (1976) In Wang,S.Y. (ed.), Photochemistry and Photobiology of Nucleic Acids. Academic Press, New York, NY, Vol. II, pp. 35-95.
