ABSTRACT
A major limitation in triple-helix formation arises from the weak energy of interaction between the
third strand and the double-stranded target. We tried to increase the stacking interaction
contribution within the third strand by extending the aromatic domain of
thymine. We report here the use of 2,4-quinazolinedione as a substitute for thymine in the canonical TA*T
triplet. The synthesis and the characterization of the quinazoline
[beta]
nucleoside Q and of its phosphoramidite derivative is described. Triple-helix-forming oligonucleotides incorporating Q have been prepared and
their ability to form triplexes has been evaluated by UV-monitored thermal denaturation measurements. The introduction of one or
multiple Q residues, either contiguous or remote from each other, slightly
destabilized triple-stranded structures, whatever the nucleic acid base composition
(pyrimidine or GT) of the third strand.
Artificial control of gene expression can be achieved by targeting a cellular
nucleic acid, usually mRNA, by a complementary antisense oligonucleotide. The
expression of numerous genes has been down-regulated in this way, either in cell-free extracts, in cultured cells or
in vivo
(for a review see
1
). The efficiency of the method is linked to the high affinity and specificity
of the interaction between the single-stranded mRNA and the antisense sequence. However mRNA self-association can reduce or prevent the binding of the antisense
oligonucleotide to the targeted region. The competition between intramolecular
and intermolecular base pairing will weaken or even abolish the expected
biological effects. This has been previously observed with antisense
oligonucleotides targeted to the mini-exon sequence of
Leishmania
, the motif present at the 5' end of every mRNA of trypanosomatids which can fold into a hairpin (
2
,
3
).
To circumvent this difficulty an alternative strategy was developed where the
antisense oligonucleotide is designed to interfere directly with a double-stranded region of mRNA, via the formation of a local triple helix (
4
,
5
). Such complexes are able to impair
in vitro
translation (Le Tinévez and Toulmé, unpublished results).
Triple helix formation obeys structural features that do not accomodate every
double-stranded sequence. At least two structural motifs have been characterized, which differ in the sequence
composition of the third strand. In the so-called `pyrimidine motif' or pyrimidine-purine-pyrimidine (YR*Y) triplexes (in this notation the third strand is written in the last position and *
indicates Hoogsteen interaction type), a pyrimidine third strand is aligned parallel to the purine strand of the Watson-Crick duplex. Sequence specificity is derived from thymine recognition of
adenine-thymine base pairs (TA*T base triplets) and protonated cytosine (C
+
) recognition of guanine-cytosine base pairs (CG*C
+
base triplets) arising through Hoogsteen hydrogen bonding
(
6
,
7
). In the purine motif or pyrimidine-purine-purine (YR*R) triplexes, a purine third strand is antiparallel to the purine strand, leading to CG*G,
TA*A and TA*T base triple combinations
(
8
,
9
; see
10
for a review). The interaction results from specific reverse Hoogsteen hydrogen bonding. Therefore
triplex formation requires the occurrence of homopurine stretches in double-stranded regions. Another limitation of triplex formation comes from the
weak energy of interaction between the third strand and the double-stranded target, leading to affinity constants at least two orders of
magnitude lower than that of the Watson-Crick duplex (
11
). Modified bases have been designed to overcome these limitations, allowing pH-independent lecture of G (
12
-
14
), increased affinity of the Hoogsteen strand
(
15
) or Watson-Crick base pair reversal lecture of homopurine track (
16
,
17
).
In an attempt to design specific bases for triple-helical structure formation, we explored the ability of extended aromatic
domain of bases in stabilizing triplexes. Stacking interactions between planar
heterocyclic ring of nucleic acids are largely involved in the stabilization of
DNA and RNA duplexes (
18
). The contribution of such interactions might also be of great importance for
triplex formation. Increasing stacking interactions through chemical
modification of nucleic acid bases may provide a means to get more stable
triplexes. As a first approach along this line we used a quinazoline nucleus
[quinazoline-2,4(1
H
,3
H
)-dione] (Q) as a substitute for thymine.
While the present work was in progress, Bhattacharya
et al
. (
19
) reported an improved synthesis of Q and its incorporation into G-rich triple-helix forming oligonucleotides. However, careful examination of
their structural data revealed conflicting assignment of the anomeric configuration of the Q nucleoside. We therefore report here the
chemical synthesis and the spectroscopical determination of [alpha] and [beta] anomers of the 2'-deoxy-d-ribofuranoside of Q. Using the
phosphoramidite derivative of the [beta] anomer we incorporated the new base Q in several oligonucleotides. The
hybridization properties of ODNs containing Q were determined by thermal
denaturation experiments (
T
m
). Duplexes are clearly destabilized by the introduction of a single Q residue
whereas triplexes are destabilized to a lesser extent depending on the third
strand sequence context.
Thin-layer chromatography (TLC) was performed on Merck silica gel 60 F254
aluminum-backed plates; visualization was done by UV illumination and by staining
with 10% perchloric acid solution (2-deoxyribose containing component only). Flash chromatography refers to column chromatography performed with Merck silica gel 60 (0.04-0.063 mm).
Melting points were determined on a Köfler bench. The NMR spectra were recorded on a Bruker AC200 spectrometer
working at 200 MHz for
1
H, 50.32 MHz for
13
C and 81.02 MHz for
31
P. The NOE, HMQC and HMBC experiments were performed on a Bruker AMX 500 spectrometer. The chemical shifts are expressed in p.p.m. using
TMS as internal standard (for
1
H and
13
C data) and 85% H
3
PO
4
as external standard (
31
P data). The IR and UV spectra were recorded on a Bruker IFS-25 and on a Kontron UVikon 940 spectrophotometer, respectively. Melting
experiments were performed on a Cary 1E UV-visible spectrophotometer with a temperature controller unit. Mass spectra
were recorded on VG Autospec spectrometer.
1-(3,5-di-
O
-
p
-toluoyl-2-deoxy-
[alpha]
,
[beta]
-d-erythro-pentofuranosyl)- quinazoline-2,4-(3H)-dione (
3a
,
b
)
. A mixture of quinazoline-2,4- (1
H
,3
H
)-dione (Aldrich; 6.5 g; 40 mmol), a few crystals each of ammonium sulfate and acetamide and a few drops of trimethylsilylchloride
was refluxed in 80 ml hexamethyldisilazane (HMDS) for 24 h under exclusion of
moisture. Excess of HMDS was removed
in vacuo
by co-evaporation with toluene. The residue was dissolved in dry CHCl
3
(500 ml) then 2-deoxy-3,5-di-
O
-
p
-toluoyl-d-pentofuranosyl chloride
2
(
20
) (6.2 g) and CuI (5 g) were added and stirred at room temperature (rt) for 19
h. The reaction was monitored by TLC (ethyl acetate:cyclohexane: 1:1) The
mixture was then poured on saturated NaHCO
3
solution (500 ml), and the organic phase was extracted by CH
2
Cl
2
, dried with MgSO
4
and evaporated. The protected nucleoside was purified by flash chromatography
in a gradient of hexane:diethyl ether (0-10%). The yield was 7.70 g (15 mmol, 96% ) of anomeric mixture (
3a
,
b
). On TLC the
R
f
= 0.44 in ethyl acetate:cyclohexane (3:7).
1
H NMR (200 MHz; mixture of anomers) (CDCl
3
), [delta] (p.p.m.): 2.4 (s, 6H, tol), 3.1 (m, 2H, H
2'
) 4.7 (m, 3H, H
4'
and H
5'
), 5.7 (m, 1H, H
3'
), 6.9 (m, 1H, H
1'
), 7.7 (m, 12H, ar);
13
C NMR (50.32 MHz) (mixture of anomers) (CDCl
3
), [delta] (p.p.m.): 21.6 (CH3, tol), 34.2-34.4 (C
2'
), 63.5-64.8 (C
5'
), 73.5-75.0 (C
3'
), 81.2-81.3 (C
1'
), 84.2-85.3 (C
4'
), 116.0 (C
8
), 116.6 (C
4a
), 123.7 (C
6
), 126.4 (C
ar
, tol), 129.0 (C
5
), 128.3-129.3-130.4 (C
ar
, tol), 134.8 (C
7
), 139.0 (C
8a
), 143.8 (C
ar
, tol), 149.9 (C
2
), 161.5 (C
4)
, 166.0 (carbonyl, tol).
1-(2-deoxy-
[alpha]
,
[beta]
-d
-
erythro-pentofuranosyl)quinazoline-2,4-(3H)- dione (
4a
,
b
).
The product
3a
,
b
(7.5 g; 14.6 mmol) was added to 1% methanolic sodium hydroxide (170 ml) at rt
and stirred for 1.5 h. The solution was concentrated and then poured on a short silica gel column
eluted with CH
2
Cl
2
:MeOH (1:1) to afford
4a
,
b
(7.1 g, 95% yield).
1-(2-deoxy-5-(4,4
'
-dimethoxytrityl)-
[alpha]
,
[beta]
-
d
-erythro-pentofuranosyl) quinazoline-2,4-(3H)-dione (
5a
,
b
)
. The product
4a
,
b
(4 g; 7.5 mmol) was carefully dried by three co-evaporations with pyridine and then dissolved in anhydrous pyridine (50 ml). Dimethoxytrityl
chloride (DMTCl, 2.8 g; 8.25 mmol) was added under an argon atmosphere and
mixed for 1 h. The reaction was poured on CHCl
3
(80 ml), the organic phase was washed, dried and evaporated. The two anomers
were then purified by flash chromatography (gradient of hexane:diethyl ether 0-100% followed by pure CH
2
Cl
2
always in the presence of 1% triethylamine), The total yield was 75% ([alpha]:[beta] = 45:55). Next, TLC was performed in CH
2
Cl
2
:MeOH:cyclohexane 8:1:1 yielding an
R
f
= 0.67 for [alpha], and 0.48 for [beta].
[alpha]
anomer (
5a
)
.
1
H NMR (200 MHz; CDCl
3
), [delta] (p.p.m.): 2.85 (m, 2H, H
2'
,
2''
), 3.3 (m, 2H, H
5'
), 3.77 (s, 3H), 3.8 (m, 1H, H
4'
), 4.42 (m, H, H
3'
), 4.82 (m, O
[beta]
anomer (
5b
)
.
1
H NMR (200 MHz; CDCl
3
), [delta] (p.p.m.): 2.16-2.9 (m, 2H, H
2'2''
), 3.5 (m, 2H, H
5'
), 3.69 (s, 3H), 3.96 (m, 1H, H
4'
), 4.7 (m, 1H, H
3'
), 5.26 (m, O
1-(2-deoxy-
[alpha]
-d-erythro-pentofuranosyl)quinazoline-2,4-(3H)- dione (
7a
)
. The product
5a
, 250 mg (0.42 mmol), was stirred for 30 min at rt with 8 ml of a 80% acetic
acid aqueous solution. Then the mixture was co-evaporated three times with MeOH and twice with toluene. The purification
by flash chromatography in CH
3
OH:CH
2
Cl
2
(1:1) afforded 100 mg of
7a
(0.35 mmol, 84% yield). m.p.: 160-162oC (crystallization from water). UV (MeOH) 274 nm ([epsilon] = 5600), 242 nm ([epsilon] = 7800), 221 nm ([epsilon] = 22900); IR (KBr) (cm
-1
) was 3400, 3042, 1714, 1680, 1604.
1
H NMR (500 MHz; DMSO) [delta] (p.p.m.): 2.39 (m, 1H, H
2''
), 2.51 (m, 1H, H
2'
), 3.53 (m, 2H, H
5'
), 4.09 (m, 1H, H
4'
), 4.35 (m, 1H, H
3'
), 4.89 (broad s, O
1-(2-deoxy-
[beta]
-d-erythro-pentofuranosyl)quinazoline-2,4-(3H)- dione (
7b
).
The product
5b
, 250 mg (0.42 mmol), was stirred for 30 min at rt with 8 ml of an acetic acid
aqueous solution (AcOH:H
2
O = 8:2). Then the mixture was co-evaporated three times with MeOH and twice with toluene. The purification
by flash chromatography (MeOH:CH
2
Cl
2
= 1:1) afforded 100 mg of
7b
(0.35 mmol; 84% yield). m.p.: 202-203oC (crystallization from water) (lit. 185-186oC); UV (MeOH) 307 nm ([epsilon] = 6400), 240 nm ([epsilon] = 12800), 227 nm ([epsilon] = 23400); IR (KBr) 3400, 3042,
1714, 1680, 1604; MS (FAB
+
) 279.1 (MH
+
);
1
H NMR (DMSO) [delta] (p.p.m.): 1.90 (m, 1H, H
2'
[alpha]), 2.61 (m, 1H, H
2'
[beta]) 3.59 (m, 2H, H
5'
), 3.67 (m, 1H, H
4'
), 4.35 (m, 1H, H
3'
), 5.1 (broad s, O
1-(2-deoxy-3-
O
-(2-cyanoethoxy(diisopropylamino)-phosphino)- 5-(4,4
'
-dimethoxytrityl)-
[beta]
-d-erythro-pentofuranosyl)quinazoline- 2,4-(
3
H)-dione (
6b
).
The product
5b
(200 mg, 0.34 mmol) was dried carefully by three co-evaporations with pyridine and dissolved in anhydrous CH
2
Cl
2
(1 ml). Then,
N,N
-diisopropylethylamine (0.225 ml) was added followed by slow addition of 2-cyanoethyl-
N
,
N-
diisopropylphosphoramidochloridite (120 mg in 0.250 ml of dry CH
2
Cl
2
) under an argon atmosphere and stirred for 1 h at rt. The reaction was quenched with MeOH (4 [mu]l) and diluted with ethyl acetate (15 ml). The solution was washed with
saturated aqueous solution of NaHCO
3
and NaCl, then dried and evaporated under reduced pressure. Crude
6b
was purified twice by flash chromatography with AcOEt:CH
2
Cl
2
:TEA (45:45:10) as eluent to afford 217 mg of pure phosphoramidite (0.27 mmol,
yield 79%). MS (FAB
-
) 779.2 (M-H)
-
;
1
H NMR (CDCl
3
) [delta] (p.p.m.): 0.95-1.21 (m, 12H, iPr), 2.34-2.8 (m, 2H, 2'), 2.6 (m, 2H, C
All oligonucleotides were synthesized at 0.2 [mu]mol-scale on a Millipore Expedite 8909 DNA synthesizer using conventional [beta]-cyanoethyl phosphoramidite chemistry. The standard and the
modified bases were dissolved in anhydrous CH
3
CN (0.1 M final concentration). The modified phosphoramidite was used with a coupling time of 15 min. The coupling efficiency was the same as that of
unmodified amidite (>98%). All oligomers were synthesized `trityl on'. After
synthesis, the solid supports were treated overnight at 55oC with fresh, concentrated NH
4
OH (3 ml), the solution was then concentrated to dryness. The crude tritylated
oligonucleotide was purified by reverse-phase HPLC (column Nucleoside 300-5 C18) using the following gradient system: A (triethylammonium
acetate 0.1 M, pH 7); B ( 0.1 M triethylammonium acetate in 80% acetonitrile).
A linear gradient of 0-60% buffer B over 60 min at a flow of 1 ml/min was used. Detection was done at 260 nm for analytical
runs and 290 nm for preparative ones. After collection, oligomers were
evaporated and detritylated for 1 h by 1 ml of 80% acetic acid solution. The
solution was then evaporated, the residue resuspended in 1 ml of water and
extracted by ethyl acetate. Finally, oligonucleotides were precipitated using
n
-butanol. When required aliquots of purified oligonucleotides were analyzed
by gel electrophoresis to confirm the expected length and purity.
The RNA oligomer was purchased from Eurogentec.
Purified oligonucleotides, 0.5 nmol of each, were dissolved in 0.5 ml of the
appropriate buffer and boiled for 2 min. In case of double-helix formation, a buffer containing 10 mM sodium cacodylate (pH 7), 50 mM
NaCl and 1 mM magnesium acetate was used, except for duplexes D8 and D9 which
were incubated in the buffer used for triple-helix formation: 10 mM sodium cacodylate (pH 7), 100 mM NaCl, 10 mM
magnesium acetate, 1 mM spermine.
Samples were kept at least 30 min at 4oC and were then heated from 4 to 90oC at a rate of 0.5oC/min, the absorbance at 260 nm was measured every 30 s.
Snake venom phosphodiesterase (SVPDE)
. The oligonucleotides (2 nmol mixed with 3 pmol of the 5'
32
P-labelled derivative) were dissolved in 0.13 ml of the following buffer
(0.1 M Tris-HCl, pH 8, 0.1 M NaCl, 14 mM MgCl
2
) and digested with 0.2 U of SVPDE (Boehringer) at 37oC. Aliquots (10 [mu]l) were removed at different times after enzyme addition and submitted
to phenol-chloroform extraction. The samples were then loaded on a 20%
polyacrylamide gel containing 7 M urea and exposed to autoradiography.
Bovine spleen phosphodiesterase (BSPDE)
. The oligonucleotides (2 nmol mixed with 10 pmol of the 3'
32
P-labelled derivative) were dissolved in 0.15 ml of 0.3 M sodium citrate, pH
6 and digested with 0.3 U of BSPDE (Boehringer) at 37oC. Aliquots (10 [mu]l) were removed at different times after enzyme addition. The analysis
was then carried out as decribed above.
Escherichia coli RNase H
. The oligonucleotides (13 pmol of DNA strand and 1.3 pmol of 5"
32
P-labelled complementary RNA) were dissolved in 0.13 ml of the following buffer (0.02 M Tris-HCl, pH 7.5, 0.1 M KCl, 10 mM MgCl
2
, 0.1 M dithiothreitol) and digested with 1.5 U of
E.coli
RNase H (Promega). Aliquots were removed at different times and transferred to
an equal volume of EDTA (50 mM) on ice. The analysis was then carried out as decribed above.
Our route to the phosphoramidite derivative of the quinazoline nucleoside is
illustrated in Figure
1
. The glycosylation of silylated quinazoline-2,4-dione (
1
) with 1-(chloro-2-deoxy-3,5 di-
O
-
p
-toluyl)-[alpha]-d-
erythro
-pentofuranose (
2
) in dry chloroform via CuI (
21
)
catalyst was found suitable to prepare appropriate quantities of the mixture of [alpha] and [beta] nucleosides (
3a
,
3b
) in 96% yield. Removal of the protecting toluoyl groups was accomplished by treatment with 1% sodium
hydroxide in methanol, and the corresponding nucleosides (
4a
,
4b
) were isolated in high yield (95%). The free nucleosides were then converted to
the corresponding 5-
O
-(4,4- dimethoxytrityl) derivatives by treatment with 4,4-dimethoxytrityl chloride (DMTCl) in anhydrous pyridine.
Purification of the reaction products by silica gel column chromatography
provided pure
5a
(45%) and
5b
(55%) in 75% total yield.
The structure of the various nucleoside derivatives (
3a
,
b
,
4a
,
b
) was checked by
1
H and
13
C NMR spectroscopy, which clearly revealed the presence of the two anomeric
configurations. The pure anomers (
5a
and
5b
) were obtained as monotritylated derivatives. A small amount of each was
deprotected using aqueous acetic acid to afford the free nucleosides
7a
and
7b
which were then submitted to spectroscopic analysis to allow the determination
of the [alpha] and [beta] configurations. The
1
H and
13
C NMR assignments are reported in Table
1
. They were reached by
1
H-
1
H and
1
H-
13
C correlation methods, COSY and HMQC
(
22
)
(data not shown). Long-range
1
H-
13
C correlation spectroscopy (HMBC) (
23
) allowed us to assign unambiguously all the protons (H-5, H-6, H-7, H-8) and the corresponding carbon atoms belonging to the
quinazoline nucleus, starting from the H-1 position. As shown in Table
1
, proton and carbon chemical shift differences between the two isomers, as well
as spin-spin coupling pattern, did not allow unambiguous determination of C-1 configurations. We therefore investigated the dipolar
interactions of nearby protons which were revealed by 2D ROESY experiments
(Fig.
2
). Thus
7a
was assigned the [alpha] configuration since a strong NOE crosspeak was detected between H-1" and H-3". This correlation was missing in the
7b
ROESY map which additionally revealed interactions between H-1" and H-4" as expected for a [beta] configuration. Additional correlations on the [beta] anomer (
7b
) led us to propose a mixture of
syn
and
anti
-conformation of the base and sugar moiety. NOE crosspeaks between H-1" and H-7, H-8 are consistent with an anti-conformation as well as H4" and H5" with H-8 whereas H-2", H-3",
H-5" cross- peaks with H-7 proton support a
syn
-conformation (Fig.
3
). The [alpha] isomer seems to have a preferential
anti
conformation since the main correlations exhibited in Figure
3
arose from aromatic protons (H-7, H-8, H-6) and the deoxyribose moiety (H4", H2").
Table 1
1
H and
13
C NMR data of compounds 7a and 7b. NMR data ([delta] p.p.m. from TMS) in DMSO-
d
6.
Table 2
Oligonucleotide sequences are listed in Table
2
. The potential ability of the quinazoline modification to form stable complexes with
complementary sequences was evaluated with single-stranded (duplexes D1-D7) and double-stranded targets (triplexes T1-T15). Duplexes D8 and D9 correspond to control
experiments. The triplexes were designed in order to allow easy analysis of the
effect of single or multiple substitutions of thymine by a quinazoline base
within purine or pyrimidine motifs. In all cases, the Hoogsteen third strand
was bridged by a pentanucleotide loop to one of the target strand, leading to
the formation of bimolecular complexes. The thermal denaturation of such
complexes was expected to occur in a single transition from triplex to random
coil (
24
,
25
), making
T
m
determination more accurate. All compounds were prepared with the same CACAC
pentanucleotide loop.
We tested first the ability of modified oligonucleotides to bind to a single-stranded DNA targets. Results show (Table
2
) that the substitution of one T by Q in one strand (D2) led to a 5.5oC decrease in
T
m
compared with the unmodified duplex (D1). Further incorporation of Q either
isolated (D4) or consecutive (D3), did not induce a significantly more
pronounced destabilization of the complex. Similar results were observed with
RNA targets (D5-D7).
The recognition of a TA base pair by Q, resulting in the formation of TA*Q base
triplets as a substitute for TA*T triplet was evaluated from
T
m
determination of triplexes T1-T6. Introduction of Q was done 1-3 times in the Hoogsteen strand of these bimolecular complexes. As
expected, the melting curves for these complexes showed a single transition
from bound to dissociated structures (not shown). Triplex formation was
revealed by comparing
T
m
data from corresponding double strand D8 (33oC) and complex T1 (41.5o). One Q substitution led to a [Delta]
T
m
of -2.5oC (T2) which was not additive since no more decrease was observed
for the second substitution (compare T2 and T5). However, consecutive
substitutions led to further destabilization: 38oC (T3) and 37oC (T4). These data can be compared with the substitution of T by G (36oC, T6) which is the least stable base triplet in this context (
26
). Thymine is also involved in the recognition of GC base pairs either in
pyrimidine or purine motifs. We therefore evaluated the binding properties of Q
to GC. We observed a decrease of melting temperature for GC*Q (35oC, T8) compared with the unmodified GC*T (37.5oC).
Similar experiments were conducted with purine motifs. Complexes T9 to T15 were
designed to form CG*G and TA*T base triplets, the Hoogsteen third strand
exhibiting an antiparallel orientation relative to the purine strand. Triplex
formation is revealed by a 8oC increase in
T
m
values from the double strand D9 to the triplex T9. Substitution of one or two
Ts by Q in the Hoogsteen strand led to a slight depression in
T
m
from 1.5-2.5oC for complexes T11 and T12, respectively. The only exception was
observed for triplex T10 where the subsitution was located 3 bp from the 3' end of the third strand which do not destabilize the structure. ([Delta]
T
m
= 0).
The ability of quinazoline Q to recognize a GC inversion in PyPu*Pu triplexes
was evaluated with complexes T13 to T15. Comparison of
T
m
data from T13 and T14 revealed a slight destabilization of the modified triplex
([Delta]
T
m
= -1oC). The GC*Q triple was much more stable than GC*G introduced in T15
as a reference and which revealed to be the worst triplet in this context (
27
).
We studied the stability of oligonucleotides incorporating the base Q towards
snake venom phosphodiesterase (SVPDE, 3' exonuclease) and bovine spleen phosphodiesterase (BSPDE, 5' exonuclease). We used the modified strand from duplex D4 which
contained two quinazoline bases at position 5 and 10 of the 12mer strand. The
oligonucleotides were either
32
P labelled at 5' using [[gamma]-
32
P]ATP and T4 polynucleotide kinase or 3' end-labelled with T4 nucleotidyl terminal transferase and [[alpha]-
32
P]dideoxy-ATP. The degradation products were analyzed by electrophoresis on a denaturating
gel. Comparative studies of the 12mer modified strand versus the unmodified
oligonucleotide showed identical time course profile for both oligomers either
when incubated with SVPDE or BSPDE enzymes. Oligonucleotides were fully degraded after 30 min incubation. No accumulation of partial degradation compounds could be
detected at the expected position of modified bases (data not shown).
RNase H cleavage, which is thought to play a key role in the mechanism of action
of antisense oligonucleotides, was also studied on duplex D7 associating the
same modified DNA strand as the one used for nuclease studies and the RNA
target. The degradation products, following incubation of the hybrids with
E.coli
RNase H, were analyzed by polyacrylamide gel electrophoresis. The analysis
realized on modified and unmodified hybrids revealed that the modification did
not alter the ability of RNase H to cleave the RNA strand, the same
fragmentation pattern was observed (data not shown).
The chemical synthesis of quinazoline-2,4-dione deoxyribofuranosides was first reported by Stout and Robin (
28
) using the direct glycosylation of silylated quinazoline with 2'-deoxyribofuranoside chloride. This way led to an unresolvable mixture of anomeric
nucleosides. Dunkel and Pfleiderer, in 1992 (
29
) synthesized the 1-(2-deoxy-[beta]-d-erythropentofuranosyl)quinazoline-2,4-dione via the chemical
deoxygenation of quinazoline-2,4-dione ribonucleoside (
30
), leading to the selective [beta] introduction of the quinazoline ring. Our synthetic pathway uses the
direct glycosylation procedure in the presence of CuI catalyst, the purification was performed on the 5' monotritylated derivatives. Battacharya and co-workers in their recent report (
19
) follow the same synthetic chemical pathway without the use of CuI as a
catalyst. Examination of TLC conditions revealed that they chose the less polar compound
as the [beta] nucleoside contrary to our own work where the [beta] nucleoside was assigned the more polar spot on the TLC plates.
Careful comparison of
1
H and
13
C NMR data of
7a
([alpha] nucleoside) and the NMR data reported by Bhattacharya
et al.
for their `[beta]' nucleoside (Table
1
) clearly confirm the identity of the two compounds. The only difference
concerns the multiplicity of H-1' proton which appeared as a doublet of doublet for
7a
and a triplet in their case. The unambiguous assignment of the [beta] configuration to
7b
lies on the NOE measurement for the two anomers
7a
and
7b
. Such two dimensional analysis has already been used in the determination of purine anomers (
31
). Therefore,
7b
is identical to the compound of Dunkel
et al.
(
29
) obtained from a stereospecific synthetic scheme as can be readily seen in
Table
1
, and corresponds to the [beta] anomer whereas Bhattacharya and co-workers wrongly assigned the [beta] conformation to a compound which corresponds to
7a
, actually the [alpha] anomer. Consequently, the incorporation of this anomer into otherwise [beta] oligomer led to chimeric [alpha][beta] oligos.
Studies on oligo [[alpha]]-deoxynucleotides have clearly established the increase in stability
of these anomers toward purified exo-nuclease (
32
). Our study on nuclease sensitivity of Q-containing oligomers which revealed that incorporation of this
modification failed to induce any resistance to exonuclease, constitute an
additional evidence for a [beta] configuration of nucleoside
7b
. It is worth noting that the [alpha]-oligonucleotides do not elicit RNase H cleavage. We might have
expected a modified cleavage pattern for [alpha]-containing oligomers.
Antisense strategy rests mainly on the high specific formation of AT and GC base
pairs. The triplex-based approach specificity is derived from the previously mentioned base
triplets formation (TAT, GCC
+
, TAA, CGG). Strategies implying formation of Watson-Crick base pairs with the single-stranded part of the target, together with formation of triplets
for the double-stranded stem of the structured mRNA, would ideally need specific bases
for single strand and double strand recognition in order to increase the
overall specificity of the antisense oligonucleotide.
The properties exhibited by the quinazoline base do not fulfill completely the
above mentioned requirements. Q seems to interact preferentially with double
strand, but it does not improve stability of triplex-forming oligonucleotides. However in contrast to what has been reported by
Bhattacharya
et al.
(
19
), Q does allow the formation of triplexes either in parallel or antiparallel
context. Their inability to observe complex formation is likely related to the
use of the wrong anomer.
Few pyrimidine derivatives are able to enhance the stability of duplexes. A
pyrimidine analogue of cytosine (pyrido-pyrimidine) with an extended aromatic ring was shown to stabilize double
helices when incorporated in a self complementary dodecamer (
33
). In contrast 5,6-dimethyl-2'-deoxyuridine derivatives destabilize double-stranded structures (
34
). As can be deduced from base stacking of B type double helices (
35
), 5 and 6 positions of pyrimidines bases do not contribute to extensive overlap
of adjacent heterocyclic rings. Among C-5 analogues of 2'-deoxypyrimidines, 5-(1-propynyl)-2'-deoxyuridine and 5-(1-propynyl)-2'-deoxycytidine are the most efficient in stabilizing double-helix complexes (
36
). The 3 carbon arm of 5 propynyl-substituted bases seems to be able to increase stacking interaction (
37
).
Substitution of T by Q in the Hoogsteen third strand of triplexes does not favor
triple-helix formation. The
T
m
decreases are however weaker than the one observed for duplexes. No clear-cut differences were observed between PyPu*Py and PyPu*Pu triplexes.
However, the introduction of a single modification in the antiparallel purine
third strand led to a weak destabilization of triplex structures and in one
case (T10) did not destabilize the complex.
Using T offers a means of recognizing GC base pairs within pyrimidine motif (
38
) or the more favorable purine context (
27
). In this last case a slight
T
m
decrease (-1oC) was found for the GC*Q base triplet. GC*Q triplet within
pyrimidine third strand is less efficient than GC*T in stabilizing the triple-helix complex.
When taken altogether the data reported here on triplex formation with a
quinazoline base lead to the conclusion of poor stacking interactions within
the third strand. This phenomenon could originate in a lack of overlap of
aromatic ring between adjacent base triplets. Optimum overlap of [pi] orbitals between consecutive triplets have been proposed to account for the significant
enhanced stabilization brought by 5-propyne uridine substituent in triplex formation (
39
). According to the stacking configuration presented in this last study, we
expect only a partial overlap of the benzo moiety of the quinazoline
heterocycle with the C
2
-O
2
bond of the adjacent thymine in the third pyrimidine strand. Additionally, a
hydrophobic contribution has been proposed (
36
) to account for duplex or triplex enhanced stability (
39
) following alkyl substitution on the 5 position of the pyrimidine ring.
According to our results, this contribution seems not of prime importance for
the quinazoline ring where the aromatic [pi] domain is largely increased in comparison to thymine. Furthermore clusters
of hydrophobic quinazoline nucleus (triplexes T3 and T4) did not yield any
contribution to the stabilization of triple-stranded structures. A recent study (
40
) has pointed out that stacking interactions between aromatic rings may not
result from classical hydrophobic effects and that dispersion and polarization
interactions have to be considered. Further improvements in modified
heterocyclic bases for optimization of triplex structures may combine extension
of overlapping domains and additional dipolar contribution to interaction
forces.
We thank C. Quéneudec for her help in nuclease studies. We are grateful to Noël Pinaud (Laboratoire de Pharmacognosie, Université de Bordeaux II), for his skilfull NMR measurements (500
MHz). We acknowledge the Etablissement Public Regional d'Aquitaine for a grant
no. 940303001.
REFERENCES
Return