ABSTRACT
The antitumor antibiotic CC-1065 binds in the minor groove of double-stranded DNA, and the cyclopropapyrroloindole (CPI) subunit of the
drug alkylates adjacent adenines at their N-3 position. We have attached racemic CPI to oligodeoxyribonucleotides
(ODNs) via a terminal phosphorothioate at either the 3
'
- or 5
'
-end of the ODNs. These conjugates were remarkably stable in aqueous
solution at neutral pH even in the presence of strong nucleophiles. When a 3
'
-CPI-ODN conjugate was hybridized to a complementary DNA strand at 37
o
C, the CPI moiety alkylated nearby adenine bases of the complement efficiently
and rapidly, with a half-life of a few minutes. The 5
'
-CPI-ODN conjugate showed very little reactivity within the duplex. CPI-ODN conjugates should be highly effective sequence-specific inhibitors of single-stranded viral DNA replication or gene selective
inhibitors of transcription initiation.
Modified oligodeoxyribonucleotides (ODNs) capable of hybridization-triggered crosslinkage in a physiologic buffer could be used as sequence
specific affinity labeling reagents and might have potential as gene selective
drugs. Ideally, these conjugates would be chemically inert until hybridization,
triplexation or synapsis, whereupon they would rapidly and efficiently
crosslink to the complexed DNA or RNA. Examples of attempts at the design of
such agents are rare. Matteucci and coworkers showed that
N
4
-ethanocytosine-containing ODNs meet some of these criteria (
1
-
3
), but are slow in the crosslinking reaction, with a half-life of ~30 h. The alkylated base was a mismatched cytosine. Rokita
et al
. used a stable silylated phenolic quinone methide precursor (
4
), which, upon hybridization, lost the silyl group to generate the reactive
species. Alkylation of the complementary strand occurred with moderate
efficiency over several hours.
Other crosslinkable ODNs fail to meet these specifications because the appended
crosslinking moiety has inherent reactivity leading to self-alkylation or to the modification of cellular nucleophiles prior to
nucleic acid binding. Such agents include the
p
-(
N
-2-chloroethyl-
N
-methylamino)-benzyl group (
5
), chlorambucil (
6
,
7
), haloacetamidoalkyl groups (
8
,
9
) and binuclear platinum complexes (
10
,
11
). Other appended crosslinking groups act under non-physiologic conditions or require unusual cofactors, such as ketal (
12
) (requiring low pH), thioether (
13
) (requiring CNBr) and silyl phenol (
14
) (requiring KF) groups. A variety of photo-crosslinkable ligands have been conjugated to ODNs, most notably psoralen
(
15
,
16
). Since these agents covalently react upon exposure to near ultraviolet light,
their use
in vivo
is problematic.
We describe the synthesis and properties of a new class of hybridization-triggered crosslinkable ODNs which are conjugated to the reactive
cyclopropapyrroloindole (CPI) subunit of the potent antitumor antibiotic CC-1065 (
17
,
18
). This DNA alkylating agent is composed of three repeating 1,2-dihydro-3
H
-pyrrolo[3,2-
e
]indole subunits (Fig.
1
). The two non-reactive N-terminal subunits (B and C) confer a high binding affinity for the
deep, narrow minor groove of A-T rich DNA. The C-terminal A subunit (CPI), which contains an electrophilic cyclopropyl
moiety, contributes additional binding affinity and is responsible for the
N
-3 alkylation of adenine (
19
). CC-1065 is very stable in neutral aqueous solution (
20
), but becomes activated when bound to the minor groove of double-stranded DNA. Alkylation of adenine occurs as a result of DNA mediated
general acid catalysis (
21
,
22
) within low affinity 5'-(A/T)(A/T)A* (* denotes preferred alkylation site) and high
affinity 5'-PuNTTA* and 5'-AAAAA* consensus sequences (
23
,
24
). The C subunit of the natural (+)-enantiomer of CC-1065 binds to the 5'-end of these sequences, thus directing the reactive CPI
subunit to the vicinity of the 3'-adenine. Alkylation of adenines in dsDNA by (+)-CPI alone is substantially less efficient and less selective
than by (+)-CC-1065 (
24
). This is a consequence of the low binding affinity of (+)-CPI for A-T rich dsDNA. While (-)-CC-1065 binds to DNA with the opposite polarity as
the (+)-enantiomer, it alkylates just as well. The (-)-enantiomer of CPI, on the other hand, reacts with a reversed
binding orientation 10-100-fold more slowly than (+)-CPI (
24
,
25
).
All air and water sensitive reactions were carried out under argon. Anhydrous
solvents were obtained from Aldrich (Milwaukee, WI). Flash chromatography was
performed on 230-400 mesh silica gel. UV-visible absorption spectra were recorded on a Lambda 2 (Perkin
Elmer) spectrophotometer with a PTP-6 temperature controller.
1
H NMR spectra were run at 20oC on a Varian 300 spectrometer, and chemical shifts are reported in p.p.m.
downfield from Me
4
Si. Racemic compound
1a
was prepared as previously described (
20
,
24
,
26
).
Racemic 3-(
tert
-Butyloxycarbonyl)-1-chloromethyl-1,2-dihydro- 3
H
-8-methylpyrrolo[3,2-e]indol-5-ol
(1a)
(0.22 g, 0.65 mmol) was converted into
1b
by treatment with HCl in ethyl acetate according to the literature procedure (
24
). The hydrochloride
1b
was dissolved in 10 ml of anhydrous DMF to which tBOC-
N
- aminocaproic acid (0.31 g, 1.3 mmol) and EDC (
N
-(3-dimethylaminopropyl)-
N
'-ethylcarbodiimide hydrochloride) (0.77 g, 4 mmol) were added. After
stirring for 3 h, the reaction mixture was concentrated
in
vacuo
to an oil and triturated with water (25 ml). The resulting solid was
centrifuged, washed with water, centrifuged again, and dried in
vacuo
. The crude material was purified by flash chromatography in
dichloromethane-methanol (9:1) to give
1c
as an off-white solid (0.195 g, 67%):
1
H NMR (CDCl
3
) [delta] 10.27 (s, 1H), 9.40 (s, 1H), 8.20 (s, 1H), 7.00 (s, 1H), 4.30 (br s,
1H), 4.4-3.8 (m, 4H), 3.4-3.1 (m, 3H), 2.7-2.45 (m, 2H), 2.42 (s, 3H), 2.0-1.7 (m, 2H),1.7-1.3 (m, 13H, partially overlapping with H
2
O and tBOC singlet).
1c
(150 mg, 0.33 mmol) was treated with 5 ml of 3 M HCl in ethyl acetate. The
reaction mixture was stirred for 15 min and then evaporated
in vacuo
to dryness. To a solution of the resulting amine hydrochloride (
1d
) in 3 ml of anhydrous DMF was added NaH (41 mg, 1.7 mmol) suspended in 1 ml of
DMF. After stirring for 30 min, a solution of
N
,
N
-diisopropylethylamine hydrochloride (150 mg, 1.1 mmol) in 0.5 ml of DMF was added to quench excess NaH. The reaction mixture containing
2a
was immediately used for the next step. Bromoacetic acid
N
-hydroxysuccinimide ester (180 mg, 0.7 mmol) was added and the mixture was
stirred for 3 h. The solvent was removed
in vacuo
and the resulting mixture was separated by reverse phase HPLC (PRP-1, Hamilton Co, 7 * 300 mm) in a gradient of acetonitrile in water (30-100%). The desired compound
2b
was obtained as a colorless solid after removal of the solvent in 15% yield
(from
1c
):
1
H NMR (CDCl
3
) [delta] 9.84 (s, 1H), 6.85 (s, 1H), 6.69 (br s, 1H), 4.15 (m, 1H), 4.06 (s, 2H),
4.00 (m, 1H), 3.34 (m, 2H), 2.89 (m, 1H), 2.52 (m, 2H), 2.02 (s, 3H), 1.99 (m,
1H, overlapping with CH
3
signal), 1.74 (m, 2H), 1.57 (m, 2H), 1.40 (m, 2H), 1.21 (m, 1H).
The unmodified target ODN was prepared from 10 [mu]mol of polymeric support (Pharmacia) on an OligoPilot DNA synthesizer (Pharmacia) using the protocol supplied by the manufacturer. Standard reagents for the [beta]-cyanoethyl phosphoramidite coupling chemistry were purchased from Glen Research. 5'- and 3'-thiophosphate modifications were
introduced using a phosphorylating phosphoramidite in combination with a sulfurizing reagent (Glen Research).
To a solution of an ODN with a terminal phosphorothioate (50 U A
260
, ~0.25 [mu]mol) in 20 [mu]l of water were added triethylamine (0.5 [mu]l) and
2b
(20 [mu]l of a 33 mM solution in DMF, 0.66 [mu]mol). After 2 h, the solution was diluted with 0.8 ml of water and
loaded onto a reverse phase HPLC column (PRP-1, 7 * 300 mm). The conjugates were resolved in 50-60% yields as cleanly separated peaks using an acetonitrile
gradient (0-30%, 50 mM triethylammonium acetate pH 8).
The 30mer target ODN was 5'-end-labeled by using T4 polynucleotide kinase and [[gamma]-
32
P]ATP and purified by 8% denaturing PAGE. CPI-ODNs (10 [mu]M) were incubated with the complementary or non-complementary 30mer target ODNs (2 [mu]M) in 10 mM HEPES, pH 7.4 and 100 mM NaCl either at 25oC for 90 min or at 37oC for 30 min. Crosslinked products were detected
by direct analysis of reaction aliquots in an 8% denaturing polyacrylamide gel.
Alkylation sites were further converted into nicks with 5'- and 3'-phosphate termini by incubation at 95oC for 30 min in the reaction buffer, followed by
treatment with 10% piperidine (
19
,
27
). Sites of cleavage in the target 30mer were mapped relative to G and G+A
sequencing ladders (
28
) in an 8% denaturing polyacrylamide gel. The percent of alkylation was
determined by quantitative phosphorimage analysis (Bio-Rad) of the cleavage products.
The conjugation chemistry is based on alkylation of a terminal thiophosphate
residue on the ODN by a CPI derivative (
2b
) with a bromoacetamide group attached. We presumed that, at neutral pH, this
latter electrophilic group would be more reactive in solution than CPI to
sulfur nucleophiles. The preparation of
2b
is shown in Figure
2
. Racemic chloromethyl derivative
1a
(
24
) was converted into amine hydrochloride
1b
, which in turn was reacted with
N-
tBOC-aminocaproic acid to give
1c
. Deprotection afforded
1d
and cyclization in the presence of NaH gave
2a
. The reaction mixture containing
2a
was condensed with bromoacetic acid
N
-hydroxysuccinimide ester to provide the desired bromoacetamide derivative
2b
in 15% yield (unoptimized) from
1c
.
Previous investigations have shown that CPI derivatives are very stable in
neutral aqueous solution. Solvolysis of these compounds is very slow above pH 3
(
20
). Realizing that covalent attachment of the CPI residue to an ODN could alter
its stability, we monitored the long wavelength band in the UV spectrum of
these conjugates over time to determine the solvolytic stability of the
cyclopropane ring. Loss of this absorbance indicates reaction of the CPI. No
detectable reaction was observed at pH 7.2 for 3 days for any of the
conjugates. In the presence of 10 mM glutathione, the conjugated CPI moiety
reacted slowly at 37oC (t
[1/2]
= 36 h) but not at room temperature (data not shown). Although
2b
was found to be stable for days in aqueous solution at pH 4.5, the CPI residue
in the conjugates underwent slow acid-catalyzed hydrolysis, with a t
[1/2]
= 170 min at 21oC in a pH 4.5 solution.
Alkylation of a complementary ssDNA by the (+/-)-CPI-ODN conjugates was initially evaluated
spectrophotometrically using 4 * 10
-6
M target ODN and 8 * 10
-6
M CPI-ODN conjugate in 100 mM NaCl and 10 mM HEPES, pH 7.4. The structures of
the target ODN and CPI-ODN conjugates used in this study are shown in Figure
3
. Two 18mer CPI-ODNs were designed to target the same oligoadenylate sequence embedded in
the middle of the complementary target ODN. One of the reactive ODNs had the
CPI residue attached at the 3'-end while the other had it attached at the 5'-end. A rapid reaction (t
[1/2]
= 2 min at 37oC) was observed when the 3'-CPI-ODN conjugate was incubated with the target strand.
Upon completion of the reaction the long wavelength absorption had been reduced
by half (Fig.
4
). When assayed by UV monitoring, no reaction was observed for the 5'-CPI-ODN.
Addition of an electrophilic moiety to a ligand such as an enzyme inhibitor or a
receptor antagonist has often proven to dramatically enhance the potency of the
ligand by formation of a covalent bond with the target macromolecule. It is
important, however, to maintain specificity by minimizing non-target reactivity of the electrophile. The best way of accomplishing this
is by design of groups that are activated in the course of reversible binding
to the target. Elegant examples of this are found in several mechanism-based enzyme inhibitors (
30
).
CC-1065 is a potent inhibitor of DNA function for exactly the same reason: it
binds first in a reversible manner to the minor groove of dsDNA and, once
bound, alkylates the nucleobases (especially adenine) of the nucleic acid much
more rapidly than it reacts randomly with nucleophiles in solution. Presumably,
general acid catalysis at the binding site is responsible for the enhanced
reactivity (
20
). Unfortunately for therapeutic purposes, it lacks the sequence specificity in
this action to allow targeting to genes of interest. As a result, it is a
highly toxic agent (
17
,
24
). This cytotoxicity is attributed to a general interference of DNA replication
and transcription. In addition, the compound has a delayed hepatotoxicity
associated with the B and C subunits (
31
).
We have now, in effect, replaced the B and C subunits of CC-1065 with an ODN to create a new class of highly sequence-specific, hybridization-triggered DNA crosslinking agents. The advantageous properties
of the reactive CPI moiety were unchanged at neutral pH when it was conjugated
to the ODN. There was no loss of the CPI chromaphore in physiologic buffer.
Most importantly, the 3'-CPI-ODN conjugate alkylated the complementary strand in a duplex
very rapidly and efficiently. Within the minor groove of the hybrid, the
tethered CPI group binds with the same polarity and preference for A-T rich
regions as observed for CC-1065.
Based on the known binding mode of the (+)-enantiomer of CC-1065, one would predict that the 3'-(+)-CPI-ODN should exhibit optimal crosslinking
activity and the 5'-(+)-CPI-ODN should react with itself in the model system used
here. This expectation was borne out by the experimental results. The 3'-(+)-CPI-ODN reacted quantitatively within a few minutes based
on the UV assay and electrophoretic analysis. Hurley, Warpehoski and coworkers
(
24
) have shown that unconjugated (-)-CPI is 10-fold less active than (+)-CPI. They have also shown that, since these two
enantiomers of CPI alkylate the same sites, they must bind in the minor groove
in opposite directions. We expected, therefore, that the (-)-CPI-ODN conjugates should be unreactive with the opposite strand
when the (-)-enantiomer was linked to the 3'-end of the ODN, but modestly reactive when it was
linked to the 5'-end. The small amount of alkylation detected by electrophoresis
(see Fig.
5
) when using the 5'-CPI-ODN could be due to this (-)-enantiomer.
These properties suggest that (+)-CPI-ODN conjugates could have application as inhibitors of single-stranded viral DNA replication (e.g., hepatitis B virus) or
as gene selective inhibitors of transcription initiation (e.g., by binding to
an open promoter complex) (
32
). These same conjugates might also alkylate double-stranded DNA, either as part of a classical triple-stranded complex (
33
) or, as we have recently shown, in a recombinase-stabilized synaptic joint (
7
). Preliminary experiments, however, have shown that the CPI-ODN conjugates described here do not efficiently alkylate a complementary
RNA target and, therefore, may not be antisense candidates.
Other approaches to improve the crosslinkage efficiency of CPI-ODN conjugates with DNA are underway and include the evaluation of
conjugates which contain the optically pure (+)-enantiomer of CPI. At the same time we intend to verify the expectation
that the sequence context which supports alkylation is less restrictive for CPI-ODNs than for CC-1065.
In summary, CPI-ODNs would appear to be ideal sequence-specific hybridization-triggered crosslinking agents which have significant
potential for use as gene modification agents.
We wish to thank Dr Vladimir Gorn for oligonucleotide synthesis. A portion of
this work was funded by grant GM52774 from the National Institutes of Health,
USPHS.
REFERENCES
Return