ABSTRACT
The target cytosines of (cytosine-5)-DNA methyltransferases in prokaryotic and eukaryotic DNA show
increased rates of C
->
T transition mutations compared to non-target cytosines. These mutations are induced either by the spontaneous
deamination of 5-mC
->
T generating inefficiently repaired G:T rather than G:U mismatches, or by the
enzyme-induced C
->
U deamination which occurs under conditions of reduced levels of
S
-adenosylmethionine (AdoMet) and
S
-adenosylhomocysteine (AdoHcy). We tested whether various inhibitors of (cytosine-5)-DNA methyltransferases analogous to AdoMet and AdoHcy would
affect the rate of enzyme-induced deamination of the target cytosine by M.
Hpa
II and M.
Sss
I. Interestingly, we found two compounds, sinefungin and 5
'
-amino-5'
-deoxyadenosine, that increased the rate of deamination 10
3
-fold in the presence and 10
4
-fold in the absence of AdoMet and AdoHcy. We have therefore identified the
first mutagenic compounds specific for the target sites of (cytosine-5)-DNA methyltransferases. A number of analogs of AdoMet and AdoHcy
have been considered as possible antiviral, anticancer, antifungal and
antiparasitic agents. Our findings show that chemotherapeutic agents with affinities to the cofactor binding pocket of (cytosine-5)-DNA methyltransferase should be tested for their potential mutagenic
effects.
S
-Adenosylmethionine (AdoMet) is a ubiquitous methyl-donor for a wide variety of methyl-transfer reactions (
1
). Furthermore, AdoMet is the substrate required for the generation of
decarboxylated AdoMet (deca-AdoMet) by AdoMet decarboxylase, which is the rate limiting step in the
synthesis of spermine and spermidine (
2
). The importance of methyl-transfer and decarboxylation reactions has made them potential targets for
chemotherapeutic intervention by competitive inhibition. A wide variety of analogs of AdoMet or
its metabolites have been evaluated as potential inhibitors of the methyl-transfer or decarboxylation reactions (
1
-
7
). Other compounds interfere with the synthesis or degradation of AdoMet and
influence the level of AdoMet, deca-AdoMet or its metabolites (
8
). Interference with the metabolism of AdoMet could have a variety of effects on
the metabolism of a cell or could change the genomic DNA methylation pattern.
DNA methylation, which is known to function as a stable gene silencing mechanism
(
9
-
14
), is essential for normal embryonic development since mouse embryos with the
(cytosine-5)-DNA methyltransferase gene homozygously deleted are inviable (
15
). A methyl-donor deficient diet lowers the genomic DNA methylation level and increases the risk for liver and colon tumors in rats (
16
-
18
). The involvement of DNA methylation in the initiation or promotion of cancer
has been well studied and altered DNA methylation patterns are found in many
tumors (
9
-
14
). Changes in DNA methylation patterns could either induce the expression of
oncogenes or silence the expression of tumor suppressor genes (
9
-
14
). Growth regulatory genes such as the p16 gene are inactivated by DNA
methylation thus allowing continuous cell division required for tumorigenesis (
13
). Epigenetic changes of DNA methylation patterns are heritable and, since they
occur with higher frequency than genetic changes, represent possible targets for altering the gene expression pattern in cancer cells (
19
).
Furthermore, since the target cytosine of the (cytosine-5)-DNA methyltransferase is mutated with a high frequency in bacteria
and in human cancer and genetic disease, the methylation reaction confers not
only the possibility of epigenetic changes but could also increase the rate of genetic changes (
20
-
24
). A high frequency of C -> T transition mutations at CpG dinucleotides occurs in the p53 tumor
suppressor gene in various tumor types and inherited diseases (Li-Fraumeni syndrome) (
25
). These mutations are assumed to arise either by the spontaneous or enzyme-mediated deamination of 5-mC -> T in the absence of completely efficient repair systems, or by
the (cytosine-5)-DNA methyltransferase induced deamination of C -> U -> T occurring at cofactor concentrations below the K
m
values of the enzyme for its cofactor AdoMet or its reaction product AdoHcy (
26
-
31
).
We have previously shown that 5-methylcytosine (5-mC) spontaneously deaminates at a rate 2-3-fold higher than C in double stranded DNA in a genetic
reversion assay (
32
). Furthermore, in the absence of methyl donor AdoMet or reaction product AdoHcy, the rate of enzyme-mediated C -> U -> T transition mutations is increased up to 10
4
-fold when compared with the spontaneous rate in water (
27
). Mutations introduced into the AdoMet binding pocket of the M.
Hpa
II (cytosine-5)-DNA methyltransferase that interfere with AdoMet binding and therefore mimic the absence of AdoMet also increase the deamination rate,
even in the presence of cofactors (
33
).
We have now tested a group of compounds analogous to AdoMet or AdoHcy for their
abilities to interfere with AdoMet in the methylation reaction of the bacterial
(cytosine-5)-DNA methyltransferases M.
Hpa
II and M.
Sss
I. Furthermore, the effects of these compounds on DNA binding affinities of M.
Hpa
II and M.
Sss
I and their abilities to increase the rate of C -> T transition mutations at the target sites of these methyltransferases in
the presence and absence of AdoMet and AdoHcy were measured.
The
Escherichia coli
strain ER2357 [
end
A1
thi
-1
sup
E44
mcr
-67
ung
-1
dut
[Delta](
arg
F-
lac
)U169 [Delta](
mcr
C-
mrr
)114::IS10
rec
A1 F'
pro
AB
lac
I
q
Z[Delta]M15
zzf
::Tn10(Tet
r
)] was kindly obtained from Dr Sha Mi (Cold Spring Harbor Laboratory).
S
-Adenosylmethionine (AdoMet), M.
Hpa
II and M.
Sss
I DNA methyltransferases were purchased from New England Biolabs.
S
-Adenosylhomocysteine (AdoHcy), 5'-methylthio-5'-deoxyadenosine (MTA) and 5'-amino-5'-deoxyadenosine were purchased from
Sigma. The syntheses of sinefungin, (
S
)-6-methyl-6-deamino- sinefungin and 6-deaminosinefungin have been described
previously (
34
).
N
4
-Adenosyl-
N
4
-methyl-2,4-diaminobutanoic acid was kindly provided by Dr A. Gall
(Oridigm Corporation, Seattle) (
35
). All analogs were dissolved in water and concentrations confirmed spectrophotometrically (OD
260
). The TMP-5-fluorouracil phosphoramidite was purchased from Glen Research
(Sterling, VA).
This assay is based on the covalent trapping of (cytosine-5)-DNA methyltransferase by the mechanism-based inhibitor 5-fluorocytosine present at the target C (
36
). The synthesis and labeling of the oligonucleotide used for covalent
interaction has been described previously (
33
). The labeled oligonucleotide (20 ng) was incubated in 1* M.
Hpa
II or 1* M.
Sss
I buffer (New England Biolabs) and DNA methyltransferase (4 U M.
Hpa
II; 2 U M.
Sss
I) with 10 [mu]M AdoMet and the various diluted cofactor analogs for 90 min at 37oC. Thereafter, SDS loading buffer was added and the samples were boiled
for 5 min and loaded on a 10% SDS- PAGE according to Laemmli (
37
). The bands representing bound and free oligonucleotides were quantitated using
a phosphorimager (Molecular Dynamics) and the ratio of (bound)/(bound + free)
was calculated. Experiments were performed twice with similar results.
The labeling method of the oligonucleotides and the gel mobility shift assay have been described previously (
38
). A new non-specific competitor oligonucleotide was used that contains no CpG dinucleotide.
Non-specific oligonucleotide:
Top strand:
5'-GGGCTCATAGGGCACCACCACACTATGT-3'
Bottom strand:
3'-CCCGAGTATCCCGTGGTGGTGTGATACA-5'
Briefly, the DNA binding reaction was done by incubating DNA methyltransferase
(4 U M.
Hpa
II; 2 U M.
Sss
I), 4 pmol of labeled oligonucleotides, 40 pmol of non-specific oligonucleotide in a 10 [mu]l buffered reaction (50 mM Tris-HCl pH 7.5, 10 mM EDTA, 13% glycerol, 0.5 mM DTT and 0.2 [mu]g/[mu]l BSA) and 100 [mu]M of the AdoMet analogs for 30 min at room temperature.
After incubation the samples were electrophoresed on a 6% non-denaturing polyacrylamide gel for 2 h at 80 V. The bands representing
bound and free oligonucleotides were quantitated using a phosphorimager and the ratio of (bound)/(bound + free) was calculated. The results shown
represent the means of two experiments.
The
in vitro
reversion assay was carried out as described previously (
27
,
33
). Briefly, 200 ng of the reporter plasmid C
C
GG-pSV2-neo
s
was incubated with the DNA methyltransferase (4 U M.
Hpa
II; 2 U M.
Sss
I) in 20 [mu]l reaction buffer (for M.
Hpa
II: 50 mM Tris-HCl pH 7.5, 10 mM EDTA, 1 mM DTT, 0.2 [mu]g/[mu]l BSA; for M.
Sss
I: 50 mM Tris-HCl pH 7.5, 1 mM EDTA, 5 mM DTT, 50 mM NaCl, 0.2 [mu]g/[mu]l BSA, which is a buffer allowing processive methylation by M.
Sss
I, since it contains no MgCl
2
;
39
) for 16 h at 37oC.
S
-Adenosylmethionine or
S
-adenosylhomocysteine was added at 10 [mu]M and the AdoMet analogs were serial diluted to the concentrations
indicated in the figures. The plasmid DNA was extracted with phenol and
chloroform and precipitated with ethanol using glycogen as a carrier. The DNA
was transformed using an electroporator (Biorad) into electro-competent
E.coli
strain ER2357 which is deficient in restriction of 5-mC (
mcr
ABC
-
) and in uracil glycosylase (
ung
-
). A part of the bacteria was diluted and plated on ampicillin plates to score
for the C
C
GG-pSV2-neo
s
transformation efficiency and the remaining bacteria plated on kanamycin plates to score for C
T
GG-pSV2-neo
r
revertants. The reversion frequency was calculated as (number of kanamycin
resistant bacteria)/(number of ampicillin resistant bacteria). Experiments were performed at least twice with similar results.
We first studied the abilities of the different cofactor analogs to bind to the
AdoMet binding pocket and to interfere with AdoMet binding and the methylation
reaction. A sensitive assay was used which is based on the covalent trapping of
the (cytosine-5)-DNA methyltransferase to an oligonucleotide containing 5-fluorocytosine at the target cytosine (
33
,
36
). Methyl-transfer from AdoMet is necessary for covalent adduct formation and
interference by AdoMet analogs prevents transfer thus giving a direct measure
of inhibition of the methylation reaction. Therefore, in contrast to an assay
only measuring competition with AdoMet for binding to the free enzyme, this
assay measures the inhibition of the enzymatic reaction composed of competition
with AdoMet, DNA binding and the methyl-transfer reaction. The assay was performed with the enzymes M.
Hpa
II or M.
Sss
I, that were incubated in the presence of AdoMet (10 [mu]M) and various concentrations of cofactor analogs. In the case of M.
Hpa
II the cofactor analogs with a complete methionine backbone (analogs
1
,
2
and
3
) inhibited the methylation reaction efficiently whereas the aminoanalog of
AdoMet (
4
) and the cofactor analogs with a truncated methionine backbone (analogs
5
and
6
) competed less efficiently with the methylation reaction (Fig.
2
). In the case of M.
Sss
I, inhibition of methylation by the analogs
1
,
2
and
3
was strong, whereas the analogs
4
and
5
showed intermediate inhibition and the analog
6
weak inhibition (Fig.
2
). For both enzymes AdoHcy competed slightly stronger than analog
1
, in contrast to previously published findings with M.
Hha
I and M.
Eco
RI, in which the opposite result was found (
45
,
46
). In summary, the analogs tested can be roughly grouped into strong competitors
(analogs
1
,
2
,
3
and AdoHcy) all containing the complete methionine backbone of AdoHcy, intermediate competitors (analogs
4
,
5
and
6
) and a weak competitor (analog
6
for M.
Sss
I). The replacement of the carbon of analog
2
with the nitrogen in analog
4
rendered it a weaker competitor. Since analog
5
(MTA) inhibited DNA methylation by M.
Hpa
II and M.
Sss
I less strongly than AdoHcy, it is unlikely that it plays a major role in
regulating DNA methylation in cancer cells, although such a role cannot be
excluded completely (
47
,
48
).
It has previously been shown that the binding of bacterial (cytosine-5)-DNA methyltransferases can be enhanced by either adding AdoHcy or by
introducing a G:U mismatch at the target cytosine (
38
,
49
-
51
). It seemed possible that the cofactor analogs used here not only inhibited the
enzymatic reaction by preventing the binding of AdoMet, but also increased or
decreased the affinities of the enzymes to their target sequences. Therefore,
the abilities of the various cofactor analogs to change the affinities of the
enzymes M.
Hpa
II and M.
Sss
I to their target sequences was measured using gel mobility shift assays. The
binding of M.
Hpa
II was increased 3-fold by AdoHcy and ~2-fold by analog
2
, but no stimulatory or inhibitory effects with analogs
1
,
3
,
4
,
5
and
6
were observed (Fig.
3
A). Introduction of a uracil instead of cytosine at the target site, thus
generating a G:U mismatch, overrode the stimulatory effect of the cofactor
(Fig.
3
B).
The results above show that the cofactor analogs all compete with different efficiencies with the methylation reaction and, depending on the analog and the enzyme, have stimulatory effects on DNA binding. We next
investigated whether any of these structural and binding characteristics
specific for each cofactor analog would influence the deamination rate, which
so far has only been described to occur in the presence of decreased
concentrations of AdoMet or AdoHcy (
27
-
30
). The plasmid C
C
GG-pSV2-neo
s
was incubated with M.
Hpa
II or M.
Sss
I in the presence of 500 [mu]M of each cofactor analog and 10 [mu]M AdoHcy, a concentration that completely prevents the enzyme-induced deamination (
27
). The reversion frequency was measured by scoring the number of kanamycin
resistant colonies harboring the reverted plasmid C
T
GG-pSV2-neo
r
. The presence of cofactor analog
1
led to a 10
2
-10
3
-fold increase of deamination to a reversion frequency similar to that in
the absence of added AdoHcy (Fig.
4
). Cofactor analog
6
, which has been found to be a weak competitor with AdoMet (Fig.
2
) also increased the reversion frequency, however, less efficiently than analog
1
. Analogs
2
,
3
,
4
and
5
as well as AdoHcy did not lead to any increase of cytosine deamination. These
analogs at 100 [mu]M suppressed deamination when no AdoHcy was added (data not shown), except
analog
5
(MTA), which only partially inhibited deamination probably since its affinity
to the AdoMet binding pocket is low (Fig.
2
).
Several steps involved in the enzymatic methylation reaction of cytosine in DNA
have been recently clarified (
52
-
56
). Each (cytosine-5)-DNA methyltransferase has unique properties such as sequence
specificity and the requirement of AdoMet or AdoHcy for efficient DNA binding,
but all share a common catalytic methyl-transfer mechanism (
57
). The target cytosine is flipped out from the double-helix after binding to the specific target sequence and positioned into
the catalytic pocket of the enzyme (
40
). Nucleophilic attack at the C6 of cytosine by the thiolate of a conserved
cysteine-residue of the enzyme activates the otherwise inert C5 of cytosine. The
activated C5 attacks the electrophilic methyl group of AdoMet which is
subsequently transferred to the cytosine moiety. Abstraction of a proton at C5
and dissociation of the covalent cysteine-C6 adduct completes the methyl-transfer reaction by [beta]-elimination, thus forming 5-methylcytosine and AdoHcy. Since the methyltransferase-AdoHcy complex has weak affinity for the
methylated target sequence, the complex rapidly dissociates.
A caveat to this reaction is the formation of the 5,6-dihydrocytidine intermediate in the absence of AdoMet or AdoHcy and the presence of solvent water that deaminates with increased frequency (
58
). In view of the high rate of transition mutations at CpGs in human cancer, it is unknown whether physiological or pathological conditions exist where the concentration of AdoMet as well as of AdoHcy is below
the
K
m
value of the enzyme for its cofactors necessary for the deamination reaction.
Considering the need to stably maintain the DNA sequence of an organism, the
(cytosine-5)-DNA methyltransferases presumably have evolved in a way that the presence of substrate AdoMet and the reaction product
AdoHcy both prevent the deamination reaction.
Our study was intended to gain further insight into the mechanism of cytosine deamination and into how such deaminations are prevented under normal physiological conditions. The affinity of the enzyme
to the target sequence appears to play no significant role in the cofactor-mediated deamination since it occurs even in the absence of any cofactor (
27
) and since both cofactor analogs
1
and
6
that increase the rate of deamination do not change the binding affinities.
Among the analogs suppressing the deamination (
2
,
3
,
4
,
5
and AdoHcy), only
2
,
3
and AdoHcy increased the DNA binding, suggesting that the affinity of the
enzyme to the target sequence does not play a role in preventing the
deamination. Therefore, similar to the methylation reaction (
59
), the deamination reaction can be separated from the conformational changes
required for target DNA binding.
We thank Dr A. Gall (Oridigm Corporation, Seattle) for the amino-analog of AdoMet. This work was supported by grant R35 CA49758.
Present addresses:
+
Department of Pathology, Box 357705, University of Washington, Seattle, WA 98195-7705, USA and
[sect]
University of California, Berkeley, College of Chemistry, Berkeley, CA 94720,
USA
REFERENCES
Return