Dnmt3aandDnmt1functionallycooperate during
de novo
methylation of DNA
Mehrnaz Fatemi*, Andrea Hermann*, Humaira Gowher and Albert Jeltsch
Institut fu
¨
r Biochemie, Justus-Liebig-Universita
¨
t, Gießen, Germany
Dnmt3a is a denovoDNA methyltransferase that modifies
unmethylated DNA. In contrast Dnmt1 shows high pre-
ference for hemimethylated DNA. However, Dnmt1 can be
activated for the methylationof unmodified DNA. We show
here that the Dnmt3aandDnmt1DNA methyltransferases
functionally cooperate in denovomethylationof DNA,
because a fivefold stimulation ofmethylation activity is
observed if both enzymes are present. Stimulation is obser-
ved if Dnmt3a is used before Dnmt1, but not if incubation
with Dnmt1 precedes Dnmt3a, demonstrating that methy-
lation of the DNA by Dnmt3a stimulates Dnmt1and that
no physical interaction ofDnmt1andDnmt3a is required. If
Dnmt1 andDnmt3a were incubated together a slightly increa-
sed stimulation is observed that could be due to a direct
interaction of these enzymes. In addition, we show that
Dnmt1 is stimulated for methylationof unmodified DNA if
the DNA already carries some methyl groups. We conclude
that after initiation ofdenovomethylationofDNA by
Dnmt3a, Dnmt1 becomes activated by the pre-existing
methyl groups and further methylates the DNA. Our data
suggest that Dnmt1 also has a role in denovomethylation of
DNA. This model agrees with the biochemical properties of
these enzymes and provides a mechanistic basis for the
functional cooperation of different DNA MTases in de novo
methylation ofDNA that has also been observed in vivo.
Keywords: DNA methylation; enzyme mechanism; DNA
methyltransferase; Dnmt1; Dnmt3a.
Cytosine residues are methylated at the 5-position for
70–80% of all CG sequences in mammalian DNA. The
pattern ofDNAmethylation serves as an epigenetic mark
in general leading to a repression of gene expression
[reviews 1,2,3]. It is used to memorize developmental
decisions of the cell and to control monoallelic expression
of genes during imprinting [review 4] and X-chromosome
inactivation [5]. Work with knockout mice has shown that
DNA methylation is an absolutely essential process in
mammals during late embryogenesis [6,7]. The methylation
pattern is created by denovomethylationand demethyla-
tion of the DNA, and maintained during mitosis by
maintenance methylation [reviews 8,9]. Denovo methyla-
tion ofDNA is most prevalent during embryogenesis,
where the methylation is restored after an almost complete
demethylation of the genome that takes place during the
first cleavage divisions [review 10]. In addition, de novo
methylation can also occur later in development and even
in adult cells to silence acquired proviral DNA or to alter
the developmental program of the cell. Aberrant de novo
methylation may lead to hypermethylation of promotor
regions of tumor suppressor genes in cancer cells and is an
important mechanism for cancer progression [11–13]. The
mechanism ofdenovomethylationofDNA is still poorly
understood and requires further investigation as this
process (together with specific demethylation events) cre-
ates the pattern ofDNAand therefore transfers the
epigenetic information to the DNA.
DNA methylation is introduced by DNA methyltrans-
ferases (MTases) which use S-adenosylmethionine as donor
for an activated methyl group [reviews 3,14,15,16]. Four
candidate DNA MTases have been identified in mammals
so far: Dnmt1, Dnmt2, Dnmt3aand Dnmt3b. Results
obtained with Dnmt1 knock-out mice have implicated this
enzyme in maintenance methylation [6], a role that is in
agreement to its pronounced preference for methylation of
hemimethylated DNA in vitro [17,18]. However, Dnmt1
also shows capabilities for denovomethylationof DNA
[15]. Interestingly, the denovo activity ofDnmt1 is
stimulated by binding of methylated DNA to an allosteric
site located in the N-terminal domain of the enzyme
[18,19]. However, denovomethylation activity is also
associated to the Dnmt3aand Dnmt3b enzymes, which
do not show a preference for methylationof hemimeth-
ylated CG sites [20–23]. It has been shown that one target
for the Dnmt3b enzyme are satellite sequences [7,24,25],
whereas specific targets for the Dnmt3a enzyme are not
yet known. In biochemical studies, it has been shown that
Dnmt3a methylates DNA in a distributive mechanism
[21]. This was a surprising observation, because it makes
the enzyme badly adapted for a fast methylationof one
domain of the DNA. However, the intriguing possibility
appeared that Dnmt3aandDnmt1 might functionally
cooperate duringdenovomethylationof DNA. This
model assumes that Dnmt3a might initiate de novo
methylation by transferring methyl groups to one region
of DNA. This would recruit and stimulate Dnmt1, which
could then methylate the whole domain of the DNA
[3,21].
Correspondence to A. Jeltsch, Institut fu
¨
r Biochemie, FB 8,
Justus-Liebig-Universita
¨
t, Heinrich-Buff-Ring, 58 35392 Gießen,
Germany. Fax: + 49 641 99 35409, Tel.: + 49 641 99 35410,
E-mail: Albert.Jeltsch@chemie.bio.uni-giessen.de
Abbreviations: Mtases, methyltransferases.
Note: *These authors contributed equally to the work.
(Received 21 June 2002, revised 8 August 2002,
accepted 21 August 2002)
Eur. J. Biochem. 269, 4981–4984 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.03198.x
EXPERIMENTAL PROCEDURES
Dnmt1 andDnmt3a were purified as described [18,21].
Methylation analyses were carried out with a 958mer and a
310mer PCR fragment which contain 95 and 27 CG sites,
respectively. Both substrates were amplified by PCR from
pAT153 plasmid. The 310mer had been selected to contain
three HpaII sites at one end. In either case, one of the PCR
primers carries a biotin group on its 5¢ end, such that the
PCR product is also biotinylated. Substrates were purified
using PCR Purification Kits (Qiagen). Concentrations were
determined from the absorbance at 260 nm. Methylation
kinetics were performed as described [26] using 3.7 l
M
[methyl-
3
H]-S-adenosylmethionine (5.55 · 10
14
BqÆmol
)1
,
Moravek Biochemicals, Brea, CA) in methylation buffer
(20 m
M
Hepes pH 7.5, 50 m
M
KCl, 50 lgÆmL
)1
BSA).
Premethylation of the 310mer PCR fragment (6 pmol) was
performed with 12 U M.HpaII (NEB) for 2 h at 37 °C. The
DNA was then precipitated and the concentration deter-
mined. To ensure complete methylationof all HpaII sites in
this experiment an aliquot of the purified methylated DNA
was subjected to a second round ofmethylation with
M.HpaII using radiolabeled AdoMet. No radioactivity was
incorporated into the DNA demonstrating that all HpaII
sites are fully methylated.
RESULTS AND DISCUSSION
It was the aim of this work to experimentally investigate a
model for denovomethylationofDNA that postulates
functional cooperation ofDnmt1and Dnmt3a. Purified
Dnmt1 andDnmt3a enzymes were employed to methylate a
958mer PCR product using labeled [methyl-
3
H]-S-adeno-
sylmethionine. We performed a series of three methylation
reactions, one with Dnmt1, the second with Dnmt3a and
the third with Dnmt1and Dnmt3a. As shown in Fig. 1, we
observed a much higher activity in the presence of both
enzymes, that is 4.7 ± 0.1-fold higher than the rate
corresponding to the sum of the rates observed with Dnmt1
and Dnmt3a alone. We conclude that a significant stimu-
lation ofdenovomethylation occurs if both enzymes are
present at the same time. Experiments with different
amounts ofDnmt1andDnmt3a showed that the level of
stimulation increases with increasing amounts of Dnmt3a
and slightly decreases with increasing amounts of Dnmt1
(data not shown). This result is reasonable, because
methylation by Dnmt3a is the initial event for de novo
methylation and due to its own denovo activity Dnmt1
might be able to stimulate itself in an autocatalytic fashion,
iftoomuchofthisenzymeisused.
We next examined if the stimulation requires a physical
interaction ofDnmt1and Dnmt3a, or if both enzymes only
interact via the methyl groups transferred to the DNA. To
this end, we have compared the methylation activities
observed when Dnmt3aandDnmt1 are present simulta-
neously in the reaction mixture with the activities obtained
when equal amounts of the two enzymes are added one after
the other (first Dnmt3a then Dnmt1) with intervening
ethanol precipitation of the DNA. In the first experiment a
physical interaction of both enzymes is possible, whereas
this is not possible in the second one. For control purposes,
we have also carried out reactions in which incubation of the
DNA with Dnmt1 was followed by incubation with
Dnmt3a or in which the DNA was incubated twice with
Dnmt3a. As shown in Fig. 2 incubation ofDnmt3a before
Dnmt1 but not Dnmt1 before Dnmt3a results in stimulation
of the DNAmethylation activity. This finding supports the
Fig. 1. Stimulation ofDNAmethylation by cooperation ofDnmt3a and
Dnmt1. A 958mer PCR product (20 n
M
) was methylated using labeled
[methyl-
3
H]-AdoMet by Dnmt1 (80 n
M
)(j), Dnmt3a (400 n
M
)(d)
and both enzymes present at the same time (r). All experiments were
carried out in triplicate. We observed a 4.7 ± 0.1-fold higher rate of
DNA methylation in the presence ofDnmt3aandDnmt1 than the sum
of the rates observed with either enzyme alone.
Fig. 2. Stimulation ofDNAmethylation observed with different mixing
protocols ofDnmt1and Dnmt3a. In the experiment labeled D3a/D3a,
methylation of 958mer (20 n
M
) was carried out with Dnmt3a (400 n
M
)
for 30 min. Then, the DNA was precipitated with ethanol and a sec-
ond methylation reaction was carried out with Dnmt3a (400 n
M
). In
the experiment labeled D1/D3a the first methylation reaction was
carried out with Dnmt1 (80 n
M
) and the second reaction with Dnmt3a
(400 n
M
). In the experiment labeled D3a/D1 the first methylation
reaction was performed with Dnmt3a (400 n
M
) and the second reac-
tion with Dnmt1 (80 n
M
). In the experiment labeled D1 + D3a, the
DNA was incubated with Dnmt1 (40 n
M
) andDnmt3a (200 n
M
)inthe
first reaction. As in the other experiments, the DNA was precipitated.
The second methylation reaction was performed with the same
amounts ofDnmt1and Dnmt3a. All experiments were carried out in
triplicate. The figure shows average values ofDNA methylation, the
error bars indicate the standard deviation of the individual data.
4982 M. Fatemi et al.(Eur. J. Biochem. 269) Ó FEBS 2002
model that Dnmt3a transfers some methyl groups to the
DNA, which in turn stimulates Dnmt1 for denovo methy-
lation of this region. In addition, it clearly demonstrates that
a physical interaction of both enzymes is not required for
stimulation to occur. We repeatedly observed a slightly
higher activity if both enzymes are present at the same time.
However, this effect is only small and close to the error
margins of our experiment. This difference could be due to a
direct interaction ofDnmt1and Dnmt3a, which has
recently been reported to occur [27].
It is known that Dnmt3a creates hemimethylated target
sites duringDNAmethylation [28]. As the activity of
Dnmt1 on hemimethylated DNA is much higher [18],
generation of one hemimethylated site will immediately lead
to a second methylation in the unmodified strand of this site
in the presence of Dnmt1. Kinetically this reaction scheme
means that two methyl groups are always introduced, one
slowly by Dnmt3a or by denovomethylation catalyzed by
Dnmt1 and the second one very fast. Therefore, this effect
can only double the rate ofDNA methylation, because
every turnover by Dnmt3a can trigger only one turnover by
Dnmt1 in the opposite strand of the DNA. Consequently,
this effect cannot account for the fivefold stimulation
observed in our experiments and additional stimulation of
Dnmt1 by pre-existing methylation must occur. To show
this effect we used a 310mer PCR fragment that contains
three fully methylated CG sites produced by methylation
with M.HpaII. As shown in Fig. 3 the premethylated DNA
is modified significantly faster by Dnmt1 confirming similar
results by other groups [29,30]. Furthermore, it is in
agreement with the allosteric activation ofDnmt1 observed
after binding of the enzyme to methylated oligonucleotides
[18,19,31,32]. A similar effect is not observed with Dnmt3a
showing that at least under these conditions no allosteric
activation occurs in Dnmt3a, although this enzyme like
Dnmt1 contains at least one additional DNA binding site in
its N-terminal part [33]. The observation that Dnmt3a is not
activated by pre-exiting methylation agrees with the fact
that the rates ofmethylationof hemimethylated substrates
by this enzyme are not higher than rates observed with
unmethylated DNA.
The results presented in this paper complement recent
in vivo data also demonstrating a functional cooperation of
Dnmt1 with Dnmt3aand Dnmt3b in maintenance as well as
in denovomethylationofDNA [34]. In this work it was
shown using Dnmt1, Dnmt3aand Dnmt3b knock-out cell
lines that Dnmt1 alone is not able to maintain methylation
levels at certain repeat sequences, but that the presence of
Dnmt3a and or Dnmt3b in addition to Dnmt1 is required
for accurate maintenance methylation in these regions. On
the other hand, Dnmt3aand Dnmt3b are not capable of
efficient denovomethylationofDNA in the absence of
Dnmt1. Whereas the maintenance methylation is not an
issue of our work, we present a mechanistic basis for the
functional cooperation ofDnmt3aandDnmt1 in de novo
methylation ofDNA that can only be provided by in vitro
experiments. In this model Dnmt3a is targeted to a domain
of the DNA that is subject to denovo methylation. As this
enzyme does not work processively, it can introduce only a
few methyl groups to the DNA. These attract Dnmt1 to the
DNA, and activate it for denovo methylation. Then, Dnmt1
spreads the methylation over the whole domain of the
DNA. Following this mechanism only a few targeting events
are required to achieve sufficient methylationof one DNA
domain whereas without the cooperation with Dnmt1,
Dnmt3awouldhavetorebindtotheDNAaftereach
turnover to achieve complete methylationof one domain of
the DNA. It should be noticed that the molecular mechan-
ism of these targeting events is still unknown, although the
interaction ofDnmt3a with some other proteins has been
shown that might be the mechanistic basis for the process
[35]. Finally, it is possible that maintenance methylation at
sequences that are not efficiently methylated by Dnmt1
alone also follows this mechanism.
Our data contribute to the view that DNAmethylation is
a complicated process, in which different enzymes are
involved: We show here that Dnmt1andDnmt3a can
cooperate in denovomethylation in vitro, a similar
phenomenon has also been observed in vivo [34]. These
authors also have shown that the Dnmt3aand Dnmt3b
enzymes (in addition to Dnmt1) are required for mainten-
ance of the methylation at certain DNA sequences. A role of
Fig. 3. Stimulation ofDnmt1andDnmt3a by fully methylated CpG sites. In this experiment the rates ofmethylationof the 310mer substrate were
compared after premethylation of the substrate on 3 HpaII sites (CCGG) located at the 5¢ end of the molecule (j) and without premethylation (d).
In the left panel 40 n
M
DNA were methylated using 120 n
M
Dnmt1, in the right panel using 400 n
M
Dnmt3a. In repeated experiments a 2.9 ± 0.2-
fold stimulation was observed with Dnmt1, whereas no significant difference was observed with Dnmt3a.
Ó FEBS 2002 Cooperation ofDnmt1andDnmt3a (Eur. J. Biochem. 269) 4983
Dnmt3a and Dnmt3b in maintenance methylation also has
been concluded from studies with cell lines which do not
express functional Dnmt1 [36] and, finally, a functional
interdependence of the Dnmt3aand Dnmt3b enzymes has
been demonstrated in mouse knock out studies [20]. Given
this complicated situation it might be recommendable to
reserve the usage of the terms Ôde novoÕ and ÔmaintenanceÕ for
the description of processes occurring on the DNAand not
to associate them with certain DNA MTases.
ACKNOWLEDGEMENTS
Thanks are due to K. Liebert for performing some of the experiments
and to A. Pingoud for discussions and support. This work was
supported by the Deutsche Forschungsgemeinschaft (JE 252/1–2, 1–3).
REFERENCES
1. Robertson, K.D. & Wolffe, A.P. (2000) DNAmethylation in
health and disease. Nat. Reviews Genetics 1, 11–19.
2. Jones, P.A. & Takai, D. (2001) The role ofDNAmethylation in
mammalian epigenetics. Science 293, 1068–1070.
3. Jeltsch, A. (2002) Beyond Watson and Crick: DNA methylation
and molecular enzymology ofDNA methyltransferases. Chem-
biochem,inpress.
4. Ferguson-Smith, A.C. & Surani, M.A. (2001) Imprinting and the
epigenetic asymmetry between parental genomes. Science 293,
1086–1089.
5. Chao,W.,Huynh,K.D.,Spencer,R.J.,Davidow,L.S.&Lee,J.T.
(2002) CTCF, a candidate trans-acting factor for X-inactivation
choice. Science 295, 345–347.
6. Li,E.,Bestor,T.H.&Jaenisch,R.(1992)Targetedmutationof
the DNA methyltransferase gene results in embryonic lethality.
Cell 69, 915–926.
7. Okano, M., Bell, D.W., Haber, D.A. & Li, E. (1999) DNA
methyltransferases Dnmt3aand Dnmt3b are essential for de novo
methylation and mammalian development. Cell 99, 247–257.
8. Hsieh, C.L. (2000) Dynamics ofDNAmethylation pattern. Curr.
Opin. Genet. Dev. 10, 224–228.
9. Wolffe, A.P., Jones, P.L. & Wade, P.A. (1999) DNA demethyla-
tion. Proc. Natl Acad. Sci. USA 96, 5894–5896.
10. Reik, W., Dean, W. & Walter, J. (2001) Epigenetic reprogram-
ming in mammalian development. Science 293, 1089–1093.
11. Jones, P.A. & Vogt, P.K. (2000) DNAMethylationand Cancer.
Springer Verlag, Heidelberg.
12. Baylin, S.B. & Herman, J.G. (2000) DNA hypermethylation in
tumorigenesis: epigenetics joins genetics. Trends Genet. 16,168–
174.
13. Warnecke, P.M. & Bestor, T.H. (2000) Cytosine methylation and
human cancer. Curr. Opin. Oncol. 12, 68–73.
14. Cheng, X. (1995) Structure and function ofDNA methyltrans-
ferases. Annu. Rev. Biophys. Biomol. Struct. 24, 293–318.
15. Bestor, T.H. (2000) The DNA methyltransferases of mammals.
Hum. Mol. Genet. 9, 2395–2402.
16. Cheng, X. & Roberts, R.J. (2001) AdoMet-dependent methyla-
tion, DNA methyltransferases and base flipping. Nucl. Acids Res.
29, 3784–3795.
17. Pradhan, S., Bacolla, A., Wells, R.D. & Roberts, R.J. (1999)
Recombinant human DNA (cytosine-5) methyltransferase. J. Biol.
Chem. 274, 33002–33010.
18. Fatemi, M., Hermann, A., Pradhan, S. & Jeltsch, A. (2001) The
activity of the murine DNA methyltransferase Dnmt1 is
controlled by interaction of the catalytic domain with the
N-terminal part of the enzyme leading to an allosteric activation of
the enzyme after binding to methylated DNA. J. Mol. Biol. 309,
1189–1199.
19. Bacolla, A., Pradhan, S., Larson, J.E., Roberts, R.J. & Wells,
R.D. (2001) Recombinant human DNA (cytosine-5) methyl-
transferase. III. Allosteric control, reaction order, and influence of
plasmid topology and triplet repeat length on methylation of
the fragile X CGG.CCG sequence. J. Biol. Chem. 276, 18605–
18613.
20. Okano,M.,Xie,S.&Li,E.(1998)Cloningandcharacterization
of a family of novel mammalian DNA (cytosine-5) methyltrans-
ferases. Nature Genet. 19, 219–220.
21. Gowher, H. & Jeltsch, A. (2001) Enzymatic properties of
recombinant Dnmt3aDNA methyltransferase from mouse: the
enzyme modifies DNA in a non-processive manner and also
methylates non-CpG sites. J. Mol. Biol. 309, 1201–1208.
22. Aoki, A., Suetake, I., Miyagawa, J., Fujio, T., Chijiwa, T., Sasaki,
H. & Tajima, S. (2001) Enzymatic properties ofde novo-type
mouse DNA (cytosine-5) methyltransferases. Nucl. Acids Res. 29,
3506–3512.
23. Hsieh, C L. (1999) In vivo activity of murine denovo methyl-
transferases, Dnmt3aand Dnmt3b. Mol. Cell. Biol. 19, 8211–8218.
24. Hansen, R.S., Wijmenga, C., Luo, P., Stanek, A.M., Canfield,
T.K., Weemaes, C.M. & Gartler, S.M. (1999) The DNMT3B
DNA methyltransferase gene is mutated in the ICF immuno-
deficiency syndrome. Proc.NatlAcad.Sci.USA96, 14412–14417.
25. Xu, G L., Bestor, T.H., Bourc’his, D., Hsieh, C L., Tommerup,
N., Bugge, M., Hulten, M., Qu, X., Russo, J.J. & Viegas-
Pequignot, E. (1999) Chromosome instability and immuno-
deficiency syndrome caused by mutations in a DNA
methyltransferase gene. Nature 402, 187–191.
26. Roth, M. & Jeltsch, A. (2000) Biotin-Avidin microplate assay for
the quantitative analysis of enzymatic methylationofDNA by
DNA methyltransferases. Biol. Chem. 381, 269–272.
27. Kim, G.D., Ni, J., Kelesoglu, N., Roberts, R.J. & Pradhan, S.
(2002) Co-operation and communication between the human
maintenance anddenovoDNA (cytosine-5) methyltransferases.
EMBO J. 21, 4183–4195.
28. Lin, I.G., Han, L., Taghva, A., O’Brien, L.E. & Hsieh, C.L. (2002)
Murine denovo methyltransferase dnmt3a demonstrates strand
asymmetry and site preference in the methylationofDNA in vitro.
Mol. Cell Biol. 22, 704–723.
29. Tollefsbol, T.O. & Hutchison, C.A.I. (1995) Mammalian DNA
(cytosine-5-) methyltransferase expressed in Escherichia coli,pur-
ified and characterized. J. Biol. Chem. 270, 18543–18550.
30. Carotti, D., Funiciello, S., Palitti, F. & Strom, R. (1998) Influence
of pre-existing methylation on the denovo activity of eukaryotic
DNA methyltransferase. Biochemistry 37, 1101–1108.
31.Bacolla,A.,Pradhan,S.,Roberts,R.J.&Wells,R.D.(1999)
Recombinant human DNA (cytosine-5) methyltransferase. II.
Steady-state kinetics reveal allosteric activation by methylated
DNA. J. Biol. Chem. 274, 33011–33019.
32. Flynn, J. & Reich, N. (1998) Murine DNA (cytosine-5-) methyl-
transferase: Steady-state and substrate trapping analyses of the
kinetic mechanism. Biochemistry 37, 15162–15169.
33. Qiu, C., Sawada, K., Zhang, X. & Cheng, X. (2002) The PWWP
domain of mammalian DNA methyltransferase Dnmt3b defines a
new family of DNA-binding fold. Nat. Struct. Biol. 9, 217–224.
34. Liang, G., Chan, M.F., Tomigahara, Y., Tsai, Y.C., Gonzales,
F.A., Li, E., Laird, P.W. & Jones, P.A. (2002) Cooperativity
between DNA methyltransferases in the maintenance methylation
of repetitive elements. Mol. Cell Biol. 22, 480–491.
35. Bachman, K.E., Rountree, M.R. & Baylin, S.B. (2001) Dnmt3a
and Dnmt3b are transcriptional repressors that exhibit unique
localization properties to heterochromatin. J. Biol. Chem. 276,
32282–32287.
36. Rhee, I., Jair, K.W., Yen, R.W., Lengauer, C., Herman, J.G.,
Kinzler, K.W., Vogelstein, B., Baylin, S.B. & Schuebel, K.E.
(2000) CpG methylation is maintained in human cancer cells
lacking DNMT1. Nature 404, 1003–1007.
4984 M. Fatemi et al.(Eur. J. Biochem. 269) Ó FEBS 2002
. for the methylation of unmodified DNA. We show
here that the Dnmt3a and Dnmt1 DNA methyltransferases
functionally cooperate in de novo methylation of DNA,
because. late embryogenesis [6,7]. The methylation
pattern is created by de novo methylation and demethyla-
tion of the DNA, and maintained during mitosis by
maintenance