Hydroxylation of methylated CpG dinucleotides reverses stabilisation of DNA duplexes by cytosine 5-methylation

Article abstract of DOI:10.1039/c0cc05671e

Cytosine-5-methylation stabilises DNA duplexes and is associated with transcriptional repression; 5-methylcytosine undergoes hydroxylation to 5-hydroxymethylcytosine, a modification of unknown biological function. Spectroscopic and calorimetric analyses show that 5-hydroxymethylcytosine introduction reverses the stabilising effect of 5-methylcytosine, suggesting that in some contexts, 5-methylcytosine hydroxylation may, along with other factors, contribute to the alleviation of transcriptional repression.

Full text of DOI:10.1039/c0cc05671e

View Article Online / Journal Homepage / Table of Contents for this issue
Dynamic Article Links
ChemComm
Cite this: Chem. Commun., 2011, 47, 5325–5327
www.rsc.org/chemcomm
COMMUNICATION
Hydroxylation of methylated CpG dinucleotides reverses stabilisation of
DNA duplexes by cytosine 5-methylationw
a
Armin Thalhammer,a Anders S. Hansen,z Afaf H. El-Sagheer,bc Tom Brownb and
Christopher J. Schofield*a
Received 20th December 2010, Accepted 16th March 2011
DOI: 10.1039/c0cc05671e
Cytosine-5-methylation stabilises DNA duplexes and is associated
with transcriptional repression; 5-methylcytosine undergoes hydro-
xylation to 5-hydroxymethylcytosine, a modification of unknown
biological function. Spectroscopic and calorimetric analyses show
that 5-hydroxymethylcytosine introduction reverses the stabilising
effect of 5-methylcytosine, suggesting that in some contexts,
5-methylcytosine hydroxylation may, along with other factors,
contribute to the alleviation of transcriptional repression.
We considered the possibility that 5mC-hydroxylation might
alter the chemical stability of DNA. In addition to the stabilising
thermodynamic effects of 5mC on DNA,10,11 precedent for this
possibility comes from observations on protein-hydroxylation,
i.e. that the collagen triple helix12 and ankyrin repeat domain
folds13 are stabilised by 2OG oxygenase-catalysed prolyl-4- and
asparaginyl-3-hydroxylations, respectively.
To investigate the effect of 5hmC, we prepared b-cyanoethyl-
protected 5hmC phosphoramidite using a procedure based on a
conversion of 20-deoxyuridine to 20-deoxycytidine in the absence
of temporary protecting groups (Fig. S5w).14 Oligodeoxyribo-
nucleotides (ODNs) were prepared by phosphoramidite
synthesis, purified and analysed by LC-MS (Fig. S1, ESIw) and
electrophoresis (Fig. S2w).
We initially performed UV melting (UVM) analyses on 2
DNA duplex forming sequences (1 and 2, Fig. 2) containing all
combinations of C/5hmC/5mC on forward and complementary
strands (Table S1w). Selection criteria included: (i) the presence
of a CpG dinucleotide, (ii) an appropriate length both to form
duplexes and investigate effects of modifications, (iii) an
absence of predicted secondary structures (e.g. hairpin loops),
and (iv) an even distribution of AT- and GC-rich regions to
enable single-stage duplex melting transitions.
Cytosine C-5 methylation of CpG dinucleotides (Fig. 1a) is a
DNA modification associated with inhibition of transcription
(Fig. 1b) and epigenetic gene silencing.1 The presence of
5-methylcytosine (5mC) can directly inhibit binding of the
transcriptional machinery (e.g. by blocking transcription
factor binding) or recruit chromatin-remodelling complexes
for histone-mediated gene silencing. Aberrant DNA methylation
is associated with cancer.2,3 Whereas direct enzymatic reversal
is well established for some protein and DNA modifications
including histone (de)acetylation and (de)methylation and
DNA N-demethylation, there have been no reports of a
single-step reversion of 5mC to cytosine.
Recently, it was reported that 5-hydroxymethylcytosine
(5hmC, Fig. 1a) occurs in human Purkinje neuron DNA.4
5hmC comprises up to B20% of total 5mC levels in some
regions, and its formation is catalysed by TET1 in human cell
lines.5 TET1 and its homologues, TET26 (implicated in myeloid
malignancies7,8) and TET3,6 are 2-oxoglutarate (2OG) and FeII-
dependent oxygenases, other examples of which have important
roles in transcriptional regulation by transcription factor hydro-
xylation and histone demethylation.9 5hmC is a chemically stable
DNA modification but its biological function is unknown.
In accord with prior work,10,11 our results (Fig. 2) show that
duplexes with CpG dinucleotides methylated on both strands
(mC/mC) have higher duplex melting temperatures (Tm) than
a Department of Chemistry and the Oxford Centre for Integrative
Systems Biology, Chemistry Research Laboratory, Oxford, UK.
E-mail: Christopher.schofield@chem.ox.ac.uk;
Fax: +44 (0)1865 285002; Tel: +44 (0)1865 275625
b School of Chemistry, University of Southampton, Southampton, UK
c Chemistry Branch, Department of Science and Mathematics,
Faculty of Petroleum and Mining Engineering, Suez Canal
University, Suez, Egypt
w Electronic supplementary information (ESI) available: Sequences,
synthesis, purification and analysis (LC-HRMS, CE, CD) of ODNs,
biophysical data (UVM/ITC), and detailed procedures. See DOI:
10.1039/c0cc05671e
Fig. 1 (a) Post-replicatory cytosine C-5 modifications in human
DNA. 5-Methylcytosine at CpG dinucleotides in promoter regions is
associated with transcriptional silencing. 5mC and 5hmC formation
are catalysed by DNA methyltransferases and 2OG oxygenases,
respectively. (b) Potential roles of 5mC and 5hmC in transcriptional
repression and activation.
z Current Address: Dept. of Chemistry and Chemical Biology,
Harvard University, Cambridge, Massachusetts, USA.
c
This journal is The Royal Society of Chemistry 2011
Chem. Commun., 2011, 47, 5325–5327 5325

View Article Online
Fig. 2 Analyses of melting temperatures of duplexes containing
single C, 5hmC or 5mC nucleotides (Table S1). x-axis, forward strand
modification; colour: complementary strand modification.
Fig. 3 Concentration-dependent UV-studies on 10mer duplexes with six
modifications (bold). (a) Effect of C/5hmC/5mC at single concentrations.
(b) Van’t Hoff analysis. Thermodynamic parameters were derived using
best-fit values of linear regression (mean ꢂ 95% CI, ESIw), (c) Dissociation
enthalpy (ꢀDbH1), (d) Dissociation entropy (ꢀDbS1). (e) Binding constants
(calcd. from DbG1, 37 1C). (*** denotes P o 0.01).
the cytosine equivalents (C/C). With the exception of duplexes
based on sequence 1 with an hmC on the forward strand, these
initial results suggested that, duplexes containing 5hmC have a
T
m resembling that of unmodified DNA, with Tm increasing in
without affecting DbH1, comparisons of the stability of
poly(GꢁC) to that of poly(Gꢁ5mC),11 and work on dye-assisted
melting-temperature-based mapping of CpG methylation.15
Analyses to date, including our work, imply that duplex
stabilisation by 5mC is predominantly due to entropic rather
than enthalpic effects. The overall results for 5mC also
demonstrate that data derived from model sequences, including
those used in our work, may be of general significance. It is
reasonable to propose that the same applies to 5hmC.
In contrast to 5mC-containing DNA, the enthalpic and
entropic terms as well as the binding constants (Kb) for
5hmC-containing DNA were remarkably similar to those of
unmodified DNA, as apparent from the van’t Hoff plot
(Fig. 3b), i.e. the stabilising effect of 5mC is substantially
reduced when substituted for 5hmC (Fig. 3e).
the order C r 5hmC o 5mC. They also suggest that, at least
with singly modified duplexes, the relative degree of stabilisation
depends on sequence context, and, when studying hetero-
geneously modified duplexes, on which strand the modification
is placed. However, Tm differences for single modifications are
small. Hence, we investigated the effect of increasing the
number of 5mC and 5hmC modifications to reduce the
significance of sequence context.
Our model sequence comprised a 10mer self-complementary
CpG dinucleotide-based sequence with modified bases on
both strands (Fig. 3). CD analyses for these duplexes were
characteristic of B-DNA rather than the A- or Z-conformations
(Fig. S3), indicating that 5hmC is a stable DNA component that
does not induce overall major conformational changes (in short
sequences). As with the results for singly modified duplexes, 5mC
caused duplex stabilisation (DTm = +1.1 1C per 5mC). In
To obtain additional insight into the determining factors of
the effects of 5hmC on duplex stability, we then performed
analyses (Fig. 4) in a different sequence context using isothermal
titration calorimetry (ITC) under more biologically relevant
conditions (37 1C, 200 mM NaCl).y The concentrations of
single-stranded DNA were chosen to enable reliable curve
fitting (non-linear regression, ESIw) and therefore a simultaneous
determination of Kb, DbH1 and DbS1 (Fig. S4a and S4bw) with
comparable accuracy, and the sequence was selected to minimise
secondary structure formation.
contrast to 5mC, 5hmC elicited a much smaller effect (DTm
=
+0.4 1C per 5hmC) relative to unmodified C. These results
suggest that the relatively large stabilising effect of cytosine
methylation may, at least partly, be reversed by hydroxylation.
To investigate the basis of the effects of 5hmC, the enthalpic
(DbH1) and entropic (DbS1) components of the free binding
energy (DbG1) were initially determined using concentration-
dependent UV melting (UVM). Analysis of the concentration-
dependent UVM data utilises the van’t Hoff equation to derive
thermodynamic parameters from the experimentally determined
relationship between total ODN concentration and Tm (Fig. 3b).
We used self-complementary 10mer ODNs with three modified
nucleotides per strand to maximise potential Tm differences and
ensure equal quantities of forward/complementary strands.
The derived DbH1/DbS1/Kb values (Fig. 3c,d,e and Table S2w)
imply that the stabilising effect of 5mC relative to 5hmC is
primarily due to a less unfavourable change in binding entropy
(Fig. 3d), which counteracts the less exothermic binding
enthalpy (Fig. 3c); 5mC causes a significant reduction in
DbS1 (Po10ꢀ4) relative to unmodified duplexes. These data
concur with observations of a methylation-induced increase10
in the Tm of the hairpin 50-CGCGCGTTTCGCGCG-30
The titration curves for unmodified (C/C, Fig. 4a),
5hmC-containing (5hmC/5hmC, Fig. 4b) and 5mC-containing
homoduplexes (5mC/5mC, Fig. 4c) were consistent with a
single-site binding model.z The derived binding constants
(Fig. 4d) show that methylation on both DNA strands causes
stabilisation of B10-fold (5mC/5mC vs. C/C, P o 10ꢀ4
)
relative to unmodified duplex, whereas for 5hmC the binding
constants were similar to those of unmethylated duplexes
(B1.6-fold stabilisation of 5hmC/5hmC vs. C/C).
As for the stabilising effect of 5mC vs. C, the destabilising
effect of 5hmC relative to 5mC was evident in multiple
sequence contexts, being apparent in the binding constants
derived from UVM (Fig. 3e) and ITC data (Fig. 4d). This
trend was not limited to fully modified duplexes, but also
c
5326 Chem. Commun., 2011, 47, 5325–5327
This journal is The Royal Society of Chemistry 2011

View Article Online
Fig. 4 ITC of duplexes with (a) unmodified C, (b) 5hmC and (c) 5mC. (a, b, c) Upper panels, data after baseline correction; Lower panels, binding
isotherms from integration. (d) Kb values (mean ꢂ s.d. of at least 2 independent expts., ESIw). Titration curves correspond to duplexes with same
modification on both strands and are representative of expts. giving the indicated Kb values. (***denotes P o 0.01).
observed within ODN sets bearing a particular modification
on the forward (groups of data bars in Fig. 4d) or reverse
strands (data bars of same colour in Fig. 4d), with the binding
constants of 5hmC-modified duplexes resembling those of
unmodified rather than methylated duplexes. In the case of
the duplex-stabilising effect of 5mC, high-level quantum
calculations suggest enhanced stacking interactions16,17
conferred by the hydrophobic methyl group may be responsible,
thus raising the possibility of the reversal of these effects by
hydroxylation.
grant [HEALTH-F4-2008-201418] READNA (A.H. E-S and
T.B.), Pfizer and the Wellcome Trust for funding. We thank
Kevin Dack for discussions.
Notes and references
y The absolute values from UVM and ITC differ, because the
techniques measure thermodynamic parameters at Tm, and at the
exptl. temp., respectively. While parameters obtained at one temp.
can be extrapolated to another, additional conformational ‘melting’
transitions of single strands occurring between these temperature need
to be taken into account, precluding direct quantitative comparison.
z Table S3 gives data for titrations/accuracy of parameter fitting;
Table S4 gives data for group averages of all unique combinations of
C/5hmC/5mC duplexes.
Overall, the combined analyses reveal the potential of 5mC
hydroxylation to, at least in part, reverse the thermodynamic
duplex stabilisation conferred by 5mC in multiple sequence
contexts, without a requirement for demethylation or further
modification. Although the roles of proteins in the regulation
of transcription have been extensively studied, to our knowledge
the effect that methylation has ‘directly’ on DNA duplex
stability has not been widely considered as a contributing
factor in transcriptional regulation. With all other factors
being equal, our results lead to the proposal that 5mC
hydroxylation is more likely associated with transcriptional
activation than repression (Fig. 1b). Consistent with our
proposal, two recent studies have associated TET activity with
transcriptional activation.18,19 Of course, as for 5mC, the
effects of 5hmC may be differentially manifested in the
presence of additional factors, including, for example,
DNA-binding20 proteins such the CXXC domains which are
part of the TET enzymes.18 Given the established regulatory
role of chromatin structure on transcription, TET-catalysed
5mC-hydroxylation of CpG islands in promoter regions of
transcriptionally inactive genes could act to reduce duplex
stability, facilitate strand separation or alleviate 5mC-con-
trolled transcriptional repression by enabling recruitment of
transcriptional machinery. Finally, given the role of the
hypoxia inducible factor 2OG-dependent hydroxylases in the
oxygen dependent regulation of transcription, it is possible
that the 5hmC formation represents a mechanism for oxygen
dependent regulation of expression, including in an epigenetic
manner.
1 J. T. Attwood, R. L. Yung and B. C. Richardson, Cell. Mol. Life
Sci., 2002, 59, 241–257.
2 A. Portela and M. Esteller, Nat. Biotechnol., 2010, 28, 1057–1068.
3 M. Esteller, Hum. Mol. Genet., 2007, 16, R50–59.
4 S. Kriaucionis and N. Heintz, Science, 2009, 324, 929–930.
5 M. Tahiliani, K. P. Koh, Y. Shen, W. A. Pastor, H. Bandukwala,
Y. Brudno, S. Agarwal, L. M. Iyer, D. R. Liu, L. Aravind and
A. Rao, Science, 2009, 324, 930–935.
6 S. Ito, A. C. D’Alessio, O. V. Taranova, K. Hong, L. C. Sowers
and Y. Zhang, Nature, 2010, 466, 1129–1133.
7 U. Bacher, C. Haferlach, S. Schnittger, A. Kohlmann, W. Kern
and T. Haferlach, Ann. Hematol., 2010, 89, 643–652.
8 U. Bacher, S. Schnittger and T. Haferlach, Curr. Opin. Oncol.,
2010, 22, 646–655.
9 M. A. McDonough, C. Loenarz, R. Chowdhury, I. J. Clifton and
C. J. Schofield, Curr. Opin. Struct. Biol., 2010, 20, 659–672.
10 L. E. Xodo, G. Manzini, F. Quadrifoglio, G. A. van der Marel and
J. H. van Boom, Nucleic Acids Res., 1991, 19, 5625–5631.
11 H. H. Klump and R. Loffler, Biol. Chem. Hoppe-Seyler, 1985, 366,
345–353.
12 D. J. Prockop and K. I. Kivirikko, Annu. Rev. Biochem., 1995, 64,
403–434.
13 A. P. Hardy, I. Prokes, L. Kelly, I. D. Campbell and
C. J. Schofield, J. Mol. Biol., 2009, 392, 994–1006.
14 A. S. Hansen, A. Thalhammer, A. H. El-Sagheer, T. Brown and
C. J. Schofield, Bioorg. Med. Chem. Lett., 2011, 21, 1181–1184.
15 C. M. R. Lopez, B. G. Asenjo, A. J. Lloyd and M. J. Wilkinson,
Anal. Chem., 2010, 82, 9100–9108.
16 C. Acosta-Silva, V. Branchadell, J. Bertran and A. Oliva, J. Phys.
Chem. B, 2010, 114, 10217–10227.
17 N. Tretyakova, R. Guza and B. Matter, Nucleic Acids Symp. Ser.,
2008, 52, 49–50.
18 S. Ito, A. C. D’Alessio, O. V. Taranova, K. Hong, L. C. Sowers
and Y. Zhang, Nature, 2010, 466, 1129–1133.
19 H. Zhang, X. Zhang, E. Clark, M. Mulcahey, S. Huang and
Y. G. Shi, Cell Res., 2010, 20, 1390–1393.
20 V. Valinluck, H.-H. Tsai, D. K. Rogstad, A. Burdzy, A. Bird and
L. C. Sowers, Nucleic Acids Res., 2004, 32, 4100–4108.
We thank the Biotechnology and Biological Sciences
Research Council, Cancer Research UK (to A.T.), the European
Community’s 7th Framework Programme [FP7/2007-2013]
c
This journal is The Royal Society of Chemistry 2011
Chem. Commun., 2011, 47, 5325–5327 5327