Biochemical reconstitution of TET1-TDG-BER-dependent active DNA demethylation reveals a highly coordinated mechanism

Abstract

Cytosine methylation in CpG dinucleotides is an epigenetic DNA modification dynamically established and maintained by DNA methyltransferases and demethylases. Molecular mechanisms of active DNA demethylation began to surface only recently with the discovery of the 5-methylcytosine (5mC)-directed hydroxylase and base excision activities of ten-eleven translocation (TET) proteins and thymine DNA glycosylase (TDG). This implicated a pathway operating through oxidation of 5mC by TET proteins, which generates substrates for TDG-dependent base excision repair (BER) that then replaces 5mC with C. Yet, direct evidence for a productive coupling of TET with BER has never been presented. Here we show that TET1 and TDG physically interact to oxidize and excise 5mC, and proof by biochemical reconstitution that the TET-TDG-BER system is capable of productive DNA demethylation. We show that the mechanism assures a sequential demethylation of symmetrically methylated CpGs, thereby avoiding DNA double-strand break formation but contributing to the mutability of methylated CpGs.

ARTICLE

Received 19 May 2015 | Accepted 22 Jan 2016 | Published 2 Mar 2016

Alain R. Weber1, Claudia Krawczyk1, Adam B. Robertson2, Anna Kunierczyk3, Cathrine B. Vgb3, David Schuermann1, Arne Klungland2 & Primo Schar1

Cytosine methylation in CpG dinucleotides is an epigenetic DNA modication dynamically established and maintained by DNA methyltransferases and demethylases. Molecular mechanisms of active DNA demethylation began to surface only recently with the discovery of the 5-methylcytosine (5mC)-directed hydroxylase and base excision activities of teneleven translocation (TET) proteins and thymine DNA glycosylase (TDG). This implicated a pathway operating through oxidation of 5mC by TET proteins, which generates substrates for TDG-dependent base excision repair (BER) that then replaces 5mC withC. Yet, direct evidence for a productive coupling of TET with BER has never been presented. Here we show that TET1 and TDG physically interact to oxidize and excise 5mC, and proof by biochemical reconstitution that the TETTDGBER system is capable of productive DNA demethylation. We show that the mechanism assures a sequential demethylation of symmetrically methylated CpGs, thereby avoiding DNA double-strand break formation but contributing to the mutability of methylated CpGs.

DOI: 10.1038/ncomms10806 OPEN

Biochemical reconstitution of TET1TDGBER-dependent active DNA demethylation reveals a highly coordinated mechanism

1 Department of Biomedicine, University of Basel, Mattenstrasse 28, Basel CH-4058, Switzerland. 2 Department of Molecular Microbiology, Oslo University Hospital, Rikshospitalet, NO-0372 Oslo, Norway. 3 Proteomics and Metabolomics Core Facility, PROMEC, Department of Cancer Research and Molecular Medicine, Norwegian University of Science and Technology, NO-7489 Trondheim, Norway. Correspondence and requests for materials should be addressed to P.S. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 7:10806 | DOI: 10.1038/ncomms10806 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10806

DNA methylation in mammals occurs at the C5 position of cytosines (5-methylcytosine, 5mC) and is found predominantly within CpG dinucleotides, affecting

6090% of such sites1. Modulating chromatin states and thereby transcriptional activity and genome stability, DNA methylation plays an important epigenetic role in various biological processes2. It is generally viewed as a static DNA modication but recent research has shown that under specic circumstances, DNA methylation can be subject to dynamic change. This is best illustrated by its genome-wide erasure during early embryonic development35 or in maturing primordial germ cells6. Locus-directed DNA demethylation has also been observed in somatic cells upon triggering transcriptional activation in various ways79. Both passive and active pathways of DNA demethylation were proposed to operate in these contexts but the mechanisms underlying active demethylation, in particular, have remained controversial10.

Recent evidence substantiates an involvement of the ten eleventranslocation (TET) family of dioxygenases11. TET proteins oxidize 5mC to 5-hydroxymethylcytosine (5hmC), 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC), all of which have been implicated as intermediates of DNA demethylation1215. Regarding active mechanisms, 5fC and 5caC appear of particular relevance, as these bases are substrates for the thymine DNA glycosylase (TDG), a DNA repair protein with an ability to excise various cytosine and 5mC base derivatives from DNA12,1618. Thus, TET and TDG together constitute catalytic activities capable of oxidation and removal of 5mC in DNA. This biochemical reasoning is supported by the phenotype of TET1922 and TDG knockout mice2325, as well as embryonic stem cells, all showing aberrations in DNA methylation23,2629.

TET and TDG thus initiate active DNA demethylation by oxidation and excision of 5mC in DNA and the anticipated downstream events will be the excision and repair of the resulting abasic site (AP-site) by the DNA base excision repair (BER) system. An engagement of the core BER system implies that the AP-site is rst incised by an AP endonuclease (that is, APE1), which generates a DNA single-strand break that then engages, through activation of poly (ADP-ribose) polymerase 1, the X-ray repair cross-complementing protein 1 (XRCC1), DNA ligase 3 (LIG3) and DNA polymerase b (POLb) for DNA gap lling with an unmethylated C and ligation30. Although this mechanism is plausible and widely accepted, there is little evidence supporting a direct link between TET and BER; a productive action of TET with the BER system on a 5mC substrate has not been shown, nor have the basic mechanistic features of such a process been addressed.

The aim of this study was therefore to reconstitute the full DNA demethylation system in vitro and to address specic properties of the DNA transactions involved. We investigated physical and functional interactions between TET1 and TDG, and tested the hypothesis that methylated DNA substrates can be converted to unmethylated DNA through oxidation and BER of 5mC. We addressed the strand specicity of the reaction, whether symmetrically modied CpGs can be demethylated without DNA fragmentation and how complex lesions such as occurring by simultaneous oxidation and deamination of opposite 5mCs within a CpG dinucleotide do affect the demethylation outcome. The data proof full functionality of a TET1TDGBER-based DNA demethylation system on hemi- and fully methylated DNA, and show that the molecular transactions involved are coordinated in a manner avoiding DNA fragmentation but creating a risk for mutation if deamination and demethylation events coincide within a CpG.

ResultsTET1 and TDG interact physically. The model of TET TDG-mediated oxidative DNA demethylation postulates a coupled action of both enzymes to facilitate an efcient removal of 5mC. To address the mode of cooperation between TET and TDG, we investigated their physical interaction, rst by co-expression and afnity purication of a full-length carboxy-terminally 6His-tagged TET1 (TET1His6) with a C-terminally glutathione S-transferase (GST)-tagged TDG (TDGGST). Although co-expression with TDG positively affected full-length TET1 expression, enrichment of TET1His6 via Ni-NTA chromatography yielded little full-length protein but prominent, presumably C-terminal fragments of B140150, 90 and 6070 kDa, possibly reecting proteolysis of poorly structured domains. Size fractionation by gel ltration then showed that two of these fragments (140150 and 6070 kDa) co-eluted with full-length TDGGST in high-molecular-weight fractions (200600 kDa) at high ionic strength (500 mM NaCl) and down to concentrations in the 100 nM range, as assessed semi-quantitatively on the basis of immunoblot signals (Fig. 1a). This indicated the formation of stable TET1TDG complexes. The 90-kDa TET1His6 fragment appeared in lower-molecular-weight fractions (90200 kDa) and only partially co-eluted with TDGGST, indicating a weaker interaction with TDG.

To further characterize the TET1TDG interaction, we used the yeast two-hybrid system. Four protein fragments spanning the entire TET1 polypeptide (Fig. 1b) were fused to the Gal4-binding domain (bait) and co-expressed separately with TDG fused to the Gal4 activation domain (prey) in yeast. Growth on media selecting for two-hybrid reporter gene activation indicated physical interactions between TDG and TET1 fragments 2 and4. These results thus indicated that TET1 harbours specic TDG interaction domains in its amino terminus (amino acids (aa) 397931) comprising the CXXC motif and in its C-terminal catalytic domain (aa 13672057) (Fig. 1b). We next performed co-precipitations from lysates of Escherichia coli cells co-expressing TDGGST with either a His6-labelled TET1 N-terminal fragment (TET1N; aa 3011366) or the TET1 catalytic domain (TET1CD; aa 13672057) (Fig. 1c). TDGGST co-eluted from the Ni-NTA resin in the bound fraction with both TET1 fragments (Fig. 1c, Ni-NTA). The outcome was the same when we enriched for TDGGST; both TET1 fragments co-eluted in the bound fraction after GST afnity purication (Fig. 1c, GST). The results of all protein interaction assays led us to conclude that TET1 and TDG physically interact through specic N-terminal and C-terminal TET1 domains.

TET1CD and TDG act in concert to release 5mC. To examine the activity of the TETTDG complex, we co-expressed the catalytic domain of TET1 (His6TET1CD) and TDGGST, as well as combinations of their catalytic-dead variants ((His6TET1CDDcat (H1652Y; D1654A) with TDGGST;

His6TET1CD with TDGDcatGST (N151A)) in E. coli and enriched the complexes by Ni-NTA chromatography (Supplementary Fig. 1). We then measured catalytic activities in a base release assay31 with two uorescein-labelled synthetic 60-bp DNA substrates containing a single 5mC or 5hmC. Incubation of the enriched His6TET1CDTDGGST fraction with both DNA substrates (1 h, 37 C) generated a substantial amount of DNA incisions at the position of the modied cytosines (Fig. 2a). This 5mC/5hmC excision activity was not detectable when either of the two proteins was mutated at its catalytic site, establishing that the excision of 5mC and 5hmC from DNA requires the catalytic activities of both TET1 and TDG.

2 NATURE COMMUNICATIONS | 7:10806 | DOI: 10.1038/ncomms10806 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10806 ARTICLE

a

b

~140 kDa

CXXC C DSH

CXXC C DSH TET1-His6

His6

2,075 aa (227 kDa)

TET1

1 3 4

2

TDG-GST

652 aa (74 kDa)

GAL4

AD BD SC-LEU-TRP

SC-LEU-TRP-HIS

-ADE 5 mM 3 AT

kDa

670 50

160

kDa

150 100

75

670 50

160

670 50

160

TET1-1

150 100

75

TDG TDG TDG TDG

200

116

97 66

45

TET1

TDG

TET1-2 TET1-3 TET1-4

-TET1

TDG

V TET1-1 TET1-2 TET1-3 TET1-4

V

kDa

V V V V

-TDG

V

lTAg

P53

c

His6-TET1N

CXXC

His6

TDG-GST

1,080 aa (117 kDa)

Cat His6-TET1CD

His6

TDG-GST

In w b In

720 aa (79 kDa)

TET1N

TET1CD

652 aa (74 kDa) 652 aa (74 kDa)

His6-TET1CD - TDG-GST

w b

His6-TET1N - TDG-GST

In w b In w b

-TET1

Ni-NTA GST

Ni-NTA GST

-TDG -TDG

-TET1 -TET1

kDa

150

100 75

50

kDa

150 100

75

50

100 75

TDG

TDG

100 75

50

50

-TDG

-TDG

-TET1

TDG

TDG

TET1CD

TET1N

Figure 1 | TET1 physically interacts with TDG. (a) Size fractionation by gel ltration at high ionic strength (500 mM NaCl) of Ni-NTA-enriched lysates ofE. coli cells co-expressing TET1His6 and TDGGST from constructs indicated. Fractions were analysed by SDSPAGE (left panel), and TET1 and TDG detected by immunoblotting (right panel); molecular weights of gel ltration standards are indicated. (b) Yeast two-hybrid analysis of the TET1TDG interaction. TET1 domains cloned into the GAL4 activation domain (AD), TET1-1 (aa 1491), TET1-2 (aa 397931), TET1-3 (aa 8701403) and TET1-4 (aa 1,3672,057) are indicated at the top. Shown is the growth of serial dilutions of strains co-expressing TET1 domains fused to the AD and TDG fused to the GAL4-binding domain (BD) and respective negative controls (TET1 domains or TDG co-expressed with the vector control (V)) on permissive and selective media. The large T antigen (lTAg) and p53 fused to the AD and BD, respectively, served as a positive control. (c) Immunoblotting of fractions obtained from Ni-NTA and GST purications using E. coli extracts co-expressing His6TET1N and TDGGST (left panel), or co-expressing His6TET1CD and

TDGGST (right panel). Expression constructs used are indicated; TET1N (aa 3011366) and TET1CD (aa 13672057); b, bound fraction, in, input; w, wash.

To conrm that the intermediates generated in these assays were oxidized 5mC or 5hmC, we examined the DNA products generated by His6TET1CD in the absence of TDG. Purication of His6TET1CD by Ni-NTA and ion exchange chromatography yielded two prominent protein fragments, both corresponding to TET1CD (Supplementary Fig. 2). Mass spectrometry identied the smaller B75 kDa fragment as an N-terminal truncation of B240 amino acids, including the conserved Cys-rich domain, which was shown to be essential for the catalytic activity32. To test the catalytic activity of this His6TET1CD preparation (Fig. 2b), we in vitro methylated highly pure plasmid DNA using the M.SssI CpG methyltransferase to completion (200 pmol mCpG sites per mg DNA), reacted this substrate DNA (200 ng, 40 pmol mCpGs)

with puried His6TET1CD (500 ng, 6 pmol) at 37 C for 1 h and detected the cytosine modications generated by immunoblot analysis with antibodies against 5mC, 5hmC, 5fC and 5caC. All detectable 5mC was fully oxidized to 5hmC, 5fC or 5caC under the reaction conditions mentioned. His6TET1CD thus carried out all predicted 5mC oxidation steps in vitro, whereby the conversion of 5mC to 5hmC appeared to be the most efcient step (Fig. 2b).

We next used separately puried TET and TDG proteins to reconstitute the 5mC release. To allow for preformation of the

TETTDG complex, we mixed His6TET1CD with His6TDG (Supplementary Fig. 2) at a ratio of 100:50 nM (most active ratio by titration), respectively, before addition of DNA substrates. A twofold molar excess of TETTDG (50 nM complex) over substrate DNA (25 nM) and an incubation of 60 min at 37 C resulted in efcient release of both 5mC and 5hmC (Fig. 2c). Notably, 5mC was nearly as efciently excised as 5hmC, which, given the single turnover setup in this assay, indicated that the oxidation of 5hmC by TET1CD was rate limiting in this assay.

5caC served as a control for TDG activity and was processed most efciently, as expected.

Together, these results establish that TET1CD and TDG

activities can be combined to act in concert to efciently excise 5mC from DNA, thereby generating alkaline labile AP-sites in DNA.

TET stabilizes TDG activity. To address whether TET1 and TDG cooperate at the level of their catalytic activities, we examined the effect of TDG on the efciency of 5mC oxidation by TET1CD. In this setup, we used puried, catalysis-decient

TDG, to limit excision of TET-generated 5fC and 5caC. We pre-incubated His6TET1CD (50 nM) with or without

NATURE COMMUNICATIONS | 7:10806 | DOI: 10.1038/ncomms10806 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10806

a

b

G G

X = 5mC, 5hmC

/

CX

F

M.SssI M.SssI

TET1CD TET1CD

G5mC

G5hmC

TET1CD-TDG

/

wt-wt

cat-wt

wt-cat

wt-wt

cat-wt

wt-cat

100 ng Plasmid DNA

2x Serial dilution

-5mC

-5hmC

60mer

60mer

32mer

Substrate 1Substrate 3

23mer

100 ng Plasmid DNA

2x Serial dilution

Product (%)

Product (%)

20

36

-5fC

-5caC

12

36

c

CX

G G

X = 5mC, 5hmC, 5caC

F

G5mC

G5hmC

G5caC

G5mC

G5hmC

G5caC

TET1CD

TDG

wt

wt cat

wt

wt

wt cat

wt wt

wt

wt

wt

wt cat

wt wt

wt

cat

cat

wt

wt

wt cat

wt

60mer 59mer

25mer

Substrate 1

Substrate 2

23mer

19 1.7

6 0.6

22 0.4

5 1.6

78 9.9

5 1.0

17 4.8

5 1.5

22 3.5

8 2.4

83 0.7

6 1.7

Product (%) Product (%)

Figure 2 | Combined TET1 and TDG activity releases 5mC through oxidized intermediates. (a) Base-excision activity of Ni-NTA-enriched His6TET1CDTDGGST on synthetic DNA substrates as indicated. The ability to generate alkaline-sensitive AP-sites in substrates containing either single

G 5mC or G 5hmC base pairs was assayed with enriched His6TET1CDTDGGST consisting of wild-type proteins or respective mutant variants

(His6TET1CDDcatTDGGST, His6TET1CDTDGDcatGST). Products were separated by denaturing gel electrophoresis, visualized with uorescent scanning and quantied; positions of the 60mer substrate DNA and product fragment are indicated. (b) Slot blot analysis of plasmid oxidation by puried

His6TET1CD. In-vitro-methylated pUC19 plasmid DNA (800 nM) was treated with His6TET1CD (125 nM) and cytosine modications were detected by immunblotting with specic antibodies against 5mC, 5hmC, 5fC and 5caC. (c) Reconstitution of 5mC/5hmC base release with puried His6TET1CD

and His6TDG proteins. DNA substrates (25 nM) containing either G 5mC, G 5hmC or G 5caC base pairs were reacted with preassembled

His6TET1CDHis6TDG (50 nM), reaction products separated by denaturing gel electrophoresis, visualized and quantied. Positions of the 60mer substrate DNA and product fragments are indicated. Shown are mean values with s.d. (n 3).

His6TDGDcat (25 nM), added 5mC substrate (25 nM), stopped the reactions at different time points and monitored the presence of 5fC and 5caC in the recovered DNA by digestion with puried active TDG (250 nM). The presence of TDGDcat had a minor effect on 5mC oxidation by His6TETCD (Fig. 3a). This result was corroborated in a methylated plasmid oxidation assay analysed by quantitative liquid chromatography tandem mass spectrometry (LCMS/MS). In this assay, we did however observe a slightly reduced conversion of 5hmC to 5fC and 5caC in the presence of TDGDcat (Fig. 3b). The reduced 5fC levels and the lack of detectable 5caC in the reactions with TDGDcat probably reect the residual activity of the TDG catalytic mutant towards these substrates, in particular in the presence of TET1CD (Fig. 3c)31.

The catalytic dead TDG may also mask 5fC and prevent further oxidation by TET1CD. These results show that under single turnover conditions, puried His6TET1CD can efciently oxidize 5mC to generate 5fC/5caC, irrespective of whether TDG is present or not, although its efciency in 5hmC oxidation may be reduced under multiple turnover conditions.

Vice versa, the presence of a twofold molar excess of TET1CD

had a positive effect on TDG activity on 5fC and 5caC excision when compared with BSA added to the same molarity. This was

most notable in reactions with TDGDcat where the glycosylase activity is rate limiting (Fig. 3c). Under these conditions, excision of both 5fC and 5caC was signicantly increased in the presence of TET1CD, in particular after prolonged incubation, indicating that the interaction of TET1CD with TDGDcat stabilizes TDG activity. Notably, the stimulation observed in 5fC excision may in part be due to the conversion of 5fC to 5caC in the presence of TET1CD and reect the activity of TET1CDTDGDcat on 5caC.

TET1CD also stabilized the fully active wild-type TDG; a twofold molar excess of TET1CD signicantly enhanced TDG-dependent excision of 5caC (Fig. 3d upper panel) and T (Supplementary Fig. 3b). As a large molar excess of BSA was required to achieve a similar stabilizing effect on TDG (Fig. 3d lower panel), we conclude that the TET1 effect is due to its specic interaction with TDG.

Reconstitution of TET-BER-mediated DNA demethylation. TET1TDG-mediated active DNA demethylation implicates the engagement of BER in the restoration of the unmethylated DNA sequence following 5fC or 5caC excision. To formally proof the functionality of such a pathway and to provide

4 NATURE COMMUNICATIONS | 7:10806 | DOI: 10.1038/ncomms10806 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10806 ARTICLE

a b

G G

C

X = 5mC

X Substrate 1

F

TET1CD vs TET1CD-TDGcat

1.0

TET1CD

TET1CD

TET1CD-TDGcat

Time (min) 15 30 60+

+ +

Mod. / total C mod.

0.8

15 30 60 + + +

0.6

TDG

TET1CD-TDGcat

0.4

60mer

0.2

0.0

23mer

0 20 40 60

Time (min)

TET1CD

TET1CD-TDGcat

40

TET1CD vs TET1CD-TDGcat

5-me(dC) 5-hm(dC)

5-hm(dC)

5-me(dC)

30

0.25

Product (%)

Mod. / total C mod.

0.15

5-f(dC) 5-ca(dC)

5-f(dC) 5-ca(dC)

0.20

TET1CD

TET1CD-TDGcat

20

10

0.10

0.05

*

0

*

15 30 60

60

0.00

Time (min)

0 20 40 60

Time (min)

c d

G G

C

X

C

X

G G

X = 5fC, 5caC Substrate 1

F

X = 5caC Substrate 1

0 0 5 10 15 20 25 30

F

TDGcat processing 5fC

Time (min)

100

TDG 5caC processing limiting BSA

Time (min)

25 TDGcat / BSA TDGcat-TET1CD

***

20

80

*** ***

*** ***

Product (%)

**

15

** **

Product (%)

60

10

40

TDG-TET1CD

TDG / BSA

5

20

0

0 0 5 10 15 20 25 30

0

0 0 15 30 60

15

30

TDGcat processing 5caC

Time (min)

TDGcat / BSA TDGcat-TET1CD

8

***

100

TDG 5caC processing excess BSA

Time (min)

80

6

Product (%)

***

**

4

Product (%)

*** ***

**

** **

60

40

2

TDG-TET1CD

TDG / BSA

20

Figure 3 | TET1CD stabilizes TDG activity. (a) Stimulatory effect of His6TDGDcat on His6TET1CD examined by base release assay. His6TET1CD (50 nM) or His6TET1CDHis6TDGDcat (25 nM) complex were incubated with DNA substrate (25 nM) containing a G 5mC base pair for the

indicated time. Recovered DNA was then assayed by a base release assay using His6TDG (250 nM) to monitor the presence of oxidized 5mC species. (b) LCMS/MS analysis of plasmid oxidation assays using His6TET1CD or TET1CDHis6TDGDcat. In-vitro-methylated pUC19 plasmid DNA (660 nM) was treated with either TET1CD (100 nM) or a preassembled TET1CDTDGDcat complex for the indicated time. DNA was analysed by LCMS/MS; shown are normalized mean values (mod/total C mod). (c) Activity of His6TDGDcat on 5fC and 5caC in the presence or the absence of His6TET1CD.

His6TDGDcatBSA (50 nM) or His6TDGDcatHis6TET1CD (50 nM) were incubated with DNA substrate (25 nM) containing a G 5fC or G 5caC at

37 C, reactions were stopped by the addition of NaOH at indicated time and analysed by denaturing gel electrophoresis. (d) The effect of His6TET1CD on

His6TDG catalysis assessed in base release assays. The time-dependent generation of AP-sites was measured after reaction of a 60mer substrate containing a single G 5caC (25 nM) base pair with a preassembled His6TDGHis6TET1CD (25 nM) or His6TDGBSA (25 nM) complex in the presence

(lower panel) or absence (upper panel) of a 60-fold molar excess of BSA. Reactions were stopped by the addition of NaOH after the indicated time and analysed using denaturing gel electrophoresis and uorescent scanning. Shown are mean values with s.d. P-values were calculated by the Students t-tests (*Po0.05, **Po0.01 and ***Po0.001).

a tool to investigate its mechanistic features, we reconstituted the entire process of active DNA demethylation with dened components. In addition to TET1CD and TDG, we puried to near homogeneity the enzymes of the core BER pathway33, APE1,

POLb and XRCC1-LIG3 (Supplementary Fig. 2). Using a 60-bp

substrate containing a single 5mC, we rst performed demethylation in step-by-step reactions to monitor the DNA intermediates generated. The combined action of His6TET1CD

and His6TDG generated an AP-site cleavable either chemically

by NaOH or enzymatically by APE1, to generate a 23-nt

NATURE COMMUNICATIONS | 7:10806 | DOI: 10.1038/ncomms10806 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10806

oligonucleotide with or without a 30-phosphate, respectively (Fig. 4a lanes 3 and 4). Following strand incision by APE1, POLb was able to insert a deoxycytidine monophosphate, thus generating a 24-nt oligonucleotide as the main product (Fig. 4a lane 5). Addition of either T4 ligase or an XRCC1LIG3 complex then efciently ligated the nicked intermediate, restoring a continuous 60mer DNA fragment (Fig. 4a lanes 6 and 7). The nearly complete re-ligation conrmed the efcient removal of the 50-dRP remains of the cleaved AP-site by POLb. These results establish that TET1 and TDG convert 5mC to DNA repair intermediates amenable to processing by the core BER system.

To test the accuracy of the reconstituted DNA demethylation process, we performed the reaction with a 59-bp DNA substrate presenting a hemimethylated CpG dinucleotide within a recognition site for the HpaII endonuclease (CCGG) (Fig. 4b). Owing to its methylation sensitivity, HpaII will not be able to cleave this substrate unless it undergoes successful and complete demethylation. We subjected the hemimethylated substrate to demethylation by the reconstituted TETTDGBER system and examined the generation of a cleavable restriction site by HpaII digestion (Fig. 4b). As expected, the asymmetrically methylated substrate was fully resistant to HpaII cleavage (Fig. 4b lane 2), as were substrates carrying the predicted intermediates 5hmC, 5fC and 5caC (Supplementary Fig. 4). Incubation in the presence of the reconstituted DNA demethylation system, however, generated HpaII-digestible DNA products, indicating that the methylated DNA fragment was converted into an intact unmethylated fragment (Fig. 4b lane3). Together, these results proof that TET1TDG-mediated oxidation and excision of 5mC generates intermediates for BER, which then acts to efciently restore the original DNA sequence in an unmethylated conguration.

Coordinated TET-TDG-BER action avoids DSB formation. CpGs in mammalian DNA are mostly symmetrically methylated, generating a potential conict for excision-repair-mediated DNA demethylation; that is, once started, a DNA demethylation event in one DNA strand would have to be completed before another event starts at the symmetrically opposite 5mC, which would otherwise lead to the formation of a DNA double-strand break

(DSB). We therefore asked whether in a symmetrically methylated CpG dinucleotide, demethylation events would generate DSBs or be conned to one strand at a time. For this purpose, we generated three 60-bp DNA substrates with either a uorescein-labelled bottom strand containing a single 5mC, a Texas Red-labelled top strand containing a single 5mC or both strands labelled and presenting a symmetrically methylated CpG (Fig. 5a). Incubation of all these substrates (25 nM) with a twofold molar excess of TET1CDTDG produced a solid 5mC release from both the bottom and the top strands, irrespective of whether the CpG was hemi- or symmetrically methylated (Fig. 5a). Activities on the top and bottom strands in hemimethylated substrates were similar, indicating the absence of sequence context effects in this setup (Fig. 5a lanes 2 and 4). Notably, the same reaction conditions applied to the substrate with 5mC modications on both strands produced approximately half the amount of incised product on each DNA strand with the total activity remaining constant (Fig. 5a lane 6). These results show that TET1CDTDG can act on both strands on a substrate containing a symmetrically methylated CpG and suggested that it does so in a sequential manner affecting only one strand at a time.

To further investigate the demethylation events at symmetrically modied CpGs, we separated TDG from TET1 activities and measured the kinetics of 5caC processing in the context of potentially arising DNA demethylation intermediates. Using equimolar substrate and enzyme concentrations34 (25 nM), we evaluated substrates containing a 5caC on the labelled DNA strand opposite an unmodied C, a 5mC or a 5hmC within the same CpG (Fig. 5b). Under the conditions applied, both initial rate and overall 5caC excision by TDG was not notably affected by the modication status of the symmetrically opposite C (Fig. 5b). 5caC was processed with appreciable efciency even in single-stranded DNA, corroborating the high afnity of TDG for this substrate. The situation when 5caC arises in both strands simultaneously is of particular interest, as it raises the possibility that TDG-initiated BER will induce DNA DSBs. We thus evaluated the behaviour of TDG in such a context, monitoring the release of 5caC from both strands in a time-course base release assay with a substrate (25 nM) carrying labels on both strands. Similar to the activity of TET1TDG on 5mC, TDG alone acted evenly on both strands carrying the 5caC

F

a b

C X

G

G

HpaII

X = 5mC

Substrate 1

F

G G

C G G C

C

X = 5mC

X Substrate 2

DSB

Substrate (50 nM)

CG/5mCG

TET1CD-TDG (100 nM) APE1 (200 nM)

POL (50 nM)

T4 lig (0.5 U) XRCC1-LIG3 (100 nM)

+

+ + + + + + + + +

+ + + +

+

NaOH (100 mM) +

Substrate

CG/CG

85 18

59 bp

CG/5mCG

TET1CD and BER +

ssDNA

dCTP (100 M) + + + +

Product (%)

60mer

23

24

23-P

Lower band (%)

3

15 14 16 4 4

Figure 4 | Full reconstitution of TETTDGBER-mediated DNA demethylation. (a) Intermediate steps of the oxidative DNA demethylation reaction were reconstituted and visualized by denaturing gel electrophoresis. Labelled 60mer substrate DNA containing one G 5mC base pair was incubated sequentially

with TETTDGBER enzymes at concentrations indicated. Reaction products were separated by denaturing gel electrophoresis and visualized by uorescent scanning; sizes of the 60mer substrate DNA and reaction products are indicated. (b) Complete DNA demethylation by the reconstituted TETTDGBER system analysed by the generation of a HpaII-sensitive restriction site. Reconstituted DNA demethylation was done with a 50-labelled 59-bp substrate containing one G 5mC base pair within a HpaII recognition site (CCGG). Recovered DNA was digested with methylation-sensitive HpaII endonuclease and

analysed by native PAGE; positions of the 59-bp substrate DNA and product fragment are indicated.

6 NATURE COMMUNICATIONS | 7:10806 | DOI: 10.1038/ncomms10806 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10806 ARTICLE

a

b

R

G G

Y

Y

G G

X Substrate 1

Y = C, 5mC X = C, 5mC

F

F

Y = C, 5mC,

5hmC X = 5caC

X Substrate 1

C* 5mC

5mC

Top (5-TR)

Bottom (5-FAM) C*

F-channel

5mC

+ + +

Product (%) 32 20

Product (%) 30 14

60mer

60mer

23mer

100

TET1CD-TDG

80

Product (%)

60

R-channel

5mC

35mer

40

20

0 0 5 10 15 20 25 30

Time (min)

ss 5caC CG/5caCG 5mCG/5caCG 5hmCG/5caCG

c

d

HpaII

X Substrate 2

G G

C G G C

R

Y

Y

G G

0 0 5 10 15 20 25 30

F

Y = 5caC X = 5caC

X Substrate 1

F

Y = C, 5caC X = C, 5caC

80

70

Substrate (25 nM)

CG/

CG

5caCG/

5caCG

60

Product (%)

50

TDG (50 nM)

CG/ 5caCG

APE1 (50 nM)

40

HpaII

+

+

++

+

+

+

+ +

30

5caC upper strand

20

10

ssDNA

DSB

Time (min)

5caC lower strand

Product (%) 83 15

59 bp

e

HpaII

G G

C G G C

Y

X Substrate 2

0.7

F

Substrate

Y = C, 5caC X = C, 5caC

CG/

CG

CG/ 5caCG

5caCG/

5caCG

BER

HpaII

+

+

+ +

+

+

59 bp

ssDNA

DSB

Product (%) 84 11

Figure 5 | Processing of differentially modied CpGs by TET1CDTDG or TDG. (a) Base release from fully methylated CpGs by His6TET1CDHis6TDG. His6TET1CDHis6TDG (50 nM) was incubated with labelled 60mer substrates (25 nM) containing a single 5mC modication on the uorescent labelled top (50-Texas Red, T) or bottom strand (50-uorescein, F) or a fully methylated CpG with labels on both strands. Product formation was monitored and quantied by denaturing gel electrophoresis and uorescent scanning (Texas Red, R-channel and uorescein, F-channel); positions of the 60mer substrate

DNA and the resulting base incision products of both strands are indicated. *Unlabelled DNA strand. (b) Release of 5caC from differentially modied CpGs by His6TDG. 60mer DNA substrates (25 nM) containing 5caC opposite C, 5mC or 5hmC in a CpG dinucleotide or in single-stranded (ss) DNA were incubated with His6TDG (25 nM) and analysed by denaturing gel electrophoresis. Shown are mean percentages of product formation with s.d. (n 3).

(c) 5caC release from a symmetrically modied CpG dinucleotide by His6TDG. A substrate (25 nM) containing 5caC on both strands within a CpG dinucleotide and labels of both strands was incubated with His6TDG (25 nM) for indicated time, analysed by denaturing gel electrophoresis and visualized by uorescent scanning of both labels. Shown are mean percentages of product formation with s.d. (n 3). (d) TDG and APE1 only generate

DNA DSBs at symmetrically modied CpGs. Base release assay using TDG and APE1 on a labelled 59-bp substrate containing either a single 5caC or a symmetrically 5caC-modied base pair within a HpaII recognition site (CCGG). Reactions were analysed by native PAGE. Substrate DNA and product fragment are indicated. (e) Full reconstitution of TDGBER on a symmetrically modied 5caC substrate. A labelled 59-bp substrate (25 nM) containing a symmetrically 5caC-modied base pair within a HpaII recognition site (CCGG) was incubated with TDG (40 nM) and BER factors (200 nM APE1, 40 nM POLb and 40 nM XRCC1-LIG3). Recovered DNA was digested with HpaII endonuclease and analysed by native PAGE. Substrate DNA and product fragments are indicated; ssDNA, free single-stranded DNA.

(Fig. 5c). The resulting plateau of single-strand incision at B50% indicated that the processing of one DNA strand by TDG largely inhibited base release from the other strand. This is a probable

consequence of TDGs tight interaction with AP-sites31,35, the coordinated dissociation of which36,37 may favour completion of the repair process to initiation of an additional repair event at the

NATURE COMMUNICATIONS | 7:10806 | DOI: 10.1038/ncomms10806 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10806

opposite strand. Only after prolonged incubation (30 min) this plateau increased above 50% for one DNA strand, indicating some turnover of TDG. To test whether the spontaneous turnover of TDG could eventually generate symmetrical AP-sites and potentially cause DSBs, we included APE1 (50 nM) in the assay (Fig. 5d). The combined action of TDG and APE1 indeed produced a notable fraction of DNA DSBs (15%). We therefore asked whether repair of a symmetrical demethylation intermediate is at all possible and can be achieved without the generation of DSBs. We thus used a substrate with a symmetrically 5caC-modied HpaII site in a reconstituted TDGBER assay and analysed the generation of cleavable HpaII sites. Both, hemi- and symmetrically 5caC-modied substrates were fully resistant to HpaII cleavage (Fig. 5e lane 4 and 6). However, incubation of the symmetrically 5caC-modied DNA (25 nM) with the TDGBER system (40 nM TDG, 200 nM APE1, 40 nM POLb and 40 nM XRCC1-LIG3) generated an appreciable amount of HpaII cleavable product, indicating that the 5caCs in both DNA strands were replaced with unmodied Cs (Fig. 5e lane 8). Notably, this process of symmetrical repair, which ultimately requires the breaking of both DNA strands did not lead to an accumulation of DSBs (o1%) (Fig. 5e lane 7), suggesting that in the presence of all repair factors repair events at both strands proceed preferentially in a sequential manner.

From these results, we conclude that DNA demethylation in vitro has no apparent strand and hence sequence-context preference. TET1TDG is capable of initiating active DNA demethylation in both strands of a fully methylated CpG. Once initiated in one strand, however, BER is completed before it restarts on the other strand, indicating that demethylation of symmetrically methylated CpGs occurs in a sequential manner.

DNA demethylation inhibits G T repair at methylated CpGs.

Another issue of BER-mediated demethylation at symmetrically methylated CpGs is the potential collision with 5mC deamination. 5mC in genomic DNA is susceptible to spontaneous hydrolytic deamination38, generating a thymine paired with a guanine. Such G T mismatches are recognized and excised also

by TDG. Enzymatic deamination coupled to BER has also been considered as a mechanism of active DNA demethylation7,24,39,40; it would replace a 5mC with an unmodied C through a mutagenic intermediate. To investigate potential interferences between deamination and oxidation-induced DNA demethylation pathways, we evaluated G T and G 5caC processing efciencies

in kinetic base release assays, using equimolar substrate/enzyme (His6TDG) concentrations (25 nM). When provided on separate DNA molecules, TDG processed the G T mismatch more

efciently than the 5caC substrate (Fig. 6a), showing that the mismatch is a preferred substrate as reported previously16. In a substrate where the G 5caC modication is next to a G T

mismatch within the same CpG dinucleotide, reecting a spontaneous deamination event on one strand while the other is being actively demethylated, TDG processes almost exclusively the 5caC, leaving the G T mismatch untouched (Fig. 6b).

The processing rate of 5caC was largely unaffected by the presence of the G T mismatch, indicating that in this

conguration 5caC is clearly the preferred substrate. The result was essentially the same when the modications were inversed within the same double-stranded substrate (Supplementary Fig. 5), thus excluding DNA strand or sequence-context effects as an explanation for the preference for 5caC.

This strong preference of TDG for the non-mutagenic 5caC next to a pre-mutagenic G T mismatch implies that TET

TDG-mediated active DNA demethylation has a potential to mutate CpG dinucleotides if it coincides with a deamination

event. To test this possibility, we used our fully reconstituted BER setup on a 59-bp substrate containing a G 5caC next to a

G T mismatch within an MscI recognition site and analysed the

generation of mutant demethylation products by endonuclease digestion. 5caC-directed sequential BER of this substrate would generate C to T mutations and thus create an MscI restriction site if two or more nucleotides were incorporated during the DNA resynthesis step (Fig. 6c). In the absence of the TDGBER machinery, no MscI cleavage products were detectable (Fig. 6c lane 3). However, full reconstitution of TDGBER generated a cleavable product, indicating that the 5caC was correctly replaced with a C but an A was incorporated opposite of T, thus manifesting the C to T transition and a loss of a CpG dinucleotide (Fig. 6c lane 4).

DiscussionRecent research on active DNA demethylation points towards a mechanism involving TET proteins and the DNA glycosylase TDG as well1214,16,17. A current model suggests that DNA demethylation through this pathway occurs in a stepwise manner via TET-catalysed oxidation of 5mC to 5fC and 5caC, which are then excised by TDG-dependent BER to restore an unmethylated DNA sequence. Despite the plausibility of this pathway, experimental evidence that directly links TET activity with TDG and BER is missing and fundamental mechanistic questions have not been addressed. The data presented here provide strong evidence for a coupling of 5mC oxidation and TDG-initiated BER in a cascade of enzymatic reactions that productively demethylates DNA. In vitro reconstitution of the active demethylation of symmetrically modied CpGs revealed a mechanism that is intrinsically coordinated to operate sequentially on both strands. Although this prevents the formation of DNA DSBs, and hence genomic instability, the process can be mutagenic if 5mC deamination and oxidative demethylation events coincide on opposite strands in a CpG dinucleotide.

In line with co-localization studies41, our work provides biochemical evidence for a direct and specic physical interaction of TET1 with TDG, implicating a link between 5mC oxidation and base excision. This interaction allowed us to enrich a functional TET1TDG complex from E. coli lysates that was highly active and capable of removing 5mC from a synthetic DNA substrate. In contrast to previous studies, showing 5mC conversion by TET and base excision by TDG in separate assays12,13,16, our data demonstrate a concerted action of both enzymes in 5mC oxidation and excision.

The relative high abundance of 5hmC in cells compared with 5fC and 5caC13,42 suggests that 5mC oxidation by TET enzymes is tightly regulated. A straightforward explanation could be that the rate of the oxidation of 5mC to 5hmC by TET enzymes is higher than that of the subsequent oxidations of 5hmC or 5fC, which may require stimulation by the presence of additional factors, such as the TDG13,43 and/or Gadd45 (refs 44,45). We examined this possibility but did not measure a stimulatory effect of TDG on TET1CD catalysis at any step of oxidation, neither did we observe such an effect for Gadd45a added to TET1CDTDG

(Supplementary Fig. 3a). These experiments were done with TET1CD, however, leaving the possibility that the missing

N terminus with its zinc nger CXXC domain may provide such regulatory function. Additional work is needed to address the mechanism of TET1 regulation, that is, to identify the factors determining the patterning of genomic 5hmC, 5fC and 5caC generation. The reconstituted demethylation assay presented here will be instrumental in this endeavor.

The engagement of a DNA glycosylase in active DNA demethylation inevitably generates a need for AP-sites

8 NATURE COMMUNICATIONS | 7:10806 | DOI: 10.1038/ncomms10806 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10806 ARTICLE

a b

R

G G

0 0 5 10 15 20 25 30

Y

C

R

Y = 5caC Substrate 1

X Substrate 1

Y

G G

C

G G

F

X = T

Y = 5caC X = T

X Substrate 1

F

100

100

80

80

Product (%)

Product (%)

60

60

40

40

Upper strand 5caC

20

20

0 0 5 10 15 20 25 30

Time (min)

CG/TG 5caCG/CG

Time (min)

Lower strand T

c

d

m

Cm

G G

C

R

G G

C

CTGGCCA

GGCCGGT

Substrate 2

F

TET

TDG

TET

TDG

m

Cm

G G Cm

BER

R

CTGGCCA

GACCGGT

Coincident deamination

Regular BER

Incomplete BER

MscI

F

POL

Substrate

CG/ CG

TG/ 5caCG

XRCC1

TET

TDG

BER

+

+

APE1 LIG3

Endo

+ +

ca

T

APE1

TET

TDG

G G

C

G G Cm

G G m

59 bp

TET

TDG

TDG

R-channel

POL

DSB

G G

C

Cm

XRCC1

16

APE1

APE1

Product (%)

F-channel

G G T

LIG3

TET

TDG

POL

59 bp

TDG

XRCC1

ssDNA

G

G

APE1

LIG3

A G

C

T G G

C

C

DSB

Product (%) 80 16

Mutation inducing DNA demethylation

Regular sequential DNA demethylation

DSB inducing DNA demethylation

Figure 6 | DNA demethylation blocks G T repair and can induce mutations. (a) Enzymatic activity of TDG on G 5caC- and G T-containing substrates.

Release of 5caC and T by His6TDG (25 nM) was monitored over time on 50-labelled 60-bp substrates (25 nM) containing either a G 5caC or G T base

pair. Reactions were stopped at indicated time, separated by denaturing gel electrophoresis, visualized with uorescent scanning and quantied. Shown are mean percentages of product formation with s.d. (n 3) (b) Base release from a substrate containing a G 5caC next to a G T mismatch. Substrate

preference of TDG (25 nM) was evaluated on a 59-bp DNA fragment (25 nM) containing 5caC on the labelled top strand (50-Texas Red) and T on the labelled bottom strand (50-uorescein) within the same CpG context as illustrated. Reactions were stopped after indicated time, separated by denaturing gel electrophoresis, and both strands visualized by uorescent scanning and quantied. Shown are mean percentages of product formation with s.d. (n 3).

(c) Full reconstitution of TDGBER on a G 5caC/G T-containing substrate. A labelled 59-bp substrate containing a G 5caC next to a G T mismatch was

incubated with His6TDG and BER factors. Correct repair of the 5caC and the introduction of an A opposite of T was monitored by MscI digestion and analysed by native PAGE and uorescent scanning. Unmodied (CG/CG) substrate DNA digested with HpaII was used as size marker; positions of the substrate DNA and product fragments are indicated. ssDNA, free single-stranded DNA. (d) Mechanistic model of TETTDGBER-mediated DNA demethylation. In the presence of all the necessary factors, DNA demethylation at fully methylated CpGs occurs in a coordinated and sequential manner to correctly re-establish the unmodied state (regular BER). Lack of coordination, for example, in the absence of downstream BER factors, repair-mediated DNA demethylation can lead to the induction of DNA DSBs (incomplete BER). Coincident oxidation and hydrolytic deamination at fully methylated CpG sites can lead to increased C to T transitions caused by the sequential repair mechanism (coincident deamination).

repair. Evidence supporting an involvement of the BER pathway has been reported for primordial germ cells, where an increase of DNA single-strand breaks and BER activity was linked to active global DNA demethylation46 and, in a more recent study, where various BER proteins were found to co-precipitate with overexpressed TET1 (ref. 41). With the successful reconstitution of TET1TDGBER-mediated DNA demethylation, we provide the rst evidence for a physical and

functional coupling of these factors in the oxidation and excision of 5mC and the resynthesis of an unmethylated C. Although such BER-mediated DNA demethylation seems mechanistically straightforward, it raises concerns regarding potential adverse effects on genome stability, in particular where the density of CpGs undergoing demethylation is high and excessive formation of DNA strand breaks might occur. It is therefore fair to assume that active demethylation in

NATURE COMMUNICATIONS | 7:10806 | DOI: 10.1038/ncomms10806 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10806

cells is a highly orchestrated process, controlled through regulatory mechanisms also involving posttranslational modications18,36,37. Our in vitro DNA demethylation system does not recapitulate regulatory actions of this kind but it does inform on intrinsic features of the mechanism regarding the potential of DNA DSB formation and the handling of complex substrates.

A distributive mode of action of TET proteins, for instance, would produce a variety of demethylation intermediates with 5caC placed opposite from 5mC, 5hmC or C within CpG dinucleotides, the precise conguration of which may then determine the efciency of initiation of BER. However, this seems an unlikely regulatory concept, as TDG processed 5caC with high efciency irrespective of the opposite C modication. Yet, our experiments indicate that although the TETTDG demethylase is capable of acting on both strands at symmetrically modied CpGs, it does so in a sequential manner without producing DNA DSBs. Even with substrates containing the efciently processed 5caC in both strands, TDG-mediated BER did not generate detectable DNA DSBs and this was not due to a preference of TDG for one strand in particular. In the case of an occurrence of symmetrical substrates within CpG dinucleotides, such as during symmetrical DNA demethylation, the high-afnity binding of TDG to AP-sites31,35 may constitute an important protective mechanism; not only will it provide an opportunity to coordinate AP-site repair but also protect the opposite strand from being processed at the same time. The importance of coupling base excision with the BER process in this delicate situation is highlighted by the observation that in the absence of POLb and XRCC1-LIG3, TDG and APE1 generated an appreciable amount of DSBs in symmetrically modied substrates (Fig. 5d). We therefore argue that BER in the context of active DNA demethylation occurs in a processive manner, where the initially attacked strand is fully repaired before processing of the opposite strand (Fig. 6d). This may explain how the replacement of symmetrical 5mC with unmodied C can occur without destabilizing the genome.

Another situation that may arise is the coincident deamination and oxidation of opposed 5mCs in CpG dinucleotides. Methylated CpGs are well known for their increased mutability, which is, to a large extent, due to the higher rate of hydrolytic deamination of methylated cytosines compared with unmethylated cytosines47. Such deamination will generate premutagenic G T mispairs within methylated CpG dinucleotides38. This

observation alone does not adequately explain the relatively high C to T mutation rates at such sites, as cells have efcient mechanisms in place to repair these G T mismatches, for

example, TDG- or MBD4-mediated BER18. Our data on G T

versus 5caC repair in CpG dinucleotides provide a plausible explanation for how G T mismatches might escape correction

and turn into mutations. Although, consistent with previous observations16, TDG processed the G T mismatch with higher

efciency than 5caC when the two lesions were analysed separately, 5caC was processed with a striking preference when both were present within the same CpG, reecting a situation where deamination occurs at a site undergoing active demethylation (Fig. 6d). This strong preference for the perfectly base-paired 5caC is consistent with a high-afnity binding of TDG to 5caC as implicated by the uniquely specic active site contacts it establishes with this base48,49. The sequential repair of both lesions, which helps avoid DSB formation, then turns into a disadvantage in this particular situation. The initiation of repair at the 5caC would mask a nearby G T mismatch for repair and

x the C to T mutation within the CpG dinucleotide whenever the resynthesis step of BER incorporates two or more nucleotides (Fig. 6d).

In conclusion, our data provide proof of functionality of an active DNA demethylation pathway based on the coupled oxidation and excision repair of 5mC; they provide insight into how intrinsic features of the mechanism allow demethylation of symmetrically methylated CpGs without the formation of DNA DSBs and how it may contribute to C to T mutagenesis within methylated CpG dinucleotides. Having a fully reconstituted DNA demethylation process established will allow future investigations into the detailed mechanism of the process, including the important aspect of TET regulation.

Methods

Bacterial expression vectors. The plasmids for the expression of TDG(GI: 37589917) (pTG-mTDGa.0, pET28-mTDGa.0 and pET28-mTDGa.1), TET1 (GI: 568968019) (pCDF-mTET1), TET1 catalytic domain (aa 13672057) (pCDF-His-mTET1CD and pCDF-His-mTET1CDDcat), TET1 N terminus(aa 3011366) (pACYC-mTET1-N), APE1 (GI:18375501) (pPRS125 and pEThis-APE1.0), POLb (GI:4505931) (pPRS112 and pQE30-6HIS-Polb), XRCC1 (GI:190684675) (pET-XRCC1) and LIG3 (GI:73747829) (pGEX4T-Lig3) were assembled by standard cloning methods based on PCR amplication with adaptor-oligonucleotides providing suitable restriction sites.

Antibodies. The following antibodies were used: TDG, rabbit polyclonal antibody 141, 1:20,000; TET1CD, rabbit polyclonal a-TET1 antibody (Millipore, catalogue number 09-872), 1:5,000; TET1-N, rabbit polyclonal a-TET1 antibody (Genetex, catalogue number GTX124207), 1:10,000; 5mC, mouse monoclonal a-5mC antibody (Diagenode, catalogue number C15200081), 1:250; 5hmC, rabbit polyclonal a-5hmC (Active motif, catalogue number 39769), 1:20,000; 5fC rabbit polyclonal a-5fC (Active motif, catalogue number 61223), 1:2,500; 5caC, rabbit polyclonal a-5caC (Active motif, catalogue number 61225), 1:2,000.

5-Carboxyethyl-N4-benzoyl-dC CE phosphoramidite. The 5caC phosphor-amidite (5-carboxyethyl-N4-benzoyl-dC CE) was synthesized in collaboration with Glen Research (USA).

Oligonucleotides. 60mer (Substrate 1 and 3) or 59mer (Substrate 2) double-stranded oligonucleotide substrates containing different modications were prepared by annealing of two complementary oligonucleotides synthesized by Adam Robertson or Microsynth (Switzerland). The upper strand was either unlabelled or carried a 50-Texas Red label, whereas the lower strand was unlabelled or carried a 50-uorescein label. Substrate 1 (standard) upper strand 50-TAGACA TTGCCCTCGAGGTACCATGGATCCGATGTXGACCTCAAACCTAGACGA ATTCCG-30 where X C, T, 5mC, 5hmC or 5caC. Substrate 1 lower strand strand

50-CGGAATTCGTCTAGGTTTGAGGTXGACATCGGATCCATGGTACCTC GAGGGCAATGTCTA-30, where X T, 5mC, 5hmC or 5caC. Substrate 2 upper

strand 50-TAGACATTGCCCTCGACGACCCGCCGCCGCGCXGGCCACC CGCACCTAGACGAATTCCG-30 where X C, T, 5mC, 5hmC or 5caC. Substrate

2 lower strand 50-CGGAATTCGTCTAGGTGCGGGTGGCXGGCGCGGCGG CGGGTCGTCGAGGGCAATGTCTA-30 where X 5mC, 5hmC or 5caC.

Substrate 3 upper strand 50-TAGACATTGCCCTCGACGGTGCCCTCXGG GCCGCGCGTCGCGCTCCCTAGACGAATTCCG-30 where X C. Substrate 3

lower strand 50-CGGAATTCGTCTAGGGAGCGCGACGCGCGGCCXGGA GGGCACCGTCGAGGGCAATGTCTA-30 where X 5mC or 5hmC.

Recombinant protein expression. The expression vectors were introduced intoE. coli BL21(DE3) cells by electroporation. Overnight precultures were diluted with fresh prewarmed LB broth medium and grown at 30 C to an OD600 level of0.60.8. Cultures were grown under selective pressure using respective antibiotics at concentrations of either 100 mg l 1 (ampicillin) or 50 mg l 1 (kanamycin and streptomycin) for single plasmid expressions and half the concentration of each antibiotic when co-expressing two plasmids. Protein expression was induced using the following conditions: TET1 (250 mM isopropyl-b-D-thiogalactoside (IPTG),25 C for 3 h), TET1TDG (250 mM IPTG, 25 C for 3 h), TDG (250 mM IPTG, 15 C for 16 h), APE1 (500 mM IPTG, 25 C for 6 h), POLb (500 mM IPTG, 25 C for 3.5 h), and LIG3 and XRCC1 were co-expressed (250 mM IPTG, 25 C for 4 h). Finally, cells were harvested by centrifugation and soluble protein fractionswere extracted by sonication (Bioruptor, Diagenode) or homogenization (Emulsiex C-3, Avestin) in lysis buffer (50 mM Na-phosphate buffer pH 7.5, 300 mM NaCl, 20% glycerol, 0.1% Tween-20, 1 mM dithiothreitol (DTT), 1 mM phenylmethylsulfonyl uoride (PMSF)), if not stated otherwise. Crude lysates were then cleared by centrifugation with 430,000 g at 4 C for 60 min.

Protein purication. For TET1CD purication, the cleared lysate was loaded onto a 1-ml HisTrap FF crude column (GE Healthcare, Germany), bound protein was eluted with 400 mM imidazole and relevant fractions dialysed against CIEX buffer

10 NATURE COMMUNICATIONS | 7:10806 | DOI: 10.1038/ncomms10806 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10806 ARTICLE

(50 mM HEPES pH 7.2, 25 mM NaCl, 20% glycerol, 5 mM DTT and 0.1 mM PMSF). Dialysed fractions were then loaded onto a 1-ml Resource S column (GE Healthcare) and bound protein was eluted with a linear salt gradient of25 mM1 M NaCl and purest fractions nally dialysed against storage buffer(50 mM HEPES pH 7.2, 100 mM NaCl, 20% glycerol and 5 mM DTT), frozen on dry ice and stored at 80 C.

BER proteins were puried as followed; in brief, APE1 and POLb were puried by Ni-NTA afnity and ion exchange chromatography34,50, and TDG was puried by Ni-NTA afnity, heparin afnity and ion exchange chromatography as follows51. Briey, E. coli lysates were prepared by sonication (12 times for 30 s on ice with intermittent chilling) and claried by centrifugation (Sorvall SS34, 18,000 r.p.m.,4 C) at 4 C. The supernatant was applied to a disposable column packed with1.5 ml preequilibrated Ni-NTA agarose (Qiagen) at a ow rate of 15 ml h 1. After washing with 120 ml sonication buffer (50 mM Na-phosphate pH 8.0, 750 mM NaCl, 20% glycerol, 1 mM imidazole, 10 mM b-mercaptoethanol and 1 mM PMSF), bound proteins were eluted with 10 ml sonication buffer containing 500 mM imidazole and dialysed against buffer H50 (50 mM Na-phosphate pH 8.0, 50 mM NaCl, 20% glycerol, 10 mM b-mercaptoethanol, 1 mM PMSF). After loading the dialysed fraction onto a 5-ml HiTrap Heparin HP column (GE Healthcare) at a ow rate of 1 ml min 1 and washing with 10 ml H50, bound protein was eluted with a liner gradient of 50-800 mM NaCl in 50 ml. Purest fractions were pooled, dialysed against buffer Q20 (50 mM Na-phosphate pH 8.5, 20 mM NaCl, 10% glycerol, 10 mM b-mercaptoethanol, 1 mM PMSF) and loaded on a 1-ml HiTrap Q HP at a ow rate of 1 ml min 1. After washing with 10 ml Q20 buffer, bound proteins were eluted with a linear gradient of 20500 mM NaCl in 15 ml. The fractions containing TDG with 498% homogeneity were pooled, dialysed against storage buffer (50 mM Na-phosphate pH 8.0, 50 mM NaCl, 10% glycerol, 10 mM b-mercaptoethanol, 1 mM PMSF), frozen in liquid nitrogen and stored at 80 C. LIG3 and XRCC1 were

puried as a complex by Ni-NTA (HisTrap HP; 50 mM Na-phosphate pH 7.5, 500 mM NaCl, 10 mM ZnCl2, 10% glycerol, 2.5250 mM imidazole, 0.1% Tween-20, 10 mM b-mercaptoethanol, 1% PMSF), glutathione (GSTrap HP; 50 mM Na-phosphate pH 7.5, 300 mM NaCl, 10 mM ZnCl2, 10% glycerol, 0.1% Tween-20, 2 mM DTT, 1% PMSF) and again Ni-NTA afnity chromatography. Highly pure fractions were dialysed against storage buffer (10 mM Tris-HCl pH8, 50 mM NaCl, 10% glycerol), snap frozen and stored at 80 C.

Gel ltration was performed using a Superdex 200 10/300GL column(GE Healthcare) and anKTA Purier 10 (GE Healthcare) according to the manufacturers instructions. Ni-NTA-enriched fractions were prepared as described above. Ni-NTA elution fractions were pooled, concentrated to 8 mg ml 1 using

Amicon Ultra Centrifugal Filters (Millipore) and buffer was changed to gel ltration running buffer (50 mM Na-phosphate pH 7.5, 500 mM NaCl, 20% glycerol). Four milligrams of the enriched fraction was then loaded onto the gel ltration column. Column washing, loading and sampling of the fractions was done according to the manufacturers instructions. Fractions (0.5 ml) were collected and 20 ml of each fraction was used for SDSPAGE and western blot analysis.

To study the interaction of TDG and TET1, Ni-NTA and GST pull-down assays were performed. TDGGST was co-expressed with a TET1 N-terminal fragment (His6TET1-N aa 3011366) or the TET1 catalytic domain (His6TET1CD aa

13672057) in E. coli as described above. Five milligrams of cleared E. coli lysate was then incubated with 25 ml of Glutathione Magnetic Beads (Thermo Scientic)

or Ni-NTA Sepharose beads (Roche) in binding buffer (50 mM Na-phosphate pH7.5, 300 mM NaCl, 20% glycerol, 0.1% Tween-20, 1 mM DTT, 1 mM PMSF) in a total volume of 1 ml at room temperature for 2 h. The beads were rinsed three times with 500 ml binding buffer and bound proteins were analysed by SDSPAGE and western blotting.

Partial purication of TET1CDTDG for activity assays was done via Ni-NTA afnity purication as described above. As catalytic mutants, His6TET1CDDcat

(H1652Y; D1654A) and TDGDcatGST (N151A) were used.

Analytical gel electrophoresis and western blotting. Protein fractions were analysed by standard SDSPAGE followed by Coomassie blue staining or by immunoblotting using chemiluminescence (WesternBright ECL, Advansta) according to the manufacturers protocol. Antibodies were diluted in 5% non-fat dry milk TBS (100 mM Tris-HCl pH 8 and 150 mM NaCl) supplemented with 0.2% Tween-20.

Yeast two-hybrid assay. To conrm the interaction between TET1 and TDG, yeast two-hybrid assay was performed using he Matchmaker yeast-two hybrid system (Clontech). TET1 was divided into four overlapping fragments (TET1-1 aa 1491; TET1-2 aa 397931; TET1-3 aa 8701403; TET1-4 aa 13672057) that were cloned into the BD (pAS2.1 BD FLAG) of the Gal4 protein and TDG was cloned into the AD (pACT2 AD) of the Gal4 protein. The Saccharomyces cerevisiae strain AH109 was co-transformed with 50500 ng of bait and prey plasmids according to the Clontech manual. Interactions were assessed by spotting serial dilutions of cells on selective medium (SC-LEU-TRP-ADE-HIS) supplemented with 2.5 mM 3AT (3-Amino-1,2,4-triazole), a competitive inhibitor of the HIS3 gene product. Cells were incubated at 30 C for 67 days.

Base release assay. The catalytic activity of TET1TDG was monitored by means of a standardized nicking assay31. Briey, the reactions were carried out in a

reaction volume of either 40 ml when using partially puried TET1TDG from Ni-NTA afnity purication fractions or 20 ml when using puried recombinant protein containing TET reaction buffer (50 mM HEPES pH 8, 50 mM NaCl, 1 mM disodium-ketoglutarate, 2 mM ascorbic acid, 75 mM Fe(II) and 1 mM ATP),0.5 pmol of substrate and 10 ml of Ni-NTA pulldown or 2 pmol puried TET1CD

and 1 pmol puried TDG (preincubated together on ice for 5 min), respectively. Reactions were incubated at 37 C for 1 h and stopped by addition of 1 M NaOH to a nal concentration of 100 mM and heating for 10 min at 99 C. After ethanol precipitation at 20 C overnight, the products were separated in a 15%

denaturing polyacrylamide gel and labelled DNA was detected using the red or blue uorescence mode of the Typhoon 9400 (GE Healthcare) and analysed quantitatively by ImageQuant TL software (v7.0, GE Healthcare).

TDG time-course reactions were carried out in 200 ml reaction volume containing nicking buffer (50 mM Tris-HCl pH 8, 1 mM DTT, 0.1 mg ml 1 BSA and 1 mM EDTA), 5 pmol of labelled substrate DNA and 5 pmol of puried TDG.

After the indicated times of incubation at 37 C, 20 ml aliquots were withdrawn and the reactions were stopped by the addition of 1 M NaOH to an end concentration of 100 mM and heating for 10 min at 99 C. Reaction products were analysed by denaturing PAGE and analysed as described above.

In vitro methylation and oxidation of plasmid DNA and slot blot analysis.

In vitro methylation of pUC19 plasmid DNA was performed using M.SssI CpG methyltransferase (New England Biolabs) according to the manufacturers instructions.

For the in vitro oxidation, 200 ng of methylated plasmid was incubated with 500 ng puried His6TET1CD from E. coli (see above). The reaction was carried out in TET reaction buffer (50 mM HEPES pH 8, 50 mM NaCl, 1 mM disodium-ketoglutarate, 2 mM ascorbic acid, 75 mM Fe(II) and 1 mM ATP) and incubated at 37 C for 1 h. Reaction was stopped with the addition of NaOH and EDTA to a nal concentration of 400 and 10 mM, respectively, and heating at 99 C for 10 min. The denatured DNA was blotted using the Bio-Rad slot blot system according to the manufacturers instruction. Hybond-N nylon membranes

(Amersham) were ultraviolet cross-linked, blocked with 5% milk and immunostaining against 5mC, 5hmC, 5fC and 5caC, and was performed using chemiluminescence (WesternBright ECL, Advansta) according to the manufacturers protocol. Antibodies were diluted in 5% non-fat dry milk TBS (100 mM Tris-HCl pH 8 and 150 mM NaCl) supplemented with 0.2% Tween-20.

LCMS/MS analysis. Plasmid DNA samples were enzymatically hydrolysed to deoxyribonucleosides in a two step reaction. First DNA was incubated at 45 C for 40 min in 10 mM ammonium acetate buffer pH 5.3 containing 5 mM magnesium chloride and 0.2 U nuclease P1 from Penicillium citrinum (Sigma, N8630). The samples were then buffered in ammonium bicarbonate to a nal concentration of 100 mM and incubated at 37 C for 30 min with 0.0002 U phosphodiesterase I from Crotalus adamanteus venom (Sigma, P3243) and 0.3 U alkaline phosphatase fromE. coli (Sigma, P5931). The reactions were stopped and contaminants, which could potentially clog the HPLC column, were precipitated by adding three volume equivalents of ice-cold acetonitrile and centrifugation at 16,000 g for 30 min. The supernatants were collected in new tubes and vacuum centrifuged at room temperature until dry. Salt residues, originating from buffers, were partially evaporated by re-dissolving the samples in 100 ml of water and vacuum drying one more time. The standards for 5-me(dC), 5-hm(dC), 5-ca(dC) and 5-f(dC) were prepared to contain the same amount of salts as the samples and followed the same desalting procedure. The samples were then nally dissolved in 50 ml of water for

LCMS/MS analysis of 5-me(dC), 5-hm(dC), 5-ca(dC) and 5-f(dC). For quantication of unmodied nucleosides (dA, dC, dG and dT), samples were diluted 1:10 with water. For some of the samples, 1:10 dilution was also used during quantication of 5-me(dC). Quantication was performed with the use of an LC-20AD HPLC system (Shimadzu) coupled to an API 5000 triple quadrupole (ABSciex) operating in positive electrospray ionization mode. The chromatographic separation was performed at 40 C with the use of an Ascentis Express C18 2.7-mm 150 2.1 mm i.d. column protected with an Ascentis Express Cartridge Guard

Column (Supelco Analytical) with an Exp Titanium Hybrid Ferrule (Optimize Technologies Inc.). The mobile phase consisted of A (water and 0.1% formic acid) and B (methanol and 0.1% formic acid) solutions. The following conditions were employed during chromatography: for unmodied nucleosides, 0.13 ml min 1 ow, starting at 10% B for 0.1 min, ramping to 60% B over 2.4 min and re-equilibrating with 10% B for 4.5 min; for 5-me(dC), 5-hm(dC), 5-ca(dC) and 5-f(dC), 0.14 m l min 1 ow, starting at 5% B for 0.5 min, ramping to 45% B over 8 min and re-equilibrating with 5% B for 5 min. For mass spectrometry detection, the multiple reaction monitoring was implemented using the following mass transitions: 252.2/136.1 (dA), 228.2/112.1 (dC), 268.2/152.1 (dG), 243.2/127.0 (dT), 242.1/126.0 [5-me(dC)], 258.1/142.0 [5-hm(dC)], 256.1/140.0 [5-f(dC)] and 272.1/ 156.0 [5-ca(dC)].

DSB assay. DSB assays were carried out in 20 ml reaction volumes containing incision buffer (50 mM HEPES pH 8, 70 mM KCl, 7 mM MgCl2, 500 mg ml 1 BSA and 1 mM DTT), 0.5 pmol of labelled substrate, 1 pmol APE1 and 1 pmol TDG.

After incubation at 37 C for 30 min, proteinase K was added to a nal

NATURE COMMUNICATIONS | 7:10806 | DOI: 10.1038/ncomms10806 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 11

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10806

concentration of 100 mg ml 1 and the reaction was incubated at 37 C for 30 min. Samples were then separated on 8% native polyacrylamide gels, and detected and quantied.

BER reconstitution. The BER reconstitution reaction was carried out stepwise to analyse individual stages of the process. The reaction mixture containing 1 pmol labelled 60 or 59 bp DNA, 5 pmol His6-TET1CD and 2 pmol His6TDG were incubated at 37 C for 30 min in TET reaction buffer (50 mM HEPES pH 8, 50 mM NaCl, 1 mM disodium-ketoglutarate, 2 mM ascorbic acid, 75 mM Fe(II) and 1 mM ATP), to generate an AP-site. The reaction mixture was then supplemented with 70 mM KCl, 7 mM MgCl2, 200 mM dCTP or dNTP, 2 mM ATP, 500 mg ml 1

BSA, 1 mM DTT and 10 pmol APE1 and incubated at 37 C for 5 min.

DNA polb (0.5 pmol) was then added and the reaction mixture incubated for a further 5 min. Finally, 2 pmol XRCC1LIG3 complex was added for a 10-min incubation. Reactions were terminated by the addition of stop buffer (50 mM Tris-Cl pH 8, 0.5% SDS and 100 mM NaBH4) and incubation on ice for 20 min.

The reaction products were analysed by denaturing PAGE and analysed as described above.

For the analysis of the endproduct with HpaII or MscI endonuclease digest, the reconstitution reaction was carried out by adding all the factors at the same time and incubation at 37 C for 1 h followed by ethanol precipitation of the labelled DNA at 20 C overnight. The recovered DNA was then treated with a total of

5 U HpaII or MscI endonuclease (New England Biolabs) at 37 C for 60 min, fragments were separated in 8% native polyacrylamide gels and detected as described above.

References

1. Bird, A. DNA methylation patterns and epigenetic memory. Genes Dev. 16, 621 (2002).

2. Jones, P. A. Functions of DNA methylation: islands, start sites, gene bodies and beyond. Nat. Rev. Genet. 13, 484492 (2012).

3. Guo, H. et al. The DNA methylation landscape of human early embryos. Nature 511, 606610 (2014).

4. Smith, Z. D. et al. DNA methylation dynamics of the human preimplantation embryo. Nature 511, 611615 (2014).

5. Oswald, J. et al. Active demethylation of the paternal genome in the mouse zygote. Curr. Biol. 10, 475478 (2000).

6. Seisenberger, S. et al. The dynamics of genome-wide DNA methylation reprogramming in mouse primordial germ cells. Mol. Cell 48, 849862 (2012).

7. Metivier, R. et al. Cyclical DNA methylation of a transcriptionally active promoter. Nature 452, 4550 (2008).

8. Kangaspeska, S. et al. Transient cyclical methylation of promoter DNA. Nature 452, 112115 (2008).

9. Kim, M. S. et al. DNA demethylation in hormone-induced transcriptional derepression. Nature 461, 10071012 (2009).

10. Schar, P. & Fritsch, O. DNA repair and the control of DNA methylation. Prog. Drug Res. 67, 5168 (2010).

11. Pastor, W. A., Aravind, L. & Rao, A. TETonic shift: biological roles of TET proteins in DNA demethylation and transcription. Nat. Rev. Mol. Cell Biol. 14, 341356 (2013).

12. He, Y. F. et al. Tet-mediated formation of 5-carboxylcytosine and its excision by TDG in mammalian DNA. Science 333, 13031307 (2011).

13. Ito, S. et al. Tet proteins can convert 5-methylcytosine to 5-formylcytosine and 5-carboxylcytosine. Science 333, 13001303 (2011).

14. Inoue, A., Shen, L., Dai, Q., He, C. & Zhang, Y. Generation and replication-dependent dilution of 5fC and 5caC during mouse preimplantation development. Cell Res. 21, 16701676 (2011).

15. Huang, Y. et al. Distinct roles of the methylcytosine oxidases Tet1 and Tet2 in mouse embryonic stem cells. Proc. Natl Acad. Sci. USA 111, 13611366 (2014).

16. Maiti, A. & Drohat, A. C. Thymine DNA glycosylase can rapidly excise 5-formylcytosine and 5-carboxylcytosine: Potential implications for active demethylation of CpG sites. J. Biol. Chem. 286, 3533435338 (2011).

17. Kohli, R. M. & Zhang, Y. TET enzymes, TDG and the dynamics of DNA demethylation. Nature 502, 472479 (2013).

18. Jacobs, A. L. & Schar, P. DNA glycosylases: in DNA repair and beyond. Chromosoma 121, 120 (2011).

19. Wossidlo, M. et al. 5-Hydroxymethylcytosine in the mammalian zygote is linked with epigenetic reprogramming. Nat. Commun. 2, 241 (2011).

20. Dawlaty, M. M. et al. Tet1 is dispensable for maintaining pluripotency and its loss is compatible with embryonic and postnatal development. Cell Stem Cell 9, 166175 (2011).

21. Dawlaty, M. M. et al. Combined deciency of tet1 and tet2 causes epigenetic abnormalities but is compatible with postnatal development. Dev. Cell 24, 310323 (2013).

22. Gu, T. P. et al. The role of Tet3 DNA dioxygenase in epigenetic reprogramming by oocytes. Nature 477, 606610 (2011).

23. Cortazar, D. et al. Embryonic lethal phenotype reveals a function of TDG in maintaining epigenetic stability. Nature 470, 419423 (2011).

24. Cortellino, S. et al. Thymine DNA glycosylase is essential for active DNA demethylation by linked deamination-base excision repair. Cell 146, 6779 (2011).

25. Saito, Y. et al. Embryonic lethality in mice lacking mismatch-specic thymine DNA glycosylase is partially prevented by DOPS, a precursor of noradrenaline. Tohoku J. Exp. Med. 226, 7583 (2011).

26. Dawlaty, M. M. et al. Loss of Tet enzymes compromises proper differentiation of embryonic stem cells. Dev. Cell 29, 102111 (2014).

27. Shen, L. et al. Genome-wide analysis reveals TET- and TDG-dependent 5-methylcytosine oxidation dynamics. Cell 153, 692706 (2013).

28. Song, C. X. et al. Genome-wide proling of 5-formylcytosine reveals its roles in epigenetic priming. Cell 153, 678691 (2013).

29. Raiber, E. A. et al. Genome-wide distribution of 5-formylcytosine in embryonic stem cells is associated with transcription and depends on thymine DNA glycosylase. Genome Biol. 13, R69 (2012).

30. Dianov, G. L. & Hubscher, U. Mammalian base excision repair: the forgotten archangel. Nucleic Acids Res. 41, 34833490 (2013).

31. Hardeland, U., Bentele, M., Jiricny, J. & Schar, P. Separating substrate recognition from base hydrolysis in human thymine DNA glycosylase by mutational analysis. J. Biol. Chem. 275, 3344933456 (2000).

32. Hu, L. et al. Crystal structure of TET2-DNA complex: insight into TET-mediated 5mC oxidation. Cell 155, 15451555 (2013).

33. Kubota, Y. et al. Reconstitution of DNA base excision-repair with puried human proteins: interaction between DNA polymerase beta and the XRCC1 protein. EMBO J. 15, 66626670 (1996).

34. Hardeland, U., Bentele, M., Jiricny, J. & Schar, P. The versatile thymine DNA-glycosylase: a comparative characterization of the human, Drosophila and ssion yeast orthologs. Nucleic Acids Res. 31, 22612271 (2003).

35. Waters, T. R., Gallinari, P., Jiricny, J. & Swann, P. F. Human thymine DNA glycosylase binds to apurinic sites in DNA but is displaced by human apurinic endonuclease 1. J. Biol. Chem. 274, 6774 (1999).

36. Steinacher, R. & Schar, P. Functionality of human thymine DNA glycosylase requires SUMO-regulated changes in protein conformation. Curr. Biol. 15, 616623 (2005).

37. Hardeland, U., Steinacher, R., Jiricny, J. & Schar, P. Modication of the human thymine-DNA glycosylase by ubiquitin-like proteins facilitates enzymatic turnover. EMBO J. 21, 14561464 (2002).

38. Shen, J. C., Rideout, 3rd W. M. & Jones, P. A. The rate of hydrolytic deamination of 5-methylcytosine in double-stranded DNA. Nucleic Acids Res. 22, 972976 (1994).

39. Rai, K. et al. DNA demethylation in zebrash involves the coupling of a deaminase, a glycosylase, and gadd45. Cell 135, 12011212 (2008).

40. Morgan, H. D., Dean, W., Coker, H. A., Reik, W. & Petersen-Mahrt, S. K. Activation-induced cytidine deaminase deaminates 5-methylcytosine in DNA and is expressed in pluripotent tissues: implications for epigenetic reprogramming. J. Biol. Chem. 279, 5235352360 (2004).

41. Muller, U., Bauer, C., Siegl, M., Rottach, A. & Leonhardt, H. TET-mediated oxidation of methylcytosine causes TDG or NEIL glycosylase dependent gene reactivation. Nucleic Acids Res. 42, 85928604 (2014).

42. Pfaffeneder, T. et al. The discovery of 5-formylcytosine in embryonic stem cell DNA. Angew Chem. Int. Ed. Engl. 50, 70087012 (2011).

43. Hashimoto, H. et al. Structure of a Naegleria Tet-like dioxygenase in complex with 5-methylcytosine DNA. Nature 506, 391395 (2014).

44. Li, Z. et al. Gadd45a promotes DNA demethylation through TDG. Nucleic Acids Res. 43, 39863997 (2015).

45. Niehrs, C. & Schafer, A. Active DNA demethylation by Gadd45 and DNA repair. Trends Cell Biol. 22, 220227 (2012).

46. Hajkova, P. et al. Genome-wide reprogramming in the mouse germ line entails the base excision repair pathway. Science 329, 7882 (2010).

47. Cohen, N. M., Kenigsberg, E. & Tanay, A. Primate CpG islands are maintained by heterogeneous evolutionary regimes involving minimal selection. Cell 145, 773786 (2011).

48. Hashimoto, H., Zhang, X. & Cheng, X. Activity and crystal structure of human thymine DNA glycosylase mutant N140A with 5-carboxylcytosine DNA at low pH. DNA Repair (Amst). 12, 535540 (2013).

49. Zhang, L. et al. Thymine DNA glycosylase specically recognizes

5-carboxylcytosine-modied DNA. Nat. Chem. Biol. 8, 328330 (2012).

50. El-Andaloussi, N. et al. Arginine methylation regulates DNA polymerase beta. Mol. Cell 22, 5162 (2006).

51. Kunz, C. et al. Base excision by thymine DNA glycosylase mediates DNA-directed cytotoxicity of 5-uorouracil. PLoS Biol. 7, e91 (2009).

Acknowledgements

We thank Professor Geir Slupphaug (PROMEC-NTNU, Trondheim) for providing access to and expertise in LCMS/MS analysis. This study was supported by the Swiss National Science Foundation (SNSF-3100A_138153), and the Liaison Committee between the Central Norway RHA and NTNU (grant: 46040500).

12 NATURE COMMUNICATIONS | 7:10806 | DOI: 10.1038/ncomms10806 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10806 ARTICLE

Author contributions

A.R.W. designed and performed experiments, and wrote the manuscript. C.K. performed the yeast two-hybrid analyses. A.B.R. and A.K. synthesized substrate DNA oligonucleotides. A.K. and C.B.V. performed LCMS/MS analyses. D.S. puried BER proteins and performed biochemical assays. P.S. designed, coordinated and supervised the study, and contributed to the writing of the paper.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Weber, A. R. et al. Biochemical reconstitution of TET1TDG BER-dependent active DNA demethylation reveals a highly coordinated mechanism. Nat. Commun. 7:10806 doi: 10.1038/ncomms10806 (2016).

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

NATURE COMMUNICATIONS | 7:10806 | DOI: 10.1038/ncomms10806 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 13