A 5-methylcytosine DNA glycosylase/lyasedemethylates the retrotransposon Tos17and promotes its transposition in riceHonggui Laa,b, Bo Dinga, Gyan P. Mishrab, Bo Zhoua, Hongmei Yanga, Maria del Rosario Bellizzia, Songbiao Chena,Blake C. Meyersc, Zhaohua Pengd, Jian-Kang Zhub,e,1, and Guo-Liang Wanga,f,1

aDepartment of Plant Pathology, Ohio State University, Columbus, OH 43210; bDepartment of Horticulture and Landscape Architecture, Purdue University,West Lafayette, IN 47907; cDepartment of Plant and Soil Sciences, Delaware Biotechnology Institute, University of Delaware, Newark, DE 19711; dDepartmentof Biochemistry and Molecular Biology, Mississippi State University, MS 39762; eCenter for Plant Stress Genomics Research, King Abdullah University of Scienceand Technology, Thuwal 23955-6900, Kingdom of Saudi Arabia; and fState Laboratory for Biology of Plant Diseases and Insect Pests, Institute of PlantProtection, Chinese Academy of Agricultural Sciences, Beijing 100193, China

Contributed by Jian-Kang Zhu, August 4, 2011 (sent for review June 15, 2011)

DNA 5-methylcytosine (5-meC) is an important epigenetic mark fortranscriptional gene silencing in many eukaryotes. In Arabidopsis,5-meC DNA glycosylase/lyases actively remove 5-meC to counter-act transcriptional gene silencing in a locus-specific manner, andhave been suggested to maintain the expression of transposons.However, it is unclear whether plant DNA demethylases can pro-mote the transposition of transposons. Here we report the func-tional characterization of the DNA glycosylase/lyase DNG701 inrice. DNG701 encodes a large (1,812 amino acid residues) DNAglycosylase domain protein. Recombinant DNG701 protein showed5-meC DNA glycosylase and lyase activities in vitro. Knockout orknockdown of DNG701 in rice plants led to DNA hypermethylationand reduced expression of the retrotransposon Tos17. Tos17 showedless transposition in calli derived from dng701 knockout mutantseeds compared with that in wild-type calli. Overexpression ofDNG701 in both rice calli and transgenic plants substantially reducedDNA methylation levels of Tos17 and enhanced its expression. Theoverexpression also led to more frequent transposition of Tos17 incalli. Our results demonstrate that rice DNG701 is a 5-meC DNAglycosylase/lyase responsible for the demethylation of Tos17 andthis DNA demethylase plays a critical role in promoting Tos17 trans-position in rice calli.

Oryza sativa | incision activity | Tos17 copy number

In eukaryotes, DNA cytosine methylation at carbon 5 of thepyrimidine ring [5-methylcytosine (5-meC)] is an important

epigenetic mark that contributes to gene silencing and playscritical roles in development and genome defense against viru-ses, transposons, and transgenes (1–3). Heavy cytosine methyl-ation usually occurs at heterochromatin and at regions rich intransposons and repetitive DNA (2, 4). Cytosine methylation ofDNA, however, is reversible through demethylation (5, 6). DNAdemethylation can be passive or active, and active DNA deme-thylation is catalyzed by one or more enzymes to remove meth-ylated cytosines and can occur independently of DNA replication(7, 8). So far, several models have been proposed to explainmechanisms of DNA demethylation in animals (7, 8). One of themodels suggests that active DNA demethylation is mediated atleast in part by a base excision repair (BER) pathway where theAID/Apobec family of deaminases convert 5-meC to T followedby G/T mismatch repair through the DNA glycosylase MBD4 orTDG with the involvement of Gadd45a (7–9). Recently, a newstudy suggested that 5-meC hydroxylase TET1 promotes DNAdemethylation in mammalian cells through a process that re-quires the BER pathway (10). Nevertheless, many aspects ofthese models have not been confirmed, and how DNA deme-thylation is carried out in animals remains controversial (7, 8).In contrast to the uncertainty about the mechanism of DNA

demethylation in animals, mechanisms of DNA demethylation in

plants are much clearer and widely accepted. In the model plantArabidopsis, research showed that the 5-meC DNA glycosylase/lyase-mediated BER pathway plays critical roles in DNAdemethylation process (3, 5, 6, 11, 12). There are four 5-meCDNA glycosylase/lyases [i.e., REPRESSOR OF SILENCING 1(ROS1), DEMETER (DME), DEMETER-LIKE2 (DML2) andDEMETER-LIKE3 (DML3)] found in the Arabidopsis genome,and genetic and biochemical analysis revealed that all four pro-teins function as DNA demethylases, and are involved in genomicDNA demethylation (1, 3, 5, 6, 12–14). ROS1 is required for theprevention of hypermethylation and transcriptional silencing of arepetitive reporter gene (1). DME was reported to be necessaryfor endosperm gene imprinting and seed viability (3, 13). BothDML2 and DML3 are required for removing 5-meC from im-properly methylated DNA in target sites (12). Biochemical evi-dence demonstrated that the DNA glycosylase activity of theseproteins removes 5-meC from DNA backbone and then the lyaseactivity cleaves the DNA backbone at the abasic site by successiveβ- and δ-elimination reactions (3, 5, 6, 12, 14). However, com-pared with these well-documented 5-meCDNA glycosylase/lyasesin Arabidopsis, no 5-meC DNA glycosylase/lyase has been wellcharacterized in other plant species.As one of the most important food crops in the world, rice

(Oryza sativa L.) has become a model monocot species for func-tional genomics research. Nipponbare, a completely sequencedjaponica rice cultivar, is now widely used for functional analysis ofgenes. Tos17, a well-characterized rice retrotransposon, is exten-sively used to generate Tos17 insertional mutants in Nipponbare(15, 16). There are two almost identical Tos17 copies, namedTos17chr7 and Tos17chr10 (located on chromosomes 7 and 10, re-spectively) inNipponbare (17).Tos17chr7 transcripts (∼4 kb) solelyaccumulate in calli and Tos17chr10 is inserted into a putative ABC-type transporter gene to form a fusion gene that generates a∼7-kbreadthrough transcript (17). Both Tos17chr7 and Tos17chr10 arehighly methylated and immobilized under normal plant growthconditions (16, 17). However, in calli the transposition of Tos17 isactivated and the copy number increases gradually during pro-longed callus culture (16). Several recent studies showed that theactivation of transcription as well as transposition of Tos17 in calliis associated with DNA hypomethylation (17–19). However, the

Author contributions: H.L., Z.P., J.-K.Z., and G.-L.W. designed research; H.L., B.D., G.P.M.,B.Z., H.Y., M.d.R.B., and S.C. performed research; and H.L., B.C.M., J.-K.Z., and G.-L.W.wrote the paper.

The authors declare no conflict of interest.1To whom correspondence may be addressed. E-mail: jkzhu@purdue.edu or wang.620@osu.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1112704108/-/DCSupplemental.

15498e15503 | PNAS | September 13, 2011 | vol. 108 | no. 37 www.pnas.org/cgi/doi/10.1073/pnas.1112704108

protein that is directly responsible for the active demethylation ofTos17 in the rice genome is as yet unknown.In this report, we identified and characterized a rice DNA

glycosylase/lyase, DNG701 (for DNA Glycosylase 701) (http://www.chromdb.org/). Enzymatic activity assays demonstrated thatDNG701 is an active 5-meC DNA glycosylase/lyase in vitro.Knockout or knockdown and overexpression of DNG701 in ricecalli and in planta showed that DNG701 is critical for Tos17demethylation, expression, and transposition in rice. Our dem-onstration of a host plant factor promoting the transposition ofTos17 suggests that the retrotransposon Tos17 may have a ben-eficial function to the rice genome during evolution.

ResultsDNG701 Encodes a Rice 5-meC DNA Glycosylase/Lyase. Among thefive predicted rice DNA glycosylases (http://www.chromdb.org/),DNG701 and its close paralog DNG704 are most closely relatedto Arabidopsis DML2 and ROS1 (Fig. S1A). As with the otherfour rice DNA glycosylases, DNG701 contains a DNA glyco-sylase domain immediately followed by an EndIII_4Fe-4S do-main, and two putative nuclear localization signals are located atthe N terminus of the protein (Fig. S1B) (20). Similar to the fourArabidopsis and other four rice DNA glycosylases, the DNG701glycosylase domain contains a helix–hairpin–helix motif anda glycine/proline-rich loop with an invariant aspartic acid (GPD)(Fig. S1C). An extremely conserved lysine (K) is located insideone of the highly conserved helices (Fig. S1C).A comparison between the cloned DNG701 cDNA (GenBank

accession no. FJ536320) and its genomic DNA sequence re-vealed that the ORF of DNG701 consists of 19 exons and enc-odes a protein of 1,812 amino acids (Figs. S1B and S2A). RT-PCR analysis indicated that DNG701 was ubiquitously expressedin the three tissues (leaves, stems, and roots) of 3-wk-old Nip-ponbare seedlings (Fig. S2B, Upper) and four tissues (leaves,stems, roots, and panicles) of 3-mo-old flowering Nipponbareplants (Fig. S2B, Lower). The fused GFP-DNG701 protein waspredominantly localized in the nucleus, whereas the GFP con-trol protein was distributed in both the nucleus and cytoplasmof rice cells (Fig. S2C).The purified recombinant maltose-binding protein-DNG701

(MBP-DNG701) fusion protein had nicking activity againstmCG- and mCCGG-containing pUBIGUS plasmids, which wasgenerated separately by CG methyltransferase (M.SssI) and byMspI methyltransferase, but did not have any nicking activityagainst unmethylated pUBIGUS (Fig. S3 A and B, Upper). Thepurified control protein MBP possessed no nicking activityagainst either methylated or unmethylated pUBIGUS (Fig. S3B,Lower). Furthermore, we conducted incision assays against 40-bpdouble-stranded oligonucleotides containing fully methylatedand hemimethylated CmCGG or mCCGG (Fig. S3C) (5). Thedouble-stranded oligonucleotides were produced by annealing32P-labeled single-stranded oligonucleotides to unlabeled com-plementary strands (Fig. S3C). Incubation of MBP-DNG701 witheither CmCGG- or mCCGG-containing oligonucleotides gener-ated one 16-nt and one 29-nt cleavage products correspondingto fully methylated and hemimethylated sites, respectively, sug-gesting that the protein has active incision activity in vitro againstfully methylated and hemimethylated oligonucleotides (Fig. 1 Aand B). In contrast, MBP showed no incision activity on the twotypes of oligonucleotides (Fig. 1A). With an increase in theamount of MBP-DNG701, more 16-nt and 29-nt cleavageproducts were produced from the methylated oligonucleotides(Fig. 1B). Furthermore, four other types of oligonucleotides,containing either fully methylated or hemimethylated CmCGGor mCCGG, were incubated with MBP-DNG701 (Fig. S3C). Theresults again showed that MBP-DNG701 had incision activityagainst the fully methylated or hemimethylated CmCGG-or mCCGG-containing oligonucleotides, but had no incision ac-

tivity against unmethylated oligonucleotides (Fig. 1C). Takentogether, these data showed that DNG701 has active DNA gly-cosylase/lyase activity against 5-meC–containing DNA in vitro.

Knockout or Knockdown of DNG701 Causes DNA Hypermethylation ofTos17 Retrotranspsons in Rice. To further elucidate the functionof DNG701 in rice, we obtained two Tos17 insertion mutantsnamed dng701-1 and dng701-2 (Fig. S2A). RT-PCR analysissuggested that DNG701 is knocked out in both the mutants (Fig.S4, Left). Meanwhile, homozygous DNG701 RNA interference(RNAi) transgenic lines PC5-9 and PC5-10 were selected as wellfor detailed analysis because the expression of DNG701 wassubstantially down-regulated although that of DNG704 was notaffected (Fig. S4, Right). DNA gel-blot analysis revealed thatthere were no apparent DNA methylation changes in the cen-tromeric repeat DNA, 45S rDNA, or Ty1 retrotransposons inboth the mutants and RNAi lines, indicating that knockout orknockdown of DNG701 does not cause global DNA hyper-methylation in these regions (Fig. S5A). However, substantialDNA hypermethylation was observed in the retrotransposonTos17 (Fig. 2A and Fig. S6A). These results indicated that loss of

B

C

A Cm

CGG oligom

CCGG oligo

40 nt

29 nt

16 nt

40 nt

16 nt

MspI MBP-DNG701

Cm

CGG oligom

CCGG oligo

40 nt

29 nt

16 nt

Fig. 1. Enzymatic activity of DNG701 on 5-meC-containing DNA substratesin vitro. (A) Incision activity assays of MBP-DNG701 and MBP on CmCGG- andmCCGG-containing radiolabeled oligonucleotides. One microgram of MBPand MBP-DNG701 proteins and 10 U of MspI were used in this assay. (B)Incision activity assays of MBP-DNG701 on CmCGG- and mCCGG-containingoligonucleotides. In the second and third lanes in each panel, 0.5 and 1 μg ofproteins were added, respectively. (C) Incision activity assays of MBP-DNG701 on hemimethylated or fully methylated CmCGG- or mCCGG-con-taining oligonucleotides. One microgram of MBP-DNG701 protein and 10 Uof MspI were used in this assay. CK, no protein added. CmCGG oligo andmCCGG oligo refer to fully methylated, as well as hemimethylated, oligo-nucleotides at the internal and external cytosines in CCGG sequence context,respectively. “Hemi” and “Full” refer to hemimethylated and fully methyl-ated oligonucleotides, respectively. CCGG, unmethylated oligonucleotides.The filled triangles above the panels indicate increasing amounts of proteins.

La et al. PNAS | September 13, 2011 | vol. 108 | no. 37 | 15499

PLANTBIOLO

GY

function of DNG701 leads to DNA hypermethylation in a locus-specific manner. Bisulfite sequencing results further verified thatmethylation levels at both CG and CHG (H is A, T, or C) sites ofTos17 were increased in the mutants and RNAi lines (Fig. 2Band Fig. S6B).We then examined expression levels of Tos17 in the mutants

and RNAi lines and found that the expression of Tos17chr10

transcripts (∼7 kb) was substantially reduced (Fig. 2C). TheTos17chr7 transcripts (∼4 kb) were not detected in either thewild-type Nipponbare or the two mutants and RNAi lines (Fig.2C). To understand if the suppression of Tos17chr10 expression iscaused by hypermethylation of its promoter, we examined themethylation level of the promoter region of ∼1-kb upstream ofthe translation start site by bisulfite sequencing (Fig. S6A). A202-bp fragment from −445 to −244 was heavily methylated in allof the genotypes and only the CHH methylation levels wereslightly increased in the mutants and RNAi lines, suggesting thatmethylation in this region may not account for the reductionof Tos17chr10 expression (Fig. S5B). Because the LTR region isknown as an enhancer-promoter of LTR retrotransposons andis usually a target of transcriptionally repressive methylation(17, 21), we determined the methylation levels of 5′ LTR regionsof Tos17chr10 and Tos17chr7 and observed much higher DNAmethylation levels in the mutants and RNAi lines than in thecontrols (Fig. 2 D and E). Taken together, these data indicatethat knockout or knockdown of DNG701 leads to DNA hyper-

methylation at 5′ LTR as well as coding regions of Tos17 andtranscriptional repression of Tos17chr10, suggesting that DNG701is an important factor for protecting Tos17 from hypermeth-ylation and silencing. In addition to Tos17, we also found thatanother retrotransposon (LOC_Os07g31690) was hypermeth-ylated in its 3′ terminal region in both the dng701 mutants, fur-ther supporting that DNG701 is an active demethylase/lyase thatremoves 5-meC at some loci in the rice genome (Fig. S5C).Interestingly, seeds harvested from mature panicles of both

mutants fell into two categories: normal seeds and wrinkled seeds(13.66% wrinkled seeds for dng701-1, 10.51% for dng701-2, and0.79% for wild-type Nipponbare). The observation suggests thatDNG701 is involved in seed development of rice (Fig. S7).

Overexpression of DNG701 Leads to DNA Hypomethylation in RiceCalli and Transgenic Plants. To further understand the functionof DNG701 in rice, we overexpressed DNG701 in Nipponbareunder the control of the maize ubiquitin promoter. Two in-dependently transformed callus lines, OX-1 and OX-2, wereobtained after a 2-mo selection on hygromycin-containing media.Simultaneously, two other independent callus lines, OX-CK-1and OX-CK-2, which were transformed with the empty vectoralone, were obtained and used as controls. Next, each callus linewas subdivided into a few smaller callus clones and continuouslycultured for another 3 mo. RNA gel-blot analysis demonstratedthat the expression levels of Tos17chr10 in the OX-1 and OX-2

0

20

40

60

80

100

CG CHG CHHTos17 chr10 5' LTR

Cyt

osin

e m

ethy

latio

n(%

) NPB WTWT CTR-1dng701-1WT CTR-2dng701-2WT CTR-3PC5-9-1PC5-10-1

0

20

40

60

80

100

CG CHG CHHTos17 chr7 5' LTR

Cyt

osin

e m

ethy

latio

n(%

) NPB WTWT CTR-1dng701-1WT CTR-2dng701-2WT CTR-3PC5-9-1PC5-10-1

E

0102030405060708090100

CG CHG CHHTos17

Cyt

osin

e m

ethy

latio

n(%

) WT CTR-1dng701-1dng701-2WT CTR-3PC5-9-1PC5-10-1

C

B

D

DNG7014kb

rRNA

Tos17 7kb

6kb

A M+BH+B

Actin

H+B M+B M+BH+B

Tos17 Tos17 Tos17

Fig. 2. Impact of knockout or knockdown of DNG701 on the methylation status of Tos17. (A) Methylation status of Tos17 retrotransposons in wild-typecontrols, mutants, and RNAi lines. A 3.3-kb Tos17 probe (Probe 1) (Fig. S6B) was used for hybridization. The same blots were reprobed with Actin. WT CTR-1and WT CTR-2 are sibling wild-type plants segregated from heterozygous DNG701/dng701-1 and DNG701/dng701-2 plants, respectively. WT CTR-3 is a seg-regated wild-type plant from a hemizygous DNG701 RNAi transgenic line PC5-9. H+B, HpaII+BstXI; M+B, MspI+BstXI. (B) Bisulfite sequencing analysis ofmethylation status of Tos17. Tos17 region d (Fig. S6B) was subjected to bisulfite sequencing. (C) Reduced expression of Tos17chr10 in the mutants and RNAilines. A 1.1-kb Tos17 probe (Probe 2) (Fig. S6B) was used for Tos17 hybridization. NPB WT, Nipponbare wild-type. (D and E) Bisulfite sequencing analysis ofmethylation levels of Tos17 5′ LTR regions. 5′ LTR regions of Tos17chr10 (region b in Fig. S6A) (D) and Tos17chr7 (region c in Fig. S6A) (E) were assayed.

15500 | www.pnas.org/cgi/doi/10.1073/pnas.1112704108 La et al.

callus lines were increased when DNG701 was overexpressed,and low levels of Tos17chr7 transcripts were detected in all gen-otypes (Fig. 3A). DNA gel-blot analysis revealed that both theOX-1 and OX-2 callus lines showed hypomethylation in Tos17relative to the control callus line, which presumably accountsfor the enhanced expression of Tos17chr10 (Fig. 3B). This hypo-methylation of Tos17 was further confirmed by bisulfite se-quencing analysis (Fig. S8). Interestingly, CG methylation levelof the 45S rDNA also declined in the OX-1 and OX-2 callus lines(Fig. 3C). Furthermore, methylation levels of the centromericrepeat DNA and Ty1 retrotransposons were also slightly reducedin the overexpression lines (Fig. 3C).DNG701-overexpressing transgenic plants and transgenic con-

trols regenerated from the 2-mo-old transformed calli were usedto study the impact ofDNG701 overexpression on Tos17 in planta.As shown in Fig. 4A, the expression levels of Tos17chr10 weremarkedly increased with the overexpression of DNG701 in theOX-1 and OX-2 transgenic lines. DNA gel-blot analysis and bi-sulfite sequencing results revealed reduced methylation levelsat its promoter region and gene body in the overexpressiontransgenic plants (Fig. 4 B–D and Fig. S6A). The transcripts ofTos17chr7 were not detected in either overexpression line despitethe hypomethylation of Tos17 (Fig. 4A). Bisulfite sequencing re-sults of the 5′ LTR regions of both Tos17chr10 and Tos17chr7 dem-onstrated that CG methylation levels in Tos17chr10 and Tos17chr7,and CHG and CHHmethylation levels in Tos17chr7 were reducedin the overexpression plants (Fig. 4 E and F). Like the over-expression calli, the overexpression plants also showed a remark-ably decreased CG methylation in 45S rDNA (Fig. S9). Thesedata suggest that overexpression of DNG701 has extensive effectson reducing DNA methylation of the rice genome.

DNG701 Is a Critical Factor That Promotes Tos17 Transposition in Calli.Because Tos17 is hypomethylated in the OX-1 and OX-2 calluslines, we wondered if transposition of Tos17 in the two lines wasaffected. DNA gel-blot analysis was performed to examine Tos17copy number in 6-mo-cultured (since callus induction) callusclones derived from the OX-1 and OX-2 callus lines and thecontrol lines. As shown in Fig. 3D (Left), an apparent increase inthe copy number of Tos17 was observed in the callus clonesderived from the OX-1 and OX-2 callus lines in which an averageof 6.8 and 6.3 Tos17 copies were present, respectively. In con-trast, there were on average 3, 2, and 3 Tos17 copies found in thecallus clones from Nipponbare, OX-CK-1 and OX-CK-2, re-spectively (Fig. 3D, Left).Furthermore, independent calli fromNipponbare and the dng701-

2 mutant seeds were induced and cultured for 10 mo. Eightindependent calli from each material were subjected to DNA gel-blot analysis to investigate the change in Tos17 copy number. Inparallel, leaf samples from both genotypes were used as thecontrols to assess native copy number of Tos17. The resultsrevealed that calli derived from dng701-2 mutant seeds gainedonly one additional Tos17 copy relative to the leaf control (Fig.3D, Right). In contrast, calli derived from Nipponbare gained 26additional Tos17 copies compared with the leaf control (Fig. 3D,Right). The increased copy number in these samples was not at-tributable to incomplete digestion because only a single band wasobserved when the blot was reprobed with Tubulin (Fig. 3D,Right). Taken together, these results indicate that DNG701 isnecessary and also a limiting factor for high transposition activityof Tos17 in rice calli.

DiscussionOur work demonstrated that DNG701 is an active rice 5-meCDNA glycosylase/lyase, and it is able to process fully methylated

A

Tos17

H+B M+B

4kb3kb2kb1.5kb1kb

0.5kb

Actin

DNG701

Tos17

rRNA

7kb4kb

6kb

Tubulin

XbaI

Tos17 4kb

6kb8kb

3kb

5kb

6kb8kb

10kb12kb

XbaI

12kb

4kb

3kb

5kb

8kb10kb

6kb

6kb8kb

HpaII MspI

45S rDNA

HpaII MspI

Centromeric repeat DNA

HpaII MspI

Ty1

B C

D

Fig. 3. Effects of DNG701 overexpression on the DNA methylation and transposition of Tos17 in rice calli. (A) The transcript levels of Tos17 in DNG701-overexpressing calli. Probe 2 was used for Tos17 hybridization. NPB WT, Nipponbare wild-type calli. OX-CK-1 and OX-CK-2, independent calli transformedwith empty vector. OX-1 and OX-2, independent calli transformed with DNG701-overexpressing construct. (B) Methylation status of Tos17 in the DNG701-overexpressing calli revealed by DNA gel-blot analysis. Probe 1 was used for Tos17 hybridization. H+B, HpaII+BstXI; M+B, MspI+BstXI. (C) DNA gel-blot analysisof methylation status of centromeric repeat DNA, 45S rDNA and Ty1 retrotransposons in the DNG701-overexpressing calli. (D) Transposition of Tos17 in theDNG701-overexpressing calli (Left) and mutant calli (Right). The open triangle denotes a newly transposed Tos17 copy in the dng701-2 calli. Probe 2 was usedfor hybridization.

La et al. PNAS | September 13, 2011 | vol. 108 | no. 37 | 15501

PLANTBIOLO

GY

or hemimethylated external or internal 5-meC efficiently (Fig. 1A–C). Like the Arabidopsis ros1 mutant (4), the rice dng701knockout mutants show hypermethylation at some transposons,indicating that loss of function of DNG701 leads to DNAhypermethylation in a locus-specific manner (Fig. 2A and Fig.S5C). We did not observe apparent methylation changes in thecentromeric repeat DNA, 45S rDNA, or Ty1 retrotransposonsbetween dng701 mutants and wild-type controls (Fig. S5A). It islikely that these regions have already been hypermethylated inthe wild-type control genomes, and any methylation-enhancingeffect of dng701 on these regions is difficult to detect usingmethylation-sensitive restriction enzymes. There is no reportabout the impact of overexpression of ROS1 or DML2 on globalDNA methylation levels at certain repetitive sequences, althoughROS1 was reported to be localized in nucleoplasm and nucleolusthat contain many copies of 45S rDNA units (5, 12, 22). Over-expression of DNG701 in rice calli and plants led to markedhypomethylation of 45S rDNA and slight hypomethylation oncentromeric repeat DNA and Ty1 retrotransposons, suggestingthat DNG701 is targeted to these regions, and overexpression ofDNG701 presumably has significant dose effect to excessivelydemethylate them (Fig. 3C and Fig. S9). Neither the Arabidopsisros1 nor the dml2 mutant produces wrinkled seeds (1, 12, 14),whereas the rice dng701 mutants produce a substantial pro-portion of wrinkled seeds (∼10–13% for mutants and ∼0.8% forNipponbare) (Fig. S7). Recently, two studies revealed that bothArabidopsis and rice endosperms are hypomethylated in allsequence contexts, and CG methylation levels are partially re-stored in both gene bodies and repeats in dme endosperm, sug-gesting the participation of DME in endosperm demethylation(23, 24). It is likely that DNG701 is also involved in endospermdemethylation in rice and, unlike its Arabidopsis coun-terparts (DML2 and ROS1), targets developmentally importantgenes. Compared with Arabidopsis, the rice genome containsa substantially higher amount of transposons and other repetitive

elements (25, 26). Therefore, it is possible that more genes inrice are subject to regulatory control by repetitive elements,and thus more genes are likely targeted for demethylation byDNG701. It will be of interest to determine the developmentallyimportant target genes of DNG701 in the future.The relationships among DNA methylation, transcription, and

transposition of transposable elements in plant kingdom havebeen explored extensively (19, 21, 27–29). In Arabidopsis, it hasbeen reported that DNA methylation is involved in controllingmobilization of transposons (27–29). In the ddm1mutant and theself-pollinated progeny, silent CACTA elements and severaltransposons like ATGP1 and ATGP2 are transcriptionally andtranspositionally activated because of DNA hypomethylation(28, 29). In the cmt3 met1 double mutant, transcription and high-frequency transposition of the CACTA elements were detected,suggesting that CG and non-CG methylation functions to im-mobilize transposons (27). Likewise, the reduced DNA methyl-ation levels in rice correlate with activation of transcription andtransposition of Tos17 (17, 18). Other research found that re-duced histone H3K9 dimethylation decreases DNA methylationof Tos17, leading to its active transcription and transposition inSDG714 RNAi transformants (19); however, the enzymes di-rectly responsible for demethylation of Tos17 have remainedunclear. Our work suggests that DNG701 is responsible for ac-tive DNA demethylation of Tos17. In dng701 mutants, the nor-mally expressed Tos17chr10 is silenced, which is accompanied byelevated DNA methylation levels at the 5′ LTR region and genebody (Fig. 2 A–E). In dng701 mutant calli, Tos17 transpositionactivity is much reduced (Fig. 3D, Right). In contrast, over-expression of DNG701 in rice calli and plants not only decreasesthe DNA methylation levels of Tos17 and up-regulatesTos17chr10 expression, but also promotes transposition of Tos17in rice calli (Figs. 3 A, B, and D, Left, and 4). Interestingly, theautonomous expression of the Tos17chr7 is not affected by itsDNA methylation levels, suggesting that DNA hypomethylation

A

rRNA

DNG701

Tos17 7kb4kb6kb

B HpaII

Tos17chr10 promoter

0

10

20

30

40

50

60

70

80

CG CHG CHHTos17

Cyt

osin

e m

ethy

latio

n(%

)

OX-CK-1

OX-1-27OX-1-31

OX-2-35OX-2-43

C

02

468

1012

1416

CG CHG CHHTos17 chr10 5' LTR

Cyt

osin

em

ethy

latio

n(%

) NPB WTOX-CK-1WT CTR-4OX-1-27WT CTR-5OX-2-35

D

0

5

10

15

20

25

30

CG CHG CHHTos17 chr7 5' LTR

Cyt

osin

em

ethy

latio

n(%

) NPB WTOX-CK-1WT CTR-4OX-1-27WT CTR-5OX-2-35

FE

Actin

M+B

Tos17

H+B M+B H+B

Fig. 4. Effects of DNG701 overexpression on the DNA methylation of To17 in transgenic plants. (A) The transcript levels of Tos17 in DNG701-overexpressingplants. OX-1–8, OX-1–27, and OX-1–31 are transgenic plants regenerated from 2-mo-cultured OX-1 calli; OX-2–34, OX-2–35, and OX-2–43 are transgenic plantsregenerated from 2-mo-cultured OX-2 calli. WT CTR-4 and WT CTR-5, segregated wild-type plants from hemizygous transgenic plants OX-1–27 and OX-2–35,respectively. Probe 2 was used for Tos17 hybridization. (B) Reduced methylation levels of the Tos17chr10 promoter in the DNG701-overexpressing transgenicplants. A 3.8-kb probe (Probe 3) (Fig. S6A) was used for hybridization. (C and D) Reduction of methylation levels of Tos17 in the DNG701-overexpressing plantsas revealed by bisulfite sequencing (region d) (C) and DNA gel-blot analysis (D). Probe 1 was used for hybridization. H+B, HpaII+BstXI; M+B, MspI+BstXI. (E andF) DNA methylation levels of Tos17chr10 5′ LTR (region b) (E) and Tos17chr7 5′ LTR (region c) (F) determined by bisulfite sequencing.

15502 | www.pnas.org/cgi/doi/10.1073/pnas.1112704108 La et al.

is not a sole determinant for the expression of Tos17chr7, andthere must be other factors to regulate its expression in riceplants (Fig. 3A). We speculate that overexpression of DNG701might generate certain immediate triggers to contribute to thetransposition of Tos17chr7. One possible trigger is the favorablechromatin state/environment. Our results showed that in theoverexpression calli the methylation levels of 45S rDNA wereobviously decreased, and the methylation levels of centromericrepeats and Ty1 transposons were also slightly reduced, in-dicating that overexpression of DNG701 has extensive effects onthe methylation levels of callus genome (Fig. 3C). Therefore, it islikely that the chromatin state/environment of the overexpressioncalli might have been altered because of the overexpression ofDNG701, which may make the chromatin prone to Tos17 in-tegration. Another possible trigger is posttranscriptional regula-tion. Overexpression ofDNG701may enhance translation/reversetranscription of Tos17 transcripts through an unknown post-transcriptional regulation. Our results suggest that DNG701 isrequired for Tos17 demethylation and expression of Tos17chr10.Importantly, DNG701 is required for the high transposition ac-tivity of Tos17 in rice calli. Our results also suggest that DNG701is a limiting factor for Tos17 transposition. The fact that plantshave evolved active DNA demethylases that promote the trans-position of transposons implies that the affected transposons alsohave beneficial effects to the plant species during evolution.

Materials and MethodsPlant Materials. All rice plants used in this study were Oryza sativa L. ssp.japonica cv. Nipponbare. Details are available in SI Materials and Methods.

Cloning of DNG701 cDNA, Subcellular Localization of DNG701, RNA Gel-Blot andRT-PCR Analysis, and DNA Gel-Blot Analysis. See details in SI Materials andMethods and Table S1.

Protein Purification and Enzymatic Activity Assays. The entire DNG701 cDNAwas inserted in-frame into the pMAL-c2× vector (New England Biolabs) to

obtain a malE-DNG701 fusion construct. Expression and purification of theMBP-DNG701 fusion protein and MBP protein were performed as previouslydescribed (5). For the nicking activity assays, 200 ng of each of these plasmidswas added to nicking reaction buffer (10 mM Tris-HCl, pH7.9, 50 mM NaCl,10 mM MgCl2, 1 mM DTT, 0.2 mg/mL BSA) plus purified MBP-DNG701 orMBP. Reactions were carried out as previously described (5). For incisionactivity assays, upper strands of the seven oligonucleotides (Fig. S3C) werelabeled with 32P at the 5′ ends, and then annealed to the correspondinglower strands, as previously described (5). Labeled double-stranded oligo-nucleotides (1 pmol) were incubated with MBP-DNG701 or MBP or MspIendonuclease in the nicking reaction buffer in a total volume of 10 μL at37°C for 2 h. The products were separated on 17% denaturing poly-acrylamide gels containing 8 M urea, and then the gels were exposed to aPhosphorImager screen. See SI Materials and Methods for details.

Genomic Bisulfite Sequencing. Two micrograms of genomic DNA was di-gested with 20 units of appropriate restriction enzymes (New EnglandBiolabs) for 3 h. Next, the digested DNA was purified by ethanol, treatedwith a sodium bisulfite solution and desalted, as previously described(19). Subsequently, 4 μL of each recovered DNA sample was subjected toPCR amplification. The PCR products were cloned into the pGEM-T Easycloning vector (Promega), and 15 to 20 clones were sequenced to calcu-late the percentage of cytosine methylation in each sample. Details are inSI Materials and Methods.

Sequence Alignments and Phylogenetic Analysis. See SI Materials and Methodsfor details.

ACKNOWLEDGMENTS. We thank Drs. Ko Shimamoto and Michael Goodinfor providing the pANDA vector and pGDG vector, respectively; Dr. StevenE. Jacobsen for providing the genomic bisulfite sequencing protocol;Dr. Jiming Jiang for providing the centromeric repeat probe; and the RiceGenome Resource Center in Japan for providing the cDNA clones and Tos17insertion lines. This research is supported by National Institutes of HealthGrant R01GM070795 (to J.-K.Z.), and US Department of Agriculture-CooperativeState Research, Education, and Extension Service Grant 2004-05425 and Na-tional Science Foundation-Plant Genome Research Program Grants 0321437and 0701745 (to G.-L.W.).

1. Gong Z, et al. (2002) ROS1, a repressor of transcriptional gene silencing in Arabi-dopsis, encodes a DNA glycosylase/lyase. Cell 111:803e814.

2. Lister R, et al. (2008) Highly integrated single-base resolution maps of the epigenomein Arabidopsis. Cell 133:523e536.

3. Gehring M, et al. (2006) DEMETER DNA glycosylase establishes MEDEA polycombgene self-imprinting by allele-specific demethylation. Cell 124:495e506.

4. Zhu J, Kapoor A, Sridhar VV, Agius F, Zhu JK (2007) The DNA glycosylase/lyase ROS1functions in pruning DNA methylation patterns in Arabidopsis. Curr Biol 17:54e59.

5. Agius F, Kapoor A, Zhu JK (2006) Role of the Arabidopsis DNA glycosylase/lyase ROS1in active DNA demethylation. Proc Natl Acad Sci USA 103:11796e11801.

6. Morales-Ruiz T, et al. (2006) DEMETER and REPRESSOR OF SILENCING 1 encode 5-methylcytosine DNA glycosylases. Proc Natl Acad Sci USA 103:6853e6858.

7. Zhu JK (2009) Active DNA demethylation mediated by DNA glycosylases. Annu RevGenet 43:143e166.

8. Wu SC, Zhang Y (2010) Active DNA demethylation: Many roads lead to Rome. Nat RevMol Cell Biol 11:607e620.

9. Rai K, et al. (2008) DNA demethylation in zebrafish involves the coupling of a de-aminase, a glycosylase, and gadd45. Cell 135:1201e1212.

10. Guo JU, Su Y, Zhong C, Ming GL, Song H (2011) Hydroxylation of 5-methylcytosine byTET1 promotes active DNA demethylation in the adult brain. Cell 145:423e434.

11. Kapoor A, Agius F, Zhu JK (2005) Preventing transcriptional gene silencing by activeDNA demethylation. FEBS Lett 579:5889e5898.

12. Ortega-Galisteo AP, Morales-Ruiz T, Ariza RR, Roldán-Arjona T (2008) ArabidopsisDEMETER-LIKE proteins DML2 and DML3 are required for appropriate distribution ofDNA methylation marks. Plant Mol Biol 67:671e681.

13. Choi Y, et al. (2002) DEMETER, a DNA glycosylase domain protein, is required forendosperm gene imprinting and seed viability in arabidopsis. Cell 110:33e42.

14. Penterman J, et al. (2007) DNA demethylation in the Arabidopsis genome. Proc NatlAcad Sci USA 104:6752e6757.

15. Miyao A, et al. (2003) Target site specificity of the Tos17 retrotransposon showsa preference for insertion within genes and against insertion in retrotransposon-richregions of the genome. Plant Cell 15:1771e1780.

16. Hirochika H, Sugimoto K, Otsuki Y, Tsugawa H, Kanda M (1996) Retrotransposons of

rice involved in mutations induced by tissue culture. Proc Natl Acad Sci USA 93:

7783e7788.17. Cheng C, Daigen M, Hirochika H (2006) Epigenetic regulation of the rice retro-

transposon Tos17. Mol Genet Genomics 276:378e390.18. Liu ZL, et al. (2004) Activation of a rice endogenous retrotransposon Tos17 in tissue

culture is accompanied by cytosine demethylation and causes heritable alter-

ation in methylation pattern of flanking genomic regions. Theor Appl Genet 109:

200e209.19. Ding Y, et al. (2007) SDG714, a histone H3K9 methyltransferase, is involved in Tos17

DNA methylation and transposition in rice. Plant Cell 19:9e22.20. Brameier M, Krings A, MacCallum RM (2007) NucPred—Predicting nuclear localization

of proteins. Bioinformatics 23:1159e1160.21. Mirouze M, et al. (2009) Selective epigenetic control of retrotransposition in Arabi-

dopsis. Nature 461:427e430.22. Zheng X, et al. (2008) ROS3 is an RNA-binding protein required for DNA demethy-

lation in Arabidopsis. Nature 455:1259e1262.23. Hsieh TF, et al. (2009) Genome-wide demethylation of Arabidopsis endosperm. Sci-

ence 324:1451e1454.24. Zemach A, et al. (2010) Local DNA hypomethylation activates genes in rice endo-

sperm. Proc Natl Acad Sci USA 107:18729e18734.25. Arabidopsis Genome Initiative (2000) Analysis of the genome sequence of the flow-

ering plant Arabidopsis thaliana. Nature 408:796e815.26. Vij S, et al. (2006) Decoding the rice genome. Bioessays 28:421e432.27. Kato M, Miura A, Bender J, Jacobsen SE, Kakutani T (2003) Role of CG and non-CG

methylation in immobilization of transposons in Arabidopsis. Curr Biol 13:421e426.28. Miura A, et al. (2001) Mobilization of transposons by a mutation abolishing full DNA

methylation in Arabidopsis. Nature 411:212e214.29. Tsukahara S, et al. (2009) Bursts of retrotransposition reproduced in Arabidopsis.

Nature 461:423e426.

La et al. PNAS | September 13, 2011 | vol. 108 | no. 37 | 15503

PLANTBIOLO

GY

Supporting InformationLa et al. 10.1073/pnas.1112704108SI Materials and MethodsPlant Materials and Growth Conditions. All rice plants used in thisstudy were Oryza sativa L. ssp. japonica cv. Nipponbare. TheTos17 insertion mutant dng701-1 and dng701-2 were obtainedfrom the Rice Genome Resource Center in Japan. T0 PC5 RNAitransgenic lines were regenerated from transformed calli carry-ing the DNG701 RNAi construct made from the pANDA vector(1). Plants used for phenotypic observations were grown ina growth chamber until maturity with 16 h of light at 26 °C and80% relative humidity followed by 8 h of dark at 24° C and 60%relative humidity.

Cloning of DNG701 cDNA. To clone the DNG701 cDNA, we am-plified a 2,404-bp cDNA fragment corresponding to the 3′ ter-minus of DNG701 (primers P1-F/P1-R) from a cDNA poolderived from the young leaves of Nipponbare. A 3,618-bp cDNAfragment corresponding to the 5′ terminus of DNG701 wasamplified (primers P2-F/P2-R) from cDNA clone AK070640obtained from KOME’s (Knowledge-based Oryza MolecularBiological Encyclopedia) cDNA library. The DNG701 cDNAwas obtained by ligating the two fragments via the EcoRV re-striction site. The primer sequences are listed in Table S1.

Subcelluar Localization of DNG701. The entire 5,439-bp DNG701cDNA was inserted in-frame into the 3′ terminus of the GFPgene in the pGDG vector (2), where the fused gene was drivenby the 35S promoter. This construct was then transfected intorice protoplasts prepared from Nipponbare 10-d-old seedlings bythe PEG-mediated transfection method (3). The pGDG vectorwas transfected simultaneously as a control. GFP fluorescencewas observed with a Nikon Eclipse E600 fluorescence micro-scope (Nikon). Excitation and emission filter Ex450-490/DM510/BA520-560 was used for GFP fluorescence (3). Images werecaptured with a SPOT 2 Slider charge-coupled camera.

RNA Gel-Blot and RT-PCR Analysis. Total RNA was isolated fromwild-type, mutant, and RNAi transgenic plants using TRIzolreagent (Invitrogen) and then treated with DNase I (Ambion).The procedures for probe labeling, blotting, hybridization, andmembrane washing were carried out as previously described(4). After hybridization was performed overnight at 65° C, themembrane was washed and then exposed to a Phosphor-Imager screen (Amersham). Probes for Tos17 and DNG701were amplified by primers P3-F/P3-R (Probe 2) (Fig. S6B)and P4-F/P4-R, respectively. For RT-PCR analysis, 1.5 μg oftreated total RNA was used to synthesize the first-strandcDNA with a reverse-transcription kit (Promega), as pre-viously described (4). Two sets of primer pairs (P4-F/P5-R forDNG701; P6-F/P6-R for DNG704) were separately applied toexamine the expression levels and patterns of DNG701 andDNG704. The Ubiquitin gene was used as an internal controlfor the RT-PCR (primers P7-F/P7-R). All primers are listed inTable S1.

DNA Gel-Blot Analysis. Genomic DNA (20 μg) isolated from dif-ferent tissues was digested with 50 units of appropriate re-striction enzymes (New England Biolabs) for 16 h. The digestedDNA was separated on a 1% agarose gel and then blotted onHybond N+ nylon membrane. The procedures for preparation of32P-labeled probe, hybridization, and membrane washing were aspreviously described (4, 5). After hybridization was performedovernight at 65 °C, the membrane was exposed to the Phos-

phorImager screen. The probe used to check the centromericrepeat was amplified from a cDNA clone (Genbank accessionno. EB086903). The probes of Tos17 (Probe 1) (Fig. S6B), 45SrDNA, Ty1, Tos17chr10 promoter (Probe 3) (Fig. S6A), actin,and Tubulin were amplified by primers P8-F/P8-R, P9-F/P9-R,P10-F/P10-R, P11-F/P11-R, P12-F/P12-R, and P13-F/P13-R,respectively. All primers are listed in Table S1.

ProteinPurificationandEnzymaticActivityAssays.The entire DNG701cDNA was inserted in-frame into the pMAL-c2X vector (NewEngland Biolabs) to obtain amalE-DNG701 fusion construct. Theconstruct was introduced into Escherichia coli strain BL21 (DE3).Expression and purification of the maltose-binding protein(MBP)-DNG701 fusion protein were performed as previouslydescribed (6). The MBP was purified in parallel by the sameprocedures and used as the control. The plasmid pUBIGUS waspurified from the strain BL21 (DE3, dcm−) using the Maxi-plas-mid purification kit (Qiagen). Fifty micrograms of plasmids weremethylated in vitro by 20 units of CG methyltransferase (M.SssI)or MspI methyltransferase according to the provided protocols(New England Biolabs). Unmethylated control plasmids weretreated simultaneously by the same procedure without addition ofthe methyltransferases. For the nicking activity assays, 200 ng ofeach of these plasmids was added to nicking reaction buffer (10mM Tris-HCl, pH 7.9, 50 mM NaCl, 10 mM MgCl2, 1 mM DTT,0.2 mg/ml BSA) plus purified MBP-DNG701 or MBP. Reactionswere carried out as described (6). The resulting mixtures werechecked in a 1% agarose gel.For incision activity assays, upper strands of the seven oligo-

nucleotides (Fig. S3C) were labeled with 32P at the 5′ ends byusing T4 polynucleotide kinase (Roche), and they were thenannealed to the unlabeled complementary strands as previouslydescribed (6). The labeled double-stranded oligonucleotides(1 pmol) were incubated with MBP-DNG701 or MBP or MspIendonuclease in the nicking reaction buffer in a total volume of10 μl at 37°C for 2 h. The products were separated on 17%denaturing polyacrylamide gels containing 8 M urea, and the gelswere exposed to a PhosphorImager screen.

Genomic Bisulfite Sequencing. For genomic bisulfite sequencing,2 μg of genomic DNA was digested with 20 units of appropriaterestriction enzymes (New England Biolabs) for 3 h. The digestedDNA was purified by ethanol, treated by sodium bisulfite solu-tion, and desalted with the Wizard DNA Clean-Up System(Promega), as previously described (7). Subsequently, 4 μL ofeach recovered DNA sample was subjected to PCR amplificationas follows: 95 °C for 5 min; followed by 15 cycles at 95 °C for 20 s,60 °C for 3 min, and 72 °C for 3 min; and finally 30 cycles at 95 °Cfor 20 s, 50 °C for 1.5 min, and 72 °C for 2 min. Five sets ofprimer pairs (P14-F/P14-R for Tos17; P15-F/P15-R for Tos17chr10

5′ LTR; P16-F/P15-R for Tos17chr7 5′ LTR; P17-F/P17-R forTos17chr10 promoter; and P18-F/P18-R for LOC_Os07g31690)were used for the PCR reactions. The PCR products were clonedinto the pGEM-T Easy cloning vector (Promega), and 15 to20 clones were sequenced to calculate the percentage of cytosinemethylation in each sample. All primers are listed in Table S1.

Sequence Alignments and Phylogenetic Analysis. Sequences ofDNA glycosylase domains were aligned using ClustalX 2.0.10 (8),and the secondary structures of these domains were predictedby PSIPRED v3.0 (http://bioinf.cs.ucl.ac.uk/psipred/). The con-served residues were highlighted using GeneDoc 2.6.02 (http://

La et al. www.pnas.org/cgi/content/short/1112704108 1 of 7

www.nrbsc.org/gfx/genedoc/index.html). To construct a phyloge-netic tree, the aligned sequences were subjected to MEGA v.4(9) with the neighbor-joining algorithm. Neighbor-joining anal-

ysis was done with the pairwise deletion option and with thePoisson correction set for distance model. Bootstrapping wasperformed with 1,000 replicates.

1. Miki D, Shimamoto K (2004) Simple RNAi vectors for stable and transient suppressionof gene function in rice. Plant Cell Physiol 45:490e495.

2. Goodin MM, Dietzgen RG, Schichnes D, Ruzin S, Jackson AO (2002) pGD vectors:Versatile tools for the expression of green and red fluorescent protein fusions inagroinfiltrated plant leaves. Plant J 31:375e383.

3. Chen S, et al. (2006) A highly efficient transient protoplast system for analyzingdefence gene expression and protein-protein interactions in rice. Mol Plant Pathol 7:417e427.

4. Zeng LR, et al. (2004) Spotted leaf11, a negative regulator of plant cell death anddefense, encodes a U-box/armadillo repeat protein endowed with E3 ubiquitin ligaseactivity. Plant Cell 16:2795e2808.

5. Vega-Sánchez ME, Zeng L, Chen S, Leung H, Wang GL (2008) SPIN1, a K homologydomain protein negatively regulated and ubiquitinated by the E3 ubiquitin ligaseSPL11, is involved in flowering time control in rice. Plant Cell 20:1456e1469.

6. Agius F, Kapoor A, Zhu JK (2006) Role of the Arabidopsis DNA glycosylase/lyase ROS1 inactive DNA demethylation. Proc Natl Acad Sci USA 103:11796e11801.

7. Ding Y, et al. (2007) SDG714, a histone H3K9 methyltransferase, is involved in Tos17DNA methylation and transposition in rice. Plant Cell 19:9e22.

8. Larkin MA, et al. (2007) Clustal W and Clustal X version 2.0. Bioinformatics 23:2947e2948.

9. Tamura K, Dudley J, Nei M, Kumar S (2007) MEGA4: Molecular Evolutionary GeneticsAnalysis (MEGA) software version 4.0. Mol Biol Evol 24:1596e1599.

DNG701 DNG704

AtDML2 AtROS1

DNG702 DNG703

AtDME AtDML3

DNGL1

100

100

100

9858

45

0.1

A B DNG701

DNG702

DNG703

DNG704

DNGL1

1812 AA

1914 AA1636 AA

1787 AA

1207 AA

DNA glycosylase EndIII_4Fe-4SNLS

C

Fig. S1. Structures of DNG701 and its homologous proteins. (A) Phylogenetic analysis of DNG701 and its homologs in rice and Arabidopsis (At). (B) Schematicdiagram of domain structures of DNG701 and its four paralogs in rice. NLS, nuclear localization signal. AA, amino acid. (C) Comparison of the DNA glycosylasedomains among DNG701 as well as its four paralogs and the four Arabidopsis 5-meC DNA glycosylase/lyases. Secondary structures of these domains werepredicted and shown. Block H1 to H9 represent helices, and the two highly conserved helices (block H6 and H7), which form a conserved helix–hairpin–helix(HhH) motif, are highlighted with black. The glycine/proline-rich loop resides behind the HhH motif. An invariant aspartic acid (D: marked by a gray square) hasa critical role for enzyme activity. An extremely conserved lysine (K: marked by a gray circle) is located inside the conserved helix.

La et al. www.pnas.org/cgi/content/short/1112704108 2 of 7

C

a b c

d e f

A

5’ 3’

ATG TGADNA glycosylase

(AG214799)

dng701-1

5’ 3’

ATG TGADNA glycosylase

(AG023538)

dng701-2

Ubiquitin

DNG701

Leaf Stem Panicle Root

Ubiquitin

DNG701

Leaf Stem RootB

Fig. S2. Gene structure of DNG701 and its expression pattern. (A) Structure of DNG701. Exons and introns are denoted as black blocks and lines, respectively.The open triangle marks the insertion site of Tos17, and the arrow indicates insertion direction of Tos17. GenBank accession numbers of flanking sequences ofTos17 are indicated above the arrows. The gray blocks show positions of DNA glycosylase domains. Both dng701-1 and dng701-2 are in Nipponbare back-ground. (B) Expression patterns of DNG701 in three different tissues of 3-wk-old Nipponbare seedlings (Upper) and in four different tissues of 3-mo-oldflowering plants (Lower) analyzed by RT-PCR. Ubiquitin was used as an internal control. (C) Subcellular localization of DNG701 in rice protoplasts. (a–c) 35S:GFP controls; (d–f) 35S:GFP:DNG701; (a and d) bright fields; (b and e) green fluorescence; (c and f) merged images of bright field and green fluorescence. (Scalebars, 20 μm.)

La et al. www.pnas.org/cgi/content/short/1112704108 3 of 7

5’ GGGAGAGAGGGAAGCmCGGAGGAAGGAACmCGGGAAAGGGGA 3’3’ CCCTCTCTCCCTTCGGCmCTCCTTCCTTGGCCCTTTCCCCT 5’

CmCGG oligo

5’ GGGGAGAGAGGGAAGCCGGAAGGAAGGAATGGGAAAGGGA 3’3’ CCCCTCTCTCCCTTCGGCCTTCCTTCCTTACCCTTTCCCT 5’ CCGG

29 nt

5’ GGGGAGAGAGGGAAGmCCGGAGGAAGGAAmCCGGGAAAGGGA 3’3’ CCCCTCTCTCCCTTCGGCCmTCCTTCCTTGGCCCTTTCCCT 5’

mCCGG oligo

16 nt29 nt

5’ GGGAGAGAGGGAAGCmCGGAAGGAAGGAATGGGGAAAGGGA 3’3’ CCCTCTCTCCCTTCGGCCTTCCTTCCTTACCCCTTTCCCT 5’ Hemi CmCGG

16 nt

5’ GGGAGAGAGGGAAGCmCGGAAGGAAGGAATGGGGAAAGGGA 3’3’ CCCTCTCTCCCTTCGGCmCTTCCTTCCTTACCCCTTTCCCT 5’ Full CmCGG

16 nt

5’ GGGGAGAGAGGGAAGmCCGGAAGGAAGGAATGGGAAAGGGA 3’3’ CCCCTCTCTCCCTTCGGCCTTCCTTCCTTACCCTTTCCCT 5’ Hemi mCCGG

16 nt

16 nt5’ GGGGAGAGAGGGAAGmCCGGAAGGAAGGAATGGGAAAGGGA 3’3’ CCCCTCTCTCCCTTCGGCCmTTCCTTCCTTACCCTTTCCCT 5’ Full mCCGG

212kD

66.4kD55.6kD42.7kD

97.2kD

MBP MBP-DNG701

C16 nt

A

UnmethylatedmCCGG mCG

B

mCCGG mCG Unmethylated

Fig. S3. Purification of recombinant MBP-DNG701 fusion protein and nicking activity assays. (A) Purification of recombinant MBP-DNG701 fusion protein andMBP protein. The open triangle indicates purified MBP protein and the filled triangle denotes purified recombinant MBP-DNG701 protein. Proteins were run ina SDS/PAGE gel and then stained with Coomassie blue. (B) Nicking activity assays of recombinant MBP-DNG701 (Upper) and MBP (Lower) on methylated andunmethylated plasmids. From the second to fourth lanes in each panel, 0.2, 0.8, and 2.4 μg of proteins were added, respectively. CK, no protein added. mCCGG,mCG, and Unmethylated denotes mCCGG-containing, mCG-containing, and unmethylated pUBIGUS plasmids, respectively. (C) Sequences of oligonucleotidesused for enzymatic activity assays of DNG701 in vitro. The filled stars indicate 32P-labeled 5′ ends on upper strands. Methylated cytosines are underlined. CCGGsites are highlighted with blue; 29 nt and 16 nt denote the number of nucleotides from methylated cytosines to the labeled 5′ ends.

UbiquitinDNG701

DNG701

UbiquitinDNG704

Fig. S4. RT-PCR analysis of dng701-1 and dng701-2 knockout mutants (Left) and DNG701 RNAi lines (Right). WT CTR-1 and WT CTR-2 are sibling wild-typeplants segregated from heterozygous DNG701/dng701-1 and DNG701/dng701-2 plants, respectively. WT CTR-3 is a segregated wild-type plant from a hemi-zygous DNG701 RNAi transgenic plant PC5-9.

La et al. www.pnas.org/cgi/content/short/1112704108 4 of 7

0

20

40

60

80

100

CG CHG CHH

Tos17chr10

promoter

Cyto

sin

e m

eth

yla

tio

n(%

)

NPB WT

WT CTR-1

dng701-1

WT CTR-2

dng701-2

WT CTR-3

PC5-9-1

PC5-10-1

0

5

10

15

20

25

30

35

40

CG CHG CHH

LOC_Os07g31690

Cyto

sin

e m

eth

yla

tio

n(%

)

NPB WT

WT CTR-1

dng701-1

WT CTR-2

dng701-2

B

C

A HpaII MspI

Centromeric repeat DNA

HpaII MspI

45S rDNA

HpaII

Ty1

MspI

Fig. S5. Methylation status of centromeric repeat DNA, 45S rDNA, Ty1 retrotransposons, Tos17chr10 promoter, and another retrotransposon. (A) Methylationstatus of centromeric repeat DNA, 45S rDNA and Ty1 retrotransposons in the wild-type controls, mutants, and RNAi lines. (Left) Genomic DNAs were digestedby HpaII or MspI and then separately blotted to different membranes for hybridization with centromeric repeat DNA probe. (B and C) Methylation status ofTos17chr10 promoter (B) and another retrotransposon (C). The region a (Fig. S6A) at Tos17chr10 promoter and 3′ terminal region of another retrotransposon(LOC_Os07g31690) were determined by bisulfite sequencing.

La et al. www.pnas.org/cgi/content/short/1112704108 5 of 7

5’LTR 3’LTRTos17chr7 4.2kb

c

Tos17chr10

ABC-type transporter

8.2kb

ATG

a(-445~-244 bp)

b

RTL’3RTL’5A

B

Probe 3 (3.8 kb)

BstXI (72)H/M(187)H/M(191)

H/M(210)

H/M(1105)H/M(1612)

H/M(3438)H/M(3838)

BstXI (4138)

569-658

RTL’3RTL’5

1 4204

3968619

3120 352427351555

Probe 1 (3.3 kb)

Probe 2 (1.1 kb)405 bp

XbaI(1490)

Tos17chr10

d

Fig. S6. Structural features of Tos17. (A) Tos17chr10 and Tos17chr7. For Tos17chr10, Tos17 is inserted in a putative ABC-type transporter gene (blue line) to forma fusion gene that is driven by the ABC-type transporter gene’s promoter (blue block). A 202-bp region from −445 to −244 (region a, shown with lime color)was found being methylated as revealed by bisulfite sequencing. A 3.8-kb Probe 3 was used for DNA gel-blot analysis to check methylation state of thepromoter. Black blocks in Tos17chr10 and Tos17chr7 denotes LTR regions. Regions b and c indicate the 5′ LTR regions examined by bisulfite sequencing. (B)Restriction map of Tos17chr10. Restriction sites of BstXI, XbaI, and HpaII/MspI (H/M) are shown above the gene. The numbers in parentheses represent positionsof the restriction sites on the gene. The green block indicates the 90-bp deletion at Tos17chr7. The red and pink lines beneath the gene represent a 3.3-kb Probe1 and a 1.1-kb Probe 2 used for hybridization analysis, respectively. The blue line shows the 405-bp region (region d) examined by bisulfite sequencing.

NPB WTdng701-1normal seeds

dng701-1wrinkled seeds

NPB WTdng701-2normal seeds

dng701-2wrinkled seeds

Fig. S7. Seed phenotypes of knockout of DNG701 in rice. Both the dng701-1 and dng701-2 mutants produced two types of seeds: normal and wrinkled seeds.Mutants were grown in a growth chamber for 4 mo to allow panicles to reach full maturity. NPB WT, Nipponbare wild-type.

0

10

20

30

40

50

60

CG CHG CHHTos17

Cyt

osin

e m

ethy

latio

n(%

) OX-CK-1OX-1OX-2

Fig. S8. Methylation levels of Tos17 in the DNG701-overexpressing calli determined by bisulfite sequencing. Tos17 region d (Fig. S6B) was subjected to thebisulfite sequencing analysis.

La et al. www.pnas.org/cgi/content/short/1112704108 6 of 7

HpaII MspI

Centromeric repeat DNA

HpaII MspI

45S rDNA

HpaII MspI

Ty1

Fig. S9. Methylation status of centromeric repeat DNA, 45S rDNA and Ty1 retrotransposons in the wild-type controls and DNG701-overexpressing plants. For45S rDNA, not only markedly decreased CG methylation but also slightly reduced CHG methylation were observed in the DNG701-overexpressing plants. WTCTR-4 and WT CTR-5, segregated wild-type plants from hemizygous transgenic plant OX-1-27 and OX-2-35, respectively. Genomic DNAs were digested by HpaIIor MspI and then separately blotted to different membranes for hybridization with the probes.

Table S1. Primer list

Primer name 5’-sequence-3’ Comments

P1-F AAATCTACATGCGACACAGTTCCAP1-R actagtTCAGTGTGTTTGCTCAGTTGCTGC Lower-case letters denote SpeI siteP2-F actagtATGGATCCCTCAGGGCTCAACCTG Lower-case letters denote SpeI siteP2-R GCTTCCCACCTGCTGAGAAGCTGTP3-F GCCCAGATGATTCATACTCAGTTCP3-R AAACCTCGCGCAACAAGGP4-F CTGATTCTACAAGTTATAATATAGAP4-R ATCTCTCTTTATCAACATCAGTTP5-R GCTTCTTGACTAAATGAAACTGTGTP6-F GATCACAACGAAAGGAGTGGP6-R ATATGGGACTAGAGCATTCGGCAAP7-F AAGAAGCTGAAGCATCCAGCP7-R CCAGGACAAGATGATCTGCCP8-F TGGGACACGGATTTTGCTTATCAAP8-R TGTAGCTCGGAGGGCACATAGTGGP9-F ATCGATGAAGAACGTAGCGAAATGP9-R GTCGCGTATTTAAGTCGTCTGCAAP10-F AGAAACTTGGCACCACTAGAACTCP10-R GCTAGGATCATAAGGTGTTGGAGAP11-F GACGGCGGTGGCATGGATCCAATTP11-R GTCTGTGCCTAGTTTAATTTAGATP12-F CGTCTGCGATAATGGAACTGGP12-R CTGCTGGAATGTGCTGAGAGAP13-F GGATGCAGACGCTTACTTGAGGTTP13-R ACAAATGTTGCACCAGTTTGTATTP14-F GTTGATTAtAGGGGATGATTTGGAGTAtATTGtTT Lower-case letters represent C-to-T substitutionP14-R CaTAAaACACaAAaCAAaTaACCATAaTaAACTa Lower-case letters represent G-to-A substitutionP15-F tTGttTTtAtTTTTtATATtA Lower-case letters represent C-to-T substitutionP15-R CAaTaaAaCAaTaaATAAATA Lower-case letters represent G-to-A substitutionP16-F TTGGtAttAATGAGttttAtT Lower-case letters represent C-to-T substitutionP17-F ttTGCAtTtGAAttATGTt Lower-case letters represent C-to-T substitutionP17-R CaACAaCaACaCaaCTaa Lower-case letters represent G-to-A substitutionP18-F tAATtAAttATTGtTtTttAT Lower-case letters represent C-to-T substitutionP18-R TAAAATaCAAACCACACACAT Lower-case letter represents G-to-A substitution

La et al. www.pnas.org/cgi/content/short/1112704108 7 of 7