Role of RNA directed DNA methylation in controlling genomic imprinting in arabidopsis thaliana

Second, our results showed that the RdDM pathway is also necessary for silencing the paternal allele of HAIKU2, which is a maternally expressed imprinted gene in endosperm.. viii List of

1

ROLE OF RNA-DIRECTED DNA METHYLATION IN

GENOMIC IMPRINTING IN ARABIDOPSIS THALIANA

VU MINH THIET

(B Sc, Vietnam National University, Hanoi)

A THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

DEPARMENT OF BIOLOGICAL SCIENCES

NATIONAL UNIVERSITY OF SINGAPORE

2010

i

ACKNOWLEDGEMENT

I would like to take this opportunity to thank everyone involved who helped

me get to this place in my life and fulfill the long process of persuading the PhD study

I especially would like to express my deepest gratitude to my supervisor Frederic Berger I would never finished my study without his unending support for the last few years I believe that I have learnt a tremendous amount from him I appreciate him for believing in me along the way I would also like to thank my supervisor Prof Davis Ng for his help that allowed me to stay and work in TLL

I have extremely lucky to work with hearty colleagues Pauline, Pei Qi, Jeanie,

Li Jing, Ramesh, Chen Zhong, Sarah, Heike, Nie Xin, Tomo, and also the former lab members Mathieu, Tadashi, Jonathan, Arnold I would like to thank my colleagues for their sharing, accompanying me all the times during my study A special thank I give

to Pauline for her excellent guidance when I started working in the lab, I have learnt a lot from her discipline in doing experiment and her critical thinking in science

I would like to thank my thesis advisory committee Dr Yu Hao, Dr Toshiro Ito, and Dr Jose Dinneny for their support

I would also like to thank DBS Graduate Office and Ms Reena, Ms Priscilla They are always there to answer my questions and help me to process my paper works

as fast as possible

I also would like to thank Tam, Long, Hang, Hung, Ngoc, Trang for their support and encouragement for last few years in Singapore

ii

I would like to thank my parents for their support over the years and not complaining me when I could not come home for several TETs A special thank to my brother for his presence in Singapore and his understanding of my hard time Thanks for your care of our parents

Last but not least, I acknowledge NUS graduate scholarship and TLL for financial support

Thank you so much everyone

September, 2010

iii

TABLE OF CONTENTS

SUMMARY vi

LIST OF TABLES x

LIST OF ABBREVIATIONS xi

CHAPTER 1: INTRODUCTION 1

Abstract 2

1 Introduction of imprinting 3

1.1.1 First reports of imprinting in plants 6

1.1.2 The impact of interploid crosses on imprinting discovery 6

1.2 Imprinted genes and their function 9

1.2.1 Arabidopsis imprinted genes 9

1.2.2 Conservation of Polycomb group imprinted genes in cereals 14

1.3 Conclusion 15

2 Molecular mechanisms controlling imprinting 15

2.1 Imprinting by DNA methylation 16

2.1.1 Maintenance of DNA methylation on the silent alleles 16

2.1.2 Two-step removal of DNA methylation in the central cell 18

2.2 Molecular controls of imprinting by Histone methylation 20

2.3 Cis-elements controlling imprinting 23

2.3.1 Cis-elements in the promoter 23

2.3.2 Evidence for imprinting regulation by long distance elements 25

2.4 Genomic imprinting and RNA-directed DNA methylation 27

2.5 Imprinting, a by-product of the global reprogramming? 30

iv

3 Biological significance and evolution of imprinting 31

3.1 Parental conflict 31

3.2 Maternal control 33

3.3 Imprinting, a factor of speciation 34

3.4 Transposons as the primer of imprinting evolution in a specific developmental context 34

Aims of the research 36

CHAPTER 2: GENOMIC IMPRINTING AND RNA-DIRECTED DNA METHYLATION IN ARABIDOPSIS 39

2.1 Introduction 40

2.2 Results 43

2.2.1 SDC is silenced in somatic tissues and expressed in seeds 43

2.2.2 SDC is a new imprinted gene in Arabidopsis 47

2.2.3 Silencing mechanism of SDC paternal allele 50

2.2.4 Mechanism of activation of SDC maternal allele 52

2.3 Discussion 59

2.4 Future work 65

Material and Methods 66

CHAPTER 3: Accession-dependent Imprinting of HAIKU2 is controlled by the RdDM Pathway 73

3.1 Introduction 74

3.2 Results 78

3.2.1 IKU2 expression in gametes 78

3.2.2 Is IKU2 an imprinted gene? 80

v

3.2.3 What mechanism controls silencing of the paternal IKU2 allele? 83

3.3 Discussion 86

3.4 Future work 90

Material and Methods 91

CHAPTER 4: GENERAL DISCUSSION 94

4.1 Main findings 95

4.2 Biological significance 96

4.2.1 Further expansion of the number of imprinted genes 96

4.2.3 Is imprinting the by-product of the asymmetrical activity of major controls of DNA methylation? 98

4.3 Future perspective 100

References 103

Annex: Maternal effect of mutation in RdDM pathway on seed development 116

vi

SUMMARY

In flowering plants and placental mammals, a subset of genes are expressed

depending on their parent of origin and defined as imprinted genes In Arabidopsis,

the non-expressed allele of imprinted genes is silenced by either DNA methylation or

Histone methylation by Polycomb repressive complex activity The Arabidopsis

imprinted genes are expressed only in endosperm, which nurtures embryo development Imprinted expression is established in two steps involving the maintenance of DNA methylation The paternal silenced allele remains marked by DNA methylation during spermatogenesis while gene silencing is released from the maternal allele by a demethylation pathway active during female gametogenesis After fertilization, the expressed allele remains active and the inactive allele remains silent by maintenance DNA methylation machinery

DNA methylation can be deposited de novo through the RNA-directed DNA

methylation pathway (RdDM) or maintained by the methyltransferase MET1 The maintenance methyltransferase MET1 plays a major role and is sufficient to establish monoallelic expression of most imprinted genes identified so far However, the function of RdDM in genomic imprinting has remained largely unknown

In contrast with MET1 activity, the RdDM pathway results in methylation of cytosine residues in any context and in absence of a hemimethylated template The RdDM

pathway comprises the de novo methyltransferase DRM2 and the RNA polymerases

POLIV and POLV In the course of this thesis, we have identified a role of the RdDM

pathway in regulating genomic imprinting in Arabidopsis thaliana

vii

First, we identified SDC (suppressor of drm1, drm2, and cmt3) as a maternally imprinted gene The SDC gene is primarily silenced by the RdDM pathway We showed that SDC is specifically expressed in endosperm from its maternal allele and

silencing of the paternal allele requires the RdDM pathway

The absence of expression of key genes in the RdDM pathway during female gametogenesis while it is maintained during spermatogenesis is sufficient to explain

the origin of the imprinted expression of SDC

Second, our results showed that the RdDM pathway is also necessary for silencing the

paternal allele of HAIKU2, which is a maternally expressed imprinted gene in endosperm The imprinted expression of HAIKU2 is observed in a genetic context- depending manner, relying on the accessions used in the reciprocal crosses HAIKU2

controls endosperm growth

In conclusion, we described two novel imprinted genes in Arabidopsis More

importantly, we identified RdDM as a new silencing mechanism functioning in imprinting establishment The loss of RdDM activity during female gametogenesis is predicted to cause a genome-wide demethylation Our findings suggest that a new class of RdDM-dependent imprinted genes remains to be characterized in plants

viii

List of Figures

Figure

number

1-5 Establishment of DNA methylation-dependent imprinted

genes throughout the plant life cycle

17

1-6 DNA methylation dependent mechanisms leading to

imprinting of maternally expressed genes in Arabidopsis

19

1-7 Polycomb Repressive Complex 2 (PRC2) dependent

mechanisms leading to imprinting of maternally expressed

genes in Arabidopsis

21

1-8 Long distance cis-elements of imprinted genes 26

2-1 De novo methylation controls SDC expression 45 2-2 Expression of SDC in gametophytes and seeds 46 2-3 Allele specific RT-PCR analysis of maternal SDC

expression

49

2-4 SDC imprint is controlled by RdDM pathway 51 2-5 Maternally expressed SDC is not controlled by DEMETER 54 2-6 Expression of NRPD2 gametophytes and developing seed 55 2-7 NRPD2a-RFP construct recues the loss of endogenous

ix

expression in central cell

2-10 A proposed model for RdDM function in controlling SDC

imprinting

63

3-1 Loss of IKU2 causes small seed size phenotype 77

3-4 Allele-specific RT-PCR results show maternal expression

of IKU2

81

3-5 Origin parental expression of IKU2 in combination of

different Arabidopsis accessions

82

3-6 Expression of IKU2 in mutation of DNA methylation and

Polycomb group

85

3-7 Silencing of paternal IKU2 allele is controlled by a

DNA-dependent RNA polymerase IV

85

MET1 DNA METHYLTRANSFERASE 1

DRM2 DOMAIN REARRANGED METHYLTRANSFERASE

FIE FERTILIZATION INDEPENDENT ENDOSPERM

FIS FERTILIZATION INDEPENDENT SEED

FWA FLOWERING WAGENINGEN

GAPDH GLYCERALDEHYDE 3-PHOSPHATE DEHYDROGENASE

H3K27 Lysine 27 residue of HISTONE3

MPC MATERNALLY EXPRESSED PAB C-TERMINAL

MSI1 MULTICOPY-SUPPRESSOR OF IRA1

NRPD1A NUCLEAR RNA POLYMERASE D1

NRPD2A NUCLEAR RNA POLYMERASE D2

NRPD1B NUCLEAR RNA POLYMERASE E

PRC2 Polycomb Repressive Complex 2

RdDM RNA-directed DNA methylation

SDC SUPPRESSOR OF drm1, drm1 and cmt3

1

CHAPTER 1: INTRODUCTION

2

Abstract

Although most genes are expressed equally from both parental alleles, imprinted genes are differentially expressed depending on their parental origin In flowering plants, imprinting is regulated by DNA methylation and histone methylation Most imprinted genes are silenced by chromatin modifications during vegetative development During gametogenesis the male or the female allele is activated by removal of chromatin modification and remains active after fertilization while the other allele remains silenced, leading to imprinted gene expression Imprinting mechanisms are conserved across plant species and to a certain extent there is evidence of convergent evolution of imprinting mechanisms between plants and mammals The physiological significance and evolutionary origin of imprinting are still unclear but in plants, imprinting may be the consequence of global epigenetic reprogramming during sexual reproduction

3

1 Introduction of imprinting

In plants male and female gametes are produced after meiosis following a series of divisions of a haploid male or female spore (Figure 1-1) In flowering plants male gametogenesis takes place in stamen to produce pollen Meiosis produces four haploid microspores Each microspore experiences two mitotic divisions producing a pollen grain comprising a large vegetative cell and two identical haploid sperm cells (Figure 1-1) Female gametogenesis takes place within the diploid tissues of the ovule (Yadegari and Drews, 2004) Meiosis produces one surviving megaspore The haploid megaspore undergoes a series of three syncytial divisions, followed by cellularisation producing the embryo sac The embryo sac contains the haploid female gamete or egg cell, and the central cell (Figure 1-1)

Plant reproduction is characterized by a double fertilization Two sperms are delivered

by the pollen tube to the egg cell and the central cell Fertilization of the egg cell leads

to embryogenesis (Figure 1-2) The second sperm cell activates division of the central cell leading to production of the endosperm The endosperm develops around the embryo and allows transfer of maternal nutrients and physical protection to the embryo (Figure 1-2) In most plant species, the central cell inherits two haploid nuclei from the syncytial gametophyte The endosperm genome thus contains two doses of the maternal genome and one dose of the paternal genome This specific parental genomic dosage attracted interest in early studies of plant reproduction, which led to the discovery of imprinting in plants After an historical account, we will review in this section the identity and function of imprinted genes in plants

4

Figure 1-1 Gametogenesis in flowering plants

In flowering plants, male gametogenesis takes place in the anther of the flower

Diploid pollen mother cell undergoes meiosis to produce a tetrad structure carrying four haploid nuclei Each of them develops into a microspore Each microspore

undergoes a mitotic division without cytokinesis to produce two haploid nuclei The two nuclei have different destiny, one become generative nucleus which later on divide into two sperm cells; whereas the other haploid nucleus remains undivided and become vegetative nucleus

Female gametophytes are produced by meiosis division of megaspore mother cell in the ovary of the flower A single megaspore mother cell undergoes meiosis to produce four megaspores which are linearly arranged Out of these four megaspores, three degenerate and the remaining undergoes three successive mitotic divisions without cytokinesis to form a large cell with eight haploid nuclei Three of them move to the micropylar end to form two sygergid cells which degenerate soon The remaining develops into the egg cell Another group of three nuclei migrate to the chalazal pole and become antipodal cells The remaining two nuclei in the centre of the embryo sac unite to become central cell

5

Figure 1-2 Double fertilization in angiosperm

Two sperm cells of the pollen grain are delivered into the embryo sac which contains two female gametes, egg cell and central cell One sperm cell fertilizes egg cell to give rise embryo The second sperm cell combine with the diploid central cell to produce a triploid (3n) tissue called endosperm which develops surrounding the embryo and supports embryogenesis Two products of double fertilization develop into the seed structure which is protected by maternally derived seed integument

6

1.1 Historical discovery of imprinting

1.1.1 First reports of imprinting in plants

The term imprinting was originally adopted to qualify the differential elimination of paternal chromosomes in the mealybug Sciara (Crouse, 1960) The first example of imprinted expression of a gene was identified in the study of pigmentation of the outer layers of the endosperm in maize (Kermicle, 1970) Irregular anthocyanin

pigmentation was linked to certain alleles of the R gene and was conferred only when

the mutation was inherited from the mother Kermicle proposed that expression of the

r allele depended on its parental origin, and he further showed that the effects

observed did not depend on gene dosage Hence, imprinting was initially described in plants several years before the concept was formulated in mice following the discovery that certain chromosomal regions can lead to developmental abnormalities when both copies are exclusively maternally or paternally derived (Cattanach and Kirk, 1985) However parent of origin expression in maize was only observed for

certain r mutant alleles and did not reflect that in the wild type only the maternal R

allele is expressed Today the mechanism causing allele-dependent imprinting is still not understood

1.1.2 The impact of interploid crosses on imprinting discovery

Evidence for parental genomic imprinting also came from the study of the seeds developing from crosses between plants with different ploidy As early as the middle

of the twentieth century crosses between tetraploid and diploid plants were shown to result in seed abortion due to endosperm failure (Cooper, 1951; Randolph, 1935) It was later shown in maize that a critical maternal : paternal genome dosage in the

7

endosperm was required for seed survival (Lin, 1984) These experiments were

repeated in Arabidopsis using tetraploid and hexaploid plant lines (Scott et al., 1998)

Increased paternal contribution caused endosperm enlargement, whereas increased maternal dosage had the opposite effect These results could be explained by different expression between paternal alleles and maternal alleles of certain genes important for endosperm development These experiments led to the model that two sets of maternally expressed imprinted genes (MEG) and paternally expressed imprinted genes (PEG) control endosperm development

MEDEA (MEA), the first imprinted gene in Arabidopsis was identified more than a

decade ago (Kinoshita et al., 1999; Vielle-Calzada et al., 1999) It was shown clearly

that MEA was actively transcribed after fertilization (Baroux et al., 2006; Calzada et al., 1999) Maternal transcription of MEA was shown using a polymorphism between two wild type strains of Arabidopsis (Figure 1-3) MEA

Vielle-maternal expression was then confirmed using transcriptional reporters (Figure 1-3) (Kinoshita et al., 1999; Luo et al., 2000; Wang et al., 2006)

8

Figure 1-3: Analysis of MEDEA imprinting

A Expression of MEA::GUS reporter construct in crosses, which involve the transgenic line carrying the reporter construct as a mother or as a father MEA::GUS is expressed only when contributed maternally

B Schematic of the analysis of the imprinted expression of MEA endogenous

locus A sequence polymorphism is used to distinguish MEA mRNA and αVPE mRNA between the wild type strains Col and Rld Seeds resulting from crosses between the two parents express mRNAs from both parental alleles and two bands are detected In contrast MEA mRNA originates only from the maternal allele and a single band is detected

9

1.2 Imprinted genes and their function

The endosperm is currently the only tissue where imprinted gene expression has been identified in Arabidopsis (Berger and Chaudhury, 2009; Kinoshita et al., 2008) rice,

and maize with the exception of Maternally expressed in embryo 1 (Mee1) in maize

which is imprinted in the embryo as well as in the endosperm (Jahnke and Scholten, 2009) (Table 1)

1.2.1 Arabidopsis imprinted genes

In Arabidopsis, amongst the first maternally expressed imprinted genes identified

MEA (Kinoshita et al., 1999; Vielle-Calzada et al., 1999) and FERTILIZATION INDEPENDENT SEED 2 (FIS2) (Jullien et al., 2006a; Luo et al., 2000; Luo et al.,

1999) are core members of the endosperm specific FERTILIZATION INDEPENDENT SEED (FIS) Polycomb group Repressor Complex 2 (PRC2) also

including FERTILIZATION INDEPENDENT ENDOSPERM (FIE) (Ohad et al., 1999) and MULTICOPY-SUPPRESSOR OF IRA1 (MSI1) (Guitton and Berger, 2005;

Guitton et al., 2004; Kohler et al., 2003b), which are not imprinted PRC2 methylates the lysine 27 residue of HISTONE3 (H3K27), and thereby represses transcription (Hennig and Derkacheva, 2009; Schuettengruber et al., 2007)

The wild-type endosperm posterior pole (also defined as chalazal pole) is distinguished from the peripheral and anterior (micropylar) domains of the endosperm

by a multinucleate structure identified as the cyst (Boisnard-Lorig et al., 2001; Brown

et al., 1999; Scott et al., 1998) (Figure 1-4) The endosperm of fis mutants is

characterized by multiple defects including enhanced proliferation, much enlarged

10

posterior structures and absence of cellularization (Guitton et al., 2004; Kiyosue et al., 1999; Kohler et al., 2003a; Luo et al., 1999) This pleiotropic phenotype might be the consequence of the maintenance of a juvenile developmental program (Ingouff et al., 2005a) Although some targets of the FIS PcG complex have been identified, the pathways downstream of this transcriptional regulation are unknown and targets

whose functions explain the fis phenotype have not been fully understood We will

detail below the function of two targets of the FIS PcG complex, which are

themselves imprinted, the ARABIDOPSIS FORMIN HOMOLOGUE 5 (AtFH5), and

PHERES1

AtFH5 encodes an actin-nucleating agent (Ingouff et al., 2005b) and is maternally

expressed in the endosperm (Fitz Gerald et al., 2009) The posterior endosperm cyst develops from the migration of nuclei from the peripheral endosperm (Guitton et al., 2004) (Figure 1-4) The early endosperm syncytial development ends when cellularization partitions the syncytium into mono-nucleate cells, but cellularization does not occur in the posterior pole (Brown et al., 1999; Sorensen et al., 2002) AtFH5 expression is confined to the posterior pole and is required for nuclear migration to

this part of endosperm (Ingouff et al., 2005b) The restricted expression of AtFH5 in the posterior endosperm depends on the FIS PRC2 In absence of FIS function AtFH5

is expressed ectopically, preventing proper development of the posterior pole (Fitz Gerald et al., 2009)

PHERES1 (PHE1) is paternally expressed (Kohler et al., 2005; Makarevich et al.,

2006) in endosperm and encodes a type one MADS-box transcription factor of the AGAMOUS-LIKE family (AGAMOUS-LIKE37) PHE1 antagonizes the role of FIS

C-TERMINAL (MPC) FWA encodes an homeodomain leucine zipper (HD-ZIP)

protein (Soppe et al., 2000), is expressed maternally only in endosperm where its function is not known (Kinoshita et al., 2004) When ectopically expressed in vegetative tissues FWA binds and inhibits the function of FLOWERING LOCUS T (FT), causing late flowering (Ikeda et al., 2007) Three other members of HD-ZIP genes (HDG3, HDG8 and HDG9) also show imprinted expression in endosperm (Gehring et al., 2009b) HDG8 and HDG9 are maternally expressed while HDG3 is expressed predominantly from its paternal allele

MPC encodes the C-terminal region of a poly(A) binding protein (PABP) (Tiwari et

al., 2008) MPC is able to bind CTC-interacting (CID) proteins but lacks the RNA

binding domain and its function is not known MPC is also expressed but not

imprinted in vegetative tissues and in the embryo

12

Table 1 List of imprinted genes and their function in plants

Maize (Zea

mayz)

R (Certain alleles only) Transcription Factor (Kermicle, 1970)

Allele MO17 of the dzr1

(Chaudhuri and Messing, 1994)

Fertilization independent endosperm 1 (Fie1)

Pc-G chromatin remodeling factor

(Danilevskaya et al., 2003)

No apical meristem related

Maternally expressed gene1

FLOWERING WAGENINGEN (FWA)

Homeobox transcription factor

(Kinoshita et al., 2004)

transcription factor

(Kohler et al., 2005; Makarevich et al., 2008)

FERTILIZATION INDEPENDENT SEED 2 (FIS2)

Pc-G chromatin remodeling factor

(Jullien et al., 2006b; Luo et al., 2000; Luo et al., 1999)

MATERNALLY

C-TERMINAL (MPC) 43

C-terminal domain of poly(A) binding proteins (PABPs);

probably controls mRNA stability and translation

MEE1

13

Figure 1-4 Developmental feature of endosperm

Endosperm is a part of the developing seed and is thought to control the maternal nutrient supply for the embryo development (A) Endosperm development starts with several rounds of synchronous syncytial divisions without cytokinesis to form three endosperm domains defined as anterior micropylar, peripheral and posterior chalazal domain (B) At heart embryo stage, the endosperm cellularization take places at anterior domain around the embryo, followed by the cellularization of peripheral domain In contrast, the posterior endosperm does not cellularise and consists of multi nucleate masses of cytoplasm, defined as the cyst (cy) at the most posterior part, and

as nodules (no) when located at the anterior part of the cyst (Scott et al., 1998) (C) AtFORMIN5 (AtFH5) encodes acting nucleating agent Its expression is restricted to the posterior pole during endosperm development AtFH5 is required for completion

of cytokinesis (Ingouff et al., 2005b)

14

1.2.2 Conservation of Polycomb group imprinted genes in cereals

In maize only maternally expressed imprinted genes have been identified and there is currently limited evidence for their function (Table 1) (Gutierrez-Marcos et al., 2006; Gutierrez-Marcos et al., 2004; Hermon et al., 2007; Kermicle, 1970) to the exception

of homologues of PRC2 members

The Arabidopsis gene MEDEA displays an imprinted pattern of gene expression and

has homology to the Drosophila Polycomb group (PcG) protein Enhancer-of-zeste

(E(z)) Amongst the three maize E(z)-like genes, Mez1, Mez2 and Mez3 only Mez1

displays a mono-allelic expression pattern in the developing endosperm tissue (Haun

et al., 2007) The two rice E(z)-like genes, OsiEZ1 and OsCLF are not imprinted (Luo

et al., 2009)

A stronger conservation of imprinting in cereals was found in the FIE homologues The maize FIE2 and sorghum FIE proteins form a monophyletic group, sharing a

closer relationship to each other than to the FIE1 protein, suggesting that maize Fie

genes originated from two different ancestral genomes (Danilevskaya et al., 2003)

The maize FIE1 gene is maternally expressed exclusively in the endosperm while

FIE2 is maternally expressed in the embryo and at lower levels in the endosperm

(Danilevskaya et al., 2003; Hermon et al., 2007) The rice genome also contains two

FIE homologues, OsFIE1 and OsFIE2 OsFIE1 is expressed only in endosperm; the

maternal copy is expressed while the paternal copy is not active (Luo et al., 2009)

The function of FIE homologues in cereal endosperm is not known

15

1.3 Conclusion

Genomic imprinting in plants primarily affects genes expressed in endosperm Most imprinted genes identified in plants are maternally expressed and only a few

paternally expressed imprinted genes have been identified in Arabidopsis

Experimental elimination of the paternal genome in fertilized central cells confirmed that maternal expression is not sufficient for early endosperm development (Aw et al., 2010) The requirement of paternal genome expression suggests the existence of yet unidentified paternally expressed imprinted genes Imprinting affects genes encoding members of a conserved PcG complex, which plays a key role in the control of several aspects of endosperm development including polarity, growth, and temporal aspects Studies using interploid crosses suggest that the overall function of imprinting related to the control of endosperm growth and seed size However, a comprehensive picture of the total number of imprinted genes is still unknown It is likely that the development of deep sequencing technologies coupled with the use of polymorphisms will enable rapid discovery of new imprinted genes in various species

2 Molecular mechanisms controlling imprinting

Parental genomic imprinting originates from epigenetic mechanisms acting during gametogenesis that differentiate the transcriptional state of the two prospective parental alleles Epigenetic regulations include a wide spectrum of mechanisms that regulate and modify phenotypes independently of the genotype (Kouzarides, 2007; Law and Jacobsen, 2010; Martienssen et al., 2008; Roudier et al., 2009) Covalent

16

modifications of the chromatin regulate the expression of the genome In all eukaryotes, histones are subjected to various types of modifications and in most eukaryotes DNA is methylated at cytosine residues DNA methylation and certain types of histone methylation can be transmitted through cell division and as a consequence constitute a form of epigenetic memory

In plants imprinted genes are not expressed when gametogenesis is initiated after meiosis Their silent state depends either on DNA methylation or H3K27 methylation

In either the male or female gamete lineage, chromatin modifications, which silence gene expression, are removed by the end of gametogenesis After fertilization the difference between the transcriptional status of the two parental alleles persists, leading to a stable imprinted expression in the endosperm

2.1 Imprinting by DNA methylation

2.1.1 Maintenance of DNA methylation on the silent alleles

Both alleles of FIS2, MPC and FWA are silenced throughout the plant life cycle until gametogenesis occurs (Figure 1-5) Silencing of FWA, FIS2, and MPC is mediated by

the DNA METHYLTRANSFERASE 1 (MET1), which maintains DNA methylation

of CpG sites (Jullien et al., 2006a; Kinoshita et al., 2004; Tiwari et al., 2008) Silent

status of FWA, FIS2 and MPC is maintained during female gametogenesis until egg

cell and central cell are differentiated (Jullien et al., 2006a; Tiwari et al., 2008)

During endosperm development, the inherited paternal copy of FWA, FIS2 and MPC

remains silenced by MET1, whereas the maternal copy is inherited as transcriptionally

17

active, resulting in monoparental expression (Gehring et al., 2006; Jullien et al., 2006a; Jullien et al., 2006b; Kinoshita et al., 2004; Tiwari et al., 2008) This mechanism is conserved in maize where analysis of the maternally expressed

Fertilization Independent Endosperm1 (FIE1) locus showed that DNA methylation is

present in sperm cells and is specifically absent in the central cell, but not in the egg cell (Gutierrez-Marcos et al., 2006; Hermon et al., 2007) This direct assessment of DNA methylation in isolated central cells shows that epigenetic marks differ between each gamete and prefigure the imprinted expression after fertilization in the endosperm

Figure 1-5 Establishment of DNA methylation-dependent imprinted genes throughout the plant life cycle

DNA methylation dependent imprinted genes are silenced by DNA methyltransferase MET1 in the vegetative tissue Maintain of DNA methylation is also required for silencing of imprinted genes in pollen Whereas DNA methylation was actively removed from the allele in the female gametes (central cell) and the gene becomes active Consequently, in endosperm, the maternally inherited allele is active, whereas the paternally inherited allele remains silence by the action of MET1

18

2.1.2 Two-step removal of DNA methylation in the central cell

The removal of the DNA methylation marks in the central cell depends on two successive mechanisms causing passive then active demethylation (Figure 1-6) Passive DNA demethylation is caused by the transcriptional repression of MET1 by

the Arabidopsis Retinoblastoma homolog RBR1 and its partner, the WD40 domain

containing protein MULTICOPYSUPPRESSOR OF IRA 1 (MSI1) (Jullien et al.,

2008) The two proteins bind to the promoter of MET1 and repress MET1

transcription causing low activity of MET1 during the late phase of female gametogenesis, while DNA still replicates The Retinoblastoma-dependent repression

of MET1 transcription is predicted to cause production of hemi-methylated DNA, the

preferred substrate for the DNA demethylase DEMETER (DME)

19

Figure 1-6 DNA methylation dependent mechanisms leading to imprinting of

maternally expressed genes in Arabidopsis

The DNA methyltransferase MET1 is expressed throughout male gametogenesis and

in the sperm cells MET1 maintains DNA methylation silencing marks (grey triangles) on the paternal allele of imprinted genes During the second part of female

gametogenesis, MET1 expression is repressed by the Retinoblastoma (pRB) pathway

This repression causes passive removal of DNA methylation on the maternal allele This mode of demethylation is not sufficient to cause expression of the target gene Only in the mature central cell, the glycosylase DEMETER (DME) is expressed and removes actively DNA methylation from the maternal allele provided to the endosperm, resulting in imprinted expression DME is not expressed in the egg cell and both the paternal and the maternal alleles remain silenced in the embryo

20

The activation of expression of FWA, FIS2 and MPC in the central cell also relies on DME (Choi et al., 2002; Jullien et al., 2006a; Kinoshita et al., 2004; Tiwari et al., 2008) DME is expressed in the central cell and encodes a DNA glycosylase that removes methylated cytosine residues (Gehring et al., 2009a) The action of DME creates single-strand DNA breaks that are repaired by the DNA ligase I (Andreuzza et al., 2010) The synergistic action of passive demethylation by repression of MET1 activity followed by active demethylation by DME might completely demethylate the cis elements in FIS2 and FWA promoters causing expression of these genes in the central cell After fertilization, the active maternal allele is inherited with a demethylated cis element while the inactive paternal allele is inherited with a fully methylated cis element (Figure 1-7) MET1 is active in endosperm and maintains the imbalanced pattern of methylation causing imprinted expression in endosperm Such a mechanism is likely to apply to all maternally expressed imprinted genes silenced by MET1 in sperm cells

2.2 Molecular controls of imprinting by Histone methylation

In vegetative tissues both alleles of MEA are silenced by H3K27 tri-methylation

mediated by Pc-G complexes (Gehring et al., 2006; Jullien et al., 2006b)

Genome-wide arrays of DNA methylation and H3K27 trimethylation have shown that the MEA

locus is covered with H3K27 trimethylation (Zhang et al., 2007b) Compromising

H3K27 trimethylation in mutants for Pc-G activity causes MEA ectopic expression in pollen and vegetative tissues (Figure 1-7) MEA imprinted status is lost in mutants for

the Pc-G complex active in endosperm (Gehring et al., 2006; Jullien et al., 2006b)

21

The silencing of MEA by H3K27 methylation implies that transcriptional activation of

MEA requires removal of tri-methylated H3K27 during female gametogenesis The

mechanism causing such a removal remains unknown

Figure 1-7 Polycomb Repressive Complex 2 (PRC2) dependent mechanisms

leading to imprinting of maternally expressed genes in Arabidopsis

PRC2 maintains H3K27 methylation silencing marks on the parental alleles of MEA

(green spheres) The two sperm cells fertilize the egg cell and the central cell During male gametogenesis, H3K27 methylation is maintained while MEA becomes expressed in the central cell Hence the endosperm inherits a silenced paternal allele (p) and an active maternal allele (m), resulting in imprinted expression The origin of

MEA expression in the central cell remains unknown

22

However DME is required for MEA transcriptional activation (Choi et al., 2002; Xiao

et al., 2006) and it is possible that MET1 and DME indirectly activate a pathway that

removes H3K27 trimethylation marks from MEA leading to its activation Alternatively, maternal MEA expression may require a transcriptional activator that is

itself directly controlled by DNA methylation and DME activity

One of the maize homologs of MEA, Maize enhancer of zeste 1 (Mez1) is imprinted

(Haun et al., 2007), and the silenced paternal allele carries H3K27 trimethylation

(Haun and Springer, 2008) Similar to the self regulation of MEA imprinting, disruption of PcG function provided by the Mez1 maternal allele causes expression of the Mez1 paternal allele (Haun et al., 2009), suggesting a conservation of the mechanisms that regulate imprinting of MEA and its homolog Mez1

PHERES1 (PHE1) is a paternally expressed imprinted gene (Kohler et al., 2005) The

silencing of the maternal allele of PHE1 is mediated by the maternal action of Pc-G in endosperm (Makarevich et al., 2006) However, the PHE1 maternal allele is

expressed at variable levels depending on the natural accessions (Makarevich et al.,

2008) The mechanisms causing transcriptional activation of PHE1 in the male gametes remain unknown and a Pc-G independent mechanism regulates PHE1 (see

2.3.2)

The transcription of the gene AtFH5 is also directly controlled by Pc-G complex activity (Fitz Gerald et al., 2009) AtFH5 expression is silenced by Pc-G complexes active in vegetative tissues prior to gametogenesis Unlike MEA, AtFH5 expression is not activated in the central cell but only the maternal allele of AtFH5 is expressed after fertilization AtFH5 expression is also confined by Pc-G activity to the posterior

23

pole of the endosperm, suggesting additional transcriptional controls The site marked

by Pc-G activity and sufficient for imprinting is contained in a 400bp domain of the

AtFH5 promoter; but the detailed mechanism causing imprinting of AtFH5 is not

known

Although histone methylation by Pc-G is involved in imprinting in mammals (Feil and Berger, 2007), it does not appear to act as the essential repressor of the silenced alleles of imprinted genes as shown for certain imprinted genes in plants A major challenge is to understand what mechanisms remove the H3K27 methylation mark

from the expressed alleles of MEA, AtFH5 and PHE1 Understanding such

mechanisms will further provide means to identify other imprinted genes controlled

by the Pc-G pathway

2.3 Cis-elements controlling imprinting

H3K27 methylation by Pc-G activity is wide spread over Arabidopsis imprinted loci and little is known about Polycomb response elements (PRE) similar to those described in Drosophila In contrast DNA methylation at CpG has been localized to

well defined cis-elements in the promoter and in 3’ of the coding sequence of

imprinted genes

2.3.1 Cis-elements in the promoter

Promoter regions of protein coding genes may contain special sequences that could

act as cis-elements for gene expression regulation In mammals, the cis-elements of

24

imprinted genes display asymmetric methylation profiles between male and female gametes that make two gametes distinct and thus required for monoallelic expression after fertilization (Constancia et al., 2004; Kaneda et al., 2004) These elements are referred as differential methylated regions (DMR) which are CpG-rich sequences or transposons/repeat derived products

In plants, the homologous cis-elements have been found in the promoters of several imprinted genes FWA is silenced by MET1 in adult tissues but maternally expressed

in the endosperm A pair of direct repeats derived from SHORT INTERSPERSED

NUCLEAR ELEMENTS (SINE) was found in the FWA promoter region (Lippman et

al., 2004) Bisulfite sequencing revealed that these repeats were hypermethylated in adult tissues and the embryo but hypomethylated in the endosperm (Kinoshita et al.,

2004) Moreover, the repeats have been shown to recruit the de novo DNA methylation machinery to silence FWA transgene The FWA transgene carrying the

direct repeats also displayed a endosperm specific expression and is imprinted in a manner similar to the endogenous gene (Kinoshita et al., 2007) Taken together, the

results suggested that SINE related repeats could serve as the DMR to mark FWA paternal and maternal differently and lead to monoallelic expression of FWA

The methylated cis-elements were mapped into the promoter regions of FIS2 and

MPC (Fitz Gerald et al., 2009; Jullien et al., 2006a; Tiwari et al., 2008) Both of them

were silenced by MET1 and loss of MET1 function caused reduced DNA methylation

on their cis-elements These results thus demonstrate that silencing of FIS2 and MPC

are likely linked with the methylation of cis-element However, the importance of

these elements in establishing the imprint of FIS2 and MPC needs to be confirmed

25

The cis-elements are not only found in the 5’ promoter region but also located at the

region downstream of 3’ end of the coding sequence and could function as distance regulatory elements for regulating gene expression Few examples of the long distance cis-elements of plant imprinted genes will be discussed in the next section

long-2.3.2 Evidence for imprinting regulation by long distance elements

Long distance regulatory elements are essential in mammalian imprinting regulation

(Bartolomei, 2009) In plants a comparable regulatory mechanism affects PHE1

imprinting (Makarevich et al., 2008) The mechanism responsible for removal of the

silencing H3K27 methylation marks from the PHE1 locus in sperm cells, despite the

presence of functional PcG, remains unknown However, MET1 also appears to

regulate PHE1 imprinting (Makarevich et al., 2008) A methylated repeat region (DMR) is located 2.2Kb downstream of PHE1 coding sequence and its methylation is required for the maintenance of the silenced state of PHE1 maternal allele (Villar et

al., 2009) (Figure1-8A and 1-9)

Similarly, MEA coding region precedes a methylated 3’ cis element called ISR (MEDEA- INTERSTITIAL SUBTELOMERIC REPEATS) and comprises 7 tandem repeats (Figure 1-8B) The ISR is demethylated on MEA maternal allele and could play essential role in MEA imprinting regulation It is possible that the methylated MEA-ISR prevents removal of the silencing H3K27 methylation marks from the MEA locus Thus demethylation of MEA-ISR by DME would be a step required for MEA

activation in the central cell

26

Figure 1.8 Long distance cis-elements of imprinted genes

(A)Triple tandem repeats were found at 2kb downstream of PHE1 The repeat region

is methylated and required for PHE1 imprinting (see Figure 1-9) (B) MEA-ISR located at 5kb downstream of MEA contains seven repeats which are targeted by RdDM pathway MEA-ISR is demethylated in maternal MEA allele; however, it is unlikely that methylated MEA-ISR is required for silencing of paternal MEA allele

PRC2 complex acts to silence paternal MEA alleles in pollen and endosperm;

however, the requirement of DNA demethylation for the expression of maternal MEA suggests possible function of MEA-ISR in controlling MEA imprinting

Figure 1-9 Regulation of PHERES1 imprinting

In central cell, the paternally expressed imprinted gene PHERES1 (PHE1) is silenced

by Polycomb Group dependent histone methylation H3K27me3 In sperm cells

H3K27me3 marks are removed by yet unknown mechanisms A cis element (DMR) located at 2.3Kb downstream the coding sequence of PHE1 contains repeats, which

are methylated on cytosine residues (coloured lollipops) and in sperm cells The removal of cytosine methylation on the maternal allele in the central cell appears

conditional to the silencing of maternal PHE1 and proper maintenance of the imprinted status of PHE1 in endosperm

27

2.4 Genomic imprinting and RNA-directed DNA methylation

Maintenance of DNA methylation at cis-elements defines the mono-expression of

MET1-dependent imprinted genes However, a major question in the genomic imprinting research is how the imprinted genes are initially methylated In plants, it

has been shown that unmodified DNA could be de novo methylated by RNA-directed

DNA methylation (RdDM) pathway which is triggered by small interfering RNAs (siRNAs) produced from the corresponding DNA loci Forward and reverse genetics have identified several essential components for RdDM pathway and also revealed that the pathway is specialized in the plant kingdoms RNA-DIRECTED RNA POLYMERASE2 (RDR2), DICER-LIKE 3 (DCL3), and DNA-DEPENDENT RNA POLYMERASE 4 (POLIV) are together required for generating 24nt siRNAs from transgenes or retro-elements (Herr et al., 2005; Onodera et al., 2005; Pontier et al., 2005) Whereas ARGONAUTE 4 (AGO4), DEFECTIVE IN RNA-DIRECTED DNA METHYLATION (DRD1), DOMAIN REARRANGED METHYLTRANSFERASE

2 (DRM2) and DNA-DEPENDENT RNA POLYMERASE 5 (POLV) participate in the DNA methylation step at the target loci Two plant-specific RNA polymerase PolIV and PolV are differentiated by the biggest subunits NUCLEAR RNA POLYMERASE D (NRPD1a) and NUCLEAR RNA POLYMERASE E (NRPD1b), respectively However, they share the same second biggest subunit NUCLEAR RNA POLYMERSE D2 (NRPD2a) (Cao and Jacobsen, 2002; Herr et al., 2005; Zilberman

et al., 2004) It is suggested that POLIV (NRPD1a/NRPD2) is directly or indirectly required for producing single stranded RNA RDR2 is involved to convert the ssRNA into the double stranded RNAs (dsRNAs) which are spliced by DCL3 to produce 24nt siRNAs PolV (NRPD1b/NRPD2) acts downstream to guide the AGO4-siRNA

28

complex together with the de novo methyltransferase DRM2 to methylate target DNA

at all cytosine context including CG, CHG and CHH RdDM has been well established in targeting and silencing of retro-elements, virus-derived sequences, transcribed inverted repeats and transgenes (Borsani et al., 2005; Herr et al., 2005) However, a few protein coding genes have been identified as a target of RdDM The

presence of tandem repeats in cis control elements of FWA and PHE1suggested that they are potentially subject to RdDM pathway SINE-related tandem repeats in FWA

promoter region have been shown to generate siRNAs, which recruit the RdDM machinery to methylate at the repeats (Chan et al., 2004; Kinoshita et al., 2007)

PHE1 contains triple repeats in the cis-regulatory region locating at 2.2 Kb

downstream of the coding region (Villar et al., 2009) Methylation of that cis-region

by DRM2 is necessary for PHE1 imprinting These results demonstrated the involvement of RdDM in regulating genomic imprinting Nonetheless, mutants for de

novo methyltransferase DRM2 and other RdDM components did not show the

reactivation of either FWA or PHE1 in the adult tissues and male gametes; therefore did not disturb the FWA and PHE imprinting (Jullien et al., 2006a; Kinoshita et al.,

2004; Kohler et al., 2005) Those results question the function of de novo DNA

methylation by RdDM in regulating genomic imprinting in Arabidopsis