Position-dependent effects of cytosine methylation on FWA expression in Arabidopsis thaliana

Gene expression can be modulated by epigenetic modifications to chromatin, and variants of the same locus distinguished by fixed, heritable epigenetic differences are known as epialleles. DNA methylation at cytosines is a prominent epigenetic modification, particularly in plant genomes, that can modulate gene expression. There are several examples where epialleles are associated with differentially methylated regions that affect the expression of overlapping or close-by genes. However, there are also many differentially methylated regions that have not been assigned a biological function despite their proximity to genes. We investigated the positional importance of DNA methylation at the FWA (FLOWERING WAGENINGEN) locus in Arabidopsis thaliana, a paradigm for stable epialleles. We show that cytosine methylation can be established not only over the well-characterized SINE-derived repeat elements that overlap with the transcription start site, but also in more distal promoter regions. FWA silencing, however, is most effective when methylation covers the transcription start site.


INTRODUCTION
Methylation of cytosine nucleotides in DNA is a prominent epigenetic mark in plant and animal genomes (1) . It is found mostly over transposons and repeat elements, consistent with its primary function in silencing their transcriptional activity in association with methylation at lysine 9 of histone 3 (H3K9) (2) . Mutants with defects in METHYLTRANSFERASE 1 ( MET1 ), encoding the major methyltransferase maintaining cytosine methylation in Arabidopsis thaliana, display various phenotypic abnormalities such as delayed flowering, dwarfism, and sterility, with increasing severity during successive rounds of inbreeding (3,4) . Genomes of met1 mutants are largely hypomethylated at CG dinucleotides that usually inherit cytosine methylation faithfully during DNA replication through MET1 activity (5) . Phenotypic abnormalities do, however, not require whole-genome changes in cytosine methylation, as a recent study describes how hypomethylation at a few select loci is sufficient to establish quantitative resistance to a pathogenic oomycete, Hyaloperonospora arabidopsidis (Hpa) (6) .
Transposable elements and repeats in the A. thaliana genome are mostly confined to heterochromatic regions, such as pericentromeres and telomeres. Others are distributed in euchromatic regions, and their proximity to protein-coding genes has been associated with constitutive or induced silencing of these proximal genes upon changes in cytosine methylation (7,8) . There are several examples of loci with alternative states of cytosine methylation and associated gene expression; the variants are known as epialleles. One example comes from the SDC ( SUPPRESSOR OF ddc ) locus with a direct tandem repeat in its promoter. When methylation at the repeat is absent, SDC is expressed, resulting in a dwarfed phenotype (9) .
Epialleles can also form following stress exposure; treatment with the 22-amino-acid peptide flg22, an immune-response inducing fragment of the bacterial flagellin protein, causes differential methylation of helitron-derived repeats lying within a 3 kb promoter region of the defense gene RESISTANCE METHYLATED GENE 1 ( RMG1 ), and ensuing activation of RMG1 expression (10) . Furthermore, cytosine methylation has been shown to modify gene expression when located farther away from the gene body, such as in the case of FLOWERING LOCUS T ( FT ), where methylation on two enhancers located 5 kb upstream and 1 kb downstream of the gene can repress transcriptional activity (11) . This was shown by experimentally targeting cytosine methylation to these enhancers using Inverted Repeat-Hairpins (IR-Hairpins ), which lead to the downregulation of FT expression and delayed flowering.
While these examples provide substantial evidence for methylation-dependent transcriptional changes, the requirements for cytosine methylation to exert this function is not well understood, as not all cytosine methylation, even when densely focused in methylated regions, triggers silencing of adjacent genes (12) .
To address such functional differences, we investigated the promoter of the well-characterized Arabidopsis thaliana fwa -1 epiallele. The FWA ( FLOWERING WAGENINGEN ) locus (At4G25530) harbors two sets of tandem repeats originating from a SINE3 retrotransposon (13) . These repeats overlap the promoter and the FWA transcribed sequence, and are covered by dense CG methylation in wild-type plants. Throughout vegetative development, FWA is transcriptionally inactive, and activated only in the female gametophyte and endosperm by maternal imprinting, when DNA methylation is erased (14) . Methylation at the repeats is also absent in fwa -1 epimutants, where the gene is constitutively active, which results in late flowering (15)(16)(17)(18) .
It is known that the presence of cytosine methylation at a specific position, the SINE -derived repeat elements, imposes transcriptional silencing on the entire locus. We made use of the unmethylated promoter in the fwa -1 epiallele and asked the reverse, whether FWA silencing can be triggered only by cytosine methylation at these repeats, or whether it can be similarly induced when cytosine methylation is artificially directed to other, non-repetitive, promoter elements.

Hairpins directing methylation to the FWA promoter
We chose three regions of 100 or 200 basepairs (bp) in length, located approximately 100, 500 and 700 bp upstream of the FWA transcription start site (Table 1 and Figure 1), and generated inverted repeat (IR)-hairpins (19) , intended to introduce cytosine methylation at these regions that are otherwise unmethylated in fwa-1 mutants.
For each region, we generated two IR-hairpins, differing in the orientation of sense and antisense sequences within the hairpin (Table 1, Figure 1, Methods).

Region-specific methylation effects on FWA activity
We transformed late-flowering fwa-1 mutants with IR-hairpin transgenes as described above (Table 1), and identified homozygous insertion lines to monitor flowering time as a proxy for

Establishment of cytosine methylation at distal promoter regions
As IR-hairpins derived from more distal regions of the FWA promoter did not induce sufficient FWA silencing to alter the timing of flowering, we asked whether cytosine methylation had been effectively established at the targeted sequences using whole-genome bisulfite sequencing.
As shown in Figure 5, all six transgenes efficiently introduced methylation in the targeted regions to comparable levels. The effect of the transgene on methylation was stable over at least three generations, and indicates that cytosine methylation at this locus can be introduced not only at the SINE -derived repeat elements. We further conclude that methylation at more distal promoter elements was less effective in FWA silencing, which may be due to the distance from the transcription start site.  (23) ). This pattern also holds true for the FWA locus,

DISCUSSION
where SINE -derived tandem repeats that overlap with the transcription start site are heavily methylated and silence the gene.
Transcription is initiated in the proximity of DNA elements providing a platform for the recruitment and assembly of polymerase II-containing complexes. Their binding is modulated by enhancers and repressors, which themselves may recognize distal sequence elements. It is likely that cytosine methylation over elements of polymerase docking may directly or indirectly hinder effective protein recruitment. When located in regions distal to the transcription start site, methylation-based silencing may be achieved by the inhibition of transcriptional enhancers (24) , and by altering short or long-distance chromatin interactions (25) , thus affecting accessibility to the transcription machinery. Such a mechanism may possibly account for our observations, that methylation targeted in Regions 2 and 3, located more than 500 bp upstream of the transcription start site, only moderately impacts FWA transcript accumulation.
On examining publicly available data from the PlantDHS Browser (26) , we observed that the 1-kb promoter region immediately upstream of the FWA transcription start, including all three regions targeted for methylation, has a uniform state of chromatin accessibility (Figure 6), not only when in the silent wild-type state, but also in the unsilenced state in ddm -1 mutants, which express FWA due to hypomethylation of the tandem repeats (27) . Therefore, it may be expected that one would see a similar transcriptional readout upon modulating different regions within the same inaccessible sequence. At least in the FWA locus, this is not the case.
Our results indicate that, while IR hairpin-induced methylation can be successfully introduced in the distal FWA promoter, this methylation can induce only moderate downregulation of FWA transcription, in contrast to very potent effects when methylating elements more proximal to the transcription start site. It will be interesting to examine the chromatin state and conformation of the FWA promoter in our transgenic lines, to understand whether they can explain the differential transcriptional downregulation that we observed.
This study was focused on the FWA locus; further investigation into differential promoter methylation at other genomic loci is required to dissect the mechanism behind methylation-dependent control on downstream transcription processes, and ultimately uncover its adaptive and biological function.

Cloning of inverted-repeat hairpin (IR-Hairpin) constructs
Three 100 bp -200 bp regions in the FWA promoter were chosen for IR-hairpin construction (Table 1 and Supplementary Table S1), and cloned in two ways: All 'A' constructs carry sequences in sense orientation relative to the open-reading frame of the FWA locus in the hairpin 5' arm, while the 'B' constructs are reversed.
Hairpins, including flanking attB gateway sites were synthesised (GeneArt), PCR-amplified and transferred to pDONR207 (Invitrogen) using BP Clonase II . Recombinant entry clones were subjected to LR Clonase II reaction with the pJawohl-ACT2 destination vector (28) and introduced into E.coli DH5 α (Invitrogen) cells by heat-shock transformation. Colonies carrying the hairpin construct in the correct orientation were verified by Sanger sequencing (oligonucleotide sequence provided in Supplementary Table S1). Table S4) were introduced into Agrobacterium tumefaciens strain GV3101(pMP90RK) (29) by electro-transformation and grown in selective LB medium. Arabidopsis thaliana fwa-1 mutants (16) were transformed with a floral dip protocol (30) . Transgenic seeds were selected with 1% BASTA ( Sigma-Aldrich).  blue) and W HB (white with ca. 25% blue), respectively a nd watered at 2-day intervals.

RT-qPCR
Ten plants belonging to each transgene were grown on soil in a randomized-block design, to reduce position effects in the growth chamber. Flowering time was recorded when the primary inflorescence meristem was approximately 1cm in height.

Bisulfite library preparation and sequencing
Genomic DNA was isolated using the DNeasy Plant Mini Kit (Qiagen) from pools of twenty 10-day old seedlings grown on the same MS-Agar plates as seedlings used for RT-qPCR. 100ng of genomic DNA was used to prepare Bisulfite libraries with the TruSeq Nano kit (Illumina, San Diego, CA, USA) according to the manufacturer's instructions, with the modifications used in (12) . The libraries were sequenced with a paired-end mode, at 125 million reads/library using an Illumina HiSeq3000 instrument.

Processing of sequenced bisulfite libraries
Raw sequencing reads were aligned using Bismark (default parameters; (32) and mapped to the A. thaliana (Landsberg erecta ) reference genome. Bam files generated after deduplication of reads were processed for identifying methylated cytosines using pipelines as previously described in (33) . Methylated cytosines in the FWA locus were loaded onto the EPIC-CoGe browser (34) for visualisation.