Promoter Variation and Transcript Divergence in Brassicaceae Lineages of FLOWERING LOCUS T

Brassica napus (AACC, 2n = 38), an oil crop of world-wide importance, originated from interspecific hybridization of B. rapa (AA, 2n = 20) and B. oleracea (CC, 2n = 18), and has six FLOWERING LOCUS T (FT) paralogues. Two located on the homeologous chromosomes A2 and C2 arose from a lineage distinct from four located on A7 and C6. A set of three conserved blocks A, B and C, which were found to be essential for FT activation by CONSTANS (CO) in Arabidopsis, was identified within the FT upstream region in B. napus and its progenitor diploids. However, on chromosome C2, insertion of a DNA transposable element (TE) and a retro-element in FT upstream blocks A and B contributed to significant structural divergence between the A and C genome orthologues. Phylogenetic analysis of upstream block A indicated the conserved evolutionary relationships of distinct FT genes within Brassicaceae. We conclude that the ancient At-α whole genome duplication contributed to distinct ancestral lineages for this key adaptive gene, which co-exist within the same genus. FT-A2 was found to be transcribed in all leaf samples from different developmental stages in both B. rapa and B. napus, whereas FT-C2 was not transcribed in either B. napus or B. oleracea. Silencing of FT-C2 appeared to result from TE insertion and consequent high levels of cytosine methylation in TE sequences within upstream block A. Interestingly, FT-A7/C6 paralogues were specifically silenced in winter type B. napus but abundantly expressed in spring type cultivars under vernalization-free conditions. Motif prediction indicated the presence of two CO protein binding sites within all Brassica block A and additional sites for FT activation in block C. We propose that the ancestral whole genome duplications have contributed to more complex mechanisms of floral regulation and niche adaptation in Brassica compared to Arabidopsis.


Introduction
Initiation of flowering is a key to the reproduction, evolution and survival of plants, and is determined by complex genetic pathways interacting with environmental and developmental signals. These are mediated through primary photoperiod, vernalization, gibberellin and autonomous pathways [1][2][3]. In Arabidopsis, FLOWERING LOCUS T (FT) expression causes early flowering independent of environmental and endogenous signals, whereas loss of FT function results in late flowering under long day (LD) conditions [4,5]. FT plays a central and indispensable role for floral induction, as an integrator of different pathways [6]. CO binds a unique sequence element containing the TGTG (N2-3) ATG motif within the FT promoter to induce expression in Arabidopsis under long day conditions [7][8][9], whereas the FLOW-ERING LOCUS C (FLC) protein binds directly to the CArG box of FT intron 1 to repress transcription [10]. Three highly conserved blocks, A, B and C were found within an approximate 5 kb upstream promoter region, and the proximal block A and distant block C were found to be essential for Arabidopsis FT activation by CO binding [11]. Although the FT protein is expressed in leaves it is mobile within the plant, combining with a bZIP transcription factor FD to form a complex of FT/FD heterodimer in the shoot apical meristem. This then activates the floral meristem identity genes APETALA 1 (AP1) and FRUITFUL (FUL) to promote flowering [7,8,12,13].
The cultivated Brassica species belong to the monophyletic Brassiceae tribe within the Brassicaceae family [14]. Among the six Brassica crops in the U-triangle [15], B. carinata (2n = 34, BBCC), B. juncea (2n = 36, AABB) and B. napus (2n = 38, AACC) are allotetraploids, which originated from pairwise hybridization of the three diploid species, B. nigra (2n = 16, BB), B. oleracea (2n = 18, CC) and B. rapa (2n = 20, AA). Comparative mapping have demonstrated the presence of numerous chromosomal regions homologous to the Arabidopsis genome, which are effectively triplicated within the diploid Brassica species, consistent with a common hexaploid ancestor in the evolution of Brassica and the tribe Brassiceae [16][17][18][19]. In contrast, partial and whole genome comparative sequence analysis indicates that gene content is variable within paralogous regions, with interspersed gene losses and insertions compared with Arabidopsis [19][20][21][22]. Interestingly, less than 10% of coding sequences of predicted gene models from Arabidopsis were recently found to be retained as systemic orthologues in each of the triplicated A subgenomes (Brassica rapa Genome Sequencing Project Consortium [23]).
Transposable elements (TEs) are ubiquitous amongst eukaryotic genomes. Of the most prevalent classes, retroelements and DNA elements [24] are believed to be major drivers of genome and gene evolution [25][26][27]. In plants, most genome size variability is associated with differences in repetitive DNA content, primarily ascribed to differential amplification of TEs [28]. The CACTA element of the En/Spm transposon system was first isolated and characterized at the molecular level in maize [29], and is estimated to be the most abundant DNA transposon family in B. oleracea [30]. B. oleracea transposon 1 (Bot1), a C genome-specific CACTA element, is believed to play a major role in the recent divergence of the B. rapa and B. oleracea genomes [31]. In Arabidopsis, the major effect of TE DNA methylation is to silence transposition. However, methylated sequences may also affect the transcription of the flanking genes, typically reducing their expression [32][33][34].
We previously identified six BnFT paralogues in B. napus [35], and showed that four of these were located within inverted duplicated blocks of homoeologous sections of chromosomes A7 and C6 (BnC6.FT.a and BnC6.FT.b; BnA7.FT.a and BnA7.FT.b). The remaining two paralogues were located on homoeologous chromosomes A2 and C2 (BnA2.FT and BnC2.FT) with deletion or variation of a putative CArG box for the FLC protein binding site that had been identified in Arabidopsis within intron 1 [10]. The phylogenetic tree based on the coding sequence clearly demonstrates that BnFT paralogues share a common ancestor with Arabidopsis FT, but had undergone triplication events consistent with the hypothesis of a common hexaploid ancestor for the Brassica species.
Here, we report the isolation and comparison of the FT promoter sequences from A2 and C2 that have a common lineage, and explore their evolutionary history both within the genus and the Brassicaceae family. FT transcripts from the distinct A2/C2 and A7/C6 lineages were then analyzed from a sample of eleven Brassica diploid and tetraploid cultivars at different developmental stages and environments, revealing divergence of expression patterns. Finally, the potential regulatory mechanisms, including DNA methylation, are further explored and discussed.

Results
Characterization of homoeologous FT genes on Brassica A2/C2 chromosomes Our previous results indicated that Brassica FT orthologues on chromosome A2 and C2 arose from a common lineage [35]. We isolated the complete promoter region of FT-A2 and partial promoter of FT-C2 from two BACs of B. napus Tapidor DH, JBnB045N08 and JBnB006F10. Based on the established upstream blocks A, B and C required for FT activation in Arabidopsis [11], we aligned the corresponding upstream regions of FT-A2 and FT-C2 from B. napus, FT-A2 from B. rapa BAC KBrB070J11 and FT-C2 from B. oleracea Scaffold000001 (unpublished data, available upon request from liusy@oilcrops.cn). This demonstrated the conservation of these blocks in the FT upstream regions of the Brassica homoeologous chromosomes A2 and C2 ( Figure 1A). Interestingly, a CACTA element of approximately 6 kb was found inserted within FT-C2 block A, and a 5.2 kb retrotransposon within block B of the same upstream region ( Figure 1A). The sequences of four promoters are available in NCBI Genbank (accessions JX193765 (FT-A2), JX193766 (FT-C2), JX193767 (FT-C6a) and JX193768 (FT-C6b)).
We then analyzed micro-synteny in the genome segments containing FT-A2 and FT-C2 from B. rapa, B. oleracea and B. napus, as well as the orthologous region from Arabidopsis. This indicated that TEs were more abundant in the neighboring region of FT-C2 than the homologous region of FT-A2. Unfortunately, we had limited access to data from the corresponding regions of FT-C2 in B. napus compared with B. oleracea ( Figure 1B). However, we did identify potential hot spots for the insertion of TEs within the intergenic region between orthologues of Atlg65290 and Atlg65295 on Brassica chromosome C2 ( Figure 1B).
We designed specific primer sets to distinguish the upstream regions containing the transposon and retrotransposon within block A and B from the A and C genome. These were used for PCR amplification of samples from a range of accessions, including B. rapa (17), B.oleracea (17), B. napus (51), B. nigra (1), B. juncea (8), B. carinata (6) and 9 species representing a range of taxa from the Brassicaceae. Based on the evidence from these marker assays, TEs were only detected within the Brassica C genome, and no insertion was found in the homologous regions of Brassica A and B genomes, nor from other Brassicaceae ( Table 1). We concluded that these transposon insertions most likely occurred after the divergence of the Brassica A and C genomes.

Evolutionary analysis of FT upstream block A in Brassicaceae
The sequences of FT upstream block A from a sample of diploid and tetraploid Brassica cultivars were determined and aligned, based on use of specific primer pairs for FT-A2 and FT-C2. The derived phylogenetic tree indicates that most B. napus and B. juncea accessions share high sequence similarity for block A with B. rapa, apart from two B. juncea accessions, Bj-6 and Bj-7 ( Figure 2A). Accessions of B. juncea and B. carinata containing sequences of B genome origin were clearly allocated to two distinct groups, with only one variety of B. nigra grouped with B. juncea ( Figure 2B). Three distinct groups were identified in the C genome phylogenetic tree, one including only B. oleracea and B. napus, and the other two including B. oleracea, B. napus and B. carinata ( Figure 2C). This is consistent with the FT-A2/C2 lineage being conserved during the allopolyploidisaton of Brassica species.
In order to explore the wider evolutionary relationships of FT genes within the Brassicaceae, we isolated the sequences of upstream block A from 9 representative genera, using the specific primer pairs for the FT-A2 orthologues. The block A sequences of the FT-A2/C2 lineage from the diploid and tetraploid Brassica species and representative Brassicaceae species were aligned with those of Arabidopsis FT, BnC6.FT.a and BnC6.FT.b. There is a clear differentiation of block A within the inverted duplication blocks of the FT-A7/C6 and FT-A2/C2 lineages. The Brassica FT-A7/C6 genes had high identity with that of Isatis indigotica and Arabidopsis ( Figure 2D), whilst the Brassica FT-A2 paralogue had higher identity with that of Orychophragmus violaceus, Rorippa indica, Capsella bursa-pastoris, Diplotaxis tenuifolia, Crambe abyssinica and Sinapi arvensis, than the corresponding Brassica B lineage ( Figure 2D). These results strongly suggest that the distinct lineages observed within the Brassica genomes arose prior to the ancestral whole genome triplication events.

Differential expression and DNA methylation analysis of FT genes in Brassica diploids and tetraploids
Since we had observed strong evidence for lineage-specific variation in upstream sequences within the paralogous FT genes, we then explored potential consequences on the pattern of gene expression. We designed specific primer sets to detect transcription from each of the FT-A2/FT-C2 and the FT-A7/C6 paralogues, using semi-quantitative RT-PCR. Within TapidorDH (B. napus, winter type DH line) and Ningyou7DH (B. napus, semi-winter type DH line) FT-A2 was transcribed in all leaf samples from different developmental stages and different photoperiod treatments, whereas FT-C2 was not transcribed ( Figure 3A). Moreover, FT-A2 was found to be transcribed in leaf 2 from three other winter type cultivars, and three spring type cultivars under vernalizationfree conditions, whilst FT-C2 was not ( Figure 3B). Interestingly, under the same conditions FT-A7/C6 was not transcribed in leaf 2 of TapidorDH and three other winter type cultivars, but was transcribed in Ningyou7DH and three other spring type cultivars ( Figure 3A-B).
We then checked FT transcription in leaves from B. rapa and B. oleracea at three different developmental stages without vernalization. FT-A2 was transcribed in all samples from two crop forms of B. rapa, yellow sarson (R-o-18) and Chinese cabbage (Chiifu-401), whereas FT-C2 was not transcribed in B. oleracea var. alboglabra, where TEs were inserted in block A and B ( Figure 3C). The FT-A7/C6 was transcribed in leaves from all diploid accessions, although the transcript level from the C genome appeared to be higher than that from the A genome ( Figure 3C).
We were interested to determine whether the TE insertion in FT-C2 had a significant effect on the neighbouring DNA methylation status. Cytosine methylation analysis of bisulphite treated DNA from leaves of diploid and tetraploid accessions focused on target regions within the FT upstream blocks A from C2 and A2 lineages ( Figure 4A). We found high levels of cytosine methylation within the TE sequence of ten randomly selected clones, and low levels in flanking sequence for FT-C2 block A from B. napus TapidorDH and Ningyou7DH, as well as from B. oleracea var. alboglabra ( Figure 4B). However, the corresponding FT-A2 upstream block A that lacked the TE insertions was free of cytosine methylation for all developmental stages tested in B. napus TapidorDH, Ningyou7DH and B. rapa R-o-18 ( Figure 4B).

Regulatory motif analysis of FT genes in Brassica
Comparison of promoter regions from FT-A2/C2 and FT-A7/ C6 lineages indicated that the latter were shorter, and lacked the upstream block B ( Figure 5). This strongly suggests that blocks A and C are more essential for the activation of Brassica FT genes. Block A appears to be highly conserved across different Brassicaceae genera ( Figure 6A). The ability of CO to promote FT expression is well established and involves binding a consensus TGTG (N2-3) ATG motif within the FT promoter [4,9]. We are now able to distinguish three distinct CO protein binding site variants within block A: type I = ''ATTGTGGTTATGATT'' in Arabidopsis, type II = ''ATTGTGGTGATGAGT'' in the Brassica FT-A2/C2 lineage, and type III = ''ATTGCGGTGATGAGC'' within the inverted duplication blocks of Brassica FT-A7/C6 lineage ( Figure 6A). It is possible that these sequence variants for the binding site correspond to the multiple CO paralogues in B. napus [36,37]. In contrast to the CO protein binding site, the putative TATA-box is more conserved within Brassciaceae, with only one base variation found in Arabidopsis ( Figure 6A).
Block C has been proposed as being another essential region for FT activation in Arabidopsis and Brassica. We therefore isolated the upstream block C of FT-A2/C2 homoeologous lineage from different species within the Brassicaceae and compared them with the corresponding region from Arabidopsis FT, B. napus FT-C6.a and FT-C6.b and B. rapa FT-A7.a. The resulting alignment demonstrated that block C is highly conserved across the Brassicaceae, although consistently located between 4.5 and 7.5 kb upstream of the start codon. We were able to designate a series of 10-30 bp regions, highly conserved within the Brassicaceae, as potential functional motifs within block C. Of seven sites (designated C-1to C-7) identified, six were universal within Brassicaceae, and another (C-2) specific to the Brassica FT-A2/ C2 lineage ( Figure 6B). This comparative taxonomic analysis provides strong circumstantial evidence for their role as potential motifs for FT activation.

Discussion
Structure and diversification of the Brassica FT-A2/C2 promoter We previously established that all extant Brassica FT genes share a common ancestry with Arabidopsis FT [35]. Here, we isolated and aligned promoter sequences from the diploid and amphidiploid representatives of the FT-A2/C2 lineage and demonstrated conservation of upstream blocks A, B and C previously identified in Arabidopsis [11] as being of key functional significance for activation of FT transcription, and hence flowering. However, since the divergence of the B. rapa A and B. oleracea C genomes around 4 MYA [20,38], two class TEs have inserted specifically within FT-C2 upstream blocks A and B. Our micro-synteny analysis indicated that TE insertion appears generally to be more abundant in the vicinity of B. oleracea FT-C2 of the homoeologous   Figure 1B) which are more abundant in the C genome, as revealed by homoeologous BAC sequencing [38,39]. The presence of a CACTA element within FT-C2 block A appears to have contributed to non-functionalization of this key floral regulator. CACTA element insertion leading to gene function diversification is common in plants [40][41][42][43]. For example, in soybean, a 20.5 kb insertion in the flavonoid 39-hydroxylase (F39H) gene led to a new allele conferring a novel phenotype of stable gray trichome color. The 20.5 kb insertion has the molecular structure of a putative autonomous transposon of the CACTA family [42].

Conservation of FT upstream block A within the Brassicaceae
FT genes are located in genomic regions corresponding to Arabidopsis chromosome 1 block E [35]. In Brassica species, extensive comparative genomic studies have shown that copies of block E are found on chromosomes A2, B2, C2, A7, B7 and C6 [19,[44][45][46]. Since block E is inversely duplicated on A7 and C6 [35,47] and there is extensive homoeology among the two sets of linkage groups A2/B2/C2 and A7/B7/C6 [45], we were confident that the large block E on B7 is also inversely duplicated. The presence of two FT lineages could be inferred within the B genome, with FT-A2/C2/B2 representing homeologues from one lineage and FT-A7/B7/C6 from the other.
The sequences of FT-A2/C2 upstream block A were grouped according to genomic content of Brassica diploid and tetraploid species within the U triangle (Figure 2A-C), which suggests that the FT-A2/C2 lineage was conserved during the process of the amphidiploid formation and domestication. A wider phylogenetic analysis of FT-A2/C2 and FT-A7/C6 upstream block A, that included a representative range of Brassicaceae species, demonstrated the clear distinction between the different groups ( Figure 2D), further confirming the independent FT lineages. It is particularly surprising that block A of I. indigotica clearly grouped with the FT-A7/C6 lineage even though the sequence was isolated with A2 specific primers and the FT-A2/C2 lineage cluster was  [14,48]. Although there is no ready explanation for these conflicting interpretations, it is recognized that genome evolution of the Brassiceae and related species is greatly complicated by the process of genome duplication, hybridization, and polyploidy. This makes it difficult to give a clearly evolutionary route for every genus and species [18,49]. Nevertheless, phylogenetic analysis of block A did uncover some of the ancient relationships between FT genes, and highlighted the coexistence within a number of species and genera of paralogues originating from different ancestral lineages. These results are consistent with the pattern of ancestral duplications identified from whole genome analysis, which suggests that Arabidopsis, and probably all core Brassicaceae taxa, experienced three ancient whole-genome duplication (WGD) events leading to paleopolyploidy [50]. The most recent At-a (,40 MYA) duplication was considered to explain the drastically increased species number in the core Brassicaceae [51]. The paleopolyploid At-a duplication was followed by later lineagespecific mesopolyploid WGD events. In Brassicaceae, at least four independent lineage-specific WGD events have been revealed and the whole genome triplication (Br-a) in Brassica was first proposed and then shown to have occurred before the radiation of the whole tribe Brassiceae [16,18,19,52]. This triplication event further complicates analysis of the evolutionary relationships between different genera and species within the tribe Brassiceae, as shown by our results above.

Transcript differentiation amongst Brassica FT genes
Duplicated genes may undergo three primary evolutionary fates, including silencing, sub-functionalization and neo-functionalization [53][54][55]. Within polyploids, and in particular allopolyploids, a wider repertoire may be available in terms of increased variation in dosage-regulated gene expression, altered regulatory interactions, and rapid genetic and epigenetic changes [56][57][58]. The ability to detect relative expression for all six Brassica FT copies is challenging due to high sequence similarity of transcripts, especially for the four FT paralogues within the duplication blocks of A7/C6 [35]. The lack of transcripts for FT-C2 in all accessions including B. napus and B. oleracea contrasted with the constitutive expression of FT-A2 in B. rapa and B. napus ( Figure 3A-C). Moreover, whereas the four FT genes within the duplication block of A7/C6 were all silenced prior to vernalization in winter type B. napus, they were expressed in at least some spring type varieties ( Figure 3A-B). These different expression patterns of FT-A2 might be attributed to the deletion or variation of FLC protein binding site within intron 1 of A2/C2 origin [10,35], which would result in constitutive expression of FT-A2. However, silencing of FT-C2 may have resulted from TE insertion and consequent conserved high levels of DNA methylation in the TE and flanking sequences which are most found at CG and CHG sites. So, the TE itself and DNA methylation appears to have generated a functional pseudogene FT-C2, given the absence of expression in all materials analyzed. TEs are recognized as a primary target of cytosine methylation in eukaryotes, where it serves primarily to silence these repeat sequences [59]. DNA methylation is able to silence genes by blocking transcription initiation either by preventing protein binding or as a consequence of DNA methylation-induced chromatin remodeling [60]. In plants, TE methylation may occur adjacent to upstream regions where TEs have inserted, with 'spreading' of methylation then affecting cisregulation, such as occurs in tomato, with the CNR gene affecting fruit ripening [61]. Although expression of FT-A7/C6 in B. oleracea appeared more abundant than in B. rapa ( Figure 3C), and may compensate for the silencing of FT-C2, this requires further investigation. The difference in transcript abundance for FT-A7/ C6 between winter and spring type B. napus under vernalizationfree conditions may be caused by different control mechanisms for flowering within these two crop types, and in particular the interactions between FLC and FT. However, to clarify this, it is essential to sample leaf at multiple developmental time points from plants subject to vernalising and non-vernalising conditions.
Recent characterization of 12 GPAT4 (sn-glycerol-3-phosphate acyltransferase 4) and seven PSY (Phytoene synthase) paralogues in B. napus and its two progenitors found that all were expressed, but with tissue specific expression exhibiting overlapping redundancy and signs of sub-functionalization [62,63]. In contrast, the expression patterns of FT genes differed with developmental stage and in different species/cultivars, consistent with selection for   winter and spring B. napus crop types. Interestingly, FT-A2 was constitutively expressed in all materials and developmental stages tested, and notably in winter-type B. napus before vernalization. FT genes have recently been found not only to regulate flowering but stomatal opening in Arabidopsis, vegetative growth in perennial poplar and storage organ formation in potato [64][65][66]. In sugar beet, two paralogues of FT have evolved antagonistic roles in the control of flowering, with the one functionally conserved with Arabidopsis FT being essential for flowering. However, the second paralogue represses flowering, with its down-regulation crucial for the vernalization response [67]. It is thus reasonable to suppose that the products of FT-A2 may participate in other physiological processes at least in winter-type B. napus, alternatively still in regulating flowering but in a different manner.

Potential FT regulatory motif e in Brassicaceae
Three conserved blocks in the promoter region of FT have been found to be essential for activation by CO in Arabidopsis, with block A subsequently found as a cis-element for promoting gene expression, block B probably enhancing FT expression in response to blue light [68] and block C as a potential enhancer for longdistance genomic regulation in conjunction with CO [11]. Interestingly, we did not detect block B in the Brassica FT-A7/ C6 lineage although these FT genes could be expressed normally ( Figure 3A-B), indicating that block B is not required for Brassica FT activation. This may be explained by the shade avoidance regulation in Arabidopsis where the blue light is an important control of the shade avoidance syndrome responses [69]. However, for Brassicas this may not be an issue, given the ecological niches it occupies compared with Arabidopsis. We found two potential motifs for CO protein binding in Brassica block A ( Figure 6A), each associated with the distinct FT-A2/C2 and FT-A7/C6 lineages, which may help to distinguish multiple CO paralogues in the Brassica genomes. Recently, another transcription factor, PHYTOCHROME INTERACTING FACTOR4 (PIF4), has been identified that binds directly to the FT promoter near the transcript start site to regulate FT expression in Arabidopsis [70]. Another level of regulation of FT by PIF4 may also exist in Brassica species which requires further investigation. We found that block C was significantly more conserved across the Brassica species and other species of Brassicaceae, even when located about 4.5-7.5 kb upstream of the FT start codon ( Figure 5), suggesting an important role in FT regulation. This finding suggests it will be important to determine whether the potential motifs within block C are combined with other regulatory proteins or small RNA, which then interact with CO to promote FT expression.

Plant materials and growth conditions
Ninety-nine plant accessions representing diploid and tetraploid Brassica and a further nine non-domesticated Brassicaceae species were sampled, with DNA isolated from the leaves of plants grown in field or glasshouse (Table S1)

Sequence acquisition and analysis
GenScan software optimized for Arabidopsis was used to identify protein-coding genes within TapidorDH BAC sequences [71]. B. rapa gene models were obtained from the Brassica database (BRAD) (http://brassicadb.org/brad/index.php) and coding sequences of B. oleracea and B. napus were compared to B. rapa and Arabidopsis genomes using Blastn and WU-BLAST against the BRAD and TIGR database (www.tigr.org/tdb/e2k1/bog1/). Transposable elements were predicted and located using ''Re-peatMasker'' (http://www.repeatmasker.org). Primers used for isolating the FT upstream blocks and estimating the insertion events are given in Table S2. For TE insertion detection, the primer sets specific to the A2 lineage are successful when there is no TE insertion in block A and block B, whereas the primers specific to the C2 lineage amplify products when TE insertions are present. ClustalW was used for multiple alignments of the block sequences (http://www.ebi.ac.uk/). Phylogenetic trees were constructed using MEGA version 4.1 [72]. Bootstraps with 500 replicates were performed to assess node support.

RNA isolation, reverse transcription and RT-PCR assays
Total RNA was extracted from leaves using the RNeasy Plant Mini Kit (Qiagen). cDNA was reverse-transcribed from total RNA (1 mg) with RevertAid TM First Strand cDNA Synthesis Kit (Fermentas) in a 20 ml reaction. Gene-specific primers were used for RT-PCR to detect Actin (30 cycles) and Brassica FT transcripts (30 cycles). The thermal cycling program was 94uC for 3 min followed by 30 cycles of 94uC for 30 s, annealing for 30 s, and extension at 72uC for 30 s, ending with a 10 min extension at 72uC. The primer sets used for RT-PCR detection are listed in Table S2.

Bisulphite sequencing
DNA was extracted from leaves using the DNeasy Plant Mini Kit (Qiagen). 450 ng genomic DNA was subjected to two successive treatments of sodium bisulphite conversion using the EpiTect Bisulphite kit (Qiagen) according to the manufacturer's instructions. The reaction was then purified once more using the PCR purification kit (Qiagen). Forward (F) and reverse (R) primers for bisulphite sequencing PCR were designed using Kismeth (http://katahdin.mssm.edu/kismeth) and listed in Table S2. The control assays of ATP1 [73] were used to ensure the complete bisulphite treatment. PCR products were cloned into the pMD18-T vector (TaKaRa), and 10 individual clones were sequenced. Percentage methylation (% C) was calculated as 1006C/(C+T).