Epigenetic regulation of photoperiodic flowering

Abstract

The cytidine analogue 5-azacytidine, which causes DNA demethylation, induced flowering in the non-vernalization-requiring plants Perilla frutescens var. crispa, Silene armeria and Pharbitis nil (synonym Ipomoea nil) under non-inductive photoperiodic conditions, suggesting that the expression of photoperiodic flowering-related genes is regulated epigenetically by DNA methylation. The flowering state induced by DNA demethylation was not heritable. Changes in the genome-wide methylation state were examined by methylation-sensitive amplified fragment length polymorphism analysis. This analysis indicated that the DNA methylation state was altered by the photoperiodic condition. DNA demethylation also induced dwarfism, and the induced dwarfism of P. frutescens was heritable.

Introduction

Flowering is regulated by environmental factors, such as night length in photoperiodic flowering, coldness in vernalization and multiple stress factors in stress-induced flowering.1–3 In vernalization, the effect of low temperature experienced by seeds or young seedlings is actualized when the plants reach the adult phase. The previously induced effect of low temperature is transmitted mitotically during the plant's development. This mitotic stability suggests an epigenetic basis for vernalization. DNA methylation/demethylation is involved in the regulation of this type of epigenetic phenomena. Therefore, DNA demethylation was hypothesized to be involved in the regulation of vernalization. Five-azacytidine (azaC) prevents the transfer of methyl residues onto DNA molecules by replacing cytosine during DNA replication; this results in demethylation of the DNA. Treatment of vernalization-requiring late-flowering mutants of Arabidopsis thaliana and Thlaspi arvense with azaC induced precocious flowering.4 The induction of flowering by DNA demethylation was subsequently reported in winter wheat5,6 and Cichorium intybus.7 All of these plants had to be exposed to a low temperature to be induced to flower. DNA demethylation was thought to be involved in the regulatory mechanism of vernalization.8–15

The effect of a favorable photoperiodic cue does not generally last long. Therefore, photoperiodic flowering has not been thought to involve epigenetics. In the short-day (SD) plant Perilla frutescens var. crispa, however, photoperiodic induction has been characterized as an irreversible and indestructible phenomenon. That is, the flowering state does not revert to a vegetative state when the photoinduced plants are exposed to non-inductive long-day (LD) conditions. This longevity of the flowering state in P. frutescens is analogous to the stability of the cold-treatment effect in vernalization, and it suggests that the same mechanism is involved in the regulation of photoperiodic flowering in P. frutescens and vernalization-induced flowering. Therefore, we hypothesized that DNA demethylation might be involved in the mechanism regulating photoperiodic flowering in P. frutescens.

Induction of Flowering by azaC in P. frutescens

Seeds of red-leafed and green-leafed P. frutescens were treated with azaC and grown under LD conditions.16 Both varieties were induced to flower by azaC. Treatment of shoot apical meristems with azaC also induced flowering. Therefore, DNA demethylation may be involved in the regulatory mechanism of photoperiodic flowering in P. frutescens. This is the first finding that photoperiodic induction can be replaced by DNA demethylation.

Photoperiodic flowering consists of two major processes: flower induction, in which the flowering stimulus is generated in leaves, and flower evocation, in which the shoot apical meristem responds to the flowering stimulus transported from the leaves. If the azaC treatment induced flower evocation, this process would skip the preceding flower induction process. However, the flowering induced by azaC was delayed compared to that induced by SD.16 This suggests that the process induced by azaC was not flower evocation, but flower induction.

Genomic DNA was extracted from the leaves of azaC-treated and non-treated P. frutescens, and each DNA sample was digested with MspI and HpaII. These restriction enzymes recognize the sequence CCGG, but they barely cleave it if the cytosine is methylated. HpaII is more sensitive to methylation. Southern hybridization was performed with a 25S-18S rDNA intergenic spacer probe. The rDNA intergenic spacer region is known to be rich in CG sequences and therefore highly methylated. When the DNA samples were digested with HpaII, the azaC-treated sample gave signals in the lower molecular weight region in which no signals were detected in the control sample. This means that the azaC treatment promoted demethylation of the rDNA intergenic spacer region. This suggests that the genes involved in the regulation of flowering were also demethylated by the azaC treatment.

Induction of Flowering by azaC in Other Plant Species

We originally assumed that flowering induction by DNA demethylation is specific to plants whose photoinduced state is stable. The number of plant species whose flowering state is stable is quite small. Included are the above-mentioned P. frutescens, as well as Xanthium strumarium, Silene armeria and Bryophyllum daigremontianum. We examined the flower-inducing effect of azaC in X. strumarium and S. armeria, as well as in Pharbitis nil (synonym Ipomoea nil), Lemna paucicostata and Lemna gibba, whose flowering states are unstable.17

S. armeria was induced to flower by the treatment with azaC as expected. However, X. strumarium was not induced to flower by azaC treatment. Unexpectedly, P. nil was induced to flower by azaC. Three strains of L. paucicostata and L. gibba G3 grown in medium containing azaC were not induced to flower. It was reported from another laboratory that azaC promoted flowering in Linum usitatissimum, in which the flowering state is unstable.18 Thus, there is no correlation between the stability of photoperiodically induced flowering states and the flower-inducing effect of DNA demethylation.

The azaC treatment induced morphologically abnormal leaves in P. nil,19 but this abnormality was observed only in the first and second leaves. This suggests that the demethylated (and activated) genes may have been methylated and inactivated again. P. nil may have a DNA remethylation activity, and therefore the flowering state may revert to a vegetative state. This would explain why the flowering state induced by SD is unstable under LD.

Inheritance of the Flowering State Induced by azaC

Epigenetically induced modifications of gene expression are generally maintained through cell division during normal development. Early-flowering lines of L. usitatissimum have been induced by treatment of germinating seeds with azaC,20,21 suggesting that azaC can induce epimutations. However, in mammals, the methylation pattern of DNA is erased during gametogenesis.22 Therefore, epigenetically induced phenotypes are usually not heritable. The flower-promoting effect of vernalization is not inherited, and neither is the flowering induced by azaC in vernalization-requiring A. thaliana and T. arvense.4 Thus, the plants can respond to vernalization in each generation.23 It is natural that plants in each generation need to be vernalized to flower. Otherwise, the regulation of flowering by environmental cues would be impossible. Artificial DNA demethylation induced by exogenous chemicals can be reverted by DNA demethylation and a remethylation systems, and therefore its effect is not heritable. The establishment of early-flowering lines by azaC in L. usitatissimum may suggest that the flowering genes in this species are not under the regulation of a DNA demethylation and remethylation system. We studied the heritability of the flowering state induced by azaC in P. frutescens, S. armeria and P. nil.24,25

The P. frutescens plants were induced to flower by azaC treatment, and seeds were harvested from the plants that flowered in response to the azaC treatment (designated azaC progeny). The azaC progeny grew normally but did not flower under non-inductive LD conditions. The azaC progeny of S. armeria did not flower under non-inductive SD conditions either. In P. nil, the flowering state reverted to a vegetative state even within the treated generation. Thus, the flowering state induced by DNA demethylation was not heritable.

The methylation state of the rDNA intergenic spacer region in the azaC and SD progeny of P. frutescens was analyzed by DNA gel blot hybridization to determine whether the DNA demethylation induced by azaC was heritable.25 No differences were observed in the band patterns between DNA samples of the two progeny when the samples were digested with both MspI and HpaII, indicating that there was no difference in the DNA methylation state of the rDNA intergenic spacer region between the azaC and SD progeny. This result suggests that the demethylated rDNA intergenic spacer region was remethylated in the next generation. The remethylation system deactivates previously activated genes and thus allows those genes to be demethylated again in response to inductive photoperiodic conditions in each generation.

Changes in DNA Methylation State by Photoperiodic Conditions

An important question is whether the DNA methylation state was altered by inductive photoperiodic conditions. We examined genome-wide methylation changes by methylation sensitive-amplified fragment length polymorphism (MS-AFLP) analysis.24,25 The MS-AFLP technique was applied to vernalization-requiring winter wheat and A. thaliana to detect DNA demethylation.26,27 Genomic DNA was extracted from SD-treated cotyledons and LD control cotyledons of P. nil after the SD treatment and digested with MspI and HpaII. DNA fragments containing the methylated CCGG would be further fragmented by this enzyme if the CCGG was demethylated. Thus, MS-AFLP can detect differences in the methylation state. Assuming that the DNA is hypermethylated under LD conditions and hypomethylated under SD conditions, the fragments detected only in the LD sample are those demethylated by the SD treatment. The DNA fragments were amplified by randomly-selective PCR using 36 primer combinations. Analysis of the pattern detected by polyacrylamide gel electrophoresis revealed many polymorphic fragments. We performed a similar MS-AFLP analysis in P. frutescens. The pattern was altered depending on the photoperiodic conditions. These results suggest that the DNA methylation state was altered by photoperiodic conditions.

In the MS-AFLP analyses, fragments that were photoperiodic condition-specific, reproducible, sharp and easy to separate from neighboring fragments were selected and extracted from the gel to clone.24,25 A homology analysis indicated that the majority of the sequences were uncharacterized. However, one of the cloned fragments from P. nil was highly homologous with partial sequences encoding ribosomal proteins of Rhizobiaceous bacteria. Identification of the genes controlling flowering that are activated by DNA demethylation has so far been unsuccessful. However, we found that many of the cloned fragments from both species contained frequent CpG sequences which can be considered as CpG islands according to the Meth Primer program (www.urogene.org/methprim er/) with the criteria; island size >100, GC percent >50.0, and Obs/Exp >0.6. A CpG island is a region with a high G + C content and frequent CpG dinucleotides.28 The methylation of CpG islands correlates with gene expression.29 Although information about the plant CpG islands, especially in dicots, is limited,30 recent studies have suggested that the methylation in CpG island-like regions is associated with gene expression.30,31 It is possible that the CpG island-like regions are involved in the regulation of gene expression in photoperiodic flowering.

Induction of Dwarfism by azaC

AzaC treatment induced not only flowering but also dwarfism in P. frutescens and P. nil.16,17 Induction of dwarfism by azaC has been reported in rice and maize.32,33 The same result was obtained with zebularine, another DNA demethylating reagent.25 The dwarfism induced by zebularine was overcome by GA₃, suggesting that the induction of dwarfism by DNA demethylation may be mediated by gibberellins, although it was previously reported that the dwarfism induced by azaC was overcome by cytokinin in rice and maize.34,35

The dwarfism induced by azaC was heritable in P. frutescens24 as in rice and maize,28,29,36 however, the induced dwarfism in P. nil was not heritable.25 Stable inheritance of a phenotype induced by DNA demethylation was reported in the early flowering of L. usitatissimum18 and other plants.37 The floral symmetry induced by hypermethylation in Linaria vulgaris is also heritable.38 Normal development proceeds more or less epigenetically because all cells in an individual have the same genome. Therefore, the epigenetic changes must be reset to repeat normal development in the next generation, as mentioned above for flowering regulation. The non-specific demethylating activity of azaC may cause demethylation even in the regions that are not programmed to be demethylated, and abnormal demethylation may not revert, resulting in epigenetic inheritance. Accidental changes in DNA methylation can be transmitted through cell divisions leading to epiallelic variation.35 The dwarfism caused by azaC may be induced by such an accidental change, and it suggests that normal vegetative growth is not regulated by DNA methylation.

It is interesting to note that the flowering state induced in P. frutescens by azaC was not heritable but that dwarfism was.24 Thus, the relevant vegetative growth genes may not be under the regulation of the remethylation system that resets the methylation state of the flowering-related genes. How the remethylation system distinguishes the genes to be remethylated (flowering-related genes) from other genes (vegetative growth-related genes) is unknown. The RNA interference machinery is responsible for distinguishing between remethylatable and non-remethylatable sequences.39 Short interfering RNAs and microRNAs cause de novo DNA methylation (RNA-directed DNA methylation).40 The DNA methylation induced by RNA-directed DNA methylation is sequence-specific.41 Some imprinted genes have frequently repeated sequences that are associated with persistent DNA methylation.42 Similarly, flowering-related genes may have sequences that are preferentially demethylated and remethylated.