Abstract
Annual dormancy-growth cycle is a developmental and physiological process essential for the survival of temperate and boreal forests. Seasonal control of shoot growth in woody perennials requires specific genetic programs integrated with the environmental signals. The environmental-controlled mechanisms that regulate the shift between winter dormancy to growth promoting genetic program are still unknown. Here, we show that dynamics in genomic DNA methylation (gDNA) levels regulate dormancy-growth cycle in poplar. We proved that the reactivation of cell division in the apical shoot that lead bud break process in spring, is preceded by a progressive reduction of gDNA methylation in apex tissue. We also identified that the induction in apex tissue of a chilling-dependent poplar DEMETER-LIKE 10 (PtaDML10) DNA demethylase precedes shoot growth reactivation. Transgenic poplars showing down-regulation of PtaDML8/10 caused delayed bud break. Genome wide transcriptome and methylome analysis and data mining revealed the gene targets of active DML-dependent DNA demethylation genetically associated to bud break. These data point to a chilling dependent-DEMETER-like DNA demethylase controlling the genetic shift from winter dormancy to a condition that promotes shoot apical vegetative growth in poplar.
Introduction
Temperate and boreal perennials are distributed around the world in those regions in which four seasons are well defined based on their specific weather conditions along the year. Hence these tree species have developed complex molecular mechanism to adapt their annual cycle with the specific environmental conditions in each season, to guarantee their survival, sexual reproduction and seed set in adult trees. Winter dormancy is the strategy used by perennials in temperate and boreal regions to avoid cold damage. During winter dormancy, growth is inactivated in all meristems of the tree, i.e., the shoot apical meristem (SAM), the axiliary meristem, and the stem vascular cambium. Winter dormancy begins in autumn with an initial stage of ecodormancy, when the growth of the SAM and vascular cambium is arrested by short days (SD) and cold temperature. During this period, the tree forms autumn apical buds to protect the SAM and simultaneously develops moderate cold tolerance. The second stage is the endodormancy, when trees acquire maximum adaptation to the cold. These two states, ecodormancy and endodormancy, were described by (1). During this period, a chilling requirement needs to be fulfilled for growth to resume. Once the chilling period is fulfilled, the trees enter in the ecodormancy again, allowing bud break and the recovery of active growth under growth-promoting environmental conditions of spring. Annual cycle of temperate trees has been reviewed in (2).
Efforts are being focus on elucidating the winter dormancy control, since it determines several economically important traits in perennials, among them tree annual growth rate and fruit development. The main molecular players in dormancy regulation identified so far are: FLOWERING LOCUS T (FT) mediating the photoperiod signaling that leads to growth cessation at dormancy entry (2–4) and FD-LIKE1 (FDL1) and ABA INSENSITIVE 3 (ABI3) controlling bud maturation and cold acclimation in poplar (5). Maintenance of dormancy require the CENTRORADIALIS-LIKE1 (CENL1) regulator in poplar (6) and the regulation of the chromatin function by histone modifications associated to the function of DAM genes in peach (7, 8). At dormancy exit, EARLY BUD-BREAK1 (EBB1) is needed to promotes bud burst in poplar (9, 10).
Energy fluxes through the TCA cycle link the metabolic status of plants to the DNA methylation status (11). Cellular dynamics in NAD+/NADH ratio affects methylation status of plants (12). The reduction of cellular metabolism during the dormancy leads to a higher NAD+/NADH ratio that potentially would promote DNA methylation (11). According to this, in chestnut (Castanea sativa), the bud formation during winter dormancy and bud break during spring are accompanied by an increment and a decrease, respectively, of gDNA methylation (13). Similarly in apple (Malus × domestica) the transition from dormant bud to fruit set is accompanied by a progressive decrease of gDNA methylation (14). It has been proposed that the mechanisms of winter dormancy in trees and vernalization in herbaceous plants show similarities (8, 15). According to this, DNA hypomethylation resulted in early flowering or a reduced vernalization requirement to promote flowering in annual plants and to bolt tolerance in sugar beet (16–18).
DNA methylation refers to the addition of a methyl group in the fifth position of cytosine (5mC) and promotes changes in gene expression (19). DNA methylation occurs and is maintained by three DNA methyltransferases in plants that recognize different methylation contexts (CG, CHG and CHH, where H = A, C or T) (20). An important feature of DNA methylation is its reversibility, which can occur passively after de novo synthesis of unmethylated strands during DNA replication, or actively through the actions of enzymes that remove the methylated bases. In Arabidopsis there are four DEMETER-LIKE DNA demethylases (DMLs) (21). Active demethylation by DMLs enables the expression of genes in response to biotic and abiotic stresses in Arabidopsis (22, 23). Moreover, DMLs have been associated with developmental transitions during fruit ripening in tomato and nodule development in Medicago truncatula (24, 25).
Here, we addressed the role of genomic DNA methylation during dormancy-growth cycle regulation in poplar. We showed that the reactivation of shoot apical growth is preceded by reduced genomic DNA methylation in apex tissue while DEMETER-LIKE DNA demethylase 10 (PtaDML10) is induced at the apex after a chilling fulfilment during ecodormancy. Our functional analysis showed that an active DML-dependent DNA demethylation is necessary for poplar bud break. Through genome wide transcriptome and methylome analysis and data mining, we identified the gene targets for the active DML-dependent DNA demethylation genetically associated to bud break. These data point to DEMETER-mediated DNA demethylation in controlling environmentally-induced developmental stage transition that shift from winter dormancy to a condition that promotes shoot apical vegetative growth.
Results and Discussion
Bud break is preceded by a reduced genomic DNA methylation in poplar apex while PtaDML10 is induced during ecodormancy and bud break
To understand the temporal pattern of genomic DNA (gDNA) methylation during the transition from dormancy to growth, we profiled DNA methylation by high-resolution HPLC in poplar trees SAM growing under natural conditions, weekly from January to the time of bud break in April, (Fig. 1A). In this profile, a stage of gDNA hypermethylation in February was followed by a period of marked hypomethylation, with minimum levels detected on March 24. Subsequently, bud break (stages 1, 2 and 3 of apical bud development, Fig. 1 A and B), coincided with increased 5mC levels. Similar pattern of 5mC confer plasticity in environmentally-controlled responses such as vernalization in sugar beet (19). Thus, this particular gDNA methylation pattern is a common readout generated by the physiological state of the shoot apical cells during poplar dormancy and sugar beet vernalization.
To elucidate whether the decrease in gDNA methylation precedes or follows cell cycle reactivation, we examined the expression of D-type cyclin PtaCYCD6:1, a marker of growth reactivation in poplar apical meristem (26). PtaCYCD6:1 mRNA levels rose from March 17 until the last time point April 14 (bud stage 3, based on the scale established in (27, 28) (Fig. 1B) indicating that most likely this decrease in gDNA methylation level is produced by an active DNA demethylation mechanism, because it occurs previously to the reactivation of cell cycle that lead bud break.
This scenario led us to identify poplar DNA demethylases, through phylogenetic and protein sequence analysis. We identified three putative DMLs genes in poplar, PtaDML6, PtaDML8 and PtaDML10, containing all domains and amino acids conserved in DMLs, essential for active 5mC excision (Fig. S1). To elucidate if one of these poplar DMLs could be involved in the reduction of gDNA methylation that precedes bud break, we analysed their annual expression pattern (monthly) in stems, and their expression pattern during winter dormancy (weekly) in stems and apices. Annual expression patterns of the three genes in poplar stems revealed that only PtaDML8 and PtaDML10 showed highest levels during vegetative growth, with PtaDML8 expression peaking in June during the highest growth period and PtaDML10 expression peaking at the time of bud break (April) (Fig. 1C and Fig. S2A). In the expression profile for poplar stems and apices, PtaDML10 induction was observed in late winter dormancy (end of February) with maximum expression levels produced at bud break (April 1 and 9), while PtaDML8 expression remained constant in stems and apices (Fig. 1 D and E and Fig. S2 B and C). These observations are consistent with previous transcriptome analyses indicating the increased PtaDML10 expression during dormancy release, the transition from endo- to ecodormancy that precedes the bud break, in poplar (29).
Additionally, PtaDML10 mRNA was localized by FISH in poplar apices. The fluorescence signal was detected in the L1 layer and central SAM, and also in the tips of leaf primordia at bud break (April 14) (Fig. 1F). This spatio-temporal localization is similar to that reported for EARLY BUD-BREAK 1 (EBB1), a putative APETALA2/ethylene responsive transcription factor in poplar (30). EBB1 overexpression causes early bud break in poplar, and activates CYCD3 genes in Japanese pear, whereas its down-regulation delays this process in poplar (30, 31). Relatively higher expression of EBB1 in L1 indicates the importance of this layer in the reactivation of SAM growth in spring.
In order to dissect whether the activation of PtaDML10 and the reduction of gDNA methylation occur during endo- or ecodormancy stages, cuttings were taken from trees grown in their natural environment in both conditions, a set of cuttings taken from trees that had not fulfilled the chilling requirement (endodormant cuttings) and a set of cuttings from trees that did fulfil the chilling requirement (ecodormant cuttings). Both sets were transferred to growth promoting-conditions, long day (LD) at 22°C, in a growth chamber and phenological observations were made at 0, 6 and 12 days after transfer (Fig. 1G). 85% of the ecodormant apical buds on day 12 reached stages 1 or 2 (27, 28), indicating the initiation of bud break (Fig. 1G). The gDNA methylation level in the ecodormant apical buds was higher than in the prechilling ones, but decreased significantly 12 days after transfer to growth-promoting environment, when bud break process began (Fig. 1H). However, the gDNA methylation levels in endodormant apical buds did not show changes during the same period (Fig. 1H). Accordingly, PtaDML10 RNA levels remained very low in endodormant apices, while it expression was gradually induced in ecodormant apices, reaching the maximum expression when the bud break began (after 12 days under growth-promoting environment) (Fig. 1I). The resumption of growth in the ecodormant samples after 12 days was indicated by the expression of PtaCYCD6 (Fig. 1J). In summary, the results of this experiment showed: in endodormant apical buds there is not initiation of bud break after 12 days under growth promoting-conditions, there are neither changes in gDNA methylation nor induction of PtaDML10 during this period. In contrast, ecodormant apical buds initiated bud break after 12 days under growth-promoting environment, and this process is accompanied by both, a significant reduction of gDNA methylation level and a huge induction of PtaDML10 expression. All together indicate that dormancy release is necessary for PtaDML10 induction during the process of bud break, and suggest that it could be implicated in the gDNA demethylation that precedes bud break.
Collectively, the expression pattern suggests that PtaDML8 would play a role during the vegetative growth period, particularly in stem tissues, and that PtaDML10 could be implicated in the gDNA methylation reduction observed in apical buds previous to the reactivation of cell division that leads to the bud break process. Induction of PtaDML10 occurs four weeks earlier than PtaCYCD6:1, together with the higher induction of PtaDML10 mRNA in L1 at this time, suggesting its role in the reestablishment of growth after winter dormancy.
Active DML-dependent DNA demethylation is required for bud break
Functional studies were performed by generating PtaDML8/10 knock-down plants (KD) (Fig. S3). RT-PCR analyses showed that both DML genes were downregulated in the transgenic plants due to the high similarity existing between PtaDML8 and PtaDML10. We examined phenological variables in two lines showing low-level PtaDMLs expression, KD2 and KD5. When grown under LD and 22°C conditions, in vitro or glasshouse cultured transgenic plants were phenotypically similar to WT plants. Moreover, these lines showed no differences compared to WT in cessation of growth elongation and bud formation when transferred from LD to short day (SD) conditions (Fig. 2 A and B). These observations rule out a role of these DMLs in dormancy entry. WT and KD plants were subjected to 0, 4 and 8 week of SD conditions at 4°C and bud burst was monitored in LD conditions at 22°C using the scale shown in Fig. 2C (27, 28). Without chilling, t=0, there was no bud break, indicating that these two DMLs do not affect dormancy release. More significant differences in bud burst were observed after 4 weeks (Fig. 2D). Thus, lower levels of DMLs expression led to a significant delay in growth reactivation in the transgenic trees after dormancy, pointing to a role of these DMLs in bud break phenology. In addition, through HPLC of genomic DNA from dissected apices, 5mC levels 4 and 2% higher were detected in KD2 and KD5, respectively than in WT (Fig. 2E).
This loss of function studies confirmed that active 5mC demethylation is necessary for bud break. Even though PtaDML8/10 knock-down plants have reduced expression of PtaDML10 and PtaDML8, the earlier described spatio-temporal expression patterns suggest that PtaDML10 mainly participates in the chilling-dependent genomic DNA demethylation needed for successful bud break after winter dormancy.
Target genes of the active DML-dependent DNA demethylation during bud break
The delayed bud break phenology observed in our PtaDML8/10 KD lines led us to hypothesize that DMLs downregulation causes inefficient reactivation of growth promoting genes or/and, that these lines were unable to turn off genes required for dormancy maintenance. To identify genes demethylated by PtaDMLs in poplar apices, WT, KD2 and KD5 lines were subjected to whole genome bisulfite sequencing and genome wide RNA sequencing transcriptome analysis. These data and the experimental flow through are provided in Fig. 3 and Dataset S1-3.
We expected that the growth-promoting gene targets of the active DML-dependent DNA demethylation would be highly expressed in young proliferating aerial tissues and induced during bud break. Thus, among the DEGs that were hypermethylated in the three contexts either in their promoter or transcribed regions, and downregulated in KD plants compared to WT, we selected those that were relatively more expressed in young leaves and SAM. Among them, we pulled out the genes induced at the time of bud break based on our weekly RNA sequencing profiles in poplar apices growing under natural conditions from Jan 13 to bud break on Apr 14 available in geneAtlas (https://phytozome.jgi.doe.gov/pz/portal.html) (Dataset S4 and S5). As shown in Fig. 4 A-C, of the 19 and 60 genes identified respectively in the two sets, almost 75% grouped in the categories of four gene ontology biological processes related to cell metabolism: protein translation, chloroplast metabolism and light reaction, secretory pathway and mitochondrial metabolism. An open question for further studies will be to explore the role in bud break of the 7 genes ascribed here to chromatin remodelling and transcription regulation biological process. Remarkably, among these genes we found poplar homologues to Arabidopsis genes with a pivotal role in meristem development and maintenance, like SHOOTMERISTEMLESS (32), ULTRAPETALA 1 (33) or KNOTTED1-like homeobox gene 6 (34), indicating the active DML-dependent DNA demethylation promotes reactivation of meristem activity.
To identify the genes required for dormancy maintenance repressed by active DML-dependent DNA demethylation, we selected from the sets of DEGs that were hypermethylated in the promoter or transcribed region, and upregulated in KD plants compared to WT, those that were suppressed during bud break in our RNAseq time series deposited in geneAtlas (Dataset S4 and S5). As shown in Fig. 4 D-F, of the 3 and 41 genes identified, respectively, more than 38% were unknown genes, indicating a lack of detailed knowledge about the regulation and maintenance of the dormant state. This may only be studied in perennial plants.
Target genes of the active DML-dependent DNA demethylation and bud break phenology are genetically associated in poplar
A genome wide single nucleotide polymorphisms (SNPs) has been associated with phenotypic variation in several traits, including bud break, in a collection of 544 Populus trichocarpa individuals spanning much of natural latitudinal range of this specie (35). To identify common candidates controlling bud break, we merged our list of the active DML-dependent DNA demethylation targets and the list of genes showing SNPs that correlates with bud flush phenotypic variation in Evans et al. work (35). We found 15 genes associated with bud break, 9 of them were identified as growth-promoting gene targets of the active DML-dependent DNA demethylation showing a marked activation during the process of bud break in our temporal RNAseq dataset, with the exception of KNAT6 which mRNA expression pattern remained unaltered during the winter (Fig. 5 A-H, Dataset S6). Among them, we identified 6 genes participating in pathways controlling protein homeostasis of the plant cell: One poplar homolog to Arabidopsis CLP PROTEASE 5 (CLPP5) that encodes a subunit of chloroplast protease complex ClpPR involved in chloroplast biogenesis and thylakoid protein homeostasis (36). One poplar homolog to Arabidopsis and Maize gene that encodes a plastidial RidA (Reactive Intermediate Deaminase A) required to the hydrolysis of the reactive and toxic intermediates generated from the serine and threonine metabolism in the chloroplast (37). Moreover, the disruption of RidA impaired photosynthesis and growth in Arabidopsis (37). One poplar homolog to Arabidopsis TRANSLOCASE OF THE INNER MITOCHONDRIAL MEMBRANE 13 (TIM13), a member of the protein import apparatus needed for protein mitochondria homeostasis (38). Additionally, we found two poplar homologs to Arabidopsis genes encoding ribosomal subunits S8e and L19e required for protein translation (39) and one poplar homolog to Arabidopsis gene that encodes a member of the ubiquitin-dependent protein catabolic process (40). Jointly, these data indicate that the active DML-dependent DNA demethylation could promote plant cell growth and bud break by the activation of key regulators of the cellular protein homeostasis. In addition to this, we identified 3 target genes, genetically associated to bud break, that participate in shoot meristem growth control: One poplar homolog to Arabidopsis NON-PHOTOTROPIC HYPOCOTYL 3 (NPH3), a blue light signaling transducer essential for shoot phototropism, leaf flattening and positioning in Arabidopsis (41-43). One poplar homolog to PROTODERMAL FACTOR 1 (PDF1), a L1 layer secreted protein suggested to play a role in L1 cell specification (44). One poplar homolog to KNOTTED1-like homeobox gene 6 (KNAT6) required for shoot meristem activity and organ boundary formation (27). Accordingly, these data show that the active DML-dependent DNA demethylation drives bud break by activating blue light and L1 specific cell signalling, and overall meristem activity.
Genetically associated with bud break, we also identified 6 downregulated target genes, showing a decreasing pattern of mRNA accumulation in our temporal RNAseq dataset (Fig.5 IK, Dataset S6). Two of them having unknown function in Arabidopsis and Poplar. Within the remaining 4 genes, it is remarkable the identification of three poplar homolog to Arabidopsis and human proteins involved in regulation of ion homeostasis. SALT OVERLY SENSITIVE 2 (SOS2) a serine-threonin protein kinase that regulate sodium ion homeostasis (45), TWO PORE CHANNEL 1 (TCP1) a nonselective cation channel on the vacuole membrane (46) and the homolog to the ORGANIC ANION TRASPORTING POLIPEPTIDE (OATP) from humans (47)Moreover, a poplar homolog to the Arabidopsis PAUSED (PSD) importin β-family of nucleocytoplasmic transport receptors involved in exporting from the nucleus to the cytoplasm tRNAs, and others RNAs and proteins (48). Therefore, the active DML-dependent DNA demethylation could play a role in shifting from a winter osmotic stress to a regular growth promoting condition, and moreover readjusting the nuclear-cytoplasm trafficking needed to promote bud break.
In apple, Kumar et al. (2016) (14) found that reduced genomic DNA methylation during bud break affects the expression of genes involved in many cellular metabolic processes. Taken together these observations suggest a conserved mechanism controlling growth-dormancy cycles in deciduous trees. Although a role for DEMETER-LIKE genes in Arabidopsis embryo development has been long established (49), mutants in Arabidopsis DMLs show little or no ulterior developmental alterations (50,51,52), suggesting that active DNA demethylation is not critical for normal development in this species. The participation of these genes in postembryonic development has recently emerged in crops. Thus, it has been shown that active 5mC demethylation by DEMETER proteins is required for transcriptional reprogramming during tomato fruit maturation (30). Further, symbiotic organogenesis uses DEMETER-dependent epigenetic regulation to reprogram gene expression in specific host cells in Medicago truncatula (23), and, as observed here, this mechanism restores sensitivity to growth-promoting stimuli for bud break in poplar. To date, the molecular regulation of bud break is poorly understood. So far, only a molecular cascade triggered by EBB1 has been implicated in the regulation of this process (10, 30). However, no overlap was found when we crossed the active DML-dependent DNA demethylation direct target genes of with known direct targets of EBB1 (30). Hence, it seems that DEMETER and EBB1 mechanisms act via separate pathways to control bud break.
Conclusion
Our data shows that a active DNA demethylation driven mainly by a chilling-dependent PtaDML10 activity is necessary for bud break and occurs as two simultaneous processes: 1) by enabling the reactivation of key genes controlling protein homeostasis checkpoints, along with the activation of blue light and L1 specific cell signalling, and overall meristem activity; and 2) a process whereby those genes that are more highly expressed during bud dormancy are downregulated, including the resetting cellular osmotic conditions and nuclear-cytoplasm export. Collectively, these results show that an active DML-dependent DNA demethylation could act as molecular switch shifting from winter dormancy to a condition that promotes shoot apical vegetative growth.
Materials and Methods
A detailed description of the materials and methods used in the paper are provided in SI Materials and Methods. The following items are included: phylogenetic analyses to identify poplar DML proteins; plant material and growth conditions; RT-PCR expression analysis; probe synthesis and fluorescence in situ hybridization (FISH); generating binary vector constructs and plant transformation; RNA sequencing in shoot apices; DNA methylation analysis and methylome analysis through whole genome bisulfite sequencing (WGBS).
Acknowledgments
This study was supported by grants AGL2011-22625, AGL2014-53352-R, KBBE PIM2010PKB-00702, and COST Action FP0905 awarded to I.A.; M.P. and D.C. were supported by the Ramon y Cajal program of MINECO (RYC-2012-10194) and PCIG13-GA-2013-631630; M.K. holds a US NSF PGRP IOS-1444543 grant. Two short stays by D.C. in the lab of S.M. in Orléans, France were funded by grants from the Consejo Social of the Universidad Politécnica de Madrid and Universidad Politécnica de Madrid. The authors also thank A. Delaunay (University of Orléans, France) for technical advice and assistance with genomic DNA methylation determination by HPLC, Phil Wigge for his helpful comments and suggestions during this work and C. Grunau and C. Chaparro (IHPE, Perpignan, France) from the RTP3E (http://rtp-3e.wix.com/rt3e) for the use of their local bioinformatics cluster.