Extensive nuclear reprogramming and endoreduplication in mature leaf during floral induction
The floral transition is a complex developmental event, fine-tuned by various environmental and endogenous cues to ensure the success of offspring production. Leaves are key organs in sensing floral inductive signals, such as a change in light regime, and in the production of the mobile florigen. CONSTANS and FLOWERING LOCUS T are major players in leaves in response to photoperiod. Morphological and molecular events during the floral transition have been intensively studied in the shoot apical meristem. To better understand the concomitant processes in leaves, which are less described, we investigated the nuclear changes in fully developed leaves during the time course of the floral transition.
We highlighted new putative regulatory candidates of flowering in leaves. We observed differential expression profiles of genes related to cellular, hormonal and metabolic actions, but also of genes encoding long non-coding RNAs and new natural antisense transcripts. In addition, we detected a significant increase in ploidy level during the floral transition, indicating endoreduplication.
Our data indicate that differentiated mature leaves, possess physiological plasticity and undergo extensive nuclear reprogramming during the floral transition. The dynamic events point at functionally related networks of transcription factors and novel regulatory motifs, but also complex hormonal and metabolic changes.
KeywordsFloral transition Leaf Arabidopsis Transcription Non-coding RNA Transcription factors DNA motif Endoreduplication
Differentially expressed gene
The transition to flowering is a decisive developmental event in the plant life cycle for reproductive success. The general understanding highlights a fine-tuned process involving a complex interplay between environmental and endogenous cues. Signals are perceived and decoded according to the plants’ lifestyle, and lead to a cascade of dramatic morphological changes at the meristem level, to produce floral organs [1, 2].
Photoperiod is a major parameter controlling the transition to flowering with intricate phototropic effects and links with the circadian clock. The light signal, perceived in the leaves, triggers the accumulation of metabolites and regulators, such as the well-conserved FLOWERING LOCUS T (FT) protein, whose expression is under the control of the CONSTANS (CO), a zinc finger transcription factor involved in photoperiod pathway . Their export, as systemic florigen signals via the vasculature to the distant shoot apical meristem (SAM), activates floral homeotic genes [4, 5, 6]. Described as a quantitative long-day (LD) species, the photoperiodic property of Arabidopsis species was exploited to induce synchronous flowering by exposure to a single LD or a single displaced short-day (SD), providing a convenient experimental inductive system . Besides photoperiod, other regulatory pathways partake to the vegetative-to-reproductive switch control [8, 9, 10].
From transcriptional and chromatin-based mechanisms, to alternative splicing and post-translational regulation, numerous regulatory levels participate to the control of the floral transition [9, 11, 12, 13, 14, 15] and its main actors, which have been gathered in the Flowering-Interactive Database (FLOR-ID) . Besides protein regulators involved in developmental transitions, an increasing number of studies have highlighted the regulatory functions of long non-coding RNAs (lncRNAs) [17, 18]. In response to vernalization, the lncRNAs COLDAIR, COLDWRAP, COOLAIR, and Antisense Long participate to the fine regulation of the key MADS-box floral repressor FLOWERING LOCUS C (FLC) via modifications of FLC chromatin environment [19, 20, 21, 22, 23]. Recently, FLORE, a Natural Antisense Transcript (NAT) of CYCLING DOF FACTOR5 (CDF5) was shown to positively regulate flowering time, repressing CDF TFs (CDF1, CDF3, CDF5), and subsequently increasing FT expression . LncRNAs are versatile regulators involved in transcriptional gene regulation, in guiding or scaffolding protein complexes involved in chromatin organization and gene regulation, or even in post-transcriptional regulatory mechanisms . Due to their large number estimated at several thousands and their diversity (intergenic ncRNAs, intronic ncRNAs, antisense RNAs, cis or trans NATs…) [19, 23, 26, 27], their functional annotations and roles in developmental phase transitions remain poorly explored.
The transition to flowering is an integrated process at the scale of the whole plant. Few studies analyzed the transcriptional behaviors of meristematic and root tissues during the floral transition at the genome level [28, 29, 30, 31, 32]. Early studies identified few CO targets differentially expressed during flowering in leaves, among which FT  which was identified as the major CO target involved in the SD to LD shift response . FT is referred as a flowering integrator with TWIN SISTER OF FT (TSF) [35, 36]. Subsequent studies increased the number of associated genes involved in the leaf response during flowering (see FLOR-ID overview and references therein). However, the dynamics of genome-wide transcriptomes in leaves during the floral transition has not been reported despite the key functions of leaves as receptors of the inductive photoperiodic signal and producer of florigenic molecules. Here, by exploiting the inductive response to a long-day (LD) shift  and disconnecting leaf growth or developmental responses from the floral inductive response, we performed a large transcriptome analysis, and identified novel loci and regulatory elements involved in flowering in mature leaves. The transcriptome dataset enabled us to highlight molecular events, providing new insights into transcriptional reprogramming in leaves accompanying the floral transition. Observations of endoreduplication events supported transcriptome data and suggested a novel function in flowering.
Flowering and organ growth in mature leaves
To complete our characterization, we monitored organ growth accompanying the photoperiodic shift. No significant difference was detected in the rosette size during the window of 0–5 dat between the continuous SD and SD-LD conditions, whereas the first six rosette leaves presented different behaviors (Fig. 1d, Additional file 2: Figure S2). The size of the first two leaves was already established 2 days before transfer (dbt), independently of the photoperiodic conditions. Leaves 3–4 showed no significant differences in growth rate between SD and SD-LD conditions from 0 to 5 dat. Leaves 5–6 presented a higher and continuous growth rate over the 15 days. Thus, during the 5-day floral transition window, developmental and growth processes are arrested in leaves 3–4, which makes this pair of mature leaves appropriate material for investigating the early molecular events associated with floral transition, independent of other developmental or signaling events.
Ploidy level changes during the floral transition window
Major changes in transcription profiles during the inductive shift
To characterize the molecular events during the floral transition, we examined RNA profiles in leaves 3–4 at different time points (T0, T2, T3, and T5) (Fig. 1c, Additional file 3: Figure S3). We identified 20,284 genes expressed at least in one of the four time points, with more than 6000 differentially expressed genes (DEGs) at the largest transition (T0/T2) and specific gene sets at the main T0/T2, T2/T3 and T3/T5 transitions. By assembling a non-redundant dataset of 14,621 long non-coding transcription units (lncTUs) based on TAIR annotations and published datasets, we also identified 531 differentially expressed lncTUs (DE-lncTUs) (Additional file 4: Figure S4, Additional file 5). These data endorse the highly dynamic transcriptional activity in mature leaves in response to the SD-LD switch.
The 24 clusters had specific GO term enrichments, even among cluster families, such as seven clusters (C20, C23, C16, C18, C19, C15, C4, C3) with strong signatures, thus supporting the cluster analysis (Fig. 3c, d, Additional file 7). This classification revealed different processes in mature leaves during the SD-LD response. For instance, C3, C15, C17 and C24 were enriched in down-regulated genes involved in photosynthesis chlorophyll biosynthesis process, light harvesting, or plastid organization, pointing at a reprogramming of the photosynthetic apparatus. C12 and C15 from CF5 and most clusters in CF3 were enriched in stress-associated terms, such as defence response, response to stimuli, response-to-wounding, highlighting a stress association with the changes in photoperiod, light and perturbation of the circadian clock. In accordance, we noticed in C19, C2 and C14, “child” GO terms associated with jasmonic acid, salicylic acid or brassinosteroids, respectively. Secondary metabolites, such as flavonoids (C18) and the defence-related glucosinolates (C4), and carbon metabolism (GO terms such as “glucan catabolic process”, “cellular polysaccharide catabolic process” in C17) were also modified. We observed enrichments in GO terms associated with cell wall, such as “cell wall organization” and “xyloglucan metabolic process” in C14 or “plant-type secondary cell wall biogenesis” and “cellulose metabolic process” in C20. For instance, XYLOGLUCAN ENDOTRANSGLUCOSYLASE/HYDROLASE 9 (XTH9) involved in cell wall loosening was up-regulated in C14 (log2 ratio 1.85, FDR 1.28E-3 at T0/T2). These data suggest that the SD-LD switch is accompanied by cell wall remodeling and some cell wall plasticity in the mature leaves in response to environmental changes. Cell wall modifications were reported in roots in response to inorganic phosphate starvation or in hypocotyl in response to light signaling . Such examples of cell wall remodeling in relation to environment signaling remain rarely reported, especially in leaf. We also noticed in C14 a GO term “response to cyclopentenones”, which are fatty acid derivatives with signaling activities. C16 cluster, belonging to CF2 with a transient down-regulation profile, had the highest enrichment terms related to translation, RNA processing and metabolism, suggesting that a strong modification of the protein metabolism at the cellular level is possibly escorting the transition of the metabolic regime occurring in the whole plant during the switch to reproductive phase. Such transient modifications of translation and associated processes were also observed during a cell dedifferentiation and re-differentiation process in Arabidopsis protoplasts . Finally, C20 was enriched in GO terms associated with cell cycle processes. The modifications of the expression of LGO/SMR1, KRP2 and KRP6 cell cycle inhibitor genes involved in endoreduplication , the CYCA2;3, a suppressor of endocycles, the major cell-cycle markers, CDKB2.1, CYCA1;1, and WEE1, a negative regulator of the entrance in the M phase [40, 41] were consistent with the onset of the observed endocycles (Additional file 8: Figure S5). In summary, leaf transcriptome during the floral transition revealed major changes in numerous processes, such as carbon and secondary metabolism, signaling events, and endoreduplication.
Flowering and hormone-related genes are differentially expressed in mature leaves during the floral induction
In the Arabidopsis Hormone Database (AHD) , we identified 331 DEGs involved in hormonal regulation from biosynthesis, metabolism, perception, and transport to hormonal responses. C23 was the most enriched cluster with these genes (Additional file 10; Additional file 11: Figure S6). We noticed that genes related to abscisic acid (ABA) and auxin were the most represented ones among the DEGs (Fig. 4c). Genes involved in the ABA biosynthesis were down-regulated in agreement with a repressive role of ABA in flowering . However, the switch also largely impacted genes related to hormone signal transduction such as auxin transport (Fig. 4d, e). Among the DEGs involved in auxin-hormone transport, most genes were up-regulated at least transiently during the switch, such as several PIN members (Fig. 4f), questioning the role of auxin transporters in the flowering time control. Genes involved in gibberellin (GA) biosynthesis were mainly down-regulated, while GAMT2, a methyltransferase involved in the GA metabolism was activated, as well as negative regulators of GA responses (RGA-LIKE1–2) (Additional file 11: Figure S6b). Indeed, genes involved in cytokinin (CK) biosynthesis and the SOB FIVE-LIKE 1, 2 genes (SOFL1, 2), which participate to CK level regulation  were activated, whereas genes involved in CK catabolism (CKK4, CKK6) were down-regulated. Consistently, the type-B ARABIDOPSIS RESPONSE REGULATOR 10 (ARR10) TF, a key player in the CK signaling pathway for the light response and shoot initiation , was up-regulated between T0/T5. Beside DEGs involved in one hormonal pathway, 36 DEGs are involved in hormonal crosstalk, with ABA being involved in most of these crosstalk (Additional file 12). Whereas GA is proposed to promote the floral transition and have antagonistic effects with ABA, our data suggest a complex hormonal interplay during the SD-LD switch, with GA, ABA but also new players such as the brassinosteroids and derivative forms, as well as IAA and CKs.
Novel regulatory actors involved in the SD-LD switch
The 32 TF set was enriched in GO terms associated with “regulation of metabolic process” (GO:0019222; p-value 5.65E-29, FDR 1.11E-26), “response to hormone stimulus” (GO:0009725; p-value 3.86E-12, FDR 3.53E-10) and “circadian rhythm” (GO:0007623; p-value 8.68E-10, FDR 6.43E-08) (PlantGSEA toolkit). The M001 motif (AAAATATCT) matched to the TFBS recognized by CCA1, LHY1, RVE1, and RVE5–8 TFs and to the “Evening Element”, involved in the control of circadian-regulated genes  and identified, for instance, in down-regulated genes such as SVP, PHYTOCLOCK 1 (PCL1) and LONG VEGETATIVE PHASE 1 (LOV1). M118 (MACGYGB) is similar to the TFBS of the MYC3 and MYC4, two TFs involved in flowering . Intriguingly, no TF could be associated with M003 (AAACCCTA) and M004 (AAACCCTAA), the two closely related motifs with the best PLM p-values (3.43E-177, 5.39E-104, respectively). Remarkably, M003 was highly similar to the (A/G/T)AACCCTA(A/G) motif, an LHP1 binding motif, related to the telo-box motif (AAACCCTA) and recognized by REPEAT BINDING PROTEIN1 (TRB1) , and to a lesser extent, to the tertiary motif of TOE1 (AACCTTAA), a TF belonging to the AP2/EREBP superfamily (E-value 0.54 using the PBMdb). Both LHP1 and TOE1 are known to repress flowering, LHP1 being a component of PRC1 complex [53, 54] and TOE1 inhibiting the CO activity in the FT activation . A majority of these 32 TFs (56%) were differentially regulated during the process, implying functional preferences of the identified motifs.
Among FLOR-ID, a functionally related subset of 64 TFs, corresponding to the differentially expressed TFs of the database (64 out of 143) (Additional file 15) was analyzed using the TF2Network tool  to decipher gene regulatory networks in the mature leaf. We identified 66 specific candidate regulators, by comparison with the subset corresponding to the FLOR-ID TFs, which were not differentially expressed (data not shown). The best-ranked regulators were HYH, ABF1, TCP21, ABI5, MYC2 and HY5 (Additional file 16: Figure S7), suggesting candidate regulators of the floral transition in mature leaf. The identification of ABI5 and ABF1 as candidate regulators, which are TFs involved in ABA responses, was in agreement with the high percentage of ABA-related genes differentially expressed in mature leaf during the floral transition (Fig. 4c).
Finally, since chromatin is a key transcriptional regulatory level, we searched for key chromatin-associated genes (CAGs) involved in flowering. We identified 90 DE-CAGs, 91% being differentially expressed at T0/T2, with a bias towards up-regulated genes (Additional files 17 and 18). We noticed that a large proportion (39%) of the DE-CAGs encoded histone variants with, for instance, the H1.3 variant (HON3), associated with stress response, 5 H3.1 variants, which are incorporated in a replication-dependent manner in agreement with the endoreduplication events or 10 H2A variants (Fig. 6c). Among the genes involved in histone post-translational modifications, such as genes encoding SDG4/ASH1-RELATED 3 and SDG13/SUVR1 histone methyltransferases, were up-regulated, whereas histone deacetylases were only weakly differentially expressed, except HDA2, which was down-regulated and associated with the floral transition for the first time here. Consistently with modifications in DNA methylation accompanying early floral transition events, we observed the expression changes of MET1, CMT3 but also of DEMETER-LIKE2 (DML2), encoding a DNA glycosylase involved in active DNA demethylation, in the mature leaves. Thus, the results suggest a rapid modification of the epigenome, concomitant with the changed TF profiles, which further endorses a dramatic reprogramming in the leaf genome, reminiscent of another developmental switch .
A set of lncTUs is differentially regulated during the floral transition
A hierarchical clustering analysis revealed that the majority of the 531 DE-lncTUs were either up- or down-regulated, whereas only 15 DE-lncTUs changed dynamically (Fig. 7b, c). This indicates that most of the of DE-lncTUs were mainly specific for either the vegetative or the reproductive phase, while fewer specific ones are involved in the transition event only.
To find out if lncTUs can change the expression of genes in cis we examined the expression of both the lncTUs and their neighboring genes. No correlation between expression and structure could be detected at the genome level. We then investigated whether lncTUs may have putative NAT regulatory function by analyzing lncTUs with overlapping genes. The FLORE NAT, for example, represses CDF5 located in antisense but also CDF1 and CDF3 located on other chromosomes and participates to the flowering time control . Whereas CDF2 was up-regulated, CDF3 and CDF5 were down-regulated in our data. Putative lncTUs that we could identified in the regions of CDF2, CDF3, and CDF5 were not differentially expressed in our experimental design, which suggests that other regulatory mechanisms control CDF expression. Alternatively, transient changes in the expression of the corresponding lncTUs could not be detected. In our dataset, we found that 655 lncTUs overlapped with a flanking gene and were transcribed in antisense (putative NAT lncTUs). Among them, 19 NAT lncTUs showed opposite transcriptional activity with the overlapping gene, in at least one of the 6 comparisons (Fig. 7d, Additional file 19). However, the expression dynamics of the gene and its NAT lncTU may be more complex and the two partners may have no synchronized expression. For instance, a lncTU formed a NAT couple with MAF5 (LNCRNA-MERGE_C-9859, named MAF5_NAT), encoding a floral repressor of the FLC clade. MAF5 was differentially regulated during the floral transition and MAF5_NAT was strongly down-regulated before the up-regulation of MAF5 (Additional file 20: Figure S8). This is different from the transient up-regulation of antisense regulatory ncRNAs which represses FLC, thus illustrating the complexity of the flowering regulation mechanism.
Finally, for each DE-lncTU, we examined the presence of the TFBSs of the SVP, FLC and SOC1 flowering regulators, in a 4-kb window (3 kb upstream - 1 kb downstream) [70, 71, 72, 73, 74]. We identified 123 TFBSs in the vicinity of 63 DE-lncTUs, with one to seven of these binding sites mainly present in the 5′ regions, suggesting some functionality of these TFBSs regarding the expression of the DE-lncTUs and putative roles of these TFs in lncTU regulation (Fig. 7e, f, Additional file 21). Furthermore, some of the DE-lncTUs with TFBSs were located in CS2 and/or involved in NAT couples. Based on these criteria (expression profile, presence of TFBS, location in specific chromatin state, NAT couple), the identified DE-lncTUs represent interesting candidates whose regulatory function in flowering will require further investigation.
Complex molecular processes in mature leaves during the floral transition
In sensing floral inductive stimuli, leaves produce the florigen signal that switches the SAM from the vegetative to reproductive phase. Key genes involved in producing the florigen have been identified, but the global molecular events in leaves during the floral transition remain poorly described. By focusing on mature rosette leaf whose growth was completed, our study completes analyses and highlights molecular events of the floral transition in this organ. Based on differential gene expression profiles, we showed that the floral transition induced by the SD-LD switch is accompanied by re-organization of photosynthetic capacity, protein synthesis, cellular metabolism, hormonal action, stress response and cell cycle regulation, with an intricate interplay between the light regime, the circadian clock and the floral transition. Our data highlight the complex role of mature leaves in the floral transition.
LD stimulates increases in leaf sucrose level as part of the florigenic signal [75, 76, 77]. However, other metabolites are also involved in the floral transition: carbon, phosphorous, nitrogen or sulphur can impact this process [4, 78]. Genes associated with flavonoids and glucosinolates contents were differentially regulated during the photoperiodic switch, whereas these secondary metabolites are usually associated with stress responses . These data support a recent study showing that the flowering regulator FLC is present in a QTL interval associated with glucosinolate contents in the brassicaceae species, Aethionema arabicum . The switch occurring in the SAM appears to require massive metabolic and physiologic reprogramming events in the leaf to further explore.
Highlight of new regulatory candidates in flowering control
Dynamics changes were reported in the transcriptional profiles of genes but also lncRNAs, highlighting the potential regulatory functions of some of them in the floral transition. Our atlas of lncTUs, with putative regulatory functions based, for instance, on their location in bivalent chromatin states or in antisense with protein-coding genes, provides promising resource for new actors in the genome regulation during the floral transition.
The analysis of the gene clusters allowed the extraction of specific putative regulatory motifs. Some of these DNA elements corresponded to binding sites of differentially expressed TFs, suggesting functionality in the transcriptional regulation of the floral transition. In parallel, a large set of differentially expressed TFs involved in various processes was identified, consistently with the molecular processes highlighted with the GO analysis. Changes in transcription of a set of FMI-FD genes, whose action is crucial in meristems, suggest other levels of gene regulation during the floral transition.
By focusing on FLOR-ID TFs, we highlighted small gene regulatory networks and potentially, new players in the floral transition in leaf, such as ABI5, ABF1, or TCP21. TCP21 is involved in the circadian clock regulation and controls CCA1 expression . The identification of TCP21 as candidate regulator here is in agreement with the large proportion (84.2%) of circadian clock–associated genes differentially expressed, during the SD-LD switch (Fig. 4b). ABI5 was reported as a floral repressor , which is consistent with its down-regulation here during the SD-LD switch and its putative interactions with the downstream deregulated TFs. ABA is involved in the control of flowering time with opposite effects according to environmental conditions [83, 84, 85]. For instance, ABA was shown to promote flowering time in a LD-dependent manner and in response to water resource availability, by modulating GIGANTEA (GI) activity on FT and TSF [84, 85]. However, the down-regulation of GI here suggests that the ABA-dependent promotion of flowering may not be involved in the SD-LD transition, but a loss of repression mediated by ABI5 may possibly occur. The hormonal contribution of the GA-dependent promoting pathway  may also play a role. Most of the hormone pathways being affected during the switch, albeit to different degrees, performing hormone dosages may help to untangle their contribution to flowering time.
Endoreduplication events accompanying the floral transition in leaves
The acceleration of the cell division was reported in the SAM during the floral transition in A. thaliana [29, 86]. Here, we report that endocycles escort the floral transition in leaves. Thus, the floral induction is accompanied by modulations of the cell cycle in both leaves and meristems, but with differences in cell cycle exits according to the organs, endocycle or mitosis, respectively. Our result is supported by the changes in expression of key cell cycle phase markers, such as CYCA2;3, but also histone variants and endocycle-related genes. Consistently with a loss of function stimulating endocycles , CYCD3 genes were down-regulated in our experiments. The spindle assembly checkpoint (SAC) genes were also shown to impact both the floral transition and the timing of the endocycle onset . Here, only BUBR1/MAD3 from the SAC family was slightly down-regulated, but its function in the Arabidopsis mitotic checkpoint control remain poorly documented.
Hormone signaling pathways participate to the control of the mitotic-to-endocycle transition: the endocycle repression is induced by high auxin contents [39, 89]. Consistently, we observed alterations of the auxin pathways during the photoperiodic inductive switch. The SUMO E3 ligase HPY2, described as an endocycle repressor, which may link auxin signaling and cell cycle program [89, 90], as well as two other negative regulators of endocycles, were up-regulated at the T0/T2 transition, suggesting that the entry into the endocycle program in the mature leaves may result from a fine dosage between the different controlling pathways.
Previous studies showed that the increase in light intensity  and UV-B radiation  are associated with changes in ploidy levels. A proposed hypothesis is that the ploidy dynamics might be an adaptive response to damage possibly induced by solar radiation. Finally, we could also speculate that the increase in ploidy level in mature leaves during the floral transition may contribute to an increase in energy production required for the developmental switch, an increase in metabolites and endogenous signaling molecules, or a modulation of transcription thresholds.
Our detailed study provides a novel molecular framework to further question the roles of new putative regulators in leaves during the floral transition, such as new putative lncRNAs, whose polyadenylation status will require further confirmation. Furthermore, it points at the relationship between flowering and endoreduplication and at the complex interplay between several plant hormones, which open new perspectives.
Plant materials and growth conditions
All Arabidopsis thaliana lines were in the Col-0 background. Seeds of the pAP1::GUS transgenic line were kindly provided by Prof. Dr. G. Angenent (unpublished material). The AP1 promoter fragment was fused to a GUS-GFP cassette as described in , using the binary pBGWFS7 vector from VIB . Plants were grown in growth chamber under SD (8 h light/16 h dark) or LD (16 h light/8 h dark) conditions. White fluorescent light was used. The photosynthetic photon flux density was 120 μmol m− 2 s− 1 in SD and LD. In SD, the temperature was 21 °C during the light and 18 °C during the dark period, and the humidity (65%) remained constant. In LD, the temperature (21 °C) and the humidity (70%) remained constant, 21 °C and 70%, respectively. Plants were cultured for 3, 4 or 5 weeks in soil, in individual pot. The transfer was done at the end of SD light, preceding the LD dark period. Flowering time indicators were recorded as previously described (Additional file 1: Figure S1a). The percentages of cauline leaf relatively to the total leaf number (CL%) quantifies the relationship between bolting and floral transition events . For plants grown 3, 4 and 5 weeks in SD and transferred in LD it similar to the CL% in continuous LD (17.6%), and higher compared to continuous SD (11.9%) . This preliminary assay showed that the SD-LD switch mimicked LD growth conditions. For a good compromise between time and material quantity, analyses were then pursued on plants grown for 4 weeks in SD.
For RNA extraction, plants were collected at Zeitgeber time 7 (ZT7) in SD, and ZT15 in LD, ZT0 marking the transition from dark to light. For leaf growth analysis, individual leaves were harvested at different time points, flattened on white paper and then digitally scanned. Leaf areas (blade and petiole) were calculated from the binary images using ImageJ software (http://rsb.info.nih.gov/ij/). Leaves from 10 to 15 plants were analyzed.
Leaves 1 to 4 were harvested, chopped with a razor blade in 800 μl of Galbraith buffer, filtered over a 30 μm mesh, and 150 μl of a propidium iodide solution (100 μg/ml) was added . The quantification of the nuclear DNA content was performed on a CyFlow® cytometer using the FloMax® software (Sysmex Partec, France) as described . The endoreduplication index was calculated by using the formula: EI = 0x(% of 2C) + 1x(% of 4C) + 2x(% of 8C) + 3x(% of 16C) + 4x(% of 32C).
Total RNAs were prepared from rosette material, treated and reverse transcribed, as previously described . Quantitative real-time PCR was performed on a BioRad CFX96 apparatus using the SYBR green Master Mix (BioRad) following manufacturer’s instructions. UBIQUITIN10 was used as reference gene. Primers are listed in Additional file 22. For GUS histochemical staining, plants were collected in the staining solution (1 mM X-Gluc (5-bromo-4-chloro-3-indolyl-ß-D-glucuronide), 0.1 M sodium phosphate buffer, pH 7.0, 2 mM potassium ferrocyanide, 2 mM potassium ferricyanide, and 0.5% Triton X-100), infiltrated under vacuum 3 times, for 5 min each, and incubated at 37 °C overnight. Samples were then washed in 70% ethanol and observed under a light microscope.
RNA extraction, library preparation and sequencing
Total RNA was extracted with the Plant RNeasy Mini kit (QIAGEN). 10 μg of RNA was treated with TURBO DNA-free kit (Ambion Ref. AM1907) and cleaned-up from enzymatic reactions with RNeasy MinElute Cleanup Kit (QIAGEN Ref. 74,204), following the manufacture instructions. RNA integrity and concentration were analyzed with the Agilent 2100 Bioanalyzer and the Agilent RNA 6000 Nano Kit (Ref. 5067–1511). For one replicate, leaves 3–4 were dissected from 20 plants and pooled. Three independent replicates were performed for each time point. Strand specific sequencing libraries were prepared from polyA RNAs using the Illumina Tru-Seq stranded RNA sample preparation v2 kit. Four libraries were multiplexed per lane and paired-end (PE) sequenced on an Illumina HiSeq 2000. Over 40 millions of 150 bp reads were generated per sample. All steps of the experiment, from growth conditions to bioinformatic analyses, were recorded in CATdb database  (http://tools.ips2.u-psud.fr/CATdb/) Project ID NGS2015_01_Transition according to the international standard MINSEQE minimum information about a high-throughput sequencing experiment.
RNA-Seq data analysis
RNA-Seq samples were processed using the following pipeline: the read pre-processing criteria included trimming library adapters and performing quality control checks using FastQC. The raw data (fastq) were trimmed using the FastX toolkit (Phred Quality Score > 20, read length > 30 bases). The Bowtie 2 mapper  was used to align reads against the A. thaliana TAIR 10 transcriptome. On average, 99% passed the quality filter and were uniquely mapped to the TAIR 10 reference genome. We extracted 33,602 genes from TAIR10 version database  with one isoform per gene corresponding to the representative gene model (longest coding sequence) given by TAIR10. The abundance of each gene was calculated by a local script, which parses SAM files and counts only paired-end reads for which both reads map unambiguously one gene, and by removing multi-hits. According to these rules, around 96% of PE reads were associated with a gene, 2% PE reads unmapped and 2% of PE reads with multi-hits were removed.
For differential expression analysis, we discarded genes, which did not have at least 1 read after a count per million (CPM) normalization, in at least one half of the samples. The library sizes were normalized using the TMM method. The count distribution was modelled with a negative binomial Generalized Linear Model (GLM) where the harvest date was considered. Dispersion was estimated by the edgeR method  in the statistical software ‘R’ (R Core team, 2015). The p-values were adjusted by the Benjamini-Hochberg procedure to control FDR. A gene was differentially expressed when its adjusted p-value was lower than 0.05.
Analysis of lncRNAs
We gathered a non-redundant dataset (IJPB_lncDB) of 14,621 putative lncRNA sequences from published lncRNAs datasets [26, 104, 105, 106]. Redundant information was removed. Datasets were organized into three subsets according to strand information (+, −, Not Available (NA)). For each subset, we merged overlapping or “book-ended” lncRNA in a single transcription unit (lncTU). Three FASTA files of lncTUs were established: one with 5055 TUs on the positive strand, another one with 4851 putative TUs on the negative strand, and the last one with 4715 TUs, without strand information. All reads were mapped against the IJPB_lncDB using the Bowtie 2 mapper  using the same count criteria. Whereas 96% of the paired-end reads mapped to the TAIR10 genome as expected, the mean mapping percentage to the lncRNA dataset was 0.78%. To establish the differentially expressed lncTUs, we used the GLM of edgeR, without or with filter using either the Bonferroni or Benjamini-Hochberg (BH) test corrections. All lncTUs differentially expressed identified using Bonferroni test were present in the list of DE lncTUs identified using BH test. We further analyzed BH DE lncTUs. No bias was observed for the distribution of the DE-lncRNAs on the two strands.
To determine overlaps between lncRNA and annotated chromatin states we used the online BEDTools suite. We established intersects for all lncRNAs, the DE lncRNAs and randomly reshuffled regions of identical size to compute the fold changes between observed and randomly distributed lncRNAs. The hierarchical clustering analysis was performed using the Multiexperiment Viewer tool (MeV 4_8) with the average linkage method, gene leaf order optimization and Pearson correlations.
Model for the co-expression analysis
Co-expression analysis was carried out on differentially expressed transcripts and lncRNAs using the R package coseq  (https://bioconductor.org/packages/devel/bioc/vignettes/coseq/inst/doc/coseq.html). We ran two clustering methods (K-means algorithm and Gaussian mixture models) for two different count data transformation functions (the centred log ratio (CLR) and logCLR for K-means; Logit and arcsin for Gaussian mixture models). Ten technical replicates were performed for each combination of method/transformation to prevent initialization problems. We computed 30 models from K = 10 to K = 40 (K = number of clusters). For each method, the best K was selected via the slope heuristics approach for K-means methods or via the Integrated Completed Likelihood (ICL) criterion for Gaussian mixture models. The transformation function, which minimizes the within clusters variability (for K-means algorithm) or the ICL criterion (for Gaussian mixture models) was retained. Since the K-means algorithm seemed more sensitive to extreme expression data, we finally retained the Gaussian mixture model method with the arcsin transformation function and K = 24. This method provided a more homogeneous number of transcripts per cluster.
For each cluster, a Singular Enrichment Analysis (SEA) of GO terms (AgriGO v2.0)  was performed (Fisher test with a FDR cut-off at 0.01 and a minimum number of mapping entries of 10), using a customized reference corresponding to expressed genes during the time course experiment. An heatmap comparing the results of individual cluster’s SEA were obtained using the SEACOMPARE program (AgriGO v2.0).
We extracted 413 genes (FLGs) from FLOR-ID , comprising the 306 core flowering time genes, genes involved in flower meristem identity and flower development (FMI-FD) and pending annotated flowering time genes. For the analysis of the TFs we used the PlantTFDB 4.0 (http://planttfdb.cbi.pku.edu.cn/). Chromatin-associated genes were described in . Hierarchical clustering was performed using the Multiple Experiment Viewer tool (MeV) with the Pearson correlation metric and average linkage clustering as linkage method . We performed functional annotation and classification using the “AgriGO” Gene Ontology tool  and the Classification SuperViewer Tool from BAR . For each cluster, we extracted the biological process (BP) GO terms with the best FDR and the best specialized and enriched “child” GO terms (Additional file 23). Venn diagrams were generated using the online tool provided by T. Hulsen (http://bioinformatics.psb.ugent.be/webtools/Venn/).
The “Preferentially Located Motifs” algorithm is based on the over-representation of a motif around the Transcription Start Site (TSS), region − 300 from TSS to 5’UTR, compared to its distribution in the region of − 1000 to − 300 (learning region) before the TSS . We also explored a list of 419 motifs merged from PLACE  and AGRIS (http://agris-knowledgebase.org/AtcisDB/bindingsites.html) to find enrichment (p-value < 0.05) around the TSS compared to all Arabidopsis genome (Additional file 23).
The accession number into the international GEO repository is GSE116123.
We thank Bruno Letarnec and Hervé Ferry for plant care in the greenhouses. We thank Prof. Dr. Gerco Angenent and Dr. Suraj Jamge for the seeds of the API::GUS transgenic line (unpublished material) in the frame of the ITN Project EpiTRAITS. We are very grateful to Dr. Ortrun Mittelsten Scheid and the VBCF platform for preliminary sequencing tests. We thank Johanne Thevenin and the plant cytology and imaging platform (PCIV)” of the IJPB Plant Observatory for the technical support on flow cytometry. We are very grateful to Dr. Jeffrey Leung for careful reading and comments. We are very grateful to and thank Prof. Dr. Ronald E. Koes (Amsterdam University) for providing SDP a PhD short-term fellowship from the University of Amsterdam. The funding of SDP is gratefully acknowledged.
SDP was supported by a PhD fellowship provided by the European Commission Seventh Framework-People-2012-ITN Project EpiTRAITS (no-316965), a short-term INRA grant and by a PhD short-term fellowship from the Swammerdam Institute for Life Sciences of the University of Amsterdam. The sequencing platform (POPS-IPS2) and the IJPB benefit from the support of the LabEx Saclay Plant Sciences-SPS (ANR-10-LABX-0040-SPS). The funding agencies played no role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.
Availability of data and materials
The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.
SDP performed experiments and analyses. LST performed libraries and sequencing. AM, DC, NB, CG, VB, and FG performed bioinformatics analyses. VG analyzed data and coordinated the analyses. SDP, PF and VG wrote the manuscript. All authors read and approved the final manuscript.
Ethics approval and consent to participate
The experimental research on plants performed in this study complies with institutional, national and international guidelines.
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 2.Pajoro A, Biewers S, Dougali E, Leal Valentim F, Mendes MA, Porri A, Coupland G, Van de Peer Y, van Dijk AD, Colombo L, et al. The (r)evolution of gene regulatory networks controlling Arabidopsis plant reproduction: a two-decade history. J Exp Bot. 2014;65(17):4731–45.PubMedCrossRefPubMedCentralGoogle Scholar
- 30.Torti S, Fornara F, Vincent C, Andres F, Nordstrom K, Gobel U, Knoll D, Schoof H, Coupland G. Analysis of the Arabidopsis shoot meristem transcriptome during floral transition identifies distinct regulatory patterns and a leucine-rich repeat protein that promotes flowering. Plant Cell. 2012;24(2):444–62.PubMedPubMedCentralCrossRefGoogle Scholar
- 52.Molitor A, Latrasse D, Zytnicki M, Andrey P, Houba-Herin N, Hachet M, Battail C, Del Prete S, Alberti A, Quesneville H, et al. The Arabidopsis hnRNP-Q protein LIF2 and the PRC1 subunit LHP1 function in concert to regulate the transcription of stress-responsive genes. Plant Cell. 2016;28:2197–211.PubMedPubMedCentralCrossRefGoogle Scholar
- 53.Gaudin V, Libault M, Pouteau S, Juul T, Zhao G, Lefebvre D, Grandjean O. Mutations in LIKE HETEROCHROMATIN PROTEIN 1 affect flowering time and plant architecture in Arabidopsis. Devevlopment. 2001;128:4847–58.Google Scholar
- 64.Kumimoto RW, Adam L, Hymus GJ, Repetti PP, Reuber TL, Marion CM, Hempel FD, Ratcliffe OJ. The nuclear factor Y subunits NF-YB2 and NF-YB3 play additive roles in the promotion of flowering by inductive long-day photoperiods in Arabidopsis. Planta. 2008;228(5):709–23.PubMedCrossRefPubMedCentralGoogle Scholar
- 68.Sequeira-Mendes J, Araguez I, Peiro R, Mendez-Giraldez R, Zhang X, Jacobsen SE, Bastolla U, Gutierrez C. The functional topography of the Arabidopsis genome is organized in a reduced number of linear motifs of chromatin states. Plant Cell. 2014;26(6):2351–66.PubMedPubMedCentralCrossRefGoogle Scholar
- 70.Gregis V, Andres F, Sessa A, Guerra RF, Simonini S, Mateos JL, Torti S, Zambelli F, Prazzoli GM, Bjerkan KN, et al. Identification of pathways directly regulated by SHORT VEGETATIVE PHASE during vegetative and reproductive development in Arabidopsis. Genome Biol. 2013;14(6):R56.PubMedPubMedCentralCrossRefGoogle Scholar
- 71.Immink RG, Pose D, Ferrario S, Ott F, Kaufmann K, Valentim FL, de Folter S, van der Wal F, van Dijk AD, Schmid M, et al. Characterization of SOC1's central role in flowering by the identification of its upstream and downstream regulators. Plant Physiol. 2012;160(1):433–49.PubMedPubMedCentralCrossRefGoogle Scholar
- 72.Mateos JL, Madrigal P, Tsuda K, Rawat V, Richter R, Romera-Branchat M, Fornara F, Schneeberger K, Krajewski P, Coupland G. Combinatorial activities of SHORT VEGETATIVE PHASE and FLOWERING LOCUS C define distinct modes of flowering regulation in Arabidopsis. Genome Biol. 2015;16:31.PubMedPubMedCentralCrossRefGoogle Scholar
- 87.Dewitte W, Scofield S, Alcasabas AA, Maughan SC, Menges M, Braun N, Collins C, Nieuwland J, Prinsen E, Sundaresan V, et al. Arabidopsis CYCD3 D-type cyclins link cell proliferation and endocycles and are rate-limiting for cytokinin responses. Proc Natl Acad Sci U S A. 2007;104(36):14537–42.PubMedPubMedCentralCrossRefGoogle Scholar
- 96.Latrasse D, Germann S, Houba-Herin N, Dubois E, Bui-Prodhomme D, Hourcade D, Juul-Jensen T, Le Roux C, Majira A, Simoncello N, et al. Control of flowering and cell fate by LIF2, an RNA binding partner of the polycomb complex component LHP1. PLoS One. 2011;6(1):e16592.PubMedPubMedCentralCrossRefGoogle Scholar
- 97.Galbraith DW, Lambert GM, Macas J, Dolezel J. Analysis of nuclear DNA content and ploidy in higher plants. Curr Protoc Cytom. 2001; Chapter 7:Unit 7.6.Google Scholar
- 102.Lamesch P, Berardini TZ, Li D, Swarbreck D, Wilks C, Sasidharan R, Muller R, Dreher K, Alexander DL, Garcia-Hernandez M, et al. The Arabidopsis information resource (TAIR): improved gene annotation and new tools. Nucleic Acids Res. 2012;40(Database issue):D1202–10.PubMedCrossRefPubMedCentralGoogle Scholar
- 108.Rau A, Maugis-Rabusseau C. Transformation and model choice for RNA-seq co-expression analysis. Brief bioinform. 2017;19(3):425–36.Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.