2.1 Plant material

Axillary and apical bud samples were collected monthly from adult and juvenile Q. suber trees located at the University of Minho campus (https://www.icampi.uminho.pt/pt/ambiente/green/, latitude: 41°33 ‘33.8” N, longitude: 8°23′54.3” W), throughout four annual cycles (summer 2015 to spring 2019). All samples were taken from the tree crown from 9:00 a.m. to 10:00 a.m. using pruning shears, if not mentioned otherwise. The monthly samples of 2015/2016, 2016/2017, 2017/2018 were constituted by a mixture of approximately twenty randomly collected buds from several trees. In 2018/2019, to investigate whether gene expression was similar in different trees, each monthly sampling was made in triplicates, which were used as biological replicates. To analyze the circadian gene expression profile, buds were also collected from adult trees at different time points over the course of the day (24 hours, 4 hour intervals), in October, January and April.

2.2 RNA extraction and cDNA preparation

Total RNA was obtained using the CTAB/LiCl extraction method (Chang et al., 1993) with some modifications (Azevedo et al., 2003). Total RNA was treated with DnaseI (Rnase-free) (New England Biolabs), and 0.5–1 ug of RNA was used for cDNA synthesis using the Invitrogen cDNA synthesis kit SuperScript® RT, according to the manufacturer's instructions.

2.3 Quantitative RT-PCR (qRT-PCR) analysis

Amplification was carried out with SsoFast™ EvaGreen®Supermix (Bio-Rad), 250 nM of each gene-specific primer (Table S1) and 1 μL of cDNA. Transcript-specific primers were designed according to data in the Q. suber transcriptome (Pereira-Leal et al., 2014). Quantitative real-time PCR (RT-qPCR) reactions were performed in triplicates on the CFX96 Touch™ Real-Time PCR Detection System (Bio-Rad). After an initial period of 3 min at 95°C, each of the PCR cycles consisted of a denaturation step of 10 s at 95°C and an annealing/extension step of 10 s at the temperature optimized for each gene-specific primers. With each PCR reaction, a melting curve was obtained to check for amplification specificity and reaction contaminations, by heating the amplification products from 60 to 95°C in 5 s intervals. Primer efficiency was analyzed with CFX Manager™ Software v3.1 (Bio-Rad), using the Livak calculation method for normalized expression (Livak and Schmittgen, 2001). Gene expression analysis was conducted using three technical replicates and normalized with the reference genes QsPROTEIN PHOSPHATASE 2A SUBUNIT A3 (QsPP2AA3) and ELONGATION FACTOR-1ALPHA (QsEF-1α) (Marum et al., 2012).

2.4 Tissue fixation, embedding and sectioning

For morphological and immunolocalization analysis, buds were collected and immediately fixed in 4% (w/v) paraformaldehyde in 1× PBS (phosphate buffered saline: 1x PBS 130 mM NaCl, 7 mM Na₂HPO₄, 3 mM NaH₂PO₄, pH 7.4) under vacuum infiltration followed by overnight incubation in the fixative solution at 4°C. Samples were then dehydrated, cleared and paraffin-embedded according to the protocol described by Coen et al. (1990) with some modifications: during the clearing procedure, the tissues were incubated 1 h at room temperature, in each of the following solutions: 75% (v/v) ethanol/25% (v/v) Histo-Clear® (VWR Chemicals), 50% (v/v) ethanol/50% (v/v) Histo-Clear®, 25% (v/v) ethanol/75% (v/v) Histo-Clear®; and incubated three times for 1 h in 100% (v/v) Histo-Clear®. Tissue sections with 8 μm thickness were obtained using a microtome (SLEE) and mounted in pre-coated poly-L-lysine slides (VWR), deparaffined by heating at 65°C for 15 min and washed with Histo-Clear® six times (15 min each), and further rehydrated with a grade of ethanol series (100, 80, 70, 50 and 30% (v/v), 5 min in each solution) ending with deionized water twice (5 min each).

2.5 Immunolocalization of 5-Methylcytosine and Post-translational Histone Modifications

Sections were permeabilized with sodium citrate buffer (10 mM Sodium citrate, 0.05% (v/v) Tween 20, pH 6.0) and by a heat-mediated antigen retrieval treatment, using the microwave at maximum power, for 4 min according to Nic-Can et al. (2013). After cooling down for 20 min, the sections were washed three times in 1x PBST (130 mM NaCl, 7 mM Na₂HPO₄, 3 mM NaH₂PO₄, 0.05% (v/v) Tween). After blocking with 8% (w/v) BSA in 1x PBS for 30 min at 4°C, the sections were incubated with primary antibodies diluted in 1x PBST: anti-5mC (1:100 dilution, Abcam AB10805), anti-H3K4me3 (1:50 dilution, Abcam AB8580), and anti-H3K18ac (1:100 dilution, Abcam AB1191) at 4°C during 48 h in a moist chamber. Negative controls were done by replacing the primary antibody with 1 x PBST. After washing in 1x PBST buffer (3 times, 5 min), samples were incubated for 3 h at 37°C, using a solution containing goat polyclonal secondary antibody to mouse or rabbit IgG – H&L conjugated to Alexa Fluor 488 (Abcam AB150113, AB150077), added in a 1:100 dilution. The slides were mounted in EverBrite™ Mounting Medium with 4′, 6-diamidino-2-phenylindole (DAPI, Biotium) and were examined with a Spinning Disk Carl Zeiss Axio Observer Z1 microscope using 10x magnification or with a Plan Apo 63x Ph3- NA1.4 –DIC HCII Oil immersion objective. Images were acquired with an AxioCamMR3 camera controlled by Axiovision 4.8.2.0 software. Zeiss fluorescence filters Set 49 and Set 38 HE were used to excite DAPI, and Alexa Fluor 488, respectively. Z-stacks with 0.2 μm were acquired to allow for the entire nuclei capture. The images of antibody detection and DAPI staining were acquired using the same microscope settings and light exposure times. Evaluation of nuclei fluorescence intensity of z-stack maximum intensity projections was performed using Fiji software (Schindelin et al., 2012) after image deconvolution using Huygens Essential (Scientific Volume Imaging B.V., Netherlands). Nuclear staining intensity was evaluated, at least, in 30 nuclei per meristem. Background fluorescence intensity was measured as the average intensity of inter-nuclear areas and divided from the nuclear staining intensity of each nucleus for correction. The intensity measurement of each epigenetic mark was obtained by dividing the labelling intensity of the specific signal in each nucleus by the correspondent DAPI intensity, as previously described (Bian et al., 2009; Leida et al., 2012; Testillano et al., 2013; Ingouff et al., 2017; Zhai et al., 2018). Data shown is representative of at least three independent experiments and the values are represented by arbitrary units.

2.6 Homology-based search and phylogenetic analysis

In order to identify Quercus suber homologs of known genes involved in dormancy in other species, pea and Arabidopsis DYL (Stafstrom et al., 1998; Tatematsu, 2005), poplar CENL1 (Mohamed et al., 2010) protein sequences were used as queries for BLASTp searches at NCBI for Q. suber (Qs), Vitis vinifera (Vv), Orysa sativa (Os), Jatropha curcas (Jc), Mallus domestica (Md), Arabidopsis thaliana (At), Quercus robur (Qr), Pisum sativum (Ps), Prunus persica (Pp) and Populus trichocharpa (Pt) homologs. The full-length amino acid sequences of each homolog (Table S2) were aligned with CLUSTALW, using the Gonnet Protein Weight Matrix. The Multiple Alignment Gap Opening penalty to 3 and the Multiple Alignment Gap Extension penalty to 1.8 (Hall, 2013) were used as alignment parameters. Phylogenetic analyses were performed with the MEGA 7.0 software program (Kumar et al., 2016) using the Jones-Taylor-Thornton (JTT) correction model, inferred by the Maximum-likelihood algorithm. The resulting trees were constructed based on 1000 bootstrap replicates.

2.7 Statistical analysis

The raw data were imported into GraphPad Prism V6.0 software to calculate descriptive statistic parameters such as means and standard deviations.

Differences in the labelling fluorescent intensity/DAPI intensity between tissues and/or developmental phases were tested through two-way ANOVA followed by Tukey's multiple comparison test at a 5% significance level.

2.8 Meteorological support data

Air temperature and accumulated cold hours (<7.2°C) were determined based on the monthly weather report bulletin (Instituto Português do Mar e da Atmosfera-IPMA) with the data monitored at the nearest meteorological station (Braga-Merelim, 41.57586944,-8.45110833, 65) from University of Minho campus (41.56236104, −8.39401066).

3.1 The complex nature of active growth and dormancy periods in Quercus suber

Quercus suber trees have meristematic tissues and organ primordia enclosed in terminal and axillary buds that cease activity and enter dormancy during autumn (Figure S1A), possibly in response to shorter days and/or lower temperatures, similar to what has been described for other Quercus species (Derory et al., 2006). The buds contain vegetative (Figure 1A, red arrows) or both vegetative and reproductive meristems (Figure 1B, highlighted in red boxes) and stay dormant throughout winter, probably due to the persistency of low temperatures. Usually, in late February/March, an increase in day length and/or temperature signals the end of the dormancy period and induces bud swelling (Figure 1C). Male flowers with well-defined anthers (Figure 1D) emerge in the axils of the swollen axillary buds (Figure S1B). In April/May, male flowers complete their development, and new shoots emerge containing young leaves and female flowers (Figure S1C). A new axillary bud emerges in the axils of the leaves that are not flanked by a female flower spike (Figures 1E and S1D). In the buds collected in June/July (Figure S1), the differentiation of male catkin primordia (Figure 1F,G, red boxes) adjacent to the apical meristem (Figure 1F, red arrow) was observed.

The vegetative and reproductive development of Q. suber shows other particularities that are frequently observed in the field. In some years with unusually warm temperatures in autumn or early winter, premature bud swelling occurs with the emergence of male flowers (but rarely female flowers). In spring, many buds fail to swell, and the timing of bud swelling fluctuates between and within trees.

3.2 QsCENL and QsDYL1 were differentially expressed throughout bud development

It has been previously suggested that CENL and DYL1 may suppress bud growth in other species (Soeda, 2004; Liu et al., 2007; Ruonala et al., 2008; Mohamed et al., 2010; Rinne et al., 2011; Lee et al., 2013; Footitt et al., 2017) but it is not known if the function is conserved in Fagaceae. Homologous sequences of CENL and DYL1 family members were retrieved from the Q. suber genome by homology-based search. CENL belongs to the FT family, together with FT TERMINAL FLOWER 1 (TFL1), TWIN SISTER OF FT (TSF) and MOTHER OF FT (MFT) (Wickland and Hanzawa, 2015). The phylogenetic tree was divided into distinct clades, each one associated with a different sub-family (Figure 2A). Two homologs belong to the TFL1 clade (QsTFL1 and QsCENL), another to the FT clade (QsFT) and the fourth to the MFT clade (QsMFT) (Figure 2A). Two homologous proteins of DYL1 were found in Q. suber but only one belongs to the DRM1/DYL1 clade (Figure 2B). The phylogenetic tree suggested that QsDYL1 is the closest homolog to PsDYL1/DRM1 and AtDRM1, considered as good dormancy markers in axillary buds of pea and in Arabidopsis seeds, respectively (Stafstrom et al., 1998; Tatematsu, 2005).

The transcription levels of CENL and DYL1 were evaluated during the bud growth-dormancy cycle for four consecutive years. QsCENL expression showed a conserved profile in all the years under study: it was higher during bud set (September/October), then decreased and remained at low levels until March (Figure 3). However, QsCENL was upregulated in the swollen buds (in April) in two of the periods (2015/2016; 2018/2019) (Figure 3). To distinguish the involvement of QsCENL in dormancy vs. flowering development, its expression was also analyzed in buds collected from Q. suber juvenile trees (not competent to flower) in October, January and March. The expression pattern was the same as the one observed in adult trees, with higher expression in October and low expression throughout the following months (Figure S2). To analyze if the QsCENL expression profile is affected by the circadian rhythm, buds were also collected from adult trees at different time points over the course of the day (24 hours, 4 hour intervals), in October, January and April. The expression of QsCENL was higher in the October bud set regardless of the time of day or night (Figure S3).

In adult trees, QsDYL1 expression increased after bud set, usually in September/October (Figure 3). Then, the expression decreased until the emergence of swollen buds in April (Figure 3). In juvenile trees, high levels of QsDYL expression were found in October, then, the expression decreased until the end of dormancy (March), similarly to what was observed in adult trees (Figure S2). However, levels of QsDYL expression had significant variation between the years of study. Similarly, bud samples collected at different time points during the day in October and January had significant variations regarding QsDYL transcript levels (Figure S3). Nevertheless, the same trend was observed: higher expression during bud dormancy (January) (Figures 3, S3).

As gene expression can be regulated by the activity of chromatin modifiers, in the following sections, we have analysed the expression of these genes and the deposition of epigenetic marks in dormant versus non-dormant bud meristems.

3.3 The expression of DNA methyltransferases and histone acetyltransferases was switched on and off throughout bud development

A likely epigenetic control of bud dormancy in Fagaceae was first reported in C. sativa with an emphasis on DNA methylation and histone acetylation (Santamaría et al., 2009, 2011). To assess the role of DNA methylation and histone acetylation during bud development, the expression of genes coding for enzymes that catalyze the deposition of these epigenetic marks was analyzed. The Q. suber epigenetic modifiers were phylogenetically identified in a previous study (Silva et al., 2020). Here, we studied gene expression of five DNA methyltransferases (QsMET1, QsDRM2, QsDNMT2, QsCMT3 and QsCMT1) and three histone acetyltransferases (QsHAF1, QsHAM1 and QsHAC1) during bud development (Figure 4 and S3). QsCMT1 transcript accumulation in the buds was very low (Figure S4). An increase in QsMET1 expression was observed prior to April bud swelling (Figure 4), while QsDNMT2 expression generally increased after October and decreased in April (Fig.4). QsCMT3 was highly expressed during late stages of bud formation when the buds were still growing (July to August/September), but the expression decreased gradually from bud set (September/October) to bud dormancy (Figure 4). The expression increased again from February to April, being highly expressed in swollen buds (April) (Figure 4). The expression of QsHAC1 was higher during growth cessation and dormancy when compared to bud swelling (Figure 4). QsHAF1 and QsHAM1 expression profiles were not conserved in the different years (Figure S4).

To infer if QsCMT3 up- or down-regulation is associated exclusively with bud dormancy or bud growth instead of reproductive development, its expression was measured in vegetative buds of juvenile trees during two different years. QsCMT3 was downregulated in both years during dormancy establishment (from October to December) and upregulated during dormancy release (March) (Figure S2) in a similar manner to adult trees. The buds collected from adult trees at different time points of the day were also used to analyze the circadian regulation of QsCMT3. The results showed that the gene is always highly expressed in the swollen opened buds (April) and less expressed in the months where bud set is occurring, or dormancy is already established (October and January), regardless of the time of the day at which the samples were collected (Figure S3).

3.4 Differential distribution of epigenetic marks in dormant and non-dormant meristems

To get insights into how chromatin modifications may affect dormancy, the distribution of epigenetic marks in buds at different developmental stages was analysed. Immunolocalization of the repressive mark 5-methylcytosine (5mC) and active marks such as acetylation of histone H3 at lysine 18 (H3K18ac) and trimethylation of histone H3 at lysine 4 (H3K4me3) (Figure 5) were performed in buds containing both reproductive and/or vegetative meristems or tissues before growth cessation in summer, in dormant winter buds, and during bud swelling in the spring.

During the late stages of bud formation in the summer, the three epigenetic marks were detected in both apical and axillary meristems (Figure 5A-I, II, III). The marks were also observed in the early stages of male catkin primordia and in the flower meristems found in the already differentiated male catkins (Figure 5B – I, III). Throughout dormancy, the marks were likewise observed in both vegetative and reproductive primordia (Figure 5C-I, II, III), in male flowers and also in the apical and axillary meristems within the swollen bud (Figure 5D-I, II, III).

The intensity level of each epigenetic mark was evaluated in apical and axillary meristems, as well as in male flower meristems, during the late stages of bud formation (summer), bud dormancy (winter) and anters during bud swelling period (spring) (Figures 6, 7 and 8). A higher intensity of 5-mC signal was detected in male flower meristems and in the apical meristems of dormant buds when compared to the corresponding structures of summer and spring (Figure 6). H3K18 acetylation and DNA methylation in the apical meristem showed an opposite pattern regarding the transition from late stages of bud formation to bud dormancy (Figures 6 and 7). In both apical and axillary meristems, H3K18ac signal intensity was higher during the late stages of bud formation (Figure 7D, F). On the other hand, male flower meristems (Figure 7B) had lower levels of H3K18ac in comparison to the terminal (Figure 7D) and axillary meristems (Figure 7F). H3K18ac appeared to be similarly represented in male flower meristems throughout the dormancy-growth cycle (Figure 7B, H, N). Tri-methylation of H3K4 was weaker in male flower meristems of dormant buds (Figure 8H) when compared with the ones found during late stages of bud formation (Figure 8B) and during male flower emergence (Figure 8B). During bud swelling period, H3K4me3 was significantly higher in male flower meristems (Figure 8N) than in axillary meristems (Figure 8R).