Gene expression during the induction, maintenance, and release of dormancy in apical buds of poplar

Abstract

The perennial lifestyle of trees is characterized by seasonal cycles of growth and dormancy. The recurrent transitions into and out of dormancy represent an adaptation mechanism that largely determines survival and, hence, the geographical distribution of tree species. To understand better the molecular basis of bud dormancy, cDNA-amplified fragment length polymorphism (AFLP) transcript profiling was used to map differential gene expression during dormancy induction, dormancy, dormancy release by chilling, and subsequent bud break in apical buds of poplar (Populus tremula×P. alba). Unexpectedly, besides poplar transcript sequences, the cDNA-AFLP profiles revealed sequence signatures originating from a complex bacterial community, which was more pronounced during dormancy and displayed temporal dynamics in composition and complexity. Based on poplar gene expression dynamics, processes and potential regulators during different phases of dormancy are described. Novel genes were linked to a crucial transitory step in dormancy induction, and to dormancy release through chilling, a molecularly unresolved phenomenon. One WRKY- and two ERF-related transcription factors were similarly expressed during the transition to dormancy in apical and axillary buds. These regulatory genes could be involved in the differentiation of stipule-like leaf organs protecting the bud, or act during the growth–dormancy transition in the meristem, revealing commonalities between para- and endodormancy.

cDNA-AFLP, dormancy, gene expression, poplar

Introduction

The seasonal cycle of growth and dormancy is a distinct feature of perennial plants and represents one of the most basic adaptations of trees to their environment. The recurrent transitions of meristems into and out of dormancy are of primary significance to plant productivity and survival. These transitions are tightly linked to the yearly dates of bud flush and bud set, and delimit the growing season. Trees use environmental cues, such as daylength and temperature, to time growth–dormancy transitions properly in order to balance maximal growth and timely protection of their meristems against hazardous frosts. Poplars are known to initiate bud set and dormancy primarily in response to short daylength (Sylven, 1940; Nitsch, 1957).

The alternating cycles of growth and dormancy have gained renewed attention, because environmental information used for proper timing of seasonal growth, particularly temperature, is subject to climate change (Menzel and Fabian, 1999). Between 1981 and 1991, the length of the growing season has inceased by 5 d per °C temperature rise on average, or by 12 d at high latitudes (Zhang et al., 2004). The prolongation of the growing season translates to both of its ends and is particularly pronounced in urban areas where temperatures are 1–3 °C higher than in other land-surface areas. In these urban areas, green-up occurs 9 d and 7 d earlier and dormancy onset 4 d and 2.5 d later in North America and Europe, respectively (Zhang et al., 2004).

Despite their great importance for productivity and survival, processes related to the transition into and out of dormancy are poorly understood at the molecular level. Likewise, little is known about the processes that occur during the dormant period in woody perennials, particularly during the release from endodormancy (Rohde et al., 2000; Arora et al., 2003; Horvath et al., 2003; Rohde and Bhalerao, 2007). So far, research on dormancy has been driven by the practical need to define chilling requirements for dormancy release in economically important species, but has barely contributed to a mechanistic understanding of the dormancy process. This lack of knowledge might be partly due to the inaccessibility of the tissues in which dormancy is imposed, or their poor amenability to molecular methods, but at the same time illustrates our ignorance about the intriguing processes that, during dormancy, predetermine the growth potential for the next season.

Here, cDNA-amplified fragment length polymorphism (AFLP) transcript profiling was used to map gene expression during dormancy induction, dormancy maintenance, dormancy release by chilling, and subsequent bud break in Populus tremula×P. alba apical buds to gain a better understanding of the molecular basis of bud dormancy. Poplar has been used as a model for perennial plants because it has many advantageous characteristics for molecular biology, and is unrivalled now by the availability of its full genome sequence (Tuskan et al., 2006). This study uncovers 141 poplar genes that are differentially expressed during dormancy induction, maintenance, or release. Unexpectedly, in addition to poplar sequences, transcript profiling revealed bacterial sequence signatures, mostly derived from bacterial species that are putative plant associates. A number of transcription factors are similarly expressed during apical and axillary bud development, suggesting their significance at the transition from growth to dormancy.

Materials and methods

Plant material and growth conditions

All experiments were done with the P. tremula×P. alba clone INRA 717.1B4 that was propagated in vitro as described (Rohde et al., 2002). For the cDNA-AFLP experiment, plantlets were acclimated to greenhouse conditions in June 2001 (∼18 °C, without additional light). Plants of 35–45 cm height were transferred to a plant growth room (Conviron, Winnipeg, Manitoba, Canada) on 7 August 2001 (see Fig. 1 for the light and temperature regimes). Light intensity was 100 μmol m⁻² s⁻¹ at plant level, temperature was kept within ±0.5 °C of the requested temperature, and humidity was 70±10%. Re-growth tests under long-day conditions were conducted in the greenhouse (with extension of the photoperiod to 16 h light). For expression analysis by quantitative reverse transcription-PCR (RT-PCR), various tissues were collected and pooled from four actively growing plants (60 cm height) in the greenhouse. Axillary buds were harvested on 3 September 2001 from 8-month-old, actively growing non-branched trees in the greenhouse; plants (176 cm height) had 65 axillary buds on average. Greenhouse conditions were as above (with extension of the photoperiod to 16 h light). Axillary buds were pooled from 11 plants for 10 consecutive positional groups, each consisting of five axillary buds from top to bottom. For light microscopy, axillary buds were harvested at positions 1, 10, 20, 30, 40, and 50 below the apex from the same set of plants. Light microscopy was done as previously described (Rohde et al., 2002). For the apices from short-day-treated plants, plants of 60 cm height were transferred to a growth chamber with an 8 h photoperiod at a temperature of 21 °C. During 7 weeks, apices/apical buds were harvested at weekly intervals and were pooled from seven plants.

Fig. 1.

Experimental design for dormancy induction, maintenance, release, and regrowth. Light and temperature conditions as well as phenological and morphological observations are given on the time scale. Cessation of internode elongation, bud set, and dormancy were observed on days 21, 35, and 42 after the onset of short days, respectively. Material was sampled throughout the 115 d of the experiment, as indicated by stars. Sampling point 1 at day 0 corresponds to long-day-grown plants directly before the start of the experiment. Sampling point 30 corresponds to cold-treated plants prior to transfer to growth-promoting long-day conditions. The lowermost bars indicate the four major phases of expression revealed by cDNA-AFLP (see Fig. 3).

RNA extraction, cDNA-AFLP, and RT-PCR

RNA was extracted with LiCl from pools of three apices or buds as described previously (Rohde et al., 2002). Equal amounts of total RNA were subsequently analysed in RNA minigels (RNA 6000 nano assay; Agilent Technologies, Palo Alto, CA, USA). RNA was free of detectable bacterial or poplar genomic DNA contamination. All RNA samples had similar ratios of the 5S, 18S, and 28S rRNAs. Bacterial contamination in RNA extracts could, furthermore, not be revealed by amplification with universal primers targeting bacterial 16S rRNA in as much as 1 μg of total RNA of the samples (5′AGAGTTTGATCCTGGCTCAG3′ and 5′AAGGAGGTGATCCAGCCGCA3′; Van Oevelen et al., 2002). A 5 μg aliquot of total RNA was reverse-transcribed into cDNA. Successful cDNA synthesis was confirmed by amplification of actin (estExt_fgenesh4_kg.C_LG_I0082) with 5′CCAAGCAGCATGAAGATCAA3′ and 5′CACCCTTGGAAATCCACATC3′, and of ubiquitin (eugene3.00040082) with 5′ATGCAGATYTTTGTGAARAC3′ and 5′ACCACCACGRAGACGGAI3′ as forward and reverse primers, respectively. Changes in gene expression were monitored with cDNA-AFLP (Breyne et al., 2003). Of the 128 possible primer combinations, 120 were scored for differentially expressed transcript-derived fragments (TDFs). To avoid putative effects of minor differences in cDNA yield and length, gene expression was scored as presence or absence rather than as fold change.

Tag identification and database searching

Of 592 TDFs, fragments were cut out from gels at two different time points, allowed to dissolve in water, re-amplified with the respective primers, and directly sequenced. Only sequences longer than 50 bp were retained for further analyses. For 405 TDFs, two clearly homologous sequences were obtained from the two different time points, and could be merged into one consensus sequence. For 80 TDFs, only one sequence was available. Altogether, 485 sequences were taken through a number of quality filters: sequences were removed that contained concatenated primer (eight TDFs), encoded fungal (17 TDFs) and human (nine TDFs) contaminations, or had multiple database hits hinting at a chimeric structure (five TDFs). TDFs originating from the same gene, but originally identified with different primer combinations, were merged on the basis of sequence homology (41 TDFs; >95% homology).

Sequences were subjected to a BLASTX against GenBank, a BLASTN against all poplar accessions (genome, chloroplast genome, unassembled sequence reads; http://genome.jgi-psf.org/Poptr1_1/Poptr1_1.home.html), and a BLASTX against the Sequence Platform for the Phylogenetic analysis of plant Genes (SPPG; http://bioinformatics.psb.ugent.be/cgi-bin/SPPG/index.htpl). Annotations for the poplar gene models were retrieved from their respective best-hit Arabidopsis thaliana (or other) gene.

Clustering

The expression data of the 405 non-redundant TDFs were originally recorded as binary data with 0 and 1 representing the absence and presence of expression, respectively. The distance metrics applied to these binary data in all cluster analyses was Pearson correlation. Sampling point 8 (20 d) was removed because it had irregular patterns in a number of amplifications. For clustering of genes into phases A, B, C, and D, average expression values were calculated over the respective periods. All hierarchical and support tree clusterings were conducted in the MultiExperiment Viewer version 3.0 of The Institute for Genome Research with average linkage (Saeed et al., 2003).

RNA extraction and quantitative PCR

Total RNA was extracted according to Chang et al. (1993) with modifications. After LiCl precipitation, total RNA was resuspended and further purified with the RNeasy Mini Kit including the on-column DNase treatment (Qiagen, Helden, Germany). Double-stranded cDNA was synthesized starting from 2.5 μg of total RNA (Breyne et al., 2003). For quantitative PCR, a LightCycler 480 System (Roche Diagnostics, Brussels, Belgium) was used with the LightCycler 480 SYBR Green I Master (Roche Diagnostics). Primer combinations were: WRKY11 (estExt_fgenesh4_pg.C_LG_VI0537), 5′CCCCCTCGTAAGTCACGATA3′ and 5′GAACTCAAGCGCGATCTACC3′; expressed protein (grail3.0001032301), 5′TCCTCCACGCGACCTAATAC3′ and 5′TTTTCAGACTTGGTTGCCCT3′; AP2/EREBP (eugene3.00031319), 5′ATCCACCCCATATGCTCTGG3′ and 5′TATCTCGGCTGCCCATTTAC3′; and ERF4 (grail3.0014007501), 5′CTCACGCGAAGAATGATATTCC3′ and 5′CGACGACTGATGATGAATCC3′. For each qPCR run, reactions were done in triplicate with 2 μl of 50× diluted cDNA in a final volume of 5 μl. Two independent qPCR runs gave identical results. Melting curves indicated unique amplification products for each primer combination. Gene-specific amplification efficiencies were calculated based on the amplification curves with a linear regression model implemented in LinRegPCR software v.7.5 (Ramakers et al., 2003). Means of gene-specific amplification efficiencies were calculated over all positive reactions for each primer combination. Crossing point values were estimated with the second derivative maximum method implemented in the LightCycler 480 software (Roche Diagnostics). Relative expression values (±SD) were calculated with qBASE v1.3.3 (http://medgen.ugent.be/qBASE), taking into account gene-specific amplification efficiencies and with actin as a reference gene (see above).

Results

Controlled conditions mimick the process from dormancy induction to dormancy release

Under natural conditions of the temperate zone, the dormant period of poplar lasts for 5–6 months. Poplars synchronize the onset of the dormant period mainly to changes in daylength that are sensed by phytochromes and transduced through the CONSTANS/FLOWERING TIME LOCUS T regulon (Howe et al., 1996; Olsen et al., 1997; Böhlenius et al., 2006). Thanks to the responsiveness of poplars to daylength, the process of dormancy induction, dormancy, and dormancy release can be mimicked under controlled growth conditions. Figure 1 depicts the conditions that were applied to steer dormancy induction, maintenance, and release in poplar within 115 d. Key to these experiments are the changes in light and temperature: short days of 8 h together with declining temperature were applied to induce dormancy; a treatment at 4 °C gradually fulfilled the chilling requirement and plants were finally transferred to a 16 h long day and 21 °C for growth resumption.

The first visible event during dormancy induction was the cessation of internode elongation. Measurements of height increment at weekly intervals indicated that internode elongation had ceased 3 weeks after the onset of short days (Fig. 1; data not shown). Primordia that had been present at the perception of short days developed into mature, though progressively smaller leaves, whereas those that were initiated after the onset of short days acquired a different developmental fate; instead of a leaf with two stipules, they generated bud scales from the stipular domains of the primordium (Rohde and Boerjan, 2001). Bud scales were visible after ∼21 short days (Fig. 1). Inside the developing bud, organogenesis continued to generate embryonic leaves and leaf primordia that both remained unelongated and would only develop in the next growing season. The bud was set after 5 weeks. Dormancy was established after 6 weeks, as monitored by the inability to resume growth upon transfer to growth-promoting long-day conditions (Fig. 1). Leaf fall, which in nature often takes place shortly after bud set, occurred under controlled conditions within the first week of exposure to 4 °C (Fig. 1). During cold exposure, plants were checked for dormancy release by returning plants to growth-promoting long-day conditions. Growth resumption was recorded in apical buds as stage 4 of bud burst, i.e. leaves diverged with the blades still rolled up [International Union for the Protection of New Varieties of Plants (UPOV), 1981]. Plants needed 21 d and 13 d to resume growth when exposed to chilling for 23 d and 48 d, respectively, showing that dormancy was gradually released. After return to long-day conditions at the end of the 48 d chilling period, plants were harvested three times at 3 d intervals during the initial stages of bud swelling (UPOV stages 1 and 2). Altogether, samples of apices or apical buds were collected at 33 points in time over the whole experimental period of 115 d, with the frequency of sampling adjusted to the pace of visible changes (Fig. 1).

An expression map of 405 transcript-derived fragments reveals sequence signatures of poplar and bacteria

To describe the molecular processes during bud development and dormancy, cDNA-AFLP transcript profiling was performed. Based on their differential expression, 592 TDFs were selected for sequence analysis. A total of 405 non-redundant TDFs were subjected to database searches and clustering (Supplementary Supplementary Data and Supplementary Sequence File available at JXB online). The TDFs were compared with the P. trichocarpa genome sequence (http://genome.jgi-psf.org/Poptr1_1/Poptr1_1.home.html), poplar expressed sequence tags (ESTs), GenBank, and SPPG (Vandepoele and Van de Peer, 2005). Best hits below an e-value of e⁻³ were collected for each TDF (Supplementary Supplementary Data, Supplementary Data, and Supplementary Data at JXB online). According to the best database hit, the 405 TDFs were categorized to either poplar (141 TDFs) or bacterial (122 TDFs) origin. A third group (142 TDFs) had no significant hit in the databases and remained unassigned (no hit; Fig. 2a).

Fig. 2.

Identity of non-redundant TDFs. (a) Of the 405 TDFs, those with homology to poplar sequences were grouped under ‘poplar’ as well as those without homology to poplar sequences, but with homology to other plant sequences. TDFs without homology to either poplar or plant sequences, but with homology to sequences of bacterial origin were grouped under ‘bacteria’, and those without any hit under the ‘no hit’ group. Within parentheses, the average sequence length (± SD) is given for each of the groups as well as the distribution over classes of GC content. Columns correspond to the three categories of sequences: dark grey, poplar; grey, bacteria; open, no hit. (b) For the category ‘poplar’, TDFs were further classified in terms of their presence in the set of predicted gene models or on the genome and the location of the TDF within the gene. The scale is the absolute number of TDFs. TDF hits within the 5′ or 3′ UTR only were considered a hit within the 500 bp before the ATG or after the stop codon, respectively. (c) For the category ‘bacteria’, tags were further classified as being derived from an opportunist, symbiotic, aquatic, surface-located or contaminating bacterial species, as identified from the most homologous gene. Ten major associated bacterial genera were found during poplar genome sequencing (grey bars; Tuskan et al., 2006) and were compared with the fractions found in the experiment (open bars).

Of the 141 poplar TDFs, 118 TDFs had a significant hit with a predicted poplar gene. Of the seven TDFs that mapped elsewhere in the poplar genome (Fig. 2b), four had an additional hit with a poplar EST or another plant sequence, suggesting that these sequences were indeed transcribed, but that a gene had not currently been predicted at the corresponding genomic loci. Of the 118 TDFs conforming to a predicted poplar gene, most TDFs were derived from a coding region (70 TDFs) or from a coding region and 3′ untranslated region (UTR) (24 TDFs; Fig. 2b). Another 16 TDFs were clearly homologous with plant-derived sequences, but were absent from the P. trichocarpa genome sequence, most probably because of its current sequence depth (Tuskan et al., 2006).

The high abundance (30%) of bacteria-derived TDFs within the selected set of differentially expressed TDFs was surprising at first (Fig. 2c). The absence of polyadenylation of bacterial mRNA and the use of an oligo(dT) primer for cDNA synthesis should have prevented the amplification of bacterial fragments. Therefore, tag-flanking sequences were analysed for stretches of adjacent thymidine residues on the coding strand that would have been transcribed into adenosine in the mRNA. For 40 of the 122 tags, the tag-flanking sequences could be unambiguously inferred from a homologous bacterial gene identified through BLASTN algorithms. Fourteen of the 40 analysed genes had stretches of 5–7 adjacent thymidines on the coding strand. Additionally, 31 of the 40 genes had adenosine stretches of 5–20 adjacent residues within the 1000 bp up- and/or downstream of the genomic position of the tag sequence. Supposedly, the adjacent adenosine residues on either mRNA or single-stranded DNA were sufficient to tether an oligo(dT) primer during cDNA synthesis.

Based on the origin of the most homologous sequence, the 122 tags were assigned to >65 different bacterial species (Supplementary Supplementary Data at JXB online). Most tags identified opportunistic bacteria that are abundant in soil, on plant surfaces, and in the rhizosphere. Seventy-six tags had a best hit to bacteria that had been described as endophytes in other plants (Fig. 2c; Supplementary Supplementary Data at JXB online; Lodewyckx et al., 2002). Seven tags identified three symbiotic genera: Mesorhizobium, Rhizobium, and Sinorhizobium. Coincidentally, during the poplar genome sequencing, 1.05×10⁶ sequence reads (out of 7.6×10⁶ in total) remained unassembled and were assigned an endophytic origin (Tuskan et al., 2006). Of these, >80% originated from 10 major bacterial species (Fig. 2c). The same 10 genera accounted for 72% of the bacterial tags in P. tremula×P. alba, indicating that the bacterial community is highly similar to that of the sequenced P. trichocarpa (Fig. 2c). Together, these tags might thus be derived from the bacterial microflora present on plant surfaces, in the apoplastic space and in the vascular system.

The group of no-hit tags was not biased towards shorter sequences (Fig. 2a). Using different codon usage models (eukaryotic/bacterial), these sequences appeared to have no coding potential. The GC content of no-hit tags was biased towards values >50% and resembled that of the tags of bacterial origin (Fig. 2a). However, the GC content of short sequences might not be reliable enough to assign a bacterial origin conclusively. Alternatively, the no-hit TDFs might be derived from the 3′ UTR of transcripts of P. tremula×P. alba that were too divergent to be mapped onto the P. trichocarpa genome sequence.

In conclusion, the sequenced tags that were differentially expressed at various developmental stages of dormant buds identified 141 poplar-derived genes, 122 bacteria-derived genes, and 142 TDFs of currently unknown origin. This proportion suggests a close association of poplar with a bacterial community.

Poplar and bacteria contribute differentially to the expression profiles

The presence of genes of poplar and bacterial origins raised the question concerning their expression dynamics with respect to each other. Therefore, the expression patterns of the three sets were clustered to time separately (Fig. 3). Firstly and most strikingly, gene expression changed significantly after 24 d, shortly after the bud scales had become apparent, in all gene sets. At this stage, the expression of a large number of poplar genes stopped or was reduced, while many bacterial and unknown genes started to be expressed (Fig. 3). Secondly, all gene sets showed three chronological, more or less distinct phases of expression after 24 d. These observations led to expression data being grouped into four major phases, hereafter called A (sampling points 1–10, i.e. days 0–24), B (sampling points 11–18, i.e. days 27–45), C (sampling points 19–24, i.e. days 48–71), and D (sampling points 25–33, i.e. days 77–115).

Fig. 3.

Four major phases of expression revealed by clustering of the expression patterns with time. The support tree method with 1000 iterations and resampling through jackknifing genes was applied to cluster time points. The tree comprises 32 sampling points; sampling point 8 was removed (see Materials and methods). All branches of the trees were 100% supported in 1000 iterations. Reducing the number of branches by imposing a distance threshold revealed clusters (shaded grey) that were subsequently assigned as expression phases A, B, C, and D.

The expression during phase A comprised ∼70% poplar-derived genes (Fig. 3). In contrast, the three later stages contained >80% tags of bacterial and unknown origin, while the plant contribution to the expression profiles dropped to 16–21% (Fig. 3). Thus, genes of bacterial origin were detected from the onset of the experiment, but became more pronounced during later stages. The higher proportion of bacterial genes during bud dormancy could either result from a decrease in plant gene expression activity, or reflect a more abundant bacterial colonization during bud dormancy.

For the genes of bacterial and unknown origin, phases B and D were more similar to each other than to the intervening phase C. Gene expression during phase C was distinct from phase B and D, for two equally important reasons: a temporary drop in abundance of gene expression and a different composition indicated by the expression of phase-specific genes (compare the number of genes in clusters ABD and BD with clusters B, C, and D; Fig. 4).

Fig. 4.

Clustering of the 122 bacterial TDFs. Hierarchical tree clustering used to group average expression values of the four phases of gene expression. The 122 outer nodes were reduced by imposing a distance threshold (shaded blue). Each cluster was named according to the major phase(s) of expression.

For the poplar genes, gene expression during phases B, C, and D was more alike and generally less abundant than in phase A. Phase C clustered separately but close to phases B and D, with the latter two not distinguished by clustering (Fig. 3). Strikingly, these phases in poplar gene expression did not coincide with the changes in temperature and light nor did they exactly correspond to the time that bud scales, bud set, dormancy, or regrowth visibly appeared (Figs 1 and 3). Clustering the expression patterns of 141 poplar genes during the four above-defined expression phases covered the whole experimental period (Fig. 5a), but a majority (89 TDFs, 55%) were expressed primarily in phases A and AB (Fig. 5a). The functional classification of the 141 TDFs resembled the relative abundance of functional categories revealed in an extensive poplar EST sequencing analysis (Fig. 5b; Sterky et al., 2004). The largest functional groups among the 141 TDFs were ‘unclassified’ (23%) and ‘metabolism’ (21%), followed by ‘transcription’ (16%) and ‘energy’ (9%; Fig. 5b). When TDFs expressed exclusively during phase A (dormancy induction up to 24 short days) were compared with those expressed only in phases B, C, or D (dormancy maintenance and release), the fractions of the categories ‘energy’, ‘transcription’, and ‘cellular communication/signal transduction’ were smaller in the phases corresponding to dormancy, in line with the assumed generally lower activity (Fig. 5c).

Fig. 5.

Clustering of the 141 poplar-derived TDFs. (a) Hierarchical tree clustering used to group average expression values of the four phases of gene expression. The 141 outer nodes were reduced to eight by imposing a distance threshold (shaded blue). Each cluster was named according to the major phase(s) of expression. (b and c) Categorization of 141 TDFs into functional classes and occurrence in the different expression clusters of (a). (b) Comparison of the relative fraction of poplar ESTs and dormancy-related TDFs in functional classes. Data on 102 019 ESTs from 19 poplar cDNA libraries were retrieved for comparison from Sterky et al. (2004). Out of the 141 non-redundant genes, 139 could be categorized in the same UPSC-MIPS classification schedule as used by Sterky et al. (2004). (c) Comparison of 68 TDFs expressed exclusively during dormancy induction (phase A) with 42 TDFs expressed exclusively during later stages (phases BCD). Two TDFs could not be assigned unambiguously to a functional category and were excluded: an actin-related protein (ARP2, CCAC220) and a lipid transfer protein (GTCT310) present in clusters AB and A, respectively. Note that 5.6% of the TDFs belonging to ‘cellular transport and transport facilitation’ were integrated into the functional category ‘metabolism’ for reasons of comparability with the data of Sterky et al. (2004) in (b), but were presented separately in (c).

Changes in metabolism and energy generation during dormancy

Of the 141 poplar genes, 40 genes encode a function in metabolism and energy generation. Genes encoding enzymes in lipid breakdown (acyl-CoA hydrolase, TCAG195), in the citric acid and the glyoxysomal cycle (isocitrate dehydrogenase, GCAT280; malate synthase, GCAA250; fumarylacetoacetate hydrolase, ACTA350; acetyl-CoA synthase, GCCC225), and in glycolysis (glyceraldehyde 3-phosphate dehydrogenase, CCCC205; pyruvate kinase, CCGC305) were expressed throughout the process, although to a lesser extent during phase C (Fig. 6). Three additional lipolytic enzymes were also mainly expressed during phases A and B (APG proteins, GCAA500 and TCCA260; GDSL-motif lipase, CCTC200; Fig. 6). Thus, conceivably, lipid, fatty acid, and isoprenoid breakdown occurred at the shoot apex and inside the forming bud already during the induction of dormancy, in contrast to previous suggestions that breakdown of lipid bodies would play a specific role in energy supply only later during dormancy (Sagisaka, 1991; Fig. 6).

Fig. 6.

Expression patterns of selected poplar genes in the cDNA-AFLP. Expression patterns are given together with an identifier and a short annotation. Complete information on all genes is available in Supplementary Supplementary Data. Expression profiles were clustered within each of the groups. The group with main expression in phases C and D only displays genes that were not included in any of the other functional categories; other genes with main expression in phases C and D were grouped in carbohydrate metabolism (TCTA190 and GCTT380), protein degradation and amino acid metabolism (ATAG200a), or transcription factors and signal transduction (GTCC360 and TCTG255).

Most of the genes related to amino acid metabolism encode enzymes for the breakdown of amino acids. Half of these genes started being expressed in phase B (Fig. 6). Because many proteins were broken down during concurrent leaf senescence, amino acids were possibly further metabolized to create new carbon skeletons for storage compounds prior to dormancy. Protein- and peptide-degrading enzymes were found during phases A and B, or B and thereafter (chloroplast-derived protease, ATCC253; peptidases, ACCC280 and CCTC365; RPN6 subunit of the 26S proteasome, GCCC195; Fig. 6).

Among the genes related to carbohydrate metabolism, the expression dynamics of genes encoding enzymes involved in polysaccharide cleavage and oligosaccharide production were remarkable. These genes were detected during phases A and B (pectate lyase, ACTA250; pectin methylesterase, ACTC375). Galactinol synthase (CTAA340), a key regulatory enzyme in the biosynthesis of the raffinose family of oligosaccharides, was expressed from 1 week after the onset of short days and remained expressed throughout the dormant period (Fig. 6). High levels of stachyose and raffinose are known to accumulate during winter dormancy in P. balsamifera and might be involved in osmoprotection (Bachelard and Wightman, 1973). Moreover, the gene encoding myo-inositol-1-phosphate synthase, expressed during early stages of growth resumption (GCTT380, Fig. 6), is closely positioned to galactinol synthase in the respective pathway. Altogether, the dynamics of lipid and carbohydrate metabolisms are in agreement with previous observations that poplar stores reserves mainly as carbohydrates (Nelson and Dickson, 1981).

Regulatory genes are identified during all phases of dormancy

Of the 40 genes with a putative regulatory function (transcription, protein fate, cellular communication, and signal transduction), the majority (31 genes) were expressed in A and AB (Fig. 5). Most of them had been already expressed in long days, but their expression was discontinued after the first 24 d of dormancy induction (phase A, 26 genes). A CCCH-type zinc finger protein (CCGC370a) was down-regulated within the first 2 weeks of short days and might be specifically associated with a process that was terminated early after the onset of short days (Fig. 6).

The most interesting group consisted of genes that were first expressed after 2 weeks of short days for up to 10 d (Fig. 6). Some of these genes might be involved in regulating the distinct change in expression observed after 24 short days and, hence, might control an important step during dormancy induction (Fig. 3). These include WRKY11 (CTAG330), an AP2/EREBP transcription factor (CTAT185), ethylene response factor 4 (ERF4, GCGG230), and AKIN β 15′-AMP-activated protein kinase (TTCT460). The consecutive expression cluster in phase B contained a C3HC4-type RING finger (GCCT400) and an RNA helicase (GTGT310; Fig. 6).

During phases B and C, a peptidyl prolyl cis/trans-isomerase (PPIase, TTGC310a) was expressed (Fig. 6). PPIases are ubiquitously present in all cells and act in protein folding, protein translocation through membranes, and signal transduction through cis/trans-isomerization of peptide bonds. Interestingly, particular PPIases, to which the identified poplar gene shows the highest homology, have been linked to cell cycle progression (Yao et al., 2001). Furthermore, a DNA-binding protein containing MYB and linker-histone domains (TCTG255; similar to MYB1 of Petroselinum crispum) and a DnaJ-class molecular chaperone (GTCC360) were expressed in phases C and D, respectively (Fig. 6). DnaJ homologues are thought to regulate the ATPase activity of Hsp70s in processes of protein folding, translocation, and protein complex assembly. Additionally, other genes with main expression during phases C and D provided interesting candidate genes for functions during the satisfaction of chilling requirement and growth resumption (Fig. 6). The role of these proteins during dormancy, particularly of those expressed in phases C and D, awaits future functional studies.

Regulatory genes with expression dynamics at the transition from growth to dormancy in apical and axillary buds

The most striking feature in the gene expression profiles was the marked change in gene expression after 24 short days (Fig. 3), suggesting that this time reflects a crucial step in dormancy induction. Four genes were identified whose differential expression coincided with this transition, suggesting a role in dormancy induction. To investigate whether their differential expression was more generally associated with the growth–dormancy transition, WRKY11 (CTAG330), an AP2/EREBP transcription factor (CTAT185), ERF4 (GCGG230), and an expressed protein gene (GTCC365; Supplementary Supplementary Data at JXB online) were selected for further experiments. The expression patterns originally observed in the cDNA-AFLP experiment with gradually decreasing temperature were largely reproducible in an independent short-day experiment with constant temperature (Fig. 7a). WRKY11 expression increased distinctly after 4 weeks of short days. The expressed protein gene (GTCC365) was transiently induced during week 1 and 2 of short days. Thus, the time of induction of these two genes was slightly shifted as compared with the cDNA-AFLP experiment, in which they were expressed at 22–24 short days (Fig. 7a). The expression of AP2/EREBP and ERF4 was similar in both independent experiments, and occurred at 13–24 short days and 15–24 short days, respectively (Fig. 7a).

Fig. 7.

Gene expression patterns of three selected transcription factors and one expressed protein. (a) Gene expression profiles during apical and axillary bud development and in different plant organs. Gene expression levels were estimated by qPCR and were given relative to the expression in the actively growing apex included within each series. Average values with the SD of triplicate qPCR runs are presented. Left panels show the temporal expression profile during short-day-induced dormancy establishment in apical buds. Apices were sampled in long-day controls (week 0) and then at weekly intervals after transfer to short days (weeks 1–7). Middle panels show the expression profile in developing axillary buds along a poplar stem. A series of axillary buds, harvested at intervals at increasing distance from the apex, represent a gradient of developmental stages of axillary buds. Right panels, expression in various organs: ap, actively growing apex; in1+2, first two internodes below the apex; in5, internode 5 below the apex; yl, young leaf at position 6 (largest leaf that is not yet fully expanded); ml, mature leaf at position 20 (fully expanded leaf); yp, petiole of young leaf at position 6; mp, petiole of mature leaf at position 20; sp, all stipules of positions 1–14; ax, all axillary buds of positions 4–23; xy, developing xylem of a debarked stem at 5–15 cm above ground; ba, bark of the stem at 5–15 cm above ground; st, the lowest 5 cm of the stem including bark, xylem, and pith; ro, washed soil-grown roots. Positions are counted from the apex downwards. (b) Light microscopy of axillary buds at positions 1, 10, 20, 30, 40, and 50 below the apex. (c) Number of organs observed within the axillary buds at positions 1, 10, 20, 30, 40, and 50 below the apex. At each position, axillary buds from six plants were examined by microscopy and results are given as the mean number (± SD).

To substantiate further a link between the expression of these genes and dormancy induction, the transcript levels of all four genes were analysed in a series of axillary buds that were harvested in groups at increasing distance from the apex of an actively growing plant. This series represents a gradient of developmental stages of axillary buds from initiation, organogenesis, and growth, to dormancy. Through microscopy, it was determined when growth, i.e. organ formation within the axillary bud, was initiated, was fully active, and finally ceased. The first distinguishable axillary bud below the apex only contained cataphylls, i.e. true bud scales that arise directly from the primordium (Fig. 7b). Leaf primordium initiation was first observed in the 10th axillary bud and occurred until position 30 below the apex (Fig. 7c). Below position 30, organ number in axillary buds did not increase further, indicating that the transition to dormancy occurred in axillary buds below position 30 (expression group 7; Fig. 7a). Expression of three out of the four investigated genes markedly increased at precisely this position (WRKY11, expressed protein, and ERF4) and these genes might thus participate in dormancy imposition in axillary buds, similar to their role in apical bud development (Fig. 7a). The expression of AP2/EREBP was 3-fold higher in all axillary buds, irrespective of their position along the shoot, when compared with the apical bud (Fig. 7a). Because the first developmental stages of axillary buds only contained cataphylls (Fig. 7b, c), it is intriguing to link this transcription factor to scale-like leaf organs.

An analysis of tissue-specific expression showed that expression of the WRKY11 gene was highest in the mature leaf, maturating xylem, and stipules (Fig. 7a). The expressed protein gene was preferentially expressed in meristems, such as the apex and axillary buds. AP2/EREBP was distinctly expressed in stipules, underlining its potential function in scale-like leaf organs (cataphylls, bud scales, and stipules). AP2/EREBP was, moreover, expressed in apex and axillary buds, where ERF4 expression was also the highest. Together, the preferential expression in tissues that can undergo growth to dormancy transitions (meristems), or in organs that are similar to bud scales (stipules), underscores a putative link of these transcription factors to bud development and dormancy.

Discussion

The transition from growth to dormancy involves a major change in gene expression

Little is known on the molecular basis of the seasonal transition from growth to dormancy, although this adaptation mechanism largely determines survival and, hence, the geographical distribution of tree species. Previous molecular studies on this developmental process have been centred on single genes or covered shorter time frames (Olsen et al., 1997; Rohde et al., 2002; Pacey-Miller et al., 2003; Espinosa-Ruiz et al., 2004; Schrader et al., 2004; Welling et al., 2004; Ruonala et al., 2006; Ruttink et al., 2007).

The most striking feature among the gene expression profiles in apical buds was the global change in expression pattern after plants had been exposed to 24 short days (Fig. 3). This time probably represents a crucial step during the induction of dormancy. Profound changes in bud structure become apparent shortly thereafter: bud scales start to develop, initiated leaf primordia cease elongating, and internode elongation has stopped (Fig. 1; Rohde and Boerjan, 2001; Rohde et al., 2002). Phenotypes of impaired bud structure, provoked by the overexpression of either ABI3 or of a dominant-negative form of ETR1, manifest from that time onwards (Rohde et al., 2002; Ruonala et al., 2006). Similarly, a microarray experiment on poplar bud set pinpointed the major change in gene expression at 3–4 weeks after the onset of short days (Ruttink et al., 2007). Thus, these gene expression changes precede the most clear-cut change in apical bud morphology. However, the gene expression changes at ∼24 d might be attributed partly to dormancy induction within the meristem that occurs concomitantly with bud formation.

A bacterial community in apical buds

Unexpectedly, the cDNA-AFLP profiles contained signatures of both poplar and bacterial origin (Fig. 2). Importantly, the association of poplar with bacteria would not have been revealed in targeted approaches for monitoring differential gene expression, such as microarrays. There was no evidence for pathogenesis during the experiment, because (i) plants were visibly healthy; (ii) none of the identified bacterial sequence signatures originated from known plant-pathogenic species; (iii) typical genes involved in plant–pathogen interaction or use of plant metabolites were absent from the bacterial genes; and (iv) no characteristic pathogen defence genes were found among the differentially expressed poplar genes.

More than half of the bacterial sequences were derived from bacteria known for their ability to interact or associate with plants. The 10 most abundant genera of associated bacteria found during poplar genome sequencing (Tuskan et al., 2006) covered 72% of all bacterial tags in our profiling, too (Fig. 2). Thus, our study confirmed the presence of a largely similar bacterial community in yet another poplar species and additionally provided evidence for changes in the composition of the bacterial community during the dormant period. Two quite similar bacterial communities have been identified in other poplars in which 21 and 53 bacterial genera have been characterized and whose structure depended on the host genotype and the external environment (Porteous Moore et al., 2006; Ulrich et al., 2007).

In general, the composition of the endophytic community will depend on its chemical and physiological environment (Kuklinsky-Sobral et al., 2004; Bailey et al., 2005). These conditions change drastically during the transition from growth to dormancy and might explain the differential accumulation of bacteria during various phases of the experiment (Figs 3 and 4). In elm (Ulmus spp.) branches, a gross comparison of the bacterial community during the growing and the non-growing season revealed significant differences in species composition and abundance (Mengoni et al., 2003). In poplar buds, a bacterial community became particularly pronounced during dormant stages (Fig. 3). This observation could either result from a decrease in plant gene expression activity, reflect a more abundant bacterial colonization during bud dormancy, or hint at a differential gene expression from bacteria in response to signals from the changing extracellular environment during dormancy induction. Altogether, the presence of bacteria, albeit often intentionally excluded from analysis as unwanted and fortuitous contamination, might be more dynamic than previously appreciated.

Cornerstones for dormancy in apical buds

Poplar-derived transcripts covered the complete experimental period and provided cornerstones for an expressional framework that underlies the physiological processes during dormancy (Figs 5 and 6). Genes of the carbohydrate metabolism and energy generation pathways underpin previous biochemical studies (Fig. 5; Bachelard and Wightman, 1973; Nelson and Dickson, 1981). Not many other hitherto dormancy-associated genes were found in our gene identification effort (Supplementary Supplementary Data at JXB online). Considering the technical limitations of the cDNA-AFLP, ∼60% of all poplar genes were monitored. Only 5–10% were constitutively expressed throughout dormancy induction, persistence, and release, during which the tremendous transcriptome changes taking place have yet to be uncovered. These changes emphasize that, besides the transition to flowering, dormancy is also one of the most distinct modifications in the life cycle of trees.

β-Tubulin, which has been suggested to be a marker for monitoring dormancy in tree buds (Bergervoet et al., 1999), is a basic structural unit of microtubules whose function is required for cell division and cell elongation. β-Tubulin was expressed up to 34 d in short days and then again at 6 d and 9 d of growth resumption in long days (Supplementary Supplementary Data and Supplementary Data at JXB online). Thus, expression of β-tubulin was found in active and not in dormant tissues. However, the ability to resume growth was completely abolished only after 42 short days (Fig. 1), suggesting that β-tubulin expression is terminated before endodormancy entry. In poplar cambium, the cell cycle is arrested sequentially, involving an ecodormant stage prior to endodormancy (Espinosa-Ruiz et al., 2004). Thus, β-tubulin expression might be terminated in an arrested (ecodormant) stage, explaining its absence prior to endodormancy in the buds.

The chilling requirement through which dormancy is released has remained a central enigma of dormancy research. A number of physiological parameters are known to change after the chilling requirement has been satisfied: respiration is restored, and oxidized glutathione is converted to reduced glutathione (Faust and Wang, 1993). Also, an increase in fermentative metabolism towards dormancy release has been noted (Or et al., 2000). Genes corresponding to these physiological processes were not present in the subset of sequenced genes during the corresponding expressional phases C and D. Instead, 19 genes mainly expressed during phases C and D, although not belonging to a particular pathway, are interesting novel candidate genes for functions during satisfaction of the chilling requirement (Fig. 6). These include six genes of unknown function and a DNA-binding protein with linker-histone domains with a potentially regulatory role in dormancy release (expression in phases C and D; Fig. 6).

Regulatory modules for dormancy induction

The most striking feature in the gene expression profiles was the change in gene expression after 24 short days (Fig. 3). Three regulatory genes, AP2/EREBP, ERF4, and WRKY11, and one expressed protein gene were associated with this crucial step in dormancy induction (Fig. 7). Based on the function of the respective Arabidopsis homologues, the three regulatory genes could act downstream of ethylene and/or abscisic acid (ABA) signals during dormancy induction. Ethylene and ABA signal transduction are sequentially activated during bud development (Ruttink et al., 2007). An increase in ABA concentration has been shown in poplar apical buds at approximately 24 short days (Rohde et al., 2002). The closest homologue of the poplar AP2/EREBP transcription factor in Arabidopsis (At5g13330, RAP2.6L) belongs to an ERF subgroup that also comprises RAP2.6 and ABR1 (AP2-like ABA repressor 1; McGrath et al., 2005; Nakano et al., 2006). RAP2.6L acts in a network that regulates many genes during shoot regeneration from root explants (Che et al., 2006). Unfortunately, two different Arabidopsis rap2.6l insertion mutants had no obvious phenotypes throughout development (Che et al., 2006; A Rohde, unpublished data). RAP2.6L expression is distinctly up-regulated upon ABA treatment (www.genevestigator.ethz.ch). The best Arabidopsis homologue of ERF4 (AtERF4, At3g15210; McGrath et al., 2005) is transcriptionally induced by ethylene, jasmonate, and ABA. ERF4 has been independently identified during short-day-induced bud set of poplar (Ruttink et al., 2007). AtERF4 acts as a transcriptional repressor capable of modulating ethylene and ABA responses. Overexpression of AtERF4 leads to ethylene insensitivity and decreased ABA sensitivity (Yang et al., 2005). Together, AP2/EREBP, ERF4, and WRKY11 are identified as putative regulators of a crucial transitory step during dormancy induction in apical buds. Future studies will reveal their function and importance.

A similar transition from activity to dormancy exists in the developmental gradient of axillary buds along a growing shoot (Fig. 7). The distinct change in expression of ERF4 and WRKY11 at positions where paradormancy is imposed on axillary buds implies that these regulatory genes participate in the establishment not only of seasonal endodormancy in the apical bud, but also of paradormancy in axillary buds (Fig. 7). This observation suggests interesting parallels between these two different types of dormancy.

Another common component of endo- and paradormancy is that protective bud scales cover the meristems in which dormancy is imposed. AP2/EREBP, expressed preferentially in stipules (Fig. 7), might act specifically in scale-like organs: the bud scales in apical buds (of stipular origin) and cataphylls in axillary buds. Cataphylls are already present in the first axillary bud below the apex, concomitant with a high expression of AP2/EREBP from the first axillary bud onwards (Fig. 7). During apical bud development, bud scales are apparent after ∼3 weeks in short days, again coinciding with a high expression of AP2/EREBP. Together, regulatory proteins that are commonly expressed in endo- and paradormancy for the growth to dormancy transition in the meristem or for the differentiation of stipule-like leaf organs have been identified.

In conclusion, differential gene expression was observed during all stages throughout dormancy induction, maintenance, release, and initial stages of bud break, reinforcing the view that active developmental changes occur during dormancy. Dormant trees accomplish important molecular changes that are required for the re-initiation of growth in the next spring. One example of such changes consists of the new set of genes that were linked to dormancy release through chilling. Throughout the process, interesting structural and regulatory genes have been identified and form a basis for future investigations. Moreover, the diverse bacterial community found in dormant buds needs to be investigated for the putative benefits of a close association of both the bacterium and the plant.

Supplementary material

The following supplementary data are available at JXB online.

Supplementary Data. Expression data for the 405 non-redundant TDFs and their assignment to clusters (cluster and row) in the hierarchical clustering analyses presented in Figs 5 and 6.

Supplementary Data. Complete annotations for 141 plant/poplar TDFs with the best-hit sequence identifiers for eight different database subsections.

Supplementary Data. Annotation for 122 TDFs of bacterial origin and their putative assignment to functional category and species.

Sequence File. Sequence file with the sequence reads of 405 TDFs in FASTA format.

Abbreviations

Abbreviations
 
• ABA

  abscisic acid

 
• AFLP

  amplified fragment length polymorphism

 
• EST

  expressed sequence tag

 
• qPCR

  quantitative PCR

 
• TDF

  transcript-derived fragment

 
• UTR

  untranslated region

The authors thank Xue Yongchang for technical assistance, Dr Johan Van Huylebroeck (Instituut voor Landbouw- en Visserijonderzoek, Melle, Belgium) for providing the growth room facilities, and Martine De Cock for help in preparation of the manuscript. This work was supported in part by the European Commission program POPYOMICS (QLK5-CT-2002-00953). TR and LS are indebted to the Institute for the Promotion and Innovation by Science and Technology in Flanders and the ‘Bijzonder Onderzoeksfonds’ of Ghent University for a postdoctoral and a predoctoral fellowship, respectively. AR is a Postdoctoral Researcher of the Research Foundation-Flanders.

Author notes

*

Present address: Institute for Agricultural and Fisheries Research, 9090 Melle, Belgium