Introduction

To survive temperate climates, overwintering plants require the ability to acclimate to freezing temperatures. This cold acclimation, taking place after exposure to chilling temperatures, involves the activation of multiple mechanisms that collectively contribute to transiently enhance freezing tolerance. Many changes at the cellular and molecular levels participate in this complex biological process. Metabolomic studies on cold stress have demonstrated that active reconfiguration of the metabolome is strongly regulated by changes in gene expression (Guy et al. 2008; Krasensky and Jonak 2012). Proteins responsible for frost tolerance mainly include enzymes involved in the biosynthesis of protective molecules, osmoprotectants, lipid desaturases, and proteins involved in protein turnover and cell detoxification (Ruelland et al. 2009; Thomashow 1999). To overcome the occurrence of frost, plants have evolved a variety of adaptive mechanisms which do not operate in all species to the same extent and some of them which may be specific to tolerant plants. Although low temperature sensing and transduction may be common to many plant species, the protective mechanisms may differ (Monroy et al. 2007). Especially, molecular studies on model plants indicate different mechanisms of cold acclimation between herbaceous and woody species (Meng et al. 2008).

In Arabidopsis, more than 700 cold-regulated genes encoding both regulatory and functional proteins have been identified (Oono et al. 2006). Collections of cold-responsive ESTs have been generated in Populus (Nanjo et al. 2004; Park et al. 2008), blueberry (Dhanaraj et al. 2004, 2007), Rhododendron (Wei et al. 2005), Poncirus (Meng et al. 2008; Sahin-Cevik and Moore 2006), Malus (Wisniewski et al. 2008), and Eucalyptus (Keller et al. 2009; Rasmussen-Poblete et al. 2008). Gene expression macro-arrays for aspen (Populus), spruce (Picea), and pine (Pinus) have provided the first near-global pictures of the autumn transcriptome in trees (Holliday et al. 2008). Some of the genes that are up-regulated appear to be involved in cold hardiness, but it is difficult to unravel gene expression related to the concomitant process of dormancy induction.

Endemic to Australia, Eucalyptus is the most widely planted forest tree worldwide because of good adaptability, fast growth rate. and excellent wood and fibre properties (Grattapaglia et al. 2012). However, as an evergreen without endodormancy, Eucalyptus is particularly exposed to frost and its survival over winter mainly relies on its ability to cold acclimate. Among the about 700 species of the genus, the hardiest (Eucalyptus gunnii) can tolerate down to −18 °C after hardening. It is cultivated in some temperate European areas or used for the introgression of frost tolerance traits into more productive Eucalyptus species (Teulieres et al. 2007). Cold acclimation was shown to be accompanied by an accumulation of sucrose, fructose and raffinose in the leaves of E. gunnii (Leborgne et al. 1995a). In addition to carbohydrate content, plasmalemma fluidity was found to be a major factor of E. gunnii frost tolerance (Leborgne et al. 1992). Global genomic technologies were more recently applied to Eucalyptus towards identifying important genes involved in cold tolerance and improving this quantitative trait. From cold-treated E. globulus, one of the most productive and cold-sensitive species, an EST collection was reported (Rasmussen-Poblete et al. 2008). Among the 3,879 singletons, the authors highlighted the significant amount of genes predicted to be involved in cell wall biosynthesis as potential targets for improving wood quality through metabolic engineering. In our hands, the expressed sequence tags generated from leaves of cold-acclimated E. gunnii by a single-pass sequencing of random cDNA clones led to the identification of 11,303 genes (Keller et al. 2009). This functional annotation provided a first indication of cold acclimation transcriptome for E. gunnii. More recently, the E. grandis genome sequence was released, providing a powerful resource to the scientific community (Myburg et al. 2011).

In the present study, a global expression analysis was carried out on a random sub-set of the initial E. gunnii EST collection to gain insights into the mechanisms underlying cold acclimation. Using macro-array technique, the transcripts were quantified during the time course of a cold acclimation programme. In the present paper, the temporal gene expression patterns are described with regard to gene identity and corresponding functional categories. The aim of this work on the frost tolerant species E. gunnii was to unravel the coordination of gene regulation in response to cold stress and more precisely to identify the main mechanisms taking place during the different steps of cold acclimation.

Materials and methods

Cold treatment of plants and frost tolerance measurements

Three-year old plants from a cold-tolerant clone of E. gunnii (line 941366), provided by AFOCEL (South Station, Montpellier, France) were grown in standard conditions (25 °C day/23 °C night, 16 h day-length and 115 μmoles m2 s−1 radian flux density (Lumilux Daylight 58 W, Osram, München, Germany). Relative humidity was about 80 % and plants were watered as needed.

Four plants were cold-acclimated as previously described (El Kayal et al. 2006). Grown under short-day photoperiod (12 h light) and a reduced radian flux density (43 μmoles m2 s−1), the plants were first transferred to 12 °C day/8 °C night for 4 days and then to 4 °C night and day for 12 days. To investigate frost tolerance of E. gunnii plants, relative leaf cell viability after freezing at −2 °C/h was evaluated by measuring electrolyte leakage on leaf discs (from five leaves per plant), at −6 °C as previously described (Leborgne et al. 1995a). Frost tolerance is given as the mean of four plant measurements and expressed as the percentage of cell viability after freezing compared with the viability of unfrozen leaf discs. Statistical analysis using the Student test was performed to determine the significantly different values for cold tolerance.

For gene expression studies, leaves were collected at different time points after temperature changes to 12 °C (15, 30 min, 2, 48, 72, and 96 h) and to 4 °C (15, 30 min, 2, 5, 10, 24, 48, 72 h, and 6 days).

Annotation and functional categorization of the sequences

2,662 cDNA clones were randomly taken from the previously analysed cold-acclimated E. gunnii cDNA library (11,303 ESTs) (Keller et al. 2009). Using BLAST, the E. gunnii sequences were compared with the whole E. grandis genome sequences available at Phytozome (http://www.phytozome.net/eucalyptus.php/). The resulting annotation was completed using BLASTx (http://blast.ncbi.nlm.nih.gov/) at an e-value (expected value) cut-off of 1e−05 against sequence from the public databases TAIR10 (www.arabidopsis.org) and Uniref 100 (www.ebi.ac.uk/uniref/). Finally, the annotation assigned to a contig was also associated with all the ESTs composing it.

Functional categories for TAIR-annotated hits were proposed using the criteria previously established (Ruepp et al. 2004) for FunCat software (ftp://ftpmips.gsf.de/catalogue/annotation_data/athaliana_funcat2008 and (ftp://ftpmips.gsf.de/catalogue/funcat-2.1_scheme), available on Munich Information Center for Protein Sequences (MIPS). The resulting classification was validated manually for all the considered hits. Using PHP scripts, the validated annotation data were associated with the corresponding expression results in a MySQL database available at (http://www.polebio.lrsv.ups-tlse.fr/eucatoul/coldexpress/).

Macro-array production

From the cDNA library, constructed from leaves of cold-acclimated E. gunnii (Keller et al. 2009), 2,662 cDNA bacterial clones were randomly picked. Using the primers 5banqseq and 3banqseq (Table 1), PCR amplifications (95 °C for 2 min, 95 °C for 30 s, 52 °C for 30 s, 72 °C for 2 min for 40 cycles, and 72 °C for 7 min) were performed twice in a 100-μL reaction volume for each of the cDNA bacterial clones carried by the pTriplEx2 vector, as well as for the empty pTriplEx2 vector as a negative control. The pBlueScript SK vectors (Stratagene, La Jolla, CA, USA) carry two control genes from E. globulus isocyanate dehydrogenase (EgIDH) and glyceraldehyde 3-phosphate dehydrogenase (EgG3PDH), which were amplified using T7 and M13R primers.

Table 1 Primer sequences used for macro-array production and RT-qPCR analysis
Full size table

PCR products were pooled, evaporated, and then resuspended to get 800–1,000 μg/μL. After denaturation by the addition of an equal volume of DMSO, they were transferred in duplicate onto Immobilon-Ny+ nylon membranes (Millipore) using a MicroGridII robot (Biorobotics). These filters were then denatured twice for 10 min on Whatmann 3 MM paper saturated with 0.5 M NaOH, 1.5 M NaCl, neutralized twice for 10 min on Whatmann 3 MM paper saturated with 1 M Tris-HCl pH7.5, 1.5 M NaCl and rinsed for 10 min on Whatmann 3 MM paper saturated with 2× SSC. After air drying for 30 min at ambient temperature and 2 h at 80 °C, they were UV-crosslinked (120,000 μJ/cm2).

Macro-array hybridization and analysis

RNA was extracted from the collected leaves using the SV Total RNA Isolation kit (Promega, Madison, USA). The PolyAtract mRNA isolation System (Promega, Madison, USA) was used for the purification of poly A+ RNA from total RNA. The RNA extracts from leaves collected at the different time points were pooled in five sets, which correspond to the very early period at 4 °C or 12 °C (“vE” = 15, 30 min, and 2 h), the early period at 4 °C (“E4” = 5, 10, 24 h), the late period at 12 °C (“L12” = 2, 3, and 4 days) and at 4 °C (“L4” = 2, 3, 6 days). The first-strand cDNA was synthesized from 10 μg total RNA using SuperScript III reverse transcriptase (Invitrogen) and 80 pmol anchored oligodT (Q-BIOgene) in presence of RNAsin (Promega). Probes were labelled by the addition of dCTP [α-33P] in the cDNA synthesis reaction mix. After RNA alkaline hydrolysis, labelled first-strand cDNA was purified by isopropanol precipitation. Labelled first-strand cDNA were denatured for 10 min at 100 °C and cooled on ice for 5 min.

To release loosely bound DNA, macro-array filters were pre-hybridized for about 10 h in hybridization buffer (5× Denhart’s, 5× SSC, 0.5 % SDS, 100 μL/mL sonicated salmon sperm). After labelled targets were added to 10 mL fresh hybridization buffer, macro-array filters were hybridized for about 20 h at 65 °C. They were then washed at 65 °C, twice in 3× SSC, 0.5 % SDS for 15 min and once in 1× SSC, 0.5 % SDS for 15 min. Filters were exposed to a phosphor screen imaging plate for 8–24 h that was scanned with Fujibas 5000 at 50 μm resolution. Detection and quantification of the signal were performed using the ArrayGauge 1.3 software (Fujifilm). Absolute values given by the software were imported in Microsoft Excel programme (Microsoft Corporation, Redmond, WA, USA) and analysed as described by Pesquet (Pesquet et al. 2005). Prior to any analysis, reproducibility of hybridization on a single membrane was assessed by comparing the signal of two duplicate spots on the same membrane. Any EST for which duplicate signals differed more than twofold were discarded from further analysis (including establishment of normalization curve). Normalization between samples was established using the linear slope defined by the blank background, unspecific hybridization, and positive hybridization controls (water, empty pTriplEx2, EgIDH, and EgG3PDH) in the different samples. The slope defined by the controls indicated a linear correlation factor of R 2 = 0.9989. After normalization, the average signal of two membrane duplicates was calculated and compared between two independent hybridization experiments. Here again, if the difference was higher than twofold, the EST was discarded from further analysis. The average signal value from the membrane duplicates and hybridization replications (four values in total) were calculated for each EST. Expression data were then transformed from raw intensity value to relative gene induction level by dividing the value for cold-treated plants (one of the five time point) by its value for control plants (25 °C). Genes that showed <0.5-fold or greater than a twofold difference in average signal values for at least two kinetic points along the acclimation programme were defined as differentially expressed.

Identification of gene expression patterns

The expression profiles of the annotated and cold-regulated ESTs were clustered using hierarchical clustering explorer 3 (HCE3) software. From the resulting 20 clusters, those exhibiting similar expression pattern (Cl-5-6, Cl-15-16-17, Cl-7-8, and Cl-11-13) were manually associated to get substantially sized groups presenting characteristic patterns (respectively, patterns I, II, III, and IV). The larger clusters (Cl-9 and Cl-10), divided into three sub-groups using HCE3 software, correspond to expression patterns V and VI. The other clusters containing <20 ESTs were not considered for further analysis. For the associated clusters, the mean induction rates of ESTs belonging were calculated at each time point (vE12, L12, vE4, E4, and L4) to get the global shape of gene expression patterns.

Real-time RT-qPCR

Using SuperScript II and random primers (InVitrogen, France), cDNAs were produced according to the manufacturer’s instructions. Specific primers for each gene were designed using the Primer Express software (version 2.0, Applied Biosystems, France) and are given in Table 1. The PCR reactions were performed in 20 μL of SYBR Green Master mix (Applied Biosystems), with 10 ng of cDNA and 300 nM of each primer. Three replicates of each PCR were run in an ABI PRISM 7900HT 295 Sequence Detection System (Applied Biosciences, France) using a programme including a first step (50 °C for 2 min and 95 °C for 10 min) followed by 40 cycles (95 °C/15 s and 60 °C for 1 min). Specific primers for 18S RNA or IDH were used as the internal control for the normalization of the RNA steady-state level, and the relative changes in gene expression were quantified using the 2−ΔΔCt method (Livak and Schmittgen 2001). The results of relative transcript abundance are presented as the mean value of three assay replicates compared with the mean of three control values.

Results

Cold acclimation of Eucalyptus plants

Eucalyptus plants (E. gunnii clone 941366) were cold-acclimated using a progressive decreasing temperature programme and frost tolerance was measured using ion leakage method after freezing down to −6 °C, as described in “Cold treatment of plants and frost tolerance measurements”. As shown in Fig. 1, the relative cell viability nearly doubled during the first step of the cold acclimation programme (12 °C day/8 °C night). An additional progressive increase was observed after 3 days at 4 °C and cell viability reached 52.36 % at the end of the cold culture period (16 days). The statistical analysis (Student test) assessed that this increase of tolerance is significant at 98 %. Since this clone exhibits a large extent of cold acclimation, it is very suitable for studying expression of genes putatively involved in this adaptive response.

Fig. 1
figure 1

Cold acclimation in E. gunnii Frost tolerance increase of three-year-old plants exposed to chilling temperatures in comparison to plants grown in standard conditions. At different time points of the cold acclimation programme (4 days at 12 °C and 8 days at 4 °C), the freezing tolerance at −6 °C is measured on leaf discs using ion leakage method. For plants grown in standard conditions, the resulting viability (7.3 %) corresponds to the basal tolerance (grey bars). For cold-cultured plants, the acquired frost tolerance (white bars) is increasing progressively until 45 % at the end of the cold acclimation. The significantly different values of tolerance (Student test, 98 %) are labelled using different letters into brackets

Full size image

Annotation and FunCat classification of the EST collection (2,662 clones)

From the cold-acclimated E. gunnii cDNA library (11,303 ESTs) (Keller et al. 2009), 2,662 cDNA clones were randomly taken to be spotted onto macro-array nylon filters. The corresponding sequences, assembled in 1,704 unigenes (878 contigs and 826 singletons) were mapped on the whole E. grandis genome sequence, available at Phytozome (http://www.phytozome.net/eucalyptus.php/). As described in “Annotation and functional categorization of the sequences”, the annotation was completed using TAIR10 and Uniref 100 public databases. Finally, after manual visualization of the alignments, a putative annotation could be proposed for 2,308 ESTs (86.7 % of the 2,662 ESTs under study) including 53 sequences corresponding to “unknown” or “uncharacterized” proteins.

The FunCat system, defined by the MIPS, allowed classifying these annotated sequences. In contrast to the GO annotation scheme used previously (Keller et al. 2009), the FunCat architecture is characterized by a simple and hierarchical structure (Ruepp et al. 2004) which leads to the description of protein function and cellular localization. The FunCat classification is particularly suitable when the set of annotated sequences is also analysed for transcriptome profiling. However, for data interpretation, it should be kept in mind that each annotated sequence can fit into more than one functional category and/or cell localization. From the 2,308 annotated sequences, 2,191 hits could be classified in functional categories and 1,166 associated with cell localization, 504 and 227, respectively, found in chloroplast and mitochondria. Since the study aims at identifying the main mechanisms involved in cold acclimation, the unclassified proteins and those classified only according to localization are not considered for further analysis. The distribution of 1,581 hits within the functional categories is shown Table 2. The main functional category, corresponding to “Information pathways” class, contains genes encoding proteins with binding function (665 hits), or involved in protein fate (234 hits) or synthesis (204 hits). Including 524 hits, the “Metabolism” category is also well represented, noticeably with 203 hits involved in carbohydrate metabolism. With a similar size (499 hits) “Perception and response to stimuli” class is mainly composed of the sub-categories “Interactions with the environment” and “Cell rescue and defence” (respectively, 347 and 288 hits). In the present study, gene expression data are analysed with regard to the described FunCat classification. This approach helps uncovering the co-regulated genes belonging to related functional groups which could be involved in the main cellular mechanisms of E. gunnii cold acclimation.

Table 2 Functional classification of E. gunnii ESTs isolated throughout the cold-acclimation process
Full size table

Representative expression patterns of cold-regulated genes

Gene expression was investigated on 2,662 cDNA clones, spotted onto macro-array filters. As described in “Identification of gene expression patterns”, the labelled targets correspond to five pools of leaf samples harvested throughout the cold acclimation programme. Ratios of hybridization signals from cold-exposed versus control plantlets grown at 25 °C provide the gene induction rates. With an expression ratio higher than two for at least one of the five time points, 1,639 sequences (61.5 %) were found to be induced by cold. In contrast, only 31 ESTs were considered as down-regulated, their expression ratio ranging between 0 and 0.5 and most often this inhibition is observed only at one time point. The whole dataset is available at (http://www.polebio.lrsv.ups-tlse.fr/eucatoul/coldexpress/).

Using HCE3 software, the 1,670 cold-regulated sequences were clustered into 20 clusters according to their kinetic of expression (Fig. 2). Using the same software, the largest ones (Cl-9-10), composed of, respectively, 345 and 687 clones were further divided into three sub-clusters. In contrast, some clusters containing a lower amount of clones were manually grouped together according to the global shape of their expression patterns (Cl-5-6, Cl-7-8, Cl-11-13, and Cl-15-16-17). Finally, the remaining clusters (Cl-1-2-3-4-12-14-18-19-20) were not further considered because they contain <20 cold-regulated ESTs and it was not possible to combine them with a similar cluster. The 11 clusters considered in this study are composed of 1,596 clones, which represent 95.6 % of the 1,670 cold-regulated sequences. To provide representative expression patterns, the mean expression value at each time point of the acclimation process (vE12, L12, vE4, E4, and L4) was calculated from the macro-array signal of ESTs belonging to associated clusters. On the resulting graphs (Fig. 3), vertical bars represent the variation between the highest and the lowest induction rates corresponding to different ESTs of each pattern. These patterns exhibiting at least one peak at each temperature phase can be characterized according to their main induction phase. Both displaying peaks at two early time points (vE12 and E4), patterns I and II represent an “early induction”. However, the highest induction rate was observed at 12 °C for pattern I and at 4 °C for pattern II. These groups, very different in size, account together for 19.7 % of the ESTs. The mean induction rates are up to three- or fourfold for most of the clones composing these clusters, except for cluster 15 peaking at sevenfold at E4. The patterns III, IV, and V (37.3 % of the ESTs) share a common first peak at 12 °C (“late-12-induction”), but they distinguish by the time course of gene regulation at 4 °C. Except for cluster Cl-9.3, the patterns III and V show the second induction peak at the latest time-point at 4 °C (L4), while the pattern IV is characterized by an earlier induction peak at 4 °C (E4). These clusters are characterized by a moderate induction ranging between three- and fourfold. Finally, the pattern VI, corresponding to the largest group (687 ESTs, 43 % of the total), exhibits a “progressive” induction. During the two temperature-phases of cold acclimation, the continuous increase of gene induction results in the highest induction rate (up to 11-fold) at the end of the cold exposure. It is worth noting that the gene induction observed for pattern VI is parallel to the increase in cold tolerance during the acclimation process. Finally, alteration in gene expression is stronger at 4 °C compared with 12 °C and at late time points compared with the early ones.

Fig. 2
figure 2

Expression patterns of cold-regulated genes throughout the cold acclimation process. Expression profiles of the 1,670 cold-regulated sequences with a predicted identity were clustered using HCE3 software. The 20 resulting clusters show the global shape of the gene expression ratio throughout the cold acclimation programme at 5 kinetic points: vE12 very early at 12 °C (15, 30 min or 2 h); L12 late at 12 °C (2, 3, and 4 days); E4 early at 4 °C (5, 10 or 24 h); L4 late at 4 °C (2, 3, 4, and 6 days)

Full size image
Fig. 3
figure 3

Characteristic expression patterns of cold-regulated genes throughout the cold acclimation process. Expression profiles of the 1,670 cold-regulated sequences with a putative annotation were clustered using HCE3 software. From the 20 resulting clusters (C1-1 to Cl-20, Fig. 2), the presented characteristic patterns correspond to clusters 9 and 10 (both composed of 3 sub-groups) in addition to manually associated clusters (Cl-5-6, Cl-7-8, Cl-11-13, and Cl-15-16-17). Based on the mean expression ratio of the ESTs belonging to associated clusters, the diagrams show the global shape of gene expression pattern throughout the cold acclimation programme. The vertical bars represent the extend of induction rate between the most and the less cold-induced ESTs (see “Materials and methods”). C25 control plantlets at 25 °C; vE12 very early at 12 °C (15, 30 min or 2 h); L12 late at 12 °C (2, 3, and 4 days); E4 early at 4 °C (5, 10 or 24 h); L4 late at 4 °C (2, 3, 4, and 6 days)

Full size image

Validation of expression data by quantitative real time RT-qPCR analysis

The validation of macro-array data by RT-qPCR was performed on 12 genes selected according to identity, expression pattern and abundance in the library. Based on the orthologous sequences in E. grandis genome, specific primers were designed for amplifying these genes (Table 1). The transcription rate of these genes was then quantified using RT-qPCR analysis on cDNA samples used for library construction and macro-array experiments.

Constans and Scarecrow which encode transcriptional factors and HSP which act as molecular chaperone are induced mainly in the first phase of cold acclimation (Cl-9.2), and are representative of “late 12 induction”. SPIRAL and MT3 (Metallothionein) genes are early cold-induced at 12 and 4 °C (Cl-15). Exhibiting a distinct regulation, the Metallothionein MT2 is co-regulated with the largest group of ESTs (Cl-10) exhibiting a “Progressive induction”. In the same group, the late embryogenesis abundant (LEA) and dehydrin gene were selected because of the high redundancy of their ESTs, β-amylase for known importance of carbohydrates in cold acclimation and Ozone stress-responsive gene for its very high induction rate.

As shown in Fig. 4, the characteristic expression patterns were overall conserved between macro-array and real-time PCR experiments for most of the genes. As expected, the RT-qPCR results provide higher induction rate than macro-array signal, except for LEA-like and β-amylase and dehydrin. Overall, the expression patterns obtained using RT-qPCR analysis match with macro-array patterns thus validating the gene clustering.

Fig. 4
figure 4

Validation of macro-array data by RT-qPCR experiments. The validation of macro-array data by RT-qPCR was performed using 12 genes belonging to the main patterns, selected according to identity, expression pattern and abundance in the library. Based on the orthologous sequences in E. grandis genome, specific primers were designed (Table 1). RT-qPCR experiments were performed on cDNA samples used for library construction and macro-array experiments. Specific primers for 18S RNA or IDH were used as the internal control for the normalization of the RNA steady-state level, and the relative changes in gene expression were quantified using the 2−ΔΔCt method. The results of relative transcript abundance are presented as the mean value of three assay replicates compared with the mean of three control values (standard deviations are indicated). To allow the comparison of the expression profiles from both analyses, macro-array and RT-qPCR data are presented in the same graph using the most suitable scale for each

Full size image

Molecular functions associated with co-regulation patterns

To draw a global picture of the mechanisms involved in cold acclimation, the extent of gene induction was analysed with regard to the molecular function of the cold-regulated genes. From the 1,596 ESTs composing the 11 selected clusters, 915 (corresponding to 313 hits) were selected according to their quantitative importance in the transcriptome (Online Resource Table 1). Distributed among the main expression patterns, these ESTs are induced by at least three- or twofold and corresponding genes are represented, respectively, by at least one or more than two ESTs. As shown in Table 3, the “early induction” group contains only 66 genes (313 ESTs) and most of them exhibit both no redundancy and low induction rate (less than fivefold). However, this small group also contains 20 genes which are strongly regulated (from 5- to 19-fold) and 19 are redundant (up to 54 ESTs). The “late-12-induction” group is composed of 157 ESTs (99 hits) from which only 13 exhibit more than a fivefold induction rate and 35 are redundant (up to 13 ESTs). In contrast, the 585 ESTs (148 hits) belonging to the “progressive induction” pattern are characterized by a strong regulation (up to 45-fold), but only a few genes are highly induced. The prominence of this group lies also in the high redundancy of these sequences since 50 hits are redundant (up to 203 ESTs). The most represented sequences (14 genes, 305 ESTs) correspond to cryoprotection of membranes and molecules or antifreeze activity. With a similar number of genes (17) but less ESTs (124), the second predicted protective mechanism consists in homeostasis and detoxification. Then, protein degradation via the proteasome is represented by nine genes (24 ESTs), osmoprotection by five genes (18 ESTs), and preservation of photosynthesis by five genes (28 ESTs). Additional genes are also involved in the stabilization of cellular structures like cell wall, membranes or microtubules.

Table 3 Comparison of the temporal groups based on gene induction rates and EST redundancy
Full size table

According to the molecular function reported in the literature for the 313 selected genes (Online Resource, Table 1), the hierarchy between the different biological mechanisms of cold acclimation was then investigated. For each gene induction pattern, the most relevant identities corresponding to predicted effector proteins were then selected and classified according to the biological mechanisms known to be involved in cold acclimation (Table 4). The analysis of these data provides an overview of the co-regulated genes throughout the E. gunnii acclimation likely to represent the main protective mechanisms leading to hardening.

Table 4 Typical cold-induced genes involved in the main biological mechanisms of cell protection
Full size table

Patterns I and II (“early induction”) include proteins associated with photosynthesis preservation like rubisco activase, auxin-induced CP12-domain and photosystem II light harvesting complex. In addition, it also includes proteins participating in carbohydrate metabolism like rubisco or galactinose synthase (GolS). The GolS activity leading to raffinose accumulation is known to participate not only in cell cryoprotection but also osmoprotection and detoxification. The cell rescue category also contains LEA which are amphipathic proteins acting as cryoprotectors of molecules and membranes, as well as dehydrin, known as the specific sub-group LEA II in Arabidopsis and the well-known cellular chaperone heat shock protein (HSP). Through the retardation of ice crystal formation, the pathogenesis-related protein (PRP) and chitinase may avoid wounding of the cell membrane through antifreeze activity. The metallothionein MT3 which is the most cold-induced of the “early induction” group (54 ESTs, 9.24-fold rate) is known to be involved in homeostasis, as well as many genes encoding proteins associated with cell detoxification like catalase, thioredoxin, rubredoxin, glutaredoxin, and adenosylmethionine decarboxylase which participates in polyamine synthesis. Beside these main functional sub-categories, LTP3, SPIRAL proteins are known to act, respectively, in cuticle wax synthesis, microtubule, and cell wall stabilization.

No typical mechanism can be revealed for the “late 12 induction” group (patterns III, IV, and V) since it is composed of genes belonging to various sub-functional categories (protein degradation, detoxification, cell rescue, and defence) and exhibiting low induction rates.

Finally, the “progressive induction” group of strongly expressed genes is characterized by the high representation of genes also present in the previous group such as LEA or dehydrin, but up-regulated to a higher level, i.e. up to 11.77 and 45.8, respectively, for 144/34 ESTs. It is important to notice that the LEA proteins are predicted to be located not only in chloroplasts but also in mitochondria. The present FunCat classification provides no indication about the sub-cellular localization of the dehydrins, which are reported in literature to accumulate during cold acclimation in the nucleus and starch-rich amyloplasts for facilitating amylase activity (Kosova et al. 2007). The metallothionein MT2 (14 ESTs, 6.81-fold rate) is also well represented in the “progressive induction” group as well as numerous genes involved in detoxification. Another redundant sequence from this group corresponds to RCI2B (29 ESTs, 4.61-fold rate) which encodes a protein with trans-membrane domains associated with membrane protection. In addition, the strong induction of a specific hit of this group (40.25-fold rate for amino alcohol P-transferase) may reflect an increase in membrane fluidity throughout acclimation. The second typical gene of this group, ELIP (5 ESTs, 8.49-fold rate) is known to encode a protein protecting photosystem against photo-inhibition (Lang et al. 2005). Like the early induction V group, this progressive induction group contains LTP3 which participate in cuticle wax synthesis, as well as genes encoding enzymes involved in carbohydrate synthesis (galactinol synthase, rubisco). However, the carbohydrate metabolism associated with enzymes catalysing starch degradation (β-amylase) is specific of this “progressive induction” group. Finally, in common with the “late 12 induction” group, genes strongly and progressively induced are involved in proteasomal degradation to achieve turnover of cold-damaged proteins and also reset the cold response.

Altogether, these analyses of the cold transcriptome allow predicting the prominence in the E. gunnii acclimation process of cryoprotection and detoxification mechanisms, which rely, respectively, on LEA/dehydrins and metallothioneins. While protection against oxidative stress appears to mainly take place through transcriptional peaks shortly after each temperature drop (12 and 4 °C), accumulation of transcripts for cryoprotective proteins seems to be a progressive process leading to maximal hardening at the end of the chilling exposure. Other protective mechanisms like osmoprotection, turnover of aberrant proteins, cuticle synthesis, and stabilization of thylakoids or microtubules likely participate in a lesser extent to the cold tolerance of this tree.

Discussion

Based on predicted identity and redundancy level, a large 11,303 EST library constructed from a cold-tolerant and acclimated E. gunnii genotype was previously suggested as enriched in stress-related genes (Keller et al. 2009). The present paper focuses on a subset of sequences from this collection which were used for investigating the expression profile throughout cold acclimation. This sample is found to be representative of the whole library since the genes previously described as the most represented are confirmed here as the most cold-regulated. Very interestingly, this in-depth analysis benefited from the recent availability of the E. grandis whole genome sequence (Grattapaglia et al. 2012) to enlarge and strengthen gene annotation. This double investigation results in a rather complete and accurate picture of the extensive transcriptome remodelling during cold acclimation and allows suggesting the associated mechanisms likely to be involved in E. gunnii cold tolerance.

The cold response is intense in this tolerant Eucalyptus species as evidenced by the large number and variety of cold-induced genes and also the very strong induction or redundancy level exhibited by some ESTs. This response is also durable, occurring throughout the cold exposure. It involves genes participating in cold perception, signal transduction, and gene regulation, and also genes encoding effector proteins likely responsible for cell protection. Actually, the progressive accumulation of the most represented transcripts like LEA including dehydrins runs in parallel to the increase in freezing tolerance. The large extent of this global gene response associated with its durability might be a prominent factor of E. gunnii cold acclimation.

Predictive identity and function of the co-regulated genes lead to decipher likely predominant processes during the different phases of cold acclimation. The early response after each temperature drop is characterized by a co-regulation of genes participating in carbohydrate metabolism (galactinol synthase) and genes responsible for strengthening or recovering photosynthesis (rubisco activase or PSII complex). These biosynthetic pathways should lead to the accumulation of soluble sugars like raffinose, fructose, and sucrose. The important link between the integrity of the photosynthetic apparatus under cold and the level of freezing tolerance was demonstrated on perennial grasses (Sandve et al. 2011). The involvement of soluble sugars in E. gunnii cold acclimation was also previously suggested when sucrose, fructose, and above all, raffinose were described to accumulate during cold hardening, especially in tolerant genotypes (Leborgne et al. 1995a; Travert et al. 1997). This accumulation was proposed to play a role in the tolerance through membrane stabilization as suggested by feeding experiments with sugars (Leborgne et al. 1995b). Both galactinol and raffinose were described to play a role in ROS scavenging (Nishizawa et al. 2008). A correlation between galactinol synthase induction, raffinose accumulation, and genetic variability in freezing tolerance was already evidenced in alfafa and Arabidopsis (Cunningham et al. 2003; Nagele and Heyer 2013). According to the different expression patterns observed for the galactinol synthase gene, raffinose accumulation would start very early and would increase throughout the acclimation process. Starch degradation mainly leads to synthesis of maltose, a compatible-solute stabilizing factor (Kaplan and Guy 2004) which may accumulate in cold-exposed E. gunnii, since β-amylase gene is up-regulated in a progressive way. All these cryoprotective sugars would accumulate within the chloroplast where galactinol synthase and β-amylase are supposed to be located.

Another feature of the early response is the representation of genes involved in various structural protection mechanisms. Among them, the SPIRAL gene encodes a plant-specific small protein that localizes to microtubules and would act in maintaining their organization (Nakajima et al. 2006; Shoji et al. 2006) preventing them from further depolymerization (Furutani et al. 2000). The strong induction of SPIRAL gene in E. gunnii very shortly after each temperature drop could represent a protection of cytoskeleton after the cold-induced destabilization, one of the main suspected cold perception mechanisms. Moreover, deposition of waxy cuticle on the leaves as insulation from frost is also suggested by the up-regulation of the LTP3 gene (Guo et al. 2013; Qin et al. 2011). Wax synthesis in E. gunnii cold response is suggested by the representation of the LTP3 in both early and progressive patterns. Interestingly, Eucalyptus transgenic lines over-expressing a cold-responsive CBF transcription factor are shown to over-accumulate cuticle wax (Navarro et al. 2011). As a major feature making Eucalyptus a sclerophyllous plants, epicuticle synthesis may participate in one of its main strategies to cope with dehydrative stresses (drought and cold).

Metallothionein genes are strongly up-regulated throughout the E. gunnii cold acclimation, MT2 and MT3 exhibiting complementary induction patterns. In addition to their ability to bind metal ions, it has been argued that metallothioneins may function as efficient scavengers of ROS production when plants are exposed to abiotic stress and also play a role in plasma membrane repair (Kohler et al. 2004; Xue et al. 2009). Different types of MT proteins are proposed to differ not only in structure but also in function (Hassinen et al. 2011). Progressively induced in E. gunnii, MT2 was proposed to function in Arabidopsis cold tolerance by mediating the ROS balance in the cytosol (Zhu et al. 2009). Up-regulated early after temperature changes, MT3 may play a distinct role in coping with cold or may act in a different cell compartment. Detoxifying response may be complemented by a set of different genes with antioxidant activity spread out through all expression profiles, e.g. glutaredoxin, rubredoxin, catalase, thioredoxin, peroxidase, or glutathione transferase. Various genes encoding PR proteins like chitinases are co-regulated with metallothionein-encoding genes in the different phases of cold acclimation. In vitro antifreeze activity has been demonstrated for most of the PR-related proteins including chitinases, β 1-3-glucanase, and thaumatin-like proteins (e.g. osmotin) (Holliday et al. 2008; Jarzabek et al. 2009; Ruelland et al. 2009). It was also suggested that PR-proteins may be used as storage proteins when growth is limited by stress (Lee et al. 2008). Chitinases which have a broad range of cell functions might also participate in the ROS scavenging and detoxifying activity as suggested in coffee plants under cold conditions (Fortunato et al. 2010).

The most obvious characteristic of the E. gunnii cold transcriptome is the high representation of ESTs identified as LEA-like including dehydrins. Highly hydrophilic, dehydrins could act as emulsifiers or chaperones in the cells by preventing proteins and membranes from unfavourable structural changes (Garcia-Banuelos et al. 2009; Kosova et al. 2007; Ruelland et al. 2009). Most identified proteins are predicted to be located in the chloroplasts of Eucalyptus cells and could participate in thylakoid preservation. The involvement of dehydrin synthesis in natural variation of freezing tolerance was already reported for rhododendron at the protein level (Lim et al. 1999) and for Picea at the gene expression level (Holliday et al. 2008). With more occasional and moderate representation, other proteins participate in membrane stabilization and fluidity preservation (SFR2, RCI2B, lecithin cholesterol acyltransferase, and aminoalcohol phosphotransferase). SFR2 is identified as essential for freezing tolerance in Arabidopsis through lipid remodelling at the outer chloroplast membrane (Moellering et al. 2010). Membrane fluidity was previously shown to be critical for E. gunnii cell survival to freezing and may constitute one parameter of the genetic variation of cold tolerance in this species (Leborgne et al. 1992). Finally, as usually observed after stress exposure, targeted protein turnover through ubiquitination seems very active in the latest steps of the acclimation for removing damaged and misfolded proteins or degrading regulatory proteins.

According to the present description of genes up-regulated during cold acclimation, it is now possible to propose a predictive hierarchical classification of the main protective mechanisms likely to participate to the increase in cold tolerance. Cryoprotection of membranes and macromolecules through LEA/dehydrin action may explain most of the hardening in late acclimation. This preservation may be completed by sugars, mainly raffinose in the early acclimation and maltose in the late acclimation. Interestingly, a large number of proteins participating to cryoprotection and detoxification are predicted to accumulate within the chloroplast, suggesting that they would be dedicated to protect this essential organelle. Red-ox regulation and membrane or macromolecule protection against peroxidation may be the most permanent mechanism, in particular through ROS scavenging, mainly with metallothioneins and a range of different antioxidant proteins (Kotak et al. 2007). A more limited part of the E. gunnii cold response seems dedicated to avoiding dehydration through osmoprotectants (polyamines and amino acids) or to avoiding frost through synthesis of antifreeze proteins or wax deposition.

One of the main goals of this study was to investigate whether an evergreen and ever-growing woody plant without endodormancy, retaining its fragile organs (buds and leaves) during winter, was developing specific cold tolerance mechanisms. The prominence of genes supposed to help in maintaining growth through the preservation of chloroplasts, photosystem, and photosynthesis is in agreement with the main biological characteristics of Eucalyptus dedicated to permanent productivity. At the plant level, deposition of cuticle wax is probably the most specific mechanism dedicated to avoiding frost in Eucalyptus. At the cellular level, avoidance mechanisms supported by fewer transcriptomic changes seem less important than tolerance (in particular through cryoprotection) which is the main strategy for E. gunnii cold acclimation.

The other published cold transcriptome of the Eucalyptus genus concerns a cold-sensitive E. globulus, but it is difficult to compare since it is focused on cell wall synthesis (Rasmussen-Poblete et al. 2008). From the present study, only one gene involved in the phenylpropanoid pathway (Cinnamoyl CoA dehydrogenase (CAD)] was found to be cold-regulated.

It is commonly hypothesized that, despite a high level of conservation in cold tolerance mechanisms between annual herbaceous plants and perennial woody plants, the perennial habit has also resulted in additional specific mechanisms. The cold-regulated transcriptome identified in this study strongly differs from data on rhododendron (Wei et al. 2006) in particular concerning phospholipid biosynthesis and desaturation as well as aquaporin regulation. More similar is the blueberry cold acclimation transcriptome which contains genes encoding cryoprotective proteins and galactinol synthase. However, differently from E. gunnii, no detoxification gene is reported (Dhanaraj et al. 2004, 2007). The main frost tolerance mechanisms, such as detoxification, cryoprotection or protein turnover seem more conserved between E. gunnii and other tree species like Poncirus trifoliata (Sahin-Cevik and Moore 2006) or Malus domestica (Wisniewski et al. 2008). The cold transcriptome of Populus deltoides (Park et al. 2008) is the most similar to the one from E. gunnii, since genes like LEA/dehydrin, metallothionein, galactinol synthase, chitinase or β1-3 glucanase are similarly highly represented. Therefore, tree species seem to share major adaptation mechanisms at the cell and gene level, regardless of their winter habit (deciduous and dormant or evergreen). However, whereas it is difficult to distinguish winter dormancy from cold hardiness in deciduous trees (Garcia-Banuelos et al. 2009; Holliday et al. 2008), our data on a broad-leaved model without dormancy only relate to cold acclimation.

Drought stress, which shares the dehydration response with frost, is the other main factor limiting Eucalyptus plantations and species selection in Australia and around the world. Cross-talk between the two responses was evidenced when E. globulus plants more resistant to dry environment were shown to also exhibit higher frost tolerance than the drought-sensitive ones (Costa e Silva et al. 2009). This difference was explained by a higher accumulation of soluble sugars in the drought-resistant clone and a higher capacity for osmotic regulation as compared with the sensitive one. Actually, a major physiological characteristic widely associated with drought resistance in Eucalyptus is osmotic adjustment (Ladiges 1974) often attributed to accumulation of the cyclic polyol (quercitol) in xeric species or sucrose and proline synthesis in mesic species (Merchant et al. 2006). RNAseq transcript profiling revealed that protection of E. globulus against drought stress mainly involves carbohydrates (Villar et al. 2011). This species was also reported to respond to water deficit by activating the antioxidant protection system, for example through an increase in glutathione reductase activity (Shvaleva et al. 2006). Recently, the leaf transcriptome under water deficit of E. camaldulensis, a drought-tolerant species, revealed a lot of common features with the E. gunnii responsive genes (Thumma et al. 2012). Genes like chitinase, galactinol synthase or LEA/dehydrin were found to be induced, representing the main mechanisms of detoxification and protection of macromolecules and membranes identified in our hands. In this case, the main noticeable difference with our data is the lack of metallothionein genes. Either this detoxification mechanism is specific to cold response or the late kinetics of expressional analysis (2 months of water stress) was not compatible with detection of MT transcripts. Therefore, adaptive response of Eucalyptus to drought or cold exhibit many common features like osmolyte accumulation or increase in antioxidant activity and a lot of candidate genes might be the same. Again, the main distinction may correspond to the relative importance of each adaptive strategy. In the case of drought tolerance, literature highlights the role of sugars and amino acids for osmoprotection when the present study evidences the prominence of cryoprotective proteins. This overall similarity between Eucalyptus genes involved in cold and dehydration responses is not so surprising, given the adaptive evolution of this tree which spread out in Australia during cool and dry period (Martin 2006).

Beyond the identification of the main mechanisms likely to participate at different extents to E. gunnii cold acclimation, this study leads to the proposal of some genes as potential candidates for improving frost tolerance and potentially drought tolerance. Taking into account the correlation between transcript accumulation and hardiness, as well as gene representation in this transcriptome, LEA/dehydrin genes are the most obvious candidates. Many studies have already shown that dehydration stress can be improved by LEA gene transfer as recently reviewed by Zhao and co-workers (Zhao et al. 2011). Moreover, in E. globulus, a freezing-resistant genotype was reported to exhibit a higher dehydrin transcript accumulation under frost compared with the sensitive genotype (Fernandez et al. 2012). Metallothionein with high induction rate and high redundancy also seems to be a good candidate, especially the early-induced MT3 gene. Over-expression of a cotton MT gene in tobacco allowed increasing abiotic stress tolerance (Xue et al. 2009). Finally, genes encoding proteins like galactinol synthases or PR-proteins, also functionally validated through transformation in other plant species (Das et al. 2011; Taji et al. 2002), could be considered as promising biotechnological tools.

In conclusion, this genomic study allows dissecting the cold response in a freezing tolerant E. gunnii clone by analysing the coordination of gene regulation during cold acclimation. The combination of functional annotation of the cold-responsive genes with the analysis of the co-regulation profiles led to draw a comprehensive picture of the different protective mechanisms likely to take place throughout the acclimation. While previous studies on this species focused on a single aspect of the response (sugars…), this global analysis allows comparing and making a hypothesis on the hierarchy between the strategies to be further checked at the metabolic level. Above all, it highlights the importance of cryoprotection and detoxification in hardening and indicates that a species especially lacking physiological adaptation against frost may compensate by a drastic reconfiguration of its transcriptome towards protecting the cells. Finally, these results provide new prospects and more precisely a range of new candidate genes for breeding.