Background

Since the construction of the first plant transcriptome map [1], gene expression atlases were published for many plants belonging to the variety of families and became a widely used tool in plant studies. By definition, transcriptome map, or gene expression atlas, is a collection of expression profiles of all genes in different organs, tissues or cells under various environmental conditions [1]. For the current moment, such collections are covering plant taxa from moss [2] and pine [3] to many species of Rosids and Asterids, including model [4, 5], agricultural [35]. Hypocotyl 27 showed photosynthetic enrichment in upregulated genes. Upregulated genes in Young Flower 27 were enriched for catabolic processes and pollen-related terms (Additional file 1: Tables S9 and S10).

The remarkable diversity of the processes leading to cold acclimation in various organs provides evidence for the adjustment of the general response to stress by organ-specific responses. This fact may limit the application of knowledge on the cold response in one organ (e.g. leaf) to another organ (e.g. seed).

Stress response in non-leaf organs does not involve many known regulators and stress-response genes

We found a surprisingly low fraction of DE genes annotated as stress responsive (GO category “GO:0006950~response to stress” and downstream categories). The fraction of genes from this category varied between 8 and 12% for both 3 h and 27 h samples; the absolute number of stress-annotated genes was the lowest in the Young Flower 3 (106 genes) and highest in Leaf 27 (628 genes). These results show that GO annotation of the Arabidopsis genome strongly underestimates the number of stress-responsive genes. This is especially pronounced in organs that are not usually the focus of stress response studies, such as flowers and seeds (Fig. 2a). GO annotation has several shortcomings and is known to be incomplete; in particular, approximately 50% of A. thaliana genes do not have biological process annotations. Additionally, many GO annotations are based only on computational predictions and are not supported by experimental data [36].

Fig. 2
figure 2

(a) Number of DE genes for each sample that are annotated or not annotated as stress-responses via Gene Ontology. (b) Number of COR genes that are DE in our data

Full size image

Searching a list of genes for which participation in the cold stress response is defined by experimental data (1322 COR genes) [21], we found that 64% of these genes were DE in at least one organ after 3 h of cold treatment (26% belong to Down, 34% to Up, and 4% to Mix category) (Additional file 1: Table S11). Considering the genes by organ, the percentage of DE genes varied from 22% in Seeds 3 to 35% in Cotyledons 3 (Fig. 2b). After 27 h, the picture was even more pronounced. Although 99% of COR genes were DE in at least one organ, the distribution of genes by organ varied greatly; 79–85% of COR genes were DE in Cotyledons 27, Leaf 27, and Hypocotyl 27, although the percentages in Flower 27, Young Flower 27, and Seeds 27 were 36, 51 and 41%, respectively (Fig. 2b, Additional file 1: Table S12). These results show that the cold stress response in non-photosynthetic organs not only involves additional genes that were not previously associated with stress but also does not recruit many known regulators.

The CBF1, CBF2 and CBF3 genes are known to be activated within a few minutes after exposure to cold [15, 18]. Despite their crucial importance in cold acclimation, these genes are not unique in the early response to stress. Using the time course experiments performed by Kilian et al. [24], Park et al. [21], identified 27 transcriptional factors as having the same behavior as CBF-encoded genes (named “first-wave” genes). In our data, the first expression measurement time was 3 h after the beginning of the low temperature conditions, which is not the earliest possible time point, although we were still able to identify all except one of the first-wave genes as DE in at least one organ. Four genes were DE in all organs after 3 h of cold treatment (including CBF3), while others had noticeable differences in their expression patterns (Additional file 1: Table S13). For 5 genes, we observed opposite changes in expression in various organs, with high differences in fold changes (e.g., the fold change for ZAT12 in Hypocotyl 3 and Leaf 3 were 0.22 and 2.96, respectively). After 27 h, the diversity of the expression changes became more notable; the first-wave genes from most organs showed downregulated expression or their expression had returned to the control value. However, in Leaf 27, there were 19 genes that were still upregulated (Additional file 1: Table S14). This variety in responses to cold from early activated transcriptional factors confirms the inadequacy of simple transfer or the results obtained on one organ to another.

Expression characteristics of stress-response genes: Shannon entropy

We assessed several parameters that are associated with organ-specific stress-response genes based on the RNA-seq transcriptome map for A. thaliana [5]. We first estimated the expression pattern width using Shannon entropy H [37, 38]. Genes with high H value are expressed ubiquitously, while those with a low H value have a narrow expression pattern. The distribution of entropy for all of the expressed genes was noticeably skewed to the right, indicating a high number of genes with wide expression patterns [5]. The second small peak appears at very low H values and corresponds to genes that are highly expressed in an organ-, tissue- or stage-specific manner. The distribution for the 15,459 genes that have expression changes in response to stress resembled the overall distribution, while genes that are common in at least 5 organs genes lacked low H peaks (Fig. 3 a and b). The entropy of unique genes for most of the samples was distributed similarly to common DE genes. However, some samples had distinct features in the H distribution. Specifically, for genes that were upregulated in Leaf 3, there was a peak at low (0–0.3) entropy values (Fig. 3c). We analyzed the expression patterns of these genes in the transcriptome map. Surprisingly, under non-stress conditions, the expression of all of these genes (with one exception) as restricted to mature anthers and whole flowers containing anthers at the same stage (Additional file 1: Table S15). These genes were differentially expressed in Leaf after 3 h of cold treatment, and the half of them (44%) were also DEs in Leaf 27. Only 33% of the genes were DE in Flower 27, and none of them had shifted expression in Flower 3. For most of these genes, their function is related to controlling cell wall conditions and pollen tube growth (Additional file 1: Table S15). Among them, the most pronounced changes were in genes encoding pectin methylesterases. Pectin is a crucial component in the cell wall, as the matrix in which other polysaccharides (cellulose and hemicellulose) are embedded. Pectins are produced in the Golgi in a highly methylesterified form and are then modified by pectin methylesterases, which catalyze deesterification [39]. The ratio of esterified to de-esterified pectins determines many cell wall properties, such as rigidity, permeability and cohesion. This increase in pectin methylesterase activity under cold stress has been found in other plants [40, 41]. It is regarded as a cold acclimation mechanism because the increase in cell wall rigidity offers a higher resistance to dehydration and inhibits organ growth. Our results indicate that implementation of this mechanism in A. thaliana occurs by the recruitment of pollen-specific genes.

Fig. 3
figure 3

Shannon entropy distribution. (a) Shannon entropy for the 15,459 stress-response genes in the transcriptome map. (b) Shannon entropy for genes that are DE in at least 5 samples. (c) Shannon entropy for unique sample genes

Full size image

Similar to Leaf 3, the H distribution for DE genes uniquely upregulated in Flower 27 has a peak at low values (Fig. 3c). All of these genes are also anther-specific. GO enrichment analysis revealed overrepresentation of terms associated with pollen tube growth and cell wall modification (Additional file 1: Table S16). These genes revealed a complex picture. In particular, we observed concerted upregulation of pectin methylesterase inhibitors (PMEI5 (AT2G31430) and others), while the expression of pectin methylesterases (PPME1, VGD1, VGDH2 (AT1G69940, AT2G47040, and AT3G62170) was also increased. Pectin methylesterase PPME1 has been shown to linearly demethylesterify pectin chains in pollen tube walls. A reduction in PPME1 activity in ppme1 mutant leads to decreased cell wall rigidity [42]. VGD1 is another gene that encodes a pectin methylesterase. VGD1 also has a linear demethylesterification activity and modifies pollen wall pectin [Genes unique to organs: Overrepresentation of regulatory elements from transcription factors outside the ERF/AP2 family

For a deeper understanding of the gene networks involved in cold acclimation in different organs, we analyzed the overrepresentation of regulatory elements from transcription factors from different gene lists (e.g., common in all or at least five organs, DE genes unique for certain samples, or genes with a certain Shannon entropy H; for a full list of the tested gene groups see Additional file 1: Table S19).

As expected, the promoter regions of DE genes upregulated in all or at least 5 organs were enriched with CBF regulatory elements. Genes which have these elements and are thus likely to be under regulation of CBF1–3 transcription factors displayed stress GO enrichment terms (Additional file 1: Table S20). As expected, promoter regions of DE genes upregulated in all or at least 5 organs were enriched with CBF regulatory elements. Genes that cause this overrepresentation and can be under regulation of CBF1–3 transcription factors have stress GO enrichment. Among these genes 12 transcription factors that also are characterized by overrepresentation in DE genes common for at least 5 organs and possibly regulated by them genes are enriched with stress GO terms too. Four of these transcription factors belong to ERF/AP2 family and were described as participants in both biotic and abiotic stress response [61].

Regarding genes that are unique for each sample, we did not find any overrepresentation of CBF regulatory elements, which shows that factors other than CBF govern organ-specific stress responses. In particular, we found an overrepresentation of regulatory elements for 9 NAC transcription factors in promoters of upregulated DE genes in Young Flower 27 (Additional file 1: Table S21).

Database

To make these data available to the plant science community, we included them in our database TraVA (https://travadb.org). The interface and options are similar to the datasets from Klepikova et al. [5] and Kasianov et al. [10]. The profiles for each gene in the cotyledons, hypocotyl, leaves, young flowers, mature flowers and seeds are represented under both the control and cold treatment conditions as the number of reads and as fold change in the expression level relative to the control (Fig. 5).

Fig. 5
figure 5

Database view. (a) Read counts and fold changes for CBF3 (DREB1A) among all samples. (b) Read counts and fold changes for AT3G13229 among all samples

Full size image

Conclusions

We analyzed gene expression in six Arabidopsis organs and tissues after 3 and 27 h of cold treatment using RNA-seq. We found that 15,459 genes were differentially expressed in at least one sample. Well-studied organs (leaf, cotyledons and the hypocotyl) showed similar results to other studies, while seeds, flowers and young flowers displayed pronounced differences. Only a small number of genes were common in all samples. These core genes were strongly enriched in stress-related GO categories. Unique sample genes were related to different processes in each organ. Some of these genes displayed expression specificities, such as peaks in Shannon entropy or DE Score distributions. Thus, while the mechanisms of cold stress response are common in all plants, in every organ they are modified in a unique fashion, including the recruitment of genes that are expressed in other organs in non-stress conditions.

Methods

Plant growth, cold treatment and sample collection

Col-0 A. thaliana (accession CS70000) plants were grown as described in Klepikova et al. [62], with the exception of vernalization. The collected samples are described in Additional file 1: Table S1. For each sample, two biological replicates with 15 individual plants were obtained and fixed in RNAlater (Qiagen, USA).

Control samples were harvested from ZT 8 to 9 and on the next day at ZT 5 temperature in a climate chamber set at + 4 °C. Samples treated with cold for 3 h were collected at ZT 8 and for 27 h at ZT 8 the next day.

RNA extraction and sequencing

Total RNA was extracted with a RNeasy Plant Kit (Qiagen, USA) following the manufacturer’s protocol. cDNA libraries for sequencing were constructed with the TruSeq RNA Sample Prep Kits v2 (Illumina) following the manufacturer’s protocol. An Illumina HiSeq2000 was used for sequencing with a 50 bp read length and a sequence depth of 20 million uniquely mapped reads.

Trimming and map** of reads and expression level determination

For read trimming, the CLC Genomics Workbench 6.5.1 was used with the following parameters: “quality scores - 0.005; trim ambiguous nucleotides – 2; remove 5’-terminal nucleotides – 1; remove 3’-terminal nucleotides – 1; and discard reads below a length of 25”. The trimmed reads were mapped using the CLC Genomics Workbench to the reference A. thaliana genome (TAIR10 genome release) with unique map** only (length fraction = 1 and similarity fraction = 0.95). For each gene, total gene reads (TGR) was determined as the sum of all the reads mapped on this gene. Sequencing and map** statistics are shown at Additional file 1: Table S22. Total gene reads and RPKM are provided for all samples at Additional file 1: Table S23 and S24, respectively.

Identification of differentially expressed genes

Differentially expressed (DE) genes were identified using the R package “DESeq2” [63]. A false discovery rate (FDR) of 0.05 and fold change of 2 were chosen as the initial threshold for significant differential expression.

Gene ontology enrichment analysis

Downregulated and upregulated DE gene lists were analyzed by Gene Ontology (GO) and other annotation (as key words or as a protein domain) enrichments using the DAVID gene functional annotation tool with an FDR value of 0.05 and fold change category representation of 2 as the threshold of significance [64, 65].

Hierarchical clustering

A hierarchical tree was obtained with the “hclust” function from the R package “stats” [66].

Identification of key transcription factors

To identify the transcription factors involved in the regulation of observed differential gene expression we used annotations of the transcription factor targets based on ampDAP-seq [67]. For each set of DE genes, we considered data for all transcription factors. We estimated the relative enrichment of targets among differentially expressed genes as the log2 of the %target (DE) / %target (non-DE). The statistical significance was assessed using the right-tailed Fisher’s exact test with 2 × 2 contingency tables (targets vs. non-targets and DE vs. non-DE) with FDR correction for multiple tested transcription factors (219 TFs).

Accession numbers

The Illumina sequence reads have been deposited into the NCBI Sequence Read Archive with project ID PRJNA411947.