The grapevine expression atlas reveals a deep transcriptome shift driving the entire plant into a maturation program

Abstract

We developed a genome-wide transcriptomic atlas of grapevine (Vitis vinifera) based on 54 samples representing green and woody tissues and organs at different developmental stages as well as specialized tissues such as pollen and senescent leaves. Together, these samples expressed ∼91% of the predicted grapevine genes. Pollen and senescent leaves had unique transcriptomes reflecting their specialized functions and physiological status. However, microarray and RNA-seq analysis grouped all the other samples into two major classes based on maturity rather than organ identity, namely, the vegetative/green and mature/woody categories. This division represents a fundamental transcriptomic reprogramming during the maturation process and was highlighted by three statistical approaches identifying the transcriptional relationships among samples (correlation analysis), putative biomarkers (O2PLS-DA approach), and sets of strongly and consistently expressed genes that define groups (topics) of similar samples (biclustering analysis). Gene coexpression analysis indicated that the mature/woody developmental program results from the reiterative coactivation of pathways that are largely inactive in vegetative/green tissues, often involving the coregulation of clusters of neighboring genes and global regulation based on codon preference. This global transcriptomic reprogramming during maturation has not been observed in herbaceous annual species and may be a defining characteristic of perennial woody plants.

Figure 1.

Overview of the V. vinifera cv Corvina Samples Used for Microarray Analysis. The photographs and diagrams show the shoot/cane organs (A) and berry cluster (B) from clone 48. The exact developmental stages are indicated by the modified E-L classification keys on each picture. Rachis, seed, berry flesh, and skin samples were taken at the stages indicated in (B). Schematic illustrations were modified from Jackson (2000).

Figure 2.

Global Gene Expression Patterns in the Different Samples. (A) Number of genes expressed in each of the 54 samples. Total: number of gene expressed in at least one organ (27,453; ∼93% of all genes on the array). Common: genes expressed in all 54 organs (2948; ∼10% of all genes on the array). (B) Number of organ-specific genes. Only samples with nonredundant organ identity were analyzed (see Supplemental Table 2 online). (C) Shared and specific expression profiles of genes expressed in multiple floral organs. (D) Shared and specific expression profiles of genes expressed at multiple bud developmental stages.

Figure 3.

Tissue Transcriptome Relationships. (A) Correlation matrix of the whole data set. The analysis was performed by comparing the values of the whole transcriptome (29,549 genes) in all 54 samples, using the average expression value of three biological replicates and Pearson’s distance as the metric. Correlation analysis was performed using R software. (B) Cluster dendrogram of the whole data set. The Pearson’s correlation values were converted into distance coefficients to define the height of the dendrogram. (C) Correlation matrix for the RNA-seq data set. Reads generated in previous experiments (Denoeud et al., 2008; Zenoni et al., 2010) were remapped on the 12x grapevine genome, V1 gene prediction. (D) Cluster dendrogram for the RNA-seq data set. Reads generated in previous experiments (Denoeud et al., 2008; Zenoni et al., 2010) were remapped on the 12x grapevine genome, V1 gene prediction. (E) HCL analysis on the whole 54-sample data set. Pearson’s correlation distance was used as the metric, and TMeV 4.3 software was used to create the transcriptional profiles dendrogram.

Figure 4.

Global Gene Expression Trends in Grapevine. (A) Variables and scores scatterplot of the O2PLS-DA model (3 + 5 + 0, UV, R²Y = 0.967, Q² = 0.868) applied to 52 samples, confirming the separation into four classes sharing similar expression signatures. Components 3 and 5 represent the predictive and orthogonal components identified by the model, whereas 0 represents the background variation (UV = unit variance scaling method). (B) and (C) The expression profiles of positive (B) and negative (C) putative molecular biomarkers were selected using an S-plot (Wiklund et al., 2008) within the first (positive) and the last (negative) percentile.

Figure 5.

Biclustering Analysis with the PLSA Algorithm. (A) Samples were divided into eight topics defined by high-level gene expression. (B) Functional category distribution of topic-specific transcripts. The V1 version of the 12x draft annotation of the grapevine genome allows the identification of ∼70% of genes. This was manually verified and transcripts were grouped into the 18 most represented functional categories, based on Plant GO Slim biological processes classification.

Figure 6.

Coexpression Distribution among Green/Vegetative Samples and Ripe/Woody Samples. Pairwise gene correlation analysis was computed by calculating the Pearson’s correlation for each gene pair in both specific subsets of organs. Curve distributions are represented by the areas under the curves normalized to 1. Green curve, green/vegetative samples; red curve, ripe/woody samples.

Figure 7.

Sliding-Window Analysis of Coexpression along Grapevine Chromosomes 2 and 16. Red and green lines correspond to positions on the chromosome where coexpression is specific for nonvegetative samples (positive variation) and vegetative samples (negative variation), respectively (see Supplemental Methods 1 online for further details on sliding-window analysis).

Figure 8.

Mutual Information of Synonymous Codon Usage in Grapevine Gene Coexpression Clusters. Each row represents a coexpression cluster, whereas each column represents a synonymous codon. Significant mutual information is shown in blue (P ≤ 10⁻⁴).