The rules of gene expression in plants: organ identity and gene body methylation are key factors for regulation of gene expression in Arabidopsis thaliana

Abstract

Background: Microarray technology is a widely used approach for monitoring genome-wide gene expression. For Arabidopsis, there are over 1,800 microarray hybridizations representing many different experimental conditions on Affymetrix ATH1 gene chips alone. This huge amount of data offers a unique opportunity to infer the principles that govern the regulation of gene expression in plants.

Results: We used bioinformatics methods to analyze publicly available data obtained using the ATH1 chip from Affymetrix. A total of 1887 ATH1 hybridizations were normalized and filtered to eliminate low-quality hybridizations. We classified and compared control and treatment hybridizations and determined differential gene expression. The largest differences in gene expression were observed when comparing samples obtained from different organs. On average, ten-fold more genes were differentially expressed between organs as compared to any other experimental variable. We defined "gene responsiveness" as the number of comparisons in which a gene changed its expression significantly. We defined genes with the highest and lowest responsiveness levels as hypervariable and housekeeping genes, respectively. Remarkably, housekeeping genes were best distinguished from hypervariable genes by differences in methylation status in their transcribed regions. Moreover, methylation in the transcribed region was inversely correlated (R2 = 0.8) with gene responsiveness on a genome-wide scale. We provide an example of this negative relationship using genes encoding TCA cycle enzymes, by contrasting their regulatory responsiveness to nitrate and methylation status in their transcribed regions.

Conclusion: Our results indicate that the Arabidopsis transcriptome is largely established during development and is comparatively stable when faced with external perturbations. We suggest a novel functional role for DNA methylation in the transcribed region as a key determinant capable of restraining the capacity of a gene to respond to internal/external cues. Our findings suggest a prominent role for epigenetic mechanisms in the regulation of gene expression in plants.

Figure 1

Classification of experiments from the NASCarrays database. Pie charts with the classification of microarray experiments according to the experimental factor categories defined by TAIR (A) or the organ used to extract RNA to perform the microarray experiments (B).

Figure 2

Global characteristics of the Arabidopsis transcriptome. A) Histogram of the number of genes (X-axis) regulated in a given number of comparisons (Y-axis). B) Average number of genes regulated by each experimental category as defined in Figure 1A. C) Histogram of the number of comparisons (X-axis) for which the specified number of genes (Y-axis) show significant regulation.

Figure 3

Correlation between methylation and gene responsiveness. (A) Plot of the frequency of methylated genes (according to Zhang et al. [24]; X-axis) within a group of genes against the number of comparisons in which that group of genes is regulated (Y-axis). The dotted line represents the regression line. B) Same as (A) except using data from Zilberman et al [25]. C) to E). Same as (A) except with the different experimental categories defined in Figure 1A, using methylome data from Zhang et al [24]. G) Same as (A) except the X-axis represents the frequency of genes that are the target of trimethylation on H3K27 [30].

Figure 4

Lack of linear correlation between expression levels and gene body methylation or TATA-box presence. (A) Plot of the median expression level across the whole NASC arrays dataset in 10% bins (X-axis) versus the frequency of methylated genes in the bin (Y-axis), as determined by Zhang et al. [24]. (B) Same as (A), except using data from Zilberman et al. [25]. C) Same as (A), except the Y-axis represents the frequency of TATA-containing genes according to the MotifSearch definition [26]. D) Same as (C), but using the PlantProm definition [27].