Nonrandom domain organization of the Arabidopsis genome at the nuclear periphery

Abstract

The nuclear space is not a homogeneous biochemical environment. Many studies have demonstrated that the transcriptional activity of a gene is linked to its positioning within the nuclear space. Following the discovery of lamin-associated domains (LADs), which are transcriptionally repressed chromatin regions, the nonrandom positioning of chromatin at the nuclear periphery and its biological relevance have been studied extensively in animals. However, it remains unknown whether comparable chromatin organizations exist in plants. Here, using a strategy using restriction enzyme-mediated chromatin immunoprecipitation, we present genome-wide identification of nonrandom domain organization of chromatin at the peripheral zone of Arabidopsis thaliana nuclei. We show that in various tissues, 10%-20% of the regions on the chromosome arms are anchored at the nuclear periphery, and these regions largely overlap between different tissues. Unlike LADs in animals, the identified domains in plants are not gene-poor or A/T-rich. These domains are enriched with silenced protein-coding genes, transposable element genes, and heterochromatic marks, which collectively define a repressed environment. In addition, these domains strongly correlate with our genome-wide chromatin interaction data set (Hi-C) by largely explaining the patterns of chromatin compartments, revealed on Hi-C maps. Moreover, our results reveal a spatial compartment of different DNA methylation pathways that regulate silencing of transposable elements, where the CHH methylation of transposable elements located at the nuclear periphery and in the interior are preferentially mediated by CMT2 and DRM methyltransferases, respectively. Taken together, the results demonstrate functional partitioning of the Arabidopsis genome in the nuclear space.

Figure 1.

Identification of chromatin located at the nuclear periphery by RE-ChIP. (A) Localization of the NUP1:GFP protein in an Arabidopsis nucleus: (scale bar) 2 µm. (B) Procedures for RE-mediated ChIP with NUP1:GFP (green). Chromatin (purple lines) fragmentation and isolation are conducted with a combination of RE (restriction enzyme) digestion and mild sonication. (C) Normalized sequence coverage (50-kb window size) on Chromosome 5 from various ChIP experiments. The horizontal bars depict pericentromeric regions, within which centromeric regions are highlighted in red. (D) NUP1:GFP RE-mediated ChIP-seq signal (50-kb window size), represented as the log₂ value of the ratio between normalized anti-GFP and IgG coverage, over all five chromosomes. Horizontal bars indicate the centromeric/pericentromeric regions, as in C.

Figure 2.

Correlation between chromatin anchored at the nuclear periphery and the Hi-C map. (A) Correlation between NUP1:GFP RE-ChIP-seq signal and Hi-C map. The Hi-C maps (normalized at 20-kb resolution) of the left and right Chromosome 1 arms are shown as Spearman correlation matrices, from which PCA was conducted; the eigenvalues of the first component are plotted below (red and blue bars) together with the NUP1:GFP signal (green lines, 20-kb window size), represented as the log₂ value of the ratio between normalized anti-GFP and IgG coverage. (B) Anti-correlation between the telomeres and NUP1:GFP RE-ChIP-seq signal. The left panel shows a Spearman correlation matrix of Chromosome 3 derived from a Hi-C map at 20-kb resolution. Arrows depict KEE regions. The right panels highlight the 6-Mb distal chromosome regions, in which their correlation with the chromosome terminus (the first 20 kb of Chromosome 3) in the Hi-C map are shown as black curves. Green curves show the NUP1:GFP signal, as in A. Due to physical linkage, chromosome termini are expected to have strong colocalization with telomeres in the nucleus. In a Hi-C experiment, chromosome termini can be used to infer the spatial interactions between telomeres and other genomic regions.

Figure 3.

Genome-wide identification of NUP1-enriched regions in various tissues. (A) Signals of NUP1:GFP RE-ChIP-seq (20-kb window size), represented as the log₂ value of the ratio between normalized anti-GFP and IgG sequence coverage over Chromosome 1. For each tissue, the solid and dotted lines depict two replicates. (B) Distribution of NUP1-enriched domains across the genome viewed with the Integrative Genomics Viewer browser (Robinson et al. 2011). (C) Percentage of NUP1-enriched genomic regions: (inf) inflorescence. (D) Venn diagram of genes enriched in four tissues.

Figure 4.

Epigenetic, genomic, and structural features of chromatin tethered at the nuclear periphery. (A) A representative genomic region from Chromosome 1 showing the distributions of NUP1-enriched chromatin identified from 7-d-old leaf tissues (shaded in green) and various epigenetic marks. Average enrichment means the percentage of regions (calculated from 100-bp windows) enriched for the respective epigenetic mark. (B,C) Epigenetic marks (B) and GC content (C) around NUP1-enriched domain borders, shown as a vertical line separating the white and gray blocks. For each plot, the area on the right indicates NUP1-enriched domains (although not all are larger than 10 kb). Average enrichment in B is defined as in A. The GC content in C is in a window size of 100 bp, with a step size of 20 bp. Because enrichment of gene bodies is found inward from NUP1-enriched domain boundaries (see Supplemental Fig. 12), for the background, we randomly picked 3000 genes with the same expression distribution profile as that of NUP1-enriched genes. For these control genes, we extracted the 20-kb regions flanking either their transcription start sites or their transcription termination sites, which were selected randomly. (D) Different types of chromatin loops associated with NUP1-enriched domains (including those in pericentromeric regions). Chromatin loops are from Liu et al. (2016). For both “intra” and “across,” the number of observed chromatin loops are significantly different (P < 2.2 × 10⁻¹⁶) relative to the permutation-based null distribution of the background, which was simulated by shifting the coordinates of NUP1-enriched domains ±50 kb or ±100 kb.

Figure 5.

Enrichment of silenced genes at the nuclear periphery. (A) Number of TE genes (left) and protein-coding genes (right) enriched in different tissues. For each column, the observed number of genes is significantly different (P < 0.001) relative to the permutation-based null distribution of the background (generated as described in Fig. 4C): (inf) inflorescence. (B) Comparison of gene expression levels, which are from a normalized tilling array data set (Laubinger et al. 2008). The P-values indicate Mann-Whitney U test results.

Figure 6.

Comparison of DNA methylation over TEs. Patterns of TE DNA methylation (CpG, CHG, CHH) in wild-type (WT) and mutants. The grouping of TEs is according to the enrichment results of NUP1:GFP RE-ChIP-seq from 30-d-old leaf tissues. The methylation ratio is calculated in 100-bp windows. The signal over each TE is linearly transformed so that the boundaries of all TEs are aligned.