In vivo mapping of arabidopsis scaffold/matrix attachment regions reveals link to nucleosome-disfavoring poly(dA:dT) tracts

Abstract

Scaffold or matrix attachment regions (S/MARs) are found in all eukaryotes. The pattern of distribution and genomic context of S/MARs is thought to be important for processes such as chromatin organization and modulation of gene expression. Despite the importance of such processes, much is unknown about the large-scale distribution and sequence content of S/MARs in vivo. Here, we report the use of tiling microarrays to map 1358 S/MARs on Arabidopsis thaliana chromosome 4 (chr4). S/MARs occur throughout chr4, spaced much more closely than in the large plant and animal genomes that have been studied to date. Arabidopsis S/MARs can be divided into five clusters based on their association with other genomic features, suggesting a diversity of functions. While some Arabidopsis S/MARs may define structural domains, most occur near the transcription start sites of genes. Genes associated with these S/MARs have an increased probability of expression, which is particularly pronounced in the case of transcription factor genes. Analysis of sequence motifs and 6-mer enrichment patterns show that S/MARs are preferentially enriched in poly(dA:dT) tracts, sequences that resist nucleosome formation, and the majority of S/MARs contain at least one nucleosome-depleted region. This global view of S/MARs provides a framework to begin evaluating genome-scale models for S/MAR function.

Figure 1.

Mapping of S/MARs in Select Regions of Arabidopsis Chr4. (A) Gene and TE coverage for chr4. Chr4 was divided into 1-kb nonoverlapping bins, and the gene and TE coverage was calculated for each bin. This data was loess-smoothed in a 100-kb window for the plot. (B) Chr4 can be subdivided into six regions based on the gene and TE coverage and the time of DNA replication (Lee et al., 2010). The position of the centromere is indicated by the yellow oval. The early-replicating distal long and short arms are the most euchromatic and are shaded light gray. The late-replicating proximal long and short arms have substantial heterochromatin and are shaded dark gray. The late-replicating constitutive heterochromatin of the knob and pericentromere are shaded black. S/MARs were not mapped in these last two regions. (C) Mapping of S/MARs in a representative late-replicating region with significant TE content. Shown are the GC (%) and the log₂ ratios for the S/MAR to genomic DNA microarrays with positive values highlighted in blue. Locations of computationally predicted S/MARs (pS/MARs) (Rudd et al., 2004) are shown for comparison. Genes and the direction of transcription are shown as green arrows, and TEs from TAIR10 Arabidopsis genome annotation are shown as brown boxes. (D) Mapping of S/MARs in a representative early-replicating region. The mapped features are as described in (C).

Figure 2.

Genomic Characteristics and Clustering of S/MARs. (A) K-means clustering of S/MARs based on genomic context. The gene, exon, and TE coverage of each S/MAR was calculated and k-means clustering (k = 5) was performed. The results for each S/MAR within these five clusters are shown as a heat map, using 10% coverage increments. Unannotated and intron coverage is shown but was not used for clustering. (B) Histograms of S/MAR AT content (%) for each cluster, using bins of 2%. (C) Histograms of S/MAR position relative to the TSS of the closest gene. S/MARs were first grouped as either TSS- (or TTS)-proximal based on their absolute distance to the TSS or TTS of the nearest (or any overlapping) gene. Distances were calculated and corrected for gene strand so that S/MARs upstream of the TSS have negative values. Histograms were calculated using 250-bp bins, and the plot is restricted to the range ±4 kb (18 outliers are not shown). Also shown is the total count for each S/MAR cluster. (D) Histograms of S/MAR position relative to the TTS of the closest gene. Distances were calculated as in (C). Two outliers are not shown. Also shown is the total count for each S/MAR cluster.

Figure 3.

Chromatin Context of S/MARs by Cluster. The colors shown in the key indicate the respective S/MAR clusters. (A) Predicted average nucleosome occupancy scores based on the Arabidopsis DNA sequence (Kaplan et al., 2009) were used to estimate the probability of nucleosome occupancy of S/MARs and flanking regions. S/MARs were aligned at their midpoint and a window of ±5 kb was delineated and divided into 500-bp bins. Upstream and downstream were defined relative to the gene closest to each S/MAR. For each S/MAR, the mean predicted nucleosome occupancy in each bin was determined and then averaged for each S/MAR cluster. (B) Similar to (A) but using experimental sequence coverage data for Arabidopsis shoot mononucleosomes (Chodavarapu et al., 2010). The units are the mean coverage for the sequence reads. (C) to (F) Epigenetic modifications of S/MARs as shown described in (A) but using microarray results for chromatin immunoprecipitations to H3K4me2/1 (C), H3K9me2 (D), H3K56ac (E), and DNA 5mC (F) in our cell line (Tanurdzic et al., 2008). The microarray data were used to define segments of chr4 marked by each epigenetic modification. We then determined the coverage of these segments within each bin and expressed this as a percentage.

Figure 4.

Correlation of S/MARs with Gene Activity. The colors shown in the key indicate the respective S/MAR clusters. (A) All genes that either overlap or are adjacent to an S/MAR were identified and categorized as having either TSS- or TTS-proximal S/MARs based on the location of the closest S/MAR midpoint. Gene activity was determined from MAS5 presence/absence calls from Affymetrix expression experiments for our cell line (Tanurdzic et al., 2008) and is the percentage of genes with detectable mRNA in each category. Totals for the TSS-proximal S/MARs are 239, 145, 243, 137, and 99 for S/MAR clusters A, B, C, D, and E, respectively, and 124, 66, 72, 74, and 21 for S/MAR clusters A, B, C, D, and E, respectively, for the TTS-proximal S/MARs. The dashed line shows the mean for all chr4 genes, and the error bars indicate the 95% confidence interval from the binomial test, and P value significance cutoffs are indicated as follows: *P ≤ 0.05, **P ≤ 0.001, and ***P ≤ 0.0001. (B) Gene activity for TF genes and non-TF genes in the functional cluster. We identified all transcription associated genes on chr4 based on the enriched GO terms in the identified functional cluster and grouped them as TFs and non-TFs based on the AtTFDB (Palaniswamy et al., 2006). Activity of the transcription-associated genes with no TSS-proximal S/MAR is shown as a dashed line. The total TF genes are 35, 8, 19, 11, and 8 for S/MAR clusters A, B, C, D, and E, respectively, and 13, 9, 18, 12, and 6 for S/MAR clusters A, B, C, D, and E, respectively, for the non-TF genes. The error bars indicate the 95% confidence interval from the binomial test. (C) to (F) Developmentally important TF genes are associated with S/MARs. Shown are AGAMOUS (C), ATH1 (D), FWA (E), and SHORT ROOT (F) with the S/MARs and pS/MARs indicated. The tracks (top to bottom) are mononucleosome sequence coverage as in Figure 3B. The locations of the S/MARs color-coded by cluster (described above). The locations of the computationally predicted S/MARs (pS/MARs) (Rudd et al., 2004) are shown in gray for comparison. Genes are shown as green segments with exons indicated by boxes and the direction of transcription by white arrows.

Figure 5.

Analysis of DNA-Motif Content of S/MARs. List of 479 sequence motifs was assembled including S/MAR-specific motifs and annotated cis-elements (Supplemental Table 11). Forty-five motifs were significantly overrepresented in at least one S/MAR cluster as determined by t tests of motif density in S/MARs relative to the density of the motif in chr4 (Supplemental Table 12). Shown is the percentage of enrichment for 37 motifs that were enriched in S/MAR clusters A, B, C, and E. Eight motifs that were enriched only in cluster D S/MARs are not shown. Motif references are included in Supplemental Table 13.

Figure 6.

Analysis of the 6-Mer Content of S/MARs and Associated Regions. (A) The 6-mer content of S/MARs and select genomic regions. The occurrence of the 2080 nonredundant 6-mers was determined for each S/MAR cluster, pS/MARs, NDRs, promoters (Pr), exons (Ex), introns (In), and TEs on chr4. The heat map shows enrichment data for the 44 overrepresented 6-mers that are in common between S/MAR clusters A, B, C, D, and E. The heat map is scaled so that blue indicates depletion, gray is no enrichment or depletion, and yellow indicates enrichment. The maximum depletion shown is 73%, while the maximum enrichment is 230%. (B) Enrichment of 6-mer classes as a function of their poly(dA:dT) content for the five S/MAR clusters. All 6-mers were classified based on their poly(dA:dT) content, as described in the text. The dashed lines are included to group the 6-mers by class. (C) As in (B) but for the following genomic regions: pS/MARs, NDRs, TEs, promoters, exons, and introns. (D) Percentage of S/MARs by cluster that overlap with a NDR and the NDR sequence coverage for these S/MARs. The percentage of NDR-positive S/MARs is shown as bars, while the NDR sequence coverage is shown by the points and solid line. (E) Comparison of the enrichment for AT and poly(dA:dT) content of S/MAR-positive and -negative NDRs. NDRs were partitioned into S/MAR-positive and -negative groups, and the AT content relative to chr4 was determined. For poly(dA:dT) content, sequence motifs comprising four or more contiguous As or contiguous Ts were identified for chr4 and for the NDRs, and the poly(dA:dT) enrichment for the NDRs, relative to chr4 was calculated. A Welch two-sample t test was used to compare the NDR groups. The error bars show the 95% confidence intervals for the means. For both comparisons, the difference between the means was highly significant (P ≤ 10⁻¹⁶) as indicated by the asterisks (Supplemental Table 16).