pubmed.ncbi.nlm.nih.gov

Widespread RNA binding by chromatin-associated proteins - PubMed

  • ️Fri Jan 01 2016

Widespread RNA binding by chromatin-associated proteins

David G Hendrickson et al. Genome Biol. 2016.

Abstract

Background: Recent evidence suggests that RNA interaction can regulate the activity and localization of chromatin-associated proteins. However, it is unknown if these observations are specialized instances for a few key RNAs and chromatin factors in specific contexts, or a general mechanism underlying the establishment of chromatin state and regulation of gene expression.

Results: Here, we perform formaldehyde RNA immunoprecipitation (fRIP-Seq) to survey the RNA associated with a panel of 24 chromatin regulators and traditional RNA binding proteins. For each protein that reproducibly bound measurable quantities of bulk RNA (90% of the panel), we detect enrichment for hundreds to thousands of both noncoding and mRNA transcripts.

Conclusion: For each protein, we find that the enriched sets of RNAs share distinct biochemical, functional, and chromatin properties. Thus, these data provide evidence for widespread specific and relevant RNA association across diverse classes of chromatin-modifying complexes.

PubMed Disclaimer

Figures

Fig. 1
Fig. 1

fRIP-Seq reveals widespread binding of chromatin-associated proteins to mRNAs and lncRNAs. a Formaldehyde cross-linking RNA–protein complexes enables identification of target RNAs by high-throughput sequencing. b We mapped RNA interaction partners for 24 proteins in triplicate and performed hierarchical clustering of log2 fRIP/input fold change over the ~25,000 genes present in at least one condition. Replicates cluster together for every protein. Binding patterns vary between proteins. Neither total RNA recovery from fRIP-Seq (orange 1–10 nanogram range, green 10–50 nanogram range, purple 50+ nanograms) or published nucleic acid binding properties (orange DNA, green both DNA and RNA, purple RNA) can explain the observed clustering solution. c The lncRNA Xist is significantly bound by HNRNPU and PRC2 components SUZ12 and EZH2 in our data, validating these known interactions. fRIP-Seq coverage suggests potential binding patterns along the transcript. The coverage scales (y-axis) have maximum coverage 500, except for CHD4 and EZH2, which have maximum coverage 1500 and 3000, respectively

Fig. 2
Fig. 2

Proteins bind RNA at various stages of RNA processing. The proteins varied on the proportion of intron alignments in the fRIP-Seq versus input. a The heat map shows the average proportion of a gene’s FPKM assigned to exon isoforms versus unspliced pre-RNA isoforms (“Materials and methods”). The scatter plots show every gene for ADAR and CHD4. Traditional RBPs ADAR, HNRNPH1, HNRNPU, and HUR likely bind co-transcriptionally; thus, they often immunoprecipitate with unspliced transcripts. b RBPs also varied on the degree to which the fRIP/input fold change correlated with input FPKM. The heat map plots the Spearman correlation of these values, and the scatter plots show every gene with a generalized additive model regression line. c Relationships between input FPKM and fold change were consistent between single and multi-exonic genes

Fig. 3
Fig. 3

Chromatin-associated proteins bind functionally coherent sets of mRNA. RBPs differed on the degree to which they preferred to bind mRNAs versus lncRNAs. a To properly compare the two, we sampled a set of low abundance mRNAs to match the distribution of lncRNAs (referred to as mRNA_lncFPKM) and plotted the FPKM distributions for each set. b The heat map plots the Z scores of Mann–Whitney U tests comparing the distributions of fold changes for lncRNAs and low abundance mRNAs. To its right, we plot the empirical cumulative distribution functions for HUR and SUZ12. c We partitioned significantly enriched genes from all fRIP-Seqs that were also enriched by twofold or more into ten distinct groups using k-medoid clustering. A gene set enrichment analysis using DAVID found significantly enriched functional annotations for each cluster (“Materials and methods”)

Fig. 4
Fig. 4

Chromatin-associated proteins prefer specific gene structure properties. The proteins had strong preferences for transcript length (a) and exon number (b). The heat maps plot the Spearman correlation between fRIP/input fold change and length or exon number. To their right, we plot empirical cumulative density functions for specific proteins exemplifying substantial correlations. c Because length and exon number are highly correlated, we isolated the role of each using semipartial correlations. We regressed fold change against exon number and computed Spearman correlation of the residual against length (Isolated fRIP vs length correlation) and vice versa (Isolated fRIP vs exons correlation). As can be seen in the resulting plot, and further explored in (d), SUZ12 is affected by exon number rather than length. e In contrast, HDAC1 correlates with length at every level of exon number plotted

Fig. 5
Fig. 5

Chromatin-associated proteins prefer specific sequence motifs. We searched for motifs that have high mutual information with the fRIP/input differential expression statistic using FIRE (“Materials and methods”). Motifs discovered for HUR (a), HNRNPH1 (b), and HNRNPU (c) matched well to known motifs in the RBPmap database [64]. For each motif, we plotted the empirical cumulative density function of the fRIP/input statistic for genes with and without the motif. Below that, we plotted the 25th, 50th, and 75th percentiles of the fRIP/input statistic for genes containing the specified number of motif occurrences. We discovered novel motifs at similar levels of significance for CAPs SUZ12 (d), CBP (e), and HDAC1 (f)

Fig. 6
Fig. 6

Protein binding to RNA relates to local chromatin. Because chromatin marks measured by RRBS (DNA methylation) or ChIP-Seq (histone modifications and modifiers) correlate with gene abundance (Fig. S16 in Additional file 1), we plotted this relationship separately for genes bound and unbound by each protein. a DNMT1-bound RNAs have less DNA methylation in their promoter, shown as a scatter plot of every gene with a generalized additive model regression. b WDR5-bound RNAs have more H3K4me3 in their promoter. c For each chromatin mark and protein, we plotted the difference between the bound and unbound regression lines as a heat map (“Materials and methods”), revealing a clear difference in the relationship of certain proteins to activating chromatin marks. d CHST2 exemplifies a WDR5-bound RNA with ample H3K4me3

Similar articles

Cited by

References

    1. Baltz AG, Munschauer M, Schwanhäusser B, Vasile A, Murakawa Y, Schueler M, et al. The mRNA-bound proteome and its global occupancy profile on protein-coding transcripts. Mol Cell. 2012;46(5):674–90. doi: 10.1016/j.molcel.2012.05.021. - DOI - PubMed
    1. Bernstein E, David AC. RNA meets chromatin. Genes Dev. 2005;19(14):1635–55. doi: 10.1101/gad.1324305. - DOI - PubMed
    1. Braunschweig U, Gueroussov S, Plocik AM, Graveley BR, Blencowe BJ. Dynamic integration of splicing within gene regulatory pathways. Cell. 2013;152(6):1252–69. doi: 10.1016/j.cell.2013.02.034. - DOI - PMC - PubMed
    1. Brockdorff N. Noncoding RNA, and Polycomb recruitment. RNA. 2013;19(4):429–42. doi: 10.1261/rna.037598.112. - DOI - PMC - PubMed
    1. Cabili MN, Cole T, Loyal G, Magdalena K, Barbara T-V, Aviv R, et al. Integrative annotation of human large intergenic noncoding RNAs reveals global properties and specific subclasses. Genes Dev. 2011;25(18):1915–27. doi: 10.1101/gad.17446611. - DOI - PMC - PubMed

Publication types

MeSH terms

Substances