Mapping of small RNAs in the human ENCODE regions

Abstract

The elucidation of the largely unknown transcriptome of small RNAs is crucial for the understanding of genome and cellular function. We report here the results of the analysis of small RNAs (< 50 nt) in the ENCODE regions of the human genome. Size-fractionated RNAs from four different cell lines (HepG2, HelaS3, GM06990, SK-N-SH) were mapped with the forward and reverse ENCODE high-density resolution tiling arrays. The top 1% of hybridization signals are termed SmRfrags (Small RNA fragments). Eight percent of SmRfrags overlap the GENCODE genes (CDS), given that the majority map to intergenic regions (34%), intronic regions (53%), and untranslated regions (UTRs) (5%). In addition, 9.6% and 16.8% of SmRfrags in the 5' UTR regions overlap significantly with His/Pol II/TAF250 binding sites and DNase I Hypersensitive sites, respectively (compared to the 5.3% and 9% expected). Interestingly, 17%-24% (depending on the cell line) of SmRfrags are sense-antisense strand pairs that show evidence of overlapping transcription. Only 3.4% and 7.2% of SmRfrags in intergenic regions overlap transcribed fragments (Txfrags) in HeLa and GM06990 cell lines, respectively. We hypothesized that a fraction of the identified SmRfrags corresponded to microRNAs. We tested by Northern blot a set of 15 high-likelihood predictions of microRNA candidates that overlap with smRfrags and validated three potential microRNAs ( approximately 20 nt length). Notably, most of the remaining candidates showed a larger hybridizing band ( approximately 100 nt) that could be a microRNA precursor. The small RNA transcriptome is emerging as an important and abundant component of the genome function.

Figure 1

SmRfrags Map in CDS ENCODE Regions, in 5′ UTR ENCODE Regions, and across the First Six Exons of ENCODE Genes in Different Cell Lines All panels show the percentage of total SmRfrags (CDS ENCODE regions [A], 5′ UTR ENCODE regions [B], across the first six exons of ENCODE genes [C]). “Encode Regions” indicates the ENCODE-array content in each category. p value was determined by Hypergeometric testing. Statistical significance is labeled by ^∗∗∗ for p values < 0.005. The analysis was performed on the forward data, but there were no identifiable differences between SmRfrags distributions on the two strands (data not shown). “RA” indicates retinoic-acid treatment (6 μM, 48 hr).

Figure 2

SmRfrags Localization Relative to TSSs, His.Pol.TAF Sites, and Proximal DNase Hypersensitive Sites The data are shown for SmRfrags mapping in the 5′ UTRs of annotated genes (TSSs [A], His.Pol.TAF sites [B], proximal DNase Hypersensitive sites [C]). Statistical significance is labeled by ^∗ for p values < 0.05, by ^∗∗ for p values < 0.01, and by ^∗∗∗ for p values < 0.005, via Hypergeometric test. “A” indicates retinoic-acid treatment (6 μM, 48 hr).

Figure 3

Correlation of Gene Expression and SmRfrags in TSSs Colors in the pie charts indicate the proportions of expressed genes (green [HeLa S3] and purple [GM06990]) and nonexpressed genes (gray) annotated in ENCODE regions. p value was determined with a Chi-square test.

Figure 4

SmRfrags in Different Cell Lines The fraction of SmRfrags detected in the indicated cell lines is shown. Percentage is expressed as the percentage of total SmRfrags.

Figure 5

Sense-Antisense SmRfrags, in Different Cell Lines In brackets, the numbers of sense and antisense pairs are indicated for each cell line. Percentage of total SmRfrags is shown in bold.

Figure 6

Three Potential New MicroRNAs in the ENCODE Regions Genomic localization of SmRfrags and Northern-blot analysis: Color bars depict the position of SmRfrags with their respective signal intensities (log2). The cutoff for the top1% positive signals ranges from 2.5 (log2) to 3.5 (log2), depending on the cell lines. Arrows indicate the strand direction. Grey bar represents the Northern-blot probe. The conservation pattern (“conservation” track) is based on the UCSC phastCons scores. This track shows evolutionary conservation in 17 vertebrates, including mammalian, amphibian, bird, and fish species, on the basis of phastCons, a phylogenetic hidden Markov model. Multiz alignments of the assemblies were used to generate this track (generated with UCSC genome browser). The conservation is visualized by a blue scale density gradient and sequence annotation specific for each species. Northern-blot validation of microRNAs: Probes are 25 mer LNA oligonucleotides (see Material and Methods). Each blot contains a positive control lane (PC), which is a 25-mer oligonucleotide with the complementary sequence used for the probe. Lanes (1), (2), (3), and (4) correspond to total RNA from HeLa S3, GM06990, SK-N-SH, and HepG2, respectively. Two specific bands of 25 and 70 nt were detected, corresponding to the mature and the precursor forms of the putative microRNA, respectively.