Genome-wide profiling of YY1 binding sites during skeletal myogenesis

Abstract

Skeletal muscle differentiation is regulated by a network of transcription factors, epigenetic regulators and noncoding RNAs. We have recently performed ChIP-seq experiments to explore the genome-wide binding of transcription factor YY1 in skeletal muscle cells. Our results identified thousands of YY1 binding peaks, underscoring its multifaceted functions in muscle cells. In particular, we identified a very high proportion of YY1 binding peaks residing in the intergenic regions, which led to the discovery of some novel lincRNAs under YY1 regulation. Here we describe the details of the ChIP-seq experiments and data analysis procedures associated with the study published by Lu et al. in the EMBO Journal in 2013 [1].

Specifications
Organism/cell line/tissue	Mus musculus/C2C12
Sex	NA
Sequencer or array type	Illumina Hiseq 2000, Illumina GA IIx
Data format	Raw data: FASTQ files Processed data: BEDGRAPH, TXT
Experimental factors	Myoblast vs myotube
Experimental features	Using ChIP-seq, we generated genome-wide maps of YY1 in skeletal myoblasts and myotubes with biological replicates. We found that a large proportion of the binding sites reside in the intergenic regions; therefore, many lincRNAs are regulated by YY1.
Consent	NA
Sample source location	Manassas, VA, USA

Direct link to deposited data

http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE45875

Experimental Design, Materials and Methods

Cell culture

Mouse C2C12 myoblast cell line was purchased from American Type Culture Collection (ATCC). The myoblasts were maintained in a growth medium (DMEM, 10%FBS and 1% penicillin/streptomycin), and induced to myotubes by culturing in a differentiation medium (DMEM, 2% horse serum and 1% penicillin/streptomycin).

ChIP assays and sequencing experiments

ChIP assays were performed as previously described [2], [3]. About 2 × 10⁷ C2C12 cells and 5 μg of antibodies were used in one immunoprecipitation. The antibodies include YY1 #1 (Santa Cruz Biotechnology, Cat# SC-1703, rabbit polyclonal), YY1 #2 (Abcam, Cat# AB58066, mouse monoclonal), Ezh2 (Cell Signaling, MA, USA, Cat# AC22), trimethyl-histone H3-K27 (Millipore, Cat# 07-449), trimethyl-histone H3-K4 (Millipore, Cat# 07-473), or normal mouse IgG (Santa Cruz Biotechnology, Cat# SC-2025) as a negative control.

For library construction, we used a protocol as described before [4]. Briefly, the immunoprecipitated DNA (~ 10 ng) were end-repaired, and A-nucleotide overhangs were then added, followed by adapter ligation, PCR enrichment, size selection and purification. The purified DNA library products were evaluated using Bioanalyzer (Agilent) and SYBR qPCR and diluted to 10 nM for sequencing on Illumina Hi-seq 2000 sequencer (YY1) (pair-end with 50 bp) or Illumina Genome Analyzer II sequencer (Ezh2, H3K27me3 and H3K4me3) (pair-end with 36 bp). Technical replicates were prepared by sequencing the same library twice. A data analysis pipeline CASAVA 1.8 (Illumina) was employed to perform the initial bioinformatic analysis (base calling). Table 1 lists all the experiments that we had performed. For MB YY1, we performed two biological replicates with the antibody SC-1703 and a third biological replicate with a second antibody AB58066. We also performed two technical replicates for each antibody (run 1 and run 2).

Table 1.

List of ChIP-seq experiments.

IP	Read length	Total readsd	Mapped readse	No. of peaks
YY1 (MB, rep1)a	50 bp	106.4	80	1820
YY1 (MB, rep2run1)b	36 bp	34.0	25.6	996
YY1 (MB, rep2run2)b	36 bp	34.5	26.0	1061
YY1 (MB, rep3run1)c	36 bp	28.9	22.4	1504
YY1 (MB, rep3run2)c	36 bp	30.9	23.9	1655
YY1 (MT)d	50 bp	86.8	62.5	626
Ezh2	50 bp	37.0	25.7	1801
H3K4me3	36 bp	10.2	6.9	21,051
H3K27me3	36 bp	26.5	20.8	10,674

^a,c,dUsing SC1703 antibody.

^bUsing AB58066 antibody.

^eTotal reads and mapped reads are reported as millions of reads. The number of uniquely mapped reads, the number of reads aligning and the number of pairs concordantly are all the same as the number of the mapped reads in this experiment based on the alignment protocol used.

Reads alignment, peak defining and motif analysis

The sequenced reads were mapped to the mouse reference genome (UCSC mm9, non-repeat-masked) using SOAP2 [5] (version 2.20, with the following parameters: “-v 2 -r 0 -m 0 -p 20”) allowing a maximum of two mismatches and only the uniquely aligned reads were kept. The protein–DNA binding peaks were identified using Model-based Analysis for ChIP-seq (MACS [6], version 2.0.9; for YY1 ChIP-seq (MB rep1); the parameters are “-g mm -m 8,30 -p 0.001” and then the peaks were filtered by q-values; for others, the parameters were “-g mm -m 8,30 -q 0.01”) with the IgG control sample as background. During the peak calling, a q-value (adjusted P-value calculated using the Benjamini–Hochberg procedure) was set under 10^− 5 for YY1; it corresponds to an empirical FDR (False Discovery Rate) of 3.4%; 10^− 2 was used for Ezh2 and H3K27me3, where the FDRs were estimated to be around 1%. This difference in data processing for ChIP-seq experiments was because the performance of MACS on a large dataset (e. g., YY1 ChIP-seq (MB rep1) sequenced on Hi-seq 2000) is not as good as on a small dataset (e.g., others on GA IIx). The mapping information and number of peaks from each dataset were shown in Table 1.

The raw ChIP-seq sequencing reads for transcription factor MyoD were downloaded from NCBI's Sequence Read Archive (SRA; http://www.ncbi.nlm.nih.gov/sra) with accession number SRX016191 and SRX016040 for MBs and MTs, respectively. The reads were aligned using the above method and the MyoD-binding peaks were identified using a q-value cutoff of 10^− 2. The processed ChIP-seq data (binding sites) for Pol II and H3K4me3 were obtained from NCBI's GEO under accession number GSE25308. When comparing peaks from different experiments, two peaks were considered as “overlapped peaks” if the distance between them was less than 1 kb.

In order to search for highly occurring motifs in the DNA sequences underlying the putative binding peaks, Discriminative Regular Expression Motif Elicitation (DREME [7], version 4.8.0) was applied on the 100 bp (± 50 bp) sequences flanking the peak summit. The analysis was run on both strands to search for motifs that are no more than 8 bp in length with E-values < 0.01.

Quality control

In peak defining, we used the IgG as a negative control and also carefully selected the q-values for a reasonable FDR. According to the ENCODE ChIP-seq guidelines [8], we calculated the Fragments in Peaks (FRiP) value using in-house programs (See Supplementary Material). Moreover, for the YY1 biological replicates, we performed Irreproducible Discovery Rate (IDR) analysis using the package developed by Li et al. [9].

Functional annotation

To identify putative YY1 target gene, each identified peak was associated with the closest RefSeq gene when it falls into the 4 kb (± 2 kb) flanking region of the gene's TSS, and these genes were considered as potentially regulated by YY1. For analysis of differentially expressed genes, we used Cufflinks [10] (version 1.3) to evaluate the expression profile (using Fragments Per Kilobase of exon model per Million mapped reads, FPKM) of all the RefSeq transcripts using the publically available RNA-Seq data obtained from − 24 h (myoblasts, MBs) and 60 h (myotubes, MTs) C2C12 [11]. Differentially expressed genes were defined as those up- or down-regulated in MTs as compared to MB. If a gene is differentially expressed and bound by YY1, we reason that it could be potentially regulated by YY1 since the YY1 level decreases during C2C12 differentiation. Up-regulated YY1 bound genes were defined if their expression in MTs is > 1.2 fold higher compared with MBs and these genes are possibly repressed by YY1 in MBs. Down-regulated YY1 bound genes were defined if their expression in MTs is less than 0.8 fold compared with MBs and they are likely activated by YY1 in MBs. Then Gene Ontology (GO) analysis was performed on both up- and down-regulated genes using Database for Annotation, Visualization and Integrated Discovery (DAVID, http://david.abcc.ncifcrf.gov/) [12], [13] for functional annotations.

Identification of YY1 bound novel lincRNAs

Since we observed that more than 1/4 of the YY1 peaks were in the intergenic regions, we suspected that YY1 may regulate unannotated lincRNAs. To validate this hypothesis, we used the list of novel lincRNA identified by Guttmanet et al. from four mouse cell types [14]. YY1-binding sites were searched in the flanking regions (± 100 kb on both sides) of these lincRNAs. The resultant list of lincRNAs was considered as YY1-associated muscle lincRNAs or Yams.

Conflict of interest statement

The authors declare no conflict of interest.

Appendix A. Supplementary data

Supplementary material

Perl script for calculating FRiP (Fragments in Peaks) value.