Stage and Gene Specific Signatures Defined by Histones H3K4me2 and H3K27me3 Accompany Mammalian Retina Maturation In Vivo

The epigenetic contribution to neurogenesis is largely unknown. There is, however, growing evidence that posttranslational modification of histones is a dynamic process that shows many correlations with gene expression. Here we have followed the genome-wide distribution of two important histone H3 modifications, H3K4me2 and H3K27me3 during late mouse retina development. The retina provides an ideal model for these studies because of its well-characterized structure and development and also the extensive studies of the retinal transcriptome and its development. We found that a group of genes expressed only in mature rod photoreceptors have a unique signature consisting of de-novo accumulation of H3K4me2, both at the transcription start site (TSS) and over the whole gene, that correlates with the increase in transcription, but no accumulation of H3K27me3 at any stage. By in silico analysis of this unique signature we have identified a larger group of genes that may be selectively expressed in mature rod photoreceptors. We also found that the distribution of H3K4me2 and H3K27me3 on the genes widely expressed is not always associated with their transcriptional levels. Different histone signatures for retinal genes with the same gene expression pattern suggest the diversities of epigenetic regulation. Genes without H3K4me2 and H3K27me3 accumulation at any stage represent a large group of transcripts never expressed in retina. The epigenetic signatures defined by H3K4me2 and H3K27me3 can distinguish cell-type specific genes from widespread transcripts and may be reflective of cell specificity during retina maturation. In addition to the developmental patterns seen in wild type retina, the dramatic changes of histone modification in the retinas of mutant animals lacking rod photoreceptors provide a tool to study the epigenetic changes in other cell types and thus describe a broad range of epigenetic events in a solid tissue in vivo.

For ChIP-Seq, nuclei were resuspended in 1ml RSB (PMSF+PI) and DNA concentrations measured spectrophotometrically. Micrococcal nuclease (MN) test digestions were carried out to determine the time interval needed to produce predominantly mononucleosomes and this was used for preparative digestion. For preparative micrococcal nuclease digestion, nuclei (0.5mg/ml DNA) was resuspended in 1 ml RSB, 0.5mM PMSF, 1mM CaCl 2 , 2.5units/ml MN, incubated at 37˚C for 45-60 min and terminated by 5mM EDTA. Nuclei were centrifuged for 7 min at 7,500rpm, pellet was resuspended in 500ul L-CHIP buffer (1% SDS, 10mM EDTA, 50mM Tris-HCl pH8.0), 1mM PMSF and PI, sonicated twice at setting 3 for 10 sec on Sonic Dismembrator (Fisher Scientific, Model 100). The pellet of mononucleosomes was subjected to ChIP after precleaning by centrifugation at 14,000rpm for 5 min. Protein concentration was adjusted to 1mg/ml with L-CHIP buffer. For ChIP-qPCR samples were treated similarly except that sonication to shear DNA to lengths of between 200 and 2,000bp was used.
Chromatin immunoprecipitation. Chromatin was diluted 10 fold in D-CHIP buffer and 5ug antibody was added and incubated with rotation overnight at 4˚C. Simultaneously 30ul protein A beads (Sigma) slurry were washed 2 times in washing buffer with 9:1 of D-CHIP (dilution buffer: 0.01% SDS, 1.1% Triton X-100, 1.2mM EDTA, 16.7mM Tris-HCl pH8.0, 167mM NaCl) and L-CHIP, resuspended in the same buffer with 500ug/ml salmon sperm DNA (Invitrogen) and 100ug/ml BSA (Invitrogen) and incubated on rotator overnight at 4˚C. Beads were washed 2 times with washing buffer, combined with the chromatin/antibody mix and rotated for 2 hours at 4˚C. Beads were washed 4 times with 1 ml LS-CHIP buffer (low salt buffer: 0.1% SDS, 1% Triton X-100, 2mM EDTA, 20mM Tris-HCl pH8.0, 150mM NaCl), 1 time with 1ml HS-CHIP buffer (high salt buffer: 0.1% SDS, 1% Triton X-100, 2mM EDTA, 20mM Tris-HCl pH8.0, 500mM NaCl) and eluted with 350ul of E-CHIP buffer (elution buffer: 1% SDS, 0.1M NaHCO 3 ) by rotating at room temp for 10min. Immunoprecipitate (IP) and 50ul of input chromatin (Input) were treated with 0.5mg/ml of Proteinase K (Roche) and RNaseA (Roche) at 37˚C for 30 min and uncrossliked at 65˚C overnight. DNA was extracted twice with phenol/chloroform, once with chloroform and ethanol precipitated with glycogen (Roche) and sodium acetate. DNA was dissolved in 50ul water and subjected to qPCR or was used to prepare libraries with Data preparation. We used program Bowtie (version 0.12.4) to map all sequence reads to the mouse genome NCBI37/mm9. Sequence reads mapped to multiple genomic locations were excluded from subsequent analysis. The mid point of each unique-mapped reads was calculated with the assumption that all fragments are 190bp in size and were used for subsequent analysis.
One Bedfile was created for each sample by counting number of unique-mapped reads in each 100bp-window along all chromosomes normalized to the total unique-mapped reads in the sample. The final Bedfiles were uploaded to a website, and custom tracks were created in UCSC Genome Browser with links pointing to the real Bedfiles, that were used for specific gene mapping.

Tag distribution analysis in defined TSS region.
To create the tag distribution around the TSS, we downloaded genomic data of RefSeq [35,56] (http://hgdownload.cse.ucsc.edu/goldenPath/mm9/database/refGene.txt), and counted the number of unique-mapped reads located in TSS +/-5Kb. Then the number of reads at each position was divided by the number of total reads in their regions. Smoothing function LOWESS in R package was used to calculate a smoothed density curve.

Detection of tag enriched region in genome.
We applied a Poisson distribution-based model to detect enriched regions. We tested all windows in whole chromosome with pace of 100bp, windows with p-value below 1.0e-5 were collected as enriched, and overlapping windows were merged into single enriched region. We also tested the same enriched region on anti-GFP control samples and removed those regions from the enriched list if they were also significantly enriched in the GFP positive samples.
Detection of tag enriched location in individual genes. The reads from each experiment were mapped to mouse genome and analyzed with NextGENe software (version 2.10). Bed format files for specific genes were made for promoter area from +2.5Kb to TSS or for whole body of the gene from TSS to TES and amount of reads were calculated for each genome interval at each developmental stage with NextGENe software. Normalized occupancy for each genome interval =(amount of reads x 5,000,000 reads x 1,000Kb)/(whole amount of reads in experiment x length of genome interval in Kb).