Brg1 modulates enhancer activation in mesoderm lineage commitment

The interplay between different levels of gene regulation in modulating developmental transcriptional programs, such as histone modifications and chromatin remodeling, is not well understood. Here, we show that the chromatin remodeling factor Brg1 is required for enhancer activation in mesoderm induction. In an embryonic stem cell-based directed differentiation assay, the absence of Brg1 results in a failure of cardiomyocyte differentiation and broad deregulation of lineage-specific gene expression during mesoderm induction. We find that Brg1 co-localizes with H3K27ac at distal enhancers and is required for robust H3K27 acetylation at distal enhancers that are activated during mesoderm induction. Brg1 is also required to maintain Polycomb-mediated repression of non-mesodermal developmental regulators, suggesting cooperativity between Brg1 and Polycomb complexes. Thus, Brg1 is essential for modulating active and repressive chromatin states during mesoderm lineage commitment, in particular the activation of developmentally important enhancers. These findings demonstrate interplay between chromatin remodeling complexes and histone modifications that, together, ensure robust and broad gene regulation during crucial lineage commitment decisions.

membrane. Membranes were blocked for 1 hour at room temperature with 5% milk Trisbuffered saline Tween (TBST). Following blocking, membranes were incubated with desired antibody in 5% milk TBST overnight at 4°C. Membranes were washed 4 times for 15 minutes at room temperature in TBST and then stained with secondary antibody in 5% milk TBST for 1 hour at room temperature. After antibody staining, membranes were washed as after primary incubation, incubated in SuperSignal chemiluminescence substrate (Thermo Scientific), and visualized. Antibodies used were anti-BRG1 (Santa Cruz sc-10768), anti-actin (Sigma A1978), and anti-FLAG (Sigma M2).

Immunofluorescence
Cultures were fixed for 30 minutes at room temperature in 3.7% formaldehyde D-PBS and washed once with D-PBS. Wells were blocked in 2% bovine serum albumin 0.1% Triton-X-100 D-PBS for 30 minutes at RT. After blocking, cultures were incubated with primary antibody at 4°C overnight. Slides were washed three times with 0.1% Triton X-100 D-PBS and incubated in secondary antibody at room temperature for 1 hour. After staining, slides were washed three times with 0.1% Triton X-100 D-PBS, stained with Hoechst 33342 (10 ug/mL) in D-PBS, and immediately imaged in 50 uL D-PBS.
Cells were pelleted at 1350 x g at 4°C in a table top centrifuge and resuspended in 0.5 mL cold ChIP lysis buffer (50 mM HEPES-NaOH, pH 7.5, 140 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, 0.1% sodium deoxycholate) and sonicated to 200-1000 bp fragments using a VirSonic sonicator. Sonicated lysates were cleared by pelleting insoluble material at 13,000 RPM at 4°C followed by incubation with 5 ug antibody overnight. Next, Protein A magnetic beads (45 uL) were added to the lysate and incubated at 4°C for 7 hrs. Prior to addition, magnetic beads were washed 3 times with block (0.5% BSA/PBS). Immunoprecipitated material was washed 2 times each with ChIP lysis buffer, high salt lysis buffer (50 mM HEPES-NaOH pH 7.5, 500 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, 0.1% sodium deoxycholate), and LiCl wash buffer (10 mM Tris-HCl pH 8.0, 250 mM LiCl, 1 mM EDTA, 0.5% NP-40, 0.5% sodium deoxycholate) and one time with TE plus NaCl, followed by elution and reverse crosslinking in 210 uL of 1% SDS in TE overnight at 65°C. 200 uL of uncrosslinked material was treated with RNase A for 2 hours, proteinase K for 2 hours, and extracted 2 times with phenol/chloroform/isoamyl alcohol. This was followed by ethanol precipitation with a glycogen coprecipitant, 80% ethanol wash and final resuspension in TE. Nucleic acid yield was determined via PicoGreen (Invitrogen). Adapter ligation and size selection (200-400 bp) were performed using a Beckman Coulter SPRI TE nucleic acid extractor, For BRG1-FLAG ChIP-seq, the protocol described above was followed except 8-10x10 6 cells were used as starting material and 10 ug antibody was incubated with Protein G magnetic beads for roughly 6 hours prior to washing and addition of the bead/antibody complex to chromatin lysate for immunoprecipitation overnight.
Brg1 ChIP-exo was performed as previously described (Serandour et al., 2013) using anti-BRG1 antibody (1mg, Abcam 110641). ChIPExo was performed on Brg1 ChIP material while still on protein G magnetic dynabeads. After incubation with the antibody, the beads were washed six times with RIPA buffer (50mM HEPES, pH7.6, 1mM EDTA, pH8.0, 0.7% Sodium deoxycholate, 1% NP-40 and 500mM Lithium chloride) and two times with Tris (10mM Tris.Cl, pH 8.0). The bead bound DNA were end polished at 30°C for 30mins in using T4 DNA polymerase, Klenow fragment of DNA polymerase and T4 polynucleotide kinase. From this point each subsequent enzymatic steps were followed by two washes with each of RIPA and 10mM Tris. P7 adapter was ligated to the DNA ends using T4 DNA ligase at 25°C for 60 mins followed by nick repair using Phi29 polymerase at 30°C for 20 mins. Samples were periodically vortexed at 900rpm in a thermomixure during enzymatic reaction. DNA was digested with λ and RecJf exonucleases at 37°C for 30 mins each. Samples were then eluted off the beads with 100µl of Elution buffer (50mM Tris.Cl, pH 8.0, 10mM EDTA, 1% SDS) by incubating at . Proteinase K was added to degrade proteins, crosslink was reversed by overnight incubation at 65°C and using Ampure beads ChIP DNA was purified. The purified DNA was denatured at 95°C for 5 mins before synthesis of the 2 nd strand by P7 primer extension in presence of Phi29 polymerase. Then, P5 adapters were ligated to the DNA ends and the DNA fragments were PCR amplified for 18 cycles using universal primers containing the index sequences. PCR products were purified using Ampure beads, size selected using 2% agarose gels in an E-Gel Electrophoresis system (Invitrogen), gel purified using minielute gel extraction columns (Qiagen) and eluted in 20µl of TE.
Samples were quantified and analysed on Qubit and Bioanalyser before sequencing on a Illumina HiSeq 2500 sequencing machine.

ChIP-seq/ChIP-exo analysis pipeline
Single end 40 bp reads were aligned to the mouse genome (mm9) using Bowtie (Langmead et al., 2009). Unique sequences were extended +200 bp and allocated in 25-bp bins. Input DNA was used as a background model. A Poissonian model was used to determine statistically enriched bins with a P-value threshold set at 1x10 -12 for H3K27me3, H3K27ac, and Suz12 and 1x10 -6 for Flag-Brg1 as described previously (Marson et al., 2008). Genomic browser tracks were generated using the Integrated Genome Viewer (Robinson et al., 2011). Browser tracks and other downstream analysis was performed on pooled data from multiple replicates. Similar trends were observed if replicates were analyzed individually.  Arid1b,Smarcd3,Smarcd2,Phf10,Dpf1,Dpf2,Dpf3,Smarce1,Arid2, and Arid1a were manually added to this list based on the literature. The median expression values of these genes for four stages of cardiac differentiation (Wamstad et al., 2012) were median centered and interquartile range scaled and clustered using the bioconductor package Hopach (http://www.bioconductor.org/packages/2.1/bioc/html/hopach.html) and a cosine angle distance metric.

Tissue and cell type expression of upregulated developmental TFs
Normalized expression data from the Gene Atlas GNF1M dataset (Su et al., 2004) for upregulated developmental TFs were averaged (mean) between replicates and visualized as a heatmap. Developmental TFs were identified as genes that are annotated as within both GO terms GO0003700 (sequence-specific DNA binding transcription factor activity) and GO0048856 (anatomical structure development).

Classification of BRG1 peaks, enhancers, and H3K27me3 domains
To identify BRG1 bound regions, we overlapped statistically enriched peaks over two replicates of FLAG ChIP-seq and required peaks in both replicates to be within 1 kb.
Calling peak overlap within 500 bp reduces the number of called peaks by less than 10%, and not much more by calls within 250 bp, indicating that concordant peaks overlap well. Peaks conserved across replicates were then merged into a single region that included both peaks and any genomic space in between. In order to classify genomic localization of BRG1 bound regions, Ensembl lists of genes and exons were Putative enhancer regions were identified by intersecting regions of H3K27ac enrichment for both biological replicates. The size of the enhancer represents the combined length of the enriched regions for each replicate. Blocks of genomic space ± 2.5 kb from the Ensembl TSSs were subtracted to yield a list of high-confidence putative distal enhancers. To determine the proportion of Brg1 peaks that fall within putative enhancers, we intersected these regions, requiring at least one base pair of overlap.
To identify enhancers associated with blocks labeled "super" enhancers, we utilized the ROSE (https:// bitbucket.org/young_computation/rose) algorithm that has been previously described (Hnisz et al., 2013). Briefly, enhancers within 12.5 kb were merged and ranked according to input-subtracted signal of H3K27ac, which is used to determine a H3K27ac signal inflection point and identify super enhancers. Enhancers within ± 2.5 kb of the TSS were excluded from this analysis. These domains of "super" enhancers were then intersected with all identified enhancers to identify enhancers associated with super enhancer domains.
Motif enrichment within BRG1+ enhancers was done by scanning the subset of BRG1bound regions that completely overlapped with a putative enhancer using TRANSFAC's 'match' algorithm for whether or not each vertebrate motif in the database was identified within 75 bp of the center of that enriched region, and the hypergeometric distribution was used to calculate the probability that a given number of each motif would be seen among BRG1-associated peaks, given its abundance among all H3K27-acetylated  , and the false discovery rates for these probabilities were controlled using the q-value (Storey and Tibshirani, 2003). Each enriched motif was annotated with the expression at each stage of in vitro cardiomyocyte differentiation using data from (Wamstad et al., 2012).

Replicate Correlation
Unique sequence were extended by 200 bp and allocated into 20 bp bins. Bin read density was calculated and compared across all genomic bins using a Spearman's rank correlation.

Visualization of ChIP-seq data
To determine the average profile of H3K27me3, H3K27ac, or SUZ12 ChIP-seq and Input signal around transcriptional start sites or enhancer regions, unique aligned reads within 5 kb of the TSS or the midpoint of enhancer regions were grouped into 50 bp bins and then normalized based on number of reads per library (reads per million). In cases where the ChIP-seq read density of many promoters or regions of interest were visualized together, normalized binned read counts for each gene were visualized as heatmaps using R. For boxplots, a normalized read count was computed ±5 kb from the TSS. For boxplots for enhancer regions, a log 2 fold change in normalized read count was computed between conditions for each predicted enhancer region or subset of enhancers investigated.

Polycomb target analysis
TSSs were extended 1 kb in each direction and compared to genome-wide data sets for H3K27me3 and Polycomb subunits in ES cells (Ku et al., 2008) to identify overlap.

Statistical Analysis
To look for statistical enrichment of BRG1 and enhancer overlap, we generated 10,000 lists of randomly generated enhancer regions of identical size to our experimentally-derived list. We then intersected each random list with BRG1 bound regions to generate a normal distribution of expected overlap. P-value represents the number of permutations over our random enhancer lists with equal or more overlap than our experimental list.
Median centered and interquartile range scaled RPKM expression values for chromatin regulators during directed cardiomyocyte differentiation of embryonic stem cells (Wamstad et al. 2012). Chromatin regulators are genes found in GO categories GO0006338 and GO00016569 and other manually curated genes.

Supplemental Table 2.
Differential gene expression between Day4 4OHT and Day4 Control treated Brg1f/f; Actin-CreER ESCs. Gene list is filtered to remove any genes not expressed above 0.5 RPKM in either condition. Includes genes significantly downregulated or upregulated at a false discovery rate of 1% and three fold change cutoffs (1.2x, 1.5x, and 2.0x).       (C) Boxplots of log 2 fold change of subsets of predicted enhancers. Enhancers are separated into Brg1 bound and unbound cohorts based on the presence or absence of a Brg1 peak respectively. These groups were subdivided into static enhancers (enhancers found in both embryonic stem cells and mesodermal cultures), activated enhancers (enhancers found only in mesodermal cultures), and "super enhancers" (predicted using methods described in (Hnisz et al., 2013)). BRG1 bound, activated enhancers show the greatest average loss in H3K27ac. "Super" enhancers bound by BRG1 had a more modest reduction in H3K27ac than BRG1-bound activated enhancers. N indicates the number of enhancers included in each set. Enhancers and promoters that gave undefined fold change values were excluded. All comparisons significant at P<0.00001 (by two-sided KS test).   (Ku et al. 2008). Upregulated genes demonstrate significant enrichment for both groups. * p = 9.03x10 -80 ; ** p=5.15x10 -82