SWI/SNF regulates half of its targets without the need of ATP-driven nucleosome remodeling by Brahma

On average, the BRM-bound genes in S2 cells were characterized by higher levels of RNA polymerase II (RNAPII) than the rest of the genes (Fig. 1c). This observation is in good agreement with previous studies showing that BRM associates preferentially with transcriptionally active loci in the polytene chromosomes of D. melanogaster [24] and Chironomus tentans [13]. However, the levels of BRM were not correlated with the levels of RNAPII at individual genes (Additional file 1: Figure S1), in agreement with the fact that many BRM target genes are repressed by BRM (see for example reference [3] and our results below). Moreover, the BRM-bound genes were characterized by a much lower nucleosome density in the gene body than the unbound genes, and by more pronounced nucleosome-depleted regions upstream and downstream of the TSS and TES, which can be seen by comparing our ChIP-seq results with results from a micrococcal nuclease study (MNase-seq) carried out by Shi et al. [25] (Fig. 1d). Interestingly, BRM was preferentially associated with long genes (> 15 kb) and was underrepresented in short genes (< 1.5 kb) (Fig. 1a). The presence of BRM in gene bodies and its preferential association with long genes suggests that SWI/SNF plays a prominent role in transcription elongation. The presence of BRM in gene bodies may also be linked to its role in pre-mRNA splicing [12, 13]. Roles in RNAPII elongation have been attributed to Swi2 in budding yeast [4] and BRG1 in mammals [5], and may provide a mechanistic basis for the regulation of alternative splicing by SWI/SNF [12].

We first analyzed the transcriptomes of the S2 cells that slightly overexpressed wild-type BRM in a depletion background, and identified 2112 genes with differential expression compared to control cells. Out of the 2112 genes, 541 were BRM-bound (245 decreased and 296 increased expression levels). This group of 541 genes occupied by BRM and affected by expression of wild-type BRM constituted a confident set of BRM target genes in S2 cells (Additional file: Table S1 and Additional file: Figure S2) and were studied further. 1 1

A gene ontology (GO) classification of biological processes showed a large functional diversity among the BRM target genes, many of which encode regulatory proteins and enzymes, and supported the involvement of BRM in important biological processes such as metabolism, cell communication, immune response and development (Additional file 1: Figure S3). The BRM target genes were more expressed in S2 cells than the rest of the genes (Fig. 2b), and were occupied by BRM in a bimodal fashion, with highest occupancy upstream of the TSS, as for the total of the BRM bound genes (compare Fig. 1a and Additional file 1). Moreover, the genes repressed by BRM in S2 cells were characterized by higher levels of BRM in the gene bodies than the upregulated genes (Additional file 1: Figure S4).

The BRM targets were also characterized by higher RNAPII levels (Fig. 2c) and lower nucleosome levels (Fig. 2d) than the average of the genome, in both the gene body and flanking regions. Interestingly, 46% of the identified targets (245 out of 541) were repressed by BRM, which points to a dual active/repressive role for BRM in transcription regulation and shows that the repressive function of BRM is very prevalent in the fly genome. The genes that were repressed by BRM showed on average higher RNAPII occupancy and lower nucleosome density in the gene body than the genes upregulated by BRM (Additional file 1: Figure S5).

We considered the possibility that the effects observed in cells that expressed the ATPase mutant could be due to the low levels of endogenous active BRM, not to the function of the mutant protein. If so, BRM depletion should reproduce the effect of expressing mutant BRM, which was not the case. Only 24 out of the 267 ATPase independent genes were differentially expressed in BRM-depleted cells compared to control dsGFP cells (Additional file: Table S2). This observation rules out the possibility that the observed effects are due to low levels of active BRM. 1

Of the 541 BRM targets identified above, 274 reacted differently to the expression of wild-type and mutant BRM. The mutant protein lacks ATPase activity, but is able to assemble into SWI/SNF complexes (reference [20] and Additional file 1: Figure S6). Moreover, ChIP-seq analysis of recombinant BRM proteins showed that both the wild-type and the mutant BRM were recruited to the genomic targets identified above (Additional file 1: Figure S7). We also analyzed the association of wild-type and mutant BRM with the ATPase-dependent and ATPase-independent genes using metagene profile comparisons and correlation analyses (Additional file 1: Figure S8). These comparisons supported the conclusion that the wild-type and the non-catalytic mutant BRM are recruited to both ATPase-dependent and ATPase-independent genes, which rules out differences due to defects in SWI/SNF targeting. Thus, we concluded that the 274 genes that reacted differently to the expression of wild-type and mutant BRM are regulated in an ATPase-dependent manner (Additional file 1: Table S1, Figure 3).

The levels of recombinant BRM, either wild-type or mutant, were higher in the gene body and lower downstream of the TES in the ATPase-dependent genes than in the ATPase-independent ones, as observed for endogenous BRM (compare the metagene plot in Fig. 5b with the plots in Additional file 1: Figure S8A). The relative abundance of the recombinant BRM proteins at the TSS was instead different from that of endogenous BRM. The fact that different antibodies were used to immunoprecipitate the recombinant and endogenous proteins could explain this difference.

The promoters of the BRM targets were highly enriched in binding motifs for different transcription factors, mostly homeobox proteins. Some of the enriched motifs were common for ATPase-dependent and ATPase-independent genes, but many others were specific (Additional file: Table S3). The enrichment of certain DNA-binding motifs in the promoters of the BRM targets could be due to secondary events caused by changes in the level of expression of the identified transcription factors. However, the expression of the majority of the identified transcription factors was not changed in cells that expressed wild-type or mutant BRM (Additional file: Table S4), which supports the conclusion that the observed effects are to a large extent direct. In summary, the motif enrichment analysis suggests that the ATPase-dependent and ATPase-independent genes are characterized by different promoter configurations. The abundance of motifs for developmental regulators in both the ATPase-dependent and ATPase-independent genes is in agreement with the GO analysis reported above and confirms that both groups of genes play roles in the regulation of developmental processes. 1 1

The average gene length was similar in the ATPase-dependent and ATPase-independent genes (8.3 and 9 kB, respectively, p = 0.76). Instead the average expression level was slightly lower for the ATPase-independent genes (average CPM values were 236 vs 127; p = 0.048). The ATPase-independent group was also characterized by higher BRM binding levels at the TSS (Fig. 5b) and higher nucleosome occupancy in the gene body than the ATPase-dependent genes (Fig. 5c). The binding and distribution of RNAPII were very similar in both groups (Additional file 1: Figure S10).

Analysis of MNase data from Shi et al. [25] revealed that BRM depletion causes a remarkable drop of nucleosome occupancy upstream of TSSs and downstream of TESs, for both groups of genes (Fig. 5d), which reveals an important role for SWI/SNF in maintaining the nucleosome configuration in genomic regions that flank transcription units. Moreover, BRM depletion caused a significant increase in nucleosome occupancy in the gene bodies of the ATPase-independent genes, which implies that a major function of BRM in this group of genes is to keep nucleosome density low through non-catalytic mechanisms. On the other hand, BRM depletion did not cause any major change in nucleosomes occupancy in the gene bodies of the ATPase-dependent genes (Fig. 5d). This difference reveals a dual function of BRM in the regulation of ATPase-dependent and ATPase-independent genes.

Next, we performed clustering analysis on the ATPase-independent genes and identified a subset of 62 genes for which the average increase in nucleosome density upon BRM depletion was most pronounced (Cluster 1 in Fig. 5d, right panel). Interestingly, the genes in Cluster 1 were on average twice as long as the average gene in the fly genome (8.4 kb compared to 4.75, p = 0.01). Taken together, these observations suggest that BRM acts to reduce nucleosome density in the gene bodies of these long genes, probably to regulate transcription elongation, and that it does so through an ATPase-independent mechanism. Cluster 1 is significantly enriched in lipid-binding proteins (GO:0008289, FDR q-value = 0.01) and includes genes that code for proteins with important roles in signal transduction and juvenile hormone function, which strengthens the conclusion that the ATPase-independent genes mediate important biological roles.

Drosophila melanogaster S2 cells were purchased from Invitrogen and cultured in Schneider’s Drosophila medium (Invitrogen) containing 10% heat-inactivated FBS, 50 U/ml penicillin and 50 μg/ml streptomycin at 28 °C.

The anti-BRM antibody used for Western blotting was raised in rabbit and has been characterized in a previous study [13]. An antibody raised against the rat BRG1 protein by Östlund-Farrants et al. [35] was used to detect BRM in S2 cells. The specificity of this antibody in D. melanogaster has been previously documented [13]. The antibodies against MOR and SNR1 were kindly provided by C. P. Verrijzer (see reference [36] and references therein). The monoclonal anti-alpha-tubulin antibody (clone B-5-1-2) was from Sigma-Aldrich and the anti-V5 antibodies were from Invitrogen (R96025↗) or Abcam (ab9116).

RNAi experiments in S2 cells were carried out as previously described by Tyagi et al. [13]. Double-stranded RNAs complementary to BRM or GFP were synthesized by in vitro transcription using the MegaScript RNAi kit (Ambion) from gene-specific PCR fragments with incorporated T7 promoters at both ends. The sequences of the PCR primers used for dsRNA synthesis are provided in Table S5 (Additional file 1). 3 × 10⁶ S2 cells were cultured in 6-well plates overnight, washed with serum-free and antibiotic-free Schneider’s medium, and treated with 30 μg of dsRNA/well. The cells were harvested 48 h after the addition of dsRNA.

The stably transfected cell lines for expression of recBRM-V5 and recBRM-K804R-V5 were constructed and cultivated as described in Yu et al. [15]. The expressions of recBRM-V5 and recBRM-K804R-V5 were under the control of the metallothionein promoter of the pMT vector (Invitrogen), which can be modulated using different concentrations of copper sulfate. For RNA-seq and ChIP-seq experiments, the expression of the endogenous BRM in the stably transfected cells was knocked down by RNAi, and the recombinant BRM proteins were expressed. To this end, the cells were first treated with dsRNA as indicated above, 100 μM copper sulfate was added to the cultures 24 h after the start of the RNAi treatment, and the incubation was allowed to proceed for additional 24 h.

Cell pellets were resuspended in SDS-PAGE sample buffer supplemented with 8 M urea and heated to 100 °C to denature and extract proteins. Proteins were separated by SDS-PAGE and transferred to polyvinylidenefluoride membrane in Tris-glycine buffer with 0.02% SDS and 4 M urea using a semi-dry electrophoretic transfer cell. The antibodies were diluted in PBS containing 0.05% Tween-20 and 1% low-fat powder milk, and antibody incubations were carried out following standard procedures. The ECL system (GE Healthcare) was used for the chemiluminiscent detection of secondary antibodies conjugated to horseradish peroxidase.

S2 cells were harvested and resuspended in PBS supplemented with 0.2% NP40, and homogenized using a glass homogenizer with a tight-fitting pestle. The homogenate was centrifuged at 1.500 g for 10 min at 4 °C. The pellet containing the nuclei was resuspended in PBS, sonicated and centrifuged at 16.300 g for 10 min at 4 °C. The resulting supernatant was supplemented with 0.1% NP40, cleared by centrifugation at 13.200 rpm for 15 min and used as input for immunoprecipitation reactions following standard procedures. Antibody incubations were overnight under rotation at 4 °C. Samples were washed in PBS supplemented with 0.1% NP40 for four times 5 min. The bound proteins were eluted with 1% SDS, precipitated with acetone and analyzed by SDS-PAGE and Western blotting.

Chromatin was extracted from S2 cells as previously described by Hessle et al. [37]. In short, the fixed cells were lysed and sonicated, the lysates were cleared by centrifugation, and antibodies against either endogenous BRM or V5 (Invitrogen) were added and incubated overnight under rotation. Protein-G coated magnetic beads were added to the samples and the incubation was allowed to proceed for one additional hour. The beads were then washed in 0.8% RIPA buffer for four times for 10 min each. The beads were resuspended in TE buffer containing 0.5% SDS, 0.1 mg/ml RNase A and 0.2 mg/ml proteinase K, and incubated at 55 °C for 3 h and at 65 °C overnight to reverse the crosslinking. The immunoprecipitated DNA was extracted with a mixture of phenol, chloroform and isoamylalcohol (Sigma). For qPCR analyses, the KAPA SYBR Fast qPCR Kit (KAPA Biosystem) was used in a QIAGEN Rotor-Gene Q system. The amplification efficiency for each primer was measured in each qPCR run. Standard curves for each primer-pair were used to quantify the immunoprecipitated DNA. The ChIP-qPCR signals were normalized using a DIG-DNA-antibody complex [38]. For ChIP-seq, 10 ng of both input DNA and immunoprecipitated DNA were used for Rubicon Thruplex-FD library preparation and sequenced in an Illumina HiSeq2500 instrument to an average depth of 39.6 million single-end 50 bp-reads per sample. Library construction and sequencing were carried out at the Science for Life Laboratory (Stockholm, Sweden). High-quality sequencing tags, as determined by inspection with FastQC, were aligned to the Drosophila melanogaster genome using build BDGP6 with Bowtie 2 [39] allowing for up to two mismatches. The alignment files were indexed using SAMtools [40]. For metagene and heatmap plotting purposes the bam files for two BRM-IP biological replicates were merged into one file and compared to a merge of the two biological replicates of their corresponding Inputs. The ngs.plot package [41] was used to calculate BRM enrichment levels throughout the genome as well as for metagene and heatmap representations. Statistical significance between different metagene distributions was calculated using the Kolmogorov-Smirnov (K-S) test in GraphPad. A fragment size of 200 bp was selected as input parameter to calculate the read coverage, based on the average sonicated DNA fragment size we observed in our Bioanalyzer runs. Average values of BRM enrichment over the genebody region and up to 200 bp upstream of the TSS and downstream of the pA site were calculated. Genes were considered as bound by BRM if the average values were ≥ 1.1 (FC IP/Input). We also used MACS2 [42] to call peaks of BRM binding. Duplicated reads were first removed from the alignment files with the filterdup function, and peaks were called using the IP alignment as Treatment and the Input as the Control with the following settings: bandwidth = 300, pvalue = 0.01, extension size of the fragments = 200, broad region calling = on, cutoff for broad region = 0.1.

Control S2 cells and S2 cells stably transfected with expression plasmids for either recBRM-V5 or recBRM-K804R-V5 were treated with BRM-dsRNA for 48 h to deplete endogenous BRM. The same cell lines were treated in parallel with GFP-dsRNA as a control. Total RNA was isolated using the RNAqueous kit (Ambion) and used to construct poly(A) libraries for massive parallel sequencing on an Illumina HiSeq2000 sequencer, high-output mode, to a depth of 10–15 M paired-end reads per sample (2 × 125 bp). Library construction and sequencing were carried out at the Science for Life Laboratory (Stockholm, Sweden) using NEBNext Ultra Directional RNA library prep kit for Illumina with 1 microgram total RNA as starting material. Two biological replicates were sequenced independently for each cell line. Illumina fastq files were inspected with FastQC to assess the quality of the paired-end reads. RSeQC 2.3.6 [43] and Preseq 2.0 [44] packages were used to check that mapping statistics, read distribution over structural parts of genes, uniformity of gene body coverage, strandedness, and library complexity were all in line with a typical strand-specific RNA-seq experiment. High-quality reads were mapped with TopHat2 [45] to the Drosophila melanogaster genome assembly, build BDGP6 (Dm3). BAM alignments were indexed with SAMtools [40]. Alignments were converted into bigWig tracks with the bamCoverage function implemented in DeepTools [46] and used for visual inspection with the IGV genome browser [47]. Read counts and FPKMs were generated using HTSeq [48] and Cufflinks tools [49], respectively. The Limma package [50] was used for differential gene expression analysis using raw reads from the independent replicates as input. Since correlation analysis of biological replicates showed high reproducibility (Additional file 1: Figure S11), the reads from both experiments were merged and indexed with SAMtools [40] and analyzed together for metagene and heatmap representations.

Gene sequences extending from nucleotide − 50 to − 10 bp upstream of the TSSs of the genes of interest were retrieved using the Regulatory Sequence Analysis Tools (http://rsat.sb-roscoff.fr/↗). Alternative TSSs were also included. We then used an unbiased exhaustive enumeration method to look for hexamers that were more enriched in the sense strand than in the antisense strand [51], using the Biostrings and Stringr packages in the software package R. The total occurrence number of each hexamer was calculated on both sense and antisense strand. The log2 ratio of the total occurrence number on the sense and antisense strands was then calculated and plotted. Sequences between − 300 bp upstream of the TSS and + 100 bp downstream of the TSS were retrieved and used for motif enrichment analysis using CentriMo (http://meme-suite.org/tools/centrimo↗).

ChIP-seq data for endogenous BRM is available in the Gene Expression Omnibus repository with accession number GSE95236↗). GFP_control and dsBRM MNase-seq datasets from Shi et al. [25] were downloaded from ArrayExpress (E-MTAB-1967). RNAPII ChIP and Input files from Lam et al. [52] were downloaded from ArrayExpress (E-MTAB-1084). The antibody used in the RNAPII ChIP-seq experiment was directed against Rpb3 [52]. All sequencing files were processed as described above for our ChIP-seq datasets. Normalized average density plots around the TSS were drawn with the ngs.plot software [41]. The statistical robustness parameter, which filters out 0.5% of genes with the most extreme expression values, was applied to all calculations.

GO classification was done using the PANTHER Classification System [53]. GO enrichment tests were carried out with GOrilla [54] using the 9872 genes detected in the RNA-seq analysis as background list. Probability values were corrected for multiple testing and the significance cutoff was p = 0.05. GO terms that were significantly enriched in the GOrilla nalaysis were summarized and visualized using Revigo [55].

Additional file 1:Figure S1. Comparison of input-corrected enrichment values for RNAPII and BRM. Table S1. List of BRM target genes in S2 cells. Figure S2. Heat map of normalized gene expression of BRM target genes. Figure S3. Gene ontology classification of BRM target genes. Figure S4. Metagene analysis of BRM in the BRM target genes. Figure S5. Metagene analysis of RNAPII and nucleosome occupancies in BRM target genes that are upregulated or downregulated by BRM. Table S2. List of genes differentially expressed in BRM-depleted cells. Figure S6. recBRM-V5 and recBRM-K804R-V5 are incorporated into SWI/SNF complexes. Figure S7. Genomic distribution of recBRM-V5 and recBRM-K804R-V5 in S2 cells analyzed by ChIP-seq. Figure S8. Distributions of recBRM-V5 and recBRM-K804R-V5 in ATPase dependent and ATPase-independent genes. Figure S9. Analysis of TATA-box occurrence in the promoters of the BRM target genes. Table S3. List of enriched motifs in the promoters of BRM target genes. Table S4. Average fpmk values for selected transcription factors. Figure S10. Metagene analysis of RNA polymerase II distribution in the ATPase-dependent and ATPase-independent genes. Table S5. Sequences of PCR primers used for dsRNA synthesis. Figure S11. RNA-seq experiments: correlation analysis of biological replicates. (PDF 4624 kb)