BAF chromatin remodeling complex subunit diversity promotes temporally distinct gene expression programs in cardiogenesis

Chromatin remodeling complexes instruct cellular differentiation and lineage specific transcription. The BRG1/BRM associated factor (BAF) complexes are important for several aspects of differentiation. We show that the catalytic subunit Brg1 has a specific role in cardiac precursors (CPs) to initiate cardiac gene expression programs and repress non-cardiac expression. Using immunoprecipitation with mass spectrometry (IP-MS), we determined the dynamic composition of BAF complexes during mammalian cardiac differentiation, and identified BAF60c (SMARCD3) and BAF170 (SMARCC2) as subunits enriched in CPs and cardiomyocytes (CM). Baf60c and Baf170 co-regulate gene expression with Brg1 in CPs, but in CMs control different gene expression programs, although still promoting a cardiac-specific gene set. BRG1, BAF60, and BAF170 all modulate chromatin accessibility, to either promote accessibility at activated genes, while closing up chromatin at repressed genes. BAF60c and BAF170 are required for proper BAF complex composition and stoichiometry, and promote BRG1 occupancy in CM. Additionally, BAF170 facilitates expulsion of BRG1-containing complexes in the transition from CP to CM. Thus, dynamic interdependent BAF complex subunit assembly modulates chromatin states and thereby directs temporal gene expression programs in cardiogenesis. Significance statement BRG1/BRM associated factors (BAF) form multi-subunit protein complexes that reorganize chromatin and regulate transcription. Specific BAF complex subunits have important roles during cell differentiation and development. We systematically identify BAF subunit composition and find temporal enrichment of subunits during cardiomyocyte differentiation. We find the catalytic subunit BRG1 has important contributions in initiating gene expression programs in cardiac progenitors along with cardiac-enriched subunits BAF60c and BAF170. Both these proteins regulated BAF subunit composition and chromatin accessibility and prevent expression of non-cardiac developmental genes during precursor to cardiomyocyte differentiation. Mechanistically, we find BAF170 destabilizes the BRG1 complex and expels BRG1 from cardiomyocyte-specific genes. Thus, our data shows synergies between diverse BAF subunits in facilitating temporal gene expression programs during cardiogenesis.

4 an embryonic stem cell complex that regulates pluripotency (9, 10), or a neural precursor and neural BAF complexes, that fine-tune neurogenesis (11,12) . Current evidence indicates that a shift in isoforms of non-essential BAF complex subunits is important for the stepwise transition from a precursor to a differentiated state in neural and muscle differentiation (11, 13-16).
Mutations in genes encoding BAF complex subunits have been associated with various cancers (17), and some have also been found to be mutated in cases of congenital heart disease (CHDs) (18). In the developing heart, subunits of the BAF complex are involved in diverse aspects of cardiac development (19,20). Brg1 is haploinsufficient in the mouse heart, and genetically interacts with genes encoding DNA-binding transcription factors associated with CHDs, indicating a potentially general importance of BAF complexes in these common birth defects (21).
Differentiation is thought to proceed during development as a continuous but highly regulated series of milestones, which include lineage decisions and linear progression towards a terminally differentiated state. These sequential events can be modeled effectively using pluripotent cells that are subjected to well-defined differentiation protocols (22, 23) . Cardiac differentiation is composed of a stereotyped set of steps, with the initial formation of cardiogenic mesoderm, subsequent specialization into multipotent cardiac precursors, and then to their differentiation into beating cardiomyocytes (24-27). The in vivo embryonic steps are well recapitulated in in vitro differentiation protocols (25, 26). It is not known which BAF complex subunits are essential for controlling temporal steps in cardiac differentiation.
Here, we define dynamic BAF complex composition during mouse cardiomyocyte differentiation.
We identify the BAF subunits BAF60c and BAF170 as enriched in cardiac precursors and cardiomyocytes. BRG1 initiates cardiac gene expression programs in precursor cells, a role shared by BAF60c and BAF170, which also maintain the cardiac program to facilitate cardiomyocyte differentiation. BAF60c and BAF170 also regulate BAF complex composition, stoichiometry and chromatin accessibility. Further, we find that BAF170 destabilizes and facilitates dissociation of BRG1 complexes from their binding sites upon cardiomyocyte differentiation. These results reveal the instructive nature that specific combination of BAF subunits attain to dictate functional outcomes during lineage commitment and differentiation.

Brg1 initiates cardiac gene expression programs during cardiomyocyte differentiation
To model early heart development, we differentiated mouse embryonic stem cells (mESCs) to cardiac troponin T positive (cTnT+), beating cardiac myocytes (25, 26). To understand the role of the BAF complex ATPase BRG1 during cardiomyocyte differentiation, we conditionally deleted Brg1 (also known as Smarca4) using an inducible Cre-loxP system ( Fig 1A) (9,28).
Loss of Brg1 at CP, but not CM, inhibited cardiomyocyte differentiation (Fig. 1B). RNAseq revealed that Brg1 regulated a total of 545 genes in CP and 125 genes in CM (p <0.05, ± 1.5fold). Reduced importance of Brg1 at the cardiomyocyte stage is consistent with its reduced expression ( Fig. 1C and S1A). In CPs, Brg1 repressed 197 (36.1%) and activated 348 (63.9%) genes (Fig. 1C). BRG1 activated genes were enriched for sarcomere organization and assembly and are essential components of cardiac cell fate establishment (Fig. 1D). Of note, Brg1 is essential for the immediate activation of these lineage-specific genes, in anticipation of the final differentiation status of the cells, but not their maintenance. Thus, Brg1 primes the celltype specific differentiation of cardiac precursors. In CMs, although BRG1 activated and repressed roughly equal number of genes, it did not enrich for any biological processes, consistent with in vivo data (29).

BRG1 binding correlates with sites of cardiac transcription factor binding 6
To distinguish between direct vs indirect role of BRG1 in gene regulation, we performed ChIPseq using a native antibody against BRG1. In CPs, we found 4791 significant BRG1 binding regions (Figs. 1E and S1B) and could detect weaker BRG1 occupancy in CMs (Fig.   S1C), presumably due to the low levels of protein at this stage. In CPs, BRG1 was bound to both transcriptional start sites (TSS) and H3K27ac marked active enhancer regions (26) (Figs. 1F and S1D) and associated with 5274 genes flanking BRG1 binding sites (within 1Mb) ( Fig   1G). 21.9% of BRG1-activated and 20.9% of the BRG1-repressed genes overlapped with genes nearest to a BRG1 binding site. Putative BRG1-activated direct targets enriched for cardiac muscle development, contraction and circulatory system processes whereas BRG1-repressed direct targets enriched for embryonic limb and skeletal system development (Fig. S1E). These results show that BRG1 facilitates of cardiac gene expression programs while preventing other developmental programs including embryonic limb development.
Motif enrichment analysis using HOMER (30) revealed GATA motifs near BRG1 activated sites and T-box motifs near BRG1 repressed sites among others (Fig. 1H). Consistently, BRG1 binding sites correlated well with GATA4 and TBX5 binding sites in cardiac precursors (Fig. 1I) (31), suggesting that BRG1 interacts with cardiac TFs to regulate gene expression during cardiac lineage commitment, as predicted from gain of function experiments (32, 33). Thus, BRG1 (and BRG1 containing complexes), in collaboration with cardiac transcription factors, direct cardiac gene expression program while preventing expression of non-cardiac genes.

BRG1 complex shows dynamic composition during cardiomyocyte differentiation
To understand the composition of BRG1-associated complexes during cardiac differentiation, we immunoprecipitated BRG1 under stringent conditions using a BRG1-3x FLAG heterozygous mouse embryonic stem (ES) cell line (34), from five different stages of cardiac differentiation: ES cells (ESC), embroid bodies (EB), mesoderm (MES), cardiac progenitor (CP) and cardiac myocytes (CM) ( Fig. 2A). An untagged control cell line similarly processed in biological triplicates at each stage served as negative control. BRG1 complexes isolated from each of these stages showed overall protein profiles similar previously reported complexes ( Fig. 2A) (10). We performed mass-spectrometry in biological triplicates and technical duplicates and compared peptide intensities after normalizing against untagged control, the bait BRG1 protein, and across stages of differentiation, and identified the composition of the BRG1 complexes ( Fig.   2B and Table S1 (Fig. 2E). However, most subunits were not developmentally regulated, and thus the assembly reflects developmental stage-specific inclusion of these subunits ( Fig   2F). For example, WDR5 is expressed mostly in ESC, while it associates with BAF complexes only in CMs.
To further understand the composition of cardiac enriched complexes, we immunoprecipitated endogenous BAF170-3xFLAG, at five different stages of cardiac differentiation in biological triplicates, again with an untagged control line for each stage, and analyzed immunoprecipitated proteins by mass spectrometry. Most of the proteins enriched at the CP and CM stages using BRG1 as bait were also enriched using BAF170 as bait, with certain exceptions (Fig. 2G). For example, ARID1b is enriched in the BRG1 IP-MS at the MES-CM stages, while in the BAF170 IP-MS it is depleted. The opposite dynamic enrichment is seen for CRABP2. WDR5 was also present in the complexes isolated by BAF170-IP. In addition, we detected the alternate ATPase BRM in the BAF170-FLAG purification, indicating that BAF170 functions within separate BRG1 and BRM complexes. These results suggest that dynamic subunit composition and subunit switch are important aspects of BAF complexes during cardiac differentiation.

BAF170 and BAF60c facilitate cardiomyocyte differentiation
We investigated the functional roles of cardiac BAF enriched subunits BAF60c and BAF170 by deleting them in ES cells using TALEN or CRISPR strategies (Fig 3A). Both BAF60c KO and BAF170 KO cells underwent cardiac differentiation as observed by beating cardiac myocytes and immunostaining of cardiac Troponin T (Fig. 3B). Cells lacking BAF170 however had a delay in onset of beating ( Fig 3C) and both BAF60 and BAF170 KO cells displayed aberrations in beating (Fig 3D), indicative of abnormal cardiac differentiation. To understand how these subunits regulate gene expression we collected cells at CP and CM stages and performed RNAseq. In CPs, BAF60c regulated a total of 474 genes (p-value <0.05, ± 1.5-fold) of which 45% were activated and 55% were repressed, while a total of 382 genes are deregulated in BAF170 KO, of which roughly equal number of genes were up or down regulated (Fig S3A).
Gene ontology analyses revealed that BRG1 and BAF60c share common functions in activating genes involved in cardiac and muscle tissue development and contraction while repressing skeletal system and limb morphogenesis (Fig S3C). Genes regulated (largely repressed) by BAF170 are mostly involved in skeletal system morphogenesis and pattern specification (Fig.   S3C). However, in CMs, BAF60C regulated a large number of genes in cardiomyocytes (2646), repressing 72% of genes and activating 28% of these genes (Fig S3B). Similarly, BAF170 also repressed a large percentage (63%) of genes in CMs (Fig. S3B). These results indicate a largely repressive function of these cardiac enriched subunits in gene regulation.
Overall comparison of the gene expression changes in CPs and CMs lacking BRG1, BAF60c, or BAF170, show that all three have shared functions in CP gene expression, but that in CPs and CMs BAF60c and BAF170 repress a cohort of genes independent of BRG1 (Figs. 3E-G). This clearly indicates divergent role of these two subunits from BRG1, suggesting their association in a different BAF complex or potential independent function outside of BAF complexes.

BAF subunits modulate temporal chromatin accessibility and facilitate cardiac gene expression programs
To understand the function of BRG1, BAF60c and BAF170 in gene expression regulation, we used ATACseq (37) to examine chromatin accessibility genome-wide in CPs and CMs. We  Fig 4F) and calcium ion transport (Cacna1c, Cacna2d4). Uniquely they also regulated chromatin near signaling and cell migration genes ( Fig 4H). Further, in absence of BAF170 we observed delayed accessibility of the Tnnt2 promoter ( Fig 4C) consistent with delayed onset of beating ( Fig 3C), and impaired accessibility at an enhancer in the Scn10a gene ( Fig 4D). This Scn10a region is a known TBX5-dependent enhancer of Scn5a, and was identified as containing a GWAS SNP associated with altered electrophysiology in humans (38, 39). These data provide evidence of the unique and shared functions that individual BAF subunits exert to regulate chromatin structure to facilitate a cardiac gene expression program while maintaining repression of non-cardiac developmental genes. These results suggest that a fine balance exists in the composition of the BAF complex and perturbation of one subunit extends to the stoichiometry of association of other subunits in the complex. It further indicates that different subunits or isoforms substitute for the absence of one or more subunits. For example, both BAF60a and BAF60b substitute for the lack of BAF60c, and depletion of BAF45c is balanced by enrichment of BAF45a (Fig. S4B). These results emphasize that subunit switching and substitution in BAF complex composition could be an important mechanism in cardiac lineage specification.

BAF170 facilitates temporal BRG1 dissociation from the genome
We explored the possibility that BAF60c or BAF170 help direct the genomic localization of BRG1. In CPs, BRG1 binds to a set of 4791 sites (Fig. S1B) and these are largely unaffected in absence of BAF170 (Fig. 5D). BRG1 binding at these site is severely reduced in cardiomyocytes in WT and cells lacking BAF60c, presumably due to the reduced predominance of a BRG1-containing complex. However, in the absence of BAF170, BRG1 binding at a large subset of these sites is retained in cardiomyocytes (Fig. 5D), consistent with the stabilization of BRG1-containing complexes. These chromatin regions are accessible in CPs, and normally subsequently inaccessible in CMs, but remain accessible in absence of BAF170, consistent with retained BRG1 complexes remodeling these regions (Fig. 5F). In CMs, BRG1 weakly bound to 3473 sites, and lack of BAF170 increased BRG1 binding to some of these ( Fig 5E), while the absence of BAF60c reduced the binding BRG1. These binding dynamics correlated with chromatin accessibility (Fig 5G). This suggests that recruitment of BRG1-inclusive complexes is regulated by cardiac specific subunits, and that concomitant dynamic expulsion of the complex is also highly regulated by similar processes.

DISCUSSION
BRG1 containing complexes respond to and modulate lineage decisions and differentiation in cardiac development by incorporating specific subunits at discrete stages. These specialized BAF complexes regulate distinct gene expression programs to drive cardiogenesis while simultaneously repressing alternate non-cardiac developmental programs. Our results suggest that BAF complexes use multiple interdependent mechanisms including switching subunits or isoforms, regulating subunit stoichiometry, altering BRG1 recruitment, and expulsion strategies, 13 to modulate dynamics of cardiac differentiation. This unanticipated complexity of chromatin complex regulation emerges as a critical determinant of differentiation.
Subunit composition switching of chromatin remodeling complexes have been reported during neural development (11, 40, 41) and skeletal myogenesis (15, 16). For example, during myogenic differentiation, BAF60c expression is facilitated over the alternative isoforms BAF60a and BAF60b, which are post-transcriptionally repressed by myogenic microRNAs. However, the systematic IP-MS-based discovery of BAF/PBAF complex subunits during the time course of a well-regulated differentiation process has to date not been achieved. Importantly, we show that BRG1-associated subunits at different stages of cardiac differentiation form specific complexes to activate or repress specific transcriptional programs. In addition, we determine that the presence of specific BAF subunits greatly dictates the overall composition, and therefore function, of the complex.
The continued roles played by BAF60c and BAF170 are intriguing considering the reduced role of BRG1 in later stages of cardiac differentiation. The incorporation of these subunits within a BRM-containing complex may explain this, although the apparent absence of a developmental or postnatal cardiac phenotype in BRM null mice might indicate minimal importance of these complexes (42). It has been suggested that BAF60c may function independently from the BAF complex (43, 44), although there is no evidence for this in cardiac cells. It is also possible that the altered BAF complex assembly in the absence of BAF60c or BAF170 creates a set of complexes with anomalous function, thus leading to aberrant gene expression.
BAF60c and BAF170 are essential for BRG1 complex composition and stoichiometry. Altered association or dissociation of several subunits in the absence of BAF60c or BAF170 indicates the presence of sub-modules within BAF complex mediated by these subunits. Our observations are consistent with reports in yeast SWI/SNF complexes where absence of specific subunits form sub-modules or altered SWI/SNF complex composition (45-47). Thus, phenotypic observations in absence of a particular subunit may be the result of disruption of a module containing many subunits. Conflicting evidence regarding this mechanism has been shown for other BAF complex subunits in mammalian cells (48, 49). We interpret our results as indicating that a stable complex with broadly altered composition exists in the absence of BAF60c or BAF170. BRG1 binding to genomic sites in CP and putative BRG1 direct targets have critical roles in promotion of cardiac gene expression programs while repressing non-cardiac fate. BRG1 binding is greatly reduced in CMs indicating diminished BRG1 function. Our finding that BAF170 largely helps BRG1 dissociation from the genome indicates a potential mechanism of BAF170 mediated gene regulation during differentiation. How BAF170 regulates genomic expulsion of BRG1 is currently not understood, but association of BAF170 with BRM containing complexes in CM might evict BRG1 from BAF complexes as both BRG1 and BRM are mutually exclusive.
Further, concomitant increase in BAF170 and decrease of BAF155 association with BRG1 containing complexes during cardiac differentiation might make BRG1 prone to posttranscriptional modification and proteasome mediated degradation, as BAF155 is known to stabilize BRG1 containing complexes (50). Consistently, we observe BRG1 complex stabilization in absence of BAF170 in CMs.
In conclusion, our study identifies morphing combinations of very specific BAF subunits that      (G) ATACseq signal over 4791 BRG1 binding sites in CP and CM for the indicated genotypes.
(H) ATACseq signal over 3473 CM BRG1 binding sites over indicated genotypes in CM.

Cardiomyocyte differentiation
Mouse embryonic stem cells (ESCs) were cultured in feeder free condition in serum and leukemia inhibitoty factor containing medium. Cardiomyocyte were differentiation as described

Nuclear extract preparation, western blot and anti-FLAG immunoaffinity purification
Nuclear extract were prepared using protocol describe before (56).

Masspectrometry
Protein samples were denatured, reduced, and alkylated before overnight trypsinization at 37° C.
Then, samples were concentrated, desalted, dried and resuspended in 0.1% formic acid for mass spectrometry analysis.

RNA-Sequencing
Total RNA were isolated from biologically triplicate samples using miRNeasy micro kit with on- were lysed, washed and tagmented for 45 mins and 3hr for CP and CM cells respectively. DNA was purified, and amplified using universal Ad1 and barcoded reverse primers (37). Libraries were purified, quantified and analyzed on bioanalyser and sequenced on a NEB NextSeq 550 sequencer using Illumina NextSeq 500/550 High Output v2 kit (150 cycles). Sequencing image 29 files were de-multiplexed and fastq generated. Reads (paird-end 75bp) were trimmed and aligned to mouse genome mm9 assembly using Bowtie (66) with a minimum mapping quality score of 30. Statistically enriched bins with a P-value threshold set to 1x10 6 were determined (67). UCSC genome browser and IGV were used to view the browser tracks. Galaxy (https://usegalaxy.org/) was used to pool multiple replicates to generate browser tracks and tornado plots. GREAT was used to generate gene lists near ATACseq sites.