Brg1 controls neurosensory cell fate commitment and differentiation in the mammalian inner ear

Otic ectoderm gives rise to almost all cell types of the inner ear; however, the mechanisms that link transcription factors, chromatin, lineage commitment and differentiation capacity are largely unknown. Here we show that Brg1 chromatin-remodeling factor is required for specifying neurosensory lineage in the otocyst and for inducing hair and supporting cell fates in the cochlear sensory epithelium. Brg1 interacts with the critical neurosensory-specific transcription factors Eya1/Six1, both of which simultaneously interact with BAF60a or BAF60c. Chromatin immunoprecipitation-sequencing (ChIP-seq) and ChIP assays demonstrate Brg1 association with discrete regulatory elements at the Eya1 and Six1 loci. Brg1-deficiency leads to markedly decreased Brg1 binding at these elements and loss of Eya1 and Six1 expression. Furthermore, ChIP-seq reveals Brg1-bound promoter-proximal and distal regions near genes essential for inner ear morphogenesis and cochlear sensory epithelium development. These findings uncover essential functions for chromatin-remodeling in the activation of neurosensory fates during inner ear development.


Introduction
The mammalian inner ear uses sensory hair cells (HCs) for mechanotransduction of vestibular and auditory stimuli and transmits this information to the brain via sensory neurons. Precursor cells for sensory cells and neurons are specified in the epithelium of the otocyst, which develops from the otic placode -thickened ectoderm that forms adjacent to the hindbrain at ~E8.5 and differentiates to form all inner ear structures. Neurosensory domain within the ventral region of the otic ectoderm is defined by transcription factors (TFs) Eya1 and Six1. A subpopulation of their daughter cells is induced to become neuroblasts from ~E9.0-9.25, which is marked by the

Brg1 controls specification of neurosensory progenitors in the otocyst
Brg1 is widely expressed in the otic placode and otocyst and in the delaminating neurons at E8.5-10.5 ( Figure 1A,B). In the developing cochlea, Brg1 is highly expressed in the prosensory primordium marked by Sox2 ( Figure 1C) and in differentiating HCs and SCs within the organ of Corti at P0 ( Figure 1D). Strong Brg1 expression was also observed in LER (lesser epithelial To study its role in neurosensory cell development during inner ear morphogenesis, Brg1 fl/fl mice (Sumi-Ichinose et al., 1997) were bred with Eya1 CreERT2 to generate Eya1 CreERT2 ;Brg1 fl/fl (Brg1 Cko/Cko or CKO) mice with specific deletion of Brg1 in Eya1 + cells. Anti-Brg1 staining confirmed selective ablation in the ventral otocyst where Eya1 is normally expressed (Figure 1figure supplement 1B-E). Brg1 Cko/Cko mice displayed a considerably smaller otocyst (n=6) than in their wild-type littermates and an absence of the VII-VIIIth ganglionic complex ( Figure 1E).
Examination of inner ears at E15.5 showed a degenerated cavity-like structure without neurosensory cells in Brg1 Cko/Cko (n=6, Figure 1F). These data indicate that Brg1 has an essential role in inner ear neurosensory structure formation.
Analysis of EdU (5-ethynyl-2'-deoxyuridine) incorporation (chased for 2 hours before harvest) revealed a reduction in proliferation rate in the mutant to ~57% of that in wild-type littermates (37.9±0.4% vs 66.1±0.3% per otocyst; n=3, Figure 1G,H). TUNEL assay showed increased abnormal cell apoptosis in the mutant (n=3, Figure 1I,J). Thus, the growth arrest and degeneration of the otocyst associated with Brg1-deficiency is due to defective cell proliferation and abnormal apoptosis.
To understand the molecular basis for the otic defects, we performed in situ hybridization (ISH) to characterize genes crucial for neurosensory cell fate specification. Eya1, Six1 and Sox2 are expressed in the ventral region of the otocyst from which all sensory organs develop ( Figure   2A-E), whereas the expression of all three genes was not observed in the CKO littermates (n=6).
Dlx5, which is expressed in the dorsolateral region of the otocyst (n=6, Figure 2F), was expressed in the ventral region of the CKO otocyst. This indicates that in the absence of Eya1-Six1-Sox2-promoted cell fate, the otic epithelium is filled with Dlx5-positive cells, suggesting acquisition of a non-sensory fate or expansion of the non-sensory area.
To determine if lack of VIIIth cochleovestibular ganglion formation is due to lack of neuroblast specification or differentiation, we examined the expression of Neurog1 and Neurod1 by ISH. At ~E9.25, Neurog1 expression in the anterior region of the otocyst was undetectable in Brg1 Cko/Cko littermates ( Figure 2G), but its expression in the Vth ganglion was preserved in the CKO. Neurod1 expression, which is detected in all cranial ganglia at E10.5 ( Figure 2H), was also lost in the VIIIth ganglion of the CKO (n=5). These data suggest that neuroblasts are not specified in the absence of Brg1. In addition, the IXth and Xth ganglia were not formed, while the VIIth and Vth ganglia were reduced in size in the mutant ( Figure 2H). Since this phenotype is similar to that seen in Eya1 -/-;Six1 -/double mutant (Zou et al., 2004), altered Neurog1 and Neurod1 expression could be due to loss of both Eya1 and Six1 expression in Brg1 Cko/Cko . Together, these data suggest that Brg1 activity is required for the expression of Eya1 and Six1 to induce otic ectodermal progenitors to adopt a sensory or neuronal fate.

Sox2 expression depends on both Eya1/Six1 function
We have previously reported that Sox2 expression is detectable in the otocyst of Eya1 -/or Six1 -/single null embryos (Zheng et al., 2003;Zou et al., 2008). Since otic neuroblast specification requires both Eya1/Six1 function, we hypothesized that Eya1/Six1 may also interact to induce Sox2 activation in the otic placode to promote a sensory fate. Indeed, Sox2 expression was not detected in Eya1 -/-;Six1 -/double mutant otic placode and otic cup between E8.5-9.25 ( Figure   2I,J). However, Dlx5 expression was similarly detected in the entire ectodermal region ( Figure   2K,L), differing from its restricted dorsal expression in wild-type littermates ( Figure 2L). In contrast, conditional deletion of Sox2 using Eya1 CreERT2 did not cause a ventral shift of Dlx5 confirming that Eya1/Six1 function upstream of Sox2. Thus, the significant reduction of Eya1/Six1 expression in the Brg1 CKO may in turn cause loss of Sox2 expression. Together, these data indicate that Eya1/Six1 interact synergistically to induce Sox2 expression in early otic progenitors to support the Eya1/Six1/Brg1 mediated prosensory fate specification.

Brg1 specifies the prosensory primordium in the cochlear epithelium
We next investigated if Brg1 has a role in auditory prosensory primordum establishment during cochlear morphogenesis. The cochlear duct elongates from the ventral region of the otocyst and the prosensory progenitors proliferate until reaching a defined number. They then exit the cell cycle to become p27 Kip1+ precursors to form a non-proliferating precursor domain in an apex to base progression between E12.5-E14.5 (Chen et al., 2002). This domain appears as a three-to four-cell layered epithelium marked by Sox2/p27 Kip1 ( Figure 3A,E). In E14.5 Brg1 CKO given tamoxifen (Tam) at E11.75-12.5, Brg1 was selective ablation in the Eya1-expressing domains including the prosensory domain, GER and SGNs (Figure 1-figure supplement 1F,G). Sox2 + progenitors were present along the entire cochlear duct, which was shortened in the CKO ( Figure   3B-D). However, only a few Sox2 + cells in the apex to medial region of the cochlea became p27 Kip1+ ( Figure 3B-D). The prosensory primordium on sections appeared slightly thinner along the basoluminal axis and Sox2 expression was maintained in the precursors in the basal layer but appeared to be downregulated in the cells in the luminal layer ( Figure 3F-H), a process normally associated with HC differentiation. Within the primordial organ of Corti, only cells in the medial region were positive for p27 Kip1 ( Figure 3B,C,F-H). As cycle exit of the progenitors occurs in an apex-to-base and medial-to-lateral direction from E12.5, the few p27 Kip1+ cells observed in the CKO cochlea could be due to the presence of Brg1 activity in those cells when CreERT2 was induced, as it takes at least 6 hrs for the CreER fusion protein to translocate to the cell nucleus and the peak marking occurs over a subsequent period of 12-24 hrs (Danielian et al., 1998).
Consistent with apex to base upregulation of p27 Kip1 , we found only very few p27 Kip1+ cells in the apical tip of the CKO cochlea after Tam injection at ~E11.75 ( Figure. 3D). Thus, Brg1 is necessary to specify the prosensory primordium by regulating p27 Kip1 activation. In the absence of Brg1, Sox2 + progenitors fail to become p27 Kip1+ precursors.
We further investigated whether the arrest of cochlear growth is due to either a decrease in proliferating cells or an increase in apoptosis. EdU incorporation demonstrated a decrease of proliferating Sox2 + cells in the CKO sensory epithelium at E13.5-14.5 after Tam injection at

Differentiation of the postmitotic precursors into HCs or SCs is disrupted in Brg1 Cko/Cko
We next asked whether the p27 Kip1+ or p27 Kip1precursors in the mutant presumptive organ of Corti are competent for HC or SC differentiation. Myo7a is a marker specific for differentiating HCs, while Prox1 is a SC marker specific for pillar cells lying between inner and outer HCs and three Deiters' cells underlying outer HCs (Figure 4-figure supplement 4). Double immunostaining showed neither Myo7a + HCs nor Prox1 + SCs present in Brg1 CKo/Cko cochlea at E18.5 (Tam at E11.75-12.5) (Figure 4-figure supplement 4). Sox2 expression had also disappeared in the mutant by this stage (data not shown). Thus, in the absence of Brg1 the prosensory progenitor cells fail to either exit cell cycle or enter terminal differentiation, eventually leading to loss of Sox2 expression.
We next administered Tam more than a day later, from ~E13.5-14.0, to examine the role of Brg1 in the activation of HC and SC fates during terminal differentiation of the postmitotic precursors, which initiates near the base from ~E14.5. By E18.0, the cochlea had reached a full 1.5 turns, and differentiation of HCs, which are aligned into one row of inner and three rows of outer HCs, had reached the apical tip as outlined by Myo7a ( Figure 4A,B). In Brg1 Cko/Cko , while the cochlea was considerably smaller and only reached ~1 turn, very few HCs that failed to align into the characteristic rows of inner and outer HCs were present in the base, and none were observed in the medial to apex ( Figure 4A,B). Since HC differentiation occurs in a medial-tolateral (inner-to-outer) gradient, the Myo7a + HCs likely represent the inner HCs, which begin to differentiate at least more than a day prior to outer HCs. Indeed, immunostaining for calretinin, a marker specific for inner HCs, confirmed that all Myo7a + HCs in the base were also calretinin + ( Figure 4C). As seen on sections, the organ of Corti was patterned into a two-cell layered epithelium with one-cell layer of HCs on the luminal side and one cell layer of underlying SCs ( Figure 4C). In Brg1 Cko/Cko , no obvious change in the epithelial thickness along the basoluminal axis was observed. However, cells in the presumptive organ of Corti were irregularly patterned and disorganized with morphological alteration in the CKO compared to those in the control littermates ( Figure 4C). Thus, Brg1 is necessary for HC fate induction and sensory epithelium formation in the cochlea.
The SCs maintain high levels of p27 Kip1 and Sox2 through adult stage ( Figures 5A,B).
Although the SCs in the CKO were irregularly shaped and the organ of Corti became slightly narrower compared to that in wild-type controls, they expressed Sox2/p27 Kip1 . Prox1 was expressed in Brg1 CKO SCs in the base but not in the medial to apex, and the Prox1 + cells were irregularly aligned ( Figure 5C). However, the neurotrophin receptor P75 NTR expression in inner pillar cells was not affected in the CKO ( Figure 4D). The glutamate-aspartate transporter (GLAST) expression in inner border and inner phalangeal cells were detectable in the CKO cochlea, but they appeared to be disorganized and dislocated on the luminal layer ( Figure   4E). The calcium-binding protein S100A marks inner HCs and surrounding inner border, inner phalangeal and inner pillar cells, as well as the Deiters' cells underlying the outer HCs ( Figure   5E,F). S100A + cells were present in the CKO, but they were disorganized with more than one cell on the luminal side. The S100A + cells on the luminal side where outer HCs are normally located were also Myo7a + cells in the base, but not in medial cochlear duct ( Figure 5F). From medial to apical end, all S100A + cells on the luminal layer were Myo7acells. In agreement with the calretinin staining ( Figure 4C), the S100A + cells on the luminal layer likely represent the inner HC precursors that failed to differentiate into Myo7a + cells. Nonetheless, our data indicate that Sox2 + p27 Kip1+ SCs in the Brg1 Cko/Cko failed to fully differentiate into Prox1 + SCs, resulting in disorganized sensory epithelium.
In contrast to Sox2 expression in the CKO cochlear sensory epithelium ( Figure 5A,B), the expression of Eya1, which is normally expressed in differentiating hair cells at E17.5-18.5 (Zou et al., 2008), was markedly reduced and only some faint signal was observed in the basal region (arrow, Figure 6A). Six1 is expressed in both hair and supporting cells (Zhang et al., 2017), but its expression in both cell types was almost undetectable in the CKO cochlea ( Figure   6A). In addition, Atoh1 expression was undetectable in the CKO ( Figure 6A). This indicates that in the absence of Eya1/Six1, Sox2 alone is insufficient to induce Atoh1 expression in Sox2 + progenitors to promote a HC fate in vivo.

Eya1-Six1 simultaneously interacts with the BAF60a/c structural subunit
The SWI/SNF complexes contain more than 10 associated BAF factors, which generate a variety of subcomplexes involved in dynamic and tissue-specific interactions with TFs and other chromatin-modifying enzymes. Some BAFs have variants that are selectively used in cell fate commitment (Ho and Crabtree, 2010). Peptides from BAF170, BAF155 and BAF60b were detected in our proteomics analysis of the Eya1 complexes purified from HEK293 cells that constitutively express a multi-tagged-Eya1 (2×HA-3×Flag-Eya1) (J. Xu et al., manuscript in preparation), leading to the hypothesis that Eya1 activity may depend on the activity of such BAFs. We therefore examined whether Eya1 interacts with such BAFs in the cochlea. We have generated the 2×HA-3×Flag-Eya1 knockin mouse line (J. Li et al., manuscript in preparation), which allows us to use anti-Flag or -HA for immunoprecipitating the multi-tagged Eya1 protein complexes. CoIP analyses using nuclear extracts from E13.5 cochleae confirmed physical association of Eya1 with Brg1, BAF170 and BAF155 ( Figure 6B). BAF60 has three variants, a, b and c, and in contrast to our mass spectrometry data showing interaction of Eya1 with BAF60b in 293 cells, Eya1 associated simultaneously with BAF60a/c but not BAF60b in the cochlea ( Figure 6B). Consistent with our previous observation (Buller et al., 2001), Eya1 formed a complex with Six1, which showed similar physical interaction with BAF60a/c but not with BAF60b ( Figure 6-figure supplement 5A). In contrast, Brg1 physically associated with all three BAF60 variants along with BAF170/BAF155 ( Figure 6B). This suggests formation of specific BAF60a/c-based SWI/SNF-Eya1/Six1 complexes.
We next investigated their expression pattern during cochlea development. BAF60a showed higher levels of expression in the prosensory primordium ( Figure 6C Both BAF170/BAF155 were also detected in the prosensory primordium ( Figure 6-figure supplement 5D,E). By E18.5, high levels of BAF170 were maintained in HCs and SCs and in GER as well as in stria vascularis and SGNs ( Figure 6-figure supplement 5F). In contrast, BAF155 was weakly expressed in HCs and SCs but strongly in Henen's cells and LER ( Figure   6-figure supplement 5G). Together, these data provide insight into how distinct BAF subunits might be used to generate a diversity of SWI/SNF complexes during cochlea morphogenesis to promote lineage-specific properties.

Role for Brg1 in global gene activation in the cochlea
To explore Brg1 genome-wide binding profile and its association with regulatory sequences in the prosensory progenitors, we performed ChIP-seq on E13.5 cochleae. MACS (Zhang et al., 2008) peak calling identified 8086 (with a P value threshold of 1e-5) and 1292 (a P value threshold of 1e-6) Brg1 peaks ( Figure 7A). ~28% of Brg1-bound regions were within 5 kb of promoter-5' UTR and ~72% were located within 5-500 kb of TSSs of the nearest gene ( Figure   7B). Analysis of the 8086 Brg1-enriched regions found 3079 (~28% promoter-proximal and To explore whether Brg1-enriched sites contain consensus motifs and to reduce the number of regions for motif analysis, we examined the 1292 peaks for the presence of known motifs. The top three motifs contain consensus sequences for bZip proteins, Zinc finger proteins and SOX factors ( Figure 7C), all of which are known to specifically interact with Brg1containing SWI/SNF and act as Brg1's transcriptional coregulators (King and Klose, 2017;Marathe et al., 2017;Trotter and Archer, 2008). Thus, this analysis provides insight into potential DNA-binding TFs that Brg1 interacts with at gene promoters and distal regulatory elements in the cochlea.
Gene Ontology (GO) analysis showed highly enriched "biological process terms" near genes associated with processes involving development and morphogenesis, including inner ear and cochlea specific processes characteristic of cochlea development ( Figure 7D  ChIP assays confirm Brg1 binding at discrete regulatory regions in the Eya1, Six1, Sox2,

Eya4 and p27 Kip1 loci and co-occupancy by BAF60a/c
At the Eya1 locus, Brg1-bound sites were recovered in promoter (I) and intergenic (II) regions and most of these regions are ECRs (evolutionary conserved regions) based on comparison of mouse with human genomes ( Figure 8A). Since Eya1 expression in the otocyst also requires Brg1 activity, we speculated that Brg1 may bind to the same elements at the Eya1 locus to activate its expression in early otic progenitors and to maintain its expression in cochlear sensory epithelium. ChIP-qPCR assays on otocysts dissected from E10.5 embryos confirmed strong Brg1 enrichment at the promoter (+610 to +429 bp; +2 to -180 bp; -373 to -532 bp) (primer positions, Supplementary file 4) in wild-type samples ( Figure 8B). For the intergenic sites, ChIP-qPCR revealed strong association of Brg1 with an ECR (-913924 to -914124 bp), but very weak enrichment at two non-ECRs (-196849 to -197024 bp and -197523 to -197695 bp) that had relatively weaker Brg1 signals in ChIP-seq (red stars). ChIP-qPCR on Eya1 CreERT2 ; Brg1 fl/fl littermates showed a significant reduction in Brg1-enrichments at these regions ( Figure 8B). To characterize if BAF60a or BAF60c is a component of the SWI/SNF complex binding at the Eya1 locus, we performed ChIP assay with anti-BAF60a or BAF60c respectively. BAF60a showed association to other three Brg1-bound sites (+610, +2 and -193924; Figure 8B), while BAF60c revealed enrichment at all four sites with different affinity levels. Brg1-deficiency led to ~2 to 4-fold reduction in BAF60a or BAF60c enrichment at these sites. The co-occupancy patterns of BAF60a/c on Brg1-bound sites could result from transient binding by separate SWI/SNF components in different subsets of Eya1 + cells. In summary, the ChIP experiments on E10.5 otocysts confirm enrichment of Brg1 at the discrete ECR regions in the Eya1 locus from early stage. Furthermore, our results support that Brg1 is the preferred ATPase subunit of the SWI/SNF complex binding at the discrete regulatory regions of Eya1 to regulate its expression.
The cyclin-dependent kinase inhibitor p27 Kip1 (Cdkn1b) is required to correctly time cell cycle exit of the prosensory progenitors prior to their differentiation and is regulated by Sox2 in cochlear SCs (Liu et al., 2012). Brg1 ChIP-seq recovered strong peaks at the discrete promoterproximal and distal regions in the p27kip1 locus ( Figure 8C). We compared these regions with public Sox2 ChIP-seq data from immortalized Sox2 + otic progenitor cells (Kwan et al., 2015) and found that Sox2-bound sites at the promoter and distal sites overlap with the Brg1-bound regions ( Figure 8C), suggesting that Sox2 may interact with Brg1-BAF complexes to regulate p27 Kip1 expression. Indeed, Sox2 physically interacts with Brg1 in the cochlea by coIP analysis (Figure 6-figure supplement 5A). ChIP assays on E13.5 cochleae confirmed association of Brg1 to the promoter-proximal and distal sites (-519, -277, +1753, +6513, +6983 and +7955) ( Figure   8C,D). Brg1-deficiency (given Tam from E11.5) significantly decreased Brg1 binding at these sites. Thus, reduced p27 Kip1 expression in Brg1 CKO cochlea (Figure 3) is strongly associated with decreased Brg1 binding to the discrete regulatory regions of p27 Kip1 gene. Brg1-deficincy also led to decreased BAF60a or BAF60c enrichment at these Brg1-bound sites. Together, these co-occupancy patterns at the ECRs further support that Brg1-based SWI/SNF complex containing BAF60a or BAF60c is necessary for p27 Kip1 expression through binding to the discrete regulatory regions in the p27Kip1 locus.

Discussion
We have previously shown that Eya1-Six1 interacts with Brg1-BAFs to convert cochlear nonsensory GER cells to SGNs, yet nothing is known about the in vivo requirement of the Brg1-BAF complexes for inner ear neurosensory cell fate commitment and their involvement in regulatory pathways controlling inner ear formation. Here we demonstrate that otic ectoderm requires the chromatin-remodeling enzyme Brg1 to form inner ear structures. The pattern of cochlear Brg1 genomic binding and its strong association with H3K27ac suggest that Brg1 is substantially involved in regulatory pathways governing inner ear development.
Ablation of Brg1 in Eya1-expressing cells blocked the activation of neurosensory cell fate inducing programs. This phenotype is linked to loss of Eya1/Six1 expression, which results in loss of Sox2 and Neurog1 expression. Eya1, Six1 and Sox2 form the core transcription circuitry for neurosensory precursor cell fate specification and subsequent differentiation in the inner ear. However, to date, nothing is known about what lies upstream of these genes in the inner ear. Our findings provide the first in vivo evidence that Brg1-based SWI/SNF chromatin remodeling factors regulate the expression of Eya1, Six1 and Sox2 at different stages of inner ear development. Our data show that Brg1 plays a crucial role in coordinating multiple processes during otic neurosensory cell specification and differentiation.
How do the Brg1-based SWI/SNF complexes help to establish and maintain neurosensory cell identities? SWI/SNF chromatin remodeling complexes are typically recruited to genomic sites via physical interactions with sequence-specific TFs to execute spatiotemporal control over specific target genes during developmental processes (Ho and Crabtree, 2010;Neely et al., 1999). Therefore, Brg1 may differentially impact the spatial and temporal expression of Eya1/Six1 in the otic ectoderm. The otic placode forms from a common pre-placodal ectoderm, which is defined by the members of Eya and Six families. However, unlike Six1/4 and Eya2 that are coexpressed in the pre-placodal ectoderm, the mouse Eya1 is expressed in the otic and epibranchial placodal ectodermal regions (Chen et al., 2009). Thus, Eya1 may have a specific role in defining an otic fate from the pre-placodal ectoderm and the timing of Eya1 activation may require a unique Brg1-SWI/SNF complex in response to external signals. In support of this, ChIP assays on E10.5 otocysts confirmed Brg1 association with the promoter and distal regulatory regions at the Eya1 locus. Future work is required to reveal exactly how the binding of Brg1 to Eya1 locus is related to establishing otic fate. This process likely involves other TFs that are expressed in the ectoderm, during the transition from the pre-placodal stage to Eya1 + multipotent otic progenitors before neurosensory cell fate induction.
Once Eya1 is activated, Brg1-BAFs may alter the structure of chromatin at lineagespecific genes to facilitate the function of Eya1, which interacts with Six1 and other factors. Our observation of a direct physical interaction between Brg1-BAFs and Eya1-Six1 suggests that Eya1-Six1 use the chromatin-remodeling properties of Brg1-BAF complexes to regulate transition from a multipotent otic ectodermal cell to a prosensory restricted progenitor cell by activating lineage-specific genes such as Sox2 or to a committed neuronal fate by activating Neurog1-Neurod1. This explains why ablation of both Eya1/Six1 also leads to a lack of Sox2 ( Figure 2) or Neurog1-Neurod1 (Ahmed et al., 2012b) expression. Since the expression of Eya1/Six1 is markedly reduced in Brg1 CKO, lack of Brg1-SWI/SNF binding at these genes could prevent their activation. This in turn simultaneously prevents expression of Sox2 and Neurog1-Neurod1 and results in a developmental block of neurosensory cell fate specification.
We thus speculate that Brg1 may bind to the regulatory elements at the Eya1 and Six1 loci in order to activate their coexpression before neurosensory fate induction in otic placode and to maintain the coexpression of these genes in otic ectoderm and cochlear sensory epithelium at later stages. Future studies using a combination of genomic and molecular approaches are required to reveal the molecular details of TFs and Brg1-BAF complexes involved in activating Eya1 or Six1 during early otic fate induction.
Our data also show an essential role for Brg1 in both cell survival and proliferation within the otic ectoderm, which is consistent with Brg1's necessity in cell survival and proliferation in various tissues. As Eya1 or Six1 is also required for normal cell proliferation and survival in the otic ectoderm (Zheng et al., 2003;Zou et al., 2006), Brg1 may cooperate with Eya1/Six1 to regulate these processes, which explains why loss of any of these genes causes a decrease in cell number in the otic ectoderm and leads to growth arrest. Ablation of Brg1 in the developing cochlea at later stages underscores the continuous contribution of Brg1-BAF remodelers during prosensory precursor cell differentiation, from the regulation of precursor proliferation and survival to the induction of cell type-specific genes.
In gain-of-function studies, we previously demonstrated the requirement of Brg1 activity and its interaction with Eya1/Six1 in inducing a robust neuronal differentiation program from cochlear nonsensory GER cells or ectodermal cells surrounding the otocyst in explants (Ahmed et al., 2012b). However, it remained unknown whether Brg1 is also necessary for HC fate induction. The present loss-of-function study now demonstrates the necessity of Brg1 in inducing auditory sensory epithelium formation. Brg1 is not only essential for specification of the prosensory domain but also for terminal differentiation of precursor cells within the domain into either HCs or SCs. Conditional deletion of Brg1 between E11.75-12.5 prevented Sox2 + prosensory progenitors from becoming p27 Kip1+ postmitotic precursors. Thus, Brg1-BAF complexes may facilitate regulation of p27 Kip1 expression and other downstream genes by Sox2, thus explaining why Sox2 alone is insufficient to induce p27 Kip1 expression. This is consistent with the physical interaction between Sox2 and Brg1 and Sox2 co-occupancy at a subset of Brg1-bound regions at the p27 Kip1 locus. The necessity of Brg1 activity in p27Kip1 regulation is further supported by the marked reduction in Brg1 enrichment at discrete regulatory ECRs in the p27Kip1 locus in the Brg1 CKO cochlea ( Figure 8D). Interestingly, when Brg1 was deleted at later stages between E13.5-14.0 after p27 kip1 activation, p27 Kip1+ cells also failed to fully differentiate into Myo7a + HCs or Prox1 + SCs. Although this developmental defect may have resulted from a failure to appropriately regulate chromatin structure, it may also be attributed to a deficiency in modification of other proteins that are crucial for HC fate induction or SC differentiation, such as Eya1, Six1, Atoh1 and Prox1, all of which are not expressed in the Brg1 mutant. In the absence of Eya1 and Six1 in Brg1 CKO cochlea, Sox2 alone appears to be incapable of activating Atoh1 or maintaining Prox1 expression in Sox2 + precursors to promote HC fate induction or SC differentiation. Although it is currently unclear whether Eya1-Six1 also use Brg1 activity to facilitate Atoh1 activation and HC induction during cochlea development, in our ChIP-seq data sets, weak enrichment of Brg1 and H3K27ac was observed at the 1.4 kb Atoh1 HC-specific enhancer located ~3.4 kb 3' of the Atoh1 coding sequence (red star, Figure S8). This enhancer mediates transcriptional regulation by Eya1-Six1-Sox2 (Ahmed et al., 2012a). It is possible that Atoh1 may not be activated yet in the cochlear samples collected from E13.5 embryos for our ChIP-seq, as suggested by the weak H3K27ac signals at this enhancer. This is consistent with the observation that Atoh1 activation is induced between E13.5-14.5 (Chen et al., 2002). We are currently performing Brg1 ChIP-seq on cochleae at later stages between E14.5 to E15.5 to examine if Brg1 enrichment at this HC-specific Atoh1 enhancer increases. Stronger Brg1 binding at later stages would suggest that recruitment of Brg1-BAF complex to the 1.4 kb HC-specific enhancer may confer temporal control of HC fate activation by Eya1-Six1-Sox2.
Nonetheless, stronger Brg1 enrichment at the promoter-proximal regions of the Atoh1 suggests that Brg1 may bind to discrete regulatory regions of the Atoh1 to temporally control its initial activation in Sox2 + progenitors and upregulation during HC fate commitment. Brg1-enrichment at the Prox1 locus but not p75 locus further indicates the essential function of Brg1-based chromatin remodeling in the activation of transcriptional programs in distinct subtypes of precursor cells during sensory epithelium development (Figure 7-figure supplement 7). Future work is necessary to elucidate how Brg1 interacts with other factors, including Eya1, Six1 and Sox2, to target its binding sites at the Sox2, p27 Kip1 and Atoh1 loci identified from our ChIP-seq data using a combination of transgenic, genomic and molecular approaches.
Our results suggest that Brg1-BAFs may have cell type-specific and context-dependent roles in inner ear development. During development, distinct BAF subunits interact with specific TFs, recruiting the complex to the sites of action of the corresponding factors to control distinct transcriptional programs (Ho et al., 2009;Lessard et al., 2007;Wang et al., 1996). Simultaneous interactions of Eya1/Six1 with BAF60a/c and BAF170/155 suggests formation of separate SWI/SNF complexes containing distinct BAF subunits in different subsets of Eya1-or Six1expressing cells, in response to differentiation cues, that activate a diverse program of lineagespecific gene expression. This is in an agreement with the differential and overlapping patterns of BAF170/155/60a/c in the cochlear sensory epithelium. As BAF60a or BAF60c acts in recruiting TFs (Chen et al., 2012;Oh et al., 2008) and nuclear receptors (Hsiao et al., 2003) to the SWI/SNF chromatin remodeling complex, the variants a/c of the BAF60 structural subunit may recruit Eya1-Six1 to the SWI/SNF complexes to regulate stage-dependent and cell typespecific programs. In response to signals, the alternative usage of specific SWI/SNF variants by the prosensory progenitors may direct distinct differentiation programs. The selective assembly of SWI/SNF with BAF60c and BAF170 may activate HC-specific genes, while Brg1 along with BAF60a-Eya1 and BAF60a-Six1 may promote both HC and SC gene activation programs during terminal differentiation. The observation of a unique pattern of BAF170 expression in the pillar cells suggests that distinct subtypes of SCs may also require cell type-specific SWI/SNF complexes. We speculate that during otic development and differentiation, otic-specific activators may both recruit and require chromatin-remodeling activities for stable binding to the regulatory regions of otic-and neurosensory-specific genes. Future molecular studies will be required to elucidate if these complexes are recruited by transient interactions with different TFs including Eya1, Six1, and Sox2, and whether association of Eya1-Six1 with BAF60a/c recruits Eya1-Six1 into a Brg1-based SWI/SNF complex that is able to remodel chromatin and activate neurosensory-specific gene transcription.
In summary, our results indicate that Brg1 activity is required for Eya1/Six1 expression and that Eya1-Six1 may use chromatin remodeling properties of the Brg1-BAFs to fine-tune control of neuronal or sensory cell fate determination. Activation of Sox2 in the otic ectoderm requires both Eya1/Six1 function. However, in vivo, Sox2 alone without Brg1 activity or Eya1/Six1 activity is insufficient to induce p27 Kip1 or Atoh1 expression in the Sox2 + progenitors in the cochlea. These findings illustrate how chromatin-remodeling enzymes are intricately involved in the transcriptional control of the master regulators of development and the differentiation of neurosensory cells in the inner ear. Since ChIP-seq failed to recover Brg1 enrichment at Neurog1 or Neurod1 loci, how chromatin remodeling of the Brg1-BAFs intersects with chromatin remodeling of these neuronal cell-specific bHLH TFs remains to be demonstrated. In conclusion, the genome-wide maps of in vivo Brg1 binding sites generated in this study provide a foundation for understanding cochlea-specific genome-wide changes in chromatin structure driven by this key chromatin remodeling factor, which is critical to understanding the role of Brg1 in development and disease. This study provides not only a novel framework for understanding how TFs might direct neurosensory cell commitment and differentiation but also implies possibilities for therapeutic intervention by targeting the SWI/SNF complex in regeneration.

Animals, genotyping and tamoxifen administration
All animal protocols were approved by Animal Care and Use Committee of the Icahn School of Medicine at Mount Sinai (protocol #06-0807).  (Arnold et al., 2011), and Eya1 +/-;Six1 +/- (Zheng et al., 2003) mice were maintained on a 129/Sv and C57BL/6J mixed background at the Icahn School of Medicine at Mount Sinai Animal Facility. Mice were bred using timed mating, and noon on the day of vaginal plug detection was considered as E0.5. For induction of the CreERT2 protein, tamoxifen (Sigma, T5648) was dissolved in corn oil (Sigma, C8267) and administrated (1.5 mg/10 g body weight) by oral gavage. Observed variations among Brg1 mutants is likely due to pre-existing developmental variation between embryos when tamoxifen was given.

Histology, immunostaining, and in situ hybridization (ISH)
Histological examination, whole-mount and section immunostaining and ISH were carried out according to standard procedures. Briefly, inner ears were fixed in 4% paraformaldehyde (PFA) for 1 hr at 4°C, dehydrated, and embedded in wax or OCT. Paraffin or frozen sections were generated at 6 µm. For ISH, tissues were fixed overnight. We used five or six embryos for each genotype at each stage for each probe and the result was consistent in each embryo. Hoechst 3342 was used for nuclear staining.

EdU and TUNEL Assays
The EdU assay was performed using a kit (catalog no. C10640, Life Technologies) following the manufacturer's instructions. EdU was co-injected with tamoxifen at 9 am of E11.5 and embryos were harvested at afternoon of E13.5. EdU was also injected at 9 am of E9.5 embryos following tamoxifen treatment at 9 am of E8.0-8.5 and embryos were harvested 2 hours after EdU injection at noon of E9.5. The TUNEL assay was performed using the Apop Tag kit for in situ apoptosis fluorescein detection (catalog no. NC9815837, Millipore) following the manufacturer's instructions.

Cell counts and calibration
EdU-incorporated cells and Hoechst-stained nuclei in E9.5 otocyst were counted respectively.
EdU-incorporated Sox2 + prosensory progenitors in the floor of the cochlear epithelium were counted in the entire cochlea. Values represent average number of EdU + cells or proliferation rate of EdU + cells/total Hoechst + cells (±standard deviations) per otocyst or EdU + Sox2 + cells (±standard deviations) per section (6 µm) or per cochlea. 15 sections per cochlea and 6 cochleae for each sample were measured. Two-tailed Student's t-test was used for statistical analysis.
For ChIP-qPCR, chromatin derived from embryonic otocysts or from E13.5 cochleae and IgG, anti-Brg1 (ab110641, Abcam), anti-BAF60a antibody (ab224229, Abcam), or anti-BAF60c (ab171075, Abcam) was used for ChIP respectively. The ChIPed DNAs were subjected to quantitative PCR (qPCR) amplification with StepOnePlus PCR system and SYBR green PCR Master Mix kit (4309155, Applied Biosystems). The enrichment fold of IP over mock IP was calculated using the comparative Ct (threshold cycle) method. Data was normalized with inputs and the enrichment of mock IP was considered 1-fold. This experiment was repeated three times and each qPCR was performed in triplicate. The primers used for ChIP-qPCR are listed in Supplementary Table 4. The DNA positions are denoted relative to the transcriptional start site (+1).

Peak calling and gene otology analysis
The ChIP-seq data were first checked for quality using the various metrics generated by FastQC (v0.11.2) (http://www.bioinformatics.babraham.ac.uk/projects/fastqc). Raw sequencing reads were then aligned to the mouse mm10 genome using default settings of Bowtie (v2.2.0) (Langmead et al., 2009). Only uniquely-mapped reads were retained and duplicates were removed. Peak-calling was performed using MACS (v2.1.1) (Zhang et al., 2008) with various pvalue cutoffs as reported in the main text. Motif enrichment analysis was performed using the Homer package (v4.8.3) (Heinz et al., 2010). The peak annotation and gene ontology analysis was performed using GREAT program (McLean et al., 2010) and Panther classification system (Thomas et al., 2003). The conservation between mouse and human genome sequences was analyzed online with ECR Brower (Ovcharenko et al., 2004).

Accession numbers.
The ChIP-seq data reported in this paper were submitted to GEO (GSE119545).                ChIP-seq (Kwan et al., 2015) and from mouse and human. Peak heights indicate degree of sequence homology; pink bars above the peaks denote ECRs; UTR (yellow) correspond thin blue box in IGV view; exons (thick blue boxes). Peak regions corresponding to ECRs (black asterisks) or non-ECRs (red asterisks) were selected for ChIP analysis. See Table S4               and Lfng-involved in Notch signaling, Prox1-a marker specific for supporting cells; Atoh1-critical for hair cell differentiation.