Anteroposterior Limb Skeletal Patterning Requires the Bifunctional Action of SWI/SNF Chromatin Remodeling Complex in Hedgehog Pathway

Graded Sonic hedgehog (Shh) signaling governs vertebrate limb skeletal patterning along the anteroposterior (AP) axis by regulating the activity of bifunctional Gli transcriptional regulators. The genetic networks involved in this patterning are well defined, however, the epigenetic control of the process by chromatin remodelers remains unknown. Here, we report that the SWI/SNF chromatin remodeling complex is essential for Shh-driven limb AP patterning. Specific inactivation of Srg3/mBaf155, a core subunit of the remodeling complex, in developing limb buds hampered the transcriptional upregulation of Shh/Gli target genes, including the Shh receptor Ptch1 and its downstream effector Gli1 in the posterior limb bud. In addition, Srg3 deficiency induced ectopic activation of the Hedgehog (Hh) pathway in the anterior mesenchyme, resulting in loss of progressive asymmetry. These defects in the Hh pathway accompanied aberrant BMP activity and disruption of chondrogenic differentiation in zeugopod and autopod primordia. Notably, our data revealed that dual control of the Hh pathway by the SWI/SNF complex is essential for spatiotemporal transcriptional regulation of the BMP antagonist Gremlin1, which affects the onset of chondrogenesis. This study uncovers the bifunctional role of the SWI/SNF complex in the Hh pathway to determine the fate of AP skeletal progenitors.


Introduction
Vertebrate limb anteroposterior (AP) patterning is controlled by a diffusible morphogen, Sonic hedgehog (Shh), that is produced from the posteriorly located zone of polarizing activity (ZPA) [1]. Cell fate marking studies on mouse limb buds have revealed that Shh signaling regulates identities of limb skeletal elements, such as the ulna and digits 2 to 5, depending on the signal concentration and time of exposure to that signal [2][3][4]. During limb bud outgrowth, Shh promotes FGF signaling in the apical ectodermal ridge (AER) by mediating the BMP antagonist Gremlin1 (Grem1) that maintains low BMP activity [5].
In vertebrates, binding of Shh to its receptor Patched1 (Ptch1) enables the signal transduction through derepression of signal transducer Smoothened, allowing Gli transcription factors (Gli1−3) to function as activators (GliA) [6]. The transcriptional upregulation of Ptch1 serves as a sensitive readout of Shh activity and is required for sequestering diffusible ligands to restrain their spread within the target range [7,8]. Notably, the spatiotemporal regulation of Ptch1 expression is important to prevent aberrant activation of Hedgehog (Hh) signaling, indicating that Ptch1 functions as a negative regulator of Hh signaling [9,10]. Meanwhile, the fulllength activators Gli2A and Gli3A contribute to the activation of Shh target genes such as Gli1, which might act as an indicator of the Shh signaling range in limb development [11][12][13]. The absence of Shh signaling allows proteolytic processing of bifunctional Gli2 and Gli3 to form the truncated repressors Gli2R and Gli3R (GliR) [14,15].
Gli3 functions as a major regulator of AP digit patterning, whereas Gli2 has compensatory roles of Gli3 activity [4,[16][17][18]. During early limb bud development, Gli3 is required to establish AP polarity through mutual antagonism with Hand2 and is involved in the formation of two signaling centers, the ZPA and AER, by restraining GliA activity [10,[19][20][21]. In addition, constitutive Gli3 expression during anterior digit patterning is mediated by repressing cellcycle genes implicated in the proliferative expansion of Shh-dependent mesenchymal progenitors and by terminating Grem1 expression to initiate chondrogenic differentiation [22,23].
Despite recent progress in identifying networks of trans-acting regulators interacting with multiple cis-regulatory modules (CRM) that orchestrate limb development, epigenetic control of the developmental process, especially the role of chromatin remodelers, remains poorly understood. The mammalian SWI/SNF chromatin remodeling complex is an ATP-dependent chromatin remodeler that uses the energy of ATP hydrolysis to alter nucleosomal structure [24]. The SWI/SNF complex is a multisubunit complex including core factors such as ATPase Brg1, tumor suppressor Snf5, and scaffolding subunit Srg3/mBaf155 (hereafter referred to as Srg3) [25]. In differentiation pathways, SWI/SNF complexes cooperate with histone-modifying factors and transcriptional regulators to mediate both transcriptional activation and repression in response to extracellular stimuli [26].
Here, we show that the SWI/SNF complex is essential for limb AP skeletal patterning. Specific inactivation of limb mesenchymal Srg3, resulting in defects in SWI/SNF complex activity [27], fails to upregulate posterior Shh/Gli target gene expression and induces the ectopic activation of target genes in the anterior limb bud after intact establishment of the ZPA. The SWI/ SNF complex-mediated modulation of Shh responsiveness and repression of the ectopic Hh pathway determine the AP identities of limb progenitors and regulate the spatiotemporal expression of Grem1. Thus, bifunctional action of the SWI/SNF complex in the Hh pathway is essential to pattern AP limb skeletal elements.

Results
Srg3 is essential for anteroposterior limb skeletal patterning To study the specific function of the SWI/SNF complex in developing limb buds, we used a conditional loss-of-function allele of Srg3 (Srg3 f/f ) [28] and a Prx1Cre transgene encoding a Cre recombinase that is activated in the early limb bud mesenchyme [29]. Prx1Cre-mediated inactivation of Srg3 in the limb bud mesenchyme was confirmed by measuring the expression of the transcript and protein in control and Srg3 f/f ;Prx1Cre (hereafter shortened as Srg3 CKO) limb buds. Whole-mount RNA in situ hybridization showed the specific clearance of Srg3 transcripts throughout the mesenchyme and western blot analysis confirmed the downregulation of Srg3 proteins with a time lapse between the fore-and hindlimb buds (S1A and S1B Fig). In addition, the downregulation of Brg1 observed in Srg3 CKO limb buds revealed the structural function of Srg3 that stabilizes the SWI/SNF complex (S1B Fig) [27].
Skeletal analysis of Srg3 CKO limbs at birth (P0) revealed the requirement of Srg3 for limb development (Fig 1). In Srg3 CKO forelimbs, the scapula developed poorly with bifurcated or enlarged foramen, aplastic clavicle, stylopod (humerus) lacking deltoid tuberosity, and radial agenesis were observed (Fig 1A and 1B). In Srg3 CKO hindlimbs, the proximal skeletons (pelvic girdle and femur) were retained normally, whereas zeugopod elements (tibia and fibula) were shortened to a similar extent (Fig 1C and 1D and S1C Fig). Both Srg3 CKO fore-and hindlimbs had rudimentary digits that were connected by ossified tissues in the anterior digital tips (syndactyly) and exhibited more severe ossification defects in anterior digits than those in posterior digits (Fig 1B and 1D and S1D Fig). Unlike predominant preaxial polydactyly in Srg3 CKO hindlimbs, digit number was variable in Srg3 CKO forelimbs (4 or less, 28%; 5, 34%; 6 or more, 38%, n = 84) ( Fig 1E). The discrepancy in severity between fore-and hindlimbs lacking Srg3 is a likely consequence of Srg3 deficiency mediated by the onset timing of Prx1Cre activity, which is first activated in the prospective forelimb bud prior to hindlimb budding [29]. Taken together, the malformation of zeugopod elements and variable digit numbers observed in Srg3-deficienct limbs suggest that mesenchymal Srg3 is involved in AP limb skeletal patterning.
Srg3 CKO forelimb buds establish distinct Hh pathways in the anterior and posterior mesenchyme Given that limb bud development requires formation of the ZPA and AER [5], we first analyzed the formation of ZPA and AER signaling centers at early stages. In E10 Srg3 CKO forelimb buds, ZPA-Shh expression levels was similar with control expression levels (n = 8 limb buds analyzed), whereas AER-Fgf8 expression was slightly reduced in Srg3 CKO forelimb buds relative to controls (n = 6) (S2A Fig). Although Srg3 inactivation did not significantly alter the formation of signaling centers, subtle changes in the AER suggest that the SWI/SNF complex functions in initial limb development. To understand the mechanism underlying Srg3-mediated limb AP patterning controlled by the counteraction of Shh and Gli3 [16,17], we examined the expression of Shh/Gli target genes. In Srg3 CKO forelimb buds, the expression domains of Gli1 and Ptch1 were normal up to at least E10 (Gli1, n = 12; Ptch1, n = 8), but were ectopically activated at E10.25 and at E10.75, respectively, in the anterior mesenchyme (Gli1 and Ptch1, n = 6) (Fig 2A and 2B). In addition, Gli1 and Ptch1 expression was activated in a graded manner along the AP axis in control forelimb buds, whereas their expression domains including ectopic regions were confined to the distal region in Srg3 CKO forelimb buds over time (Fig 2A and 2B; Gli1, n = 5; Ptch1, n = 6). Importantly, Ptch1 transcripts were not detected in the core mesenchyme of Srg3 CKO forelimb buds (Fig 2B). Gli1 was ectopically activated from around E11 in Srg3 CKO hindlimb buds, but its expression was comparable to control hindlimb buds in the posterior region (S2B Fig). These data suggest that Srg3 both activates and represses Shh/Gli target gene expression in distinct regions.
To define whether bifunctional action of Srg3 in the Hh pathway affects the interlinked signaling between the ZPA and the AER [30], we examined the expression pattern of epithelialmesenchymal signaling genes during limb bud outgrowth. In Srg3 CKO forelimb buds, the size shows another scapula phenotype. Red arrowheads denote hypoplastic scapulae and black arrowheads indicate the loss of clavicle, deltoid tuberosity, and radius in the Srg3 CKO forelimb. The insets in (C) and (D) show a dorsal view of a hindlimb autopod marked with digit numbers. Red arrows point to the fused digits with soft tissues. cv, clavicle; dt, deltoid tuberosity; fe, femur; fi, fibula; hu, humerus; pg, pelvic girdle; r, radius; sc, scapula; ti, tibia; u, ulna; 1−5, digit identity. of the Shh expression domain was subtly reduced (Fig 2C; n = 6). Grem1 expression expanded anteriorly (n = 7), whereas Bmp4 expression was reduced in the anterior and posterior mesenchyme of Srg3 CKO forelimb buds (n = 4) (Fig 2D and 2E). AER-Fgf4 expression shifted anteriorly in Srg3 CKO forelimb buds (Fig 2F; n = 6). Taken together, these data suggest that distinct  Hh pathways established by Srg3 deficiency differentially impacted epithelial-mesenchymal signaling in the anterior and posterior mesenchyme.
The polarization of nascent limb mesenchyme and establishment of the ZPA are controlled by antagonistic interactions between Hand2 and Gli3 in the posterior and anterior regions, respectively [19][20][21]. To assess whether Srg3 deficiency in the limb bud mesenchyme affects AP polarity at the prepatterning stage, we examined the expression domains of Hand2 and Gli3 at E9.5. Consistent with the formation of an intact ZPA up to at least E10 (S2A Fig), the expression domains of these positional markers remained comparable to controls in Srg3 CKO forelimb buds (Fig 2G and 2H; Hand2, n = 8; Gli3, n = 9). During limb bud outgrowth, the distribution of posterior markers Hand2 and Hoxd13 was more posteriorly restricted or reduced in Srg3 CKO forelimb buds than in control limb buds, whereas their expression was activated in the anterior region at E10.75 (Fig 2I and 2J; Hand2, n = 7; Hoxd13, n = 5). By contrast, the expression domains of anterior markers Alx4 and Pax9 exhibited progressive decreases in Srg3 CKO forelimb buds (Fig 2K and S3A Figs; Alx4, n = 6; Pax9, n = 6) [31]. Consistently, the expression of anterior markers was mildly downregulated in Srg3 CKO hindlimb buds (S3B Fig), suggesting that the loss of anterior identity in Srg3-deficient limb buds correlates with the timing of Srg3 inactivation. Taken together, these data indicate that Srg3 deficiency progressively decreased the AP identities of limb progenitors, leading to a disruption of asymmetry after early specification of the AP axis.
Mesenchymal Srg3 deficiency induces ectopic Shh expression and distalizes epithelial-mesenchymal signaling at late stages Inactivation of Srg3 in the limb bud mesenchyme caused progressive alterations in Shh/Gli target gene expression and in AP identity (Fig 2). To gain further insights into the regulation of Shh/Gli target genes by the SWI/SNF complex, we reexamined the distribution of epithelial-mesenchymal signaling genes at subsequent stages. In Srg3 CKO limb buds, Shh expression was ectopically induced in the anterior margin and subsequently expanded along the distal margin ( Fig 3A and  S4A Fig; n = 6 per stage). Ectopic Shh signaling reduced Gli3R protein levels by inhibiting Gli3 processing in the anterior mesenchyme of Srg3 CKO forelimb buds ( Fig 3B). To test whether the SWI/SNF complex is directly implicated in repressing Shh through the regulation of limb-specific Shh enhancer ZRS (ZPA regulatory sequence), which is responsible for localized expression of Shh [19,32], we performed a chromatin immunoprecipitation (ChIP) assay. We did not observe the enrichment of Srg3 at any regions on the ZRS (S4B Fig). This suggests that ectopic Shh expression is indirectly induced in Srg3 CKO limb buds.
After ectopic Shh expression was activated, the anteriorly expanded domain of Grem1 at E10.5 was divided into two parts: the anterior domain and the posterior domain (Figs 2D and 3C; n = 7). In E11.5 Srg3 CKO forelimb buds, the derepressed expression of Grem1 in the anterior was remarkably reduced in the distal mesenchyme, whereas its posterior domain was distally shifted ( Fig 3C). As the posterior domain of Grem1 closer to the AER reflects loss of FGF signaling repressing Grem1 Srg3-containing SWI/SNF complexes are required for the transcriptional activation and repression of Gli1 and Ptch1 in developing limb buds To verify whether Srg3 directly regulates the expression of Shh/Gli target genes in developing limbs, we examined the effects of Srg3 deficiency by transducing a Cre-expressing retroviral vector into Srg3 f/f mouse embryonic fibroblasts (MEFs). We focused our analyses on the transcriptional regulation of Shh/Gli target genes Gli1 and Ptch1. Quantitative real-time PCR (qPCR) showed that Srg3-deficient MEFs expressed higher levels of Gli1 and Ptch1, suggesting that the SWI/SNF complex represses Shh/Gli target genes ( Fig 4A). To exclude the possibility that the SWI/SNF complex indirectly represses Shh/Gli target genes by other factors in the MEFs, we treated Srg3-deficient MEFs with the Hh pathway inhibitor cyclopamine [34]. Although cyclopamine reduced Gli1 and Ptch1 expression in control and Srg3-deficient MEFs, Srg3-deficient MEFs expressed higher levels of Gli1 and Ptch1 than control MEFs ( Fig 4A). This indicates that the Srg3-containing SWI/SNF complex represses Shh/Gli target genes in, at least, a Hh-free condition. Thus, this finding could corroborate the derepression of Shh/Gli target genes in the anterior mesenchyme of Srg3 CKO limb buds. Next, we examined whether Srg3 is involved in the activation of Gli1 and Ptch1 expression upon Shh stimulation. In the presence of Shh-conditioned medium, Srg3-deficient MEFs displayed severely reduced activation levels of Gli1 (101vs. 16.1-fold) and Ptch1 (7.79-vs. 2.96-fold), relative to controls ( Fig 4B). These data suggest that Srg3 is required for responses to Shh, supporting findings that the distribution of Shh/Gli target genes was confined to the distal-posterior mesenchyme in Srg3 CKO forelimb buds.
During limb development, Gli1 expression requires both transcription factors Gli2 and Gli3 [11], and Gli proteins regulate the expression of Ptch1 [23,35]. We asked whether the bifunctional action of Srg3 requires an interaction with Gli2 and Gli3 to regulate Gli1 and Ptch1 expression in developing limbs. Reciprocal coimmunoprecipitation of Srg3 with Gli2 and Gli3 from E11.5 limb bud lysates revealed that Srg3 formed a complex with endogenous Gli2, Gli3FL, and Gli3R ( Fig 4C). Using previously reported Gli-binding sites [23], we assessed Brg1 and Srg3 occupancy at the regulatory regions of Gli1 and Ptch1 by performing chromatin immunoprecipitation followed by qPCR (ChIP−qPCR) in E11.5 limb bud extracts. ChIP −qPCR analysis showed that both Brg1 and Srg3 proteins were enriched at the promoter regions of Gli1 and Ptch1 around Gli-binding regions in control limb buds, whereas their occupancies were considerably diminished in Srg3 CKO limb buds ( Fig 4D). Furthermore, Brg1 and Srg3 were also enriched near the limb specific enhancer of Ptch1, which might be required for sensing graded Shh activity (Fig 4D, region p5) [35].
We next investigated whether loss of Srg3 affects the recruitment of Gli2 and Gli3 proteins to the regulatory region of Shh target genes. The occupancy of Gli2 and Gli3 proteins was not significantly changed at the regulatory regions of Gli1 and Ptch1 in Srg3-deficient MEFs (Fig 4E). However, we found that Gli2 occupancy of Gli-binding sites was increased and the occupancy of Gli3 was reduced in E11.5 Srg3 CKO limb buds at the promoter regions of Gli1 and Ptch1 relative to controls (Fig 4F, regions g1−g2 and p1−p4). These data indicate that Gli proteins bound to the Gli-binding sites were not affected by Srg3 deficiency and suggest that their enrichment was differentially influenced by ectopic Shh activity in E11.5 Srg3 CKO limb buds. By contrast, we also found the decreased occupancy of Gli2 and Gli3 proteins near the limb specific enhancer of Ptch1 in E11.5 Srg3 CKO limb buds (Fig 4F, region p5), suggesting that GliA contributed by ectopic Shh signals might have no significant effect on this region. We hypothesized that Srg3 deficiency affects histone modification at the promoter regions of Gli1 and Ptch1 because the expression domains of Gli1 and Ptch1 were not expanded throughout Srg3 CKO forelimb buds at E11.5, despite the high GliA and the low GliR condition. Indeed, SWI/SNF complexes functionally interact with histone modifying proteins [36][37][38]. Furthermore, Shh signaling induces a loss of a repressive mark, trimethylation of histone 3 at lysine 27 (H3K27me3), by switching histone modifiers from methyltransferase Ezh2 to demethylase Jmjd3 [39]. To test our hypothesis, we compared H3K27me3 enrichment at the promoter regions of Gli1 and Ptch1 upon Srg3 deficiency against a previously reported distribution of H3K27me3 in MEFs [40]. Although Ezh2 and Suz12, components of the Polycomb repressive complex 2 (PRC2), were immunoprecipitated with both Gli2 and Gli3 in developing limbs (S5A Fig), there was no global change in H3K27me3 levels in Srg3-deficient MEFs or in the anterior and posterior mesenchyme of Srg3 CKO limb buds (S5B and S5C Fig). At the enriched regions of H3K27me3 on Gli1 and Ptch1 promoters, however, Srg3 deficiency resulted in decreased H3K27me3 level in a basal condition (Fig 4G, regions g1−g2 and  p1−p2). Upon Shh stimulation, on the contrary, H3K27me3 levels at these regions in Srg3-deficient MEFs were significantly higher than in controls (Fig 4H). Taken together, these data suggest that Srg3-containing SWI/SNF complexes contribute to the activation and repression of Shh target genes through changes in the chromatin status of Gli binding regions.

Loss of mesenchymal Srg3 disrupted BMP signaling and caused defective chondrogenesis in forelimb buds
Posterior Shh signaling establishes limb skeletal structures including posterior zeugopod elements (ulna/fibula) and digits 2 to 5 [2,4]. By contrast, loss of Gli3R or ectopic Shh signaling is detrimental to the formation of anterior skeletal structures [17,31,41]. To determine the role of bifunctional Srg3 in skeletal patterning, we assessed BMP activity, which promotes chondrogenesis at late stages [42]. In the anterior mesenchyme of Srg3 CKO forelimb buds, Msx2 expression, which marks BMP activity [30], was reduced at E10.75 and greatly abolished in the proximal region excluding the distal mesenchyme after ectopic Shh was induced (Fig 5A; n = 8 per stage). By contrast, posterior BMP activity remained low in Srg3 CKO forelimb buds ( Fig  5A). Among Bmp ligands, the expression of both Bmp2 and Bmp4, but not Bmp7, was diminished in the posterior mesenchyme of Srg3 CKO forelimb buds at E11.75 (Fig 5B and S6 Fig;  n = 6 per gene). Concurrent downregulation of Bmp2 and Bmp4, which are required to form the ulna and posterior digits 4 and 5 [43], could be causally implicated in the hypoplastic posterior skeletal elements of Srg3 CKO forelimbs (Fig 1). These results indicate that Srg3 deficiency disrupted BMP activities in the anterior and posterior mesenchyme.
We next examined the distribution of Sox9, which marks the condensation of chondrogenic progenitors [44]. Consistent with partially developed proximal skeletal elements, Sox9 expression was diminished in the stylopod primordia of Srg3 CKO forelimb buds prior to E10.75 (Fig 5C, left panel; n = 6). The expression of Hoxa9, Hoxd9, and Hoxd10 of paralogous hox groups Hox9 and Hox10, which control the formation of proximal skeletal elements [45,46], was reduced in the proximal region of Srg3 CKO forelimb buds (S7A Fig; n = 4 per gene). The expression of Irx3 and Irx5, which are essential for patterning proximal and anterior skeletal structures [31], was also downregulated in the proximal anterior region of Srg3 CKO forelimb buds (S7B Fig; n = 4 per gene). These data suggest that the Srg3-containing complexes might be required to pattern proximal skeletons. As limb bud outgrowth distally proceeds, Sox9-expressing progenitors were also decreased in the zeugopod and autopod primordia of Srg3 CKO forelimb buds (Fig 5C, middle and right panel; n = 8). Particularly, Sox9-expressing autopod progenitors in Srg3 CKO forelimb buds did not initiate mesenchymal condensation. Furthermore, Hoxa13, which delineates the presumptive autopod territories [47], was distalized and relatively enhanced in the anterior region (Fig 5D; n = 5). Taken together, deficiency of mesenchymal Srg3 progressively resulted in the loss of Sox9-positive progenitors in zeugopod and autopod primordia and this loss was paralleled by alterations in BMP activity.

Genetic interaction between Srg3 and Twist1 reveals synergism in the anterior zeugopod development
We tested whether ectopic Shh activity impacts anterior zeugopod development, as demonstrated by the absence of the radius and hypoplastic tibia in Srg3 CKO limbs. We introduced a single conditional allele of Twist1 (Twist1 f/+ ), which represses Shh expression in the anterior mesenchyme [41], into the Srg3 CKO background. Twist1 f/+ ;Prx1Cre forelimbs were phenotypically similar to Srg3 CKO forelimbs, except for more severe defects in the scapula (n = 13/13) (compare S8A Fig with Fig 1B). However, Twist1 f/+ ;Srg3 f/f ;Prx1Cre hindlimbs displayed ossification defects and syndactyly in the anterior autopods similar to those of Srg3 CKO hindlimbs ( Fig 1D and S8B Fig, arrow), but Twist1 f/+ ;Srg3 f/f ;Prx1Cre hindlimbs exhibited a complete absence of tibia (S8B Fig, arrowhead). Consistent with this skeletal phenotype, Sox9-positive progenitors of the tibia primordia were reduced in Twist1 f/+ ;Srg3 f/f ;Prx1Cre hindlimbs relative to Twist1 f/+ ;Prx1Cre hindlimbs (S8C Fig; n = 6). In addition, ectopic expression of Gli1 in Twist1 f/+ ;Srg3 f/f ;Prx1Cre hindlimb buds was activated earlier than in Twist1 f/+ ;Prx1Cre and Srg3 CKO hindlimb buds (compare S8D Fig with S2B Fig; n = 7). However, early activation of ectopic Shh expression was not detected in Twist1 f/+ ;Srg3 f/f ;Prx1Cre hindlimb buds (S8E Fig;  n = 6). These data suggest that anterior zeugopod development might be affected by SWI/SNF complex-mediated epigenetic changes including the ectopic Hh pathway and not simply because of the Shh ligand-dependent pathway. This finding supports the idea that the fate of anterior skeletal progenitors is progressively determined.

Bifunctional action of SWI/SNF complex in the Hh pathway regulates the spatiotemporal expression of Grem1
The low to high transition of BMP activity by the timely termination of Grem1 expression is required to initiate condensation and chondrogenic differentiation of proliferative digit progenitors [30,43]. To determine the effect of bifunctional action of the SWI/SNF complex on chondrogenic differentiation, Grem1 expression and BMP activity were analyzed at later stages. By E11.75, Grem1 expression began to be downregulated throughout the entire mesenchyme of control forelimb buds, but its decline was not observed and its separated domains became closer than those at E11.5 in Srg3 CKO forelimb buds (compare Fig 6A left panel with Fig 3C, black brackets; n = 8). At E12.5, Grem1 expression was cleared from the presumptive digit territories and confined to the interdigital mesenchyme in control forelimbs, but these spatial pattern changes were not observed in the autopods of Srg3 CKO forelimbs (Fig 6A, right panel; n = 6). Although Msx2 expression in the anterior margin of Srg3 CKO forelimb autopods was comparable to that in controls, it was undetectable in the interdigital mesenchyme (Fig 6B; n = 8). By contrast, the increased expression of Grem1 and low BMP activity were observed in the anterior region of Srg3 CKO hindlimb autopods (S9A Fig). We next examined whether delayed temporal kinetics of Grem1 in Srg3 CKO autopods is correlated with chondrogenesis of digit primordia and with digit separation. In Srg3 CKO autopods, the distributions of Sox9 and its target gene Col2a1 revealed delayed mesenchymal condensations, and anterior digit progenitors were relatively less condensed than posterior ones (Fig 6C and S9B Fig; n = 8 per gene). In addition, the comparison of Col2a1 distributions in Srg3 CKO fore-and hindlimb autopods revealed that both the extent of Grem1 propagation and its anterior upregulation caused the sequential onset of chondrogenesis in the posterior and anterior autopods. At this stage, Lysotracker Red staining in Srg3 CKO forelimb autopods showed increases of apoptotic cells in the distal mesenchyme underlying the AER, likely as a consequence of growth defects (Fig 6D, left panel; n = 6 per stage). Furthermore, the reduction of cell death in the interdigital mesenchyme, resulting in soft tissue syndactyly, was observed in the anterior autopods of Srg3 CKO forelimbs at E13.5 and hindlimbs at E12.5 (Fig 6D, right panel and S9C Fig). Taken together, these data indicate that the spatiotemporal regulation of Grem1 by the SWI/SNF complex is involved in digit determinative processes as well as in cell survival of expanding autopod progenitors.

Discussion
Our genetic analysis has shown that the SWI/SNF complex is required to modulate Shh responsiveness and repress the ectopic Hh pathway. Although specification of the AP limb bud axis is not affected by conditional inactivation of Srg3 in the limb bud mesenchyme, Srg3 CKO posterior progenitors fail to respond to graded Shh activity, leading to the redistribution of epithelial-mesenchymal signaling to the distal region. In parallel, loss of Srg3 causes the activation of ligand-independent and subsequent ligand-dependent Hh pathway in the anterior mesenchyme, resulting in the loss of anterior identity over time. Our analysis also reveals the dual requirement of the SWI/SNF complex in the Hh pathway for spatiotemporal regulation of Grem1.
Posterior limb skeletal elements are patterned depending on Shh signaling [2,4]. By contrast, recent reports have shown that formation of proximal and anterior limb skeletons is inhibited by early Hh activity prior to establishment of the ZPA and by activation of the anterior Hh pathway during limb patterning [10,31]. Skeletal phenotypes in Srg3 CKO forelimbs suggest that the Srg3-containing SWI/SNF complex is required for these distinct responses to Hh signaling. It has been known that SWI/SNF complexes and Polycomb group (PcG) proteins have antagonistic functions in repressing differentiation-related genes of embryonic stem cells [38]. In anterior limb buds, however, the SWI/SNF complexes appear to function synergistically with PcG proteins to repress the basal expression of Shh target genes. Consistent with our findings, deletion of H3K27 methyltransferase Ezh2, a catalytic subunit of PRC2, leads to ectopic expression of Shh target genes in anterior limb buds as well as derepression of Shh target genes in MEFs [39,48]. Given that the PRC2 interacts with Gli proteins in developing limbs, PRC2 complexes are also likely to be involved in Gli-mediated repression of Shh target genes in anterior limb buds. In addition to the repressive function in the anterior limb bud, it is assumed that the SWI/SNF complexes also act cooperatively with H3K27 demethylases in activating Shh-induced target genes. It has been demonstrated that the SWI/SNF complexes functionally interact with H3K27 demethylases such as Jmjd3 and Utx in various tissues such as developing lungs and hearts [36,37]. Particularly, a recent report showed changes in the epigenetic environment by switching Ezh2-PRC2 to Jmjd3 for Shh-induced target gene activation [39]. This implies that cooperative action between the SWI/SNF complex and Jmjd3 might be required for Shh target gene activation during limb development. Previous studies regarding SWI/SNF components have demonstrated that Snf5 deficiency leads to ectopic expression of Gli1 in developing limbs [49], and ATPase Brg1 is involved in the regulation of Shh target genes in an ATPase activity-independent manner during neural development [50]. However, we have presented genetic evidence showing bifunctional action of the SWI/SNF complex in distinct territories of limb bud mesenchyme. We do not exclude the possibility that the SWI/ SNF complex acts cooperatively with other chromatin regulators such as histone deacetylase (HDAC) that is associated with Shh/Gli pathway in developing limbs [50,51]. In addition, the phenotypes observed in Srg3 CKO limbs raise the possibility that the SWI/SNF complex likely controls the expression of other transcriptional regulators not specific to the Shh signaling pathway, such as Bmp and Hox genes. Further studies, including genome wide mapping of a H3K27Ac enhancer mark from the anterior and posterior limb buds of Srg3 CKO embryos, will help to elucidate the distinct regulatory functions of the SWI/SNF complex in chondrogenic differentiation and proximal patterning.
In Srg3 CKO forelimbs, one notable phenotype is the formation of variable digits, unlike polydactyly in hindlimbs. Concomitant deletion of Gli2 and Gli3 completely eliminates Gli1 expression but does not lead to digit loss in developing limbs [4,11,18]. Prx1Cre-mediated early deletion of Ptch1, however, causes oligodactyly and is accompanied by activation of the Hh pathway, whereas late Ptch1 depletion causes polydactyly [9,10]. Importantly, we have uncovered the requirement of the SWI/SNF complex for robust expression of Ptch1. Thus, the core mesenchymal deficiency of Ptch1 expression, resulting from its posterior restriction, may lead to reduced Shh activity sensing and restrain posterior digit formation in Srg3 CKO forelimbs. In Srg3 CKO forelimb buds, the reduced sensing of Shh causes distalization of epithelialmesenchymal signaling and Hoxa13/Hoxd13-positive presumptive autopod regions, markedly similar to limb buds conditionally lacking Ptch1 [9,35]. Recent studies on the mammal species with two to four digits might support variable digit patterning by altered Ptch1 expression observed in Srg3 CKO forelimb buds [35,52]. We assume that the extent of digit loss might be dependent on the integrity of the SWI/SNF complex controlled by Srg3. Meanwhile, ectopic Shh expression was induced in Srg3 CKO limb buds, although there is no enrichment for Srg3 on the ZRS. It has been shown that ectopic expression of Hoxd13 and Hand2 leads to misexpression of Shh in anterior limb buds [53][54][55]. These molecular changes observed in Srg3 CKO limb buds may result in ectopic expression of Shh, causing preaxial polydactyly. Taken together, variable digit patterning in Srg3 CKO forelimbs appears to occur through combinatorial actions of altered Ptch1 expression and ectopic anterior Hh activity.
Both the proximal and distal BMP activities in the anterior mesenchyme of Srg3 CKO forelimb buds are distinct from those of Gli3-deficient limb buds [22]. The comparison of anterior zeugopod development and digit numbers between Srg3 CKO fore-and hindlimbs showed that the dose and exposed duration of ectopic Hh activity negatively impact the differentiation of anterior prechondrogenic progenitors. Our data and previous reports have demonstrated that the expansion of Hh signaling has an inhibitory effect on the formation of proximal and anterior skeletal elements [10,31,41]. In this regard, the proliferative expansion of anterior progenitors negatively controlled by Gli3 might require time to ensure a sufficient population such as both Irx3-and Irx5-positive early progenitors [22,31]. Particularly, the genetic interaction between Srg3 and Twist1 showed synergism in limb skeletal formation such as in anterior zeugopod development. Twist1 not only functions as a Shh repressor but also controls the onset of osteoblast differentiation [41,56]. It is possible that the repressive roles of Twist1 in developmental processes might contribute to recruit chromatin regulators such as the PcG, for example, in promoting the epithelial-mesenchymal transition and in suppressing mesenchymal stem cell senescence [57,58]. The functional interaction of the SWI/SNF complex with transcriptional regulators acting either as activators or as repressors, which can recruit enzymes that modify active or repressive histone marks, may reveal synergistic and antagonistic actions of gene regulation at the chromatin level.
Derepression is one of the regulatory mechanisms underlying limb bud patterning. Our data highlight the sustained requirement of the SWI/SNF complex for transcriptional regulation of Grem1, a major Gli target gene controlled by derepression [23]. The expression of Grem1 in the limb bud is severely reduced in Shh −/− mutants and symmetrically expanded in both Gli3 −/− and Shh −/− ;Gli3 −/− mutants [16,17,59]. Compared with previous observations, Grem1 expression in Srg3 CKO forelimb buds is dynamically redistributed, possibly a consequence of the reconstitution of the GliA/GliR gradient by low Shh responsiveness and ectopic Shh activity. Consistently, it has recently been suggested that limb-specific enhancers integrated by multiple posterior GliA-and anterior GliR-dependent CRMs regulate the transcriptional activity of Grem1 [60]. Furthermore, the combined region of Grem1 expression domains in Srg3 CKO forelimb buds indicates that the definitive digit identity in this region could be progressively determined by altered Hh activity (Fig 6). Thus, our analysis suggests that bifunctional action of the SWI/SNF complex in the Hh pathway is essential for spatiotemporal regulation of Grem1 that mediates AP skeletal patterning elicited by GliA and GliR functions [18,22].
We have demonstrated that the SWI/SNF complex plays decisive roles in conferring graded Shh signaling upon developing limb progenitor cells. The SWI/SNF complex influences the progression of interlinked morphogen signaling pathways by modulating Shh responsiveness in the posterior limb bud and by repressing the Hh pathway in Shh-free regions. Our study showing the effects of epigenetic regulation by the SWI/SNF chromatin remodeling complex on limb patterning provides insights into deciphering developmental processes directed by morphogen gradients.

Ethics statement
All experiments with animals were performed according to the guidelines established by the Seoul National University Institutional Animal Care and Use Committees (SNUIACUC). SNUIACUC approved this study (approval number: SNU-130503-2). CO 2 gas was used for animal euthanasia.

Mice and embryos
Generation of mice carrying a conditional allele of Srg3 (Srg3 f/f ) was previously described [28].

Whole-mount in situ hybridization
The transcript distributions were assessed by whole-mount in situ hybridization according to the standard procedures as described [61] with the following minor modifications: embryos were permeabilized in proteinase K (10 μg/ml) in PBST at room temperature for 11 min (E9.5 −E10.5), 14 min (E10.5−E11.5) or 17 min (E11.5−E12.5) for analysis of limb mesenchyme and briefly for 3 min regardless of age for analysis of AER. All probes were linearized with the appropriate restriction enzyme and labeled using digoxigenin RNA labeling mix (Roche) with the appropriate polymerase (T7, T3 or SP6). Shh, Gli1, Bmp2, Bmp4 and Bmp7 probes were kindly provided by Y. Kong (Seoul National University). Fgf4 (Addgene plasmid #22085) [62] and Fgf8 (Addgene plasmid #22088) [63] probes were gifts from G. Martin. Hoxa9, Hoxd9 and Hoxd10 probes were generously provided by D. Wellik and Irx3 and Irx5 probes were provided by C. Hui. Other probes were amplified by PCR from cDNA fragments encompassing at least two exons (about 400−600 bp) of target genes and cloned into pGEM-T vectors (Promega). All representative expression patterns were obtained by analyzing at least three independent embryos per probe.

Skeletal staining and detection of apoptotic cells
Skeletal preparations and detection of apoptotic cells were performed as previously described [19,30]. For analysis of skeletal structures, samples were collected at E14.5 and P0 and cartilages and bones were stained with Alcian Blue and Alizarin Red, respectively. Distribution of apoptotic cells in whole limb buds was analyzed using Lysotracker Red (Molecular Probes L-7528, Invitrogen).

Cell culture
Primary Mouse Embryonic Fibroblasts (MEFs) prepared from E13.5 Srg3 f/f embryos, HEK293T, and Phoenix-eco cells were grown in DMEM medium (WelGENE) supplemented with 10% fetal bovine serum (FBS). For generation of Srg3-deficient MEFs, Phoenix-eco packaging cells were transfected with retroviral vectors expressing GFP alone (Empty) as a control or Cre-recombinase (Cre) by calcium phosphate method and their retroviral supernatants were harvested 2 d after transfection. MEFs were infected with the retroviral supernatant by spin infection for 90 min at 2500 rpm in the presence of 8 μg/ml polybrene. For inhibition of Hh signaling, MEFs were treated with 5 μM cyclopamine dissolved in ethanol vehicle for 24 h. For Shh stimulation, HEK293T cells were transiently transfected with ShhN expressing vector (kindly provided by M. Kang, Korea University Guro Hospital). Shh conditioned medium produced from transfected HEK293T cells was replaced with DMEM containing 2% FBS 24 h before harvesting and filtering of medium, and then added to MEFs for 24 h. Shh stimulated or cyclopamine treated MEFs were harvested for qPCR.