Combinatorial CRISPR/Cas9-approach to elucidate a far-upstream enhancer complex for tissue-specific Sox9 expression

Summary

SRY-box 9 (SOX9) is a master transcription factor that regulates cartilage development. SOX9 haploinsufficiency resulting from breakpoints in a ~1 Mb region upstream of SOX9 was reported in acampomelic campomelic dysplasia (ACD) patients, suggesting that essential enhancer regions of SOX9 for cartilage development are located in this long non-coding sequence. However, the cis-acting enhancer region regulating cartilage-specific SOX9 expression remains to be identified. To identify distant cartilage Sox9 enhancers, we utilized the combination of multiple CRISPR/Cas9 technologies including enrichment of promoter-enhancer complex followed by next-generation sequencing and mass spectrometry (MS), SIN3A-dCas9 mediated epigenetic silencing and generation of enhancer deletion mice. As a result, we could identify a critical far-upstream cis-element and Stat3 as a trans-acting factor, regulating cartilage-specific Sox9 expression and subsequent skeletal development. Our strategy could facilitate definitive ACD diagnosis and should be useful to reveal the detailed chromatin conformation and regulation.

eTOC paragraph

Mochizuki et al. develop combinatorial and systematic CRISPR/Cas9-based approaches to identify tissue-specific enhancers. They apply this to exploring Sox9 regulation in chondrocytes and identify a cartilage-specific enhancer important for SOX9 expression and skeletal development. CRISPR/dCas9-ChIP-mass spectrometry analysis further implicated STAT3 in acting at the enhancer to regulate SOX9 expression.

Introduction

SOX9 plays a critical role in various developmental processes, including chondrogenesis, sex determination, and pancreas development (Furuyama et al., 2011; Kronenberg, 2003; Ono et al., 2014; Wright et al., 1995). SOX9 mutations cause campomelic dysplasia (CD) and acampomelic campomelic dysplasia (ACD), which is a subtype of CD (Fonseca et al., 2013; Foster et al., 1994; Wagner et al., 1994). CD is a severe congenital disease and a form of skeletal dysplasia that is characterized by the bowing and shortening of the long bones, a bell-shaped thorax, narrow iliac wings, respiratory distress, and, often, disorders in sexual development (Foster et al., 1994; Meyer et al., 1997; Wagner et al., 1994). By contrast, ACD is not accompanied by bowed limbs and presents milder skeletal features (Fonseca et al., 2013). Consistent with these characteristics, Sox9 haploinsufficiency in mice leads to a severe phenotype that resembles CD and ACD (Bi et al., 2001).

The phenotypes observed in CD, ACD, and cases of SOX9 deletion are the result of impaired cartilage development, indicating that Sox9 is a “master transcription factor” required for chondrocyte differentiation and subsequent skeletal development (Kronenberg, 2003; Wright et al., 1995).

Previous studies have identified Sox9 enhancers for several distinct organs in regions that are relatively close to Sox9 promoter sites. These Sox9 enhancers were scattered between +95 kb and −251 kb for the gonad and testis (Sekido and Lovell-Badge, 2008), the telencephalon and midbrain, the node, notochord, gut, bronchial epithelium, pancreas, cranial neural crest and inner ear (Bagheri-Fam et al., 2006), and limb chondrocytes (Yao et al., 2015). As a far distant enhancer, the Sox9 enhancer for the mandible region was suggested to be located around –1.5 Mb based on a genome analysis of PRS patients (Figure 1B) (Benko et al., 2009). Considerable efforts have been aimed at identifying the cartilage-specific enhancer of Sox9. However, the critical enhancer region of Sox9 in chondrocytes has not yet been identified. The difficulty in identifying the cartilage Sox9 enhancer may be explained by the following factors, which are potentially unique to Sox9 regulation in chondrocytes: (1) multiple enhancer regions may contribute to Sox9 expression in chondrocytes and (2) these potential enhancers may be scattered over long distances in the upstream non-coding sequence.

Figure 1. Elucidation of the Sox9 promoter–enhancer complex using a CRISPR/Cas9 system and ChIP-seq of the non-coding region located between Sox9 and Kcnj2.

(A) Costal cartilages of 16.5-day-old embryonic mice (approximately 30 embryos) were harvested, and chondrocytes were cultured. After introduction of gRNA and HA-dCas9, ChIP was performed using an anti-HA antibody, and the obtained DNA were subsequently processed for MiSeq sequencing. ‘g’, ‘P’ and ‘E’ represent designed gRNA, promoter and enhancer, respectively. (B) ChIP-seq data obtained using an anti-HA antibody in the region between Sox9 and Kcnj2 in costal chondrocytes and MEF cells (*P<10⁻⁵, **P<10⁻⁸, P***<10⁻¹⁵). One peak, in addition to the gRNA region, remained (P-value < 10^–8). The lower panel shows the locations of the breakpoints in CD, ACD and PRS patients and other previously identified SOX9 enhancers. The locations in humans were replaced with the locations in mice between Sox9 and Kcnj2. 1, Testis (TESCO) 2, Node, notochord, gut, bronchial epithelium and pancreas (E1). 3, Most Sox9-expressing somatic tissues (SOM). 4, Chondrocyte (E84). 5, Chondrocyte (E195). 6, Prechondrocytes (E250). 7, Cranial neural crest cells, inner ear (E3). 8, Two mandibular mesenchyme enhancers. (C) Evolutionarily conserved RCSE regions in the mice [chr11, 111,838,889–111,839,334 (mm10)] with those of the genomes of chickens (71.5%), opossums (70.9%), and humans (80.9%). (D) ChIP-qPCR with H3K27ac and H3K4me1-antibody in all five peaks. Error bars represent SEM (*P < 0.05, **P < 0.01, ***P < 0.001, two-tailed Student t-test).

A ~2-Mb non-coding region is located between SOX9 and KCNJ2, which is a nearby coding gene upstream of SOX9. To survey this long DNA sequence for tissue-specific enhancers were used to generate transgenic (TG) mice carrying BAC and YAC clones that spanned the Sox9 far-upstream region using LacZ as a reporter. This approach demonstrated that even a 0.6-Mb region located upstream of the Sox9 transcription start site (TSS) may not be sufficiently long enough to include cartilage-specific enhancers (Wunderle et al., 1998). Conversely, human genetic data showed that breakpoints around the 1-Mb region upstream of SOX9 can cause Pierre Robin syndrome (PRS) and ACD, a subtype of CD (Benko et al., 2009; Fonseca et al., 2013; Leipoldt et al., 2007), implying that an essential cis-acting element of Sox9 for cartilage exists around this 2-Mb far-upstream region. Indeed, breakpoints within the 1-Mb region at 50–375 kb and 789–932 kb upstream of SOX9 are known to cause CD, whereas breakpoints that occur further upstream result in PRS and ACD (Benko et al., 2009; Fonseca et al., 2013; Leipoldt et al., 2007).

Recently, the CRISPR/Cas and transcription activator-like effector nuclease (TALEN) systems have been developed as new genome editing technologies (Boch et al., 2009; Cong et al., 2013; Jinek et al., 2012; Mali et al., 2013). These systems are based on the molecular mechanism whereby any endogenous DNA sequence can be recognized via the custom generation of a guide RNA (gRNA) for the CRISPR/Cas system or a TALE construct for the TALEN system. These systems can be applied to enrich a target chromatin region or the recruit artificial transcription mediators in order to analyze enhancer functions (Fujita and Fujii, 2013; Mendenhall et al., 2013). Here, we utilized the CRISPR/Cas9 system to explore the distant, cartilage-specific enhancers of Sox9 and their dynamic functions during cartilage development.

Design

During development, spatiotemporal gene expression is tightly regulated by cis-acting enhancer DNA regions and trans-acting proteins. However, there are few successful and available approaches to identify far-upstream enhancers and associating transcription factors. For example, although the cartilage-specific SOX9 enhancer has been expected to be ~1 Mb upstream of SOX9 based on genomic analyses of ACD patients (Benko et al., 2009; Fonseca et al., 2013; Leipoldt et al., 2007), both the far-upstream enhancers and transcription factors have not yet been well characterized.

Here, we propose combinatorial and systematic CRISPR/Cas9-based approaches to identify and characterize the far-upstream enhancers and associated transcription factors using Sox9 expression in cartilage as a model.

First, the promoter chromatin region with interacting far-upstream enhancers was enriched by CRISPR/Cas9 targeting of the Sox9 promoter in chondrocytes. The following sequence analysis of the precipitated DNA samples revealed the critical far-upstream enhancer that exhibited an activity that was partially confirmed by TG mouse generation and LacZ staining..

Second, the function of the identified enhancer was examined using CRISPR/Cas9 fused to a transcriptional repressor system in vitro. Sox9 expression in costal chondrocytes and chondrogenesis in C3H10T1/2 cells were significantly repressed by targeting the enhancer with SIN3A-dCas9 fusion protein.

Next, the critical role of the identified region was confirmed by generating enhancer deletion mice by CRISPR/Cas9. The rib cages of rib cage specific enhancer (RCSE)-null mice were significantly narrower and shorter compared to those of wild-type mice by alcian blue/alizarin red staining and micro CT.

Furthermore, transcription factors that regulate Sox9 expression were screened using a CRISPR/Cas9-ChIP-MS system targeting the RCSE region in chondrocytes; Stat3 was identified as a transacting factor for Sox9 enhancer.

Thus, our combinatorial analysis using the CRISPR/Cas9 system enables the elucidation of promoter-enhancer transcriptional complexes, including enhancers and transcription factors.

Results

Identification of candidate enhancers for Sox9 in primary chondrocytes using CRISPR/dCas9

Sox9 expression during embryogenesis is observed at around E9.5 in chondrogenesis sites and continues in chondrocytes embedded in the cartilage of body regions, such as the limbs, ribs, and jaws. To enrich the Sox9 promoter with associated cis-acting enhancer regions and identify putative enhancer regions that are associated with the Sox9 promoter, we designed seven gRNAs that cover the Sox9 TSS near the promoter in a retroviral system (Figure S1A). Each gRNA and HA-tagged dCas9 were introduced in a chondrogenic cell line, ATDC5 (Atsumi et al., 1990), followed by selection on geneticin and puromycin for 1 week and chromatin immunoprecipitation (ChIP) using an anti-HA antibody. Precipitation efficiencies of the target chromatin regions were determined via quantitative PCR, and the results confirmed that gRNAs specifically recognized and pulled down target chromatin regions. Among the gDNAs used, the C3 gRNA probe showed the highest efficiency and specificity and therefore was selected for subsequent analysis (Figure S1B). The reason why C3 gRNA showed the best pull down efficiency might be a consequence of the relationship between off-target effects and gRNA binding efficiencies.

To examine the physiological expression patterns of Sox9, primary chondrocytes from dissected rib cages of E16.5 mice were purified (Figure 1A); results showed strong expression of Sox9 and Col2a1 (Figure S2). However, after the second passage, Sox9 and Col2a1 expression levels were significantly decreased, indicating that chondrocyte properties were not maintained after multiple passages (Figure S2). Using the above optimized method, we performed ChIP-seq of enriched Sox9 promoter–enhancer complexes in 5 × 10⁷ primary chondrocytes and murine embryonic fibroblasts (MEF) cells (negative control). Specifically, C3 gRNA and dCas9 were introduced into primary chondrocytes and MEF cells using a retroviral system. After selection on geneticin and puromycin for 1 week, ChIP was performed using an anti-HA antibody on infected chondrocytes without passage (Figure S1C). The precipitated chromatin samples were then subjected to DNA purification and processed for next-generation DNA sequencing analysis. Several sequence peaks appeared within the 1.7-Mb Sox9 upstream region of the mouse genome in primary chondrocytes, while there were no significant peaks in MEF cells. Statistical validation of these sequence peaks yielded significant peaks in addition to the gRNA region in the Sox9 upstream region with a P-value < 10^–5 and only one peak with a P-value < 10^–8 (Figure 1B). The candidate enhancer region (Rib-Cage Specific Enhancer: RCSE) [chr11, 111,838,889–111,839,334 (mm10)] was determined as well conserved region among mice, chickens, opossums, and humans, nearby the strongest peak (Ovcharenko et al., 2004) (Figure 1C). We next performed the ChIP-qPCR using anti-H3K4me1 and H3K27ac antibody in all five regions and found that RCSE regions were significantly enriched in both H3K4me1 and H3K27ac ChIP, suggesting the potential enhancer (Figure 1D).

The RCSE region forms a transcriptional complex with the promoter of Sox9 in primary chondrocytes

To test for chromatin interactions between the Sox9 promoter and RCSE region, which corresponded to the highest peak for rib-cage-derived chondrocytes located around the region 1-Mb upstream of Sox9, chromosome conformation capture (3C) was performed in primary chondrocytes. Specific interactions were detected by PCR between the Sox9 promoter and RCSE region and the detected band included sequences of both the Sox9 promoter and the RCSE region in primary chondrocytes (Figure S5), while no bands were detected in MEF cells (Figure 2A).

Figure 2. Assessment of the RCSE region.

(A) 3C analysis of the RCSE region. “PC” and “MEF” represent primary chondrocytes and MEF cells. “G” represents mouse genomic DNA. “W” represents an equal volume of H2O substituted for a 3C template. “+” represents the ligated template, while “-”represents a non-ligated control template. (B) Scheme of the ChIP-qPCR and MS evaluating the physical association between the RCSE region and the Sox9 promoter. (C) ChIP-qPCR using an anti-HA antibody after the introduction of both the RCSE gRNA and dCas9. In primary chondrocytes, the Sox9 promoter was enriched in the RCSE enhancer–chromatin complex, whereas association between the RCSE enhancer and the Sox9 promoter was not observed in MEFs. (D) Scheme showing the repression of the RCSE region by a SIN3A repressor. (E) Sox9 expression in SIN3A-repressed primary chondrocytes. (F) Sox9 and Col2a1 expression in SIN3A-repressed C3H10T1/2 cells during chondrogenesis. “dCas9” denotes induction using only dCas9 with RCSE gRNA, whereas “dCas9+SIN3A” denotes induction using both SIN3A-dCas9 and RCSE gRNA. d0, d3, and d6 indicates 0, 3, and 6 days after cartilage differentiation, respectively. Error bars represent SEM (*P < 0.05, **P < 0.01, ***P < 0.001, two-tailed Student t-test).

Next, to confirm the physical association between the Sox9 promoter and RCSE region, we designed a gRNA near the RCSE region (Figure 2B). ChIP-qPCR was performed using an anti-HA antibody after the introduction of both the RCSE gRNA and HA-tagged dCas9 into primary chondrocytes and MEF cells. The Sox9 promoter was enriched in the RCSE enhancer–chromatin complex in primary chondrocytes, whereas no association was observed between the RCSE enhancer and the Sox9 promoter site in MEF cells (Figure 2C). These data suggest that there is a long-range interaction between the Sox9 promoter and RCSE region in primary chondrocytes.

The RCSE region plays an essential role in Sox9 expression in chondrocytes during chondrogenesis

To further characterize the role of this enhancer in Sox9-dependent chondrogenesis, we used C3H10T1/2 cells as an in vitro differentiated chondrocyte model. As previously reported, bone morphogenetic protein 2 (BMP-2) promotes chondrogenesis in C3H10T1/2 cells, which is accompanied by Sox9 upregulation (Denker et al., 1999). We introduced a gRNA recognizing the RCSE enhancer and dCas9 fused with a SIN3A repressor element, which recruits a transcriptional repressor complex to a specific enhancer region anchored by the gRNA via the SIN3A repressor domain (Figure 2D). We tested this CRISPR/Cas9-mediated repressor recruitment system in rib-derived chondrocytes and found that targeting this enhancer by SIN3A significantly repressed Sox9 expression (Figure 2E). Moreover, as expected, recruitment of the SIN3A repressor domain to the RCSE enhancer region inhibited Sox9 and Col2a1 upregulation and subsequent chondrogenesis in C3H10T1/2 cells (Figure 2F).

The RCSE region functions as a costal and sternum cartilage-specific enhancer

To test whether the RCSE enhancer region can act as tissue-specific enhancer, we generated TG mice carrying the putative enhancer region using LacZ as a reporter (Figure S4A). As expected, TG mice with the Sox9 promoter and 4× RCSE region showed cartilage-specific LacZ staining, particularly at the costal cartilage in whole-mount images in the E12.5 and E14.5 stages (Figure 3A), in addition to weaker signals at the spine, shoulder, forelimbs and cranium. By contrast, TG mice with only the Sox9 promoter showed no LacZ staining in cartilage (Figure S4B). Moreover, sterna showed LacZ staining in E14.5, but not E12.5 mice. Similar to the results obtained using whole-mount images, frozen sections showed costal cartilage- and sternum-specific LacZ staining in E14.5 mice (Figure 3B). On the other hand, other Sox9-expressing tissues (brain, kidneys, and limbs) showed no LacZ staining (Figures S4C-E). These data suggest that the RCSE region functions as a costal- and sternum cartilage-specific enhancer.

Figure 3. Activities of RCSE TG mice in E12.5 and E14.5.

(A) TG mice carrying the 4× tandem RCSE region, in addition to a Sox9 promoter region combined with a LacZ reporter gene. (B) Frozen sections of sterna and ribs. Staining of mice was confirmed in three independent samples to assess reproducibility, and results yielded identical patterns.

RCSE null mice have narrower and shorter rib cages than wild-type mice

To demonstrate the critical role of the RCSE enhancer in proper cartilage development, we generated RCSE, Sox9 far-distant enhancer deletion mice using the CRISPR/Cas9 system (Figure 4A, S5A and B). First, we confirmed that Sox9 and Col2a1 expression is downregulated in the costal cartilage of RCSE^+/− mice and RCSE-null mice when compared to wild-type mice (Figure 4B). Next, characterization of the bone/cartilage developmental phenotype in E16.5 embryos was performed via alcian blue/alizarin red staining. We observed a distinct defective rib cage phenotype in RCSE^+/− mice and RCSE null mice. In particular, the rib cages of RCSE^+/− mice were significantly narrower and shorter compared to those of wild-type, and those of RCSE null mice were further narrower and shorter compared to those of wild-type and RCSE^+/− mice (Figure 4C); furthermore, rib cage volumes (calculated via ellipsoid approximation) were significantly smaller in RCSE^+/− and RCSE null mice (Figure 4D). On the other hand, wild-type, RCSE^+/− and RCSE null mice showed no clear differences in the length of limbs.

Figure 4. Deletion of the RCSE region in vivo.

(A) Scheme of deletion of the RCSE region using two different gRNAs and Cas9. The RCSE region is located at chr11, 111,838,889–111,839,334 (mm10). (B) Comparison of Sox9 and Col2a1 expression in costal chondrocytes of wild-type, RCSE^+/− mice and RCSE null mice. (C) Skeleton specimens of wild-type, RCSE^+/− mice and RCSE KO E16.5 mouse embryos. (scale bar, 2 mm). (D) Thoracic volume was approximated by calculating an ellipsoid volume. RCSE^+/− and null mice were compared with wild-type as controls in 4 littermates (n = wild-type 5, RCSE^+/− type 6, KO mice 5). (E) Micro CT of rib cages from wild-type, RCSE^+/− mice and RCSE null mice (10 weeks old). (scale bar, 5 mm). (F) Thoracic volume was approximated by calculating a conic volume. RCSE^+/− and null mice were compared with wild-type as controls in 3 littermates (n = wild-type 4, RCSE^+/− type 6, KO mice 5). Error bars represent SEM (*P < 0.05, **P < 0.01, ***P < 0.001, two-tailed Student t-test).

Using micro computed tomography (CT), skeletal formation in adult (10 weeks old) wild-type and RCSE^+/−, RCSE null mice was also evaluated. The rib cages of RCSE^+/− adult mice (10 weeks old) were significantly narrower and shorter compared to those of wild-type, and those of RCSE null adult mice were further narrower and shorter compared to those of wild-type and RCSE^+/− mice (Figure 4E, F). Moreover, RCSE-null mice had shorter sterna and thinner sternebrae (Figure 4E). The observed sternal and rib cage defects were consistent with the bell-shaped chest or pectus excavatum observed in Robin sequence and other milder forms of CD and ACD (Fonseca et al., 2013; Moog et al., 2001). On the other hand, there were no clear differences in terms of body weight and the length of limbs between wild-type and RCSE-null mice.

The epiphyseal growth plates of the ribs and sternum of 10-day-old wild-type and RCSE-null mice were also evaluated via histological analysis. While no differences in the femoral growth plates were observed, rib growth plates of RCSE null mice showed enlargement (~2.8-fold) of the hypertrophic zone and reduction (~0.7-fold) of the proliferating zone (Figures 5A, B). In the sternum, cartilage layer and sternal width were less in RCSE null mice compared with those in wild-type mice, whereas the hypertrophic zone and proliferating zone in sternum growth plates of RCSE-null mice appeared normal (Figure 5C). These phenotypes in the hypertrophic zone are consistent with those in the Sox9^+/− mutant mice (Bi et al., 2001) and conditional deletion mice (Akiyama et al., 2007; Kist et al., 2002). Genetic evidence further supports the critical role of the far-distant enhancer of Sox9 in rib cage cartilage development, as determined using CRISPR/Cas9-dependent Sox9 promoter chromatin enrichment.

Figure 5. Comparison of the longitudinal sections of the ribs and femoral and sternum growth plates between 10-day-old wild-type and RCSE null mice.

(A) Rib growth plates (scale bar, 200 μm). (B) Femoral growth plates (scale bar, 200 μm). (C) Sternum growth plates (scale bar, 500 μm, enlarged images, 100 μm). “P” and “H” represent proliferating chondrocytes and hypertrophic chondrocytes.

Identification of Stat3 in the Sox9 transcription complex in chondrocytes

Finally, to obtain more detailed insights into the molecular mechanisms underlying the Sox9 promoter-enhancer complex in chondrocytes, we performed large-scale Sox9 promoter-enhancer complex enrichment using an RCSE enhancer-targeting gRNA and used the obtained chromatin precipitate for MS analysis (Figure 6A). Specifically, we introduced both the gRNA near the RCSE region and HA-tagged dCas9 into primary chondrocytes. Subsequently, we performed ChIP using an anti-HA antibody on chondrocytes expressing dCas9 and gRNA. Samples were then subjected to MS to identify the proteins associated with the Sox9 transcriptional complex. As negative controls, ChIP-MS using the CRISPR/Cas9 system was performed using the RCSE region-targeting gRNA in MEF cells with a similar technique.

Figure 6. Identification of proteins associated with the enriched RCSE enhancer–chromatin complex via ChIP-MS.

(A) Scheme of the ChIP-MS showing the physical association between the RCSE region and the Sox9 promoter. “A” indicates transcriptional factors that we would like to identify. (B) Details of the proteins identified using ChIP-MS analysis (detected using primary chondrocytes twice, not detected using MEF cells twice). (C) siRNA screening in 6 TFs in mouse chondrocytes. Error bars represent SEM (*P < 0.05, **P < 0.01, ***P < 0.001, two-tailed Student’s t-test).

Using this strategy, we observed a set of proteins that was present in the enriched Sox9 enhancer–chromatin complex. As expected, a series of histone proteins and dCas9 (CSN1) were detected in all MS analyses. Here, we focused on 506 proteins that were detected in ChIP-MS twice using chondrocytes, and also that were not detected using MEF cells twice (Table S1).

The analysis detected six transcriptional factors (TFs) and activators (Figure 6B). To determine the functional roles of these TFs in Sox9 gene regulation, we introduced siRNAs targeting each TF into primary mouse chondrocytes and investigated their corresponding effects on Sox9 expression. Among all of the siRNAs tested, treatment with siRNA targeting signal transducer and activator of transcription 3 (Stat3) was found to reduce Sox9 expression by 50–60% (Figure 6C). We additionally performed Multiple Reaction Monitoring (MRM)-targeted proteomics four times to ensure stabilized reproducibility of Stat3 using RCSE-targeting primary chondrocytes and could confirm a very large quantity of Stat3 in four independent experiments, compared with MS targeting a different sequence in chr11 and a different chromosome (chr12) (Figure S6). Based on these findings, we further examined the potential role of Stat3 in Sox9 gene expression.

Stat3 is a member of the STAT family and is expressed in most tissue types. Stat3-knockout mice succumb to embryonic lethality at E6.5 (Takeda et al., 1997). Stat3 can be transcriptionally activated by phosphorylation of tyrosine 705 or serine 727 residues in response to various cytokines, including interleukin-6 (IL-6) (Zhong et al., 1994). STAT3 has been reported to be a positive regulator of chondrogenic differentiation (Hall et al., 2017; Kim and Sonn, 2016; Kondo et al., 2015). However, the exact functions of Stat3 in Sox9 expression via enhancers in chondrocytes and cartilage tissues remain unclear.

Stat3 regulates Sox9 expression via the RCSE enhancer region

First, although Stat3 overexpression in ATDC5 did not lead to increased luciferase activity in the presence of the Sox9 promoter alone, the addition of a 1× RCSE region to the Sox9 promoter increased luciferase activity, and including the 4× RCSE region in the Sox9 promoter further increased luciferase activity induced by Stat3 (Figure 7A). Moreover, three putative binding motifs of Stat3 were identified within the RCSE region using the JASPAR database (Mathelier et al., 2016) (Figure S7A). We constructed reporter vectors in which the RCSE region was cloned upstream of the Sox9 promoter (Figure S7B), and a luciferase assay was performed in ATDC5. Our results revealed that deletion of the S2 (ΔS2) Stat3 binding motif abrogated STAT3-dependent upregulation of luciferase activity, while deletion of the S1 (ΔS1) or S3 (ΔS3) motif did not alter the upregulation of luciferase activity (Figure S7C). ChIP-qPCR using an anti-Stat3 antibody revealed an increased Stat3 concentration in the RCSE and Sox9 promoter regions (Figure 7B). In addition, nuclear staining of primary chondrocytes was performed via immunostaining using an anti-Stat3 antibody (Figure 7C). Stat3 knockdown was performed by stably expressing a short-hairpin RNA (shRNA) in C3H10T1/2 cells, which led to Sox9 and Col2a1 downregulation during chondrogenesis (Figure 7D). Finally, we investigated the role of Stat3 in Sox9 expression in vivo. Given that Stat3-knockout (KO) mice showed embryonic lethality at E6.5, Stat3 conditional KO (cKO) mice (Stat3^f/f; Mx-Cre) were analyzed. Sox9 and Col2a1 showed significant downregulation in the costal chondrocytes of Stat3-cKO mice (Figure 7E). Moreover, Stat3-cKO mice exhibited severe growth defects with delayed endochondral ossification at E15.5. Interestingly, Stat3-cKO mice exhibited systemic chondrodysplasia, whereas RCSE-null mice showed a skeletal defect only in the rib cages (Figure 7F). These phenotypic discrepancies could be explained by the possibility that ubiquitous expression of STAT3 may also regulate not only RCSE, but also the promoter and other enhancers that are involved in whole cartilage-specific regulation of Sox9 expression (Figure 7G).

Figure 7. Regulation of Sox9 expression by Stat3.

(A) Luciferase assay under Stat3 overexpression with Sox9P, Sox9P + 1× RCSE and Sox9P + 4× RCSE Luc reporter. STAT3 expression was analyzed by western blotting. (B) ChIP enrichment in primary chondrocytes using anti-Stat3 antibody. (C) Immunostaining of primary chondrocytes using an anti-Stat3 antibody (scale bar, 100 μm). (D) shRNA knockdown of Stat3 in C3H10T1/2 cells during chondrogenesis. (E) mRNA expression levels of Stat3, Sox9, and Col2a1 in costal cartilage in Stat3^f/f and Stat3^f/f; Mx-Cre E15.5 mouse embryos. (F) Skeleton specimens of Stat3^f/f and Stat3^f/f; Mx-Cre E15.5 mouse embryos. (scale bar, 2 mm) (G) Scheme showing the regulation of Sox9 transcription. Error bars represent SEM (*P < 0.05, **P < 0.01, ***P < 0.001, two-tailed Student’s t-test).

Discussion

As previously described, cartilage-specific enhancers of Sox9 have been reported by several studies. To date, the longest cis-acting regulatory region that has been validated using targeted genome elimination is a 1-Mb upstream enhancer of mouse limb-specific sonic hedgehog expression (Sagai et al., 2005). However, there have been no successful reports that systematically surveyed the far-upstream enhancers, and none have demonstrated the function of the Sox9 enhancer by generating enhancer deletion mice.

A strategy for the enrichment of targeted chromatin regions using CRISPR/Cas9 or TALEN systems has been proposed as en-ChIP and CAPTURE (Fujita and Fujii, 2013; Liu et al., 2017). Here, we could determine the critical far-upstream enhancer (i.e., over 1 Mb) and trans-acting factors for Sox9 expression in cartilage by CRISPR/Cas9-based promoter region enrichment. The function of the identified enhancer could be successfully determined by CRISPR/Cas9 fused to a transcriptional repressor system in vitro. Furthermore, the critical role of the region has been confirmed by generation of the enhancer deletion mice by CRISPR/Cas9. Our findings revealed by a combinatorial and systematic CRISPR/Cas9-based approach were consistent with genetic alternations in human ACD (Fonseca et al., 2013).

While this RCSE region involves four SNPs (rs4141924, rs7221327, rs12453384, and rs9912158), there have been no reports of associations between human diseases and these SNPs in the RCSE region. Interestingly, the RCSE enhancer region, which showed the highest peak in the rib-cage chondrocyte ChIP-seq enriched with the Sox9 promoter, corresponded to costal cartilage and sternum-specific expression as determined via LacZ enhancer reporter analysis. Consistent with the above results, targeted deletion of this enhancer region in E16.5 mouse embryos and 10-week-old adult mice significantly reduced the thoracic volume, owing to underdeveloped costal cartilage and sternum, which resembles a primary symptom of CD and ACD (Fonseca et al., 2013; Moog et al., 2001). These findings strongly suggest that the RCSE enhancer identified via our CRISPR/Cas9 analysis should be responsible for one of the major symptoms of ACD.

Moreover, the epiphyseal rib growth plates in RCSE-null mice showed the enlargement of the hypertrophic zone and sternum growth plates in RCSE null mice showed the narrowness of cartilage-layer and sternal width. This phenotype is very similar to the enlargement of the hypertrophic zone in Sox9^+/− mutant mice (Bi et al., 2001). This discrepancy of phenotype between costal cartilage and sternum cartilage may be partly explained by the difference in the timing of Sox9 expression in both cartilages via RCSE. LacZ expression via RCSE was observed in only costal cartilage in E12.5, whereas LacZ expression in sternum started appearing at E14.5. These data suggest that the RCSE region functions as the critical enhancer and promotes endochondral ossification in ribs-cage and subsequently extends the lengths of ribs and sternum. On the other hand, Sox9 and Col2a1 expression in costal cartilage remained approximately 50% in RCSE null mice, suggesting there are other additional cartilage enhancers.

Several studies have previously investigated Sox9 gene expression. CCAAT-binding, CREB, hypoxia-inducible factor 1a, Sonic hedgehog, and Sp1 have been demonstrated to induce Sox9 expression via the Sox9 promoter (Amarilio et al., 2007; Colter et al., 2005; Piera-Velazquez et al., 2007; Tavella et al., 2004), while extracellular stimuli, such as epidermal growth factor, fibroblast growth factor, or transforming growth factor-β signals, would enhance Sox9 expression (Fyodorov and Kadonaga, 2002; Liu et al., 2015; Rodriguez-Esteban et al., 1999). However, the precise regulatory mechanisms that control Sox9 expression in the cartilage remain to be fully elucidated. Our present findings showing the interaction between Sox9 and Stat3 via RCSE enhancer should add new insights into mechanisms of cartilage-specific Sox9 gene expression. Here, we utilized Mx-Cre mice, lacking mesenchyme specificity, to delete Stat3 expression. To confirm the precise function of Stat3 in early stage of chodrogenesis, it may be useful to apply mesenchyme specific Cre line, such as dermo1-Cre, to delete Stat3 expression in future. Additionally, other molecules present in the Sox9 promoter–enhancer complex should be also investigated, for example, Far upstream binding protein 3 (Fubp3) is associated with a 50% reduction of Sox9 expression and is a strong candidate transcriptional factor that regulates Sox9 expression.

Combinatorial analysis using the CRISPR/Cas9 system enabled a detailed investigation of the transcriptional mechanisms and genome dynamics of inherited diseases. The comparison between this strategy and other chromatin status analyses, such as 4C, Hi-C, ATAC-seq, should provide us more detailed information of the whole transcriptional apparatus of Sox9. Combining these strategies, the identification of this new RCSE enhancer and its corresponding molecular functions will aid in the establishment of definitive diagnoses and identification of potential targets for the treatment of these diseases (Attanasio et al., 2013; Kronenberg, 2003; Leung et al., 2015).

Limitations

Chromatin conformation analysis, such as 4C or Hi-C, has been successfully utilized to reveal genome wide promoter-enhancer interactions, indicating the transcriptional active domains. It is still no alternative strategy to reconstruct genome wide chromatin 3-D structures; however, these methods are based on random DNA fragment ligation frequency, which could limit the resolution and be affected by PCR bias. On the other hand, CRISPR/dCas9-ChIP-seq in this report could simply concentrate the targeted interacting chromatin complex without amplification, resulting in identification of detailed direct enhancer/promoter association with low background. The potential problems of this method might be caused by dCas9 occupancy to the targeted region, off target of CRISPR or introduction of CRISPR/dCas9 to the cell. Thereby these strategies could be complementally utilized to understand the detailed chromatin conformation and subsequent gene regulation.

STAR Methods

Detailed methods are provided in the online version of this paper and include the following:

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for reagents and resources should be directed and will be fulfilled by the Lead Contact, Hiroshi Asahara (asahara.syst@tmd.ac.jp).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

ATDC5 culture

ATDC5 cells were cultured in DMEM-F12 Ham (Sigma–Aldrich) supplemented with 5% fetal bovine serum, 100 U/mL penicillin, and 100 μg/mL streptomycin (Wako Pure Chemical Industries) at 37°C in 5% CO₂.

Primary chondrocyte isolation

The anterior rib cages and sterna of E16.5 C57BL/6J embryos (Sankyo Lab Service) were harvested and incubated with 3 mg/mL collagenase D (Roche Applied Science) in Dulbecco’s modified Eagle’s medium (DMEM) for 1 h at 37°C. Subsequently, embryos were further incubated in DMEM containing 3 mg/mL collagenase D at 37°C overnight.

The next day, the digested anterior rib cages and sterna were filtered through a 100-μm cell strainer (BD Biosciences). Isolated chondrocytes were cultured in DMEM-F12 Ham (Sigma–Aldrich) supplemented with 10% fetal bovine serum, 100 U/mL penicillin, and 100 μg/mL streptomycin (Wako Pure Chemical Industries) at 37°C in 5% CO₂ and were used without passage.

C3H10T1/2 chondrogenesis

C3H10T1/2 cells were cultured in DMEM (Sigma-Aldrich) supplemented with 10% fetal bovine serum, 100 U/mL penicillin, and 100 μg/mL streptomycin (Wako Pure Chemical Industries) at 37°C in 5% CO2. Trypsinized cells were resuspended in DMEM with 10% fetal bovine serum at a concentration of 10⁷ cells/ml and a 10-μl drop of this cell suspension was placed in the center of a well in a standard 24-well culture dish. After 2 h at 37°C and 5% CO2, 1 ml medium containing BMP-2 was added to the culture in a concentration of 100 ng/ml. Medium was changed every 3 days (Denker et al., 1999).

Mice

Wild-type mice were in the C57BL/6J background. In transgenic mice, the constructs including the RCSE region was injected into fertilized mouse eggs and implanted into pseudopregnant foster females. F0 embryos were collected and analyzed at E12.5, E14.5.

In the generation of RCSE knockout mice, the concentration of the injected RNAs was maintained at 500 ng/ μL. Two gRNAs, which were located near the RCSE region, and hCas9 mRNA were mixed at a 1:1:2 ratio. The final concentration of the two gRNAs was 125 ng/ μL each, while that of hCas9 was 250 ng/ μL in this experiment (Inui et al., 2014). RCSE knockout mice were analyzed at E16.5 and 10-weeks-old.

In the Generation of Stat3 flox mice, the inducible conditional Stat3 KO mice (Mx1-Cre;Stat3f/f) were generated by crossing Stat3f/f and Mx1Cre transgenic mice. Mx1-Cre;Stat3f/f, and Stat3f/f mice were crossed, and pregnant mice received a total of three intraperitoneal injections of 250 μg of polyI:C (Sigma–Aldrich) administered every three days starting from day eight after the vaginal plug was observed. Stat3-cKO mice were analyzed at E15.5.

Because it is difficult to determine the sex of embryos, we did not sort genders prior to experiments. All mice studies were performed in accord with protocols approved by the committee at the Tokyo Medical and Dental University.

METHOD DETAILS

gRNA design

Seven gRNAs within the CpG island located upstream of Sox9 and two gRNAs located near the RCSE region were designed using the University of California, Santa Cruz genome browser (http://genome.ucsc.edu/) (Figure S1A, Table S2).

Plasmid construction

Plasmids expressing Cas9, nuclease-dead Cas9 (dCas9) and gRNAs were obtained from Addgene. Construction of the gRNA expression vector was performed as previously described (Inui et al., 2014). HA- and FLAG-tagged dCas9 were PCR-amplified using primers that contain the HA and FLAG sequences, and the resulting products subsequently cloned into the retroviral vector pMIGR-IRES-Puro. gRNA sequences were cloned downstream of the U6 promoter of a pSIREN-RetroQ Neo vector (Clontech) using the Gene Art Seamless Cloning and Assembly Kit (Invitrogen) according to the manufacturer’s instructions. The retroviral vector encoding the dCas9-SIN3A fusion protein was constructed using the Multisite Gateway Three-Fragment Vector Construction Kit (Invitrogen) according to the manufacturer’s instructions. Briefly, oligonucleotides encoding the HA tag, dCas9, and human SIN3A (from positions 1714 to 3417 of NM_001145358.1) were cloned into pDONR P4-P1r, pDONR 221, and pDONR P2R-P3, respectively, to generate pENTR plasmids. Subsequently, the generated oligonucleotides were further cloned into the retroviral destination vector pMIGR-IRES-Puro R4R3. shRNA oligonucleotides targeting mouse Stat3 and its control were cloned into pSIREN-RetroQ Puro. The following shRNAs were used: control, 5’–GCCCAGATTTAGAGACAAT–3’; shStat3#1, 5’–GGTATAACATGCTGACCAATA–3’; and shStat3#2, 5’–GGTACATCATGGGTTTCATCA–3’. The Sox9 promoter sequence [chr11:112780809–112782537; Dec 2011, (GRCm38/mm10)] was inserted into the pBluescript II vector (Agilent Technologies). The RCSE region was amplified from mouse genomic DNA via PCR (F:AAAACCATTGCTGCACCCAG, R: GCACCCTCTAACCACGATCA), and four tandem copies of the RCSE region were cloned downstream of the Sox9 promoter-driven LacZ site (kindly provided by Dr. S. Takada, National Institute of Child Health, Tokyo, Japan) or upstream of the Sox9 promoter-driven luciferase vector (pGL4; Promega) using the Gene Art Seamless Cloning and Assembly kit. The coding sequence of Stat3 was cloned into pcDNA3.1 (Invitrogen) using the Gene Art Seamless Cloning and Assembly Kit.

Retroviral infection

PLAT-E cells (Morita et al., 2000) were seeded at a density of 1 × 10⁷ onto poly-D-lysine (0.02 mg/mL; Sigma–Aldrich)-coated 15-cm dishes in DMEM containing 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. After 24 h, cells were transfected with 15 μg of the p10A1 vector (Clontech) and 45 μg of pSIREN-gRNA, pMIGR-HA-dCas9, or pMIGR-dCas9-SIN3A using Fugene HD (Promega) and subsequently cultured for 48 h. Supernatants containing the virus were collected, passed through a 0.45- μm filter, and infected into primary chondrocytes in the presence of 8 μg/mL hexadimethrine bromide (Sigma–Aldrich). The following day, chondrocytes were selected using puromycin (1 μg/mL; InvivoGen) and G418 (500 μg/mL; Wako Pure Chemical Industries) in DMEM F12-Ham for 1 week.

Chromatin immunoprecipitation (ChIP)

ChIP was performed as follows. Briefly, Dynabeads Protein A (Thermo Fisher Scientific) was washed twice in western blocking reagent (Roche Applied Science) and incubated with blocking buffer and 10 μg of anti-HA antibody (Abcam), anti-Stat3 (Santa Cruz Biotechnology) or normal rabbit IgG (as control; Santa Cruz Biotechnology) overnight at 4°C. The next day, 5 × 10⁷ primary chondrocytes expressing dCas9 and gRNA were fixed with 1% formaldehyde for 10 min at room temperature (RT) to allow chromatin cross-linking, followed by quenching of the reaction with 0.125 M glycine. Cells were washed in phosphate-buffered saline (PBS) and collected. After centrifugation, the pellet was lysed in lysis buffer 1 (50 mM Hepes-KOH pH 7.5, 140 mM NaCl, 1 mM EDTA pH 8.0, 10% glycerol, 0.5% NP-40, 0.25% Triton X-100, and protease inhibitors) and rotated gently for 10 min at 4°C. After centrifugation, the pellet was lysed in lysis buffer 2 (10 mM Tris-HCl pH 8.0, 200 mM NaCl, 1 mM EDTA pH 8.0, 0.5 mM EGTA pH 8.0, and protease inhibitors) and rotated gently for 10 min at RT. After centrifugation, the pellet was lysed in lysis buffer 3 (10 mM Tris-HCl pH 8.0, 100 mM NaCl, 1 mM EDTA pH 8.0, 0.5 mM EGTA pH 8.0, 0.1% Na-deoxycholate, 0.5% N-lauroylsarcosine, and protease inhibitors) and sheared by sonication (Covaris) until the average length of the DNA fragments was ~1 kb. Two percent of the chromatin fragments were stored at –30°C for later use as whole-cell extracts. The remaining chromatin fragments were equally divided and rotated with Dynabeads protein A containing antibodies overnight at 4°C. Beads were washed repeatedly and suspended in elution buffer [1% sodium dodecyl sulfate (SDS)] for 1 h at 65°C. The supernatant was collected in fresh tubes and incubated overnight at 65°C to reverse the cross-links. Proteins were digested using proteinase K, and purified DNA was obtained using the MinElute PCR Purification kit (Qiagen).

ChIP-seq library preparation and data analysis

DNA libraries for next-generation sequencing were prepared using the TruSeq ChIP Sample Preparation kit (Illumina) using 10 ng of immunoprecipitated DNA or input DNA following the manufacturer’s instructions. The resulting libraries were sequenced on a MiSeq (Illumina) instrument. Each paired-end sequence read contained 100 bp, and only unique sequences were mapped to the mouse genome (mm10) using the Bowtie software (http://bowtie-bio.sourceforge.net/index.shtml) version 1.0.1. The mapped sam format files were converted into bam files using Samtools (http://samtools.sourceforge.net/) version 0.1.19, and immunoprecipitated DNA-enriched regions were detected using MACS (http://liulab.dfci.harvard.edu/MACS/) version 1.4.2 with default parameters. Sequence data were visualized using IGV (https://www.broadinstitute.org/igv/).

We detected 8.5 million raw sequence reads in the anti-HA PC ChIP-seq library and 4.2 million raw sequence reads in the anti-HA MEF ChIP-seq library using MiSeq. Next, we detected 6.2 million uniquely mapped reads in the anti-HA PC ChIP-seq library and 3.1 million uniquely mapped reads in the anti-HA MEF ChIP-seq library using bowtie software. CRISPR/dCas9 ChIP-seq enriched regions were detected by MACS with P-value cutoff for peak detection using P = 10–5, 10–8 and 10–15. All ChIP-seq experiments were performed in single replicates.

ChIP-MS sample preparation

ChIP for MS (2 × 10⁸ chondrocytes per assay) was performed using the protocol as described above in the ChIP section, and cross-linking to Dynabeads Protein A using BS³ (Thermo Fisher Scientific) was additionally performed before immunoprecipitation. Briefly, HA or IgG-coupled Dynabeads Protein A were washed twice in conjugation buffer (20 mM Sodium Phosphate, 0.15 M NaCl), after which the Dynabeads were resuspended in 5 mM BS³. After incubation at RT for 30 min with rotation, the cross-linking reaction was quenched by adding the quenching buffer (1 M Tris-HCl pH 7.5). After incubation at RT for 15 min with rotation, the cross-linked Dynabeads were washed twice with lysis buffer 3. The solution containing the chromatic fragments was equally divided and rotated with the cross-linked Dynabeads Protein A at 4°C overnight. The Dynabeads were washed repeatedly and suspended in elution buffer (1% SDS) for 1 h at 65°C. The supernatant was allowed to react at 95°C for 5 min and subjected to methanol–chloroform precipitation. The resulting pellets were dissolved in 0.1 M ammonium bicarbonate (pH 8.8) containing 7 M guanidine hydrochloride, reduced using 50 mM TCEP [Tris(2-carboxyethyl)phosphine], and subsequently alkylated using 100 mM iodoacetamide. After alkylation, samples were digested with lysyl-endopeptidase (Wako Pure Chemical Industries) for 4 h at 37°C and then further digested with trypsin (Thermo Fisher Scientific) for 14 h at 37°C. Digested samples were applied to a C18 spin column and eluted with 50% acetonitrile containing 0.1% formic acid. Eluted samples were dried and dissolved in 0.1 M ammonium bicarbonate (pH 8.8) containing 1.4 M guanidine hydrochloride.

LC-MS/MS analysis

Digested peptide samples were analyzed using a nanoscale LC–MS/MS system as previously described (Araki et al., 2016; Natsume et al., 2002). The peptide mixture was applied to a Mightysil-PR-18 (Kanto Chemical) frit-less column (45 × 0.150 mm ID) and separated using a 0–40% gradient of acetonitrile containing 0.1% formic acid for 80 min at a flow rate of 100 nL/min. Eluted peptides were sprayed directly into a mass spectrometer (Triple TOF 5600+; AB Sciex). MS and MS/MS spectra were obtained using the information-dependent mode. Up to 25 precursor ions above an intensity threshold of 50 counts/s were selected for MS/MS analyses from each survey scan. All MS/MS spectra were searched against protein sequences of RefSeq (NCBI) using the Protein Pilot software package (AB Sciex). Relative quantification of identified proteins was performed based on iBAQ (intensity-based absolute quantification) method (Schwanhausser et al., 2011) without calibration with the Universal Proteomics Standard.

Protein identification and quantification by mass spectrometry

Protein identification was performed using the ProteinPilot software (AB Sciex) using the NCBI non-redundant mouse protein data set (NCBI nr RefSeq Release 71, containing 124,899 entries). Identified proteins (identified peptide confidence > 95%) are listed in Table S1. Relative quantification of identified proteins was performed based on the iBAQ method without calibration using universal protein standards (Schwanhausser et al., 2011). Obtained iBAQ values for each protein are also listed in Table S1.

MS analysis using MRM (Multiple Reaction Monitoring)

MRM analysis was performed with a QTRAP5500 instrument (SCIEX) operated in the positive-ion mode as previously described (Kitazawa et al., 2017). The parameters were set as follows: spray voltage of 1350 V; curtain gas setting of 30; collision gas setting of 12; and interface heater temperature of 20°C. Collision energy (CE) was calculated with the following formulae: CE = (0.044 × m/z1) + 5.0 and CE = (0.05 × m/z1) + 4.0 (where m/z1 is the m/z of the precursor ion) for doubly and triply charged precursor ions, respectively. Collision cell exit potential (CXP) was calculated according to the formula: CXP = (0.0391 × m/z2) − 2.2334 (where m/z2 is the m/z of the fragment ion). The declustering potential was set to 50, and the entrance potential was set to 10. The resolution for Q1 and Q3 was set to “unit” (half-maximal peak width of 0.7 m/z). The scheduled MRM option was used for all data acquisition with a target scan time of 1.0 s and MRM detection windows of 400 s. Typically, 50 fmol of internal standard peptides from C13/N15 incorporated STAT3 was added to chromatin immunoprecipitate sample digest, and applied to the column. The peak area for the transitions was quantified using MultiQuant 2.0 (SCIEX). The peaks of light peptide was manually detected on the basis of retention time of heavy peptide (internal standard) peaks. The MQ4 algorithm was used to integrate and score peak groups. The absolute number of protein molecules was calculated by multiplication of the ratio of the light (sample) and heavy (internal standard) intensities summed for each transition and the known amount of recombinant protein.

Quantitative RT–PCR

Total RNA was isolated using the Reliaprep RNA Cell Miniprep System (Promega) following the manufacturer’s instructions. Reverse transcription was performed using ReverTra Ace (Toyobo). Quantitative real-time RT–PCR (qRT–PCR) was performed with the Thunderbird SYBR qPCR mix (Toyobo). Gapdh was used as internal control. Gene expression levels were quantified using the delta-delta CT method. Primer sequences are listed in Table S2.

Chromosome conformation capture (3C)

3C was performed as previously described (Hagege et al., 2007). Briefly, 1× 10⁷ primary chondrocytes were fixed with 1% formaldehyde for 10 min at room temperature, followed by quenching of the reaction with 0.125 M glycine. Cells were washed in PBS and collected. After centrifugation, the pellet was lysed in lysis buffer (10 mM Tris-HCl pH 7.5, 10 mM NaCl, 5 mM MgCl₂, 0.1 mM EGTA pH 8.0, 1× complete protease inhibitor) and incubated for 10 min on ice. After centrifugation, the pellet was lysed in 1.2× B buffer (Wako) containing 0.3% SDS and incubated for 1 h at 37°C while shaking. The solution was incubated with 2% Triton X-100 for 1 h at 37°C while shaking and incubated with 400U of HindIII at 37°C overnight while shaking. On the following day, the solution was incubated with 1.6% SDS for 25 min at 65°C and incubated with 1.15× ligation buffer and 1% Triton X-100 for 1 h at 37°C while shaking. After that, the sample was incubated with 5600 U of T4 DNA ligase (Takara) for 8 h at 16°C. After proteins digestion using proteinase K and RNA digestion using RNAase A, DNA was obtained by ethanol precipitation.

siRNA knockdown experiments

Small interfering RNAs (10 pm; Qiagen) were transfected using Lipofectamine RNAi Max (Invitrogen) into primary chondrocytes. After 36–48 h, total RNAs were prepared and subjected to qRT–PCR analysis.

Immunofluorescence staining

Primary chondrocytes were rinsed twice with PBS and fixed for 10 min at RT with 4% paraformaldehyde (PFA). Cells were rinsed rapidly twice with PBS and soaked in 0.2% Triton X-100 in PBS for 15 min at RT. Cells were then rinsed twice and soaked with Blocking One (Nacalai Tesque) for 1 h at RT, rinsed twice with PBS, and incubated with an anti-pStat3 (Tyr705) antibody (Cell Signaling) at 2 μg/mL in 5% Blocking One/PBS overnight at 4°C. The following day, cells were rinsed twice with PBS and incubated with Alexa Fluor 488 donkey anti-rabbit IgG (H+L) (Life Technologies) diluted to 1:1000 and Hoechst (Lonza) solution at 1 μg/mL in PBS for 1 h at RT.

Luciferase assay

ATDC5 cells were transfected with pcDNA3.1 Stat3 or its control, together with pGL4-Sox9 promoter or pGL4-Sox9 promoter bearing the 1×, 4× RCSE enhancer and pGL4.74 (Promega), using Fugene HD (Promega). Luciferase activity was measured 36 to 48 h after transfection using a Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer’s instructions.

Also, the RCSE and Sox9 promoter were cloned into pNL1.1 vector (Promega). Deletion of putative Stat3 binding sites was performed via inverse PCR. ATDC5 cells were transfected with pcDNA3.1 Stat3 or its control, together with pNL1.1 RCSE or its mutants. Nano luciferase activity was measured 48 h after transfection using a Nano-Glo Luciferase Assay System or Nano-Glo Dual-Luciferase Reporter Assay (Promega) following the manufacturer’s instructions.

Western blot analysis

After measurement of luciferase activity, cell lysates were collected and denatured by Sample buffer at 95ºC for 5min. Proteins were separated by SDS-PAGE and transferred to PVDF membrane. After blocking with Blocking One (Nacalai tesque) for 1h at RT, membranes were incubated with anti-β-actin (SIGMA, 1/4,000 dilution) or anti-STAT3 (Cell Signaling Technology; 1/2,000 dilution) at 4ºC overnight, followed by HRP-conjugated anti-mouse IgG (GE Healthcare) or anti-rabbit IgG (GE Healthcare). Bound antibodies were visualized by Pierce ECL Western Blotting Substrate (Thermo Fischer Scientific).

X-gal staining

Embryos were fixed for 30 min at 4°C with 4% PFA. After washing thrice with 1 mM MgCl₂/PBS, embryos were stained with 0.1% X-gal in staining buffer [0.01% sodium acetate, 0.02% NP-40, 1 mM MgCl₂, 5 mM K₃Fe(CN)₆, and 5 mM K₄Fe(CN)₆] at 37°C for 5 h to overnight. The following day, embryos were rinsed thrice with 1 mM EDTA/PBS and fixed again with 4% PFA for 1.5 h at 4°C. Embryos were observed after washing thrice with PBS. For reproducibility, staining was performed on three independent embryos.

Histochemical analysis

For histological analysis, 10-day-old mice were euthanized using CO2. Thoracic samples were fixed in 4% PFA/PBS overnight, dehydrated in a series of increasing EtOH concentrations, and subsequently embedded in paraffin. Samples were then processed for paraffin sectioning (5 μm) and stained with hematoxylin and eosin (H&E) or alcian blue and eosin to observe the outline of the cartilage region.

Alcian blue/alizarin red skeletal staining

Mouse embryos were fixed overnight at RT in 100% MtOH and then replaced for 4 days at RT with 95% EtOH. Embryos were stained with alcian blue solution (0.015%; Sigma–Aldrich) overnight at RT. After washing twice with 95% EtOH, embryos were replaced with 2% KOH overnight at RT and subsequently stained with alizarin red solution (0.0075%; Sigma–Aldrich) overnight at RT. Skeletons were replaced with 1% KOH/20% glycerol for 3 days and passed through increasing concentrations of 1:1 glycerol/ethanol solution (20%, 50%, 100%) each for 1 day. Samples were stored in glycerol.

Micro CT

Wild-type or RCSE-null mice (10 weeks old) were analyzed via micro CT (SMX-100CT, Shimazu).

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical details and statistical significance values are indicated in the text, figure legends, or Method Details. Error bars represent standard error of the mean (SEM) from either independent experiments or samples.

DATA AND SOFTWARE AVAILABILITY

All raw and processed ChIP-seq data are available in the DDBJ BioProject database: PRJDB5833.

KEY RESOURCE TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-HA tag antibody – ChIP Grade	Abcam	Cat#ab912; RRID:AB_307019
ANTI STAT3 antibody	Santa Cruz Biotechnology	Cat#C-20 (sc482); RRID:AB_632440
ANTI STAT3 antibody	Cell Signaling Technology	Cat#4904; RRID:AB_331269
Phospho - Stat3 (Tyr705) antibody	Cell Signaling Technology	Cat#9131; RRID:AB_331586
Anti-β-actin antibody	SIGMA	Cat#A2228; RRID:AB_476697
Anti-Histone H3 (mono methyl K4) antibody – ChIP Grade	Abcam	Cat#ab8895; RRID:AB_306847
Histone H3K27ac antibody (pAb)	Active Motif	Cat#39133; RRID:AB_256216
Bacterial and Virus Strains
Competent Cells (DH5α)	ZYMO Research	Cat#T3002
Biological Samples
Mouse: costal chondrocytes (E16.5)	Sankyo lab Service	N/A
Mouse: murine embryonic fibroblasts (MEF) (E16.5)	Sankyo lab Service	N/A
Chemicals, Peptides, and Recombinant Proteins
G418	Thermo-Fisher	Cat#1181231
Puromycin	InvivoGen	Cat#ant-pr-1
Dynabeads protein A	VERITAS	Cat# DB2002
HindIII	NIPPON GENE	Cat#311–01163
T4 DNA ligase	New England Biolabs	Cat#M0202S
BMP-2	Pepro Tech	Cat#120–02
Critical Commercial Assays
MiSeq Reagent Kit v2	Illumina	Cat#MS-22–2002
MiSeq Reagent Kit v3	Illumina	Cat#MS-22–3001
NEBNext ChIP-seq library prep kit	New England Biolabs	Cat#E6240L
TruSeq ChIP Library Preparation Kit	Illumina	Cat#IP-202–212
mMessage mMachine T7 Transcription Kit	Ambion	Cat#AM1344
Multisite Gateway Three-Fragment Vector Construction Kit	Invitrogen	Cat#12537023
Deposited Data
Raw and processed sequencing data	This paper	DDBJ BioProject: PRJDB5833
iBAQ-based mass spectrometry	This paper	Table S1
Experimental Models: Cell Lines
Mouse: ATDC5 cells	Atsumi et al., 1990	N/A
Mouse: C3H/2T1/2 cells	Denker et al., 1999	N/A
Human: PLAT-E cells	Morita et al., 2000	N/A
Experimental Models: Organisms/Strains
Mouse: C57BL/6J	Sankyo lab Service	N/A
Mouse: RCSE knockout	This paper	N/A
Mouse: Sox9 promoter transgenic	This paper	N/A
Mouse: 4xRCSE + Sox9 promoter transgenic	This paper	N/A
Mouse: Mx-Cre/STAT3 flox	Oike et al., 2017.	N/A
Oligonucleotides
Primers for qRT-PCR, 3C and genotyping	This paper	Table S2
gRNAs	This paper	Table S2
shRNAs	This paper	Table S2
siRNAs	This paper	Table S2
Recombinant DNA
pSIREN-RetroQ-sgRNA-Neo	This paper	N/A
pMIGR-dCas9-IRES-Puro	This paper	N/A
pMIGR-dCas9-SIN3A-IRES-Puro	This paper	N/A
pBluescript II-Sox9 promoter-LacZ	Dr. Shuji Takada, Department of Systems BioMedicine, National Research Institute for Child Health and Development.	N/A
pBluescript II-Sox9 promoter-RCSE-LacZ	This paper	N/A
pcDNA3.1-Stat3	This paper	N/A
pGL4 RCSE-Sox9 promoter	This paper	N/A
pNL1.1 RCSE-Sox9 promoter	This paper	N/A
p2A1	Chlontech	Cat#631530
Software and Algorithms
Bowtie	Langmead and Salzberg, 2012	http://bowtie-bio.sourceforge.net/index.shtml
Samtools	Li et al., 2009	http://samtools.sourceforge.net/
MACS	Zhang et al., 2008	http://liulab.dfci.harvard.edu/MACS/
IGV	James et al., 2011	https://www.broadinstitute.org/igv/
Other
Mightysil-PR-18	Kanto Chemical	https://products.kanto.co.jp/web/index.cgi?c=t_product_table&pk=209
Triple TOF 5600+	Sciex	https://sciex.com/products/mass-spectrometers/qtof-systems/tripletof-systems/tripletof-5600-system
QTRAP5500	Sciex	https://sciex.com/products/mass-spectrometers/qtrap-systems/qtrap-5500-system

Highlights.

• CRISPR/Cas9-ChIP-seq approach revealed Sox9 enhancer candidate.

• SIN3A-dCas9 mediated epigenetic silencing and 3C confirmed far-upstream Sox9 enhancer.

• CRISPR/Cas9-mediated enhancer deletion mice showed ACD-like phenotype

• CRIPSR/dCas9-ChIP-MS identified STAT3 as a trans-acting mediator of Sox9 enhancer