Chd4 and associated proteins function as corepressors of Sox9 expression during BMP-2–induced chondrogenesis


  • Fenyong Sun,

    1. Department of Clinical Laboratory Medicine, Tenth People's Hospital of Tongji University, Shanghai, China
    Search for more papers by this author
    • The first two authors contributed equally to this work.
  • Qingyuan Yang,

    1. Department of Clinical Laboratory Medicine, Tenth People's Hospital of Tongji University, Shanghai, China
    Search for more papers by this author
    • The first two authors contributed equally to this work.
  • Wenhao Weng,

    1. Department of Clinical Laboratory Medicine, Tenth People's Hospital of Tongji University, Shanghai, China
    Search for more papers by this author
  • Yue Zhang,

    1. The Central Laboratory, Tenth People's Hospital of Tongji University, Shanghai, China
    Search for more papers by this author
  • Yongchun Yu,

    1. The Central Laboratory, Tenth People's Hospital of Tongji University, Shanghai, China
    Search for more papers by this author
  • An Hong,

    1. Institute of Biology Engineering, Key Laboratory of Ministry of Education for Genetic Engineering, College of Life Science and Technology, Jinan University, Guangzhou City, Guangdong Province, China
    Search for more papers by this author
  • Yuhua Ji,

    1. Key Laboratory of Neuroregeneration, Nantong University, Nantong City, Jiangsu Province, China
    2. Institute of Tissue Transplantation and Immunology, College of Life Science and Technology, Jinan University, Guangzhou City, Guangdong Province, China
    Search for more papers by this author
  • Qiuhui Pan PhD

    Corresponding author
    • The Central Laboratory, Tenth People's Hospital of Tongji University, Shanghai, China
    Search for more papers by this author

Address correspondence to: Qiuhui Pan, PhD, The Central Laboratory, Tenth People's Hospital of Tongji University, No. 301, Yanchang Middle Road, Zhabei District, Shanghai 200072, China. E-mail: Yuhua Ji. E-mail:


Mouse embryonic fibroblasts (MEFs) differentiate into fully functional chondrocytes in response to bone morphogenetic protein-2 (BMP-2). However, the comprehensive proteomic aspect of BMP-2–induced chondrogenesis remains unknown. We took advantage of quantitative proteomic analysis based on isobaric tag for relative and absolute quantitation (iTRAQ) and on-line 2D nano-liquid chromatography/tandem mass spectrometry (LC/MS/MS) to identify proteins differentially expressed during BMP-2–induced chondrogenic differentiation of MEFs. We found 85 downregulated proteins, and ingenuity pathways analysis (IPA) revealed a protein-protein network with chromodomain-helicase-DNA-binding protein 4 (Chd4) in the center. Chromatin immunoprecipitation (ChIP) and nuclease hypersensitivity assays showed that Chd4, interacting with Hdac1/2, cooperates with its related proteins Kap1 and Cbx1 to bind at −207/−148 of the Sox9 promoter. We also provided evidence that let-7a targets the 3'UTR of Chd4 to promote chondrogenesis of MEFs. Together, our findings indicate that BMP-2 induced the upregulation of let-7a, targeting Chd4 and positively controlling the chondrogenic differentiation of MEFs. These findings illustrate epigenetic regulation of the chondrogenic differentiation process and also expand the understanding of the involved intracellular mechanisms.


Mouse embryonic fibroblasts (MEFs) possess the ability of self-renewal and multipotential differentiation. In vitro, MEFs can be induced into several cell types, including adipocytes,[1] myocytes,[2] and chondrocytes.[3] Compared with embryonic stem cells (ESCs) from the inner cell mass of blastocytes, MEFs are easier to isolate and culture. For this reason, MEFs are an ideal cell model for the study of cell commitment and differentiation. It has been reported that, in response to BMP-2, MEFs can enter and complete the program of chondrogenic differentiation ex vivo, changing from undifferentiated progenitor cells to mature hypertrophic chondrocytes.[3] In our previous study,[4] we elaborated that Sox9 is activated through the BMP pathway and a CCAAT box in the proximal promoter. However, despite our findings, the exact molecular events occurring during this process remain largely unclear, especially in regard to proteomic interactions and regulation.

Chromodomain helicase DNA-binding protein 4 (CHD4), also known as Mi-2beta, belongs to the class II subfamily of CHD ATPases and is well characterized as a key catalytic subunit of the NuRD transcriptional repressor complex.[5, 6] Coupling of helicase and HDAC functions makes NuRD unique among chromatin remodeling complexes.[7] In humans, the Mi-2/NuRD complex is involved in the negative regulation of lymphocyte differentiation and cell fate.[8, 9] Furthermore, studies have shown Mi-2/NuRD is of great importance, both for maintaining hematopoietic stem cell (HSC) pools and in normal lineage progression.[6]

MicroRNAs (miRNAs) are a class of approximately 22 nt single-stranded non-protein coding RNAs that play critical roles in controlling cell processes such as cell proliferation, apoptosis, and differentiation.[10, 11] Chondrogenic differentiation is a complex process under the concerted regulation of various cytokines, growth factors, and multiple differential factors. In support of this view, some studies have explored the mechanisms of miRNA regulation of chondrogenesis. miR-140 and its validated downstream target gene, histone deacetylase 4 (HDAC4), play an important role in chondrocyte proliferation and differentiation.[12] Other research has shown that miR-199* might affect its target gene, Smad1, to regulate early chondrogenic differentiation.[13] Collectively, current evidence suggests that some portion of miRNAs is likely to be of functional importance in the regulation of cell differentiation.

Therefore, to gain a comprehensive understanding of the molecular mechanisms underlying BMP-2–induced chondrogenesis, we applied iTRAQ labeling coupled with on-line 2D LC/MS/MS proteomic technology to quantitatively assess the protein expression profile of these cells in vitro. As a consequence, we identified a proteomic interaction network with Chd4 at the center. To our knowledge, this is the first report that Chd4 functions as a repressor in BMP-2–induced chondrogenesis of primary MEFs. Our data also showed that Chd4 is recruited to the TGGCTG box in Sox9 promoter and modifies the chromatin structure to initiate and maintain gene repression in primary MEFs. Interestingly, we also found that during BMP-2–induced chondrogenesis, sustained upregulation of let-7a specifically suppressed Chd4 expression, resulting in the activation of the Sox9 gene, a known key transcription factor required for chondrogenesis.[4] Overall, these findings deepen our understanding of the underlying intracellular mechanisms that modulate chondrogenic differentiation of MSCs and, in turn, help promote their potential application in human cartilage diseases.

Materials and Methods

Methods for cell culture, chondrocyte differentiation, and Alcian blue staining are all referred to in our previous report.[4]

iTRAQ labeling, sample cleaning, and desalting

Cultures after treatment with BMP-2 were washed thoroughly with ice-cold phosphate-buffered saline (PBS). The subcellular nuclear fraction was enriched using a Qproteome Cell Compartment Kit (Qiagen, Hilden, Germany). iTRAQ labeling and sample cleaning were performed as described.[14] In brief, 100 µg of nuclear proteins from MEFs and chondrogenic-differentiated MEFs were precipitated with ice-cold acetone overnight at −20°C, and pellets were dissolved, denatured, alkylated, digested with trypsin (1:20 [w/w], 37°C for 18 hours), and labeled according to the manufacturer's instructions (Applied Biosystems, Foster City, CA, USA). MEFs and chondrogenic-differentiated MEFs were labeled with 114 and 117 iTRAQ reagents, respectively, and three independent biological replications were performed. Before on-line 2D nanoscale LC/MS/MS analysis, iTRAQ-labeled samples were cleaned up and desalted. A cation exchange cartridge system (Applied Biosystems) was used to remove the reducing reagent, SDS, excessive iTRAQ reagents, undigested proteins, and trypsin from the labeled sample mixtures that would interfere with the LC/MS/MS analysis. Subsequently, elutes of cation exchange were desalted on a C18 reversed-phase column (4.6 mm inner diameter × 150 mm, 5 µm, 80 Å; Agilent, Waldbronn, Germany).

On-line 2D nano LC/MS/MS

Elutes were separated by gel electrophoresis, and whole lanes were excised into several regions, digested, and analyzed by 2D LC/MS/MS. As previously described,[14] 2D nano-LC/MS/MS analyses were conducted on a nano-HPLC system (Agilent Technologies, Santa Clara, CA, USA) coupled to a hybrid Q-TOF mass spectrometer (QSTAR XL) equipped with a nano-ESI source (Applied Biosystems) and a nano-ESI tip (Picotip; New Objective, Woburn, MA, USA). Analyst 1.1 software was used to control the QSTAR XL mass spectrometer and nano-HPLC system and to acquire mass spectral data. Vacuum-dried iTRAQ-labeled and purified peptides were reconstituted in phase A and injected at a flow rate of 10 µL/minute onto a high-resolution strong cation exchange (SCX) column (Bio-SCX, 300 µm inner diameter × 35 mm; Agilent Technologies), which was on line with a C18 precolumn (PepMap, 300 µm inner diameter × 5 mm; LC Packings, Vernon Hills, IL, USA). After loading, the SCX column and C18 precolumn were flushed with a 12-step gradient of sodium chloride solution (0, 20, 30, 40, 50, 60, 70, 80, 90, 100, 200, and 400 mM) for 5 minutes and phase A for 10 minutes at a flow rate of 10 µL/minute. Afterwards, the precolumn was switched on line with a nanoflow reversed-phase column (100 µm i.d. 150 mm, 3.5 µm, 100 Å; Agilent Technologies). The peptides concentrated and desalted on the precolumn were separated on this nanoflow C18 column at a flow rate of 300 nL/minute, and the gradient induced a linear increase of 4% to 40% acetonitrile in 0.1% formic acid over 90 minutes. Eluted peptides were electrosprayed through a noncoated silica tip (Picotip; New Objective).

Protein identification and relative quantitation

The MS raw data were analyzed predominantly as described.[14] In brief, ProteinPilot Software 3.0.1 (Software Revision Number: 67476; Applied Biosystems) was used to identify and quantify the peptides and proteins. The complete set of raw data files (*.wiff) were searched against the nonredundant International Protein Index (IPI) database (mouse v3.62, 56,733 entries) using the Paragon and ProGroup algorithms (Applied Biosystems). In this study, a protein with an “unused” confidence threshold (ProtScore) >1.3% was reported, and the corresponding False Discovery Rate (FDR) was less than 1%. The relative abundance of proteins was calculated based on the individual peptide ratio (iTRAQ 117 to iTRAQ 114) identified in three independent iTRAQ analyses. The threshold values generally set for significant upregulation and downregulation were ≥2.0 or ≤0.5.

Computational and systems-level analysis

The bioinformatic analysis of differentially expressed proteins was achieved using Ingenuity Pathways Analysis (IPA) software (version 6.3; Ingenuity Systems, Redwood City, CA, USA). First, IPA was used to assign the differential proteins into different molecular and cellular functional classes based upon the underlying biological evidence from the IPA literature database. Then, the differential proteins and the corresponding expression values were uploaded into the IPA software for generating networks and overlaid onto a global molecular network developed from information contained in the Ingenuity Pathways Knowledge Base. For the functional annotation analysis of components in the Chd4-centered interactome, Gene Ontology analysis[15] was put into practice.

siRNA and miRNA transfection

SMARTpool siRNAs against Chd4 (L-045446-01-0005; Thermo Fisher Scientific, Rockford, IL, USA) and miRNA mimics were transfected into the primary MEFs using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA), as described in the manufacturer's instructions.

Construction of reporter plasmids and luciferase assays

The Chd4 3'UTR, which contains a putative let-7a binding site, was amplified from primary MEF cDNA. PCR fragments were cloned into the XhoI/NotI sites of the psiCHECK vector (Promega, Madison, WI, USA). Substitution mutations of the binding site constructs were generated using a PCR-based site-directed mutagenesis kit (TaKaRa Bio, Shiga, Japan), using the corresponding plasmids as the templates. The reporter plasmids were transiently transfected into MEFs using Lipofectamine 2000. psiCHECK plasmids containing the Chd4 3'UTR were cotransfected with let-7a mimics, with or without the let-7a inhibitor. After further cultivation in an incubator at 37°C, 5% CO2, for 24 hours, the transfected cells were harvested, lysed, and centrifuged, and the pellet was subjected to luciferase assays. Luciferase activity was measured as chemiluminescence in a luminometer (PerkinElmer, Waltham, MA, USA) using the dual-luciferase reporter assay system (Promega) according to the manufacturer's protocol. All transfections were performed in triplicate, and the results were expressed as the means ± standard deviations.

Western blot analyses and antibodies

The methods for Western blot analysis were the same as used in our previous report.[4] Briefly, cells were harvested and lysed in 0.5-mL lysis buffer (10 mM Tris-HCl pH 7.6, 5 mM EDTA, 50 mM NaCl, 30 mM sodium pyrophosphate, 50 mM NaF, 0.1 mM Na3VO4, 1% Triton X-100, 1 mM PMSF, and protease inhibitor cocktail tablet [Roche Molecular Biochemicals, Mannheim, Germany]). A total of 30 ug of protein was processed for SDS-PAGE, followed by electrophoretic transfer to Immobilon P membranes (Millipore, Bedford, MA, USA). The blots were blocked with 5% nonfat milk in Tris-buffered saline (TBS, pH 7.4) for 1 hour and then incubated with targeting antibodies (Supplemental Table S1) diluted in 5% nonfat milk. The secondary antibodies were visualized using enhanced chemiluminescence and recorded on X-ray films (Fuji Medical, Tokyo, Japan).

Electrophoretic mobility shift assay (EMSA)

Nuclear extracts of MEFs were collected using a kit from Active Motif (Tokyo, Japan). Nuclear extracts (3 mg) were incubated on ice for 15 minutes in 15 mL reaction mixture containing 4% glycerol, 1 mM MgCl2, 0.5 mM dithiothreitol (DTT), 0.5 mM ethylenediaminetetraacetic acid (EDTA), 50 mM NaCl, 10 mM Tris-HCl (pH 7.5), and 2.0 mg poly(dI-dC) with or without molar excess of unlabelled DNA competitors, followed by addition of the radiolabeled probe. Chd4 binding was confirmed using a specific antibody against Chd4. Nuclear extracts (3 mg) were incubated with or without antibody (1 mg) at room temperature for 25 minutes, and the probe was then added and incubated for an additional 20 minutes. All DNA–protein complexes were resolved by electrophoresis on 5% native polyacrylamide gels. Sequences of the variable probes are shown in Fig. 3B.

Chromatin immunoprecipitation (ChIP) and Re-ChIP analyses

ChIP analysis was performed using a kit from Active Motif. Briefly, cells were fixed with 1% formaldehyde, washed with precooled PBS, and lysed in buffer. The lysates were pelleted and precleared. The protein–DNA complexes were incubated with protein A beads, the protein–DNA complexes were eluted in 1% SDS/0.1 M NaHCO3, and the cross-links were reversed at 65°. In the Re-ChIP experiments, complexes were sequentially immunoprecipitated with different antibodies. In the end, the cross-linked DNA was recovered by phenol–chloroform extraction and ethanol precipitation and then subjected to semiquantitative PCR analysis. Sequences of specific primers are provided in Supplemental Table S2.

Nuclease hypersensitivity assays

These tests were performed the same as in our previous report.[16] In short, aliquots of nuclei were digested with NheI (5 U/mL). The reaction was stopped by addition of stop buffer. After purification, DNA (15 ug) was digested completely with SalI or DraI. The digestion products were detected by Southern blot analysis using the indirect end-labeling method.

Statistical analysis

Statistical evaluations were conducted using t test or ANOVA analysis, and p values that were less than 0.05 were considered statistically significant.


Identification of proteins differentially expressed during BMP-2–induced chondrogenesis

To investigate differentially expressed proteins during BMP-2–induced chondrogenesis, we performed a proteomic screen on the basis of novel iTRAQ labeling and LC/MS/MS analysis (Supplemental Fig. S1). The quantitative proteomic analysis resulted in the identification of 502 nuclear proteins, among which 495 were quantified. The corresponding FDR for protein identification was less than 1%, and detailed information for the identified proteins is provided in Supplemental Table S3. The thresholds for upregulation and downregulation were ≥2.0 and ≤0.5, respectively, resulting in 85 proteins downregulated and 4 proteins upregulated during BMP-2–induced chondrogenic differentiation of MEFs.

Subsequently, the downregulated proteins were assigned to different molecular and cellular functional classes based on the IPA literature database. Interestingly, the top category was RNA post-transcriptional modification (36/85, p = 1.94 × 10−37 to 4.05 × 10−2; Supplemental Fig. S2A). Moreover, the top physiology category was skeletal and muscular system development and function (Supplemental Fig. S2B), indicating an essential role for these downregulated proteins in chondrogenic differentiation. Next, MS results from each lane were searched against IPI, as shown in Supplemental Fig. S3A. The sequence IEENSLK[IT4]EEESTEGEK[IT4] was identified as Chd4, and the quantitation of this Chd4 sequence indicated it was downregulated, as illustrated in Supplemental Fig. S3B. By uploading the corresponding expression values of the differential proteins into IPA to generate interaction networks, we obtained a signaling network supported by the available literature and/or canonical information stored in the database. Interestingly, Chd4 was at the center of these diverse proteins (Fig. 1A). To further confirm these results, 20 differentially expressed proteins were verified by Western blot analysis (Fig. 1B).

Figure 1.

Predicted Chd4-centered proteomic network. The predicted signaling network, with a central role for Chd4, was obtained from the Ingenuity Pathways Knowledge Base by uploading the corresponding expression values of the differentially expressed proteins into the IPA. (A) In this map, various shapes represent different functional classes of the products, and the deeper the color, the more reduced the proteins are during the process. Moreover, the solid lines indicate direct interactions between proteins, whereas dashed lines indicate indirect interactions. All lines are supported by at least one reference from the literature or from canonical information stored in the Ingenuity Pathways Knowledge Base. Arrows indicate activation. (B) Western blot verification of some of the downregulated proteins identified during BMP-2–induced chondrogenesis.

We then obtained the functional annotation for those 30 proteins (Supplemental Table S4) in the interactome by GO analysis. First, we analyzed these proteins for molecular function through PANTHER database, and the results indicated that 80% (24/30, p = 9.61 × 10−16) of the proteins were involved in nucleic acid binding, 33% in mRNA processing (10/30, p = 4.11 × 10−13), and 30% in mRNA splicing (9/30, p = 1.99 × 10−12; Supplemental Table S5). Then we determined the biological processes frequently involved through GOTERM database, and detected 53% (16/30, p = 3.81 × 10−6) involved in regulation of transcription, 40% (12/30, p = 2.75 × 10−10) in RNA processing, and 33% (10/30, p = 1.06 × 10−10) in RNA splicing (Supplemental Table S6). We also analyzed the domain composition of the Chd4-centered proteins via the INTERPRO database (Supplemental Table S7). Highly represented domains were RNA recognition motif, DNA binding SAP, chromo domain, zinc fingers, and DZF domains, most of which were either involved in ATP-dependent chromatin remodeling or binding RNA related to the regulation of gene expression in cell differentiation.[17-20]

Chd4 inhibits chondrogenic differentiation by downregulating Sox9 by binding the TGGCTG box

Our proteomic results indicated that Chd4 was downregulated during the chondrogenic induction process after BMP-2 treatment, whereas the expression of Sox9, a critical transcription factor involved in multiple steps of chondrogenesis, has been reported to be upregulated during chondrogenic differentiation.[16] We thus investigated whether there was a potential link between Chd4 and Sox9 during the chondrogenic differentiation process. We first designed three different siRNAs against Chd4. Western blot analysis demonstrated that the third siRNA against Chd4 was the most efficient (Fig. 2A); therefore, we chose to transfect this siRNA for subsequent experiments. We detected the expression of Sox9 by immunoblot analysis. The results showed that Sox9 was greatly upregulated in MEFs after deletion of Chd4 after BMP-2 treatment (Fig. 2B), suggesting that Chd4 plays a functional role in chondrogenic differentiation of MEFs.

Figure 2.

Chd4 functions as a repressor of chondrogenic differentiation. (A) Western blot analysis confirms the efficiency of three different siRNAs against Chd4. (B) Primary MEFs were treated with a negative control (NC) siRNA or the Chd4 siRNA with or without BMP-2 treatment for 5 days. The nuclear extracts were then subjected to Western blot analysis for the detection of Sox9. (C, D) Primary MEFs were transiently transfected with Chd4 siRNAs or an NC siRNA in the absence/presence of 200 ng/mL BMP-2. The whole-cell extracts were then subjected to Western blot analysis for the detection of Col2a1 (B) or Alcain blue staining (C) for identification of sulfated glycosaminoglycans.

To further study the biological functions of Chd4 during chondrogenesis, we detected the expression of Col2a1, a typical marker of cell differentiation, by Western blot analysis and also performed Alcian blue staining to identify sulfated glycosaminoglycans. These experiments were performed after MEFs were transiently transfected with Chd4 siRNAs or a negative control (NC) siRNA in the absence/presence of 200 ng/mL BMP-2 as previously reported.[4] The results demonstrated that knockdown of Chd4 remarkably promoted chondrogenic differentiation in the presence of BMP-2, implying a negative role for Chd4 in MEF chondrogenic differentiation (Fig. 2C, D).

Subsequently, we analyzed the promoter of Sox9 via ChIP assays to evaluate whether Chd4 was bound to the Sox9 promoter in primary MEFs. As shown in Fig. 3A, less Chd4 was enriched at the region containing the core promoter of Sox9 (+1k region)[16] when MEFs were treated with the Chd4 siRNA compared with the NC siRNA. However, no signals were detectable at the other regions (−3k, +2k, 3'UTR regions).

We then aimed to identify the exact binding site of Chd4 in the Sox9 promoter via EMSA. We prepared several oligonucleotide probes covering nts −207 to +17 (Fig. 3B), including the Sox9 core promoter. As shown in Fig. 3C, lane 2, after incubation, a shift band was detected using the probe −207/−148, which was specifically opposed by a 20-fold competitor (Fig. 3C, lane 3). No binding proteins were detected when using other probes (data not shown).

Figure 3.

Chd4 inhibits chondrogenic differentiation by binding the TGGCTG box in the Sox9 promoter. (A) The diagram shows the locations of the different PCR primers used for ChIP assays (top panel). Formaldehyde cross-linked chromatin was prepared from the BMP-2–induced MEFs and immunoprecipitated with antibodies to NF-Y, p300, or Chd4. PCR products from the ChIP assays were visualized by gel electrophoresis. As a negative control, the protein–DNA complexes were incubated with nonspecific IgG. (B) Schematic representation of the different probes used to locate the exact binding site of Chd4 within the Sox9 promoter using the indirect end-labeling EMSA technique. The NF-Y binding site is within the box. (C) The probe located at –207/–148 generated a shift band that was opposed by 20-fold excess of unlabeled probe (see the solid arrow). The following incubation of the complex with Chd4 antibody resulted in a supershift band (see the hollow arrow), taking nonspecific IgG as a control. (D) The probe –207/–148 was divided into three detailed screening probes, A, B, and C, and the probe A was further mutated at a series of regions, resulting in seven mutant probes, namely MA1, MA2, MA3, MA4, MA5, MA6, and MA7. (E) The probe A generated a specific shift band, and its corresponding competitor could specifically weaken the band intensity. Meanwhile, the other two probes and their competitors generate no shift band. The protein-probe A complex was further verified by supershift assays; a supershift band was specially generated by Chd4 antibody, as the hollow arrow points. (F) Seven mutant probe As were applied to EMSA assays to establish the exact binding site of Chd4 in the Sox9 promoter. The MA4 detected no shift band.

To further explore if Chd4 was involved in the formation of the protein–DNA complexes, we performed supershift assays using an anti-Chd4 antibody, taking IgG as a control. As illustrated in Fig. 3C, lanes 4 and 5, a supershift band was specifically formed in the presence of this antibody. Moreover, we divided the −207/−148 probe into three shorter probes, A, B, and C, to further investigate the more accurate binding site for Chd4 (Fig. 3D). Interestingly, we found that only probe A (−207/−183) could detect a shift band, and the band intensity was significantly weakened by the specific competitor to probe A (Fig. 3E, lanes 1 to 6). Additionally, we corroborated that Chd4 was involved in this binding complex (Fig. 3E, lanes 7 to 9). Furthermore, we applied mutant assays to establish the specificity of this interaction. We designed seven mutated probe As (Fig. 3D and Supplemental Fig. S4), and the EMSA results (Fig. 3F) showed that no shift band was detectable for probe MA4 (mutant probe A4), indicating that the TGGCTG box might be the specific binding site for Chd4 in the Sox9 promoter. Collectively, we concluded that Chd4 transcriptionally inhibited MEF differentiation by downregulating Sox9 through the TGGCTG box.

Afterwards, to further functionally confirm the binding specificity, we designed a Sox9-promoter-luc reporter and transiently transfected the plasmids into MEFs with BMP-2 treatment. Unexpectedly, the plasmids showed the proper basal promoter function, but upregulated promoter activity was undetectable (data not shown).

Chd4 condenses the chromatin structure of the Sox9 promoter region

Chd4 is a key component of the NuRD complex that contains ATP-dependent nucleosome disruption activity as well as histone deacetylase activity, which is usually associated with transcriptional repression.[21, 22] In our previous work,[16] we found that the binding of the NF-Y-p300 complex to the Sox9 promoter along with Pcaf and RNA polymerase II regulated chromatin remodeling and promoted BMP-2–induced chondrogenesis. Interestingly, here we discovered that knockdown of Chd4 could slightly increase the binding of NF-Y and p300 to the core promoter of Sox9 (Fig. 3A, bottom panel). Additionally, ChIP assays revealed the binding of Hdac1/2 with the promoter of Sox9 in vivo (Fig. 4A). It is reported that Hdac1/2 are also components of the NuRD complex. Therefore, we inferred that Chd4 perhaps functioned negatively during chromatin remodeling and histone modification at the core promoter of Sox9 in a complex including Hdac1/2. Consistent with this conjecture, results from our ChIP analyses demonstrated a remarkable decrease in the levels of the Sox9 promoter immunoprecipitated using antibodies to Hdac1 or Hdac2 after knockdown of Chd4, with or without BMP-2 treatment (Fig. 4A). Moreover, knockdown of Chd4 combined with BMP2 induction resulted in the increased binding of euchromatic markers with the Sox9 core promoter, including tetra-acetyl-H4 (ac-H4), trimethyl-H3-K4 (H3K4Me3). Consistently, these conditions also resulted in reduced binding of the heterochromatin marker dimethyl-H3-K9 (H3K9Me2),[23] which functions in chromatin silencing (Fig. 4A), implying an essential role of Chd4 in the epigenetic regulation of chromatin structure.

Figure 4.

Chd4 condenses the chromatin structure of the Sox9 promoter region (+1k). (A) Primary MEFs were treated with NC siRNA or Chd4 siRNA in the absence/presence of BMP-2. After DNA recovery, the precipitates were evaluated by real-time PCR for the relative enrichment over the negative control antibody at the +1k region. All results represent the means of three independent experiments ± standard deviation. (B) Primary MEFs were treated with NC siRNA or Chd4 siRNA, with or without BMP-2 treatment. Nuclei were then purified and digested with 25 to 200 U of NheI. Purified DNA was digested with DraI and detected by Southern blot analysis. (C) Nuclear extracts from primary MEFs were precipitated with anti-Chd4 or anti-Hdac1/2 antibodies. Co-immunoprecipitated proteins were analyzed with anti-Hdac1/2 or anti-Chd4 antibodies. Nonspecific IgG was used as the control for precipitation, and whole-cell lysis was used as the control for the Western blot analysis.

We also performed an NheI accessibility assay to evaluate the status of the chromatin structure of Sox9. The NheI restriction site is located −79 bp upstream of the transcription starting site and adjacent to the core promoter of the Sox9 gene.[16] The results implied that knockdown of Chd4 made MEFs more sensitive to NheI, illustrating an open status of the chromatin structure in this region (Fig. 4B, lane 3). Furthermore, BMP-2 could significantly strengthen the effect of the Chd4 siRNA (Fig. 4B, lane 5). In addition, IP assays of extracted proteins from primary MEFs showed that Chd4 could interact with both Hdac1 and Hdac2 in vitro, which have crucial roles in chromatin compaction and transcriptional repression (Fig. 4C). With regard to the above results, we concluded that within the Sox9 promoter region, binding of Chd4 condenses the chromatin structure by recruiting Hdac1/2 and competitively inhibit the recruitment of the NF-Y-p-300 complex to the core promoter, thus leading to the repression of Sox9 to protect MEFs from differentiation.

Chd4-associated proteins synergistically regulate the expression of Sox9

KAP1 (KRAB-associated protein 1, also known as tripartite motif-containing protein 28 [TRIM28]) is required to directly recruit the CHD3/Mi2 component of the NuRD complex via SUMO-interacting motifs.[24] In our proteomic data, Cbx1 (chromobox homolog 1) and Kap1 were downregulated during BMP-2–induced chondrogenic differentiation; therefore, we considered that the two proteins might synergistically regulate the expression of Sox9 along with Chd4 during MEF chondrogenic differentiation. Thus, we used Re-ChIP analyses to evaluate whether Chd4, Cbx1, and Kap1 are bound simultaneously to the Sox9 promoter in primary MEFs. We performed two sequential IPs of the cross-linked complexes from extracts of primary MEFs with antibodies to Chd4, Cbx1, and Kap1. These procedures were followed by PCR analyses to determine whether sequences of the Sox9 core promoter (+1 region) were present in the precipitates. Using an anti-Chd4 antibody in the primary IP and an anti-Cbx1 antibody in the secondary IP, we found that the Sox9 core promoter was present in the precipitates (Fig. 5A, lane 16). This result indicated that both proteins were bound simultaneously to the Sox9 promoter. A similar result was obtained when the cell extracts were immunoprecipitated first with anti-Cbx1 and second with the Chd4 antibody (Fig. 5A, lane 6). Interestingly, when the chromatin extracts were sequentially immunoprecipitated with anti-Chd4 and anti-Kap1 antibodies, the Sox9 promoter was also detected (Fig. 5A, lane 14). This result indicated that Chd4 and Kap1 were also simultaneously bound to the Sox9 core promoter. These interactions, however, only occured within the context of the Sox9 core promoter because the distal Sox9 promoter region (+3k, as shown in Fig. 5A) was not detected in these precipitates. Deepening our knowledge of the relationship among Chd4, Cbx1, and Kap1 during chondrogenic differentiation, we discovered that knockdown of Chd4 leads to a decrease in the level of the Sox9 core promoter immunoprecipitated by anti-Cbx1 or Kap1 antibodies (Fig. 5B), reflecting the fact that both Cbx1 and Kap1 are likely to jointly regulate the transcriptional repression of Sox9 along with Chd4. Unfortunately, we could not detect a direct interaction among Chd4, Cbx1, and Kap1 (data not shown).

Figure 5.

Chd4 recruits the associated proteins Cbx1 and Kap1 to the Sox9 promoter. (A) ChIP and Re-ChIP experiments were performed with anti-Chd4, anti-Kap1, and anti-Cbx1 antibodies on primary MEFs treated with 200 ng/mL BMP-2 for 24 hours, using nonspecific IgG as the control. After DNA recovery, the precipitates were evaluated by electrophoresis. (B) Primary MEFs were treated with NC siRNA or Chd4 siRNA alone or in the presence of BMP-2. After DNA recovery, the precipitates were evaluated by real-time PCR to determine the level of enrichment of Cbx1 and Kap1 over the Sox9 core promoter relative to the negative control antibody. All results are presented as the means of three independent experiments ± standard deviation.

Chd4 is targeted by let-7a during BMP-2–induced chondrogenic differentiation

Previous reports showed that miR-145 is a key negative regulator of chondrogenic differentiation by directly targeting Sox9 at the early stage of chondrogenic differentiation.[25] Thus, we determined whether there were any miRNAs mediating the biological functions of Chd4 during BMP-2–induced chondrogenesis. For this purpose, we first searched the mouse TargetScan database for predicted binding miRNAs within the Chd4 3'UTR. We found three highly conserved binding sites among vertebrates, namely miR-194, let-7/98/4458/4800 (Supplemental Fig. S5). We then attempted to determine if these miRNAs were involved in BMP-2–induced chondrogenic differentiation. We assessed the relative expression of these miRNAs in MEFs with or without BMP-2 induction by real-time PCR. The results revealed that the expression of let-7a increased prominently after BMP-2 treatment, and the peak level (up to a fivefold increase relative to the control group) was reached at 5 days and sustained throughout the 21 days of BMP-2 treatment (Fig. 6A).

Figure 6.

Let-7a promotes BMP-2–induced chondrogenic differentiation by targeting the Chd4 3'UTR. (A) The relative enrichment of let-7a in the presence of BMP-2 induction as based on real-time PCR. Error bars indicate the standard deviation based on at least three independent experiments. (B, C) psiCHECK vectors containing Chd4 3'UTR (B) or mut Chd4 3'UTR (C) were transiently transfected into primary MEFs. Luciferase assays were performed 24 hours post-transfection. Error bars indicate the standard deviation based on at least three independent experiments. Bottom panels indicate the binding sites predicted by TargetScan, PicTar, and miRanda or the mutant site (bold red). (D) Chd4 expression levels in MEFs were detected by Western blot analysis 5 days after transfection with the let-7a mimics or let-7a inhibitor in the presence or absence of 200 ng/mL BMP-2. (E, F) Primary MEFs were transiently transfected with the let-7a mimics or let-7a inhibitor every 7 days for 3 weeks in the presence of 200 ng/mL BMP-2. The cell extracts were then subjected to Western blot analysis of Col2a1 (E) or Alcain blue staining for detection of sulfated glycosaminoglycans (F).

We next verified the predicted interactions between let7a and the Chd4 3'UTR. Experiments using a dual-luciferase reporter assay system carrying a Renilla luciferase reporter were performed. The 3'UTR segment of Chd4 containing the let-7a binding site was amplified by RT-PCR and inserted into the psiCHECK plasmid (downstream of the luciferase stop codon). In parallel, we generated a mutant construct of the Chd4 3'UTR using a site-directed mutagenesis method. As shown in Fig. 6B, the ratios of renilla/firefly luciferase after transfection with let-7a mimics were significantly lower compared with the control. Moreover, addition of a let-7a inhibitor could again elevate this ratio, indicating the direct binding of let-7a with the Chd4 3'UTR. Compared with the wild-type Chd4 3'UTR, the mutant insert affected let-7a binding and conferred resistance to let-7a (Fig. 6C). We then used immunoblot analysis to confirm the results of the luciferase assays. Introduction of let-7a alone to primary MEFs significantly reduced Chd4 protein expression, and this suppression was further strengthened when cells were exposed to BMP-2 (Fig. 6D). Additionally, this suppression could be rescued by the introduction of a let-7a inhibitor, which provided further evidence of a direct interaction between let-7a and Chd4.

To detect the biological functions of let-7a in the process of cell differentiation, we transiently transfected primary MEFs with let-7a mimics in addition to BMP-2 treatment. As shown in Fig. 6E, F, a notable promotion of chondrocyte differentiation was observed when compared with the NC control. In contrast, the presence of the let-7a inhibitor, which can completely block endogenous let-7a, resulted in no change of Col2a1 expression and Alcian blue staining. Together, these results indicated that BMP-2 induced the upregulation of let7a to positively affect chondrogenic differentiation by targeting Chd4.

We also analyzed the potential regulation of the Chd4-centered interactome by let-7a using five different bioinformatic software programs: Miranda, Mirwalk, TargetScan, RNA22, and miRDB. The results indicated 16 of the 30 downregulated protein-coding genes contained target sites for let-7a (Supplemental Table S4). Therefore, we speculated that let-7a is likely to promote chondrogenic differentiation of MEFs by downregulating Chd4 and its associated proteins.


For planarian regeneration and tissue homeostasis, the Smed-CHD4 gene, which is predicted to encode a chromatin-remodeling protein similar to CHD4/Mi-2 proteins, is required for neoblasts to produce progeny cells committed to differentiation, suggesting a crucial role for CHD4 proteins in stem-cell differentiation.[26] In hematopoietic stem cells, SNF2-like ATPase Mi-2beta is required for maintenance of multilineage differentiation in the early hematopoietic hierarchy.[27] In the study, our bioinformatic analysis using the IPA database revealed a Chd4-centered interactome composed of various proteins downregulated in the process of BMP-2–induced chondrogenic differentiation. Subsequent biological analysis revealed that Chd4 and the associated proteins in the network functioned in the control of transcription, chromatin modification, and RNA splicing, directly or indirectly regulating cell differentiation. In addition, knockdown of Chd4 resulted in a remarkable upregulation of Sox9 and promoted chondrogenic differentiation in the presence of BMP-2. These results were consistent with our previous report[4] indicating that BMP-2 could efficiently induce Sox9 expression along with chondrogenic differentiation, thus illustrating a negative role for Chd4 in the regulation of cell differentiation. In addition, promoter studies and mutant assays revealed that Chd4 negatively controlled the expression of Sox9 through the TGGCTG box at −207/−148 adjacent to the transcription start site of Sox9. It was worth noting that, in the study, we did not detect upregulated promoter activity in our reporter assay; the results were in agreement with our previous reports.[28] We thought that unlike stably transfected plasmids, the chromatin structure of transiently transfected plasmids was usually incomplete or even in a “naked” state, so even though Chd4 complexes could bind the target region (TGGCTG) of Sox9, they were unable to remodel chromatin structure to block the open status of the target gene as they usually did in intracellular chromatin.

Previous reports have demonstrated that Chd4 is a core component of the NuRD complex, which contains both histone deacetylase and nucleosome remodeling activities and has been implicated in the silencing of subsets of genes involved in various stages of cellular development.[29] Mbd3, an essential component of the NuRD corepressor complex, plays an important role in the development of pluripotent cells in vivo and in their ex vivo progression into embryonic stem cells.[30] Moreover, downregulation of HDAC1, which is also widely reported to be a member of the NuRD complex, is important during osteogenesis.[31] In a genetic study of HDAC1 and HDAC2 in embryonic stem cells, Dovey and colleagues[32] identified a unique requirement for HDAC1 in the optimal activity of HDAC1/2 corepressor complexes and cell fate determination during differentiation. IP assays of nuclear extracts from primary MEFs in our current study demonstrated the interaction of Chd4 with both HDAC1 and HDAC2 in vitro, and additional ChIP analysis revealed the binding of Hdac1 and Hdac2 to the promoter of Sox9 in vivo. Consequently, Chd4 might interact with Hdac1 and Hdac2 to cooperatively repress the differentiation of MEFs. Interestingly, deletion of Chd4 leads to a slightly increased binding of the NF-Y-p-300 complex (which we previously showed to bind the core promoter to open the chromatin structure) and advance the transcription of Sox9 and chondrogenic differentiation in the presence of BMP-2. Taken together, it is possible that Chd4 closes the chromatin structure by interacting with Hdac1 and Hdac2, thus competitively inhibiting the binding of the NF-Y-p-300 complex to the promoter of Sox9 and preventing chondrogenic differentiation via Sox9 downregulation.

To confirm this hypothesis, we employed nuclease hypersensitivity assays, which can be used to indirectly indicate the status of the chromatin structure. The results showed that nuclei from MEFs treated with Chd4 siRNA were more sensitive to NheI, illustrating the open status of the chromatin. Also, results of ChIP assays revealed the upregulation of signals for euchromatic markers and, inversely, downregulation of signals for chromatin silencing, corroborating the role of Chd4 in condensing the chromatin of the Sox9 promoter region. However, the exact mechanics of competitive interaction between the Chd4 and the NF-Y-p-300 complex still require further study to determine their precise role during cell differentiation.

Further analysis of the associated proteins of the Chd4 network identified by the ChIP and Re-ChIP assays indicated the combined binding of Cbx1 and Kap1 with the core promoter of Sox9. In the Chd4-centered interactome, there was no direct interaction identified between Chd4 and either Cbx1 or Kap1. Consistent with this, we also did not detect a direct interaction between Chd4 and Cbx1 or Kap1 (data not shown). In other reports, Cbx1 was shown to directly interact with Kap1 during heterochromatin-mediated gene silencing.[33] However, our IP assays did not indicate a direct interaction between Cbx1 and Kap1. It is possible that Cbx1 and Kap1 might synergistically promote the binding of the histone protein H3 with the Chd4 complex and, simultaneously, cooperate with other proteins in the network to transcriptionally regulate the expression of Sox9, preventing the chondrogenic differentiation of MEFs. This would reflect the key role of Chd4 in the regulation of MEF differentiation.

In terms of miRNAs regulating cell differentiation, we predicted the binding sites within the Chd4 3'UTR using TargetScan, which revealed a highly conserved domain for let-7. This binding site was confirmed based on the results of dual-luciferase reporter assays. Further functional studies showed an opposing effect of let-7a relative to Chd4 in the regulation of chondrogenic differentiation. Consistently, let-7b has been previously reported to play an essential role in the process of neural stem cell proliferation and differentiation.[34] Likewise, in lung cancer stemlike side population cells, reduced let-7 promotes differentiation of these cells, indicating that let-7 is involved in maintaining the balance between differentiation and quiescence.[35] Additionally, in the bipotent K562 human leukemia cells and human CD34+ hematopoietic progenitor cells, let-7 may promote megakaryocytic differentiation.[36] Hence, we can draw the conclusion that let-7 affects cell differentiation by affecting a series of physical process. Therefore, in this current work, we provide the first evidence that, after BMP-2 induction, upregulation of let-7a promotes MEF differentiation by targeting the master regulator Chd4. Interestingly, we identified predicted binding sites for let-7a in nearly half of the proteins in the Chd4-centered network. Therefore, we conjectured that let-7a is involved in the promotion of cell differentiation by transcriptionally inhibiting proteins associated with Chd4. However, further evidence elucidating the relationship between these proteins and let-7a needs to be provided.

In summary, to our knowledge, we are the first to recognize the central role of Chd4 in chondrogenic differentiation. In brief, our model indicates that BMP-2 induces the upregulation of let-7a, followed by the degradation of its target gene Chd4 and the separation of Hdac1/2 with Chd4. Subsequently, the competitive complex NF-Y-p-300 increasingly binds with the core promoter of Sox9, resulting in an open access for other positive transcription regulators of Sox9, thus advancing the chondrogenic differentiation of primary MEFs, as summarized in Fig. 7. However, because of the limitations of in vitro study, further in vivo studies are needed to consolidate the ubiquitous role of Chd4 and its associated proteins in repressing cell differentiation. To some extent, our findings illustrate an epigenetic regulation of the chondrogenic differentiation process as well as expand our understanding of the intracellular mechanisms that modulate chondrogenic differentiation and self-renewal of MSCs. These results also further promote the application of MSCs in the treatment of human disease.

Figure 7.

Schematic indicating the detailed mechanism of Chd4 and associated proteins in the regulation of chondrogenic differentiation as described in this study.


All authors state that they have no conflicts of interest.


This work was supported by the Ministry of Science and Technology of China (2012CB966904) and the National Natural Science Foundation of China (90919050, 30971631, 30971466, 81071524).

Authors' roles: Study design: QP, YJ, and FS. Study conduct: QY, WW, and YZ. Data collection: FS, QY, WW, and YZ. Data analysis and data interpretation: YY, AH, YJ, and QP. Drafting manuscript: FS and QY. Revising manuscript content: YJ and QP. Approving final version of manuscript: FS, QY, WW, YZ, YY, AH, YJ, and QP. QP takes responsibility for the integrity of the data analysis.