hmgb1 can facilitate activation of the matrilin-1 gene promoter by sox9 and l-sox5/sox6 in early...

Biochimica et Biophysica Acta 1829 (2013) 1075–1091

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta

j ourna l homepage: www.e lsev ie r .com/ locate /bbagrm

Hmgb1 can facilitate activation of the matrilin-1 gene promoter by Sox9and L-Sox5/Sox6 in early steps of chondrogenesis

Tibor Szénási a,1, Erzsébet Kénesi a,1, Andrea Nagy a,b, Annamária Molnár a, Bálint László Bálint c,Ágnes Zvara b, Zsolt Csabai a, Ferenc Deák a, Beáta Boros Oláh c, Lajos Mátés b, László Nagy c,d,László G. Puskás b,e, Ibolya Kiss a,e,⁎a Institute of Biochemistry, Biological Research Centre, Hungarian Academy of Sciences, H-6701 Szeged, Hungaryb Institute of Genetics, Biological Research Centre, Hungarian Academy of Sciences, H-6701 Szeged, Hungaryc Department of Biochemistry and Molecular Biology, Medical and Health Science Center, University of Debrecen, H-4032 Debrecen, Hungaryd MTA-DE “Lendulet” Immunogenomics Research Group, University of Debrecen, H-4032 Debrecen, Hungarye Avidin Ltd., H-6726 Szeged, Hungary

Abbreviations: CEC, chicken embryo chondrocyte; CChIP, chromatin immunoprecipitation; Dpe1 and Dpe2,ments 1 and 2; ECM, extracellular matrix; EMSA, electrGP, growth plate; GST, glutathione S-transferase; HDMHMG, high-mobility-group; Hmgb, HMG box; Ine, initiamesenchyme; LM-PCR, ligation-mediated PCR; Nfi, nuelement 1; RCS, rat chondrosarcoma; SI and SII, silencer⁎ Corresponding author at: Institute of Biochemistry, B

Hungarian Academyof Sciences, Temesvári krt. 62., H-672599 782; fax: +36 62 432756.

E-mail addresses: sztibor@brc.hu (T. Szénási), kenesnagya@brc.hu (A. Nagy), molnar.annamaria@gmail.comlbalint@med.unideb.hu (B.L. Bálint), zvara@brc.hu (Á. Z(Z. Csabai), deak.ferenc@brc.mta.hu (F. Deák), bejjus010mates.lajos@brc.mta.hu (L. Mátés), nagyl@med.unideb.h(L.G. Puskás), kiss@brc.hu (I. Kiss).

1 These authors contributed equally to the work andfirst authors.

a b s t r a c t

Article history:Received 23 November 2012Received in revised form 8 July 2013Accepted 9 July 2013Available online 13 July 2013

Keywords:Growth plateCartilage-specific regulationMatrilinTransgenic miceChromatin immunoprecipitationSilencing

The architectural high mobility group box 1 (Hmgb1) protein acts as both a nuclear and an extracellularregulator of various biological processes, including skeletogenesis. Here we report its contribution to theevolutionarily conserved, distinctive regulation of the matrilin-1 gene (Matn1) expression in amniotes. Wepreviously demonstrated that uniquely assembled proximal promoter elements restrict Matn1 expressionto specific growth plate cartilage zones by allowing varying doses of L-Sox5/Sox6 and Nfi proteins tofine-tune their Sox9-mediated transactivation. Here, we dissected the regulatory mechanisms underlyingthe activity of a conserved distal promoter element 1. We show that this element carries three Sox-bindingsites, works as an enhancer in vivo, and allows promoter activation by the Sox5/6/9 chondrogenic trio. Inearly steps of chondrogenesis, declining Hmgb1 expression overlaps with the onset of Sox9 expression. Unlikerepression in late steps, Hmgb1 overexpression in early chondrogenesis increasesMatn1 promoter activationby the Sox trio, and forced Hmgb1 expression in COS-7 cells facilitates induction of Matn1 expression by theSox trio. The conserved Matn1 control elements bind Hmgb1 and SOX9 with opposite efficiency in vitro. TheyshowhigherHMGB1 than SOX trio occupancy in established chondrogenic cell lines, andHMGB1 silencing greatlyincreases MATN1 and COL2A1 expression. Together, these data thus suggest a model whereby Hmgb1 helps re-cruit the Sox trio to the Matn1 promoter and thereby facilitates activation of the gene in early chondrogenesis.We anticipate that Hmgb1 may similarly affect transcription of other cartilage-specific genes.

EF, chicken embryo fibroblast;distal promoter upstream ele-ophoretic mobility shift assay;, high density mesenchyme;

tor element; LDM, low densityclear factor I; Pe1, promoterelements I and II

iological Research Centre of the6 Szeged, Hungary. Tel.:+3662

ie@gmail.com (E. Kénesi),(A. Molnár),vara), csabai0911@gmail.com2@gmail.com (B. Boros Oláh),u (L. Nagy), pusi@brc.hu

should be considered as equal

rights reserved.

1. Introduction

SRY-related high-mobility-group (HMG) box (Sox) proteins andcanonical high-mobility-group box (HMGB) proteins have distantly re-lated DNA-binding domains that regulate gene expression by diversemechanisms during development and disease. Several Sox proteinscontrol cell-fate decisions in early steps of endochondral bone formation[1], whereas the Hmgb1 protein was shown to regulate later steps [2].We determined here the role of Hmgb1 in early steps by focusing onthe control region of the matrilin-1 gene (Matn1) [3].

The Sox and Hmgb protein families have similar as well as distinctfeatures [1,4,5]. Their HMG boxes show only 20% identity, but bothinteract with the minor groove of the DNA helix and induce a sharpbend of this helix upon binding. While Hmgb proteins are abundantand widely expressed non-histone chromatin components, Sox pro-teins are expressed at a low level and only in certain cell types. Hmgbproteins bind distorted DNA transiently and without sequence specific-ity [4,6]. Lacking a transactivation domain, they act only as architectural

1076 T. Szénási et al. / Biochimica et Biophysica Acta 1829 (2013) 1075–1091

proteins to alter the chromatin structure and modulate transcription.Hmgb1 can promote transcription by displacing the histone H1 andthereby increasing the accessibility of chromosomal DNA to regulatoryfactors [7,8]. By bending DNA, it helps recruit general transcriptionfactors to the promoter and, via protein–protein interactions, it canenhance the binding of classical transcription factors to their cognatesites and promote the recruitment of additional interacting factors [4].By contrast, Sox proteins bind DNA specifically, although with lowersequence specificity and efficiency than classical transcription factors[5]. In addition to a postulated architectural function, some Sox proteins(e.g. Sox9) feature a potent transactivation domain, and thereby act asclassical transcription factors. They interact with many partner factorsto facilitate enhanceosome formation and control cell fate and differen-tiation during development [1]. The Sox9, L-Sox5 and Sox6 proteins arereferred to as chondrogenic Sox trio, as they act cooperatively to directthe lineage commitment and differentiation of chondrocytes duringendochondral bone development.

Endochondral ossification is the processwhereby vertebrate embryosfirst form cartilaginous skeletal structures and then remodel them intobone and bone marrow tissue. It involves strict spatial and temporalcoordination of morphogenetic and differentiation steps [9–11]. Firstchondrogenesis takes place, which starts with the condensation ofmesenchymal precursors and their differentiation into chondroblaststhat proliferate and deposit an extensive cartilaginous extracellularmatrix (ECM) [10,12,13]. This matrix is primarily made of collagen-2(encoded by Col2a1) and aggrecan (Acan), which determine its physicalproperties, while matrilins and other minor components modulatethe ECM organization. Then, in the middle of cartilaginous primordia,proliferative chondroblasts exit from the cell cycle, differentiate intoprehypertrophic and hypertrophic cells, and eventually undergo apopto-sis. Blood vessels, osteoclasts and osteoblasts then invade and replacethe mineralized hypertrophic cartilage by bone tissue. This sequentialdifferentiation of chondrocytes results in the formation of distinct tissuezones, collectively referred to as the growth plate (GP), in line withtheir essential role in skeletal elongation.

The multi-step chondrocyte differentiation process is marked bysequential changes in gene expression [12,14,15]. Col2a1 starts tobe expressed in prechondrocytes, whereas Acan and most othercartilage ECM genes are turned on in early chondroblasts andupregulated in the columnar or proliferating GP zone. By contrast,Matn1 shows such a slow and delayed activation in earlychondroblasts [15–18] that is expressed almost exclusively incolumnar and prehypertrophic chondrocytes [19–21]. Sox9 directsthe commitment of osteochondroprogenitors to the chondrocytelineage [22,23], and together with L-Sox5 and Sox6, directly controlsthe subsequent steps of chondrocyte differentiation [24–26]. The Soxtrio namely activates Col2a1, Acan and most other cartilage ECM genes[12,22,27,28]. Sox9 binds as a homodimer to inverted pairs of Soxmotifs in cartilage-specific enhancers to activate transcription via itstransactivation domain. L-Sox5/Sox6 binds more variable Sox motifs.Lacking a transactivation domain, they synergize with Sox9 by increas-ing its binding efficiency [29]. Various patterning factors, signalingmolecules (e.g. Ihh/PTHrP, FGF, BMP), and hormones influence theshape and size of skeletal elements and bone growth [10–12,30,31].Endochondral ossification is defective in Hmgb1−/− mice, as Hmgb1 issecreted by hypertrophic chondrocytes and acts as a chemoattractantfor cell invasion [2].

Matrilins are multidomain filament-forming proteins, which serve asadaptors in ECM assembly and in mechanotransduction of chondrocytes[32–34]. Unlike other matrilins, Matn1 is expressed exclusively incartilage, where it mediates connections between aggrecan, collagen-2and other molecules. Pericellular Matn1 and Matn3 are needed for Ihhsignaling and mechanical stimulation of chondrocyte proliferation anddifferentiation [35]. MATN1 polymorphism has been linked to mandibu-lar prognatism in human and Matn1 upregulation was implicated invertebral fusion of Atlantic salmon [36,37]. We previously reported that

the 2-kb Matn1 promoter region directs transgene expression to the GPzones that specifically express endogenousMatn1 [20,21]. This promoterregion features several sequence blocks that are highly conserved in am-niotes [38], andwe showed that themost proximal elements, i.e., a highlyconserved promoter element 1 (Pe1), an initiator element (Ine), andtwo silencers (SI and SII), function together to restrict promoter activityin a stage- and GP zone-specific manner [3,38,39]. This is made possibleby a dose-dependent modulation of the activity of Sox9, bound the Pe1,by L-Sox5/Sox6, bound to Ine, and nuclear factor I (Nfi) family proteins,bound to SI and SII.

Here we extend our model demonstrating that a distal promoterelement 1 (Dpe1), which functions as an enhancer in transgenic mice,interacts with the short promoter via the Sox trio. Furthermore, weshow that Hmgb1, whose expression overlaps with that of Sox9 inearly chondrogenesis, is capable of increasing the activation of theMatn1 promoter by the Sox trio, whereas it repressesMatn1 expressionin later steps or in established cell lines.

2. Materials and methods

2.1. Cell culture

Chicken embryo fibroblasts (CEF), chondroblasts (CEC), and mesen-chymal cells were prepared and cultured as described previously [39].Low density (LDM) and high-density mesenchyme (HDM) cultureswere made by plating 1 × 106 cells and 5 × 106 cells, respectively, in35-mm plates as described previously [3]. COS-7 cells were culturedunder standard conditions. HDMcultures consisting of early proliferative(stage Ia) chondroblasts and CEC cultures rich in late proliferative (stageIb) chondroblasts represented the low and high Matn1-expressingcell types, respectively [17,18,39]. LDM, CEF, COS-7, and human 293Tcultures were used as Matn1-nonexpressing controls.

The C-28/I2 immortalized human costal chondrocyte [40], theSW1353 human chondrosarcoma (ATCC HTB-94) and the RCS (ratchondrosarcoma) [41] cell lines were cultured in DMEM supplementedwith 10% FCS (GIBCO).

2.2. Oligonucleotides and plasmid constructions

As previously, all positions are given in bp from the first T of theTATA motif of the chicken Matn1. Luciferase reporters FO15Luc andAC8Luc driven by the short and long Matn1 promoters, respectively,were described [38] as well as their mutant derivatives ΔPe1M1-,ΔPe1M4-, ΔIneM1-, ΔIneM2-, ΔIneM3- and ΔPe1M1-ΔIneM2-AC8Luc[3]. ΔDpe1ABC-, ΔDpe1BC- and ΔDpe2-AC8Luc were made by delet-ing sequences between positions −1879/−1791, −1848/−1791and −1745/−1642, respectively, from the long Matn1 promoter.

Luciferase reporters harboring multiple copies of the Dpe1 elementupstream of the Matn1 short promoter were made by inserting oneto eight copies of the PCR-amplified Dpe1 fragment into FO15Luc.4×Dpe1(−)FO15Luc was generated by inserting blunted four copies ofDpe1 into FO15Luc in reverse orientation. PCLuc and 4×Dpe1(+)PCLucwere generated by replacing the Matn1 short promoter of FO15Lucand 4×Dpe1(+)FO15Luc, respectively, with the Col2a1 short promoterfragment between positions−309/+118.

For transgenic experiments, LacZ fusion construct 8×Dpe1(+)NAD1 was produced by inserting eight copies of the Dpe1 elementupstream of the short promoter (−334/+67) of NAD1, the Matn1short promoter-LacZ reporter [38].

Structures and sequences of all constructswere verified by restrictionmapping and sequencing.

2.3. Transient expression assays

CEC and CEF cultures were transfected with 2 μg reporters, whileHDM, LDM, and COS-7 cultures were transfected with 5 μg reporters

1077T. Szénási et al. / Biochimica et Biophysica Acta 1829 (2013) 1075–1091

using the Ca-phosphate coprecipitationmethod as described previously[3]. pRL-TKVector (Promega) served as an internal control to correct fortransfection efficiency, but parallel plates were also transfected withFO15Luc. Firefly and renilla luciferase activities were measured in aLuminoscan Ascent (ThermoLabsystem 2.6) using Luciferase AssaySystems (Promega) according to supplier's instructions 72 h (HDMand LDM) or 48 h (other cells) posttransfection. Relative luciferaseactivities were expressed in fold as compared to values of FO15Luctaken as 1, unless noted otherwise.

Cotransfection assays were made essentially as described previ-ously [3] using effector plasmids pcDNA5′UT-FLAG-L-Sox5 (pFSox5),pcDNA5′UT-FLAG-Sox6 (pFSox6), and pCDMA-SOX9 (pSOX9) orpcDNA5′UT-FLAG-SOX9 (pFSOX9) [28] and pHmgb1 expressing ratHmgb1 [42]. In a typical experiment, 125 ng pFSox5, 125 ng pFSox6and 250 ng pSOX9 or 250 ng pFSOX9 were added with or without100 ng pHmgb1. Some experiments were performed with increasingamounts of pHmgb1 (0–500 ng) or pFSox5 and pFSox6 (0–250 ng).All transfections were done in duplicates or triplicates and repeated3–10 times with at least two different DNA preparations. Results arepresented as means ± standard error of the mean (SEM).

Statistical analysis was carried out using one-way analysis ofvariance (ANOVA) and Dunnett's test with KyPlot version 2.0 beta 15.*p b 0.05, **p b 0.01, ***p b 0.001 vs. reporters cotransfected withempty vector(s) or mutants vs. similarly cotransfected AC8Luc;#p b 0.05, ##p b 0.01, ###p b 0.001 as indicated.

2.4. Generation and histological analysis of transgenic mice

All animal experiments were conducted according to the ethicalstandards of the Animal Health Care and Control Institute, CsongrádCounty, Hungary. C57BL/6, CBA, CD-1 and FVB mice were obtainedfrom Charles River Laboratories, Hungary. Transgenic mice were gen-erated by microinjection of the purified insert of 8×Dpe1(+)NAD1and analyzed as described previously [3,21]. On embryonic day 15.5(E15.5), the foster mothers were killed by cervical dislocationand the transgene was detected by PCR in founder (G0) embryos.The fixed G0 embryos were whole mount stained with X-gal andphotographed with a Leica MZFLIII stereomicroscope equipped witha DC300F camera. Cryosections counterstained with eosin wereanalyzed using a Nikon Eclipse E600 microscope equipped with aSpot RT Slider camera. Figures were made with Adobe Photoshop8.0 and CorelDraw X4 software.

2.5. Electrophoreticmobility shift assay (EMSA) and supershift experiments

Double-stranded oligonucleotides were synthesized for the Dpe1element comprising positions −1879/−1791: 5′-GAG TCC AGT GTTTTC GTT TTT GGA GGC CCG GGG AA-3′ (Dpe1A), 5′-GGA AAA ATTATG TTT CAT ATA TTA AAA ATA AAC A-3′ (Dpe1B), and 5′-AAA TAAACA CTA CTT TTA CAG AGG TAT AAA TGC-3′ (Dpe1C). Nucleotidesequences for Ine and Pe1 were described previously [3,38]. Codingregion of Hmgb1 was inserted in frame into pGEX expression vector.GST-tagged L-Sox5, SOX9 and Hmgb1 were expressed and purified,and crude cell extracts were made as described [38]. 20–30 fmolend-labeled DNAprobeswere incubated either with 0.6–3.2 μg purifiedGST-fused Hmgb1, SOX9, L-Sox5 or 3 μg crude CEC or CEF cell extracts inthe presence of 100–500 ng poly(dG–dC) · (dG–dC) and separated onpre-run 5% or 6.6% PAGE. Supershift experiments were performed usingHmgb1-specific antisera (gift of D.P. Edward).

2.6. Quantitative real-time PCR (QRT-PCR)

Total RNA was isolated from CEC, CEF or HDM cultures at subse-quent days of chondrogenesis using the RNA isolation kit (Macherey-Nagel) according to the manufacturer's instructions. The quantity ofisolated RNA was measured by spectrophotometry (NanoDrop). QRT-

PCR was performed on a RotorGene 3000 instrument (Corbett Re-search) with gene-specific primers (Suppl. Table S1) and SybrGreenprotocol at a final primer concentration of 250 nM as follows: 15 minat 95 °C, 45 cycles of 95 °C for 15 s, 60 °C for 25 s and 72 °C for 25 sto monitor gene expression changes as described earlier [3]. Each indi-vidual Cτ valueswere normalized to the average Cτ values of three inter-nal control genes (Gapdh, 18S rRNA, and 28S rRNA). The final relativegene expression ratioswere calculated as either 2−ΔCτ values (comparedto the internal control genes) or 2−ΔΔCτ values (comparison of the nor-malized ratios) as indicated in the figure legend.

2.7. Immunofluorescence

Acetone-fixed 10-μm cryosections were used for immunofluores-cence. Nonspecific binding of the antibodies was blocked with 10% goatnormal serum. The specimens were incubated at 4 °C overnight withthe following primary antibodies in combinations: rabbit affinity purifiedantisera specific for Matn1 [43] (1:200 dilution) and SOX9 (Abcam,ab3697, 1:50) and mouse monoclonal antibody for HMGB1 (MBL,M137-3, 1:200). The appropriate secondary antibodies were applied atroom temperature for 1 h in the dark: Alexa 488-labeled anti-rabbitIgG antiserum (Molecular Probes, 1:400) and Cy3-conjugated anti-mouse IgG antibodies (Jackson Immunoresearch, 1:400). Nuclei werestained with 1 μg/ml Hoechst in PBS for 5 min. The specimens weremounted with fluorescent mounting medium (Dako), viewed with aNikon Eclipse E600 microscope equipped with epifluorescence and Panflour objectives, and photographed with a Nikon digital camera D5000.After immunofluorescence, the coverslips were removed and the sec-tions were restained with hematoxylin and eosin. The images wereprocessed using SPOT software (version 4.0.9 for Windows; DiagnosticInstruments) and figures were made with Adobe Photoshop 8.0 andCorelDraw X4 software.

2.8. Forced expression assays combined with western analysis andQRT-PCR

We made effector plasmid pFHmgb1 by inserting the Hmgb1coding region [42] into pcDNA5′UT-2FLAG. To estimate the relativeexpression levels of Sox and Hmgb1 proteins, we cotransfectedCOS-7 cells with 10 μg AC8Luc, 1 μg each of pFSOX9, pFSox5 andpFSox6, and increasing amounts of pFHmgb1. Transfected cells werelysed and supernatants were used to measure luciferase activitiesand for western analysis with rabbit anti-FLAG (Sigma) antisera asdescribed previously [3].

To test the induction of the endogenous Matn1 in forced expres-sion assays, we cotransfected COS-7 cells with 50 ng pFSOX9, 75 ngpFSox5, and 75 ng pFSox6 without and with 800 ng pFHmgb1 using2 μl TurboFect (ThermoScientific, R0531). Transfection mixtureswere adjusted with empty vectors to the same amount of total DNA.Transfections were made in duplicates and repeated 3 times. RNAwas isolated from the cells and the Matn1 mRNA level was deter-mined by QRT-PCR using the SybrGreen protocol and gene-specificprimer pairs (Suppl. Table S2). Cτ values were normalized to that ofGapdh. Data are presented as mean ± SEM.

2.9. Hmgb1 silencing

Silencing experiments were performed in chondrogenic celllines C-28/I2, SW1353, and RCS with siRNAs purchased from BioneerCorporation (Daejeon, Republic of Korea) for human HMGB1:5′-caggaggaauacugaacau-3′; for rat Hmgb1: 5′-cugucaacuucucagaguu-3′; for human GAPDH: 5′-gugugaaccaugagaagua-3′, and for neg-ative control siRNA: 5′-ccuacgccaccaauuucgu-3′. 1.2–2.0 × 105 cellswere plated in 6-well plates and transfected with 100–400 pmol ofsiRNA duplexes 24 h after plating using X-tremeGENE siRNA Trans-fection Reagent (Roche Applied Science) as suggested by the supplier.

Cultures were harvested 30 h (RCS) or 42 h (C-28/I2, SW1353) aftertransfection. RNA was isolated from the cells and marker gene expres-sion levels were determined by QRT-PCR using the SybrGreen protocoland gene-specific primer pairs (Suppl. Table S2). Gene expressionlevels were normalized to the invariant Rps18 mRNA levels. Data arepresented as mean ± SEM from three independent experiments.

2.10. Chromatin immunoprecipitation (ChIP)

ChIP studies were performed in chondrogenic cell lines C-28/I2,SW1353 and RCS and in the non-expressing 293T cells, as describedearlier [44] with modifications. Briefly, cells were crosslinked withDSG for 30 min, fixed with 1% formaldehyde for 10 min, and washedtwice with PBS and nuclei were isolated. Sonication was performedwith Bioruptor (Diagenode SA). Immunoprecipitation was performedwith rabbit ChIP grade antisera (Abcam) raised against HMGB1(ab18256), SOX9 (ab3697), SOX5 (ab94396), and SOX6 (ab30455),which specifically recognize other vertebrate orthologs. Immunopre-cipitation with rabbit IgG was used as a negative control. We usedmicroarray data deposited in NCBIs Gene Expression Omnibus (GEO)(accession number GSE10024) to search for potential ChIP controlgenes of low expression level in chondrocytic cells. From the three con-trol genes we selected, TUBB3/Tubb3 codes for the neuronal β-tubulinisotype III, while IL1B encoding the proinflammatory cytokine IL-1βand CRP/Crp encoding the acute phase C-reactive protein can also be in-duced in chondrocytic and tumorigenic cells, respectively. Protein–DNAcomplexes were collected withmagnetic beads (Invitrogen, DynabeadsProtein A cat no.100.01D). Beads were washed, captured complexeswere eluted and DNA was purified with Qiagen MinElute columns.QRT-PCR measurements were performed on a Roche LC480 QPCRinstrument. The recovered DNA was normalized to the input materialused for the ChIP reaction. ChIP experiments were repeated 3–6times. Sequences of the primer sets used for QRT-PCR are listed inSuppl. Table S3.

3. Results

3.1. Delineating promoter upstream elements in transientexpression assays

We recently reported that distal promoter regions are needed toenhance the low activity of the proximalMatn1 promoter and to directtransgene expression in the developing cartilaginous elements witha pattern characteristic of that of endogenous Matn1 [3]. To identifythe distal promoter elements (Dpe) involved in this regulation, wefirst assessed the influence of upstream regions of increasing lengthon the short promoter activity in chondrogenic cultures. Sequences−2011/−1741 did not exert significant activation, whereas sequences−2011/−1394 and−2011/−948 elevated the short promoter activity7.2–9.2-fold, up to 38–48% of the long promoter activity, in CEC culturesrich in late proliferative chondroblasts (Fig. 1A). The increasewas below2-fold in HDM culture consisting of early proliferative chondroblastsand in non-expressing CEF culture, indicating marked chondroblaststage-specific activation for upstream sequences −2011/−1394. Asthis region includes the putative Dpe1 and Dpe2 elements conservedin amniotes [38], we next examined if deletion of these elementsaffected the long promoter activity. Deletion of the entire Dpe1 elementor its subfragments B and C decreased the long promoter activity by 71%and 49%, respectively, in CEC culture, while deletion of Dpe2 hadmildereffect (Fig. 1B). Thus, Dpe1 is a major contributor of the stage-specificactivity of the long promoter.

To get direct evidence of the positive role of Dpe1, we fused it tothe short promoter. One Dpe1 copy had only a slight effect in agree-ment with the low activity of PH-FO15Luc (Fig. 1B), but the promoteractivity increased sharply in positive correlation with the copy num-ber of Dpe1 in CEC culture (Fig. 1C). By contrast, multiple copies of

Dpe2 hardly increased the activity of the short promoter (data notshown).

3.2. Dpe1 enhances the short Matn1 promoter activity in transgenic mice

Next we studied the spatiotemporal activation of the Matn1 shortpromoter by eight copies of the Dpe1 element in transgenic mice(Fig. 2). By monitoring the LacZ expression in the developing skeletonof founder embryos, we observed great enhancement as compared tothe low activity of the transgene driven by the short promoter alone(NAD1) [38] (Fig. 2B–N). The expression pattern resembled that ofthe transgene driven by the long promoter [21], with similar GPzone specificity, craniocaudal increase in vertebral bodies and evenmore pronounced proximodistal differences in developing limbs(Fig. 2C–N). Thus, Dpe1 enhances the short promoter activity intransgenic mice with the zonal and distal structure-specific patternof the endogenous Matn1.

3.3. Chondrogenic Sox transcription factors bind Dpe1 in vitro

As the Dpe1 element includes putative Sox motifs conserved inamniotes [38], next we asked if it can interact with purified Sox factorsin vitro. The chicken sequence harbors three putative sites with tandemand inverted paired motifs, which share 5/7 or 6/7 nucleotide identitywith the Sox consensus sequence (Fig. 3A). These motifs also show5/10 to 8/10 nucleotide identity with AGAACAATGG, the preferredSox9-binding site [45]. We demonstrated that Dpe1 carries at leastthree sites, which can bind purified Sox proteins in EMSA (Fig. 3). BothGST-fused SOX9 and L-Sox5 recognized each subfragment of Dpe1, butwith inverse binding efficiency. SOX9 exhibited the strongest bindingto Dpe1C forming two complexes and weaker binding to Dpe1A andDpe1B forming one and three diffuse complexes, respectively (Fig. 3B).On the other hand, L-Sox5 most preferably recognized Dpe1B, followedby A and C (Fig. 3C). In fact, SOX9 bound the Matn1 control elementswith highly variable efficiency. It showed more potent complex forma-tion in vitro with each of the Dpe1 Sox sites than with those of Ine, butit bound Pe1 even more powerfully (≥5-fold) than Dpe1C (Fig. 3B).Thus, Dpe1 can directly interact with SOX9 and L-Sox5 in vitro.

3.4. Dpe1 participates in Sox trio-mediated promoter enhancementin culture

To test the contribution of Dpe1 to the Sox-mediated promoteractivation, we co-transfected reporters driven by wild-type or mu-tant long promoters with Sox expression plasmids (Fig. 4A). Deletionof Dpe1 decreased ~3-fold the transactivation of the promoterby SOX9 in LDM culture and close to half in CEC culture. It alsodecreased to half the synergistic activation by the Sox trio in LDMculture or in COS-7 cells forced to express L-Sox5/Sox6 in optimaldose relative to SOX9, but not in CEC culture expressing the Soxtrio at high level. This suggests that SOX9 binding to Dpe1 is impor-tant for promoter activation in late proliferative chondroblasts. Inearly steps of chondrogenesis, however, SOX9 binding to Dpe1 aswell as synergistic activation of SOX9 by an optimal dose of L-Sox5/Sox6 is also needed for promoter enhancement.

Next we assessed if Dpe1 elements can exert Sox trio-mediatedenhancement to homologous and heterologous promoters in COS-7cells forced to express constant amount of SOX9 and increasingamounts of L-Sox5 and Sox6 (Fig. 4B–D). Four copies of Dpe1 in-creased the activity of the Matn1 and the Col2a1 promoters 3.1- and6-fold, respectively, which was hardly increased further by forcedexpression of SOX9 or L-Sox5/Sox6 alone. L-Sox5/Sox6, however,synergized with SOX9 in a dose-dependent manner to mediate a17- and 23-fold activation of the Matn1 and Col2a1 short promoters,respectively, from the Dpe1 elements. As the Dpe1 elements workedefficiently in both orientations (Fig. 4C) and enhanced even the

Fig. 1. Role of Dpe1 in the promoter activity in chondrogenic cultures. (A) Effect of promoter upstream regions of increasing length on the shortMatn1 promoter activity. ConservedDNA elements located in the short promoter (Pe1 and Ine) and in the distal promoter upstream region (Dpe1 and Dpe2) are depicted. (B) Effect of Dpe deletions on the activity ofthe long wild-type (wt)Matn1 promoter. (C) Enhancement of the short promoter activity by Dpe1. (A–C) Luciferase activities of reporters were measured in HDM and CEC cultures,expressingMatn1 at low and high levels, respectively, and in the non-expressing CEF culture. Luciferase activities are presented as fold values relative to that for FO15Luc. *p b 0.05,**p b 0.01, ***p b 0.001 vs. FO15Luc; #p b 0.05, ##p b 0.01, ###p b 0.001 vs. AC8Luc or as indicated.

heterologous Col2a1 promoter (Fig. 4D), we concluded that Dpe1 canfunction as a cartilage-specific enhancer element.

Collectively these data support and further extend our model forthe regulation of Matn1 by demonstrating that Dpe1 is an importantenhancer element that can mediate activation of homologous andheterologous promoters by the Sox trio.

3.5. Hmgb1 is expressed in early chondrogenesis in inverse correlationwith chondrogenic marker genes

Hmgb1 was reported to facilitate late steps of endochondral ossi-fication [2], but its function in early steps has not been studied yet.To address the possible involvement of Hmgb1 in chondrogenesis,we used double immunofluorescence to monitor Hmgb1 expression inthe developing limbs of mouse embryos (Fig. 5). We observed that theubiquitous Hmgb1 immunosignal started to decrease in early steps ofchondrogenesis, showing overlapwith Sox9 in condensedmesenchyme,

prechondrocytes and early chondroblasts (Fig. 5A, B, and D). In line withformer data [12,21,26], Matn1 had a narrower spatiotemporal expres-sion pattern than Sox9, being first detectable only in early chondroblastswith some delay (Fig. 5C and E). Thus only a very limited overlap wasseen between Hmgb1 and Matn1 in the latter cells at the onset ofMatn1 (Fig. 5E). Hmgb1 expression, however, dropped as chondrogene-sis progressed, exhibiting a complementary pattern to that of Sox9(Fig. 5B and D) and Matn1 (Fig. 5C and E) in overtly differentiated carti-laginous elements.

Next we compared the kinetic changes in the expression ofHmgb1 and various marker genes by QRT-PCR during in vitro chon-drogenesis in HDM culture (Fig. 6). This culture faithfully mimicsthe early steps of chondrogenesis as it differentiates to early prolifer-ative chondroblasts characterized by elevated Col6a1 expression(Fig. 6C). CEC culture, expressing the genes for Sox trio and cartilageproteins at high levels (Fig. 6B–D), represented a later stage. CEFculture served as a negative control.

Fig. 2. Dpe1 elements enhance the short promoter activity in vivo. (A) Schematic of 8×Dpe1(+)NAD1 with eight copies of Dpe1 fused to the short promoter-LacZ construct (NAD1).(B) Expression of the transgene in founder embryos (FE) whole mount stained with X-Gal at E15.5. (C–N) Histological analysis of embryo cryosections. In the developing limbs, LacZactivity increases proximodistally, e.g. from humerus (hu), radius (ra) and ulna (ul) to the highest level in metacarpals (mc), metatarsals (mt) and phalanges (ph) (C–J). Distalepiphysis (de) exhibits stronger staining than the proximal one (pe) in the developing radius, ulna, metacarpals, metatarsals and phalanges (C–E and G–J). Cranial (cran) vertebralbodies (vb) are not stained, but staining increases in the lumbar (lumb) and caudal regions (caud) (L–N). LacZ expression is highest in the zones of columnar chondroblasts (cc) andprehypertrophic chondrocytes (pc), but it is also elevated in the early, epiphyseal (ec), and source (sc) chondroblasts of distal epiphyses of phalanges, metatarsals and radius(D, E and G). np, nucleus pulposus; t, tarsal; tb, trabecular bone. Bars, 2 mm (B); 200 μm (C–N).

The Hmgb1 mRNA level was high in CEF culture, declined duringdifferentiation in HDM culture and was low in CEC culture (Fig. 6A).In fact, the down-regulation of Hmgb1 during chondrogenesis wassimilar, but even greater than that of Hmgn1, which was implicatedin the activation of Sox9 [46]. In our experiments, both Hmgb1 andHmgn1 levels showed an inverse correlation with those of Sox9,Sox5 and most cartilage ECM genes (Col2a1, Col11a1, Matn1, andMatn3) (Fig. 6B–D). Matn4 level, however, peaked in HDM culturesuggesting a function in early stage of chondrogenesis (Fig. 6D). Re-markably, from a very low expression level compared to the internalcontrol genes measured in committedmesenchyme (Suppl. Table S1),Matn1 expression showed the highest relative increase (2000-fold) inCEC culture, in contrast to the lower (180-fold) increase in the level ofCol2a1 and other cartilage ECM genes (b80-fold, compared to HDMday 0) (Fig. 6C and D).

In agreement with former observations [17], activation of Matn1was first detected in HDM culture at day 4. At this time, the sharplydeclining Hmgb1 mRNA level (1.5 × 10−3) relative to the internalcontrol genes was only 2.3-fold higher than the increasing Sox9mRNA level (6.6 × 10−4) (Suppl. Table S1). This ratio decreased to1.3 by day 7 in HDM culture. In CEC culture consisting mostly of lateproliferative (stage Ib) chondroblasts, however, the mRNA levels forSox9 and L-Sox5 were 21.5-fold and 15.5-fold higher, respectively,than that for Hmgb1.

To sum up, Hmgb1 expression gradually declined in an inverse cor-relation with the activation of the chondrogenic Sox and ECM genesduring limb development and in chondrogenic cultures, exhibiting asmall overlap in their expression domains in early steps of in vivo andin vitro chondrogenesis. Coexpression ofHmgb1 and Sox9 at comparablelevels just before and at the time of the onset of Matn1 raises the

Fig. 3. Interaction of Dpe1 with purified Sox proteins in vitro. (A) Nucleotide sequence of Dpe1. Arrows mark the subfragments and staggered arrows depict the Sox motifs. Bindingof GST-fused SOX9 (B) and L-Sox5 (C) to the Dpe1 subfragments in EMSA, as compared to their binding affinity to Ine and Pe1. F, free probe.

possibility that, in addition to Sox9, Hmgb1 may also contribute to thegene regulation in early steps of chondrogenesis.

3.6. Hmgb1 increases Matn1 promoter activation by the Sox trio

Next we assessed whether Hmgb1 can influence the Matn1 pro-moter activity in cotransfection assays. We observed transient activa-tion of the long promoter by increasing the amount of Hmgb1expression plasmid with the highest peak (2.6-fold) in HDM cultureand a lower increase in LDM and CEF cultures (Fig. 7A). The smalltransient increase, however, turned to inhibition at higher doses ofHmgb1 in CEC culture. Smaller (1.6–1.8-fold) transient increase wasalso seen for the short promoter in CEF and HDM cultures but hardlyany in CEC culture (Suppl. Fig. S1A).

Thus, Hmgb1 can markedly increase the long promoter activity inmesenchymal cultures, where the Sox trio expression is low (Fig. 6B)and the in vivo occupancy of Sox-binding sites of Pe1 and Ine is verylimited [3,38]. In CEC culture, however, where the Sox trio expressionis high and Pe1 and Ine are occupied by in vivo bound factors [3,38],elevated amount of Hmgb1 does not activate, but rather inhibitstranscription (Fig. 7A). Therefore, we hypothesized that Hmgb1 mayincrease the promoter activity in early steps of chondrogenesis byacting as an architectural protein and facilitating the synergisticactivation by the Sox trio.

To test this hypothesis, we measured the long promoter activity bycoexpressing Sox proteins with optimal amount of Hmgb1. In increas-ing dose, SOX9 gradually increased (2.5–2.7-fold), whereas L-Sox5/Sox6 gradually decreased (by 29–58%) the promoter activity in mesen-chymal cultures (Suppl. Fig. S1B and C).When constant amount of SOX9and increasing amount of L-Sox5/Sox6 were coexpressed with andwithout optimal amount of Hmgb1, Hmgb1 exerted only a small signif-icant increase in the transactivation by SOX9, but it doubled thedose-dependent synergistic activation of SOX9 by L-Sox5/Sox6 in LDMand HDM cultures, by compensating the decline caused by higherdoses of L-Sox5/Sox6 relative to SOX9 (Fig. 7B). In CEF culture, Hmgb1activation (2.1-fold) peaked at optimal dose of L-Sox5/Sox6 versusSOX9, thus raising the promoter activity up to 9.3-fold (Fig. 7B and C).

To further prove that Hmgb1 can facilitate the synergistic promoteractivation by the Sox trio, we forced to express constant amount of Soxproteins and increasing amount of Hmgb1 in COS-7 cells andmonitoredthe expression of the FLAG-tagged proteins in western blots (Fig. 7D).We found that the Sox trio-mediated 17.6-fold enhancement of the pro-moter activity increased further to 38-fold at ~3:1molar ratio of Hmgb1to SOX9.

We concluded that optimal dose of Hmgb1 may facilitate thedose-dependent synergistic activation of the Matn1 promoter by theSox trio in early steps of chondrogenesis or in cultures expressingthe Sox trio at a low level. The activation, however, turns to inhibitionin CEC culture exhibiting high endogenous Sox mRNA levels.

3.7. Hmgb1 facilitates induction of endogenous Matn1 by the Sox trio inforced expression assays

Next we performed forced expression experiments in COS-7 cellsto test whether optimal levels of the chondrogenic Sox proteins aresufficient for ectopic induction of endogenous Matn1 and whetherHmgb1 can enhance the induction. As shown in Table 1, forcedexpression of the Sox trio in optimal L-Sox5/Sox6 dose relative toSox9 increased the very low expression level of the endogenousMatn1 by 26-fold. Forced expression of Hmgb1 together with theSox trio increased the Matn1 expression level by 79-fold as comparedto the vector-transfected control. This confirms the ectopic inductionof endogenous Matn1 by forced expression of Sox trio in the non-expressing COS-7 cells and an additional 3-fold activation by forcedexpression of Hmgb1. We concluded that Hmgb1 can facilitate notonly the Sox trio-mediated activation of the reporter gene driven bythe long Matn1 promoter (Fig. 7D), but it can also increase activationof endogenous Matn1 by the Sox trio.

3.8. Nucleoprotein complex formation on the Matn1 promoter elementswith Hmgb1 in vitro

Since Hmgb1 binds distorted DNA, such as 4-way junction [6], andsince both Pe1 and Ine harbor paired Sox sites within palindromicsequences, which may form a hairpin structure, we assessed Hmgb1

Fig. 4. Functional analysis of the Dpe1 element in forced expression assays. (A) Effect of Dpe1 deletion on the synergistic activation of the long promoter by L-Sox5/Sox6 and SOX9coexpressed at optimal (2.7:1) molar ratio in COS7 cells and at higher ratio in LDM and CEC cultures. The schematic (not drawn to scale) symbolizes factor binding to the short promoter[3] and upstreamelements based on former and present data, respectively. The transcription efficiency in early stage of chondrogenesis (E) is shown (thin arrow). (B)Map of the reportersdriven by four copies of Dpe1 fused to the homologous Matn1 or the heterologous Col2a1 short promoters in direct or reverse orientations as indicated. (C and D) Dose-dependentsynergistic activation of reporters by forced expression of the Sox trio. Luciferase activities are presented as fold values relative to that for AC8Luc (A), FO15Luc (C) or PCLuc (D). *p b 0.05,**p b 0.01, ***p b 0.001 mutants vs. similarly cotransfected wild-type AC8Luc (A) or vs. vector-cotransfected 4×Dpe1-reporters (C and D); #p b 0.05, ##p b 0.01, ###p b 0.001 as indicated(A, C, and D).

binding to the promoter elements in vitro. When we performedsupershift experiments on the Pe1 element, we found that the slow-ly migrating, Sox-specific complex II formed with CEC nuclear pro-teins was supershifted with the Hmgb1-specific antibody (Fig. 8A),indicating that Hmgb1 participated in formation of the Sox-specificcomplex on Pe1. Similar results were obtained on the Ine element.From the 4 complexes formed on Ine with CEC nuclear proteins,the closely migrating complexes I and II were partially supershiftedwith the Hmgb1-specific antibody (Fig. 8B). In consistency with pre-vious data [3], the same complexes were competed withSox-specific oligonucleotides and supershifted with anti-Sox9 anti-body (Fig. 8B and data not shown). We concluded that besides Soxproteins, Hmgb1 can also bind the Pe1 and Ine elements from CECextract.

We next studied the in vitro interaction of the conserved DNAelements with GST-fused Hmgb1. In inverse correlation with theirSOX9-binding efficiency, Dpe1A and Dpe1B bound more stronglywhereas Pe1 and Dpe1C bound weakly to Hmgb1 (Fig. 8C). Ineshowed weak interaction with both proteins. When Hmgb1 wasadded in increasing amount together with Sox factors, it modulated

the binding of Sox factors to the elements in a dose-dependent man-ner (Fig. 8D and E). Curiously, it decreased the binding of SOX9 andL-Sox5 to Dpe1A or, after an inhibition at a low dose, it transientlyincreased the formation of L-Sox5 and SOX9 complexes on Dpe1B athigher doses of Hmgb1 relative to Sox factors. By contrast, Hmgb1increased the binding of Dpe1C, Ine and Pe1 to SOX9 and L-Sox5 atan optimal dose, followed by a decrease at higher dose. Hmgb1 alsomodulated the binding efficiency of SOX9 in the presence of L-Sox5:either decreased both (Ine) or dose-dependently increased the for-mation of SOX9-specific complexes (Pe1 and Dpe1C) or transientlyincreased the formation of L-Sox5 complexes, while transientlyinhibited the SOX9 complexes (Dpe1A and Dpe1B). Pe1M1 mutationdisrupting the paired Sox9motifs, but not the palindrome structure ofPe1, extremely reduced SOX9 binding to the element, but it did notalter the weak recognition by Hmgb1 (compare Figs. 3B, 8C and F).Hmgb1 somewhat increased SOX9 binding to Pe1M1, while it decreasedL-Sox5 binding.

These data show that Hmgb1 can interact with the Matn1 controlelements Dpe1, Pe1 and Ine in vitro and modify the DNA binding effi-ciency of Sox proteins.

Fig. 5. Expression of Hmgb1 is complementary to those of Sox9 and Matn1 in the developing mouse limb. Double immunofluorescence on consecutive cryosections of E12 (A) andE14.5 mouse embryos (B–E). Hematoxylin–eosin staining of the same cryosections is shown below for comparison. The expression domains of Hmgb1 and Sox9 overlap with eachother in early steps of chondrogenesis just before and at the time of activation of Matn1. Note the overlapping Hmgb1 and Sox9 signals e.g. in condensed mesenchyme orprechondrocytes (asterisk) and early chondroblasts of developing metatarsals (arrowhead) (A, B and D), which exhibit no or yet low Matn1 immunosignal, respectively (C and E). Inthe overtly differentiated cartilaginous elements, however, the Hmgb1 signal ceases, complementary to the high Sox9 and Matn1 signals (B–D). Symbols are as defined for Fig. 2. Bars,200 μm (A, D and E); 500 μm (B and C).

3.9. Hmgb1 compensates for the effect of promoter mutations

Next we tested how point mutations in the Sox or other sites ofPe1 and Ine [3] affected the Hmgb1-facilitated activation of theMatn1 long promoter by the Sox trio (Fig. 9). As opposed to the lackof any significant effect in CEC culture (Fig. 9C), Hmgb1 increasedthe Sox trio-mediated synergistic activation of the wild-type promoterand some of its mutants in LDM culture (Fig. 9B). Thus, it doubled theactivity of derivatives ΔPe1M1 or ΔIneM1, carrying mutations in theSox site of Pe1 or the 5′ Sox site of Ine, respectively. It also elevatedthe activity of ΔPe1M4 lacking factor binding to the Sox spacer. Bycontrast, mutation IneM2, disrupting both Sox sites and an unknownfactor binding site of Ine, obstructed the positive effect of Hmgb1,decreasing the activity to half. In spite of the significant, but smallactivation by Hmgb1, deletion of Dpe1ABC or mutation in the 3′ Soxsite of Ine (IneM3) or the double mutant Pe1M1/IneM2 decreased thetransactivation by 60–76%.

Thus in early stage of chondrogenesis, Hmgb1 can partly or fullycompensate for the lack of factor binding to the Sox site of Pe1 andits spacer or to the 5′ Sox site of Ine. However, it could not compen-sate for the lack of factor binding to the 3′ or both Sox sites of Ineor to Dpe1. Pe1M1, Pe1M4, and IneM1 did not destroy the palindromestructure of Pe1 and Ine, whereas IneM3 and IneM2 disrupted one orboth palindromes of Ine, respectively (Fig. 8F) [3,38]. Dpe1 alsoharbors palindrome sequences (Fig. 3). This suggests that Hmgb1facilitated the Sox trio-mediated activation of wild-type and mutantpromoters that maintained the palindrome structure. One possibleexplanation is that Hmgb1 may bind to such palindrome sequencesand by prebending the DNA, it may facilitate the binding of Soxand partner factors to the Matn1 control elements in early chondro-genesis, thereby compensating for the negative effect of pointmutations.

3.10. Effect of Hmgb1 silencing on the expression of cartilage-specific genes

To test the effect of Hmgb1 on the endogenousMatn1 expression inlater steps of chondrogenesis, we performed silencing experiments inchondrogenic cell lines. The established human cell lines of either costalchondrocyte (C-28/I2) or chondrosarcoma origin (SW1353) are knownto express cartilage-specific genes at a relatively low level compared toprimary cultures [40,47], whereas the chondrogenic marker gene ex-pression of the RCS cell line is more similar to that of primary cultures[41]. In consistencywith these data, wemeasured lowmRNA levels rel-ative to that of the S18 ribosomal proteinmRNA in the human cell linesfor MATN1 (3.99 × 10−5 in C-28/I2 and 9.74 × 10−5 in SW1353 cells)and COL2A1 (4 × 10−5 in C-28/I2 and 1 × 10−5 in SW1353 cells)(Fig. 10A and B). The mRNA levels for SOX9 were also relatively low(7.16 × 10−2 in C-28/I2 and 4.74 × 10−2 in SW1353 cells), in sharpcontrast to their highly elevated HMGB1 mRNA expression (1.59 inC-28/I2 and 1.46 in SW1353 cells). RCS cells exhibited much higherrelative mRNA levels for Col2a1 (8 × 10−2), Sox9 (2.51) and Matn1(3.56 × 10−4), whereas Hmgb1 mRNA expression (1.39 × 10−1) waslower in the rat than in the human cell lines (Fig. 10C).

As these cell lines differed from primary cultures in their highHMGB1 and low chondrogenic marker gene expressions, they providedexcellent tools to test the hypothesis in silencing experiments whetherthe highly elevated expression of HMGB1 compared to that of SOX9may contribute to the repression of the chondrogenic marker genesin later stage of chondrogenesis. We found that silencing of HMGB1by 2.5-fold in C-28/I2 cells increased the expression level for MATN1and COL2A1 by 3.6–4.3-fold and 4.9–5.3-fold, respectively (Fig. 10A).Silencing of HMGB1 by 2.4–2.7-fold in SW1353 cells increased the ex-pression level for MATN1 and COL2A1 by 9.6–32-fold and 10–13-fold,respectively (Fig. 10B). The activation was specific, as it affected neitherthe GAPDH nor the SOX9mRNA levels (Fig. 10A and B). When silencing

Fig. 6. Complementary expression ofHmgb1 and cartilage-specific genes in chondrogeniccultures. (A–D) Marker mRNA levels were determined by QRT-PCR in HDM cultureundergoing chondrogenesis in vitro and compared to mRNA levels of non-expressingCEF and high Matn1-expressing CEC cultures. Cτ values were normalized to the averageCτ values of three internal control genes. Relative expression levels are presented as foldvalues relative to the HDM day 0 values.

decreased the relative Hmgb1 mRNA level of RCS cells by 42–57-foldless than that of Sox9, theMatn1 and the Col2a1mRNA levels increasedby 59–74-fold and 9.3–15.9-fold, respectively (Fig. 10C). Hmgb1 silenc-ing did not significantly alter theHprt or Sox9 levels, suggesting a specificeffect.

As Hmgb1 silencing activated Matn1 and Col2a1 in RCS and simi-larly their orthologs in human chondrogenic cell lines, we concludedthat elevated Hmgb1 versus Sox9 expression may inhibit the cartilageprotein gene expression in these cells. In consistency with the repres-sion of the Matn1 promoter activity by high dose of Hmgb1 in CECculture (Fig. 7A), these data support the hypothesis that large abun-dance of Hmgb1 may interfere with the Sox trio-mediated activationof Matn1 and Col2a1.

3.11. ChIP analysis of cell type-dependent in vivo binding of HMGB1 andSOX trio to the conserved MATN1 control elements

Next we assessed the interaction of Hmgb1 and the Sox trio withMatn1 control elements in vivo. As Dpe1, Pe1 and Ine elements arewell conserved in amniotes [3,38], and the in vivo occupancy of theMatn1 promoter elements has already been reported based on geno-mic footprinting in chicken primary cultures [3], we decided to per-form ChIP analysis in human and rat chondrogenic cell lines, usinghuman 293T cells as non-expressing controls. Due to the resolutionof the ChIP technique, the MATN1 Pe1 and Ine1 elements could notbe dissected.

Our results showed cell type-specific in vivo binding of the SOXtrio to the MATN1 DNA elements in SW1353 and C-28/I2 cells, whilethe low occupancy of these elements resembled to that of theTUBB3, IL1B and CRP control genes in 293T cells (Fig. 11A). InSW1353 cells, Dpe1 bound to SOX9 and SOX5 with ≥3.7-fold higherefficiency than the combined Pe1–Ine elements, which did not exceedthe occupancy of TUBB3 and IL1B controls. In C-28/I2 cells, Dpe1 wasalso bound by the SOX trio about 2–3-fold more efficiently thanPe1–Ine or the TUBB3 control. Interestingly, however, similar oreven higher (≥2-fold) SOX5 and SOX9 occupancy was seen on theIL1B and CRP sequences, suggesting that their putative SOX motifscan bind SOX factors and may contribute to the regulation of thesegenes in the latter cell line. In consistency with our data, IL1B can beinduced in human articular chondrocytes [48] and various cancersalso have raised CRP levels [49]. In agreement with the high HMGB1expression of SW1353 and C-28/I2 cells (Fig. 10A and B), eachestablished human cell line exhibited ~10-fold higher HMGB1 thanSox9 binding to the Matn1 DNA elements (Fig. 11B). Interestingly,however, except the low occupancy of the IL1B control in C-28/I2cells, we also observed elevated but highly variable HMGB1 bindingto the control genes, especially in SW1353 cells. TUBB3 and CRP arealso expressed in certain tumors [50] and HMGB1 occupancy mayaffect binding of SOX and other factors to their DNA elements. Inline with our data in 293T and SW1353 cells (Fig. 11B), HMGB1 bind-ing to the IL1B promoter was also reported in another cell [51].In C-28/I2 cells, however, the low HMGB1 occupancy of the ILIB con-trol sequence may not impede SOX factor binding to the element(compare Fig. 11A and B). Dpe1 presented the strongest binding rel-ative to the three controls in SW1353 followed by 293T, while thecombined Pe1–Ine elements bound to HMGB1 with similar or lowerefficiency than the control genes.

In RCS cells, Sox9 bound Dpe1, Pe1 and Ine most strongly andwith similar affinity among the Sox factors, but it recognized theTubb3 and Crp elements even more efficiently (Fig. 11C). Hmgb1 alsobound ~10-fold more strongly the Matn1 DNA elements than Sox9(Fig. 11D). It occupied theMatn1 DNA elements with the same efficien-cy as the Tubb3 control, but with lower efficiency as the Crp control.

The high occupancy of the conserved MATN1 DNA elements byHMGB1 and the low occupancy of the Pe1 key element by the SOXtrio can explain the low level of cartilage-specific gene expression inhuman chondrosarcoma and immortalized chondrocyte cell lines. InRCS cells, Sox9-binding efficiency of the Pe1 key element was similarto that of Dpe1, but the Hmgb1 occupancy of the elements was stillhigh, in agreement with the low Matn1 expression of this tumorigeniccell line and with the sharp induction of Matn1 upon Hmgb1 silencing.

Fig. 7. Effect of Hmgb1 on the synergistic activation of Matn1 promoter by the Sox trio. (A) Transient activation of the long promoter by Hmgb1 in cotransfection assays. (B and C)Effect of coexpressed Hmgb1 and Sox trio on the long promoter activity in mesenchymal cells. (D) Forced expression of increasing amount of FLAG-tagged Hmgb1 and constantamount of Sox trio in COS-7 cells combined with western analysis using anti-FLAG antibody. Nonspecific bands are marked with asterisk. Luciferase activities are presented asfold values relative to that for AC8Luc (A, B and D) or for Hmgb1-cotransfected AC8Luc (C). *p b 0.05, **p b 0.01, ***p b 0.001 vs. vector-cotransfected reporter (A–D); #p b 0.05,##p b 0.01, ###p b 0.001 as indicated (B–D).

3.12. Model for the interplay between Hmgb1 and Sox trio bound to theconserved Matn1 control elements at subsequent steps of chondrogenesisin amniotes

Based on the data, we suggest a model for the contribution ofHmgb1 to the activation of the gene at the onset of chondrogenesisin amniotes (Fig. 11E). Hmgb1 is bound to the Dpe1, Pe1 and Ineelements in fibroblasts and in committed mesenchyme (Fig. 11Ea).

Table 1Induction of the endogenous Matn1 in the nonexpressing COS-7 cells by forced expres-sion of the Sox trio and Hmgb1.

Transcription factorsexpressed by force

Matn1 expression level

2−ΔCτ ± SEMa Fold p value

Empty vector 1.95E − 06 ± 1.47E − 07 1Hmgb1 2.42E − 06 ± 2.87E − 07 1.24 9.90E − 01Sox trio 5.03E − 05 ± 5.64E − 06 25.76 7.08E − 03Sox trio + Hmgb1 1.54E − 04 ± 2.65E − 05 78.68 2.25E − 05

a Matn1 mRNA levels are given relative to the invariant Gapdh mRNA level.

The architectural protein may fluidize the chromatin and bend theDNA to facilitate sequence-specific binding of the Sox trio to theseDNA elements in early steps of chondrogenesis (Fig. 11Eb). Thetranscriptional activity of the gene is increasing as Hmgb1-bindingis replaced by Sox9-binding and L-Sox5/Sox6 increases thetransactivation by Sox9 in a dose-dependent synergistic manner inlate proliferative chondroblasts, as reported previously [3] (Fig. 11Ec).However, highly elevated Hmgb1 expression in transfected late prolifer-ative chondroblasts or oncogenically transformed chondrogenic cells(e.g. chondrosarcomas), can repress the gene expression, likely becauseHmgb1 present in large abundance may compete with Sox9- and Soxtrio-binding to the conserved DNA elements (Fig. 11Ed).

4. Discussion

This work reveals a previously unknown role for Hmgb1, and givesnew insight into cartilage-specific gene regulation, by extending our for-mer model [3] on the unique transcriptional control mechanismsunderlying Matn1 expression in amniotes. This model implies thatSox-mediated interaction between the proximal promoter and upstream

Fig. 8. Hmgb1 binds to conserved Matn1 control elements in vitro. (A and B) Anti-Hmgb1 antibody can supershift the Sox-specific complexes formed on Pe1 and Ine with CEC nuclearproteins. (C) Purified GST-fused Hmgb1 binds the DNA elements with variable efficiency in EMSA. Binding of purified SOX9 and L-Sox5 to Dpe1 (D) and short promoter elements Ineand Pe1 (E) in the presence of increasing amount of purified Hmgb1. (F) Hmgb1 binds with similar efficiency to Pe1 and Pe1M1 harboring palindrome sequences. Ab, antibody; F, freeprobe; H, Hmgb1 complex; PI, preimmune serum; S5, L-Sox5 complex; S9, SOX9 complex; S9c, Sox9-specific competitor.

enhancer elements results in high spatiotemporalMatn1 activity. Formerreports revealed that a unique set of conserved promoter elements play adominant role in restricting the cartilage-specific activity, with a centralrole for Pe1 [3,21,38]. Chondroblast stage-dependent regulation isachieved by dose-dependentmodulation of the Pe1-bound SOX9 activityby L-Sox5/Sox6bound to Ine andNfiproteins bound to SI and SII. Hereweextended thismodel by demonstrating that the conserved Dpe1 is an im-portant upstream enhancer element that works in transgenic mice andmay account in large part for the high Sox-mediated enhancement ofthe Matn1 promoter in late proliferative chondroblasts for the followingreasons. Dpe1 features three Sox sites binding SOX9 and L-Sox5with op-posite efficiency in vitro. It is needed for the high chondroblaststage-specific promoter activity and transactivation by the Sox trio.Dpe1 elements can exert a Sox trio-mediated, dose-dependent synergis-tic enhancement to theMatn1 and Col2a1 promoters.

Furthermore, we also extended this model by proposing a role forHmgb1 at the onset of chondrogenesis based on the following obser-vations. Hmgb1−/− mice were previously shown to suffer from pleio-tropic defects. They die from hypoglycemia after birth, but survivorsrescued by glucose treatment show severe developmental retardationand abnormalities in skeletal development [52]. Detailed analysis re-vealed a delay in endochondral ossification, largely because the lackof Hmgb1 secretion by hypertrophic chondrocytes impaired cartilageinvasion by blood vessels, osteoclasts, and osteoblasts [2]. The size of

the developing cartilaginous elements stained by alcian blue was alsosmaller in E16.5 Hmgb1−/− embryos than in wild-type littermates [2],suggesting that early steps of endochondral bone formation may alsobe affected by themutation. Furthermore, Hmgb1 andHmgb2 are need-ed for posterior digit development [53]. In line with in situ hybridiza-tions by others [2,53], we found that Hmgb1 expression declinedduring in vivo and in vitro chondrogenesis in inverse correlation withthe activation of chondrogenic Sox and ECM genes, showing some over-lap with raising Sox9 expression just before and at the time of Matn1onset in early chondroblasts. Hmgb1 and SOX9 recognized the Dpe1,Pe1 and Ine elements with a reciprocal binding efficiency in vitro. Inforced expression assays, optimal dose of Hmgb1 augmented the Soxtrio-mediated activation of the Matn1 promoter in early chondrogene-sis as well as the ectopic induction of the endogenous Matn1 in COS-7cells by 2–3-fold. Similar effect of Hmgb1 transfection was reported inother systems [54]. Hmgb1 could in large part compensate for the effectof mutations in Pe1 and in the 5′ Sox site of Ine in early chondrogenesis.Former genomic footprinting demonstrated the increasing occupancyof Sox motifs of the Pe1 and Ine elements of Matn1 at nucleotideresolution in chicken primary cultures representing subsequent stepsof chondrogenesis [3,38]. In agreement with these observations, ChIPexperiments revealed specific binding of Sox9, followed by Sox5/Sox6,but even stronger binding of Hmgb1 to the conserved Dpe1, Pe1 andIne elements ofMatn1 in RCS cells and likewise in human chondrogenic

Fig. 9. Effect of Hmgb1 on the Sox trio-mediated transactivation of the long promoter and its mutant derivatives. (A) The schematic (not drawn to scale) illustrates factor binding tothe short promoter and upstream elements based on previous [3] and present data. Arrows indicate transcriptional activities at early stage of chondrogenesis (E). Sox proteins werecoexpressed with and without Hmgb1 in LDM (B) and CEC cultures (C) and reporter activities were expressed in fold values relative to that for AC8Luc cotransfected with vectors.*p b 0.05, **p b 0.01, ***p b 0.001 mutants vs. similarly cotransfected AC8Luc; #p b 0.05, ##p b 0.01, ###p b 0.001 as indicated.

cell lines. This was consistent with the high Hmgb1 but low Matn1 andCol2a1 expression of these cells. Elevated Hmgb1 level may be due tothe oncogenic transformation of these cells, since Hmgb1 is upregulatedin cancer cells and plays a role in tumor growth [55]. Silencing ofHMGB1 specifically and markedly increased the expression of cartilageprotein genes but not that of SOX9. When silencing decreased theHmgb1 mRNA level below 2% of that of Sox9 in RCS cells, the Matn1and Col2a1 expression increased even more dramatically, indicatingthat large abundance of Hmgb1 can inhibit the Sox9-mediated activa-tion of cartilage protein genes. Hmgb1 overexpression in CEC alsoinhibited the Matn1 promoter activity, indicating a conserved mecha-nism in amniotes. Our data strongly suggests that, by modulating theccess of Sox factors to evolutionarily conserved DNA elements, Hmgb1can facilitate the synergistic activation of Matn1 by the Sox trio in earlychondrogenesis, which turns to inhibition later.

Considering the present and previous results [3,21,38], we pro-pose the following model for the distinctive control mechanism ofMatn1 in amniotes (Fig. 11E). Lack of in vivo footprints in CEF andslow occupancy of DNA elements during chondrogenesis in HDM cul-ture [3,38] suggest that the chromatin structure is largely closed andclassical transcription factors are not bound to theMatn1 promoter inembryonic mesenchyme. However, based on the high Hmgb1 expres-sion in CEF and committed mesenchyme as well as its very efficient tomoderate binding to Dpe1, Pe1 and Ine in human fibroblasts and CEFin ChIP assays, Hmgb1 may bind the conserved Matn1 DNA elementslikely covered by nucleosomes in these cells (Fig. 11Ea). Since Hmgb1can displace histone H1 from nucleosomes [8], it may fluidize thenucleosome structure, and thereby help recruit Sox factors to theircognate sites on Pe1, Ine and Dpe1, when the Sox genes are turnedon at the onset of chondrogenesis (Fig. 11Ea).

As Sox9 level raises, while Hmgb1 level declines during chondrogen-esis, Sox proteinsmay gradually displace Hmgb1 from theDNA elementsdue to their stronger and sequence-specific binding (Fig. 11Eb). Pe1 andIne can preferably bind SOX9 and L-Sox5/Sox6, respectively, whereas SIand SII can interact with Nfi proteins in vitro [3,38,39]. In genomicfootprinting, the Sox motifs of Pe1, the 5′ Sox site of Ine and the Nfi

motifs of SI and SII were first occupied by in vivo bound transcription fac-tors, in line with the transient activation of Nfi genes in early chondro-genesis [3]. As Nfi proteins can interact with histones and polymeraseII [56], Nfi binding to SI near TATA may also help to disrupt the nucleo-somes and to recruit the preinitiation complex to the promoter(Fig. 11Ea and b). We found that Sox9 binding to Pe1 plays a crucialrole in this regulation and L-Sox5/Sox6 binding to Ine may increase theefficiency of Sox9-binding to the nearby Pe1 [3], similar to the observa-tions on the Acan and Col2a1 enhancers [29]. Dpe1may enhance the pro-moter activity via Sox9 likely bound to site C first at a low L-Sox5/Sox6dose to Sox9. L-Sox5/Sox6 bound to sites A and B at higher doses maylikewise secure Sox9 binding to Dpe1 (Fig. 11Eb). Our data support akey role for Dpe1 in the Sox-mediated promoter enhancement, likelyvia interaction with Pe1. In early proliferative chondroblasts, however,the transcription activity is still low, due to the low abundance of Soxand Nfi factors and the low occupancy of the binding sites.

In late proliferative chondroblasts expressing L-Sox5/Sox6 andNfi inoptimal ratio relative to Sox9, optimal binding of Sox factors to Ine andDpe1may result in full occupancy of Pe1 by Sox9 and efficient enhance-ment by Dpe1 and other distal elements (Fig. 11Ec). At the same time,Hmgb1 level and binding to the elements drop. Nfi proteins alsodose-dependently modulated the activation by SOX9 in forced expres-sion assays [3]. Since they can interact with general transcription factorsand coactivators [56], optimal occupancy of the Nfi sites may also pro-mote the preinitiation complex assembly, enhanceosome formationand transactivation. As a result, the transcription activity is highlyincreased in late proliferative chondroblasts (Fig. 11Ec). Based on tran-sient expressions and transgenic mice data in this and previous studies[3,21], besides Dpe1, cartilage-specific DNA elements between −1791and −334 may also contribute to the promoter activation. Pe1 andIne also interacted with other unidentified factors, including those af-fecting proximo-distal patterning [3 and data not shown] (Fig. 11Eband c). Such factors may also bind Dpe1, thus explaining the strongdistal structure preference of the 8×Dpe1(+)NAD1 transgene (Fig. 2).

If the Hmgb1 abundance relative to Sox9 greatly increases for anyreason (e.g. transfection, tumorigenesis, oncogenic transformation,

Fig. 10. Effect of Hmgb1 silencing on chondrogenic marker gene expression. (A and B) Established human chondrogenic cell lines were transfected with increasing amount ofHMGB1 siRNA (100 pmol and 200 pmol for C-28/I2; 200 pmol and 400 pmol for SW1353). Parallel plates transfected with control siRNA (Ctrl) and GAPDH siRNA served as neg-ative and positive controls, respectively, to indicate the efficiency and specificity of silencing. (C) Similar transfection experiments were performed in the RCS cell line using increas-ing amount (200 pmol and 400 pmol) of rat Hmgb1 siRNA and negative control siRNA. (A–C) Marker gene expression levels were determined by QRT-PCR and plotted relative tothe invariant RPS18/rps18 mRNA levels. *p b 0.05, **p b 0.01, ***p b 0.001 vs. nontransfected samples.

etc.) in overtly differentiated chondroblasts, Hmgb1 may interferewith the binding of Sox trio to the conserved Matn1 control elementsby competing with Sox proteins for binding to their cognate sites onDpe1, Pe1 and Ine (Fig. 11Ed). This can explain the decreased Matn1promoter activity in CEC culture overexpressing Hmgb1 (Fig. 7A)and the low endogenous MATN1/Matn1 expression level in perma-nent chondrogenic cell lines (Fig. 10). High abundance of Hmgb1may interfere with Sox trio binding to control elements of Col2a1and other cartilage protein genes, thus explaining the low chondrogenicmarker gene expression of permanent cell lines relative to primarycultures in consistency with other reports [40,41,47].

Synergistic activation of the promoter by the Sox trio resembles themode of action of the Col2a1 and Acan enhancers [28,29]. The controlmechanism utilized by Matn1, however, differs from those of othercartilage-specific genes in the unique assembly of DNA elementsaround the TATA motif, which leads to dose-dependent fine tuning of

the activity of the Pe1-bound Sox9by L-Sox5/Sox6 andNfi [3]. As the ac-tivation of Sox andNfi genes follows a different kinetics, the abundancesof L-Sox5/Sox6 and Nfi relative to Sox9 vary during chondrogenesis,resulting in stage-dependent alterations in theMatn1 promoter activity.Sox9 is known to interact with many partner factors [5,12], some ofthese may affect proximo-distal patterning. Partner factor GLI was re-ported to inhibit the transactivation by SOX9 to repress Col10a1 beforehypertrophy [57].

Hmgb1 is known to enhance and stabilize the binding of variousgene-specific and basal transcription factors and steroid receptors totarget sequences and regulate the transcription of many cellular andviral genes [4,6,7,58]. Considering the role of Sox proteins in tissue-specific gene regulation [1,12], our model suggests an important novelmechanism: depending on their relative abundance, Hmgb1may regu-late transcription by modulating the access of Sox factors to their DNAelements. The effect is dual. In early chondrogenesis or in optimal

Fig. 11. ChIP analysis of Hmgb1 and Sox trio binding to the conservedMatn1 control elements in amniotes. ChIP study was performed with SOX trio- (A and C) and HMGB1-specific(B and D) antibodies in SW1353, C-28/I2 and 293T human cell lines (A and B) and in rat RCS cells (C and D). The immunoprecipitated DNA was measured with QRT-PCR assayscovering the combined Pe1–Ine and Dpe1 elements of MATN1 and the separate Dpe1, Pe1, and Ine elements of the rat Matn1. Occupancy of the control genes TUBB3/Tubb3, IL1Band CRP/Crp is shown for comparison. Columns represent results measured with SOX5-, SOX6-, SOX9-, and HMGB1-specific antibodies and IgG controls. Values represent enrich-ment relative to input DNA. Data are presented as mean ± SEM. (E) Model for the regulation of theMatn1 promoter by Hmgb1 and Sox trio. Schematic illustration of factor bindingto the conserved DNA elements at the onset of chondrogenesis (a), in early (b) and late proliferative chondroblasts (c) and in oncogene-transformed cell lines (d). See text for de-tailed description.

dose relative to the Sox trio in COS-7 cells, Hmgb1 may fluidize the nu-cleosome structure and facilitate the Sox trio-mediated activation of theMatn1 promoter. In overtly differentiated chondroblasts, Hmgb1 levelfalls off and the Sox trio can exert promoter activation. High dose ofHmgb1 in later steps of chondrogenesis or in permanent chondrogeniccell lines, however, inhibits the Matn1 promoter, possibly by competingfor Sox binding sites, and thereby represses the Matn1 expression.

Activation ofHmgb1 in prehypertrophic chondrocytes [2]may contributeto the repression of Matn1 during hypertrophy.

Hmgb1 may also regulate other cartilage ECM genes by modulatingSox factor binding to their control elements. As Hmgb1 also works asalarmin and its level is high in the joints of patients with rheumatoidarthritis [59,60], it may inhibit the activity ofMatn1 and other cartilageECM genes in such patients.

Acknowledgements

We are grateful to V. Lefebvre for critical reading of the manuscript,to D.P. Edward for the gift of Hmgb1-specific antibody and expressionplasmid, to M. Paulsson for the anti-Matn1 antibody, and to B. deCrombrugghe and M.B. Goldring for the gift of the RCS and the C-28I/2cell lines, respectively. We thank A. Simon, K. Hegedűs, E. Horváth,M. Balogh, and K. Kávai for excellent technical assistance and M. Tóthfor the artwork.

This work was supported by grants OTKA T049608 (to I.K.),PD50006 (to E.K.) and PD101421 (to L.M.) from the HungarianNationalScientific Research Foundation and by grants GVOP-3.1.1.-2004-05-0290/3.0 (to I.K.) and GOP-1.1.1-11-2011-0003 (to Avidin Ltd)from the Economic Competitiveness Operative Programme of the Na-tional Development Plan. Á.Z. and B.L.B. were supported by János Bolyaifellowship of Hungarian Academy of Sciences (BO/00781/11 and BO/795/08). B.L.B. was also supported by Szodoray Fellowship of the Uni-versity of Debrecen. ChIP experiments were performed at the Centerfor Clinical Genomics and Personalized Medicine of the University ofDebrecen.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.bbagrm.2013.07.004.

References

[1] V. Lefebvre, B. Dumitriu, A. Penzo-Mendez, Y. Han, B. Pallavi, Control of cell fateand differentiation by Sry-related high-mobility-group box (Sox) transcriptionfactors, Int. J. Biochem. Cell Biol. 39 (2007) 2195–2214.

[2] N. Taniguchi, K. Yoshida, T. Ito, M. Tsuda, Y. Mishima, T. Furumatsu, L. Ronfani, K.Abeyama, K. Kawahara, S. Komiya, I. Maruyama, M. Lotz, M.E. Bianchi, H. Asahara,Stage-specific secretion of HMGB1 in cartilage regulates endochondral ossifica-tion, Mol. Cell. Biol. 27 (2007) 5650–5663.

[3] A. Nagy, E. Kénesi, O. Rentsendorj, A. Molnár, T. Szénási, I. Sinkó, A. Zvara, S.T.Oommen, E. Barta, L.G. Puskás, V. Lefebvre, I. Kiss, Evolutionarily conserved,growth plate zone-specific regulation of the matrilin-1 promoter: L-Sox5/Sox6and Nfi factors bound near TATA finely tune activation by Sox9, Mol. Cell. Biol.31 (2011) 686–699.

[4] M.E. Bianchi, A. Agresti, HMG proteins: dynamic players in gene regulation anddifferentiation, Curr. Opin. Genet. Dev. 15 (2005) 496–506.

[5] Y. Kamachi, M. Uchikawa, H. Kondoh, Pairing SOX off: with partners in the regu-lation of embryonic development, Trends Genet. 16 (2000) 182–187.

[6] A. Agresti, M.E. Bianchi, HMGB proteins and gene expression, Curr. Opin. Genet.Dev. 13 (2003) 170–178.

[7] M. Stros, HMGB proteins: interactions with DNA and chromatin, Biochim. Biophys.Acta 1799 (2010) 101–113.

[8] J.O. Thomas, K. Stott, H1 and HMGB1: modulators of chromatin structure, Biochem.Soc. Trans. 40 (2012) 341–346.

[9] M.B. Goldring, K. Tsuchimochi, K. Ijiri, The control of chondrogenesis, J. Cell. Biochem.97 (2006) 33–44.

[10] E.J. Mackie, L. Tatarczuch, M. Mirams, The skeleton: a multi-functional complexorgan: the growth plate chondrocyte and endochondral ossification, J. Endocrinol.211 (2011) 109–121.

[11] L. Quintana, N.I. zur Nieden, C.E. Semino, Morphogenetic and regulatory mecha-nisms during developmental chondrogenesis: new paradigms for cartilage tissueengineering, Tissue Eng. B Rev. 15 (2009) 29–41.

[12] V. Lefebvre, P. Smits, Transcriptional control of chondrocyte fate and differentia-tion, Birth Defects Res. C Embryo Today 75 (2005) 200–212.

[13] H. Shimizu, S. Yokoyama, H. Asahara, Growth and differentiation of the developinglimb bud from the perspective of chondrogenesis, Dev. Growth Differ. 49 (2007)449–454.

[14] R. Cancedda, F. Descalzi Cancedda, P. Castagnola, Chondrocyte differentiation, Int.Rev. Cytol. 159 (1995) 265–358.

[15] N.S. Stirpe, P.F. Goetinck, Gene regulation during cartilage differentiation: temporaland spatial expression of link protein and cartilage matrix protein in the developinglimb, Development 107 (1989) 23–33.

[16] S. Mundlos, B. Zabel, Developmental expression of human cartilage matrix protein,Dev. Dyn. 199 (1994) 241–252.

[17] S.Muratoglu, C. Bachrati, M.Malpeli, P. Szabó,M. Neri, B. Dozin, F. Deák, R. Cancedda,I. Kiss, Expression of the cartilage matrix protein gene at different chondrocytedevelopmental stages, Eur. J. Cell Biol. 68 (1995) 411–418.

[18] V. Szűts, U. Mollers, K. Bittner, G. Schurmann, S. Muratoglu, F. Deák, I. Kiss, P.Bruckner, Terminal differentiation of chondrocytes is arrested at distinct stagesidentified by their expression repertoire of marker genes, Matrix Biol. 17 (1998)435–448.

[19] A. Aszódi, N. Hauser, D. Studer, M. Paulsson, L. Hiripi, Z. Bősze, Cloning, sequencingand expression analysis of mouse cartilage matrix protein cDNA, Eur. J. Biochem.236 (1996) 970–977.

[20] A. Aszódi, L. Módis, A. Páldi, A. Rencendorj, I. Kiss, Z. Bősze, The zonal expressionof chicken cartilage matrix protein gene in the developing skeleton of transgenicmice, Matrix Biol. 14 (1994) 181–190.

[21] I. Karcagi, T. Rauch, L. Hiripi, O. Rentsendorj, A. Nagy, Z. Bősze, I. Kiss, Functionalanalysis of the regulatory regions of the matrilin-1 gene in transgenic micereveals modular arrangement of tissue-specific control elements, Matrix Biol.22 (2004) 605–618.

[22] H. Akiyama, V. Lefebvre, Unraveling the transcriptional regulatory machinery inchondrogenesis, J. Bone Miner. Metab. 29 (2011) 390–395.

[23] W. Bi, J.M. Deng, Z. Zhang, R.R. Behringer, B. de Crombrugghe, Sox9 is required forcartilage formation, Nat. Genet. 22 (1999) 85–89.

[24] H. Akiyama, M.C. Chaboissier, J.F. Martin, A. Schedl, B. de Crombrugghe, The tran-scription factor Sox9 has essential roles in successive steps of the chondrocytedifferentiation pathway and is required for expression of Sox5 and Sox6, GenesDev. 16 (2002) 2813–2828.

[25] P. Smits, P. Dy, S. Mitra, V. Lefebvre, Sox5 and Sox6 are needed to develop andmaintain source, columnar, and hypertrophic chondrocytes in the cartilagegrowth plate, J. Cell Biol. 164 (2004) 747–758.

[26] P. Smits, P. Li, J. Mandel, Z. Zhang, J.M. Deng, R.R. Behringer, B. de Crombrugghe, V.Lefebvre, The transcription factors L-Sox5 and Sox6 are essential for cartilageformation, Dev. Cell 1 (2001) 277–290.

[27] T. Ikeda, S. Kamekura, A. Mabuchi, I. Kou, S. Seki, T. Takato, K. Nakamura, H.Kawaguchi, S. Ikegawa, U.I. Chung, The combination of SOX5, SOX6, and SOX9(the SOX trio) provides signals sufficient for induction of permanent cartilage,Arthritis Rheum. 50 (2004) 3561–3573.

[28] V. Lefebvre, P. Li, B. de Crombrugghe, A new long form of Sox5 (L-Sox5), Sox6 andSox9 are coexpressed in chondrogenesis and cooperatively activate the type IIcollagen gene, EMBO J. 17 (1998) 5718–5733.

[29] Y. Han, V. Lefebvre, L-Sox5 and Sox6 drive expression of the aggrecan gene incartilage by securing binding of Sox9 to a far-upstream enhancer, Mol. Cell.Biol. 28 (2008) 4999–5013.

[30] G. Karsenty, H.M. Kronenberg, C. Settembre, Genetic control of bone formation,Annu. Rev. Cell Dev. Biol. 25 (2009) 629–648.

[31] Q. Wu, Y. Zhang, Q. Chen, Indian hedgehog is an essential component ofmechanotransduction complex to stimulate chondrocyte proliferation, J. Biol.Chem. 276 (2001) 35290–35296.

[32] F. Deák, R. Wagener, I. Kiss, M. Paulsson, The matrilins: a novel family of oligomericextracellular matrix proteins, Matrix Biol. 18 (1999) 55–64.

[33] A.R. Klatt, A.K. Becker, C.D. Neacsu,M. Paulsson, R.Wagener, Thematrilins: modulatorsof extracellular matrix assembly, Int. J. Biochem. Cell Biol. 43 (2011) 320–330.

[34] R. Wagener, H.W. Ehlen, Y.P. Ko, B. Kobbe, H.H. Mann, G. Sengle, M. Paulsson, Thematrilins—adaptor proteins in the extracellular matrix, FEBS Lett. 579 (2005)3323–3329.

[35] K. Kanbe, X. Yang, L. Wei, C. Sun, Q. Chen, Pericellular matrilins regulate activationof chondrocytes by cyclic load-induced matrix deformation, J. Bone Miner. Res. 22(2007) 318–328.

[36] J.Y. Jang, E.K. Park, H.M. Ryoo, H.I. Shin, T.H. Kim, J.S. Jang, H.S. Park, J.Y. Choi, T.G.Kwon, Polymorphisms in the matrilin-1 gene and risk of mandibular prognathismin Koreans, J. Dent. Res. 89 (2010) 1203–1207.

[37] M.E. Pedersen, H. Takle, E. Ytteborg, E. Veiseth-Kent, G. Enersen, E. Faergestad, G.Baeverfjord, K.O. Hannesson, Matrilin-1 expression is increased in the vertebralcolumn of Atlantic salmon (Salmo salar L.) individuals displaying spinal fusions,Fish Physiol. Biochem. 37 (2011) 821–831.

[38] O. Rentsendorj, A. Nagy, I. Sinkó, A. Daraba, E. Barta, I. Kiss, Highly conservedproximal promoter element harbouring paired Sox9-binding sites contributesto the tissue- and developmental stage-specific activity of the matrilin-1 gene,Biochem. J. 389 (2005) 705–716.

[39] P. Szabó, J. Moitra, A. Rencendorj, G. Rákhely, T. Rauch, I. Kiss, Identification of anuclear factor-I family protein-binding site in the silencer region of the cartilagematrix protein gene, J. Biol. Chem. 270 (1995) 10212–10221.

[40] M.B. Goldring, J.R. Birkhead, L.-F. Suen, R. Yamin, S. Mizuno, J. Glowacki, J.L.Arbiser, J.F. Apperley, Interleukin-1β-modulated gene expression in immortalizedhuman chondrocytes, J. Clin. Invest. 94 (1994) 2307–2316.

[41] K. Mukhopadhyay, V. Lefebvre, G. Zhou, S. Garofalo, J.H. Kimura, B. de Crombrugghe,Use of a new rat chondrosarcoma cell line to delineate a 119-base pair chondrocyte-specific enhancer element and to define active promoter segments in the mousepro-α1(II) collagen gene, J. Biol. Chem. 270 (1995) 27711–27719.

[42] W. Doppler, M. Windegger, C. Soratroi, J. Tomasi, J. Lechner, S. Rusconi, A.C. Cato, T.Almlof, J. Liden, S. Okret, J.A. Gustafsson, H. Richard-Foy, D.B. Starr, H. Klocker, D.Edwards, S. Geymayer, Expression level-dependent contribution of glucocorticoidreceptor domains for functional interaction with STAT5, Mol. Cell. Biol. 21 (2001)3266–3279.

[43] N. Hauser, M. Paulsson, D. Heinegard, M. Morgelin, Interaction of cartilage matrixprotein with aggrecan. Increased covalent cross-linking with tissue maturation,J. Biol. Chem. 271 (1996) 32247–32252.

[44] B.L. Bálint, P. Gábor, L. Nagy, Genome-wide localization of histone 4 arginine 3methylation in a differentiation primed myeloid leukemia cell line, Immunobiology210 (2005) 141–152.

[45] S. Mertin, S.G. McDowall, V.R. Harley, The DNA-binding specificity of SOX9 andother SOX proteins, Nucleic Acids Res. 27 (1999) 1359–1364.

[46] T. Furusawa, J.H. Lim, F. Catez, Y. Birger, S. Mackem, M. Bustin, Down-regulation ofnucleosomal binding protein HMGN1 expression during embryogenesis modulatesSox9 expression in chondrocytes, Mol. Cell. Biol. 26 (2006) 592–604.

[47] M. Gebauer, J. Saas, F. Sohler, J. Haag, S. Soder, M. Pieper, E. Bartnik, J. Beninga, R.Zimmer, T. Aigner, Comparison of the chondrosarcoma cell line SW1353with primaryhuman adult articular chondrocytes with regard to their gene expression profile andreactivity to IL-1β, Osteoarthr. Cartil. 13 (2005) 697–708.

[48] K. Andreas, C. Lubke, T. Haupl, T. Dehne, L. Morawietz, J. Ringe, C. Kaps, M. Sittinger,Key regulatorymolecules of cartilage destruction in rheumatoid arthritis: an in vitrostudy, Arthritis Res. Ther. 10 (2008) R9.

[49] T. Nakamura, R.J. Grimer, C.L. Gaston, M. Watanuki, A. Sudo, L. Jeys, The prognosticvalue of the serum level of C-reactive protein for the survival of patients with aprimary sarcoma of bone, Bone Joint J. 95-B (2013) 411–418.

[50] M. Mariani, S. Shahabi, S. Sieber, G. Scambia, C. Ferlini, Class III β-tubulin (TUBB3):more than a biomarker in solid tumors? Curr. Mol. Med. 11 (2011) 726–731.

[51] M. El Gazzar, B.K. Yoza, X. Chen, B.A. Garcia, N.L. Young, C.E. McCall,Chromatin-specific remodeling by HMGB1 and linker histone H1 silencesproinflammatory genes during endotoxin tolerance, Mol. Cell. Biol. 29 (2009)1959–1971.

[52] S. Calogero, F. Grassi, A. Aguzzi, T. Voigtländer, P. Ferrier, S. Ferrari, M.E.Bianchi, The lack of chromosomal protein Hmg1 does not disrupt cell growthbut causes lethal hypoglycaemia in newborn mice, Nat. Genet. 22 (1999)276–280.

[53] J. Itou, N. Taniguchi, I. Oishi, H. Kawakami, M. Lotz, Y. Kawakami, HMGB factorsare required for posterior digit development through integrating signaling pathwayactivities, Dev. Dyn. 240 (2011) 1151–1162.

[54] M.E. Bianchi, M. Beltrame, Upwardlymobile proteins, EMBORep. 1 (2000) 109–114.[55] R. Hock, T. Furusawa, T. Ueda, M. Bustin, HMG chromosomal proteins in development

and disease, Trends Cell Biol. 17 (2007) 72–79.[56] R.M. Gronostajski, Roles of theNFI/CTF gene family in transcription anddevelopment,

Gene 249 (2000) 31–45.[57] V.Y. Leung, B. Gao, K.K. Leung, I.G. Melhado, S.L. Wynn, T.Y. Au, N.W. Dung, J.Y. Lau,

A.C. Mak, D. Chan, K.S. Cheah, SOX9 governs differentiation stage-specific geneexpression in growth plate chondrocytes via direct concomitant transactivationand repression, PLoS Genet. 7 (2011) e1002356.

[58] T. Ueda, M. Yoshida, HMGB proteins and transcriptional regulation, Biochim. Biophys.Acta 1799 (2010) 114–118.

[59] D.S. Pisetsky, H. Erlandsson-Harris, U. Andersson, High-mobility group box protein 1(HMGB1): an alarmin mediating the pathogenesis of rheumatic disease, ArthritisRes. Ther. 10 (2008) 209.

[60] Y. Shi, S. Sandoghchian Shotorbani, Z. Su, Y. Liu, J. Tong, D. Zheng, J. Chen, Y. Xu, Z. Jiao,S. Wang, L. Lu, X. Huang, H. Xu, Enhanced HMGB1 expression may contribute to Th17cells activation in rheumatoid arthritis, Clin. Dev. Immunol. 2012 (2012) 295081.

hmgb1 can facilitate activation of the matrilin-1 gene promoter by sox9 and l-sox5/sox6 in early...

Documents

trps1 plays a pivotal role downstream of gdf5 signaling in...

the expression and function of micrornas in chondrogenesis...

interaction of tgfβ and bmp signaling pathways during...

okadaic acid-induced inhibition of protein phosphatase 2a...

the use of biodegradable plga nanoparticles to mediate sox9...

effects of cyclic compressive loading on chondrogenesis of...

2013 cai jci (erg-sox9) supplementary

regulation and role of sox9 in cartilage formation

electromagnetic fields counteract il-1 β activity during...

matrilin-2, a large, oligomeric matrix protein, is expressed...

effects of cd44 antibody– or rgds peptide–immobilized...

sox9 and nfia coordinate a transcriptional regulatory...

mutations in sox9, the gene responsible for campomelic...

sox9 is highly expressed in nonampullary duodenal adenoma...

the mir-124-sox9 paramutation: rna-mediated epigenetic...

sox9 is expressed in normal prostate basal cells and...

the transcriptional activity of sox9 in chondrocytes is...

regulation of the human sox9 promoter by sp1 and creb

hypoxic conditions increase hypoxia-inducible transcription...

regulatory mechanisms of chondrogenesis and implications for...