The study was performed using two batches of human chondrocyte primary cultures derived from seven TDI and four control fetuses without skeletal pathology, aged 18 to 25 weeks. The radiological features of TDI such as narrow trunk, curved femurs and platyspondyly were observed in all pathological fetuses. The chondrocytic phenotype of the primary cells was assessed before performing the microarray experiment. Cultured cells retained chondrocytic morphology and a high expression of cartilage specific genes such as aggrecan, collagen type II compared to collagen type I and SOX9 was observed by RT-PCR (data not shown). Four heterozygous Arg248Cys and three Tyr373Cys mutations were identified in the FGFR3 extracellular domain by genomic DNA sequencing of TD primary chondrocytes. The control primary chondrocytes did not contain mutations in the FGFR3 domains (Supplementary data S1).

High levels of proliferative chondrocytic markers together with reduced or negligible expression of hypertrophic markers confirmed that the primary normal and TDI chondrocytes were not dedifferentiated and exhibited a proliferative phenotype.

The cell cycle is a highly coordinated process, which involves expression of a host of genes in the different cell cycle phases. Most genes essential for S, G2, and M phases, were found to be down-regulated in the human TDI primary chondrocyte cultures. Conversely, genes promoting G1 progression, especially cyclin D1 were up-regulated, in agreement with previous studies [32], [36]. Inhibitors of G1 progression, such as p18 were down-regulated. These data, together with the reduced frequency of chondrocytes undergoing cytokinesis observed in TDI cultures, are consistent with a decrease in TDI chondrocyte proliferation and a G1 block as observed in RCS lines upon FGF stimulation [32], [36].

Interestingly, the observed increase in cyclin D1 expression is consistent with the typical FGF mitogenic effect on most cell types [37]. Thus, the unique cell cycle inhibition response observed in TDI chondrocytes may occur during a later phase of the cell cycle, e.g. the G1/S transition as cyclin A (−2.8 fold change) can be rate limiting for S-phase entry. Alternatively, it may also result from the regulation of p107 and p130 pocket proteins, which undergo dephosphorylation upon FGF stimulation [35], [36]. PTPRG, a member of the protein tyrosine phosphatase family previously shown to induce pRB dephosphorylation [38], is up-regulated in TDI chondrocytes, suggesting potential involvement in this process.

Early cell cycle exit and premature chondrocyte differentiation may also be due to modulation of the ratio between activator and repressor E2Fs. In support of this hypothesis, 38% of modulated E2F targets were found to be down regulated in the current dataset, and mainly associated with the cell cycle process, whereas up-regulated E2F targets are involved in developmental and differentiation processes. Down-regulation of E2F8 in TDI chondrocytes (−2.2 fold change) may contribute to such a mechanism [39].

Several basement membrane and ECM structural components are down-regulated in TDI chondrocytes, suggesting an altered matrix organization and turnover. A high number of modulated genes are involved in glycosaminoglycan (GAG) and proteoglycan biosynthesis, diversification and sulfation. Among these, hyaluronan synthase 3 (HAS3) was up-regulated. Over-expression of this gene product may result in shortened high molecular weight hydrophilic hyaluronan molecules, possibly reducing ECM viscosity [40] and cell movement throughout the cartilage. Likewise, changes in relative quantities of xylosyltransferases, glucuronyltransferases, galactosyltransferases and glycosyltransferases may also alter biochemical properties and to some extent, the function of these proteoglycans in cell adhesion, receptor activation and long-range diffusion of signaling proteins. The present data also provides an indication of a putative defect in GAG sulfation, which may lead to dramatic biological effects, as exemplified by mutations in PAPSS2, SLC26A2, gPAPP or SUMF1, all causing skeletal dysplasias [41]–[44].

Cartilage integrity and maturation depends on well-tuned interactions with the ECM, which acts jointly with the cytoskeleton to determine chondrocyte shape and control motility [14]. In the TDI primary chondrocyte cultures, several differently modulated genes are involved in adhesion and cytoskeleton dynamics and regulate or act along the Rac1 and CDC42 RhoGTPase signaling pathways. This is in agreement with previous studies showing dwarfism, abnormal cell shape and disrupted columnar organization in mice lacking β1 integrin [9], ILK [13] and Rac1 [45].

Although functions of some of these modulated genes are somehow controversial, depending on cellular context (e.g. ARHGDIB can both inhibit and positively regulate Rac1), these observations collectively suggest that regulation of Rac1 and Rho/ROCK pathways may be impaired, possibly resulting in reduced motility. As the columnar arrangement of the proliferative zone requires chondrocytes to undergo a gliding movement with one cell moving on top of the other [46], these findings may provide clues about growth plate disorganization. The protein products of RND3, IQGAP3 and ARHGDIB, acting at the interface of ERK1/2 and RhoGTPase pathways, may be involved in the pathological process and would warrant further examination.

The transcription factor binding site (TFBS) enrichment analysis suggests that TDI mutations affect almost all signaling pathways involved in chondrocyte biology. This is further supported by the modulation of 24 transcription factors that belong to, or act downstream of, the MAPK, Wnt, Hedgehog, BMP/TGF-β, NFκB, PI3K/AKT and calcium signaling pathways.

Likewise, the expression of genes coding for several key components of many signaling pathways is modified in TDI primary chondrocytes, including DUSP6, MAP3K5, ZAK and MAP2K6 in ERK1/2 and p38MAPK signaling [47]–[49], ID2 and ID3 in BMP and TGFβ signaling, PIK3R1, PLCG2 and PLCB4 in inositol phosphate signaling, FOXM1 and MAPK8 [50] in JNK signaling, DKK1 [51] in WNT signaling, (HIP) [52], RAB23 [53], IFT74 in hedgehog signaling or NKX3.2 in PTHrP [18] and NFkB [54] signaling. In addition, some modulated genes code for scaffold proteins which are essential for the appropriate subcellular location of signaling molecules. For instance, down-regulation of CNKSR2 in TDI chondrocytes may impair the crosstalk between MAPK and Ral pathways which is essential for proper receptor endocytosis, cytoskeleton remodeling and DNA synthesis [55]. Likewise, down-regulation of IQGAP3 and LSP1 as well as over-expression of NME1 may modify ERK1/2 activation upon PKC or Ras activation [56]–[58].

The potentially reduced JNK signaling may lead to G1 cell cycle arrest in TDI chondrocytes, as FOXM1 (−3.1 fold change) is known to promote proliferation and control the G1/S transitions by acting through MAPK8 (−2 fold change) [50]. Simultaneously, increased BMP signaling and up-regulation of FOSL2, EIF4A and RPS23 combined with ATF5 and NKX3.2 down-regulation, may induce differentiation [18], [59]–[62]. Altogether, these findings are suggestive of reduced proliferation, premature cell cycle exit and induction of differentiation.

In conclusion, signal transduction pathways represent networks that process and integrate information from the cell environment to regulate the spectrum of downstream targets in a context specific manner. This study suggests that FGFR3 activating mutations in human chondrocytes may disturb this finely tuned process at many levels. Firstly, FGFR3 activating mutations affect input signals (modified expression of other growth factors and membrane bound receptors, biochemical properties of the ECM). Secondly, FGFR3 activating mutations alter the expression of several components of signaling pathways, possibly modifying the responsiveness to a number of other external stimuli. Finally, FGFR3 activating mutations alter cytoskeleton and scaffold dynamics, resulting in improper compartmentalization and insulation. Thus, TDI pathogeny may result from the interplay of several signaling pathways and from a defective combination of signaling modules. This study provides new insight into molecular events triggered by FGFR3 mutations during endochondral ossification. Further in vivo studies using compound transgenic mice will now be required to study these hypotheses.

All cartilage specimens in this study were consented under an IRB-approved protocol of informed consent through Hôpital Necker-Enfants Malades. All consents were written and signed by the participants in accordance with the French ethical standards.

Cartilage samples were obtained from 11 medically aborted fetuses following informed consent of parents. Chondrocytes were isolated from the growth plate as described previously [63], plated in flask (25 cm²) and cultured at 37°C for 4 days in DMEM supplemented with 10% fetal calf serum (Invitrogen). Chondrocytes were collected after 24 hours depletion. Total RNA was extracted using the RNeasy isolation kit (Qiagen) and treated with DNase I. Integrity and purity of total RNA were analyzed using a Bioanalyser 2100 (Agilent).

cDNA synthesis and labeling were performed on 3 µg of total RNA using the GeneChip one-Cycle Target Labeling and Control Reagents kit, according to the manufacturer's protocol. Human Genome U133plus2.0 GeneChips (Affymetrix) were hybridized, revealed and washed according to the Affymetrix protocol. GeneChips were scanned using a 7G scanner (Affymetrix) and images (DAT files) were converted to CEL files using GCOS software (Affymetrix). Quality Control of cRNA synthesis, hybridization and data acquisition was performed according to the manufacturer's protocol completed with personal QC and data visualization.

Data normalization and differential analyses were performed with the R-Bioconductor statistical environment [64]. After low level diagnostics with the affyQCReport package, raw data were normalized, and log₂ intensity expression summary values for each probe set was calculated using the gcRMA algorithm. Probe sets corresponding to control genes or having a low intensity signal (median log₂ of intensity <4) were discarded, yielding a total of 21836 probe sets for further analyses.

Preliminary analyses highlighted a strong batch effect, maybe related to changes in culture conditions having occurred during the sample collection phase. A two way ANOVA with robust variance estimation was performed with the LIMMA package [65] to evaluate effects of mutation, batch and the interaction between them. Probe sets showing significant batch and interaction effects (p-value< = 0.001) were discarded from further analysis. Modulated probe sets were then defined as having a mutation effect (p-value< = 0.05), a range₂ > = 1.5 and a fold change > = 1.5.

Analysis of genes associated with cell functions was carried out using Ingenuity Pathways Analysis and FATIGO+ to identify biological processes and pathways which may be associated with modulated gene expression. FATIGO+, as well as whole genome rVista, were used for TFBS analysis. The Genomatix Suite was used to identify targets of E2Fs and some modulated TFs. Statistical significances were evaluated using the set of 21836 expressed probe sets as reference. All data are MIAME (minimum information about a microarray experiment)-compliant and were formatted and exported to the ArrayExpress database, according to MIAME guidelines (ArrayExpress accession number: E-MEXP-2276).

cDNA were generated from 1 µg total RNA using the Verso RT PCR kit (ABgene AB-1453). RNA samples were heated at 70°C for 5 min, placed 1 min on ice before adding a pool of random hexamers and anchored oligo-dT 3∶1 (v/v) and reaction components, according to the manufacturer's protocol. Reverse transcription was performed during one hour at 50°C, followed by two minutes at 95°C.

Quantitative PCR primers were defined using the Primer Express software. Primers were chosen to generate 100–120 bp PCR products spanning at least one intron, making it possible to detect latent genomic DNA contaminations. Primer sequences are available as supplementary data (supplementary data S6).

QPCR were performed using an ABI Prism 7300 and the Absolute QPCR SYBER Green Mix (Abgene AB-1158). Reactions were set up in 25 µl, using 50 ng of cDNA and 70 nM of each primer, according to the manufacturer's protocol. A three step cycling method was used (95°C 15 s, 60–62°C 15 s, 72°C 30 s) for 40 cycles and melt curves were analyzed to confirm the specificity of the reaction.

Standard samples (3 quantity obtained by 1/100 serial dilution) were added to measure amplification efficiencies and all experiments were carried out in triplicate on the same cDNA preparations. Quality control and relative quantity calculations were performed using the qBase software [66]. Normalization was performed with two reference genes in order to improve result accuracy and reliability. RPLP13A, TBP, ACTB and 18S primer sets were evaluated as potential control transcripts, based on previous results [67] and ACTB and RPL13 were finally chosen as endogenous control genes.

Primary and immortalized chondrocytes [68] were seeded in 8 wells Lab-Tek chamber slides NUNC field with DMEM+10% fetal calf serum (Invitrogen) in 5% CO₂ at 37°C. Three different controls and four TDI cultures (Arg248Cys or Tyr373Cys mutations) were cultured on each slide, leaving one empty negative control chamber. Each slide was then analyzed with a different antibody. This design ensured that all samples were subjected to the same conditions, thus reducing experimental bias.

For mitosis analysis, three control and three TDI primary cultures stained by DAPI were analyzed. A total of 1200 cells in 10 random fields were counted for each category.

For immunocytochemistry, cells were fixed in 4% PFA for 20 min, permeabilized in 0.1% TritonX-100/PBS and blocked in 10% sheep or donkey serum, depending on primary antibody. Immunocytochemistry was performed as previously described [69], using Alexa488 or Alexa568 (Molecular Probes) secondary antibodies. Primary antibody references and working dilutions are provided as supplementary data S8. No background was detectable in empty wells and when the primary antibody was omitted. Cells were covered with mounting solution (Vector), analyzed field by field and examined using an Olympus IX2-UCB microscope using 20x and 40x objectives.

Immunofluorescence pictures for a given antibody were selected as good representatives of the samples and were taken using the same camera settings in order to allow semi-quantitative signal comparisons.