Introduction

Cleidocranial dysplasia (CCD) is an autosomal-dominant inherited skeletal disease with high penetrance and variable expressivity, characterized by hypoplastic/aplastic clavicles, large fontanelles, dental anomalies, and delayed skeletal development (OMIM #119600). It has been established that CBFA1/RUNX2 is the gene responsible for an equivalent phenotype in heterozygous mice and molecular studies have identified several mutations of this gene in patients with a clinical diagnosis of CCD (Mundlos 1999). RUNX2 encodes for a transcription factor that belongs to the CBFα family, which is characterized by the 128 amino acids long evolutionary conserved, runt domain, and is homologous to the Drosophila pair-rule gene runt. RUNX2 controls osteoblast differentiation, specifically through heterodimerization with CBFβ and by binding to specific DNA sequences (Ducy et al 1997). CBFβ stabilizes the RUNX2–DNA interaction (Tahirov et al 2001). Several mutations impairing RUNX2 activity have been identified and characterized in patients with CCD, including both familial and sporadic cases (Quack et al 1999; Otto et al 2002). Most of these mutations are missense mutations in the runt domain that impair both specific DNA binding and promoter transactivation (Lee et al 1997; Yoshida et al 2003).

The aim of this study was the analysis of a novel RUNX2 missense mutation found in a family with CCD.

Materials and methods

Patients and mutation screening

The family analyzed was initially referred for genetic counseling to the Istituto di Genetica of Policlinico Universitario in Udine (Italy). Informed consent was given for us to obtain blood samples for genetic analysis.

Total DNA was purified from peripheral blood leukocytes of patients, and the mutation screening was performed on the eight exons (from zero to seven) of the RUNX2 gene, as previously described (Tessa et al 2003).

Plasmids, mutagenesis, and expression of recombinant proteins

To express the recombinant runt domain we used the pGST-OSF2 plasmid encoding the RUNX2 cDNA cloned in frame with GST in the pGEX4T3 plasmid (Pharmacia, Piscataway, NJ, USA). For transactivation studies, the following plasmids were used: pCMV-OSF2 encoding the full length mouse RUNX2 cDNA under the control of CMV promoter in the pCMV5 plasmid; p6OSE2-luc encoding six OSE2 target sequences upstream of the luciferase (luc) reporter gene (Ducy et al. 1999). The mutation G192R was introduced into both eukaryotic and prokaryotic vectors using a Quick-change kit (Stratagene, La Jolla, CA, USA), following the manufacturer’s instructions. Recombinant proteins were obtained in a BL21 Escherichia coli strain by IPTG induction as previously described (Damante et al 1996).

Gel retardation assay and cell transfection

The oligonucleotide sequences used as probes were as follows. The wild-type, 5′-CGCAGCTCCCAACCACA TATCCTCT-3′ (top strand) is derived from the human osteocalcin promoter (−141 to −165) and contains an OSE2 motif (AACCACA). The mutant sequence is 5′-CGCAGCTCCCAGACACA TATCCTCT-3′ (top strand) (Ducy et al. 1997). Double-stranded sequences labeled at the 5′ end with 32P were used as probes in gel retardation assays, which were performed as previously described (Damante et al 1996). HeLa cells were cultured in DMEM medium with 10% calf serum. The calcium phosphate coprecipitation method was used for transfections, as described elsewhere (Puppin et al 2003).

Computer simulations of RUNX2 wild-type and G192R mutant structures

The sequence for Homo sapiens RUNX2 was obtained from GenBank (Acc. NP_004339.3). The fragment selected for modeling starts at amino acid 97 and ends at amino acid 210. The mutation addressed in the present study occurs at position 178 in this sequence, but we will use the number 192 for consistency with previous studies (Lee et al 1997). This fragment, covering most of the runt domain, was created via homology modeling on the runt domain of the RUNX1 domain, solved by X-ray crystallography at 1.7 Å resolution (chain A, pdb id. 1ean) (Bäckström et al 2002). Model building for the native form and the G192R mutant was performed using the program Nest (Petrey et al 2003). Model building was straightforward because of the high degree of identity (104/114 residues) and overall similarity (110/114 residues) between the RUNX1 and RUNX2 runt domain sequence. The system was prepared and simulated as previously described for β-lactoglobulin (Fogolari et al 2005). In summary, the system was energetically minimized and ions were added in the most stable regions according to the solution of the Poisson–Boltzmann equation, in order to make the system electrostatically neutral. The system (protein and ions) was placed in a box of water of size 82×66×63 Å3. Water was energetically minimized and 100 ps of molecular dynamics (keeping the protein and ions fixed) were simulated in order to let the water soak the protein. Then the whole system was energetically minimized and equilibrated to 300 K for 100 ps. Finally, 3.6 ns of molecular dynamics were simulated and used for later analysis.

Results and discussion

The family analyzed was composed of a 55-year-old mother and two children: a 27-year-old male and a 33-year-old female. All patients had a clinical diagnosis of CCD and displayed a classic phenotype with aplasia of the clavicles, large fontanelles, short stature, skeletal anomalies, and supernumerary teeth. Aplasia of the clavicles was monolateral in the son. The mother suffered aseptic nontraumatic necrosis of the femoral head at the age of 7. At the age of 55 she was still menstruating. In this subject, at the age of 48, BMD data indicated the occurrence of osteopenia (L2–L4 T and Z scores −2.07 and −1.41, respectively). The mutation, initially found in the mother, was also confirmed in both children. The mutation affects position 574G>A in RUNX2 exon 2, causing the glycine at position 192 to change into arginine (G192R) in loop 9 of the runt domain, and it has been never described before. Gel retardation experiments, performed with wild-type and mutant recombinant proteins, show that the G192R mutant maintains a DNA binding activity that is about 50% of that of the wild-type protein (Fig. 1). The mutant protein retains DNA binding specificity, as demonstrated by the inability to bind the OSE2 mutant oligonucleotide.

Fig. 1
figure 1

DNA binding of wild-type and G192R mutant to a target sequence of RUNX2 (OSE2) and to a mutant form of it (OSE2 mutant). B and F indicate protein-bound and free DNA, respectively

Full size image

An analysis of the transactivation properties of the G192R mutant was performed via cotransfection assays in HeLa cells, using the RUNX2-responsive osteocalcin promoter linked to the luc gene (p6OSE2-luc) as reporter. As shown in Fig. 2a, the G192R mutation abolishes the specific transactivation properties of RUNX2 on the specific target promoter. In Fig. 2b, we can see that the G192R mutant, when overexpressed, exerts a dominant-negative effect. In fact, an excess of the expression vector for the mutant protein (9 μg versus 1 μg), reduces the transactivating effect exerted by the expression vector of the wild-type protein. This effect is not due to a squelching phenomenon (Gill and Ptashne 1988), as demonstrated by using 10 μg of the wild-type expression vector. A simple explanation for these results is that the mutant protein is able to interact with CBFβ, reducing the possibility of interactions between CBFβ and the wild-type protein. The interaction between the mutant protein and CBFβ is not productive—the latter is not able to induce any structural modification that contributes to the stabilization of the protein–DNA complex. Accordingly, structural modification of loop 9 of the runt domain occurs upon interaction with CBFβ (Tahirov et al 2001; Bäckström et al 2002). From these data, it cannot be concluded whether the patient’s phenotype is influenced by the dominant-negative effect of the mutation. However, it is important to note that the older family patient (the mother), though menstruating, suffers from osteopenia (which has only been reported in a few CCD patients). This finding fits with the observation that transgenic mice expressing a dominant-negative allele of RUNX2 after birth develop an osteopenic phenotype (Ducy et al 1999).

Fig. 2a,b
figure 2

Transcriptional effects of the wild-type RUNX2 and the G192R mutant on osteocalcin promoter. a The G192R mutant is not able to activate the osteocalcin promoter. b The G192R mutant exerts a dominant-negative effect on the wild-type protein. In both a and b, each bar indicates the mean value ± SD of three independent transfections

Full size image

To support the hypothesis delineated above, we have performed molecular dynamics simulations on the wild-type and G192R mutant uncomplexed protein. As expected, the backbone conformation at position 192 is lost when native glycine is replaced by an arginine. This conformational rearrangement, which takes place during the 100-ps equilibration simulation, is the likely molecular basis for all the observed effects. For further analysis, snapshots were taken at 100 ps intervals along the dynamics. In order to simplify the analysis, snapshots were superimposed pairwise and the structure possessing the smallest average root mean square deviation (RMSD), computed on Cα atoms, among the snapshots, was taken as the most representative conformation of the ensemble. The structures chosen were those at 2.0 ns (average RMSD=1.13±0.19 Å) and 3.3 ns (average RMSD=1.20±0.41 Å) for native and mutant simulations, respectively.

The RMSD between the Cα atoms of the two most representative conformations is reported in Fig. 3a. A very large deviation (over 5 Å) is observed at the region encompassing the mutated residue. The conformational change occurring in loop 9 is able to propagate by a network of hydrogen bonds to neighboring loop 5. It is expected that any change in one of these two loops will have an effect on the other one. Here, mutation of Gly 192 to Arg induces a conformational rearrangement at loop 9 which in turn has direct hydrogen bonds with loop 5. The two loops move in a coordinated way. Although these loops are known to undergo conformational changes upon binding to CBFβ and DNA, we have superimposed the two most representative structures of the wild-type and mutant protein on the corresponding protein in complex with CBFβ and DNA (pdb id. 1h9d; Bravo et al 2001).

Fig. 3a,b
figure 3

a Root mean square deviation (in Å) for Cα atoms after global superposition of the most representative structures for the native form and the G192R mutant of the runt domain. b Loops 5 and 9 of the native form (blue) and the G192R mutant (red) of the runt domain. Residues 192 and 160, important for loop 5 and 9 interactions, are displayed in bond representation. The two most representative structures of the native and mutant ensembles are superimposed on the crystallographic complex of CBFβ factor (yellow) along with the runt domain (orange) and DNA (green)

Full size image

The side chain of Arg 192, in the mutant protein, still points toward DNA and steric hindrance with DNA could possibly be accommodated by the overall flexibility of this protein loop (Fig. 3b). However, the conformational changes needed at loop 9 in order to bind DNA are expected to have an impact on further CBFβ binding, through loop 5. This would explain why the mutant still retains some DNA binding capability but is not able to activate osteocalcin promoter.