Background. The mineralized skeleton is a major evolutionary novelty that has contributed to the impressive morphological diversifications of the vertebrates. Essential to bone biology is the solidified extracellular matrix secreted by highly specialized cells, the osteoblasts. We now have a rather complete view of the events underlying osteogenesis, from a cellular, molecular, genetic, and epigenetic perspective. Because this knowledge is still largely restricted to mammals, it is difficult, if not impossible, to deduce the evolutionary history of the regulatory network involved in osteoblasts specification and differentiation. In this study, we focused on the transcriptional regulators Runx2 and VDR (the Vitamin D Receptor) that, in mammals, directly interact together and stabilize complexes of co-activators and chromatin remodellers, thereby allowing the transcriptional activation of target genes involved in extracellular matrix mineralization. Using a combination of functional, biochemical, and histological approaches, we have asked if the interaction observed between Runx2 and VDR represents a recent mammalian innovation, or if it results from more ancient changes that have occurred deep in the vertebrate lineage. Results. Using immunohistochemistry and in situ hybridization in developing embryos of chick, frog and teleost fishes, we have revealed that the co-expression of Runx2 and VDR in skeletal elements has been particularly strengthened in the lineage leading to amniotes. We show that the teleost Runx2 orthologue as well as the three mammalian Runx1, Runx2 and Runx3 paralogues are able to co-immunoprecipitate with the VDR protein present in nuclear extracts of rat osteoblasts stimulated with 1,25-dihydroxyvitamin D₃. In addition, the teleost Runx2 can activate the transcription of the mammalian osteocalcin promoter in transfection experiments, and this response can be further enhanced by 1,25-dihydroxyvitamin D₃. Finally, using pull-down experiments between recombinant proteins, we show that the VDR homologue from teleosts, but not from ascidians, is able to directly interact with the mammalian Runx2 homologue. Conclusions. We propose an evolutionary scenario for the assembly of the molecular machinery involving Runx2 and VDR in vertebrates. In the last common ancestor of actinopterygians and sacropterygians, the three Runx paralogues possessed the potential to physically and functionally interact with the VDR protein. Therefore, 1,25-dihydroxyvitamin D₃might have been able to modulate the transcriptional activity of Runx1, Runx2 or Runx3 in the tissues expressing VDR. After the split from amphibians, in the lineage leading to amniotes, Runx2 and VDR became robustly co-expressed in developing skeletal elements, and their regulatory interaction was incorporated in the genetic program involved in the specification and differentiation of osteoblasts.