BACKGROUND AIMS Fractures with a critical size bone defect (e.g., open fracture with segmental bone loss) are associated with high rates of delayed union and non-union. The prevention and treatment of these complications remain a serious issue in trauma and orthopaedic surgery. Autologous cancellous bone grafting is a well-established and widely used technique. However, it has drawbacks related to availability, increased morbidity and insufficient efficacy. Mesenchymal stromal cells can potentially be used to improve fracture healing. In particular, human fat tissue has been identified as a good source of multilineage adipose-derived stem cells, which can be differentiated into osteoblasts. The main issue is that mesenchymal stromal cells are a heterogeneous population of progenitors and lineage-committed cells harboring a broad range of regenerative properties. This heterogeneity is also mirrored in the differentiation potential of these cells. In the present study, we sought to test the possibility to enrich defined subpopulations of stem/progenitor cells for direct therapeutic application without requiring an in vitro expansion. METHODS We enriched a CD146+NG2+CD45- population of pericytes from freshly isolated stromal vascular fraction from mouse fat tissue and tested their osteogenic differentiation capacity in vitro and in vivo in a mouse model for critical size bone injury. RESULTS Our results confirm the ability of enriched CD146+NG2+CD45- cells to efficiently generate osteoblasts in vitro, to colonize cancellous bone scaffolds and to successfully contribute to regeneration of large bone defects in vivo. CONCLUSIONS This study represents proof of principle for the direct use of enriched populations of cells with stem/progenitor identity for therapeutic applications.