In this study we could show for the first time that MSC maintain their undifferentiated phenotype after ex vivo long-term expansion. The results of the microarray assays showed that MSCs did not change significantly their gene expression profiles during long-term expansion since only nine out of almost 30.000 genes were differentially expressed between passage 2 and passage 5 in ten donors. In addition, no changes in the expression profiles after additional 5 passages could be observed in five of the ten donors. These surprising results can only be partially explained by the limitations of the microarray technology. Due to the lower sensitivity and uncomplete coverage of the transcriptome, some genes required for stem cell maintenance might have been missed. Additionally, individual differences between the donors and slightly different conditions during tissue extraction and sample treatment could increase the number identified genes which were otherwise filtered out during averaging. However, the microarray results were supported by the analyses of the effects of extensive long-term ex vivo cultivation of MSC on their proliferation capaCity, their morphology, surface marker profile, and differentiation ability. During the long cultivation time we could observe, in concordance with other groups [37, 38], a progressive lost of proliferation ability and a slight shift in the cell-morphology of MSC of higher passage. However, their surface marker profile determined by flow cytometry (CD44+, CD90+, CD73+, CD105+, CD166+, CD45-, CD117-, HLA DR) and by Real-time RT-PCR was not altered. Over and above, the stem cell markers like CD90 and STRO1 showed stable expression levels, which further supports our observation that MSC maintained their undifferentiated phenotype and remained capable of osteogenic differentiation during ex vivo expansion at all tested passages (p2-p10).
Additionally to this clinically relevant evidence, we provide novel biological insights derived from the global view of the molecular processes during osteogenic development. First, we were able to illustrate the different phases of osteogenic development in a human model. Temporal expression of cell growth and osteoblast phenotype-related genes in a rodent model were first described by Stein et al. . In our study, using large-scale expression profiling of clinical samples we could associate gene clusters to the three distinct phases of the development of osteoblast phenotype: proliferation, matrix maturation and mineralization.
Second, we could identify markers of late osteogenesis: ID4, CRYAB and SORT1:
a) ID4 is one of the four members of the ID (inhibitor of DNA binding) transcription factor family, which belongs to the basic-helix-loop-helix (bHLH) family. The ID members play a key role in differentiation processes and are involved in cell cycle control and cellular senescence [40–43]. It has been shown that ID proteins, by heterodimerization with bHLH proteins (MyoD, E12, E47), inhibit their binding to DNA, thus acting as dominant negative regulators, and have to be down-regulated to continue the differentiation process [44, 45]. To the best of our knowledge only ID1, ID2 and ID3 have been identified as most significantly up-regulated early targets of the osteogenic BMP2, BMP6 and BMP8 . In our study we provide evidence that also ID4 exhibits a specific role during the osteogenic differentiation.
b) The second identified gene is crystallin-αB (CRYAB). Alpha crystallins are small heat shock proteins and composed of two gene products: alpha-A and alpha-B, for acidic and basic. Alpha-A is preferentially restricted to the vertebrate eye lens where it maintains the transparency and refractive index of the lens, whereas alpha-B is widely expressed in many tissues and organs . The here observed up-regulation of CRYAB suggests an involvement in osteogenic differentiation but its specific role is still unknown. This involvement is supported by Furushima et al. , who performed a linkage study with knowledge-based candidate genes to detect genetic determinants associated with OPLL (ossification of the posterior longitudinal ligament of the spine), a predominant myelopathy among Japanese. The candidate genes for examination were obtained by cDNA microarray analysis of gene expression profiles during osteoblastic differentiation of MSC. Among the 24 genes identified in cDNA microarray analysis which could be associated with bone metabolism, CRYAB was the only gene which showed significant evidence of linkage.
c) Further, we could show the role of sortilin1 during osteogenic differentiation. Sortilin 1 (SORT1) represents a multi-ligand type-1 receptor which binds a number of unrelated ligands that participate in a wide range of cellular processes. The expression of SORT1 was increasing until day seven and then slowly decreasing. Sortilin1, also known as neurotensin receptor-3, is a glycoprotein originally purified from human brain . The expression of sortilin1 is ubiquitous; it has been observed in heart, brain, placenta, lung, skeletal, muscle, pancreas, proState, testis, small intestine, thyroid, and spinal cord, whereas its expression in bone marrow could not be detected . The function of srotilin1 in bone metabolism is poorly understood. Maeda et al.  reported first differential expression during osteogenic differentiation and performed further experiments, which showed that sortilin 1 is promoting extracellular matrix maturation. Our results presented here strengthen this hypothesis, since sortilin 1 shows a similar expression progression as other well known osteogenic genes, such as ALPL or BMPs.
In summary, we have identified by expression profiling and confirmed by quantitative RT-PCR three markers of late osteogenesis. Further studies are necessary to quantify the expression of ID4, SORT-1 and CRYAB in primary human osteoblasts and other cell types and to identify the role of these genes in osteogenesis.
Third, by mapping the obtained expression data on curated biomolecular pathways, we were able to illustrate the distinct activation of the different Smad pathways induced by BMPs, TGF-β and inhibin, which were summarized as TGF-β pathways. The mapping of the expression data demonstrated that the fate of cell differentiation was simultaneously controlled at three different points: 1) BMP2, BMP6 and BMP8 activated the Smad-signaling pathway, consisting of the main components Smad 1, Smad 5 and Smad 8 resulting in osteoblast differentiation; 2) TGF-β2, regulating Smad 2 and Smad 4, was down regulated and the subsequent mechanisms were suppressed; and 3) inhibin was, like TGF-β2, down regulated. These findings have been confirmed by recent reports describing the diverse Smad-signaling pathways, but have never been illustrated in a general overview before this study [52–55].
Finally, the comparison of in vitro differentiated osteoblasts with in vivo developed osteoblasts on morphological and genetic level proved the usefulness of MSC as in vitro model for the investigation of the osteogenic development.