In this work we have isolated from bone marrow aspirates human mesenchymal stem cells characterized by their surface marker profile [1, 17] and their osteogenic, chondrogenic and adipogenic potential [5–7, 22, 23] to study the expression profile of genes involved during maturation and adipogenic differentiation. These cells showed all the important characteristics of human MSC and were further used for gene expression profiling analysis during adipogenic development. In studies of human mesenchymal tissue development, the analysis of expression of specific genes, carried out over a distinct time period, is relevant to understanding the maturation process. Since the applied cocktail of adipogenic inducers, dexamethasone, 3-isobutyl-1-methylxanthine, indomethacin and insulin, was sufficient to promote differentiation of human MSC to adipocytes, these MSC were further used for gene expression profiling during adipogenesis. Our data helped identifying gene sets characterizing the sequential differentiation steps from stem cells to fully differentiated adipocytes. In our study, we observed differentially expressed genes coding for molecules that have not yet been described in the context of adipogenesis. For example the expression of human receptor gene SCARA5 and the endoplasmic reticulum protein gene MRAP were highly up-regulated, suggesting new candidate genes effective in adipocytic differentiation. Also we have confirmed most of the already described marker genes involved in adipogenesis .
The PPARG signalling pathway has been intensively studied over several years and scenario through which various adipocyte transcription factors interact with adipogenic target genes have already been proposed . PPARG is a hormone receptor, which has already been demonstrated to be the master regulator of adipogenesis [25, 26]. The dependent downstream-regulation cascade can be controlled by PPARG itself or by PPARG in combination with the controlling factor C/EBPA (also up-regulated in our study).
It is very likely that C/EBPA and PPARG direct the final phase of adipogenesis by activating expression of adipocyte-specific genes, such as fatty acid synthetase, fatty acid binding protein, leptin and adiponectin. However, the mainstream interaction between PPARG as well as the RXRA receptor, another gene product was found to be up-regulated during adipogenesis in our study via ligand-binding, results in a conformational change, namely in the formation of a special heterodimeric protein structure. The remaining functional domain within the complex binds to various peroxisome proliferator response elements (PPREs) and therefore activates transcription of adipogenic target genes via phosphorylation events, or, alternatively, causes induction of co-activator and co-repressor complexes [27, 28]. It has been claimed that about 200 proteins can be controlled by PPARG, probably via the PPARG-RXRA signalling pathway . Therefore, it was not surprising that various biochemical subgroups belonging to carbohydrate metabolism, citrate cycle and energy transfer or ion transport could be reformed on the basis of the identified adipogenic candidate genes. Interestingly, we found the PPARG target CES1, an enzyme located in the endoplasmic reticulum and responsible for cellular detoxification  to be highly up-regulated. Another PPARG target example is presented by the addition of a second GPD1 enzyme type. The PPARG-RXRA signalling expression pattern leads to several additional downstream adipogenic target genes and corresponding pathways, which are either up- or down-regulated . Of course, we also found other up-regulated PPARG targets, such as the transcripts coding for fatty acid-binding FABP4  and PLIN, which binds to fatty acids in such a way that lipid droplets are formed and fatty acids are protected against degrading enzymes . Furthermore, the mobilization of the stored triglyceride is thought to be controlled by interactions among intracellular lipases during hormone-mediated lipolysis and other proteins that coat lipid storage droplets . In our study, the mean value of PLIN was greatly increased after 17 days as was that of the hormone ADIPOQ, which is also a known downstream target of the PPARG-RXRA-regulator. Recently, PPARG has received attention as a possible pharmacological target for the thiazolidinedione class of anti-diabetic drugs as it is essential for the final phase of adipocyte differentiation and the pro-adipogenic effects of thiazolidinediones have spurred interest in identifying therapeutic compounds that retain anti-diabetic activity without promoting adipogenesis. Inhibitors of PPARG activity that reduces adipogenesis and thus serve as the basis for the development of effective anti-obesity drugs  have already been identified.
Further highly expressed genes (all named by the gene name) were grouped together according to their known biological function and pathway. The lipogenesis group includes fatty acid synthase (FAS), which synthesizes fatty acids (responsible for lipid accumulation). This group also includes the enzymes ACACB, SRD5A2L, HADHSC, and DGAT2. ACACB is a biotin-containing enzyme, which catalyzes the carboxylation of acetyl-CoA to malonyl-CoA, the rate-limiting step in the fatty acid synthesis. But ACACB is also a complex multifunctional enzyme system, which controls fatty acid oxidation through the ability of malonyl-CoA to inhibit carnitine-palmitoyl-CoA transferase I.
As expected we found the adipokine and adiponectin metabolism (ADIPOQ) to be up-regulated. ADIPOQ encodes a central cytokine, which is released by adipocytes. It is involved in the control of fat metabolism and insulin sensitivity, with direct anti-diabetic, anti-atherogenic and anti-inflammatory activities. In the human body, adiponectin stimulates the phosphorylation of the AMP activated protein kinase (AMPK) in liver and the skeletal muscle, antagonizes TNF-α by negatively regulating its expression in various tissues such as liver and macrophages, and also by counteracting its effects in adipocytes by enhancing glucose utilization and fatty-acid production . The inhibitory potential of TNF-α is mediated using the endothelial NF-kappa-B signalling cAMP-dependent pathway [36, 37]. ADIPOQ also plays a role in cell growth, angiogenesis and tissue remodelling by binding and sequestering various growth factors. It is very likely that adiponectin interacts with one of the collagens or is part of the extracellular matrix system . At least one of the known collagen receptor genes, COL5A3, was slightly up-regulated in our study. We also detected the up-regulation of typical cell-cell signalling factor genes with adhesive/cytokine like function, such as CLEC1A and LGALS3. It has been shown that LGALS3 can protect mitochondrial membrane integrity and models suggested that it acts as an anti-apoptotic factor, because it prevents cytochrome C release, thereby blocking the effector stage of apoptosis . It was discussed in literature that APOL6 can contribute to the formation of ion channels through intracellular membranes and is therefore involved in mechanisms triggering programmed cell death . In any case, the regulator hormone ADIPOQ consists of a collagen domain by itself and can pass through different cellular membranes without any alteration of the domain. ACACB is also part of the adipocytokine signalling pathway and seems to play a central enzyme role in rats as well as in man and alterations of the expression pattern can possibly be responsible for obesity, diabetes , and other metabolic pathway activities.
In addition, cytokines such as tumor necrosis factor alpha (TNF-α) and other transforming growth factors interfere with adipocyte differentiation. As expected, not all typical examples for adipocytokines (e.g. leptins, TNF-α, Interleukin-6 (IL6) or resistin) are expressed in all types of tissues identically  - and some adipocytokines were not found in this study. Nevertheless, we detected members that are involved in the adipocytokine-signalling pathway such as FLRT3, of which the protein product might also function as a receptor signalling protein as well as a cell adhesion protein .
During the adipogenic maturation process, media containing insulin may have resulted in the development of insulin-independent mechanisms . This phenomenon probably was the reason why the insulin-independent cellular membrane receptor, which is called glucose transporter-member 1 (GLUT1), was missing in our candidate gene list. Instead we found up-regulation of the lipoprotein gene stomatin (STOM). Over-expression of this lipoprotein often results in depressing the basal rate of glucose transport by decreasing the intrinsic activity of GLUT1 . However, we also detected insulin-dependent genes, such as INSIG1, which have been suggested as a target of PPARG .
During adipogenic differentiation, we found that the expression of IGFBP7 was down-regulated. The insulin-like growth factor binding family is thought to modify most IGF1 actions  and is very important for human growth, while insulin by itself can bind to the IGF1 receptor to activate the phosphatidyl-inositol-3-phosphat kinase (PI3K)/Akt (proteinkinase B)-signal mechanism of the cellular structure . However, a few members of IGFBP family, such as IGFBP7, do have other biochemical functions. For example IGFBP7 also binds directly to insulin and in an artificial cell system it was recently demonstrated that glioma cell growth can be mediated by expressing IGFBP7 . This observation fits very well with our second finding, and therefore we assume again that IGFBP7 has a negative effect on adipocyte differentiation. IGFBP7 can modulate the stimulatory effect of vascular endothelial growth factors (VEGF) on angiogenesis by interfering with VEGF expression . On the other hand, IGFBP5 is thought to be associated with proteins of the extracellular matrix (ECM) like fibronectin depending on bound molecules, which might interfere with IGF1 and further growth activities [46, 49].
Strikingly, we detected increased expression of members of the angiotensin II modulator family, such as the angiotensin II type 1 receptor (AGTR1A) gene, coding for a receptor whose action is mediated by association with G proteins, which activate the phosphatidylinositol-calcium second messenger system. The human rennin-angiotensin system can be activated by AGTR1A, while angiotensin II is the main regulator or effector of the cellular proliferation and survival . Recently, many publications demonstrated that tissue specificity is achieved by modulation of the angiotensin receptor proteins together with effector proteins , which would fit well with our data and implies that AGTR1A is involved in the adipocyte development. We also found up-regulation of a further G protein-coupled receptor gene, the cannabinoid receptor 1 (brain), called CNR1 . The CNR1 receptor is the most abundant G protein-coupled receptor expressed in the brain. However, CNR1 has also been detected in adipocytes, which is consistent with our data, suggesting that CNR1 is directly or indirectly effective on adipogenesis.
The regulatory factor lipopolysaccharide-induced TNF-α factor (LITAF), which was down-regulated during adipogenesis until day 17 and expressed at much lower levels when compared with native fat cell tissue, might play an important modifying role in TNF-α gene expression either through induction of lipopolysaccharides (LPS) or via using the apoptotic TP53/p53 pathway. Moreover, some fatty acids act as signalling molecules regulating the differentiation into adipocytes or cell death.
We also found up- or down-regulation of a number of transcripts coding for factors implicated in apoptosis such as LITAF, PRKC, PAWR, BID, MORF4L2, and HNRPC. It has been shown very recently that LITAF is a transcription factor, which plays an important role in regulating the expression of TNF-α and various other inflammatory cytokines in response to LPS stimulation, negatively affecting adipocytic regulation , but this observation still has to be confirmed. With our comparison approach we demonstrated that several other apoptotic factors or mediators have been increased or decreased during adipogenesis of human MSC. The PAWR (PRKC, apoptosis, WT1, regulator) expression was down-regulated as well. It has regulatory control functions via down-regulating the B-cell lymphoma 2 (BCL-2) gene expression. The BH3 interacting domain death agonist (BID) was also down-regulated and is responsible for the release of mitochondrial cytochrome c. Two further down-regulated transcription factor candidates, which do both belong to a part of the apoptosis pathway, are the transcription factor genes mortality factor 4 like 2 (MORF4L2) and the heterogeneous nuclear ribonucleoprotein C (C1/C2) (HNRPC).
We detected the up-regulation of genes, whose encoding factors yet are unrelated to adipogenesis, such as SCARA5 and MRAP. It is already known that SCARA5 is a ferritin receptor mediating non-transferrin iron delivery, which is essential for cell growth , and that MRAP might control cellular trafficking with the help of other transmembrane proteins . Our data suggest their positive effect on adipogenesis, which should be further investigated in gain and loss of function studies. We also observed up-regulation of central genes coding for enzymes of the glycolytic pathway such as hexokinase 2 (HK2), which is integrated in mitochondrial membranes instead of working in the cytosol , as well as many stress proteins such as the peroxisome control enzyme catalyze (CAT).
As a follow-up strategy it will be important to identify regulatory networks related to specific transcription factors. Search tools like rVISTA  suggest further regulatory proteins or transcription factors based only on potential binding sites upstream of the genomic sequence of the named genes and calculate the probability that individual binding sites are identified with a significantly increased frequency. For example, the list of our up-regulated candidates (Additional file 1) returned GATA3, HFH3, AP4, FOXO4, POU6F1, FOXO1, GRE, ISRE, LBP1, FOXP3 as potential candidates involved in the regulation of these genes. However, this approach can only generate hypotheses, which require further confirmation by other techniques like chromatin immunoprecipitation followed by array hybridisation. It also will be interesting to check, if some of the detected adipogenic genes of the defined clusters or from one biochemical group (see Table 1) perform a familiar pattern of transcriptional regulation or to find de novo binding strategies on the promotor site of the genes in further experiments.