The transgenic BMP2 ES cell lineage was generated with the linearized pBMP2^p-puro IRES2 EGFP construct by stable transfection. Like its parental wild type CGR8, the BMP2 ES cells do not express BMP2 in the undifferentiated state (Figure 1a(i) and 1a(ii)). Expression of BMP2 during progressive differentiation induced by the hanging drop protocol (see Materials and methods) starts in the three-day-old embryoid bodies (EBs), gradually increases to a maximum in the five-day-old EBs and, thereafter, gradually decreases to a minimum in ten-day-old EBs, in the same manner as that seen in the RT-PCR results (Figure 1a). During the course of differentiation induced by the hanging drop protocol, the EGFP-expressing cells in the three- and four-day-old EBs were found to be scattered (Figure 1b). As differentiation continues, the EGFP fluorescence peaks in the five-day-old EBs and the EGFP-expressing cells are localized to a particular region in every EB, as shown in Figure 1b. The RNAs isolated from these EBs were analyzed for the expression of other candidate markers (T-bra, flk1, smooth muscle α-actin, neurofilament-H (NF-H) and also α-fetoprotein (AFP)) to demonstrate that these EBs were differentiating in the normal way as per their parental wild-type EBs (Additional data file 1). Isolation and further characterization of the BMP2⁺, puromycin-resistant cells were optimized according to the protocol described in Figure 1c. Briefly, a single cell suspension of BMP2⁺ ES cells was seeded in bacteriological dishes for two days to form two-day-old EBs. These were then transferred into gelatine coated tissue culture dishes and cultured for a further two days. Thereafter, plated EBs were treated with 3 μg/ml puromycin for three days. After trypsinization of puromycin-resistant seven-day-old BMP2⁺ cells, fluorescence-activated cell sorting (FACS) analysis was performed. As demonstrated in Figure 1d,e, after 3 days of puromycin treatment, EGFP fluorescing and puromycin resistant BMP2⁺ cells (hereafter called BMP2⁺ cells) accounted for 93% of the cells in the EBs, whereas in the control EBs without puromycin treatment (hereafter called control EBs (seven-day-old EBs)) only 11% of the cells were BMP2⁺ cells. This result demonstrates a nearly 8.5-fold enrichment of BMP2⁺ cells in EBs treated with puromycin. As demonstrated in Figure 1f, plating of the BMP2⁺ cells in gelatine coated tissue culture dishes for another three days in the presence of puromycin results in a bright EGFP-positive BMP2⁺ cell population. Furthermore, the BMP2 protein was detected by immunostaining using BMP2-specific antibodies. The undifferentiated BMP2 ES cells were included as a negative control (Figure 1h). As demonstrated, BMP2 is detected only in the cytosol, and specifically in vesicles, of the BMP2⁺ cells (Figure 1g).

The Affymetrix data obtained were validated by quantitative real time PCR (qPCR; Additional data file 2). To identify Gene Ontology (GO) 10 categories, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways 11 and BioCarta pathways 12 specifically enriched in BMP2⁺ cells, we first analyzed genes that are upregulated in BMP2⁺ cells in comparison to control EBs. Moreover, to identify BMP2⁺ cell-specific genes, a three condition comparative analysis of the BMP2⁺ cells to control EBs and to BMP2 ES cells was made (Tables 1 and 2).

Table 1 indicates different GO terms (level 5) and one KEGG pathway (mitogen-activated protein kinase (MAPK) signaling pathway) that are enriched in the BMP2⁺ cells compared to control EBs. In the GO categories 'biological process' (GOTERM_BP), 'molecular function' (GOTERM_MF) and 'cellular component' (GOTERM_CC), several relevant categories are pooled, such as transcriptional processes (for zinc ion binding and transcription) and apoptotic processes (for apoptosis, cell death, death, programmed cell death and regulation of apoptosis) (Table 1). Genes for transcriptional activity and associated with apoptosis are also found to be specifically enriched in the BMP2⁺ cells (Table 2). These results suggest that BMP2 causes direct or indirect induction of apoptotic processes. This hypothesis is supported by the observations that apoptotic effects of BMP2 promote cavitation in EBs and in mouse embryos 13. The MAPK signaling pathway involved in apoptotic processes 14 seems to be specific for the BMP2⁺ cell population (Table 2). It is well established that the MAPK signaling pathway is involved in several processes, including cell cycle progression, cellular transformation apoptosis and differentiation (for a review, see 14). These results suggest that both apoptosis and transcription genes are characteristic gene expression signatures for the BMP2⁺ cells.

Additional data files 3-5 list all the genes and include the change factor (CF) values belonging to the GO categories 'transcription' and 'apoptosis' and the KEGG 'MAPK pathway', respectively. Among genes highly upregulated specifically in the BMP2⁺ cells, Gm397 (gene model 397) and Tbx4 are identified (Additional data file 3). Tbx4 has been shown to be expressed in the lateral mesoderm and is involved in limb outgrowth in the mouse 15 whereas the function of Gm397 is unknown.

Interestingly, among the MAPKs, several kinases belonging to both the classical and the c-Jun amino-terminal kinase (JNK) and p38 MAP kinase pathway were overexpressed in BMP2⁺ cells (Additional data file 5, KEGG pathway scheme). The JNK and p38 MAPK pathway is known to be stimulated by serum and stress factors 14. The most striking gene specifically upregulated in BMP2⁺ cells was Hspa1a (heat shock protein 1A; Additional data file 5), which belongs to the Hsp70 family of stress response genes. Members of this family participate in the process of folding and refolding of misfolded proteins and in the transport of proteins across membranes 16. Hsp1a is also found to be upregulated in chondrons, which includes the chondrocyte and its pericellular matrix, compared to chondrocytes 17.

Table 3 lists the developmental genes that are overexpressed in the BMP2⁺ cells compared to control EBs. Among these, Dnmt3l (DNA (cytosine-5-)-methyltransferase,3-like), Fgf4 (fibroblast growth factor 4), Tdgf1 (teratocarcinoma derived growth factor), Zic1 (zinc finger protein of the cerebellum 1), Ifrd1 (interferon-related developmental regulator 1), Tbx4 (T-Box 4), and Neurod1 (neurogenic differentiation 1) were highly expressed in the BMP2⁺ cells. DNA methylation of the genome is essential for mammalian development and plays crucial roles in a variety of biological processes including genomic imprinting 1819. Dnmt3l^mat-/- mice die before mid-gestation due to an imprinting defect 18. In addition, Dnmt3L is required for differentiation in the extra-embryonic tissue 18. Molecular and genetic data indicate that FGF signaling plays a major role in regulating trophoblast proliferation and differentiation 20. Fgf4 is expressed in early embryos, becoming restricted to the inner cell mass (ICM) of the blastocyst and later to the epiblast of the early post-implantation embryo 20.

Teratocarcinoma-derived growth factor (encoded by Tdgf1, also known as Cripto-1) plays a pivotal role as a multifunctional modulator during embryogenesis and oncogenesis, and may be involved in stem cell maintenance 21. NeuroD1 is a member of the basic helix-loop-helix transcription factor family and has been shown to play a major role in development of the nervous system and formation of the endocrine system 22. The transcription factor ZIC1 plays important roles in patterning the neural plate in early vertebrate development. Zic1 expression was detected in the neural plate border, dorsal neural tube, and somites 23. Moreover, Zic1 plays an important role in early patterning of the Xenopus presumptive neurectoderm 24.

Interferon-related developmental regulator 1 (IFRD1; also known as PC4, Tis7) is a chromatin-associated protein that induces chromatin condensation and plays multiple roles in cellular processes, including transcription, DNA replication and repair 25. It is expressed early in the mouse embryo and extra-embryonic tissues during gastrulation and at mid-gestation in restricted structures (such as the central nervous system, kidney, and lung primordia), whereas it is ubiquitously expressed at late gestation 26. IFRD1 has been shown to act as a coactivator of myogenic differentiation 1 (MyoD1) and myocyte enhancer factor 2C (MEF2C) during myogenesis 27.

The three condition comparative analysis results in a set of seven BMP2⁺ cell-specific genes (Table 4). Among these, the most prominently regulated genes are Zic1, Ifrd1 and Tbx4, which have been discussed previously. Ciliary neurotrophic factor (CNTF) is of particular interest. CNTF is a cytokine with neurotrophic and differentiating effects on central nervous system cells and myotrophic effects on skeletal muscle 28.

To identify overrepresented GO categories or KEGG pathways specifically downregulated in BMP2⁺ cells, we analyzed the data with the DAVID bioinformatics resource 29. Comparative analysis of the expression level of genes in BMP2⁺ and in control EBs shows downregulated genes belong to several overrepresented GO categories, such as focal adhesion, TGF-β signaling pathway, extracellular matrix (ECM)-receptor interaction and shh signaling pathway (Table 5). Some overrepresented categories are related to the developmental processes (for example, development, organ development, embryonic development and brain development; (Table 5, entries in bold). This is not surprising, since the seven-day-old control EBs can still develop into various somatic precursor cells, as indicated in the tables (for example, vasculature development and brain development). Notably, GO categories associated with impaired developmental processes appear not to be characteristic of BMP2⁺ cells when the expression levels of these genes in undifferentiated ES cells are also taken into account. These results clearly show that the BMP2⁺ cells are more closely related to the undifferentiated ES cells than to the control EBs with regard to their developmental potential and plasticity.

Genes belonging to GO categories related to the proliferative processes, such as M phase of mitotic cycle and DNA metabolism are specifically downregulated in the BMP2⁺ cells. Additional data files 6-8 list the genes belonging to the GO 'development' category, TGF-β KEGG pathway and the GO 'M phase' category. The most strikingly downregulated genes from the TGF-β KEGG pathway (Additional data file 7) are Bmp5, Fst (follistatin), Id1 (inhibitor of DNA binding 1) and Tgfβ2 (TGF-β). BMPs are members of the TGF-β superfamily of signal molecules, which mediate many diverse biological processes ranging from early embryonic tissue patterning to postnatal tissue homeostasis 30. WNT, Notch, FGF, Hedgehog and BMP signaling pathways act together during embryogenesis, tissue regeneration and carcinogenesis 31. Follistatin is a BMP antagonist that regulates the actions of the TGF-β superfamily members 32.

We analyzed those transcripts that are differentially expressed in BMP2⁺ cells and involved in selected GO categories of the 'biological process' branch (Additional data file 9). SOURCE 33 was used to obtain GO annotations for the category 'biological process'. The Genesis GO browser (version 1.7.0) 3435 was used to identify transcripts of interest belonging to the biological process categories adhesion, cell cycle, cell death, cell-cell signaling, cellular metabolism, development, stress response, signal transduction, transcription, and transport. Numbers of these transcripts for each selected category are displayed as separate up- and downregulated groups (Additional data file 9, parts B and D).

More stress-related and less developmental genes are identified when the gene expression levels in undifferentiated ES cells are taken into account ('three condition comparative analysis') than when expression levels are compared between BMP2⁺ cells and control cells alone (pairwise comparison). In the three condition comparative analysis, all 16 cell death-related transcripts are upregulated in BMP2⁺ cells. When analyzed further, most of them are apoptosis-related genes. This annotation suggests that cell death during ES cell differentiation mainly involves apoptosis. When ES cells are not taken into account in the two condition comparison, some (18 out of 49) cell death transcripts are downregulated in BMP2⁺ cells. In reference to signal transduction during BMP2⁺ cell differentiation, more transcripts are differentially upregulated than downregulated in BMP2⁺ cells compared to ES cells and control EBs. However, the ratio is reversed when BMP2⁺ cells are compared only with control EBs. The control EBs differentiat to various cell populations and, thus, more signaling pathways are activated than in BMP2⁺ cells, which eventually contribute to signaling pathways limited to development.

Hierarchical clustering of genes identified as differentially expressed and involved in development in the pairwise comparison illustrates how transcripts distribute into co-regulated groups and show good reproducibility between experimental replicates (Additional data file 9, part E). Interestingly, the experimental conditions 'BMP2⁺ cells' and 'undifferentiated cells' are more closely related to each other than to the condition 'control EBs', indicating an earlier developmental stage of BMP2⁺ cells compared to control EBs of the same age (Additional data file 9, part E).

The BMP2⁺ cells showed enriched expression of ectodermal markers neurofilament (NF)-H and NF-M, but it is evident that the BMP2⁺ cells shared more mesodermal characteristics as described in the previous sections. Further investigation of this contradictory phenomenon led to the conclusion that these ectodermal markers may be more likely expressed by NCSCs. In agreement with our results, it was repeatedly reported that NCSCs share more ectodermal and less mesodermal characteristics 55565758. Expression of the NCSC-specific p75^NTR and Nestin transcripts at higher levels compared to the control EBs (Figure 3a,b) confirmed the increased presence of NCSCs in BMP2⁺ cells. In addition, astrocyte-specific GFAP and melanocyte-specific tyrosine phosphatase 1 (Tyrp1) in BMP2⁺ cells (Figure 3a) confirmed the presence of NCSC-derived lineages in the BMP2⁺ lineage as well. Expression of p75^NTR in BMP2⁺ cells was further confirmed by immunostaining with an antibody against p75^NTR in BMP2⁺ cells (Figure 3c, left panel). Furthermore, the presence of glial cells in BMP2⁺ cells has been confirmed by immunostaining with an antibody against GFAP (Figure 3c, right panel). Notably, it has been reported that the differentiation of NCSCs into their lineage fates is mainly dependant on the presence of BMP2 at the required threshold level and also the availability of other factors, such as TGFβ1, Wnt1, Ihh and BMP4, in combination 56. It was well demonstrated that exogenous addition of recombinant BMP2 to cultured NCSCs isolated from chicken explants of cranial and trunk dorsal neural folds from stage 8/9 embryos resulted in the differentiation of NCSC into glia, melanocytes and smooth muscle cells 56. Expression of smooth muscle α actin (SMA) was also detected in the differentiated BMP2⁺ cells (Figure 4).

It is surprising to note that the local concentration of BMP2 secreted by BMP2⁺ cells themselves is able to induce the differentiation of NCSCs into their lineages without the exogenous addition of BMP2.

During vertebrate embryonic development, when the notochord induces the transformation of surface ectoderm to neuroectoderm, a multipotential middle cell layer develops with characteristics of both cell types. These cells are the neural crest cells. They migrate dorsolaterally to form the neural crest, a flattened irregular mass between the surface ectoderm and neuroectoderm. This layer of cells separates into right and left portions and then migrates to various locations within the embryo to give rise to most structures of the peripheral nervous system, such as Schwann and glia cells of the autonomic and enteric nervous systems, endocrine cells, such as the adrenal medulla, and C-cells of the thyroid as well as non-neural tissues, such as pigment cells of the skin and internal organs, smooth muscle of the cardiac outflow tract and great vessels, pericytes, craniofacial bones, cartilage and connective tissues 5960.

Wnt1 and c-kit are well known mediators of melanocyte differentiation of NSCS 6162 and are found to be expressed in the BMP2⁺ population. The tyrosine kinase c-kit has been found in the cell membranes of haematopoietic stem cells, primordial germ cells and presumptive subepidermal melanocytes 63. Intriguingly, wnt-1 and BMP2 act synergistically to suppress differentiation and to maintain NCSC marker expression and multipotency by combinatorial Wnt1/BMP2 signaling 64.

The presence of Wnt1 transcripts in the BMP2⁺ cell population may be interpreted in two ways: first, Wnt1 may be involved in the maintenance of NCSCs in their pluripotency state in combination with BMP2; and second, Wnt1 may be involved in driving the differentiation of NCSCs into melanocytes. Both possibilities cannot be ruled out in the BMP2⁺ cell population since it includes proliferating NCSCs on the one hand (increased cell number when subjected to immunostaining) and melanocytes in the same culture on the other hand. The local BMP2 and/or Wnt1 gradient may drive the NCSCs to produce smooth muscle cells, pericytes, or melanocytes or to remain in their pluripotent state, respectively.

This is the first study that enables us to selectively obtain ES cell-derived NCSCs and their derivatives all at the same time via a BMP2 promoter-based lineage selection approach. The co-expression of Wnt1 and BMP2 indicates the existence of an environment to both keep the NCSCs in stemness and to enable ongoing differentiation of NCSCs to form melanocytes. Thus, the study of these BMP2-expressing cells during early differentiation of ES cells will pave the way for a better understanding of NCSCs and their differentiation into lineages. The BMP2⁺ cells derived from the ES cells may serve as an ideal model for neural crest stem biology in the future since the NCSCs and their derivatives can be selectively enriched by the BMP2 promoter-driven lineage selection approach. In addition, it provides a valuable system where the enriched NCSCs prime themselves to differentiate into their cell specific lineages, since the enriched NCSCs secrete BMP2 and cause a BMP2 gradient, which negates the need for supplying exogenous BMP2. Noggin, an inhibitor of BMP2, can be used to keep the NCSCs in a plastic state. Once the procedures for maintaing BMP2-derived NCSCs in a state of plasticity have been fine tuned, they will be potential candidates for cell replacement therapy, since they can differentiate into any tissue depending upon the local environment of the tissue in which they are injected.

As indicated in Figure 4a, expression of SMA in the BMP2⁺ cells is enriched compared to the control EBs. In addition, the microarray data confirm the upregulation of SMA and other smooth muscle specific genes, such as those encoding calponin and SM22-α, in the BMP2⁺ cell population compared to control EBs (Figure 4b). The immunostaining of smooth muscle cells has been performed with an antibody against SMA, 1 day after plating the BMP2⁺ cells (7 + 1 days in total), 8 days after plating with puromycin (7 + 8 days in total), 18 days after plating with puromycin (7 + 18 days in total) and 18 days after plating without puromycin (7 + 18 days in total) (Figure 4c). Detection of smooth muscle cells even 18 days after puromycin treatment indicates that smooth muscle cells express BMP2 and, therefore, survive the puromycin treatment 65. It is noteworthy that the number of SMA positive cells was less in the culture in which puromycin treatment continued compared to the culture in which puromycin treatment was discontinued. The inhibitory effect in these cultures is more likely due to BMP2 since recent reports demonstrated that BMP2 inhibits proliferation of smooth muscle cells 6667. Simultaneously, the formation of smooth muscle cells from their precursors is crucially dependant on the phenotypic inductive role of BMP2 68.

The expression of the cardiac marker genes NKx2.5, MLC-2a, α-cardiac actin and Mef2c is reduced in the BMP2⁺ cells compared to the control EBs (Figure 5a,e). This suggests that cardiomyogenesis is repressed in the BMP2⁺ cells. Accordingly, Mesp1 was repressed (Figure 2a).

EBs prepared from a mixture of BMP2⁺ cells with wild-type ES cells in ratios of 1:1, 1:2, 1:4, 10:1, 50:1, 1:10 and 1:50, respectively (using the hanging drop protocol and applying the differentiation protocol as outlined in Figure 5b and also previously described 69), did not augment/delay the onset of contractile activity in comparison to control wild-type EBs as observed on day 12. Also, there were no significant differences in terms of the magnitude of the intensity of the contractility (data not shown). This corresponds to the observation that BMP2 added during the differentiation of ES cells did not enhance cardiomyogenesis 70. Compared to the CGR8 wild-type and β-actin control cells, culturing of the BMP2⁺ cells for even a further 28 days (35 days in total) in the presence or absence of puromycin did not result in beating clusters of cardiac cells (data not shown). These results suggest that the plated BMP2⁺ cells did not contain mature beating cardiomyocytes at this stage. In order to investigate the cardiomyogenic potential of the BMP2⁺ cells to differentiate into cardiac beating cells, EBs were made from the BMP2⁺ cells using the hanging drop protocol and the differentiation process was observed in comparison to the EBs formed by cells from control EBs. The secondary EBs made from the BMP2⁺ cells were contracting on day 11, similarly to the secondary EBs formed by cells from seven-day-old control CGR8 wild-type EBs that were not treated with puromycin, as well as EBs made from the β-actin CGR8 clone transfected with β-actin promoter-driven puromycin resistance and EGFP expression cassettes that were treated with puromycin in the same way as the secondary BMP2⁺ EBs. Notably, the intensity of contraction in the BMP2⁺ EBs was significantly stronger compared to that in both controls (Additional data file 10). The whole EB was contracting in a jellyfish-like fashion compared to the controls (Figure 5c; Additional data files 11 and 12). The contracting areas persisted for more than a week, which was longer than in the control EBs. In order to investigate whether the increased beating activity of the cardiomyocytes generated from the BMP2⁺ cells correlates with increased expression of cardiomyocyte specific transcripts, we determined the expression of α-cardiac actin and α-MHC in secondary EBs at day 11. Increased expression levels of both cardiac specific genes was observed in the EBs generated from BMP2⁺ cells compared to both the control secondary EBs (Figure 5d). Thus, BMP2⁺ cells have the capacity to develop into cardiomyocytes. Interestingly, the cardiomyogenic potential apparently seems to be regulated by the BMP2⁺ lineage cells only. These findings suggest that the BMP2⁺ cells are primed to become cardiomyocytes independently of the other, BMP2 negative lineage cells. However, neither the secondary EBs maintained on puromycin nor the contractile secondary EBs when treated with puromycin contained cardiomyocytes. The question of whether the cardiomyocyte precursors contained in the BMP2⁺ cell population are the transient derivatives of the NCSCs or another cell lineage needs to be investigated, since NCSCs are also capable of differentiating into cardiomyocytes 5960. However, although the BMP2⁺ cells are capable of differentiating into beating cardiomyocytes, further extensive investigations are required to demonstrate the cardiogenic potential of these cells in detail, as previously described for ES cell-derived cardiomyocytes 71.

Expression of E-cadherin, Flk1, Flt1, Pecam1 and Runx1 shows the presence of vascular endothelial progenitors in the BMP2⁺ population (Figure 6a). As demonstrated in Figure 1j, BMP2⁺ cells are able to differentiate into cells with an epithelial/endothelial morphology after being cultured for eight days in the presence of puromycin. Detection of epithelial-like cells has been confirmed by immunostaining. As indicated in Figure 6c, epithelial-like cells were detected even after 11 days of culturing in the presence of puromycin.

Expression of c-kit and CD34 and Runx1 indicates the presence of haematopoietic stem cells in the BMP2⁺ population (Figure 6a,b). Ihh, which is implicated in haematopoiesis and vasculogenesis 72 is also expressed in the BMP2⁺ cells but its expression level is lower than that of the control cells. Flt-1, also known as vascular endothelial growth factor receptor 1 (VEGFR-1), is a high-affinity tyrosine kinase receptor for vascular endothelial growth factor and is normally expressed only in vascular endothelial cells. However, the Flt-1 transcript was recently found to be expressed in human peripheral blood monocytes 73. Monocytes are known to differentiate into a variety of cell types, such as osteoclasts in bone, dendritic cells in the immune system and mature macrophages in a number of tissues, for example, Kupffer cells in liver 73. Expression of Flt-1 and c-fms in the BMP2⁺ population indicates the possible existence of a monocyte-macrophage lineage in these cells.

Expression of osteopontin, Cbfa, osteocalcin and alkaline phosphatase in BMP2⁺ cells showed the possible occurrence of osteoblasts (Figure 7a). BMP2⁺ cells plated and maintained in the differentiation medium with puromycin showed Alizarin red positive clusters after 18 days (Figure 7b), which indicates the first appearance of osteoblasts. After 35 days of culturing in the presence of puromycin, areas brightly stained with Alizarin red were observed, confirming the occurrence of osteoblasts in culture (Figure 7c). These results show that the puromycin resistant BMP2⁺ cells per se are capable of differentiating into osteoblasts. It may also be possible that the osteoblasts might develop from the BMP2⁺ derived NCSCs in the presence of an increased BMP2 concentration for longer periods of time 74.

Expression of the transcription factor Sox9, the proteoglycan aggrecan, and collagen II (with two alternative splicing forms) was detected in the BMP2⁺ population, indicating the existence of chondrocytes 75 (Figure 7a). Sox9 is a key transcriptional factor for chondrocytic differentiation of mesenchymal cells via chondrocyte-specific enhancer of the pro alpha1(II) collagen 76. Expression of Pax-7 and myoD shows the presence of satellite cells 7778. Satellite cells represent a distinct lineage of myogenic progenitors responsible for the maintenance of skeletal muscle 79. MyoD is expressed only when satellite cells are activated to proliferate and differentiate into primary myoblasts, which will in turn differentiate into cells of the myofibres of skeletal muscles 78. An elevated level of MyoD indicates either the proliferation or the differentiation of satellite cells into primary myoblasts in the BMP2⁺ population. But the expression of myogenin and myostatin was not detected by RT-PCR, even after 40 cycles in the BMP2⁺ cell population, suggesting the absence of skeletal muscle cells. This hypothesis is supported by Affymetrix data that show that myogenin is very weakly expressed; myostatin (Gdf8) is not present on the Affymetrix chip.

Interestingly, it has been reported that the satellite cells can spontaneously differentiate into adipocytes in an alternative mesenchymal pathway 80. Expression of FSP-1 shows the occurrence of fibroblasts in the BMP2⁺ cell population. FSP-1 has been implicated in the epithelial-mesenchymal transition and is a representative marker for mesenchymal cells 81.

The RT-PCR analysis of the BMP2⁺ population showed the expression of the pre-adipocyte gene marker pref-1 82, the pan-adipocyte markers PPARγ 83, aP2, Glut4 84 and LPL as well as the pan-macrophage marker c-fms 828385 (Figure 8a,b). Although the markers PPARγ, aP2 and Glut4 are unique to adipocytes, these genes are also expressed in macrophages. Similarly, the trophoblast lineage also expresses c-fms 82. Notably, the cells derived from the BMP2⁺ cells cultured in the continual presence of puromycin were not positive for either Sudan Red or anti F4/80 stainings. However, when puromycin treatment was discontinued, there were cells stained with Sudan Red and F4/80, a marker for macrophages, as shown in Figure 8c,d. These findings suggest that macrophages are derived from differentiated BMP2⁺ cells in which BMP2 expression is downregulated at a later time. There were no cells stained with Oil Red O in both cases (with discontinued puromycin treatment), even when cultured for a prolonged period of days. Thus, the BMP2⁺ population might contain the pre-adipocytes expressing only PPARγ, aP2, Glut4 and Pref-1 but not adipsin. Apparently, other factors secreted by BMP2 negative cells are required for maturation of pre-adipocytes to adipocytes.

Murine CGR8 ES cells (ECACC 95011018) were cultured in the absence of feeder cells in Glasgow minimum essential medium (GMEM) supplemented with 10% fetal bovine serum (FBS), 2 mM L-glutamine, 100 units/ml leukemia inhibitory factor (LIF) and 50 μM β-mercaptoethanol (ME) in 0.2% gelatine coated flasks as previously described 69. To induce differentiation, an ES cell suspension of 1.6 × 10⁴ cells/ml was made in Iscove's modified Dulbecco's Medium (IMDM) supplemented with 20% (FBS), 1% non-essential amino acids (vol/vol), 2 mM L-glutamine and 100 μM β-ME. The cells were cultured in bacteriological dishes for two days, then the two-day-old EBs were transferred into a 0.2% gelatine coated 10 cm tissue culture dish. On day 4, the EBs were treated with 3 μg/ml puromycin for the next 3 days. Medium containing puromycin was refreshed every day. On day 7, the puromycin resistant BMP2⁺ cells were used for the experiments.

The hanging drop differentiation protocol as described in Figure 1 was carried out as follows. Briefly, 500 undifferentiated BMP2⁺ ES cells were spotted in 20 μl differentiation medium in the upper lid of a bacteriological dish and cultured for 2 days. On day 2, the EBs were transferred into suspension until assays were performed, with medium changes in between as outlined in Figure 5b.

pIRES2 EGFP was purchased from Clontech (Heidelberg, Germany). Human cytomegalovirus (CMV) immediate early promoter and enhancer were removed by a double digestion with NheI and AseI and then subsequently blunt end ligated to get the pIRES2 EGFP φ CMV construct. Puromycin cDNA flanked on both ends by BamH1 restriction sites, which was PCR amplified from pIRESPuro3 by Pfu DNA polymerase (Promega, Mannheim, Germany), was inserted at the SmaI site of the pIRES2 EGFP φ CMV construct to get pPuro IRES2 EGFP. A 2.9 kb BMP2 promoter fragment with both proximal and distal transcription start sites, excised with HindIII and SacI digestion from pBMP2-GL3 86, were blunt end ligated to EcoRI digested and klenowed pPuro IRES2 EGFP to generate pBMP2^p Puro IRES2 EGFP. This BMP2 reporter construct drives the expression of both puromycin resistance and EGFP under the control of the BMP2 promoter by the use of an IRES sequence. This construct was electroporated in CGR8 with 500 μF and 240 V in a Bio-Rad Gene Pulser™ (Bio-Rad, Hercules, CA, USA). The transfected clones were selected using 400 μg/ml neomycin and, after selection, were maintained with 200 μg/ml neomycin. During EB generation, the neomycin selection was discontinued.

One day prior to the sample processing, 50,000 cells per each well of Lab-Tek Permanox slide Chambers (Nalge Nunc International, Naperville, IL, USA) were seeded and cultured in the absence of puromycin. After 24 hours, the samples were fixed with either 4% paraformaldehyde in phosphate-buffered saline for 10 minutes or -20°C precooled methanol-acetone (1:1) solution, permeabilized with 0.1% Triton X-100, and labeled with the following antibodies: mouse anti-SMA (1:200) (Sigma, Taufkirchen, Germany); mouse anti-pan cytokeratin (1:50) (DakoCytomation, Hamburg, Germany); mouse anti-Ksp-cadherin (Zymed, South San Francisco, CA, USA); mouse anti-myoD1 (Santa Cruz, Heidelberg, Germany); anti-actinin antibodies (Sigma, Taufkirchen, Germany); mouse anti-GFAP (1:100) (Sigma, Taufkirchen, Germany); rabbit anti-p75 NTR (1:100), (Chemicon, Hampshire, UK); rabbit anti-S100 (1:400), (DakoCytomation, Hamburg, Germany); and mouse anti-actinin (1:100). This was followed by labeling with secondary antibodies, either Cy3-labeled anti-rabbit IgG (Sigma), rhodamine labeled anti-mouse IgG (Sigma), or Cy3-labeled donkey anti-rat IgG antibodies. The specificity of the antibodies has been tested using the appropriate tissues (Additional data file 13). For a control of SMA immunostaining, vessels of mouse liver were stained. Heart and skeletal muscles of embryonic mouse tissues served as controls for the specifity of the anti-actinin and the myoD1 antibodies, and murine brain tissue for the specifity of antibodies against neuronal markers such as p75-NTR and GFAP. Mouse kidney and small intestine were used as control tissues for Ksp-cadherin or E-cadherin and pan-cyokeratin staining, respectively.

Sudan Red staining was performed on paraformaldehyde fixed embryonic cell clusters or on cytospins of human lymphocytes. The paraformaldehyde-fixed samples were treated with Alizarin Red S for five minutes and then excess Alizarin Red was washed off by dipping in acetone:xylene (1:1), followed by a wash in xylene. They were then stained for haematoxylin for ten minutes and the excess dye was washed off in running water.

Total RNA was extracted using RNeasy Mini Kit (Qiagen, Hilden, Germany) with on-column DNase I (Qiagen) digestion according to the manufacturer's instructions. Total RNA (5 μg) was reverse transcribed using SuperScript II Reverse transcriptase (Invitrogen GmbH, Karlsruhe, Germany) with random primers according to the manufacturer's recommended protocol. PCR amplification was done with REDTaq ReadyMix (Sigma) with 0.4 μM of each primer. GAPDH was used as an internal control. The following conditions were used. An initial denaturation at 95°C for 2 minutes, followed by 22-35 cycles of 30 s denaturation at 95°C, 30 s annealing at 60°C and 60 s of elongation at 72°C. A final extension at 72°C for 5 minutes was included. Electrophoretic separation of PCR products was carried out on 2% agarose gels with 0.001% ethidium bromide. The primer pairs included in this study are listed in Additional data file 14.

A single cell suspension was prepared by trypsinization. Cell clumps were removed by passing through the cell strainer cap of a round bottom tube from BD Falcon^® (Heidelberg, Germany). Propidium iodide staining (Sigma) was included to exclude dead cells. Acquisition of 10,000 live (PI negative) cells was done with a FACScan (BD Biosciences, Heidelberg, Germany) and the data analysis was done with CellQuest software (Becton Dickinson, Heidelberg, Germany). The wild-type EBs were used as the control on the same day as the sample EBs.

Total RNA was extracted from undifferentiated ES cells and EBs using the RNeasy total RNA isolation kit (Qiagen GmbH, Hilden, Germany). The preparation quality was assessed by agarose formaldehyde gel electrophoresis. Three independent total RNA preparations, each 15 μg from the BMP2⁺ cells, the mixed cell population and the undifferentiated ES cells, were labelled with the One-Cycle Target Labeling and Control Reagent package (Affymetrix, High Wycombe, UK) as described in the manufacturer's instructions. Briefly, double-stranded cDNA was synthesized using the one-cycle cDNA synthesis module. Biotinylated cRNA was synthesized with the IVT labeling kit and cleaned up using the sample cleanup module.

After fragmenting of the cRNA for target preparation using the standard Affymetrix protocol, 15 μg fragmented cRNA was hybridized for 16 h at 45°C to Mouse Genome 430 2.0 arrays, which carry probe sets representing 45,101 probe sets. Following hybridization, arrays were washed and stained with streptavidin-phycoerythrin in the Affymetrix Fluidics Station 450 and further scanned using the Affymetrix GeneChip Scanner 3000 7G. The image data were analyzed with GCOS 1.4 using Affymetrix default analysis settings. After RMA normalization 87, three pair-wise comparisons were performed using the Student's t-test (unpaired, assuming unequal variances). A Student's t-test p value < 10^-2 and a fold change >2 were used to identify and restrict the number of differentially expressed genes. Hierarchical clustering was performed for an intersection of genes differentially expressed between undifferentiated BMP2 ES cells and BMP2⁺ EBs as well as differentially expressed between seven-day-old control EBs and BMP2⁺ EBs (three condition comparative analysis) to determine and differentiate treatment and developmental aspects. The cluster analysis (see Additional data file 9, part E) was done using cluster version 2.11 88, applying mean-centering and normalization of genes and arrays before average linkage clustering with uncentered correlation.

Differentially expressed genes were analyzed according to predefined pathways or functional categories annotated by KEGG 11, BioCarta 12, and GO 10 using the DAVID bioinformatic resource 29. For an overrepresented GO, Biocarta or KEGG pathway, a cut-off p value of 0.1 was chosen. In general, it should be noted that one gene can participate in more than one KEGG or BioCarta pathway or GO category.

Validation of the Affymetrix data was performed by qPCR analysis with the ABI Prism 7900 HT Sequence Detection System (Applied Biosystems, Foster City, CA, USA). RNA (1 μg) from BMP2 ES cells, seven-day-old control EBs and BMP2⁺ cells were reverse transcribed with ThermoScript™ Reverse Transcriptase (Invitrogen). Then, qPCR was perfomed in triplicate for every sample using TaqMan Gene Expression Assays (Applied Biosystems). The Gene Expression Assays used for validation were Brachyury (T) (Mm00436877_m1), NF-H (Mm01191456_m1), BMP2 (Mm01340178_m1), GAPDH (Mm99999915_g1), and Nanog (Mm02019550_s1). Averaged C_t values of each qPCR reaction from the target gene were normalized with the average C_t values of the housekeeping gene, GAPDH, which ran in the same reaction plate, to obtain the ΔC_t value. The fold change was calculated as follows: fold change = 2−(ΔCtgene1−ΔCtgene2) MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfeBSjuyZL2yd9gzLbvyNv2Caerbhv2BYDwAHbqedmvETj2BSbqee0evGueE0jxyaibaiKI8=vI8tuQ8FMI8Gi=hEeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciGacaGaaeqabaqadeqadaaakeaacaaIYaWaaWbaaSqabeaacqGHsislcaGGOaGaeuiLdqKaae4qamaaBaaameaacaqG0baabeaaliaabEgacaqGLbGaaeOBaiaabwgacaaIXaGaeyOeI0IaeuiLdqKaae4qamaaBaaameaacaqG0baabeaaliaabEgacaqGLbGaaeOBaiaabwgacaaIYaGaaiykaaaaaaa@46CC@. Since the genes included are not expressed in at least one of the three conditions (BMP2 ES cells, seven-day-old control EBs or BMP2⁺ cells), the ΔC_t of the gene in the sample with the lowest expression was used as ΔCt gene2 to calculate the fold change using the above formula. The resulting fold change is expressed as the percentage of the maximum fold change.