Mesenchymal stem cells (MSCs) are present in a wide variety of tissues during development of the human embryo starting as early as the first trimester. Gene expression profiling of these cells has focused primarily on the molecular signs characterizing their potential heterogeneity and their differentiation potential. In contrast, molecular mechanisms participating in the emergence of MSC identity in embryo are still poorly understood. In this study, human embryonic stem cells (hESs) were differentiated toward MSCs (ES-MSCs) to compare the genetic patterns between pluripotent hESs and multipotent MSCs by a large genomewide expression profiling of mRNAs and microRNAs (miRNAs). After whole genome differential transcriptomic analysis, a stringent protocol was used to search for genes differentially expressed between hESs and ES-MSCs, followed by several validation steps to identify the genes most specifically linked to the MSC phenotype. A network was obtained that encompassed 74 genes in 13 interconnected transcriptional systems that are likely to contribute to MSC identity. Pairs of negatively correlated miRNAs and mRNAs, which suggest miRNA-target relationships, were then extracted and validation was sought with the use of Pre-miRs. We report here that underexpression of miR-148a and miR-20b in ES-MSCs, compared with ESs, allows an increase in expression of the EPAS1 (Endothelial PAS domain 1) transcription factor that results in the expression of markers of the MSC phenotype specification.
- human embryonic stem cells
- microarray analysis
mesenchymal stem cells (MSCs), identified several decades ago as bone-forming progenitor cells in the bone marrow (21), were subsequently defined phenotypically as CD29+, CD44+, CD90+, and CD105+ cells, negative for hematopoietic lineage markers and HLA-DR (68). They have more recently attracted enormous attention because of their presence in a large number of other tissues in the adult, including adipose tissue, peripheral blood, dental pulp, dermis, synovial liquid, and skeletal muscle (16, 50, 68, 74, 75). Of similar interest has been the demonstration of MSCs not only in the adult but also at almost all stages of development, as early as the first trimester in the human embryo (14), in the amniotic liquid (30), in the placenta (43), and in the umbilical cord blood (68). Those populations of self-renewable multipotent MSCs share a large set of phenotypic markers, although none has been shown to be specific to date, and they exhibit differentiation abilities along the osteogenic, chondrogenic, and adipogenic pathways.
MSCs appear, therefore, as a unique multipotent stem cell population with functional specificities that distinguish them from differentiated cells as well as from other stem cells, either the immortal pluripotent blastocyst-derived embryonic stem cells (ESs) or the well-localized less multipotent tissue-specific stem cells. Analysis of their gene expression profiles, focusing especially on the molecular correlates of the heterogeneity among MSCs of different origins (25, 66, 68) and their differentiation potential, has provided important clues to specialization routes for these cells (60, 73). Recent studies reported that gene expression profiling of MSCs derived from human (h)ESs was very similar to that of MSCs derived from bone marrow (BM-MSCs) (17) or from fetal tissues (32). In contrast, molecular mechanisms participating in the emergence of MSC identity in embryo have attracted less attention.
In the present study, we have compared the genetic pattern of the pluripotent hESs and their MSC derivatives, seeking molecular clues specific for MSC identity. We have taken advantage of recent protocols that trigger in vitro differentiation of MSCs from hESs (hereafter called ES-MSCs) (5, 6, 26, 37, 40, 45). Phenotypically characterized cell populations obtained at near-homogeneity in any desired amount allowed us to perform whole genome differential transcriptome profiling under the best technical conditions (59) correlated with analysis of microRNA (miRNA) signatures. Extraction of negatively correlated pairs of miRNA and mRNA then pointed to miRNA-target relationships.
Isolation and amplification of hESs.
The hES cell line VUB01 (XY, passage 80) was derived at the Vrije Universiteit Brussels (41) and H9 (XX, passages 50–60, WA09) by the National Stem Cell Bank. The hES cell line SA01 (XY, passage 40) is distributed by Cellartis. VUB01 and H9 were maintained on a feeder layer of mitomycin C-inactivated murine embryonic fibroblast (MEF) cells, in a humidified 10% CO2 incubator at 37°C in Knock-Out (KO)-DMEM supplemented with 20% KO Serum Replacement (KSR), 1 mM l-glutamine, 1% nonessential amino acids, 0.1 mM β-mercaptoethanol and 4 ng/ml basic (b)FGF (all from Invitrogen). SA01 was maintained on an inactivated human foreskin fibroblast feeder in DMEM-F-12 supplemented with 20% KSR, 1 mM L-glutamine, 1% nonessential amino acids, 0.1 mM β-mercaptoethanol, and 4 ng/ml bFGF in a humidified 5% CO2 incubator at 37°C.
For the three cell lines, culture medium was changed daily and routine passages were performed by mechanical cutting of ESs on a fresh feeder layer every 4–5 days.
Isolation, purification, and expansion of hMSCs from the bone marrow.
Bone marrow cells were obtained from iliac crest aspirates from healthy donors giving cells for allogeneic transplantation purposes after informed consent. They were used in accordance with the procedures approved by the human experimentation and ethics committee of the Hopital St Antoine (Paris, France). Isolation, purification, and expansion were performed as previously described (8).
Differentiation of hES.
Mesenchymal differentiation was obtained based on a protocol described by Barberi et al. (5). Briefly, differentiation was induced by plating 2 × 104 ESs/cm2 on 0.1% gelatin-coated dishes in the presence of KO-DMEM medium supplemented with 20% fetal bovine serum (FBS, Invitrogen), 1 mM l-glutamine, 1% nonessential amino acids, 1% penicillin-streptomycin, and 0.1 mM β-mercaptoethanol. Medium was changed every other day. Confluent cells were passaged with trypsin-EDTA 1× (Invitrogen) in new gelatin-coated dishes.
To induce osteoblastic differentiation, cells were plated at a density of 30,000 cells/cm2 in a specific medium from Cambrex containing dexamethasone, ascorbate, and β-glycerophosphate. After 21 days cells were analyzed by alkaline phosphatase activity with an enzyme kit from Sigma-Aldrich.
Adipogenic differentiation was induced by culturing the cells in the same medium as that used for differentiation supplemented with 100 μM linoleic acid. Adipogenesis was detected by the presence of neutral lipids in the cytoplasm stained with Oil Red O.
Immunophenotyping was carried out with a FACScalibur and Cell Quest software (Becton Dickinson Biosciences). More than 10,000 events were acquired for each sample and analyzed. Cells were collected with trypsin (trypsin-EDTA 1×; Invitrogen) and resuspended at 5 × 105 cells in PBS-2% FBS. Cells were stained for 20 min at room temperature with one of the following anti-human antibodies: CD73-PE (SH3/NT5E), CD44-PE, CD54-PE (ICAM-1), CD29-PE (integrin-β1), CD106-PE (VCAM), CD166-PE (ALCAM), CD14-PE, CD31-PE (PECAM-1), CD56-PE (NCAM), HLA-ABC-PE, HLA-DR-PE, CD34-APC, and CD45-FITC (all from Becton Dickinson Biosciences/Pharmingen) and CD105-PE (SH2/Endoglin; Caltag); primary monoclonal antibody FORSE1, vimentin, and Stro1 were used with mouse IgG- or IgM-Alexa as secondary antibody. Mouse isotype antibodies served as respective controls (Becton Dickinson).
For immunohistochemistry, cells were fixed in 4% paraformaldehyde solution for 20 min at room temperature, washed with PBS three times, and exposed to blocking buffer (1% BSA-5% goat serum) and 0.1% Triton X-100 when permeabilization was required. Cells were stained for 2 h with either the primary monoclonal antibody Stro1 or α-smooth muscle actin (α-SMA) and next with the appropriate secondary antibody and DAPI for 1 h.
The proliferation potential of MSCs was measured with a commercial kit (Cambrex) according to the manufacturer's protocol.
Total RNA was extracted from the cells with the RNeasy Mini Kit (Qiagen) according to the manufacturer's protocols. The quality of RNA was controlled with a BioAnalyzer 2100 (Agilent). For DNA microarrays, each biological sample was processed in triplicate for both cell lines (VUB01 and SA01), i.e., altogether six RNA preparations were analyzed for undifferentiated ESs and six for ES-MSCs. Analysis was performed with an Affymetrix platform (Institut Curie Paris, Réseau National des Genopoles, France). RNA samples were processed for labeling and then hybridized on Affymetrix HG_U133_Plus_2 human oligonucleotide arrays; the DNA chips were scanned according to the Affymetrix protocols.
Real Time RT-PCR
For quantitative RT-PCR, total RNA (500 ng) was reverse transcribed with SuperScript III (Invitrogen) as described in the manufacturer's protocol.
Real-time PCR was performed with SYBR Green Core Reagents (Applied Biosystems) according to the manufacturer's protocol. The incorporation of the SYBR Green dye into the PCR products was monitored in real time with the Chromo4 system (Bio-Rad). The efficiency of the amplification was determined for each pair of primers by comparison with a standard curve generated with serially diluted cDNA. Target genes were quantified relative to a reference gene (β-TUBULIN) with the mathematical model described by Pfaffl (49). All PCR reactions were performed in triplicate. The complete list of primers used is presented in Supplemental Table S7.1
Array Assist 4.1 software (Stratagen) was used to achieve the statistical treatment of the expression data. The GC-RMA algorithm was used for normalization. Next, a one-way ANOVA was performed to eliminate genes that presented a significant variance between biological samples (P value < 0.05 corrected with Benjamini and Hochberg false discovery rate) (7).
Paired Student's t-tests were performed to compare the expression changes between samples. The significantly modulated genes were defined as genes that showed a fold change (FC) ≥ 2 and a P value ≤ 0.01 (P value corrected with Benjamini and Hochberg false discovery rate). Ingenuity software (www.ingenuity.com) was used to build gene networks.
Establishment of List of MSC Markers from the Literature
Analysis was restricted to genes related to those previously described in the literature as differentially attached to the MSC phenotype (12, 19, 27, 28, 45, 58, 65, 68). This generated a list of 262 genes (Supplemental Table S4), of which 2 were found in four of those studies, 14 in three, 31 in two, and 215 in one.
All listed genes that were upregulated in ES-MSC samples were used as “leads” to identify in silico all transcription regulators that positively regulated them. Then, all relevant transcription regulators upregulated in ES-MSC samples were sorted out. Another search in databases identified all genes that were positively regulated by the selected set of transcription regulators, and this list was then cross-matched with upregulated genes in ES-MSC samples. Altogether, this procedure produced networks that contained all genes that were both upregulated in ES-MSC samples (compared with undifferentiated ESs) and linked, directly or indirectly, to MSC markers previously identified in the literature.
miRNA profile analysis was performed with TaqMan Array encoding for human miRNA panel V1.0 (Applied Biosystems) on three independent RNA preparations for the undifferentiated hESs and the ES-MSCs for each cell line (VUB01, H9, and SA001). After induction of hESs toward the neural lineage, neural precursors (ES-NPs) were purified from neural rosettes (48) by cell sorting using the panneuronal surface marker neural cell adhesion molecule (NCAM or CD56).
Eighty nanograms of total RNA was extracted with the mirVana Isolation Kit (Ambion) and reverse transcribed with the TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems) in a multiplex RT system consisting of eight predefined RT primer pools containing up to 48 RT primers each. All 365 miRNA targets were reverse-transcribed in eight separate RT reactions, and each RT reaction was pipetted into one of the eight filling ports on the TaqMan Array.
Real-time PCR was performed with no UNG no AmpErase TaqMan MASTER MIX (Applied Biosystems) on an ABI 7900, and results were normalized against RNU 48, a small nucleolar RNA (snoRNA).
Data are expressed as means and SD in the three independent RNA preparations. Student's t-test was used to identify miRNAs differentially expressed between hESs and ES-MSCs (or ES-NPs) with a FC ≥ 2 and a P value ≤ 0.01.
Transfection of microRNA Precursors
MSC cells were seeded at 30,000–40,000 cells per P24 well plates and transfected with Pre-miRT miRNA Precursor (Ambion) at a final concentration of 100 nM with the Lipofectamine RNAiMAX transfection reagent (Invitrogen). Total RNA and protein were collected 72 and 96 h after transfection. A CyTM3-labeled Anti-miR Negative Control (Ambion) was used as a control.
Extraction of total RNA, reverse transcription, and analysis of miR expression level by real-time PCR were performed as described above.
Luciferase Reporter Assays
The GeneCopoeia pEZX-MT01 target sequence expression clone containing the entire human EPAS1 (Endothelial PAS domain 1) 3′-untranslated region (UTR) sequence (accession no. NM_001430.3) that had firefly luciferase as the reporter gene was provided by LabOmics. A tracking reporter gene, Renilla luciferase, was also provided in the vector system, which allows indicators for successful transfection and expression of these miRNA target sequence constructs in target cells.
For luciferase reporter assays, HEK 293 cells were plated at a density of 15,000 cells/well in 96-well plates (Corning). Transfection of the miRNAs by Lipofectamine RNAiMAX was performed 48 h after plating. Cells were then transfected 24 h later with the pEZX-MT01 containing the EPAS1 3′-UTR sequence plasmid or by the empty vector used as a control by Lipofectamine LTX. Cells were lysed 72 h after the first transfection, and assay for firefly and Renilla luciferase activity with the Dual Glo Luciferase Assay System (Promega) was performed according to the manufacturer's instructions and measured on the Analyst GT (Molecular Devices).
Differentiation of ES-MSCs displaying a phenotype similar to that described by previous authors (5, 6, 45) was readily obtained after ∼20 days of culture (2 passages). CD73+ cells amounted to ∼65% of the population at day 13 and reached >98% at passage P2. ES-MSCs expressed endoglin (CD105), integrin-β1 (CD29), CD44, ALCAM (CD166), Vimentin, Stro1, and HLA-ABC (Fig. 1A). They were negative for hematopoietic markers (CD34, CD45, and CD14), neuronal markers (NCAM/CD56 and FORSE1), and the endothelial marker CD31. Cells were immunoreactive for Stro1 and α-SMA (Fig. 1B). A profound decrease in expression of OCT4 and NANOG was observed by quantitative RT-PCR compared with undifferentiated hES, demonstrating the absence of undifferentiated cells in the ES-MSC population (Fig. 1C). The ability of the cells to differentiate into osteoblasts and adipocytes was assessed by the expression of both alkaline phosphatase and Oil Red O staining after treatment with the inductive medium (Fig. 1D). ES-MSCs retained the same phenotype for up to 20 passages or 2 cycles of freezing (Supplemental Table S1), and their doubling time, evaluated with bromodeoxyuridine (BrdU), remained constant (Supplemental Fig. S1).
A set of 5,543 genes was identified in ES-MSCs with a FC superior to 2.0 compared with hESs and a P value < 0.01, including 2,018 upregulated and 3,525 downregulated genes (Supplemental Table S2). Nine hundred thirty-six (17%) of these have no known functions, among which 25 downregulated genes had a FC ≥ 20, suggesting that they may be associated with pluripotency. The most marked decrease in expression was observed for known pluripotency transcription factors such as NANOG (FC = 260.8), SOX2 (FC = 232.2), and OCT4/POU5F1 (FC = 127.4) (Supplemental Table S3) (11). Expression of TDGF1/CRIPTO, another transcription factor associated with ESs (62), even showed a negative FC of 761.
Expression of MSC-Associated Genes in ES-MSCs
A multistep in silico strategy was applied to these data in order to sort out gene expression patterns that would be specifically linked to the phenotypic MSC features. Upregulated genes already identified as MSC markers in the literature allowed us to retrieve all transcription regulators positively associated with them from databases. In this general list, transcription regulators that were upregulated in ES-MSCs were then sorted out. Transcriptional networks were constructed by identifying upregulated genes in ES-MSCs that are known target genes of those transcription regulators.
By comparison with a list of 262 genes derived from a literature search (see methods), 95 upregulated genes were identified in ES-MSCs (Supplemental Table S4). These comprised typical MSC markers, such as CD44 (FC = 110), CD73/NT5E (FC = 83.7), ALCAM/CD166 (FC = 14.5), INTEGRIN-αV/ITGAV (FC = 10.5), INTEGRIN-β1/ITGB1 (FC = 8.8), VIMENTIN/VIM (FC = 4.8), and CD105/ENDOGLIN (FC = 4.3), as well as FN1 (FC = 152.02), COL1A1 (FC = 90.28), COL6A3 (FC = 86.01), COL1A2 (FC = 55.22), COL4A2 (FC = 21.94), CTGF (FC = 20.72), RUNX2 (FC = 10.84), PLOD2 (FC = 10.73), FTH1 (FC = 2.44), and TMSB4X (FC = 2.03). PPARG and SOX9 just missed statistical significance (FC = 1.6- and 2.2-fold, P value 0.05 and 0.02, respectively).
Thirty-six upregulated transcription regulators were positively linked to 61 of those 95 “MSC marker genes” in ES-MSCs (Supplemental Table S5). Gene networks were completed by adding 68 genes that were both listed in databases as positively controlled by one or more of those 36 transcription regulators and upregulated in ES-MSCs (Table 1), among which were 9 more transcription regulators (DSCR1, MBD2, FOSL2, NFKB1, BCL3, PCAF, SQSTM1, CBFB, ETV6). This led to a total of 165 genes with which networks were built with the Ingenuity software. This analysis identified three networks. Two were not further studied because they seemed to associate in a nonspecific manner to differentiation. The first network comprised mostly genes associated with the cell cycle, which is known to be different in the undifferentiated stage (20). The second network was centered on the ubiquitous transcription factor JUN (Supplemental Fig. S2).
In contrast, the third network—which consisted of 74 of the 165 selected genes—comprised the largest panel of genes linked to the extracellular matrix and differentiation pathways toward main MSC progeny (Fig. 2 and Table 2). This network included altogether 13 transcription factors, 32 known MSC markers, and 29 other genes. Interestingly, the expression level of transcription factors upregulated in ES-MSCs was quite similar compared with BM-MSCs (Supplemental Fig. S3).
miRNA Profile and Function in ES-MSCs
To elucidate the molecular mechanism that contributes to MSC identity, miRNAs that targeted the transcription regulators implicated in the upregulation of MSC marker genes were searched.
Of 367 human miRNAs from which expression was compared in hESs and ES-MSCs, 40 were upregulated and 127 downregulated in ES-MSCs (Supplemental Table S6). Among these, the most downregulated were the miRNAs known to be implicated in the maintenance of hES pluripotency such as miR-372, miR-302a,b,c,d, miR-367, miR-371, miR-373, and miR-520g. As a demonstration of the specificity of the analysis, miR-9, miR-33, and miR-124a, which are known to be implicated in neural differentiation, as well as miR-133a and miR-133b expressed in cardiac and skeletal muscles were also downregulated in ES-MSCs. Conversely, upregulated miRNAs comprised miR-27a, miR-27b, miR-148b, miR-210, and miR-143,145, which are known to be implicated in the MSC differentiation toward osteoblasts and adipocytes, respectively (Supplemental Table S6). To be further able to discriminate among the modulated miRNAs those that could be more specifically associated to the establishment of the MSC phenotype, a parallel “countertest” was performed during the differentiation of hESs toward hES-derived neural precursor cells (ES-NPs).
The miRNAs that target the 13 upregulated transcription regulators in ES-MSCs were searched by TargetScan (Table 3), which identified miRNAs that were preferentially downregulated with a strong FC ratio (>5) in ES-MSCs while upregulated or unaffected in ES-NPs (Table 3). Twelve miRNAs matched those stringent criteria and were therefore good candidates for a role in MSC specification: miR-9, which targets both FLT1 and EPAS1, the cluster that comprises miR-17p5, miR-20a, miR-20b, miR-93, miR-106b, and miR-148a, which targets EPAS1, miR-18a and b, which target ETV6, and miR-15a and miR-195, which target TFAP2A.
EPAS1 thus appeared as the most frequently targeted gene. We therefore specifically looked for the functional effects of its control by miRNAs, using Pre-miRs for miR-148a and miR-20b. miR-429, which binds the unrelated transcription factor BHLHB3, was used as a control (Supplemental Fig. S3). To test directly whether miR-148a and miR-20b can repress translation through binding to EPAS1, we used the luciferase reporter construct pEZX-MT01 with the entire human EPAS1 3′-UTR immediately following the Renilla coding sequence. HEK293 cells were cotransfected sequentially with miRNAs (miR-148a, miR-20b, miR-429, or scramble miRNA) followed by the luciferase/EPAS1 3′-UTR reporter construct or the empty vector. In cells transfected with miR-148a and miR-20b, a decrease of ∼40% of the luciferase expression was observed compared with scramble miRNA, whereas miR-429 had a weaker effect (Fig. 3B). These data indicated that miR-148a and miR-20b can downregulate expression from the EPAS1 mRNA carrying the 3′-UTR.
The expression level of MSC markers linked to EPAS1 was then analyzed by quantitative RT-PCR. Control levels of these marker genes were similar in ES-MSCs and BM-MSCs, whereas they were strongly upregulated compared with hESs (Fig. 3C). Pre-miR treatment that induced overexpression of either miR-148a or miR-20b led 72 h after transfection (maintained or increased at 96 h) to a significant decrease in expression levels of the EPAS1-related genes GBE1, HIF1A, and PLOD2 (Fig. 3C). PLAU and PLAUR, which are present but more remote in the composite network drawn from transcriptome profiling results, were also significantly affected. Conversely, DUSP1 and LOX, although closely associated to EPAS1 in the network, were not. Pre-miR treatment for miR-429 did not affect any of these genes compared with the Anti-miR Negative Control (Fig. 3D). On the contrary, hypoxia conditions did not affect the expression of these MSC markers (not shown).
The main result of this study is the identification, in human cells, of a network of 13 interconnected transcription factors, corresponding to a specific gene expression pattern of MSCs compared with ESs. Concurrent analysis of miRNA expression profiles revealed that miR-148a and miR-20b might be implicated in MSC identity by controlling the expression of EPAS1.
Specific Patterns of Gene Expression Associated with MSC Phenotype
In our study, we have identified in our ES-MSC population expressed genes already known to be associated with the MSC phenotype in various types of MSCs (Supplemental Table S4).
Only a few studies have reported gene expression profiling of MSCs by comparison of hES-derived MSCs to hESs (17, 37, 45). Lian et al. (37) have identified highly expressed genes that encode for membrane proteins that could be used for the isolation of MSCs from differentiating hESs. Here, we focused on transcription regulators upregulated in ES-MSCs compared with hESs.
The most differentially expressed transcription factor of the 13 that together formed a network associated to the embryonic to MSC transition was ARID5B. Although the association of this gene with the mesodermal lineage has been shown in the mouse, its role was suggested up to now in later stages of differentiation. The knockout of Arid5b has thus been associated with a profound defect in adipogenesis in mice, suggesting an essential role for accumulation of lipid stores in postnatal life (71). A role in smooth muscle differentiation has additionally been suggested (70). In the mouse, in contrast, Arid3b is essential for mesodermal and mesenchymal differentiation (64), whereas it was strongly downregulated in ES-MSC samples (FC: −30). These results may reflect species differences. It cannot be excluded, however, that an earlier step in the embryonic to mesenchymal transition was missed during which ARID3B was transiently overexpressed. In any case, our results suggest that ARID5B is a key player in the specification of a MSC identity, before acting in differentiation pathways toward MSC derivatives.
It is interesting to mention that this conclusion of a dual role in MSC identity and in later stages of differentiation may apply to most transcription factors in the network associated to the MSC phenotype in the present study. Indeed, while ARID5B was the only transcription factor in this network with a major role in adipogenesis, others participate in essentially all potential differentiation pathways of MSCs. This includes osteogenesis, in particular with TFAP2A, TWIST1, and FOSL1. TFAP2A contributes to patterning mesenchymal cells of neural crest origin that form the craniofacial skeleton (54). Conversely, Twist proteins are “antiosteogenic” as they transiently inhibit Runx2 function during skeletal development in mice (9). Accordingly, TWIST1 mutations provoke Saethre-Chotzen syndrome, which is characterized by craniostenosis and various skeleton abnormalities. FOSL1 is necessary for bone formation (18), by acting on expression of the bone matrix genes osteocalcin, collagen 1A2, and matrix G1a protein. Also of interest was the upregulation of MYOCD, which is the only transcription factor known to be both necessary and sufficient for vascular smooth muscle cell differentiation (36), as it underlined the potential role of the ES-derived MSCs in the patterning of blood vessels early on in development. EPAS1—which is detailed below—and CITED2 appear to participate, in part associated with TFAP2A, in various aspects of morphogenesis, in particular for the heart (2, 3, 55). In addition, FLI1 (and EPAS1) is associated to the support function of MSCs for the hematopoietic system because its knockout in mice also provokes aberrant hematopoiesis and hemorrhaging (61) associated with disruption of the basement membrane and mesenchymal tissues, in which Fli1 is normally expressed.
Altogether, gene expression profiling in ES-MSCs revealed the expression of a number of transcription factors that had previously been associated with later differentiation stages leading to a diversity of MSC cell progenies. In the absence of analyses at the single-cell level, a bias in the present results cannot be excluded, originating from the combination of expression patterns from already differentiating MSCs. This, however, appears unlikely given the overall similarity of the cells' phenotype, both morphologically and in FACS analyses, and the possibility to maintain it unaltered for a large number of passages. The present results rather suggested that each of these transcription factors may contribute to a dual role by taking part in MSC identity and then by supporting differentiation into one specific derivative.
Implication of miR-148a and miR-20b in MSC Identity
In the search for additional molecular mechanisms that would control the transcription systems associated with MSC identity, miRNAs were additionally analyzed because their roles are more and more demonstrated in cell specification (22, 39, 44, 51) and, in particular, stem cell differentiation (4, 13, 34, 35, 53, 63). In hESs, lineage-specific transcription factors can indirectly determine the fate of differentiated cells by modulating the levels of lineage-specific miRNAs (10, 23, 24, 29, 38, 67). Although miRNA patterns of expression have been described in MSCs (23, 33, 53), they have, to our knowledge, been associated with further types and stages of differentiation of those cells rather than comparison to a less-differentiated stage to MSCs as in the present study. By taking the embryonic stage as a reference, specifically downregulated miRNAs were sought, the association of which with specifically upregulated transcription factors may indicate a role in the MSC identity itself.
The present analysis revealed a particular number of miRNAs associated with EPAS1 that were downregulated specifically in MSCs compared with ESs, and two of these, miR-148a and miR-20b, negatively controlled EPAS1 effects on MSC marker genes.
miR-148a had previously been associated with processes involved in the specification of hematopoietic stem cell phenotype (42) and shown to be upregulated during bone formation in association with miR-15b (46) as well as miR-20b (47).
The functional effect of miR-148a and miR-20b on the MSC specification by targeting EPAS1 has not been yet described.
EPAS1 (Endothelial PAS domain 1), also called HIF-2α (Hypoxia-inducible factor 2α), is a widely expressed transcription factor that is strongly induced by hypoxia. It has been demonstrated to play a critical role in embryonic development (1, 55). In stem cells, in which the oxygen level in the immediate environment can influence stem cell function and differentiation, EPAS1 may participate to regulate stem function through activation of Oct-4 (15).
EPAS1 appears to be instrumental in the differentiation of many MSC progenies, possibly because of a particular functional importance in the specification of the MSC identity itself. Indeed, EPAS1 has been shown to promote adipogenesis and chondrogenesis under specific conditions (31, 57). Recent studies demonstrated that EPAS1 also stimulates transcription of genes involved in the pathological transformation of osteoarthritic chondrocytes (52, 72). It may also contribute with FLI1 to support hematopoiesis in the bone marrow, because loss of Epas1 in mice provokes hypocellularity in the bone marrow (55) and is an essential regulator of murine erythropoietin production (56). At the molecular level, EPAS1 exerts a direct positive regulation on a number of genes expressed in MSCs, such as LOX, GBE1, NDRG1, PLOD2, CITED, and PKIB (69). It also interacts with genes involved in the remodeling of the extracellular matrix, PLAU and PLAUR. In our study, all these genes were found to be upregulated in MSCs compared with hESs, whereas their expression levels were similar in BM-MSCs. We have demonstrated that this regulation might be controlled by a direct interaction of miR-148a and miR-20b, alone or together, with EPAS1 by targeting its 3′-UTR and was independent of an hypoxic induction. Consequently, the forced expression of these miRNAs led to a significant decrease of the expression of some of them, such as GBE1, NDRG1, PLOD2, PLAU, and PLAUR.
In summary, our data delineate two miRNAs participating in the MSC identity. The decrease of the expression level of miR-148a and miR-20b observed in MSCs compared with hESs would result in the overexpression of one of their targets, the transcription regulator EPAS1, which allows expression of MSC genes contributing to the determination of the MSC phenotype.
This work was supported by the Association Française contre les Myopathies (AFM), by the Agence Nationale de la Recherche, and by the IngeCELL program of the cluster Medicen Paris Region. C. Rochon-Beaucourt was the recipient of a fellowship from Ministère de la Recherche et de la Technologie.
The gene expression data have been deposited in GEO Data Bank with the accession number GSE7879.
No conflicts of interest, financial or otherwise, are declared by the author(s).
The authors thank Dr. Marc Peschanski for continuous support and input during this study. We thank Drs. Karen Sermon for providing the VUB01 hES cell line, Morad Bensidhoum for the hMSCs from bone marrow, and Ileana Mateizel and Christian Jorgensen for helpful contributions and comments on the manuscript. We thank the platform of RNG, Institut Curie, Paris, France, for performing transcriptome experiments.
Present address of C. Rochon-Beaucourt: Cellectis BioResearch Parc Biocitech 102, Av. Gaston Roussel, 93235 Romainville Cedex, France.
↵1 Supplemental Material for this article is available online at the Journal website.
- Copyright © 2011 the American Physiological Society