Biological Activities Encoded by the Murine Mesenchymal Stem Cell Transcriptome Provide a Basis for Their Developmental Potential and Broad Therapeutic Efficacy



We used serial analysis of gene expression to catalog the transcriptome of murine mesenchymal stem cells (MSCs) enriched from bone marrow by immunodepletion. Interrogation of this database, results of which are delineated in the appended databases, revealed that immunodepleted murine MSCs (IDmMSCs) highly express transcripts encoding connective tissue proteins and factors modulating T-cell proliferation, inflammation, and bone turnover. Categorizing the transcriptome based on gene ontologies revealed the cells also expressed mRNAs encoding proteins that regulate mesoderm development or that are characteristic of determined mesenchymal cell lineages, thereby reflecting both their stem cell nature and differentiation potential. Additionally, IDmMSCs also expressed transcripts encoding proteins regulating angiogenesis, cell motility and communication, hematopoiesis, immunity and defense as well as neural activities. Immunostaining and fluorescence-activated cell sorting analysis revealed that expression of various regulatory proteins was restricted to distinct subpopulations of IDmMSCs. Moreover, in some cases, these proteins were absent or expressed at reduced levels in other murine MSC preparations or cell lines. Lastly, by comparing their transcriptome to that of 17 other murine cell types, we also identified 43 IDmMSC-specific transcripts, the nature of which reflects their varied functions in bone and marrow. Collectively, these results demonstrate that IDmMSC express a diverse repertoire of regulatory proteins, which likely accounts for their demonstrated efficacy in treating a wide variety of diseases. The restricted expression pattern of these proteins within populations suggests that the cellular composition of marrow stroma and its associated functions are more complex than previously envisioned.


Mesenchymal stem cells (MSCs) resident in adult bone marrow are best characterized by their capacity to differentiate into various connective tissue cell lineages [1]. MSCs and their progeny also produce a wide array of cytokines, chemokines, and adhesion molecules that regulate hematopoiesis [2, 3]. More recently, the cells have been shown to suppress T-cell immunoreactivity in response to alloantigens [47] and to benefit the treatment of neurological disorders [8], cardiac disease [9], and pulmonary fibrosis [10]. Accordingly, various clinical trials are ongoing to evaluate whether these properties can be exploited for a therapeutic intent [1115]. Despite these efforts, many basic aspects of MSC biology, as well as their mechanism of action in many disease models, remain indeterminate. MSCs are known to express the surface antigens Stro-1, NGFR, CD44, CD73, CD105, and CD106. However, expression of these antigens is not predictive of the differentiation potential of the cells [16]. Consequently, MSCs are typically enriched from bone marrow by attachment to tissue culture plastic [17], but studies have shown that these adherent populations are phenotypically and functionally heterogeneous [1820]. However, it is unclear if this heterogeneity contributes to or detracts from the potential clinical utility of MSCs.

A large number of studies aimed at evaluating the therapeutic potential of MSCs use murine models, but MSC cultures established from mouse bone marrow are replete with hematopoietic cells, which adhere to plastic, stromal cells or the matrix molecules they secrete [2, 2123] and persist in the cultures after serial passage [24, 25]. The latter is demonstrated by the fact that these plastic adherent cells afford hematopoietic reconstitution when transplanted to lethally irradiated recipients [22, 26, 27]. Several methods to fractionate plastic adherent murine marrow cultures have been described but have not gained widespread acceptance [28, 29]. Alternatively, long-term expansion of adherent populations in vitro has been purported as a novel method to isolate murine MSCs in several recent studies [3033] despite the fact that dozens of stromal cell lines have been established previously by similar means [2]. As expected, these MSC preparations share many characteristics with stromal cell lines, including aberrant expression of certain cluster designation antigens, a propensity to differentiate into adipocytes, loss of chondrogenic potential, and lack of cellular senescence. Previously, our laboratory developed a method based on immunodepletion to fractionate stromal and hematopoietic cell lineages from 6- to 8-day-old plastic adherent cultures [34, 35]. Immunodepleted murine MSCs (IDmMSCs) lack expression of hematopoietic and endothelial markers, differentiate into adipocytes, chondrocytes, and osteoblasts in vitro, and osteocytes in vivo. They also require hematopoietic cells for growth in vitro, and their capacity for multilineage mesenchymal differentiation is inhibited by fibroblast growth factor 2 (FGF2) [35]. Consequently, immunodepletion produces a cell population that, based on phenotype and function, more closely recapitulates properties of the bona fide MSC.

In this report, we used serial analysis of gene expression (SAGE) to catalog the transcriptome of IDmMSCs. Its interrogation revealed that IDmMSCs express a diverse repertoire of mRNAs, including those that reflect the developmental potential of the cells and their varied functions in bone and marrow. The latter include transcripts encoding regulatory proteins that modulate angiogenesis, bone turnover, cell motility and communication, hematopoiesis, immunity and defense as well as neural activities. Our analysis revealed that different classes of regulatory proteins are expressed within specific subpopulations of IDmMSCs. Moreover, some of these proteins are absent or expressed at reduced levels in other MSC preparations and cell lines. Collectively, these results suggest that the heterogeneity of IDmMSC populations with regard to encoded biological activities likely contributes to their demonstrated efficacy in a wide variety of disease models. Further characterization of these unique subpopulations may provide more efficacious cellular vectors tailored for treating specific diseases.

Materials and Methods

Isolation of Murine MSCs

Murine MSCs were isolated from bone marrow and purified by immunodepletion as previously described [35]. Briefly, whole bone marrow purged from the long bones of FVB/N mice was cultured in alpha minimum essential medium (Invitrogen Corporation, Carlsbad, CA, supplemented with 10% fetal calf sera (lot no. F0091; Atlanta Biologicals, Atlanta, and PenStrep (Gibco-BRL, Gaithersburg, MD, at 37°C with 5% CO2 for 72 hours, and then the nonadherent cells were removed by aspiration. Cells were cultured an additional 5–7 days with a single media change and then harvested by gentle scraping after incubation in 0.25% trypsin and 1 mM EDTA. Cells were dispersed by gentle agitation, filtered through a 70-μm filter (BD Biosciences, San Jose, CA,, suspended in Hanks' balanced salt solution (HBSS) at 40 × 106 cells per ml, incubated on a rotator for 1 hour at 4°C, and then successively for 50-minute intervals with M-280 Dynabeads (5 beads/cell) (Dynal Biotech LLC., Brown Deer, WI, conjugated to anti-CD11b, anti-CD34, and anti-CD45 antibodies (10 μg/mg beads). The immunodepleted cells were diluted in 5 ml of HBSS and counted. Total RNA was prepared from approximately 1 × 106 cells using the RNeasy Mini Kit (Qiagen, Inc., Valencia, CA, according to the manufacturer's instructions. All RNA samples were treated with DNase I during the purification procedure to ensure removal of contaminating DNA. The cell lines D1 ORL UVA (no. CRL-12424) and M2–10B4 (no. CRL-1972) were purchased from the American Type Culture Collection (Manassas, VA, and cultured according to the manufacturer's instructions. C57BL/6 and DBA1 murine MSCs were obtained from the Center for Gene Therapy at Tulane University.


SAGE was conducted using the I-SAGE kit (Invitrogen Corporation) according to the manufacturer's instructions. Concatenated ditags were cloned into pZero and transduced into electro-competent bacteria (Invitrogen Corporation) by electorporation (BTX, Holliston, MA, Bacterial colonies were screened by blue/white selection to identify those that harbored vectors containing ditags. Plasmid DNA was isolated using the Genomic DNA isolation kit (Millipore, Bed-ford, MA, and the Robosmart 384 automated workstation (MWG Biotech, High Point, NC, Plasmids were sequenced using the BigDye Terminator Cycle Sequencing Reaction kit and analyzed using a 377 ABI automated sequencer (Applied Biosystems, Foster City, CA, Sequence files were analyzed using the SAGE program group [36].

cDNA Library Construction

A cDNA library was constructed from the same RNA used for SAGE via the SMART® cDNA Library Construction Kit (Gibco-BRL). The resulting nonamplified library contained approximately 2.6 × 106 independent clones. After amplification, the library was aliquoted into pools of decreasing complexity to facilitate screening by polymerase chain reaction (PCR). Integrity of the cDNA library was measured using control reverse transcription (RT)–PCR primers for genes of varying abundance and size (Gibco-BRL). To validate SAGE tags, 50 μl of the primary pool of the amplified phage library was boiled for 5 minutes, and then aliquots (1 μl) were used as input in PCR reactions (100 μl) containing 100 pmoles of forward and reverse gene-specific primers, 1× PCR buffer, 0.2 mM dNTPs, and 0.5 U Taq polymerase (Sigma, St. Louis, After an initial denaturation step at 94°C for 3 minutes, reactions were amplified for 30 cycles at 94°C for 30 seconds, 55–65°C for 45 seconds, and 72°C for 90 seconds, followed by a final incubation at 72°C for 5 minutes. PCR product were electrophoresed through a 1% agarose gel, excised from the gel, purified using GeneElute columns (Sigma), and then cloned using the AdvanTAge PCR cloning kit (Clontech, Palo Alto, CA, Plasmid DNA was isolated and sequenced as described above to confirm the identity of each product.

Real-Time PCR

Total RNA (25 ng) was converted to cDNA and amplified by the PCR process using the TaqMan® EZ RT-PCR kit (Applied Biosystems) according to the manufacturer's instructions. Reactions were performed on a 7900 HT sequence detector (Applied Biosystems), and transcript levels were quantified using the relative Ct method by employing glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA as an internal control. The following Assay-On-Demand Taqman® probes (Applied Biosystems) were used for analysis: Dusp1, Mn00457274-gl; Irx3, Mm00500463-ml; Mef2d, Mm00504929-ml; Omd, Mm 00449589-ml; Spry4, Mm00442345-ml; Timp3, Mn00441826-ml; and Twist, Mm00442036-ml.

Flow Cytometry

Aliquots (2.5 × 105) of IDmMSCs were suspended in 50 μl of wash buffer (0.1% sodium azide, 1.0% bovine serum albumin [BSA] in phosphate-buffered saline [PBS]) containing a rat anti-mouse CD16/CD32 antibody (Fc Block; BD Pharmingen, San Diego, at a concentration of 1 μg per 1 × 106 cells and incubated for 3–5 minutes at 4°C in the dark. Wash buffer (50 μl) containing 5 μg of the appropriate primary antibody (BD Pharmingen) was added, and the cells were incubated for an additional 20 minutes. Cells were washed twice with 200 μl of wash buffer and, where necessary, were incubated for 20 minutes in wash buffer (100 μl) containing 5 μg of a fluorochrome-conjugated secondary antibody (BD Pharmingen). The extent of cell labeling was evaluated using a Beckman Coulter Model Epics XL flow cytometer (Beckman Coulter, Fullerton, CA, Isotype controls were run in parallel using the same concentration of each antibody tested.


Aliquots (2 × 104) of IDmMSCs were plated in eight-well chamber slides (BD Biosciences) and after 24–48 hours fixed with 2% paraformaldehyde/0.2% glutaraldehyde for 15 minutes at room temperature. Cells were then incubated for 10 minutes in blocking buffer (PBS containing 0.1% BSA, 5% Tween-20, and 20% of the appropriate animal sera), washed, and then incubated overnight at 4°C with a 1:100 dilution of an anti–IL-15, anti-Cyr61, or anti-MIF antibody (Santa Cruz Biotechnology, Santa Cruz, CA, Cells were then washed, incubated with the appropriate fluorescein isothiocyanate–conjugated secondary antibody for 10 minutes at room temperature, washed, and then a cover slip was applied using mounting media containing 4′,6-Diamidino-2-phenylindole (DAPI). Alternatively, fixed cells were incubated in 0.3% hydrogen peroxide for 1 minute at room temperature, incubated with a 1:100 dilution of an anti-VEGFB, anti-IRX3, anti-PTN, or anti-TWIST antibody (Santa Cruz Biotechnology) followed by a biotinylated anti-goat IgG antibody, and then staining was visualized using the Tyramide Signal Amplification kit according to the manufacturer's instructions (Invitrogen Corporation). Fluorescent micrographs were obtained using a Leica RX-DMV upright fluorescent microscope (Meyer Instruments, Inc., Houston, TX, attached to a digital camera (Cooke Sensicam High Performance; Hamamatsu Corp., Bridgewater, NJ, and rendered using Slidebook® software (Intelligent Imaging Innovations, Denver,

SAGE Database Analysis

Catalogued SAGE tags were annotated in Microsoft® Access (Microsoft, Redmond, WA, as described in the I-SAGE manual (Invitrogen Corporation). The resulting database was interrogated using a collection of macros we developed that run in Microsoft® Excel. Annotated SAGE databases were exported to Excel, and the hierarchical relationships embedded in the data were removed using the macro Telescope, which parcels each gene into a separate row of the spreadsheet and repeats its associated DNA tag and frequency as many times as there are associated genes. Telescope then collapses the tag and frequency information into a single bin. Descriptors for nonunique tags that match two genes are concatenated and placed in a single bin, and those matching three or more genes are abbreviated as “multiple match.” Once in this format, the macro Recognize Machinery was used to interrogate the database using a list based on gene ontologies. The macros Master, Align, and Set Logic were used for comparative analysis of SAGE databases. Briefly, Master generates a master list of all unique tags contained within a specified number of databases. The macro Align places tags from each library in register with those of the master list while maintaining the register between the tag and its frequency of occurrence. Finally, Set Logic determines the union or disjunction of the libraries being analyzed and compiles the resulting information on a separate spreadsheet.


Composition of the 500 Most Abundant Expressed SAGE Tags

Previously, we described a method based on immunodepletion to enrich MSCs from murine bone marrow (34, 35]. IDmMSCs lack expression of the hematopoietic and endothelial cell lineage markers CD11b, CD31, CD34, CD45, and CD117 but express markers typical of MSCs, including CD9, CD29, CD44, CD81, CD106, and Sca1. IDmMSCs also exhibit poor growth in vitro, and their capacity for multilineage differentiation is reversibly inhibited by FGF2. These latter characteristics distinguish the cells from other MSC cell lines and populations isolated by long-term propagation in vitro [3033].

To better characterize their biology, we catalogued the IDmMSC transcriptome via SAGE by sequencing 59,007 SAGE tags, 15,815 of which were unique. The most highly expressed transcript based on SAGE tag abundance was that corresponding to the β-galactoside binding lectin galectin 1 (Lgals1), which plays a role in cell adhesion, migration, and proliferation [37] as well as modulates T-cell function and apoptosis [38, 39]. Also included among the 500 most abundant SAGE tags were those that mapped to mRNAs expressed principally in connective tissues, such as secreted acidic cysteine rich glycoprotein (osteonectin), fibronectin, matrix metalloproteinase 2, matrix gamma-carboxyglutamate protein, cadherin 11, extracellular matrix protein 1, and collagens type I, III, and VI (Supplemental Database 1). In addition, tags corresponding to mRNAs encoding beta and gamma actin, as well as the actin binding proteins destrin, transgelin, transgelin 2, calponin 2, thymosin beta 10, and tropomyosin α and β, were also represented, suggesting that IDmMSCs are highly motile cells. Other abundant tags corresponded to macrophage migration inhibitory factor (MIF), interleukin-1 receptor antagonist (IL-1RN), and peroxiredoxin 1, regulatory proteins involved in immunity and defense. Lastly, the list also included Nanog, a member of the homeobox family of transcriptional regulators that is known to maintain pluripotentcy of embryonic stem cells [40]. By screening a cDNA library generated using the same RNA as for SAGE analysis, we confirmed that catalogued SAGE tags correspond to expressed transcripts. PCR products amplified from the cDNA library were cloned and sequenced to confirm their identity (Fig. 1A).

Figure Figure 1..

Validation of catalogued SAGE tags. Polymerase chain reaction products amplified using gene specific primers from an IDmMSC cDNA library were electrophoresed through a 1% agarose gel, cloned, and their corresponding sequences analyzed via BLAST to confirm their identity. (A): Expressed transcripts corresponding to the top 500 most abundant SAGE tags that map uniquely to a single mRNA. (B): Validation of a representative sampling of other expressed transcripts catalogued by SAGE. Abbreviations: Actg, actin, gamma, cytoplasmic; Agpt, angiopoietin; BLAST, Basic Local Alignment Search Tool; Ccl7, Chemokine ligand 7; Clec2d, C-type lectin domain family 2, member d; Csrp1, cysteine rich protein 1; Dsn, destrin; Dusp1, dual sensitivity phosphatase 1; Fos, FBJ osteosarcoma oncogene; Fstl, follistatin-like; Gpdh, glyceraldehyde dehydrogenase; Hdgf, hepatocyte derived growth factor; Ilk, integrin linked kinase; Il1rn, Interleukin-1 receptor antagonist; Irx3, Iroquois related homeobox 3; Jmj, Jumonji; Lgals1, galectin 1; Ly75, lymphocyte antigen 75; Mif, macrophage migration inhibitory factor; Mmp2, matrix metalloproteinase 2; Npnt, nephronectin; Nspe1, nuclease sensitive element binding protein 1; Ogn, osteoglycin; Omd, osteomodulin; Osn, osteonectin; Pfn1, profilin 1; Prdx1, peroxiredoxin 1; SAGE, serial analysis of gene expression; Six1, sine oculis-related homeobox homologue 1; Std, stanniocalcin; Tagln, transgelin; Timp3, tissue inhibitor of metalloproteinase 3; Tubb5, tubulin, beta 5; Vegfb, vascular endothelial growth factor b.

SAGE Tags Corresponding to Mesenchymal Lineage–Specific Transcripts

Further interrogation of the IDmMSC transcriptome revealed that the cells expressed mRNAs characteristic of determined mesenchymal cell lineages. For example, 2.9% of all SAGE tags catalogued corresponded to skeletal-specific transcripts and these mapped specifically to 91 separate mRNAs (Supplemental Database 1). These included various matrix molecules such as secreted acidic cysteine rich glycoprotein (osteonectin), fibronectin, matrix Gla protein, cadherin-11, collagens type I, II, V, IX, and XI, decorin, fibromodulin, neurochondrin, and osteoglycin. Also represented were several growth factors and transcriptional regulators important for bone development and function. The former included transforming growth factor β2 and β3, connective tissue growth factor, insulin-like growth factor, and bone morphogenetic proteins 1, 4, and 5. The latter included the FBJ osteosarcoma gene (Fos), which is highly expressed in embryonic and adult bone tissue, as well as Twist 2, which plays a role in mesoderm specification during development [41, 42] and also regulates osteoblast differentiation [43, 44].

In addition, approximately 1.8% of all SAGE tags corresponded to muscle-specific transcripts, and these mapped specifically to 78 mRNAs encoding contractile proteins, calcium binding proteins, calcium transporting ATPases, voltage-sensitive calcium channels, and skeletal-specific, smooth muscle–specific, and cardiac muscle–specific isoforms of myosin. Other catalogued transcripts encoded the muscle-specific proteins desmin, smoothelin, myocyte enhancer factor 2D, acidic syntrophin 1, titincap, and troponin 1, T1, and T3. Lastly, we also identified 76 transcripts mapping to unique tags encoding various cytokines, chemokines, and adhesion molecules that regulate aspects of hematopoiesis and 22 transcripts that regulate adipocyte differentiation or function. The presence of mesenchymal lineage–specific transcripts within the IDmMSC transcriptome reflects their differentiation potential as defined by functional assays. The diversity and abundance of such transcripts further suggest that mesengenesis is an active process in adherent cultures even in the absence of exogenous agents that promote cellular differentiation.

Categorization of the IDmMSC Transcriptome Based on Gene Ontologies

Using gene ontologies defined by the Celera® Discovery System Panther classifications [45], we also determined the percentage of SAGE tags that mapped to transcripts categorized according to their molecular function or biological process (Fig. 2). We then ranked those transcripts that mapped uniquely to a given tag in order of their relative abundance (Supplemental Database 2 and 3). This analysis catalogued numerous mRNAs not previously reported in MSCs, providing novel insight into their biology. For example, categorizing the transcriptome based on molecular function revealed drebrin as one of the most abundant cytoskeletal proteins. Drebrins are actin-binding proteins that play a role in the formation of cell processes, including neurites, dendritic spines, lamellipodia, filipodia, and foot processes in podocytes [46]. Consequently, expression of drebrin may account for the ability of stromal subtypes to reticulate cell processes deep into the hematopoietic cords in bone marrow. SAGE tags mapping to tubulin β5 were also identified by this analysis, consistent with the fact that expression of this protein is restricted to tissues active in hematopoiesis [47]. In addition, dual specificity phosphatase 1 (Dusp1) and AXL receptor tyrosine kinase (Axl) were among the most abundant phosphatases and kinases, respectively. Dusp1 specifically inactivates mitogen-activated protein kinase, which has been shown to regulate osteogenic and adipogenic differentiation of human MSCs [48]. Axl has been shown to mediate the osteogenic differentiation of pericytes [49] and play an important role in the control of chondrocyte growth and survival [50]. Furthermore, the transcription factors nuclear receptor coactivator 3 (Ncoa3) and nuclear receptor-binding SET-domain protein 1 (Nsd1) were also highly expressed. Ncoa3 is a cofactor for MyoD-stimulating transcription during muscle differentiation [51] and possesses histone acetyltransferase activity [52]. Alternatively, mutations in Nsd1 results in Sotos syndrome, a childhood overgrowth syndrome characterized by distinctive craniofacial features, developmental delay, and advanced bone age [53]. In contrast, expression of the transcription factor Runx2 (Cbfa1), which promotes osteoblast maturation [54], ranked 53rd in abundance out of the 261 transcription factors catalogued. IDmMSCs also expressed low levels of nuclease sensitive element binding protein 1 (Nsep1), a negative regulator of HLA class II gene expression [55].

Figure Figure 2..

Categorization of the IDmMSC transcriptome via gene ontologies. Pie chart illustrating the percentage of SAGE tags that map uniquely to specific transcripts classified according to their (A) molecular function or (B) biological processes based on the Celera Panther Gene Ontology classification system. Abbreviations: IDmMSC, immunodepleted murine mesenchymal stem cell; SAGE, serial analysis of gene expression.

Categorizing transcripts based on their biological process revealed that the majority of SAGE tags mapped to mRNAs were involved in protein modification (12.9%), followed by signal transduction (8.91%), immunity and defense (7.8%), cell adhesion (6.5%), nucleic acid metabolism (5.86%), and cell structure and motility (5.76%). The transcriptome also contained numerous transcripts encoding proteins that regulate neural activities, including various axon guidance and neural cell adhesion molecules, neurite-inducing factors, neurotransmitter receptors, neurotrophins, and a large number of proteins involved in vesicle transport, synaptogenesis, and synaptic transmission (Supplemental Database 1). Other transcripts encoding proteins common to neuronal cells or that regulate nervous system development were also identified, including the transcription factor neurogenic differentiation 1 (Neurod1), the neurofilament proteins Nef1 and Nef3, and the growth suppressor necdin, which is expressed in all postmitotic neurons in the brain [56]. The IDmMSC transcriptome also contained mRNAs encoding proteins with proangiogenic activity, including vascular endothelial growth factor B, cysteine rich protein 61 (Cyr61), and connective tissue growth factor, as well as factors affecting endothelial cell growth and migration, such as angio-associated migratory cell protein, angiopoietin, and hepatoma-derived growth factor (data not shown). Approximately 3.24% of the SAGE tags also mapped specifically to transcripts that function in developmental processes, and 40% of these tags encoded proteins that play a role in the specification and differentiation of mesoderm (Fig. 3 and Supplemental Database 1).

Figure Figure 3..

Number of transcripts expressed by IDmMSCs that regulate developmental processes. The pie chart illustrates the total number of transcripts that mapped uniquely to a given SAGE tag(s) and that function in specific developmental processes. Abbreviations: IDmMSC, immunodepleted murine mesenchymal stem cell; SAGE, serial analysis of gene expression.

Restricted Expression of Regulatory Proteins in IDmMSC Populations

The diverse repertoire of mRNAs expressed by IDmMSCs implicates the cells in a wide array of biological activities important for the function of bone and marrow. To validate these results, we analyzed the expression profile of different classes of regulatory proteins in IDmMSC populations. Immunofluorescence staining confirmed that the cells express the cytokine interleukin-15 (IL-15), the angiogenic factors cysteine rich protein 61 [CYR61) and vascular endothelial growth factor B [VEGFB), the transcriptional regulator iroquois-related homeobox 3 (IRX3), and TWIST, as well as the multifunctional proteins pleiotrophin (PTN) and MIF. However, expression of these proteins was restricted to specific subpopulations of cells and the number of cells expressing each protein appeared highly variable. For example, whereas IL-15 and IRX3 were detected in approximately half of the population, only a small percentage of cells expressed CYR61, VEGFB, PTN, and TWIST. Furthermore, the subcellular location of IRX3 and TWIST varied between cells. IRX3 was localized exclusively in the cytoplasm of some cells (Fig. 4E) but was confined to the nucleus or evident in both compartments in other cells (Fig. 4F). TWIST showed a similar expression pattern and appeared to be highly expressed in the nucleus of a small number of cells (Fig. 4O). Fluorescence-activated cell sorting (FACS) analysis confirmed that the number of cells expressing these different regulatory proteins was highly variable within IDmMSC populations (Fig. 5).

Figure Figure 4..

Immunohistochemical localization of specific regulatory proteins to subpopulations of immunodepleted murine mesenchymal stem cells (IDmMSCs). IDmMSCs were stained with fluorescein isothiocyanate (FITC)–or Cy3-conjugated antibodies specific for (A–C) interleukin-15 (IL15), (D–F) iroquois related homeobox 3 (IRX3), (G–I) cysteine rich protein 61 (CYR61), (J–L) vascular endothelial growth factor B (VEGFB), (M) pleiotrophin (PTN), (N) macrophage migration inhibitory protein (MIF), and (O) TWIST and counterstained with 4′,6-Diamidino-2-phenylindole (DAPI). Arrows indicate localized expression of IRX3 (F) and TWIST (O) within the nucleus.

Figure Figure 5..

Percentage of cells within different murine mesenchymal stem cell populations that express different classes of regulatory proteins. Histograms showing the percentage of immunodepleted murine mesenchymal stem cells (IDmMSCs) and C57BL/6, M210B4, and CRL-12424 cells that reacted with antibodies against interleukin-15 (IL15), cysteine rich protein 61 (CYR61), macrophage migration inhibitory protein (MIF), interleukin-1 receptor antagonist (IL1RN), and pleiotrophin (PTN).

Due to the fact that IDmMSCs undergo limited expansion in vitro prior to immunodepletion, we also anticipated that the cells may express a broader repertoire or a greater abundance of transcripts as compared with other murine MSC populations. Therefore, we compared mRNA and protein levels of different classes of regulatory proteins in IDmMSCs with those of the MSC cell lines D1 ORL UVA and M2–10B4, as well as MSC populations isolated from C57Bl/6 and DBA1 mice by long-term propagation in vitro [32]. As shown in Figure 5, approximately 66% of IDmMSCs expressed IL-15 but only 3.3% and 2.8% expressed PTN and CYR61, respectively, consistent with our immunostaining results. Approximately 21.4% and 24% of cells expressed MIF and IL-1RN, respectively, as well. Similarly, a large percentage cells in the other MSC populations expressed IL-15. However, the number of cells expressing MIF or IL-1RN was greatly diminished in C57BL/6 and D1 ORL UVA populations as compared with IDmMSCs. Furthermore, these proteins were not detected in M2–10B4 populations. A slighter higher proportion of IDmMSCs also expressed CYR61 and PTN as compared with other populations. Similarly, real-time PCR revealed that the transcription factors Irx3 and Twist and the signaling proteins Dusp1 and Sprouty 4 (Spry4) were expressed at significantly higher levels in IDmMSCs as compared with other murine MSC populations (Fig. 6). Specifically, Twist mRNA levels were 3-fold, 19-fold, 11-fold, and 7-fold, and Spry4 levels were 58.8-fold, 9.0-fold, 9.3-fold, and 21.7-fold higher in IDmMSCs as compared with C57BL/6, M2–10B4, D1 ORL UVA, and DBA1 populations, respectively. In contrast, M2–10B4 cells expressed equivalent levels of Timp3 mRNA as IDmMSCs, and C57Bl/6 cells expressed the highest levels of osteomodulin, a BMP2-inducible extracellular skeletal protein [57].

Figure Figure 6..

mRNA levels of different regulatory factors were higher in immunodepleted murine mesenchymal stem cells (IDmMSCs) as compared with other murine mesenchymal stem cell preparations. Total RNA prepared from IDmMSCs, C57BL/6, and DBA1 cells and the M210B4 and CRL-12424 cell lines was analyzed by real-time polymerase chain reaction to quantify expression levels of dual specificity phosphatase 1 (Dusp1), iroquois related homeobox 3 (Irx3), osteomodulin (Omd), sprouty 4 (Spry4), tissue inhibitor of metalloproteinase 3 (Timp3), and Twist 1. Transcript abundance was determined using the relative Ct method and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as an internal control. All samples were run in triplicate, and the mean ± SD was plotted.

Molecular Fingerprint of IDmMSCs via Comparative Genomics

To determine a molecular fingerprint for IDmMSCs, we compared their transcriptome to that of 17 different murine cell types and tissues for which SAGE databases are available in the public domain (Table 1). The 18 SAGE databases contained a total of 668,445 tags that occurred with a frequency of two or greater, and 37,479 of these were unique. A total of 575 tags were found exclusively in the IDmMSC database. Strikingly, 90 (15.6%) tags corresponded to RIKEN cDNAs, 49 (8.5%) failed to match any known gene sequences, 45 (7.8%) corresponded to uncharacterized expressed sequences, and 7 (1.2%) encoded hypothetical proteins. Therefore, approximately 33% of all SAGE tags unique to IDmMSCs as defined by the criteria outlined above encoded proteins of indeterminate function. Gene annotations corresponding to transcripts mapping uniquely to the remaining tags were filtered against all transcriptomes to remove redundancies, revealing 43 transcripts specific to IDmMSCs (Supplemental Database 1). These included the skeletal proteins osteoglycin, osteomodulin, and parathyroid hormone receptor 1, as well as chemokine ligand 7 (Ccl7) and lymphocyte antigen 75 (Ly75). Ccl7 mediates macrophage recruitment during inflammation but, after cleavage by gelatinase A, functions as a general chemokine antagonist that dampens inflammation [58]. Ly75 is involved in antigen presentation [59] but also has antiproliferative activity against B lymphocytes [60]. Other IDmMSC-specific transcripts included various developmentally regulated genes, such as Iroquois related homeobox 3 (Irx3), sine oculis-related homeobox 1 (Six1) and 4 (Six4), as well as filamin b (Flnb). Irx3 is involved in regionalization of the otic vesicle, bronchial epithelium, and limbs [61], and Six1 and Six4 are also expressed in limb buds during development [62]. Flnb plays a role in vertebral segmentation, joint formation, and endochondral ossification [63]. Also represented was the multifunctional protein stanniocalcin, which stimulates osteoblast differentiation [64], inhibits macrophage chemotaxis and chemokinesis [65], and modulates angiogenesis [66]. Consequently, the repertoire of IDmMSC-specific transcripts appears to reflect the unique functions associated with MSCs and/or their progeny, including skeletogenesis and immunity and defense.


Our SAGE analysis is consistent with other gene profiling studies indicating that MSCs are characterized by an abundance of mRNAs encoding various connective tissue proteins [6769]. Our studies also revealed that IDmMSCs express high levels of mRNAs encoding several potent regulatory factors. For example, the most abundant catalogued SAGE tag corresponded to Lgals1, a glycoprotein known to bind to CD45 [70] and function as a growth inhibitor [71]. A defining characteristic of IDmMSCs is their lack of growth in vitro. One explanation for this is that removal of CD45-expressing hematopoietic cells by immunodepletion increases the amount of galectin 1 bound to IDmMSCs via engagement of the fibronectin receptor [72], thereby inhibiting their proliferation. Accordingly, galectin 1 may function as a sensor to coordinate hematopoiesis with expansion/contraction of stromal elements in marrow. IDmMSCs also expressed high levels of IL-1RN, a natural product thought to counteract the affects of the proinflammatory cytokine IL-1. Production of this protein may explain the ability of IDmMSCs to ameliorate the effects of bleomycin-induced lung injury in mice [10]. IL-1 is also a potent stimulator of bone resorption. Therefore, expression of IL-1RN together with Clec2d, a C-type lectin that inhibits multinucleate osteoclast formation, may enable IDmMSCs to regulate bone turnover, as well.

Our SAGE analysis also provides a framework to reconcile the molecular and functional heterogeneity of MSC populations (Fig. 7). For example, previous reports have shown that MSCs enriched from marrow by plastic adherence represent a collection of cells with varying potentials that follow a deterministic differentiation program in vitro [20]. Consistent with these findings, the IDmMSC transcriptome was replete with mesenchymal lineage–specific transcripts, suggesting the population was comprised of cells at varying stages of differentiation. In addition, the identification of developmentally regulated mRNAs known to affect mesoderm specification and differentiation implies that the population also contained primitive progenitors and/or stem cells. Accordingly, the IDmMSC transcriptome reflects a hierarchy of cellular differentiation similar to the mesengenic process proposed by Caplan [73] and experimentally demonstrated by various groups [1, 18, 20].

Figure Figure 7..

The IDmMSC transcriptome reflects the plasticity of MSCs as well as the varied functions they and their progeny perform in bone marrow. Schematic summarizes the nature of catalogued transcripts in the IDmMSC transcriptome. These include transcripts that regulate mesoderm specification and development or that are characteristic of determined mesenchymal cell lineages, which reflect the developmental potential of MSCs. Also identified were transcripts that participate in angiogenesis, cell motility and communication, hematopoiesis, immunity and defense, as well as neural activities, which reflect the varied biological activities that MSCs and progeny participate in within bone and marrow stroma. Abbreviations: IDmMSC, immunodepleted murine mesenchymal stem cell; MSC, mesenchymal stem cell; SAGE, serial analysis of gene expression.

Additionally, identification of transcripts regulating angiogenesis, hematopoiesis, cell motility and communication, immunity and defense, as well as neural activities reflects the characteristics and/or associated functions of stromal cell subtypes known to exist in marrow. For example, a number of specialized cell types in stroma are highly motile, including reticular cells and adventitial reticular (AR) cells. Reticular cells are so named because they extend or reticulate long, cytoplasmic processes into the hematopoietic cords, and AR cells mediate egress of hematopoietic cells into the vasculature by contracting or physically migrating to and from the sinus wall. Furthermore, motility of these cells is coordinated by communication via gap junction formation with periarterial adventitial (PAA) and hematopoietic cells [74]. Osteoblasts also use gap junctions to establish communication networks to coordinate the remodeling of bone tissue [75].

Bone and marrow are also innervated by nervous tissue. Efferent nerve terminals from fibers that track into the hematopoietic cords terminate directly onto AR cells. Because the PAA and AR cells are connected by gap junctions, they are indirectly coupled to each other forming a circuit termed the “neuroreticular complex,” providing a means by which nervous input can alter stromal function to regulate hematopoiesis [76]. This close association with nervous tissue provides a rationale for stromal elements to express neuroregulatory factors. Some of these factors may guide nerve innervation into bone and marrow during growth, remodeling, and repair. Others are known to affect the activity of both neurons and hematopoietic cells. For example, neuropilin 1 functions as an axon guidance molecule but was recently shown to mediate interaction between hematopoietic cells and the stromal cell line MS-5 [77]. Pleiotrophin, a neurite-inducing factor, stimulates osteoprogenitor chemotaxis, proliferation, and differentiation [78], and the neurotrophin NGF induces proliferation of hematopoietic precursors [79]. Interestingly, vascular endothelial growth factor (VEGF) and angiopoietin 1 (Ang1) have been shown to mobilize hematopoietic stem cells from marrow and induce their proliferation, respectively [80, 81]. Co-operatively, these factors induce capillary proliferation and expansion of the sinusoidal space. Expression of these and other angiogenic factors likely modulates vessel growth and remodeling as well, processes essential for bone growth.

The large number of expressed transcripts encoding unique regulatory factors revealed by our SAGE analysis greatly expands the repertoire of biological activities that can be ascribed to stromal elements within marrow. This diversity of function is also reflected by the cohort of transcripts deemed unique to IDmMSCs, which included several small leucine-rich proteins unique to skeletal tissue and proteins that affect immune cell activity, proliferation, and chemotaxis, as well as angiogenesis. IDmMSC-specific transcripts encoding homeotic genes important in limb morphogenesis further substantiate the existence of early mesodermal progenitors or stem cells within plastic adherent populations. Comparative studies suggest that expression levels of specific regulatory proteins may be highly variable between preparations of murine MSCs. IDmMSCs express higher levels of these factors than other MSC populations, likely because long-term propagation of cells in vitro selects against specific subpopulations. Differences in the expression profile of regulatory proteins between murine MSC populations may produce disparate outcomes in both in vitro and in vivo experiments.


Our comprehensive SAGE analyses are the first to provide a molecular basis for the unique biological characteristics of murine MSC populations enriched from marrow by plastic adherence. Furthermore, the finding that IDmMSCs express a diverse array of regulatory proteins, expression of which is restricted to specific subpopulations of cells, indicates that the cellular complexity and biological functions of marrow stroma are more diverse than previously envisioned. Based on these results, we propose that the heterogeneity of IDmMSC populations likely attributes to their demonstrated therapeutic efficacy in a wide variety of disease models. Exploiting the unique characteristics of these subpopulations may provide more efficacious cellular vectors tailored for treating specific diseases.

Table Table 1.. List of transcriptomes used for comparative analysis to generate a genetic footprint for immunodepleted murine mesenchymal stem cells
  1. a

    Abbreviation: ESC, embryonic stem cell.

Library IDCell type/TissueTotal tags (2)
GSM3830Dendritic cells derived from the ESC line ESF 11623,604
GSM3833Dendritic cells from mouse bone marrow16,251
GSM3829ESC line ESF 11610,562
GSM3683CD4+, CD25 spleen cells14,379
GSM3686CD4+, CD25 spleen cells9,400
GSM767Granular cell precursor cells43,255
GSM55E11.5 developing forelimb (B6CH3 strain)52,552
GSM56E11.5 developing hind limb (B6CH3 strain)53,349
GSM1618Heart from 3-month-old C57BL/6 mice66,755
GSM1682P19 embryonal carcinoma cells48,557
GSM580R1 ESCs106,721
GSM5435Adult testis somatic cells (no germ cells)66,165
GSM4871Whole brain of 8- to 12-week-old C57BL/616,883
GSM3825T-regulatory cell clone after CD3 stimulation10,408
GSM3837Mouse splenic B cells (T cell depleted)5,688
GSM787Primary granular cell precursor cells71,152
GSM2415Mouse cerebellum from P23 C57BL/612,523


This research was supported in part by a grant from the National Institutes of Health to D.G.P. (R01-AR44210-01A1), the Louisiana Gene Therapy Research Consortium (New Orleans), and HCA-the Health Care Company (Nashville, TN). The authors would like to thank Peggy Wolfe for providing the C57BL/6 and DBA1 murine MSCs.


The authors indicate no potential conflicts of interest.