PPARγ2 Regulates a Molecular Signature of Marrow Mesenchymal Stem Cells

Bone formation and hematopoiesis are anatomically juxtaposed and share common regulatory mechanisms. Bone marrow mesenchymal stromal/stem cells (MSC) contain a compartment that provides progeny with bone forming osteoblasts and fat laden adipocytes as well as fibroblasts, chondrocytes, and muscle cells. In addition, marrow MSC provide an environment for support of hematopoiesis, including the development of bone resorbing osteoclasts. The PPARγ2 nuclear receptor is an adipocyte-specific transcription factor that controls marrow MSC lineage allocation toward adipocytes and osteoblasts. Increased expression of PPARγ2 with aging correlates with changes in the MSC status in respect to both their intrinsic differentiation potential and production of signaling molecules that contribute to the formation of a specific marrow micro-environment. Here, we investigated the effect of PPARγ2 on MSC molecular signature in respect to the expression of gene markers associated exclusively with stem cell phenotype, as well as genes involved in the formation of a stem cell supporting marrow environment. We found that PPARγ2 is a powerful modulator of stem cell-related gene expression. In general, PPARγ2 affects the expression of genes specific for the maintenance of stem cell phenotype, including LIF, LIF receptor, Kit ligand, SDF-1, Rex-1/Zfp42, and Oct-4. Moreover, the antidiabetic PPARγ agonist TZD rosiglitazone specifically affects the expression of “stemness” genes, including ABCG2, Egfr, and CD44. Our data indicate that aging and anti-diabetic TZD therapy may affect mesenchymal stem cell phenotype through modulation of PPARγ2 activity. These observations may have important therapeutic consequences and indicate a need for more detailed studies of PPARγ2 role in stem cell biology.


INTRODUCTION
PPARγ, an essential regulator of lipid, glucose, and insulin metabolism [1], is expressed in bone marrow mesenchymal stem cells (MSC). PPARγ is expressed in mice and humans in two isoforms, PPARγ1 and PPARγ2, which originate from up to seven different transcripts due to alternative promoter usage and alternative splicing [2][3][4][5]. PPARγ2 differs from PPARγ1 by 30 additional amino acids on its N-terminus, which constitute AF-1 domain of ligand-independent geneactivating function [6]. While PPARγ1 is expressed in a variety of cell types, including osteoblasts, PPARγ2 is expressed in cells of adipocyte lineage and serves as an essential regulator of adipocyte differentiation and function [7,8].
Osteoblasts and adipocytes are derived from a marrow mesenchymal cell compartment which also serves as a source of progenitors for marrow fibroblasts and cartilage cells and functions as hematopoiesis-supporting stroma [9,10]. Commitment of marrow MSC toward adipocyte and osteoblast lineage occurs by a stochastic mechanism, in which lineagespecific transcription factors (such as Runx2 for osteoblasts and PPARγ2 for adipocytes) representing intrinsic determinants of this process are activated [8,11]. Embryonic stem cells with a null mutation in PPARγ spontaneously differentiate to osteoblasts and are unable to differentiate to adipocytes [12]. In marrow MSC, PPARγ2 acts as a dominant negative regulator of osteoblast differentiation [8,13]. Using a model of marrow MSC differentiation (U-33/γ2 cells), we have previously demonstrated that activation of the PPARγ2 isoform by the highly specific agonist and antidiabetic thiazolidinedione (TZD), rosiglitazone, converted cells of osteoblast lineage to terminally differentiated adipocytes 2 PPAR Research Regulation of progression through cell cycle, angiogenesis, development, cell proliferation, cell differentiation a Affymetrix probe ID b fold change c gene ontology [28] and irreversibly suppressed both the osteoblast phenotype and the osteoblast-specific gene expression [8]. The expression of PPARγ2 in marrow MSC increases with aging [14]. Moreover, bone marrow derived from old animals produces unknown PPARγ activator(s) that stimulates adipocyte differentiation and suppresses osteoblast differentiation [14]. These changes cause alterations in the milieu of intrinsic and extrinsic signals that determine MSC lineage allocation. For instance, this contributes to the preferential MSC differentia-tion toward adipocytes and decreased differentiation toward osteoblasts that leads to the development of senile osteopenia.
PPARγ plays an important role in the maintenance of bone homeostasis as demonstrated in several animal models of either bone accrual or bone loss depending on the status of PPARγ activity [12,[15][16][17][18][19]. A decrease in PPARγ activity resulted in increased bone mass due to increased osteoblast number [12,18], whereas increased PPARγ activity   [28] due to TZD administration led to the bone loss [15][16][17]19]. TZD-induced bone loss was accompanied with changes in the cellular composition of the bone marrow, such as decreased numbers of osteoblasts and increased numbers of adipocytes, and changes in the MSC phenotype characterized by a loss of MSC plasticity. These changes are characteristics for aging bone marrow [20]. Recently, several human studies have demonstrated that TZD use is associated with decreased bone mineral density and an increased risk of fractures in postmenopausal diabetic women [21][22][23]. This prompted US Food and Drug Administration to issue a warning of possible adverse effects of TZD on human bone. The development of high throughput analysis of gene expression using microarrays has advanced studies on genes and signaling pathways controlled by a single gene product. The transcriptional role of PPARγ in either differentiated cells or functional tissues has been studied using DNA microarrays, mostly to determine its role in the physiology during disease and as a result of therapeutic treatment with TZDs of these target tissues [24][25][26]. None of these studies, however, were designed to test for the effect of the PPARγ2 isoform on the molecular signature of MSC. Using a model of marrow MSC differentiation under the control of the PPARγ2 transcription factor, we found that both the presence of PPARγ2 and its activation with the antidiabetic TZD, rosiglitazone, resulted in gene expression changes for multiple genes that characterize the stem cell phenotype and their phenotypic lineages. Even though our model was originally developed to study the mechanisms by which PPARγ2 suppressed osteoblastogenesis and promoted adipogenesis, our studies suggest that PPARγ2 has a profound effect on the expression of signature genes for cell "stemness."

Cell cultures and RNA isolation
Murine marrow-derived U-33 (previously referred to as UAMS-33) cells represent a clonal cell line spontaneously immortalized in the long term bone marrow culture conditions. To study the effect of PPARγ2 on marrow mesenchymal stem cell differentiation, U-33 cells were stably transfected with either PPARγ2 expression construct (referred to as U-33/γ2 cells) or an empty vector control (referred to as U-33/c cells) as described previously [8]. Several independent clones were retrieved after transfection and carefully analyzed for their phenotype. Clone 28.6, representing U-33/γ2 cells, and clone γc2, representing U-33/c cells, were used in the experiments presented in this manuscript. Cells were maintained in αMEM supplemented with 10% FBS heat-inactivated (Hyclone, Logan, UT), 0.5 mg/ml G418 for positive selection of transfected cells, 100 U/ml penicillin, 100 μg/ml streptomycin, and 0.25 μg/ml amphotericin (sigma) at 37 • C in a humidified atmosphere containing 5% CO 2 . Media and additives were purchased from Life Technologies (Gaithersburg, MD).
Cells were propagated for one passage and than seeded at the density of 3 × 10 5 cells/cm 2 . After 48 hours of growth, when cultures achieved approximately 80% confluency, cells were treated with either 1 μM rosiglitazone or the same volume of vehicle (DMSO) for 2, 24, and 72 hours, followed by RNA isolation using RNeasy kit (QIAGEN Inc., Valencia, CA). The replicate experiment was performed independently on a fresh batch of cells. Two replicates were used for microarray analysis. The factorial design of experiment was 2×3×2 which corresponded to two cell lines (with and without PPARγ2), three time points (2, 24, 72 hours), and two treatment regiments (rosiglitazone and vehicle).

Microarray experiments
RNA quality was assessed using the Agilent Model 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). Five micrograms of total RNA were processed for use on the microarray by using the Affymetrix GeneChip one-cycle target labeling kit (Affymetrix, Inc., Santa Clara, CA) according to the manufacturer's recommended protocols. The resultant biotinylated cRNA was fragmented then hybridized to the GeneChip Mouse Genome 430 2.0 Array (45,000 probe sets used to analyze over 39,000 mouse transcripts and variants from over 34,000 well-characterized mouse genes; Affymetrix, Inc.). The arrays were washed, stained,  Regulation of progression through cell cycle, angiogenesis, positive regulation of neuroblast proliferation, development, positive regulation of cell proliferation, organ morphogenesis a Affymetrix probe ID b fold change c gene ontology [28] and scanned using the Affymetrix Model 450 Fluidics Station and Affymetrix Model 3000 scanner using the manufacturer's recommended protocols by the University of Iowa DNA Core Facility. Raw gene expression measurements were generated using the microarray suite (MAS) version 5.0 software (Affymetrix, Inc.). Statistical assessment of differential gene expression is described in Lecka-Czernik et al. [27].

RESULTS AND DISCUSSION
An essential role of PPARγ2 in the regulation of marrow MSC lineage allocation, together with the evidence of its increased activity in MSC with aging [14], prompted us to study the effect of PPARγ2 on the expression of stem cell gene markers. Two aspects were examined: the effect of the presence of PPARγ2 in U-33 stem cells and the effect of PPARγ2 activation with rosiglitazone on stem cell phenotype.
Here we used a model of marrow MSC differentiation under the exclusive control of a single protein, PPARγ2. This system allows for relatively unambiguous studies of the unique effects of PPARγ2 isoform on MSC phenotype. The model of PPARγ2-dependent MSC differentiation consists of two cell lines derived from the same parental cell line (U-33 cells), which either express the PPARγ2 protein (U-33/γ2 cells) or do not express the PPARγ2 protein (U-33/c cells) [8,29]. To assess the effects of the presence of PPARγ2 on the phenotype of U-33 cells in nontreated conditions, we compared gene expression in U-33/γ2 and U-33/c cells maintained in basal growth conditions (this is referred to as the "P versus V" analysis). This comparison provides information about PPARγ2 activities, which are either ligand independent or acquired as a result of activation with natural ligands present in the growth media or endogenously produced by tested cells. The results of "P versus V" analysis may provide information on a role of PPARγ2 in a continuum of changes that occur in stem cells during aging. To assess an effect of rosiglitazone on the expression of stem cell-related genes, we compared gene expression in U-33/γ2 cells treated with rosiglitazone and nontreated U-33/γ2 cells (this is referred to as the "PR versus P" analysis). This analysis provides important information on the effects of rosiglitazone on the stem cell phenotype. Finally, comparison of the results of both analyzes provides information on differences between endogenous and artificially induced PPARγ2 activities in respect to stem cell gene expression.
To avoid differences in the cell phenotype due to different rates of cell growth, we chose the 72-hour time point for the analysis of gene expression (see Section 2). In basal growth conditions at this time point, cell cultures of U-33/γ2 and U-33/c were in state of confluence, cells acquired fibroblastlike appearance and cell cultures were indistinguishable morphologically from each other. In contrast, U-33/γ2 cells There are no known exclusive markers for MSC. However, based on extensive work with MSCs and other stem cell populations, several proteins have emerged as candidate markers associated with a stem cell phenotype. These entities include ATP-binding cassette g2 (Abcg2), cell surface antigen CD44, stem cell factor or kit ligand (SCF/Kitl), epidermal growth factor receptor (Egfr), early growth response factor 2 (Egr2), leukemia inhibitory factor (Lif), leukemia inhibitory factor receptor (Lifr), and stromal-derived factor/CXCchemokine ligand 12 (SDF-1/CXCL12). Based on the available published information for stem cell gene expression for the analysis, we arbitrarily chose 135 genes that represent markers of either early or lineage committed stem cells [9,[30][31][32][33][34]. The analysis showed that the expression of 38% of analyzed genes was not affected by activation state of PPARγ2 (see Table 4), the expression of 28% genes was exclusively affected by the presence of PPARγ2 ("P versus V" analysis) (see Table 1(a)), and the expression of 10% genes was exclusively affected by rosiglitazone-activated PPARγ2 ("PR versus P" analysis) (see Table 1(b)). The genes whose expression was affected by both rosiglitazone-activated and nonactivated PPARγ2 constituted 24% of the total genes studied; their expression was affected in equal proportion either similarly (see Table 2) or in the opposite direction in these two conditions (see Table 3).
Comparison of the two cell lines indicates that a majority of analyzed genes are up-regulated in U-33/γ2 versus U-33/c cells (see Tables 1(a) and 3). Most of these genes are characteristic for stem cells of hematopoietic and neural lineages while some of them are expected to be up regulated in hematopoiesis supporting stromal cells (e.g. Kitl, RANKL (Table 1(a)), and the CXCL family (Tables 1(a) and 3)).
These interesting observations have at least two reasonable interpretations. The first interpretation suggests that observed differences are a reflection of different phenotypes of the two individual parental cells from which each of the two clones originated. Hence, differences in gene expression between both cell lines are PPARγ2-independent. The second possibility suggests that these differences are PPARγ2-dependent and result from either PPARγ2 ligandindependent activity or activity acquired from endogenous ligand. Several lines of evidence suggest a correlation between the adipocyte-like phenotype of marrow stroma cells and support for hematopoiesis [35,36]. Hematopoiesis depends heavily on the microenvironment provided by mesenchymal cell compartment in the marrow and the ability of these cells to produce growth factors and cytokines that act in a paracrine fashion to influence the differentiation of hematopoietic progenitors. In the long term bone marrow cultures, an in vitro system of hematopoietic cell differentiation, stroma cell support for myelopoiesis, is provided by cultures consisting mostly of adipocytes [35,37]. Similarly, in vivo studies in a model of SAMP6 mice that are characterized by senile osteopenia due to a diminished number of osteoblasts and increased myelopoiesis, correlates positively with an increased number of marrow adipocytes [38]. Interestingly, U-33/γ2 cells support osteoclastogenesis much better than U-33/c cells (unpublished observation), in part due to relatively higher RANKL (9-fold in "P versus V," Table 3) and lower OPG (−34.6-fold in "P versus V"; Table 1(a)) expression. Another important regulator of bone marrow hematopoiesis, including osteoclastogenesis, is represented by the chemokine CXCL12 or SDF-1 [39,40]. Growing experimental evidence indicates that CXCL12 and its receptor CXCR4 axis is not only required for hematopoietic stem cell signaling but also has a crucial role in the formation of multiple organ systems during embryogenesis as well as adult nonhematopoietic tissue regeneration and tumorigenesis [39]. According to our analysis, an expression of CXCL12, but not CXCR4, is up regulated in U-33/γ2 cells ("P versus V") and suppressed by PPARγ2-activated with rosiglitazone ("PR versus P") (see Table 3). Thus, it is conceivable that mesenchymal cells which express PPARγ2 acquire the adipocyte-like phenotype typified by the production of number of cytokines and support hematopoietic stem cell differentiation.
While PPARγ2 has a positive effect on the stromal phenotype supporting hematopoiesis, it has a negative effect on the expression of "stemness" genes. The expression of LIF cytokine and its receptor, a regulatory system required for the stem cell self renewal, is significantly suppressed in U-33/γ2 cells as compared to U-33/c cells (see Table 1(a)). Interestingly, activation of PPARγ2 with rosiglitazone did not affect the expression of these genes. The presence of PPARγ2 in U-33/γ2 cells suppresses the expression of Egr2/Krox20, a stem cell-specific transcription factor with a role in the development of nervous system and endochondrial bone formation [41]. Egr2/Krox20 also regulates osteoblast differentiation and osteocalcin expression [42]. Again, rosiglitazone does not affect Egr2/Krox20 gene expression (see Table 1(a)). PPARγ2 cellular presence also affects expression of Zfp42 transcription factor, which is a marker of human and murine embryonic stem (ES) cells. Expression of Zfp42 is down regulated during ES cell differentiation [43]. An artificial knockdown of Zfp42 with RNAi resulted in spontaneous differentiation of ES cells toward endoderm and mesoderm lineages, whereas its overexpression led to the loss of self-renewal capacity of ES cells [44].
The expression of ABCG2, a well recognized stem cell marker [45], was down-regulated in "PR versus P" (−3.1 fold) (see Table 1(b)) and slightly in "P versus V" (−1.3 fold, P < .01) conditions (not shown). ABCG2 represents an ATP-binding cassette (ABC) transporter which serves to efflux certain xenobiotics (including anticancer drugs) that can lead to the development of multidrug resistance syndrome. This is a significant obstacle in cancer treatment [46]. This gene is also considered to be a marker of primitive pluripotent stem cells, termed "side population," which were identified based on their ability to exclude Hoest dye [45]. The ability to exclude a variety of substances may comprise a mechanism that protects stem cells from exogeneous and endogeneous toxins. Finding that ABCG2 expression is down regulated by PPARγ2, especially after activation with rosiglitazone, implicates PPARγ2 as a negative regulator of stem cell phenotype as well as a negative regulator of multidrug resistance. Similarly, Egfr a marker of early stem cells is down regulated by PPARγ2 when activated with rosiglitazone [47].
Interestingly, however, the expressions of Oct-4 (POU5f1) and FGF4, well recognized embryonic stem cell markers highly expressed in the totipotent and pluripotent ES cells [48,49] are up regulated in U-33/γ2 cells compared to U-33/c cells and are not affected in U-33/γ2 cells treated with rosiglitazone (see Table 1(a)).
Another interesting grouping consists of genes whose expression is differentially regulated by both activated and nonactivated PPARγ2 (see Table 2). A number of genes implicated in early stem cell maintenance and recruitment, among them CD44, H2-D1, PCNA, CD109, Spred1 and 2, and Stag1 and 2, are down regulated in U-33/γ2 cells in both basal conditions and upon rosiglitazone treatment.
The last category represents gene markers specific for terminally-differentiated cells.Consistent with the proadipocytic and antiosteoblastic activities of PPARγ2 activated with rosiglitazone, the expression of the gene encoding FABP4 increases, whereas an expression of the gene underlying alkaline phosphatase decreases. Markers of the neuronal phenotype are either decreased (S100b, Table 2) or not affected (nestin and NCAMs, Table 4), and the expression of CD34, a bona fide marker for cells of hematopoietic lineage, is not affected (see Table 4). However, the expression patterns of gene markers characteristic for embryonic stem cells and a large number of markers that are associated with a nonmesenchymal phenotype, including markers of different hematopoietic and neuronal lineages, indicates that marrow mesenchymal U-33 cells possess a mixed phenotype with some characteristics of early primitive pluripotent stem cells and lineage oriented mesenchymal cells.
In conclusion, PPARγ2 is a powerful modulator of the stem cell phenotype and its activation with antidiabetic TZDs affect the expression of "stemness" genes. It is unclear at this time whether, and to what extent, PPARy2 is expressed in stem cells in vivo and whether this key transcription factor plays a significant role in stem cell biology. However, the findings presented here, together with previously published evidence of increased PPARγ2 expression in MSCs with aging [14] and a loss of marrow MSC plasticity or ability to convert between phenotypes as a result of aging and TZD therapy [20], suggest that aging and TZD therapy may affect stem cell phenotype through modulation of PPARγ2 activity. These observations may also have important therapeutic consequences and indicate a need for more detailed studies of PPARγ2 role in stem cell biology.