Author contributions: S.S., Y.J. and L.W.: conception and design, writing manuscript, and final approval of manuscript; L.W., Y.Z., Y.L., K.A., C.C. and C.Q.: collection and assembly of data; S.S., L.W., Y.Z. and Y.L.: data analysis and interpretation.
Disclosure of potential conflicts of interest is found at the end of this article.
first published online in STEM CELLSEXPRESS April 4, 2013.
An inflammatory microenvironment may cause organ degenerative diseases and malignant tumors. However, the precise mechanisms of inflammation-induced diseases are not fully understood. Here, we show that the proinflammatory cytokines interferon-γ (IFN-γ) and tumor necrosis factor α (TNF-α) synergistically impair self-renewal and differentiation of mesenchymal stem cells (MSCs) via nuclear factor κB (NFκB)-mediated activation of mothers against decapentaplegic homolog 7 (SMAD7) in ovariectomized (OVX) mice. More interestingly, a long-term elevated levels of IFN-γ and TNF-α result in significantly increased susceptibility to malignant transformation in MSCs through NFκB-mediated upregulation of the oncogenes c-Fos and c-Myc. Depletion of either IFN-γ or TNF-α in OVX mice abolishes MSC impairment and the tendency toward malignant transformation with no NFκB-mediated oncogene activation. Systemic administration of aspirin, which significantly reduces the levels of IFN-γ and TNF-α, results in blockage of MSC deficiency and tumorigenesis by inhibition of NFκB/SMAD7 and NFκB/c-FOS and c-MYC pathways in OVX mice. In summary, this study reveals that inflammation factors, such as IFN-γ and TNF-α, synergistically induce MSC deficiency via NFκB/SMAD7 signaling and tumorigenesis via NFκB-mediated oncogene activation. STEM Cells2013;31:1383–1395
Degenerative diseases, such as osteoporosis, usually show elevated inflammatory activity [1, 2]. Postmenopausal osteoporosis is the most common form of osteoporosis, in which estrogen deficiency-induced T-cell activation promotes osteoclastogenesis to reduce bone mineral density (BMD) [3–5]. Although antiresorptive therapies have been utilized in clinical treatment, concerns have been raised by the emergence of side effects, such as bisphosphonate-related osteonecrosis of the jaw [6, 7]. Thus, it is critical to further understand the underlying mechanisms of osteoporosis and develop novel therapeutic strategies accordingly.
Mesenchymal stem cells (MSCs) are a population of stromal progenitor cells with self-renewal and multipotent differentiation capabilities, possessing the potential to replace damaged and diseased tissues [8, 9]. Accumulating evidence shows that deficiency in MSC/osteoblast lineage consistently contributes to bone degenerative phenotypes in osteoporosis in both mice and humans [2, 10–13]. However, it remains unclear how MSCs are impaired in these immune disorder-associated diseases. If left untreated, degenerative diseases may become a risk factor for cancer occurrence [14, 15], in which chronic infections may play an important role . Moreover, anti-inflammatory treatments show effectiveness in reducing the risk of cancer metastasis . It was speculated that either osteosarcoma or Ewing's sarcoma may directly originate from spontaneous mutation of bone marrow-derived MSCs [18–21]. Therefore, it is possible that a long-term inflammatory environment may eventually induce MSC tumorigenesis.
It has been proved that MSCs closely interact with the host immune system. MSCs are able to inhibit proinflammatory cytokines by targeting several subsets of immune cells, including T cells, B cells, dendritic cells, and natural killer cells [22–24], and such immunoregulatory property of MSCs provides the foundation for clinical applications in treating several autoimmune diseases [2, 25]. Conversely, immune components target MSCs. For example, activated T cells can induce bone marrow MSC apoptosis via the tumor necrosis factor (TNF) receptor superfamily member 6 (FAS)/FASL and CD40/CD40L pathways, respectively [10, 26]. Lymphocytes can impair survival of implanted MSCs by secreting the proinflammatory cytokines interferon-γ (IFN-γ) and TNF-α, thereby negatively affecting MSC-mediated tissue regeneration [27, 28]. However, it remains unknown whether specific inflammatory factors, such as IFN-γ and TNF-α, affect MSC properties in an immune disorder-associated model, including ovariectomized (OVX) osteoporotic mice. Among many inflammatory cytokines, IFN-γ and TNF-α have been proved to play crucial roles in estrogen-deficient osteoporosis [29, 30]. Previous studies have shown that OVX enhances production of TNF-α by T cells, which in turn, augments macrophage colony-stimulating factor and receptor activator of nuclear factor κB (NFκB) ligand-induced osteoclastogenesis . Furthermore, it has been demonstrated that OVX upregulates IFN-γ-induced class II transactivator, resulting in enhanced T-cell activation and prolonged lifespan of activated T cells .
Mothers against decapentaplegic homolog 7, or SMAD7, is an important inhibitory factor of the transforming growth factor β/SMAD pathway, which is one of important signaling pathways regulating osteogenic differentiation. The relationship between SMAD7 and NFκB is not clear. Although it was reported that SMAD7 inhibits nuclear translocation and transcriptional activity of the cell survival factor NFκB [31, 32], other studies demonstrated that proinflammatory cytokines, such as interleukin-1β (IL-1β) and TNF-α, upregulated SMAD7 through the activation of the NFκB pathway in fibroblasts and chondrocytes [33, 34]. Therefore, IFN-γ and TNF-α may upregulate SMAD7 via the activation of the NFκB pathway, thereby inducing the deficiency of osteogenic differentiation of MSCs. In this study, we determine that IFN-γ and TNF-α, as critical inflammatory factors, synergistically induce MSC deficiency via NFκB/SMAD7 signaling, whereas their long-term effect can result in the induction of MSC tumorigenesis by NFκB-mediated oncogene activation.
MATERIALS AND METHODS
Female C3H/HeJ, C57BL/6, TNF-α knockout (KO) mice (B6;129S-Tnftm1Gkl/J) and IFN-γ KO mice (B6.129S7-Ifngtm1Ts/J) were purchased from Jackson Lab (Bar Harbor, ME, http://www.jax.org). Generation of ovariectomized (OVX) mice was performed as described previously  and age-matched mice receiving sham operation served as control. Female immunocompromised mice (Beige nude/nude XIDIII) were purchased from Harlan (Indianapolis, IN, http://www.harlan.com). These animal experiments were performed under institutionally approved protocols for the use of animal research (University of Southern California # 10941, 11141, and 11327).
Anti-mouse TNF-α, SMAD7, alkaline phosphatase (ALP), TNF receptor-1 (TNFR1), TNFR2, IFN-γ receptor α (IFNGRα), vimentin (VIM), pan-Cytokeratin (pan-CK), proliferating cell nuclear antigen (PCNA), c-FOS, and c-MYC antibodies were purchased from Santa Cruz Biotech (Santa Cruz, CA, http://www.scbt.com); IFN-γ antibody was purchased from Abbiotec (San Diego, CA, http://www.abbiotec.com); phosphorylated nuclear factor κB (p-NFκB), NFκBp65, p-NFκBp50, NFκBp50, phosphorylated nuclear factor κ B inhibitor (p-IκB), IκBα, p-SMAD1, SMAD1, runt-related transcription factor 2 (RUNX2), and signal transducers and activators of transcription factor 1 (STAT-1) were purchased from Cell Signaling (Danvers, MA, http://www.cellsignal. com). Anti-β-Actin antibody was purchased from Sigma-Aldrich (St. Louis, MO, http://www.sigmaaldrich.com).
Measurement of TNF-α and IFN-γ in Blood Serum, Bone Marrow, Derma, and Fat
Peripheral blood was collected from retro-orbital venous plexus, and IFN-γ and TNF-α enzyme-linked immunosorbent assay (ELISA) (BioLegend, San Diego, CA, http://www.bio legend.com) was performed to detect in blood serum according to their manufactures' instructions. Total protein from bone marrow, derma, and subcutaneous fat was extracted using M-PER mammalian protein extraction reagent (Thermo, Rockford, IL, http://www.thermoscientific.com), and IFN-γ and TNF-α were measured by ELISA, respectively.
Isolation of Mouse MSCs
For isolation of bone marrow MSCs, bone marrow cells were flushed out from bone cavity of femurs and tibias of mice with 2% heat-inactivated fetal bovine serum (FBS; Equitech-Bio, Kerrville, TX, http://www.equitech-bio.com) in phosphate-buffered saline (PBS). For isolation of dermal MSCs and subcutaneous fat MSCs, dermis and subcutaneous fat were carefully separated from the full skin under a dissecting microscope, cut into small pieces, and digested with 1.5 mg/ml collagenase type I (Worthington, Lakewood, NJ, http://www.worthington-biochem.com) and Dispase (Roche, Indianapolis, IN, http://www.roche-applied-science.com) in PBS 2 mg/ml for 1 hour at 37°C to release the individual cells. A single-cell suspension of all nucleated cells was obtained by passing all bone marrow cells through a 70-μm cell strainer (BD Bioscience, San Jose, CA, http://www.bdbiosciences. com), respectively. All the single cells were seeded into 100-mm culture dishes (Corning, Corning, NY, http://www. corning.com) and initially incubated for 48 hour at 37°C and 5% CO2. To eliminate the nonadherent cells, the cultures were washed with PBS twice on the second day. The attached cells were cultured for 16 d with alpha minimum essential medium (α-MEM, Invitrogen, Grand Island, NY, http://www.invitrogen.com) supplemented with 20% FBS, 2 mM L-glutamine (Invitrogen), 55 μM 2-mercaptoethanol (Invitrogen), 100 U/ml penicillin, and 100 μg/ml streptomycin (Invitrogen). To confirm mesenchymal stem cell character, flow cytometric analysis was used to show that these MSCs were positive for CD73, CD90, Sca-1 and SSEA-4 and negative for CD11b, CD31, CD34, and CD45.
Population doubling (PD) for each passage was determined using the formula PD≈log2[Nc/No], where No is the original cell population and Nc is the number of cells at confluence, and by adding the PD from each passage together to obtain the PD values. The PD assay for mouse MSCs began with passage 1.
Osteogenic Differentiation of MSCs
For osteogenic induction in vitro, mouse MSCs were seeded on 35-mm dishes (Corning) and cultured in the growth medium until the cells reached confluence. To induce osteogenic differentiation, the medium was changed to an osteogenic medium consisting of L-ascorbic acid phosphate and β-glycerophosphate. After 2 weeks of osteogenic induction, total protein was extracted and analyzed for osteogenic markers by Western blot. After 4–6 weeks, calcium deposits formed by osteoblasts on the dishes were detected by staining with 2% Alizarin Red (Sigma-Aldrich).
Histological and Histomorphometric Analysis
Femurs and skins were fixed with 4% paraformaldehyde, followed by decalcification with 10% EDTA (pH 8.0) for femurs, and embedded in paraffin. For histological and histomorphometric analysis, sections were deparaffinized and stained with hematoxylin and eosin (HE), followed by trabecular percentage and derma thickness calculation using ImageJ software.
For the detection of TNF-α, IFN-γ, VIM, pan-CK, PCNA, p-NFκBp65, c-FOS, and c-MYC, the sections were blocked with serum matched to secondary antibodies, incubated with the first antibodies for 30 min at room temperature, and stained using VECTASTAIN Elite ABC Kit (Vector, Burlingame, CA, http://www.vectorlabs.com) and DAB Kit (Invitrogen), according to the manufacturers' instructions.
Administration of IFN-γ and TNF-α Neutralizing Antibodies in OVX Mice
To chronically deplete IFN-γ and TNF-α, at 1-day before OVX, 12-week-old female C3H mice were injected i.p. with 250 μg of IFN-γ and TNF-α neutralizing antibodies (BioLegend) in 0.5 ml sterile saline every 3 days until 21 days post-OVX, respectively. In control groups, 250 μg isotype control antibodies in 0.5 ml sterile saline were identically administered.
Administration of IFN-γ and TNF-α in Nude Mice
To mimic IFN-γ- and TNF-α-induced MSC deficiency in an in vivo model, 8-week-old female nude mice were injected with 4 μg recombinant mouse IFN-γ (BioLegend) and 1 μg recombinant mouse TNF-α (BioLegend), either alone or in combination, in 0.2 ml sterile saline via tail vein every 3 days for 21 days, respectively.
Inhibition of Osteogenic Differentiation by IFN-γ and TNF-α Treatment
Half a million MSCs were seeded onto each 6-well culture dish (Corning) with or without treatment using IκB kinase α (IKKα), SMAD7, TNFR1, or IFNGRα small interfering RNA (siRNA) (Santa Cruz) according to the manufacturer's instruction. IFN-γ and TNF-α, either alone or incombination, were used to treat the MSCs with normal growth media or osteogenic media for 14 days. Total protein was extracted for detection of NFκB pathway and osteogenic markers using Western blot. For some experiments, the culture plates were stained with 2% toluidine blue O and 2% paraformaldehyde to evaluate cell survival.
Western Blot Analysis
Total protein was extracted using M-PER mammalian protein extraction reagent (Thermo). Nuclear protein was obtained using NE-PER nuclear and cytoplasmic extraction reagent (Thermo). Protein was applied and separated on 4–12% NuPAGE gel (Invitrogen) and transferred to ImmobilonTM-P membranes (EMD Millipore, Billerica, MA, http://www.emd millipore.com). After blocking with 5% nonfat dry milk for 1 hour, the membranes were incubated with the primary antibodies (1:200–1,000 dilution) at 4°C overnight. Horseradish peroxidase-conjugated IgG (Santa Cruz) was used to treat the membranes for 1 hour and subsequently treated with a chemiluminescent substrate (Thermo). The bands were detected on BIOMAX MR films (Kodak, Rochester, NY, http://www. kodak.com). Each membrane was also stripped using a stripping buffer (Thermo) and reprobed with anti-β-Actin to determine the loading amount.
Sarcomagenesis of BMMSCs in MCA-Treated Mice
To determine the susceptibility to tumorigenesis of bone marrow MSCs (BMMSCs) in vivo, we locally applied 3-methycholanthrene (MCA) to the bone marrow of tibias as a carcinogen to induce osteosarcoma according to the method described in previous studies with modifications [36, 37]. One month after OVX or sham operation, a small hole was created in the metaphysis of right tibia using a dental bur after anesthesia, and a single dose of 0.5 mg MCA in 10 μl of dimethyl sulfoxide (DMSO) was injected into the bone marrow of tibia, followed by sealing using bone wax. Control mice received identical procedures except that DMSO without MCA was implanted into tibias. MCA-treated mice were closely observed for tumor occurrence for up to 4 months. MCA-treated tibias were harvested, demineralized with 10% EDTA, and embedded for HE, Trichrome staining, and immunostainings for VIM, pan-CK, and PCNA.
MSC Spontaneous Malignant Transformation Assay
To evaluate the effects of inflammatory microenvironment to MSC malignant transformation in vitro, we treated mouse BMMSCs with long-term IFN-γ (50 ng/ml) and TNF-α (5 ng/ml), with or without treatment of NFκB inhibitor BAY 11-7082 (1 μM, Sigma-Aldrich), and continuously passaged to gradually acquire unlimited population doublings as previously described . Giemsa staining was performed to monitor the cell shape.
Transplantation and Injection of BMMSCs
To assess tumorigenesis and migration potential, 2 × 106 and 1 × 106 of BMMSCs were injected into 8-week-old nude mice subcutaneously and intravenously through tail vein, respectively. Transplants and lungs were harvested 8 weeks after transplantation, and the paraffin-embedded sections were stained with HE, Trichrome, or incubated with VIM, pan-CK, and PCNA antibodies (Santa Cruz) for immunohistochemistry analysis.
Semiquantitative Real-Time PCR
Total RNA was isolated from the cultures using SV total RNA isolation kit (Promega, Madison, WI, http://www. promega.com) following the manufacturer's protocols. The cDNA was synthesized from 100 ng of total RNA using Superscript III (Invitrogen). Polymerase chain reaction (PCR) was performed on a CFX 96 real-time PCR thermocycler (BioRad, Hercules, CA, http://www.bio-rad.com) using gene-specific primers and Cybergreen supermix (BioRad). Primer pairs are as follows: mouse c-Fos, forward primer (AAACCGCATGGAGTGTGTTGTTCC) and reverse primer (TCAGACCACCTCGACAATGCATGA); mouse c-Myc, forward primer (GCCACGTCTCCACACATCAG) and reverse primer (TCTTGGCAGCAGGATAGTCCTT). The conditions for amplification were as follows: 94°C for 5 min, followed by 40 cycles of 94°C for 20 s; 60°C for 20 s; and 72°C for 25 s. Data from the real-time PCR reactions were analyzed using CFX Manager Software, Version 2 (Bio-Rad). Relative changes in mRNA expression were analyzed using the 2–ΔΔCt method, with Gapdh serving as an internal reference to correct for differences in RNA extraction or reverse transcription efficiencies. These standardized data were used to calculate fold-differences in gene expression. All real-time PCR amplifications were performed in triplicate.
Aspirin Administration in OVX Mice
To determine the effects of aspirin to prevent MSC tumorigenesis and MSC deficiency in vivo, mice were fed with aspirin (1 mg/ml, Sigma-Aldrich) dissolved in water or placebo for 2 months in OVX mice and 4 months in MCA-treated OVX mice. BMMSCs and DMSCs were harvested for PD and osteogenic differentiation analysis.
SPSS 13.0 was used to perform statistical analysis. Significance was assessed by independent two-tailed Student's t test or analysis of variance. The p < 0.05 were considered significant.
Elevated Levels of IFN-γ and TNF-α Synergistically Impair MSCs in OVX Mice
Of many immune disorder-associated degenerative diseases, postmenopausal osteoporosis is an important example. It features overactivation of proinflammatory cytokine-producing T cells and decreased osteogenic differentiation capabilities in bone marrow MSCs (BMMSCs), resulting in reduced bone mineral density (BMD) [2–5, 10]. IFN-γ and TNF-α have been proved to play crucial roles in estrogen-deficient osteoporosis [29, 30]. Blockage of either IFN-γ or TNF-α can abolish OVX-induced osteoporotic phenotype [3, 5]. Moreover, OVX enhances T-cell activation and prolonged lifespan of active T cells through IFN-γ-induced class II transactivator . Our recent study also showed that IFN-γ and TNF-α synergistically impaired implanted MSCs in vivo . Based on these observations, we asked whether IFN-γ and TNF-α would affect MSC properties in OVX-induced osteoporotic mice. We hypothesized that IFN-γ and TNF-α, as two key representatives in the inflammatory microenvironment, might synergistically play an important role in inducing MSC deficiency in OVX-induced osteoporotic mice. Concentrations of both IFN-γ and TNF-α in serum, bone marrow and derma, but not in subcutaneous fat, were markedly increased in OVX mice when compared with sham-operated mice by ELISA and immunohistochemical staining analysis (Fig. 1A; Supporting Information Fig. S1). BMMSCs derived from OVX mice showed markedly decreased population doublings when compared with sham mouse-derived BMMSCs, indicating that BMMSC self-renewal capability was impaired in OVX mice (Fig. 1B). Importantly, osteogenic differentiation of BMMSCs was also impaired in OVX mice, as indicated by Alizarin red staining to show markedly decreased mineralized nodule formation after induction of osteogenic media (Fig. 1C). Because derma and subcutaneous fat, as neighboring tissues, showed different alteration in the levels of IFN-γ and TNF-α following OVX procedure (Fig. 1A), we speculate that OVX may affect derma and subcutaneous fat MSCs differently from bone marrow MSCs. Although OCT-4 is not a specific marker for MSCs, it has been proved to be an important marker to indicate that MSCs possess self-renewal capacity and can retain their undifferentiated status . Therefore, we used OCT-4 as a marker of functional stem cell assurance for MSCs. We found that the derma, but not the fat, from OVX mice harbored markedly less OCT-4-positive MSCs (Fig. 1D), contributing to a markedly decreased collagen deposition in the derma as shown by Trichrome staining, when compared with sham-operated mice (Fig. 1E). To further confirm decreased MSC number in OVX mice, we performed histoimmunostaining to show markedly reduced number of dermal MSCs, which express the MSC markers CD105, CD44 and CD73 in OVX mice (Supporting Information Fig. S2). Moreover, OVX-derived dermal MSCs (DMSCs), but not fat MSCs (FMSCs), showed markedly decreased population doublings (Fig. 1F) and osteogenic differentiation (Fig. 1G), which served as a representative to determine the differentiation capabilities of MSCs in the experiments thereafter. Consequently, OVX mice showed markedly deteriorated trabecular bone structures and reduced dermal thickness when compared with sham-operated mice (Supporting Information Fig. S3A, S3B).
We next asked whether elevated levels of IFN-γ and TNF-α would synergistically cause MSC deficiency. To clarify this, we first systemically administered either IFN-γ or TNF-α neutralizing antibody to OVX mice to test whether depletion of IFN-γ or TNF-α could prevent OVX-induced MSC deficiency. Indeed, administration of IFN-γ neutralizing antibody was able to markedly reduce IFN-γ level in the serum of OVX mice, without affecting TNF-α level; similarly, administration of TNF-α neutralizing antibody could markedly reduce the serum TNF-α level in OVX mice, without changing IFN-γ level (data not shown). Interestingly, histological analysis showed that either IFN-γ or TNF-α neutralizing antibody was able to markedly inhibit bone loss in OVX mice (Fig. 2A). MSC characterization showed that either IFN-γ or TNF-α neutralizing antibody treatment in vivo markedly improved population doublings and osteogenic differentiation in OVX-derived BMMSCs (Fig. 2B). Also, either IFN-γ or TNF-α neutralizing antibody was able to prevent derma atrophy in OVX mice (Fig. 2C) and improve population doublings and differentiation capabilities in OVX-derived DMSCs (Fig. 2D). To further dissect whether IFN-γ and TNF-α synergistically cause MSC deficiency in OVX mice, we confirmed that OVX was not able to induce BMD loss in either IFN-γ KO or TNF-α KO mice (Fig. 2E, 2F). Importantly, population doublings and differentiation capabilities were not decreased in BMMSCs or DMSCs in IFN-γ KO or TNF-α KO mice after OVX (Fig. 2E, 2F). To further confirm the synergistic effect of IFN-γ and TNF-α in vivo, we systemically infused IFN-γ and TNF-α, either alone or in combination, to immunocompromised mice via tail vein. Interestingly, infusion of IFN-γ and TNF-α in combination, but not separately, caused a dramatic reduction in population doublings and differentiation capabilities of both BMMSCs and DMSCs (Fig. 3A–3D), resulting in both osteopenia phenotype and dermal atrophy with poor collagen deposition as shown by Trichrome staining (Fig. 3E–3G; Supporting Information Fig. S4A, S4B). Furthermore, when IFN-γ and TNF-α, either alone or in combination, were subcutaneously transplanted with hydrogel as a controlled release system, DMSC population doublings and differentiation capabilities were reduced by IFN-γ and TNF-α in combination, but not separately (Supporting Information Fig. S5A, S5B), thereby causing derma atrophy (Supporting Information Fig. S5C). These in vivo data suggest that elevated levels of IFN-γ and TNF-α synergistically, but not independently, cause MSC deficiency.
IFN-γ and TNF-α Synergistically Induce NFκB/SMAD7 Pathway Activation, Causing Differentiation Deficiency in MSCs
To examine the mechanism by which IFN-γ and TNF-α synergistically induce impairment of MSCs, we demonstrated that BMMSCs and DMSCs derived from OVX mice showed the upregulation of p-NFκB, p-IκB, and inhibitory SMAD7, but the downregulation of osteogenic genes, including phosphorylated SMAD1 (p-SMAD1), RUNX2, and ALP (Fig. 4A; Supporting Information Fig. S6A). To mimic in vivo microenvironment in OVX bone marrow, we treated BMMSCs with IFN-γ and TNF-α in different concentrations, either alone or in combination, and found that TNF-α-induced NFκBp65 activation in a dose-dependent manner with a maximal activation at concentrations of 5 and 20 ng/ml (Supporting Information Fig. S6B), whereas the combination of IFN-γ (50 ng/ml) and TNF-α (5 ng/ml) induced more significant upregulation of NFκBp65 signaling (Fig. 4B). Interestingly, IFN-γ and TNF-α in combination, but not separately, caused a significant reduction of mineralized nodule formation in BMMSCs (Fig. 4C) and DMSCs cultured in osteogenic media (Supporting Information Fig. S6C, S6D), without dramatically affecting MSC survival (Fig. 4E). Furthermore, Western blot showed that IFN-γ/TNF-α-induced NFκB activation upregulated inhibitory SMAD7, but downregulated osteogenic genes, including p-SMAD1, RUNX2, and ALP (Fig. 4F). When NFκB expression was knocked down by IKKα siRNA, IFN-γ/TNF-α-induced upregulation of SMAD7 and downregulation of p-SMAD1, RUNX2, and ALP were abolished, leading to the rescue of IFN-γ/TNF-α-induced reduction of mineralized nodule formation in BMMSCs (Fig. 4G, 4H) and DMSCs (Supporting Information Fig. S6E, S6F). Under osteogenic induction, IFN-γ and TNF-α treatment in combination was able to dramatically upregulate IFN-γ receptor IFNGRα and TNF-α receptor TNFR1, but not TNFR2, when compared with IFN-γ and TNF-α treatment alone (Fig. 4D). Combination knockdown of IFNGRα and TNFR1 resulted in marked inhibition of IFN-γ-/TNF-α-induced activation of NFκB pathway when compared with knockdown of TNFR1 alone, suggesting that IFN-γ activates TNF-α-induced NFκB activation via IFNGRα (Supporting Information Fig. S7A). Moreover, IFN-γ, either alone or in combination with TNF-α, was able to markedly upregulate p-STAT-1 (Fig. 4D), whereas knockdown of STAT1 dramatically reduced the TNF-α-/IFN-γ-induced activation of NFκB pathway, suggesting that IFN-γ/IFNGRα pathway enhances TNF-α-triggered NFκB activation via STAT-1 activation (Supporting Information Fig. S7B). We confirmed the efficiency of IKKα, IFNGRα, TNFR1, STAT-1 and SMAD7 knockdown by their spe-cific siRNAs using Western blot analysis (Supporting Information Fig. S7C, S7D). Additionally, knockdown of SMAD7 expression by siRNA assay resulted in rescuing TNF-α-/IFN-γ-induced reduction of mineralized nodule formation and downregulation of p-SMAD1, RUNX2, and ALP in BMMSCs and DMSCs, without affecting p-IκB and p-NFκB (Supporting Information Fig. S8A–S8D). These data suggest that SMAD7 acts as a downstream target of NFκB signaling in TNF-α-/IFN-γ-induced MSC osteogenic deficiency.
To further confirm the role of the NFκB pathway in TNF-α-/IFN-γ-induced MSC deficiency, we showed that NFκB inhibitor Bay 11-7082 treatment was able to rescue osteogenic differentiation of OVX-derived BMMSCs by downregulating SMAD7 and upregulating p-SMAD1, RUNX2, and ALP (Fig. 4I, 4J). To confirm the synergistic effects of IFN-γ and TNF-α that resulted in MSC deficiency in vivo, we demonstrated that OVX failed to induce MSC NFκB activation in IFN-γ and TNF-α KO mice, respectively (Fig. 4K, 4L). Moreover, either IFN-γ or TNF-α neutralizing antibody treatment in vivo was sufficient to downregulate NFκB pathway in BMMSCs to the level observed in the sham-operated group (Fig. 4M). These in vivo results confirm that IFN-γ and TNF-α work synergistically to induce osteogenic deficiency in MSCs via NFκB pathway.
IFN-γ and TNF-α Increase the Tendency Toward MSC Malignant Transformation via NFκB-Mediated Upregulation of Oncogenes c-Fos and c-Myc
Given the facts that degenerative diseases may be a risk factor for cancer occurrence  and that NFκB activation may upregulate oncogenes c-Fos and c-Myc expression , we next asked whether IFN-γ-/TNF-α-induced NFκB activation could increase the tendency toward MSC malignant transformation. When we extended the observation time to 1-year post-OVX, the aging OVX mice showed a markedly increased tumor incidence when compared with aging sham-operated mice (Fig. 5A). Interestingly, ELISA showed that both IFN-γ and TNF-α concentration in serum remained markedly higher in OVX mice when compared with sham-operated mice (Fig. 5B), suggesting that long-term overexposure of MSCs to IFN-γ and TNF-α may be associated with higher susceptibility to tumorigenesis in OVX mice. To confirm that such higher susceptibility of tumorigenesis involves MSCs, we locally applied 3-methycholanthrene (MCA) as a carcinogen to induce osteosarcoma [36, 40], and found that the MCA-treated OVX mice showed a higher incidence of sarcoma with earlier sarcoma occurrence in femurs when compared with sham-operated mice (Fig. 5C). In tumors that occurred in MCA-treated OVX mice, but not sham-operated mice, histological analysis showed the production of osteoid by atypical tumor cells with high mitotic activity and abundant neovascularization, as well as osteogenic differentiation of the osteosarcoma-like tumor cells, as indicated by Trichrome staining (Fig. 5D). Immunohistochemical costaining showed that CD73 was highly overlapped with VIM or PCNA (Fig. 5E), confirming that the highly proliferating tumor cells mainly originated from mesenchymal cells, while excluding the possibility of epithelial cell origin, as indicated by negative staining of pan-CK (Supporting Information Fig. S9A). Furthermore, MCA-treated OVX mice showed a markedly lower incidence of sarcoma with delayed sarcoma occurrence in IFN-γ KO and TNF-α KO mice, respectively, when compared with the sham-operated mice (Fig. 5F), suggesting again that IFN-γ and TNF-α work synergistically to, in this case, induce sarcomagenesis.
These findings provided a strong rationale for further exploring the mechanisms by which long-term IFN-γ/TNF-α overexposure could induce MSC malignant transformation. To achieve this, we used an established spontaneous malignant transformation assay  by continuously passaging mouse BMMSCs to gradually acquire unlimited population doublings. When treated with long-term IFN-γ/TNF-α in vitro, the BMMSCs showed a markedly quicker increase in population doublings after passage 10, whereas NFκB inhibitor was able to delay such increase in population doublings (Fig. 6A). At passage 12, IFN-γ-/TNF-α-treated BMMSCs showed markedly smaller morphology with frequent abnormal mitosis and the generation of sarcoma-like tumors on s.c. inoculation, which, when infused to nude mice via tail vein settled in the lungs with in situ tumorigenesis (Fig. 6B). In contrast, neither NFκB inhibitor-treated mice nor control mice showed any change in cell morphology or generated any tumor on s.c. inoculation or infusion into nude mice (Fig. 6B). Immunohistochemical staining of the tumors derived from s.c. inoculation of IFN-γ-/TNF-α-treated BMMSCs revealed that the highly proliferating tumor cells originated from mesenchymal cells, as indicated by high expressions of VIM and PCNA and negative staining of pan-CK (Supporting Information Fig. S9B).
To further dissect the molecular mechanism of such IFN-γ-/TNF-α-induced MSC tumorigenesis, we demonstrated that the expression of c-MYC and c-FOS, was markedly elevated in long-term IFN-γ-/TNF-α-treated BMMSCs when compared with control BMMSCs (Fig. 6C), but that knockdown of NFκB by IKKα siRNA assay was able to abolish the elevation in c-MYC and c-FOS (Fig. 6C). To confirm this result in vivo, we showed that c-Myc and c-Fos were markedly elevated in BMMSCs derived from aging OVX mice when compared with sham-operated ones using real-time PCR (Fig. 6D), which was confirmed by Western blot and immunohistochemistry showing that c-MYC and c-FOS expression was enhanced along with NFκB activation in aging OVX mice when compared with sham-operated mice (Fig. 6E, Supporting Information Fig. S9C).
Aspirin Blocks OVX-Induced MSC Deficiency and Tumorigenesis by Inhibiting IFN-γ and TNF-α
Next, we investigated whether blocking the levels of IFN-γ and TNF-α by pharmacological approach could rescue MSC deficiency and tumorigenesis in vivo. As aspirin has been reported to inhibit the function of IFN-γ and TNF-α [28, 41], we examined the effect of systemic aspirin treatment in OVX mice which had undergone MCA injections in their tibias. Interestingly, aspirin treatment dramatically reduced the concentrations of IFN-γ and TNF-α in both serum and bone marrow, resulting in a decreased and delayed osteosarcoma occurrence following MCA treatment (Fig. 7A, 7B), with marked inhibition of NFκB activity in BMMSCs, as shown by Western blot analysis (Fig. 7C). The data suggest that aspirin treatment may decrease the tendency toward MSC malignant transformation by inhibiting IFN-γ-/TNF-α-induced NFkB activation.
To determine the efficacy of aspirin administration to rescue IFN-γ-/TNF-α-induced MSC deficiency in organ degenerative diseases, we systemically applied aspirin to OVX mice and found that aspirin could markedly improve BMD and rescue the deteriorated trabecular bone structures and derma atrophy in OVX mice (Fig. 7D; Supporting Information Fig. S10A, S10B), at least partly by reducing the concentrations of IFN-γ and TNF-α in serum, bone marrow and derma (Fig. 7E). Importantly, in vivo aspirin administration was able to markedly improve population doublings and differentiation capacities of both BMMSCs and DMSCs in OVX mice (Fig. 7F–7H), suggesting that aspirin treatment may provide an ideal MSC-based therapy for OVX mice by preventing MSC deficiency which is induced by medium-term IFN-γ and TNF-α overexposure.
In summary, we have uncovered a previously unrecognized mechanism whereby elevated levels of IFN-γ and TNF-α synergistically, but not independently, cause MSC deficiency via NFκB-mediated activation of SMAD7 in OVX mice, whereas long-term elevated levels of IFN-γ and TNF-α synergistically result in increased tendency toward MSC malignant transformation through NFκB-mediated upregulation of the oncogenes c-Fos and c-Myc (Supporting Information Fig. S11). Thus, as a therapeutic regimen, systemic administration of aspirin provides a promising approach in rescuing MSC deficiency in organ degenerative diseases, as well as preventing MSC tumorigenesis through reducing the levels of IFN-γ and TNF-α.
It has been well documented that estrogen deficiency increases levels of IFN-γ and TNF-α thereby enhancing osteoclast activity in OVX mice [3–5]. However, it has been unclear whether elevated levels of IFN-γ and TNF-α contribute to the impairment of MSCs/osteoblasts in OVX mice or osteoporotic humans [2, 10–13]. In this study, we found that elevated levels of IFN-γ and TNF-α synergistically induce osteogenic deficiency of MSCs through NFκB/SMAD7, whereas long-term overexposure of IFN-γ and TNF-α can induce malignant transformation of MSCs.
TNF-α is widely accepted as a proinflammatory cytokine, serving as a prerequisite in several models of inflammatory and autoimmune diseases. However, IFN-γ is believed to possess both proinflammatory and anti-inflammatory properties, such as inhibiting IL-1 production, and, as such, it also plays a complex role in inflammation and autoimmune diseases . The respective effects of IFN-γ and TNF-α in osteogenic differentiation of MSCs have remained controversial [28, 43–47]. The discrepancies in the effects of IFN-γ and TNF-α among these studies may be attributed to the different doses of IFN-γ and TNF-α, treatment, variations between the microenvironments and MSC growth status. Nonetheless, IFN-γ and TNF-α do not present independently in vivo; instead, they are simultaneously secreted by activated Th1 and other immune cells and cooperatively create proinflammatory microenvironments. Indeed, our data demonstrated that high levels of IFN-γ and TNF-α synergistically play a crucial role in causing MSC deficiency in OVX mice, thereby causing the phenotypes of osteoporotic bone or derma atrophy in OVX mice. Blockage of either IFN-γ or TNF-α with their neutralizing antibodies in vivo is sufficient to effectively rescue MSC deficiency. Furthermore, OVX is not able to induce MSC deficiency in IFN-γ and TNF-α KO mice respectively, suggesting that IFN-γ and TNF-α act together, not independently, to induce MSC deficiency in the OVX model. As OVX can induce elevation of multiple inflammatory cytokines, it is possible that other cytokines, with the exception of IFN-γ and TNF-α, may contribute to OVX-induced MSC impairment in OVX mice. For example, IL-1 has been proved to enhance osteoclastogenesis in OVX mice . Thus, further study is required to clarify the role of other inflammatory cytokines in OVX-induced osteoporosis. In this study, we demonstrate that IFN-γ and TNF-α play a crucial role in synergistically inducing MSC deficiency in OVX mice. Given the fact that MSCs derived from derma and subcutaneous fat have different changes in population doublings and osteogenic differentiation after OVX, the effects of IFN-γ and TNF-α in inducing MSC deficiency may be associated with tissue levels of IFN-γ and TNF-α.
It has been demonstrated that IFN-γ and TNF-α may activate STAT-1 and NFκB pathways, thereby interacting with their common basal transcription complexes to affect coactivator recruitment in fibroblasts and asthmatic airway smooth muscle cells, respectively [49, 50]. IFN-γ and TNF-α have also been reported to synergistically activate STAT-1/IFN regulatory factor-1 and c-Jun NH2-terminal kinase (JNK) pathways to induce apoptosis of pancreatic β-cells in autoimmune diabetes [51, 52]. However, it is unclear how overexposure of stem cells to the synergistically activity of IFN-γ and TNF-α affects their behavior during the course of degenerative diseases. In this study, we demonstrated that IFN-γ binds its receptor IFNGR1 to strongly enhance TNF-α/TNFR1-upregulated NFκB, which, in turn, mediates inhibitory SMAD7 to downregulate the osteogenic master gene Runx2  and, hence, induce osteogenic deficiency in MSCs, both in vivo and in vitro. In our previous studies, we proved that TNF-α converts the signaling of the IFN-γ-activated, nonapoptotic form of FAS in BMMSCs to a caspases-associated proapoptotic cascade, resulting in the apoptosis of implanted BMMSCs . Such difference in the final fates of MSCs may be caused by high levels of IFN-γ and TNF-α produced by more intensive local infiltration of Th1 cells in tissue-engineered transplants when compared with those in a degenerative disease model, such as OVX mice. The synergistic dose-dependent effects of IFN-γ and TNF-α indicate that proinflammatory microenvironments in degenerative diseases and tissue engineering are hierarchical and complex for the alteration of MSCs.
It has been suggested that TNF-α may drive malignant tumor progression and that NFκB signaling may be involved in tumor development [16, 54, 55]. However, it has been unknown whether long-term overexposure of synergistically active IFN-γ and TNF-α could cause malignant transformation of MSCs. In this study, we extended the period of IFN-γ and TNF-α overexposure and found that the ultimate effect was upregulation of the oncogenes c-Fos and c-Myc via NFκB signaling, thereby causing increased tendency toward MSC tumorigenesis, both in vitro and in vivo. Furthermore, carcinogen-treated OVX IFN-γ and TNF-α KO mice showed a markedly lower incidence of sarcoma with delayed sarcoma occurrence, indicating that IFN-γ and TNF-α must work together, not independently, to upregulate NFκB-mediated oncogenes c-Fos and c-Myc. Thus, it is plausible that both organ degenerative diseases and malignant tumor might be “stem cell diseases”, in which IFN-γ and TNF-α overexposure plays a dynamic role in inducing MSC deficiency and MSC tumorigenesis through activation of NFκB-mediated SMAD7 and the oncogenes c-Fos/c-Myc, respectively.
As a widely used anti-inflammatory drug, aspirin is able to inhibit osteoclastogenesis and improve osteogenesis in OVX mice by inhibiting cyclooxygenases and upregulating telomerase in BMMSCs [10, 56]. As aspirin can block Th1 cell development  and reduce TNF-α- and IFN-γ-producing cells in implanted tissue-engineered bones , we hypothesized that aspirin should be able to block TNF-α and IFN-γ in OVX mice and therefore rescue MSC deficiency and MSC tumorigenesis. Indeed, we demonstrated that systemic administration of aspirin was able to decrease concentrations of TNF-α and IFN-γ, both in serum and bone marrow, and prevent MSC deficiency and carcinogen-induced MSC tumorigenesis in OVX mice. Therefore, aspirin may be a promising drug to treat estrogen deficiency-induced MSC deficiency and reduce the risk of tumor occurrence. However, cautious evaluation is needed before these findings in mice would accurately translate to humans. On the one hand, although aspirin was shown to be associated with higher BMD at multiple skeletal sites in a cohort study of men and women , Bauer et al.  only observed a modest beneficial effect on BMD in another cohort study of postmenopausal women. On the other hand, although daily aspirin has been proved to decrease the risk of adenocarcinoma metastasis in a randomized controlled trial , a recent study indicated that regular administration of aspirin and other nonsteroidal anti-inflammatory drugs was associated with a slightly higher incidence of follicular lymphoma incidence . Thus, the daily use of aspirin or nonsteroidal anti-inflammatory drugs and the risk of cancer in human should be investigated carefully using large cohorts of patients. Also, large cohort studies are required to determine the effects of aspirin in treating post-menopausal osteoporosis and other organ degenerative diseases. Moreover, it is known that MSCs possess such immunomodulatory properties as immunosuppression . Therefore, it will be interesting to examine whether host MSCs show any immunosuppressive effects, such as inhibition of activated T cells and NK cells, thereby serving as a self-protecting mechanism against potential autoimmune reactions by the host immune system.
In summary, this study demonstrated that TNF-α and IFN-γ, as two important representative inflammatory factors, play a crucial role in synergistically inducing MSC deficiency and eventual MSC tumorigenesis via the NFκB pathway in OVX-induced osteoporotic mice. Anti-inflammatory treatment, such as aspirin administration, prevents MSC deficiency and malignant transformation in OVX mice.
This work was supported by grants from National Institute of Dental and Craniofacial Research, National Institutes of Health, Department of Health and Human Services (R01DE017449, R01DE019932, and R01DE019413; to S.S.), grant from California Institute for Regenerative Medicine (RN1-00572; to S.S.), grant from the National Natural Science Foundation of China (No. 81020108019; to Y.J. and S.S.), and grant from the National Basic Research Program (973 Program) of China (No. 2011CB964700; to Y.J.).
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors declare that they have no competing interests.