Spontaneous Expression of Embryonic Factors and p53 Point Mutations in Aged Mesenchymal Stem Cells: A Model of Age-Related Tumorigenesis In Mice

Introduction

As an organism ages, repeated rounds of cell division may lead to the accumulation of genetic damage sufficient to drive the malignant phenotype. Aging is the strongest risk factor for cancer; however, the target cell of malignant transformation is not precisely known. Although a peripheral tissue stem cell is a likely candidate, recent evidence suggests that cells arising from the bone marrow (BM) may also be targets of transformation.

BM-derived cells (BMDC) have a surprising degree of plasticity ( 1). They are identified within all tissues of the body as committed cells ( 2) or as peripheral stem cells ( 3), enabling them to remain in the tissue for extended periods. Whether BMDCs transdifferentiate directly or acquire a peripheral phenotype through fusion with differentiated cells is debated. There is strong evidence both for ( 4– 6) and against ( 7, 8) fusion, and the decision to fuse with a peripheral cell may depend on the differentiated cell type and the presence and severity of tissue damage. The notion of fusion is not trivial. Severely damaged or aged tissue may contain genetic defects, and BMDCs themselves may acquire genetic damage from repeated rounds of replication. Fusion of damaged cells offers another mechanism for cell transformation and cancer initiation seen with aging and with chronic inflammation. Conversely, depending on the genetic mutations involved, fusion may rescue a damaged cell and attenuate the malignant phenotype.

In tissue culture, human mesenchymal stem cells (MSC) are susceptible to mutations, transforming spontaneously after extensive culture ( 9) or after transduction with the telomerase hTERT gene ( 10). Similarly, murine MSCs transform in culture after extended replication and form tumors in immunocompromised mice ( 11). It is not clear how these cells behave in an immunocompetent host or how the genetic changes acquired in vitro correlate with changes seen in vivo.

There is now a substantial body of evidence that BMDCs contribute to cancer in both human ( 12– 15) and mouse models ( 3, 16– 19) as tumor cells and as stromal cells. In the mouse model of chronic Helicobacter felis–induced gastric cancer, gastric cancer originates from a marrow-derived cell ( 3) and requires that the BM stem cell spends an extended period replicating within the peripheral tissue ( 3). We hypothesize that, in the mouse model, aging alone, by extending replication time, would be sufficient for spontaneous transformation of BM-derived MSCs and that genetic alterations responsible for transformation would mirror events occurring in vivo.

Using sex mismatched, green fluorescent protein (GFP)-labeled or β-galactosidase (β-gal)-labeled whole BM- or MSC-transplanted mice and age-matched controls, we show that BM-derived MSCs reside within peripheral tissues of young mice and persist throughout the life of the animal. Fibrosarcoma is the most common age-related cancer in the C57BL/6 mouse. Examination of age-related fibrosarcomas in our marked marrow-transplanted mice shows that these tumors can derive from MSCs and carry mutated p53. We tested the hypothesis that MSC would transform with aging in vitro and acquire characteristics of naturally occurring tumor cells. MSCs isolated from young mice spontaneously transform in culture after repeated replications, acquiring clinically significant p53 mutations and gene expression profiles consistent with in vivo tumors. When placed in a syngeneic mouse model, these spontaneously transformed MSCs (stMSC) form complex soft tissue tumors composed of tumor cells and recruited host cells. MSCs contribute to the tumor directly and through fusion with host BM-derived MSC. stMSCs, but not fused cells, retain tumorigenic potential and can be transferred to second host animals. stMSCs provide a novel model system for studying tumor initiation and progression and testing targets for tumor therapy.

Materials and Methods

All animal work was done at the University of Massachusetts Medical School with Institutional Animal Care and Use Committee approval in accordance with institutional guidelines.

BM transplantation. BM was isolated from the femurs and tibia of male or female 6- to 8-week-old C57BL/6-TgN[ACTbEGF](GFP) mice, C57BL/6J/Gtrosa26 (ROSA26), or C57BL/6J wild-type (WT) mice and processed as described previously ( 3). Whole marrow was used directly for transplantation or used to isolate MSCs. For isolation of MSCs, total marrow was placed into culture and adherent cells were collected at 4 days and expanded in MesenCult (Stemcell Technologies, Inc.) medium according to the manufacturer's direction. Recipient C57BL/6J female mice were irradiated with 900 rads from a Gammacell cesium-137 irradiator and reconstituted with 5 × 10⁶ donor whole marrow cells or 500,000 expanded MSCs plus unmarked female BM as support cells via a single tail vein injection. Mice were used after 4 weeks in experiments or allowed to age without further intervention.

Evaluation of spontaneous tumor formation. Eighteen to 24 months after BM transplantation, mice were euthanized followed by intracardiac perfusion with 10 mL PBS followed by 30 mL of 4% paraformaldehyde and 0.1% glutaraldehyde (pH 7.4). Necropsy was done and tumors were identified. Tumors were postfixed for 30 min to 4 h in 4% paraformaldehyde and 0.1% glutaraldehyde (pH 7.4) at 4°C, embedded in OCT (Sakura), snap frozen, sectioned on a cryostat, or processed for conventional paraffin sections. Sections were stained using routine H&E. For immunohistochemistry, sections were labeled using antibodies directed against bacterial β-gal (Promega) or GFP (Abcam).

Culture and characterization of MSC. Total marrow was isolated as described above. Once confluent, cells were cultured with DMEM/20% FCS. At each passage, cultures were split as follows: one plate was held in culture at confluence for detection of foci, one plate was split into the next passage culture, and the remainder of the plates was frozen. Transformation was assessed using standard assays for foci formation, growth in soft agar, and growth in liquid culture followed by growth in an animal model (see below). Surface expression of CD44 and CD45 (BD PharMingen) was assessed by fluorescence-activated cell sorting (FACS) analysis. Assessment of lineage-specific markers, cytokeratin, vimentin, desmin, CD34, CD31, S100, and smooth muscle actin (SMA; BD PharMingen), was determined by immunohistochemistry or reverse transcription-PCR (RT-PCR). Selected stMSCs were stably transfected with the plasmid pDS-Red-monomer-hyg-C1, red fluorescent protein (RFP) expression was verified by RT-PCR and FACS, and single-cell clones were isolated. Freshly isolated MSCs and late-passage and transformed MSCs were analyzed by OligoGEArray cancer and metastasis cDNA arrays (SuperArray Bioscience) according to the protocol of the company. For growth analysis, 21 60-mm tissue culture plates were seeded with 10,000 cells each. Daily for 7 days, three plates of each group were harvested and trypan blue–negative viable cells were counted by hemocytometer and graphed as the mean ± 1 SD. Standard bromodeoxyuridine (BrdUrd) incorporation, cell cycle analysis, and carboxyfluorescein diacetate succinimidyl ester (CSFE) dilution for doubling time evaluation were measured by FACS. These experiments were repeated a second time with similar results.

p53 immunoblotting and sequencing. Cells were lysed in cold lysis buffer [50 mmol/L Tris-HCl (pH 7.4), 250 mmol/L NaCl, 10 mmol/L MgCl₂, 10 μmol/L ZnCl₂, 1% Triton X-100, 0.5% sodium deoxycholate, complete protease inhibitor tablets (Roche)] and 20 μg of lysate/lane were run on 4% to 12% Nupage Bis-Tris gel (Invitrogen). Antibodies used for immunoblotting were αp53-FL393 (Santa Cruz Biotechnology) and αGAPDH-6C5 (Advanced ImmunoChemical). For sequencing cDNA of the DNA-binding domain of Trp53, total RNA was extracted using RNeasy (Qiagen) and cDNA was synthesized using StrataScript RT (Stratagene) and Pfu high fidelity DNA polymerase (Invitrogen) with Trp53F (96–115), 5′-AGATATCCTGCCATCACCTCACTG-3′, and Trp53R (918–900), 5′-GGGCAGCGCTCTCTTTGC-3′, primers. The PCR product of 823 bp was sequenced directly or cloned into pcDNA 3.0 by blunt-end cloning and sequenced with T7 and BGH-rev primers. The sequencing data were analyzed using Chromos 2.31 software.

In vivo tumor formation and analysis. stMSCs or RFP-stMSCs actively growing in culture were isolated, washed, and resuspended in PBS and viable cells were counted. Between 5 and 5 million cells were injected s.c. on the upper back of female C57BL/6 mice, C57BL/6(GFP) mice, or C57BL/6 mice transplanted with marrow from female C57BL/6(GFP) mice. Mice were anesthetized, exsanguinated by right atrial puncture, and perfused with 10 mL PBS and 30 mL of 4% paraformaldehyde and 0.1% glutaraldehyde (pH 7.4). Tumors were postfixed for 30 min to 4 h and embedded in OCT, snap frozen, sectioned on a cryostat, or processed for paraffin embedding followed by direct visualization of fluorescent protein or immunohistochemistry. Primary antibodies anti-CD31 (BD PharMingen) and anti-FABP4 (Cayman) and FITC-conjugated secondary antibodies (Chemicon International and Santa Cruz Biotechnology). Sections were mounted with antifade Vectashield with 4′,6-diamidino-2-phenylindole (Vector Laboratories), fields were viewed and captured with the Zeiss Axiopath system, and serial three-color images were overlayed using the Olympus MicroSuite B3SV imaging software.

In vivo confocal tumor imaging. Mice were anesthetized and imaged live, or for large tumor imaging, mice were euthanized and excised tumors were imaged immediately. Images were viewed with GFP excitation of 488 nm, emission of 500 to 550 nm, RFP excitation of 543 nm, emission of 565 to 615 nm, and confocal reflectance of 488 nm laser. Images were viewed with the Zeiss C-Apo 40× objective lens and recorded with LSM image browser software.

Single-cell preparation. Four weeks after tumor implantation, mice were anesthetized, euthanized, and perfused with a two-step collagenase digestion solution following published protocols ( 3). Tumors were mechanically disassociated through a wire sieve and FACS immediately into RFP-, GFP-, or dual-expressing populations or analyzed by FACS after varying times in culture.

Results

MSCs reside in peripheral tissue. Eight- to 12-week-old female C57BL/6 mice were transplanted with total BM or MSC from male C57BL/6J transgenic mice expressing a nonmammalian β-gal enzyme (ROSA), chicken β-actin-enhanced GFP, or control WT littermates. Mice receiving MSC also received unmarked female C57BL/6 total BM as support for hematopoietic cell engraftment. Transplanted mice, along with age-matched controls aged without any intervention, were euthanized at 2, 6, and 12 months after transplantation. FACS analysis of peripheral blood showed that 78% to 90% of nucleated cells were GFP⁺ in mice receiving total GFP⁺ BM and <1% of circulating cells in GFP⁺ MSC-transplanted mice. Peripheral tissue engraftment of cells was evaluated beginning at 2 months after transplantation. Rare β-gal⁺ or GFP⁺ BMDCs (excluding leukocytes) were identified within the connective tissue of the skin, gastrointestinal tract, mesentery, and genitourinary tract. Peripherally engrafted cells were recovered with similar frequency in mice receiving total marrow or MSC and were fibroblastoid in appearance.

Fibrosarcomas arising in aged mice are derived from BM MSC. For tumor analysis, mice and age-matched controls were euthanized 18 to 24 months after transplantation. The number and type of tumors identified did not differ between groups ( Table 1 ), suggesting that transplantation protocol did not induce excess tumors. The most common tumors detected among all groups were fibrosarcomas, presenting as a firm, seemingly painless flank or back mass. Microscopically, tumor cells were thin, arranged in a partially swirled herringbone pattern ( Fig. 1A ), and were BM derived as identified by a combination of anti-GFP immunohistochemistry or anti-β-gal immunohistochemistry ( Fig. 1A) with Y chromosome fluorescent in situ hybridization (Y-FISH; Fig. 1B, arrows highlight Y chromosome within nuclei). Soft tissue tumors in transplanted mice were similar in size, shape, and histology to those from control aged mice. Soft tissue tumors that were BM derived formed in mice receiving either marked MSC or whole BM; therefore, we reasoned that repeated replication was driving transformation of a cell found within the MSC population.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Fibrosarcomas from female C57BL/6 mice transplanted with male GFP-expressing or β-gal–expressing marrow are derived from BM. A, anti-GFP antibody (top) or anti-β galactosidase antibody staining in tumors from transplanted mice as indicated. IHC, immunohistochemistry. B, Y-FISH analysis shows Y chromosomes in tumor cells. DAPI, 4′,6-diamidino-2-phenylindole. C, tumors derived from stMSC show a similar morphology to naturally occurring tumors, neovascularization, and invasion.

MSCs spontaneously transform as a result of repeated replication. To evaluate the effect of repeated replication on MSC, we isolated MSCs from young male C57BL/6 mice ( 3). Cells were CD44 positive and CD45 negative, consistent with published data. Cells were placed in culture and continuously passaged. After 12 months in culture, 13 distinct foci were isolated arising from a homogenous fibroblast-like cellular background. Transformation was assessed by secondary foci formation, growth in soft agar, growth in liquid culture, and the ability to form tumors in C57BL/6 mice. Cells of all 13 foci fulfilled at least three of the criteria, and the four cell lines tested fully fulfilled all criteria. stMSCs had one of three phenotypes: large round epithelial-like cells with centrally located nuclei, thin fibroblast-like cells, and small round cells with scant cytoplasm. The most common tumors recovered in the aged mice were fibrosarcomas; therefore, we investigated stMSC-A3, a fibroblast-like cell. At low cell density, stMSC-A3 cells closely resembled nontransformed MSC cultures with a fibroblast-type appearance without a discernible pattern. At confluence, cells assume a whirled arrangement, foci are formed, and spheroids are detached from the culture plate ( Fig. 2A ). Detached spheroids are viable in liquid culture and, when removed, adhere and grow rapidly as a monolayer in a new dish. stMSCs have increased growth kinetics as determined by standard growth curve ( Fig. 2B), BrdUrd incorporation ( Fig. 2C), and CSFE doubling time ( Fig. 2D). stMSC-A3 cells remain CD44 positive and CD45 negative ( Table 2 ) and do not express the lineage-specific markers cytokeratin (epithelium), desmin (muscle), CD31 (endothelium), S100 (melanocytes, select neural tissue, and antigen-presenting cells), or SMA (muscle, myofibroblasts, and activated fibroblasts) by immunohistochemistry or RT-PCR, and they do not contain oil droplets or stain by oil red O (adipocytes; negative data not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Characteristics of stMSC. A, at subconfluence, early-passage MSC (P2), late-passage MSC (P11), and stMSC (A3) have a similar morphology, although A3 cells are smaller. At confluence, P2 is larger, P11 is spindyloid and forms swirls and whirls, and A3 is disordered and forms foci. B, stMSCs grow rapidly compared with early- or late-passage nontransformed MSC. Standard growth curve, cells as labeled. C, BrdUrd incorporation of the cell lines and cell cycle FACS analysis of early-passage MSC, late-passage MSC, and stMSC as indicated. D, CSFE doubling time curves. Cell lines and times as indicated. *, P < 0.02.

stMSC and age-related human tumors share genetic mutations and gene expression profiles. We analyzed the gene expression pattern of early- and late-passage nontransformed MSC compared with stMSC using a combination of focused microarray, RT-PCR, and Western blot analysis and compared the alterations as cells transformed to the gene expression pattern in the MSC-derived fibrosarcomas, naturally occurring soft tissue tumors, and age-related cancers of humans, focusing on the six essential traits that dictate malignant growth ( 20). stMSCs acquire alterations in growth factor production (Fgf1), factors governing extracellular matrix modulation, invasion, and metastasis (MMP9, Timp1, Timp3, Spl1, Itga5, and Muc1), and aberrant functioning of tumor suppressor genes [p21, p53, and itga5 (disrupted Her-2 signaling)]. We also found reactivation of Klf4, Oct4, Sox2, and c-myc, factors responsible for maintaining stem cell pluripotency ( 21), which are commonly found overexpressed in aggressive cancers (summarized in Table 2). Immunofluorescence staining for p53 produced a similar pattern of distribution in the spontaneous tumors derived from ROSA-positive BM and the tumors derived from the stMSC ( Fig. 3C ). p53 was elevated in the stMSC lines consistent with a mutated protein ( Fig. 3B) and lower levels were detected in the late-passage nontransformed cells. Early-passage cells had WT sequence and no stainable protein ( Fig. 3A and B, top). Late-passage cells had both WT and a 307 Thr-Ile mutation with rare p53-positive cells by immunofluorescence ( Fig. 3A and B, second row panels). Five distinct point mutations in the DNA-binding domain were found in late-passage stMSC line A3. Mutations at 250 Thr-Ala (human position 253), 248 Ile-Asn (human position 251), and 307 Thr-Ile (human position 312) are clinically relevant mutations, found frequently in a variety of human cancers, whereas 247 Pro-Arg and 253 Thr-Ser do not seem to be functionally significant. Subcloning of A3 shows 307 Thr-Ile mutation to be the dominant mutation. B2 and A2 stMSC lines, isolated 4 months apart, both contained the 307 Thr-Ile mutation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

p53 alterations in naturally occurring and experimental fibrosarcomas. A, p53 immunofluorescence (IF; green) in MSC cultures. B, Western blot analysis of protein in the corresponding cell lines; mutation by sequence analysis listed below blot. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. C, tissue immunofluorescence of p53 in naturally occurring fibrosarcoma (top) and stMSC-induced fibrosarcoma (bottom).

stMSCs are cancer stem cells and form aggressive tumors in immunocompetent C57BL/6 mice. Single cells were isolated and grown as clonal populations; three distinct lines were used for the experiments presented here. Five million cells were implanted s.c. on the right upper back of male or female mice. Twenty of 20 mice displayed visible palpable tumors 6 to 10 days after implantation. At 14 days, most tumors exceeded 7 mm, and by 28 days, the majority of tumors ulcerated through the overlying skin necessitating euthanasia. Serial dilutions of cells were injected to determine the minimum cell number for tumor formation. As few as five cells formed tumors, which first became palpable from 38 to 98 days after implantation (5 of 10 mice). There were no differences in tumor characteristics between tumors derived from 5 or 5 million cells; all were histologically similar, composed of whirls of thin spindyloid cells arranged in a herringbone pattern ( Fig. 1C), and similar in appearance to naturally occurring fibrosarcomas ( Fig. 1A). Tumors contained adipose cells ( Fig. 1C, second box, arrow) and developed an extensive vascular network ( Fig. 1C, third panel, arrows, vascular channels containing blood). Tumor cells stain positive for α-SMA ( Fig. 1C, fourth panel, arrows) and negative for desmin, vimentin, S100, CD34, CD31, and cytokeratin (negative data not shown). CD31 identified tumor vasculature but did not stain tumor cells. Tumors were highly aggressive, invading muscle, overlying skin, and vascular structures ( Fig. 2C, second row). Large vessels contained tumor thrombi but there was neither gross nor light microscopic evidence of metastatic lesions to the liver, lung, brain, spleen, or regional lymph nodes. We next examined if tumor cells could be serially passaged from host to host, fulfilling a criteria of tumor stem cells. Tumor fragments isolated at necropsy were minced, cultured in vitro for several weeks before reimplantation, or directly reimplanted into secondary hosts. These cells formed tumors with the same frequency and characteristics as the parent cell line. The process was repeated through eight serial hosts without alteration in tumor-forming ability or growth characteristics.

stMSCs recruit host BMDC to tumors and fuse with host cells. To address the relative contribution of stMSC to the tumor and tumor stroma, we transfected stMSC-A3 cells with a plasmid carrying the hygromycin resistance gene along with RFP driven by the cytomegalovirus promoter (pDs-Red-monomer-hyg-C1; cell line termed RFP-stMST-A3), isolated single-cell clones, and confirmed that RFP-stMSC-A3 and A3 cells had identical growth characteristics in culture. RFP-stMSC-A3 cells or unlabeled stMSC-A3 cells were injected into female C57BL/6 mice. Excised tumors were evaluated by regular light and fluorescent microscopy. Dual-photon imaging and conventional visible light confocal imaging were used to image tumor cells within live animals. Tumor size and frequency were the same for both the stMSC-A3 and the RFP-stMSC-A3 cells, producing palpable tumors in 6 to 10 days in 100% of mice. RFP was contained in intracytoplasmic perinuclear vacuoles of varying sizes within tumor cells. Many of the cells within the tumor were RFP positive; however, a large number of cells did not express RFP protein but were morphologically indistinguishable from the RFP-positive cells. To determine if this represented a loss of RFP expression from tumor cells, or if RFP-negative cells within the tumor were of host origin, we implanted RFP-stMSC-A3 cells within two distinct GFP-labeled hosts. Five million male RFP-stMSC-A3 tumor cells were injected into female C57BL/6-GFP mice, ubiquitously expressing GFP driven by the chicken actin promoter, providing us with a model whereby all host tissue expresses GFP, or into female C57BL/6 mice transplanted with marrow from female C57BL/6-GFP mice such that only BMDCs carry the GFP label. Tumors from GFP mice and mice transplanted with GFP BM were evaluated by conventional fluorescent microscopy, in vivo confocal microscopy, or in vivo two-photon microscopy.

Host BM cells readily home to the tumor and remain separate from, or fuse with, the RFP-positive tumor cells. BM-derived GFP cells within the area of tumor are detectable as early as day 3 after tumor cell injection. These cells are thin spindyloid cells phenotypically indistinguishable from the RFP-stMSC-A3 cells. In both models, 10% to 20% of the cells within the tumor contained both GFP and RFP, suggesting fusion between RFP-stMSC-A3 and host BM-derived MSC ( Fig. 4A , live in vivo confocal imaging and Supplementary Material). Large tumors oftentimes had necrotic centers. At the center of these tumors, RFP cells remained viable with unchanged appearance when compared with cells in the periphery of the tumor; however, GFP-positive BMDCs and fused cells either had become smaller round cells or had fragmented, suggesting host BMDC and fused cells may not tolerate the relative hypoxia of the necrotic tumor center as well as the RFP cells ( Fig. 3A, second panel and Supplementary Material).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Fusion and transdifferentiation of stMSC in vivo. A, in vivo confocal imaging of RFP-stMSC in a GFP host at the tumor periphery (first panel) and necrotic tumor center (second panel) and blood vessel within the tumor (third panel) and peritumoral fat (fourth panel). Arrows, fused cells in fat. B, cells of a disaggregated tumor placed in culture have unfused (first panel: thick arrow, GFP host cell; thin arrow, RFP stMSC) and fused (second panel: arrows, two fused cells) cells. C, X-FISH (red) and Y-FISH (green) images of tumor stroma superimposed on light microscopic view for identification of tissue structure. XY chromosome within endothelial cell (top box) and Y chromosome within adipose cell (bottom box). D, coexpression of endothelial marker CD31 (FITC, green) with RFP in endothelial cells (first panel) and coexpression of fat marker FABP4 (FITC, green) with RFP in fat cells (second panel, arrow).

Fusion of RFP-stMSC to host cells results in loss of malignant phenotype. To test the growth characteristics of the different populations of tumor cells, RFP-stMSC-A3 cells were again implanted in GFP⁺ transgenic mice or C57BL/6 mice transplanted with GFP⁺ BM. Tumors were enzymatically digested and mechanically disaggregated and analyzed immediately for fluorescent protein expression by FACS and fluorescent microscopy or placed in culture and analyzed after 3 to 10 days. Fluorescent microscopy of cultured cells shows distinct color signals in unfused cells ( Fig. 4B, first panel: thick arrow, GFP; thin arrow, RFP) and the merged color signals in fused cells ( Fig. 4B, second panel: arrows, two fused cells). FACS analysis of cell suspensions into GFP⁺, RFP⁺, and GFP⁺/RFP⁺ populations shows that, of the cells retaining fluorescent protein expression after isolation, half were RFP positive and ∼25% were either GFP or GFP/RFP expressing. As the time in culture progressed, the RFP⁺ population increased and the GFP⁺ and RFP⁺/GFP⁺ populations declined, suggesting a growth advantage of the RFP⁺ cells (Supplementary Material). When placed separately in culture, host-derived GFP⁺ cells and fused GFP⁺/RFP⁺ were viable with a fibroblast phenotype and proliferated very slowly; neither formed tumors when injected s.c. into mice (n = 10). In contrast, RFP-positive cells proliferate rapidly in culture, form foci, and establish tumors when injected into mice, supporting the notion that unfused RFP-stMSC-A3 cells are the cancer stem cell in this model and fused cells and host BMDC function as cancer-associated fibroblasts (CAF) and stromal cells. Data obtained from the GFP⁺ and the C57BL/6 transplanted with GFP⁺ marrow gave similar results, suggesting that the majority of host cells within the tumor are BM derived.

Fused cells contain DNA of both the host and the stMSC line. GFP and RFP can be released from dying cells into surrounding tissue and phagocytosed by neighboring viable cells. This is an unlikely explanation for our findings because both the GFP and RFP are seen within live tissue, in tissue sections by fluorescent microscopy, in disaggregated whole tumor cells, as well as in tumor cells propagated in culture. It is expected that GFP and RFP expression would be lost in tissue culture if its presence were due solely to bystander uptake. However, we used two different approaches to confirm that the RFP/GFP signal was due to fusion of cells. First, PCR analysis of GFP⁺, RFP⁺, and GFP⁺/RFP⁺ cell populations confirmed that the GFP⁺/RFP⁺ population contained both GFP and RFP DNA sequences. Second, pure populations of cells were grown with or without hygromycin (resistance gene carried by the RFP vector). As expected, both the RFP⁺ and the GFP/RFP⁺ cells were resistant to hygromycin, whereas GFP-positive cells were susceptible to hygromycin, confirming the integrity of the RFP cassette within the fused cells.

stMSCs directly or through fusion transdifferentiate into tumor-associated adipocytes and endothelial cells. BM cells have been shown to contribute to tumor stroma ( 13, 16– 18) in other experimental systems and in our model system presented here. We next determined the contribution of stMSC-A3 cells to tumor vasculature and tumor adipocytes (see also Supplementary Material). X chromosome FISH (X-FISH) and Y-FISH analysis identified Y chromosomes within endothelial cells ( Fig. 4C, top box, FITC-labeled Y-FISH and Texas red–labeled X-FISH show X and Y chromosome within the nuclei of an endothelial cell) and adipose cells ( Fig. 4C, bottom box, FITC-labeled Y-FISH and Texas red–labeled X-FISH show the Y chromosome within the nuclei of an adipocyte; X chromosome not detected). To determine if RFP-stMSC-A3 tumor cells differentiate directly to endothelium or adipose tissue, or fuse with host marrow-derived precursor cells, we examined RFP-stMSC-A3 tumors in both the GFP⁺ host and the C57BL/6 mouse transplanted with GFP marrow. Our data support that endothelial structures are composed of three distinct populations of cells. The majority of endothelial structures were derived from peripheral host cells or host marrow-derived cells, and between 1% and 10% of endothelial cells are derived from stMSC with the percent varying widely between mice. Fused endothelial cells were found predominantly within smaller vascular structures (Supplementary Material). In vivo confocal imaging identified endothelial structures composed of a combination of host and RFP-stMSC-A3 cells interspersed and entire structures composed of stMSC-A3 cells ( Fig. 4A, third panel). In vivo confocal imaging of fat surrounding tumors shows fat cells to also be composed of three distinct populations: GFP-positive marrow-derived cells, RFP-stMSCA3–derived cells, and RFP-stMSC-A3 fused with marrow-derived host cells ( Fig. 4A, fourth panel). The majority of the fat within and surrounding the tumors originates from fused cells ( Fig. 4A, fourth panel, arrows). We next verified that the vascular and endothelial structures containing RFP granules expressed appropriate lineage markers. Using RFP-stMSC-A3 cells in a C57BL/6 host, colocalization of RFP and anti-CD31-FITC or anti-FABP4-FITC shows lineage appropriate protein expression for RFP-stMSC-A3–derived endothelial cells and fat, respectively ( Fig. 4D).

Discussion

Previous work from our laboratory shows that gastric cancer arises from BMDC in the mouse model of Helicobacter-induced gastric cancer ( 3). Whereas our in vivo experiments used whole BM, our in vitro experiments suggested that a cell in the MSC pool was responsible for initiating cancer. Indeed, malignant degeneration of MSC has been shown in vitro by other groups ( 9, 11) and, in some reports, occurs in the absence of carcinogen exposure or experimental manipulation (other than culturing), suggesting that MSC may be the target cell of in vivo transformation. However, it is not clear if events in vitro mirror in vivo events; the demonstration that in vitro–transformed MSCs form tumors in vivo is suggestive of but does not prove that MSCs transform in vivo to form tumors. The experiments described here show that, in the C57BL/6 mouse, aging alone is associated with the formation of BM-derived fibrosarcomas. Culture-expanded MSCs undergo similar gene expression changes seen in native tumors, supporting the notion of in vivo transformation of a BM-derived MSC. In culture, forced replication of mouse MSC results in spontaneous transformation of cells with a phenotype and surface marker profile similar to in vivo fibrosarcomas. Unlike most tumors derived from cell lines, stMSC-derived tumors are complex in structure with extensive vascular networks and tumor stroma, resembling naturally occurring tumors. Most importantly, when placed in an immunocompetent model, these cells form tumors indistinguishable from fibrosarcomas occurring in the aged mouse by direct contribution to the tumor mass and via recruitment and fusion with host BMDC to form cells of the supporting stroma. This model recapitulates recruitment of CAFs from the BM, which have substantial regulatory effect on tumor growth (reviewed in ref. 22), providing a model to further study the role of CAFs in tumor formation and progression. Our model is consistent with the reports from animal models of chronic inflammation where BMDC contributes to both tumors ( 3) and tumor stroma ( 18) directly or through fusion with host cells ( 19). Graft-versus-host disease after human BM transplant has been associated with increased peripheral stem cell engraftment and increased tumor formation, supporting peripheral growth of BM stem cells may lead to transformation.

The role for fusion between BMDC and peripheral tissue cells has been debated, with several studies supporting a role for fusion in phenotypic determination especially in organs known to contain multinucleated cells, such as the liver and muscle. Other studies do not support a role for fusion and instead assign phenotypic alterations of BMDC to genetic reprogramming or transdifferentiation of cells based on environmental signals. Although it is difficult to reconcile seemingly disparate results, BMDC fusion with peripheral tissues or other BMDC may depend on multiple factors, including the cell types involved, the local environment within the tissue, or the genetic composition of both fusion partners. The effect of fusion seems to depend heavily on the experimental system used, and fusion may contribute very different phenotypes depending on the situation. In our model system, fusion restores a nontransformed phenotype to cells, as fused cells are unable to form secondary tumors in mice and cannot be propagated extensively in culture. In this model, cancer stem cells are recovered as RFP-positive cells and can be propagated in culture as well as passaged through serial immunocompetent mice. Conversely, one can envision fusion of a BMDC with a damaged peripheral cell. Although the peripheral cell likely lacks the longevity for tumor formation, it is conceivable that fusion with a long-lived stem cell could provide the immortality to a dysregulated cell, thus forming cancer.

In summary, we show that MSCs spontaneously form fibrosarcomas in mice as a result of aging. Genetic changes, such as p53 mutations, and gene expression changes common to de novo cancer formation occur in cultured MSC as a result of proliferation, strengthening the support for MSC as the origin of cancer in the mouse model. What remains to be studied is whether MSCs residing in the periphery accumulate damage and transform or cells that have collected sufficient mutations are in circulation and, once a suitable niche is found, are able to engraft and initiate tumor formation. The implications for human disease are substantial, as identifying the mechanisms for growth control within these cells will yield novel targets and approaches to cancer therapy. This cell line model, which forms complex tumors in vivo, will be useful for testing anticancer therapies and identifying mechanisms for cell invasion and differentiation. Determining conditions that favor formation of other complex tumors will drastically alter our approach to testing therapeutic agents in animal models.