Passage-dependent cancerous transformation of human mesenchymal stem cells under carcinogenic hypoxia

Cell culture

hMSCs (Lonza, Walkersville, MD, USA) were maintained in α-modified Eagle's medium (αMEM; Gibco Cell Culture, Carlsbad, CA, USA) with 20% fetal bovine serum (FBS; Gibco) with 1% penicillin-streptomycin (P/S; Gibco) at 37°C and 5% CO₂. For all experiments, cells were seeded at a density of 5000 cells/cm². Treatment conditions included sterile-filtered nickel sulfate (100 μM, Sigma-Aldrich, St. Louis, MO, USA) and/or hypoxia (0.5% O₂, 5% CO₂, 94.5% N₂) in a hypoxia incubator chamber (Stemcell Technologies, Vancouver, BC, Canada). Cells were cultured for 3 d before endpoint evaluations. The two passage groups, passage 7–9 (P7–9) and passage 12+ (P12+), were selected to represent early and advanced stages of abnormality acquisition and senescence, respectively (15).

Western blot

Western blot analysis was done according to standard protocols (34). Primary antibodies (1:1000) include mouse anti-human p53 (recognizes both mutant and wild-type forms of p53; Abcam, Cambridge, MA, USA), mouse anti-human HIF1A (Abcam), rabbit anti-human c-MYC (generously provided by Dr. Steve Hann, Vanderbilt University Medical Center), and mouse anti-human β-actin (Santa Cruz Biotechnology, Santa Cruz, CA, USA). Horseradish peroxidase (HRP)-conjugated secondary antibodies (1:1000, Santa Cruz Biotechnology) were used; relative protein expression against β-actin was quantified via densitometric analysis using ImageJ software (U.S. National Institutes of Health, Bethesda, MD, USA; n≥3 independent experiments).

Cell proliferation

Cells were incubated with 5-bromo-2′-deoxyuridine (BrdU, Sigma) at 20 μM for 16 h, fixed in 4% paraformaldehyde (PFA; Sigma), and blocked with 1% bovine serum albumin (BSA; Sigma). DNA was denatured with HCl (2 N for 12 min on ice, 1 N for 20 min at 37°C) and neutralized with borate buffer (pH 9.0). Incorporated BrdU in proliferating cells was detected by rat anti-BrdU antibody overnight at 4°C (1:100; Abcam), followed by secondary DyLight594-conjugated goat anti-rat antibody for 2 h at room temperature in the dark (1:50, Jackson Immunoresearch, West Grove, PA, USA). Nuclei were counterstained with Hoechst (5 μg/ml; Sigma) and cells were imaged under a Nikon Eclipse Ti inverted fluorescence microscope (Nikon Instruments Inc., Melville, NY, USA). Cell proliferation was represented by the percentage of BrdU-positive cells relative to the total number of cells per image (n≥6 images).

ROS activities

To measure intracellular hydrogen peroxide (H₂O₂) or superoxide (O₂^·−), hMSCs were incubated with dichlorofluorescein diacetate (DCFDA; Molecular Probes, Carlsbad, CA, USA) or dihydroethidium (DHE; Molecular Probes), respectively, for 30 min at 5 μg/ml at 37°C. Cell nuclei were counterstained with Hoechst. Fluorescence intensity was measured on a plate reader (Tecan Group Ltd., Männedorf, Switzerland) and normalized to the corresponding cell number (excitation/emission: DCFDA, 490/540 nm; DHE, 340/535 nm; Hoechst, 360/488 nm). To visualize and quantify mitochondrial superoxide (mROS), cells were incubated with MitoTracker Red (CM-H₂XROS; Invitrogen, Carlsbad, CA, USA) for 30 min at 2.5 μg/ml. The average fluorescence intensity for individual cells was calculated using ImageJ. Reduction potential was quantified by measuring the reduction of the resazurin sodium salt (Sigma) to resofurin, following the manufacturer's instructions. Cells were incubated with resazurin at 5μM for 2 h at 37°C. Reduction activity was measured on a plate reader (Tecan; excitation/emission: 535/595 nm) and normalized to cell number. For all experiments, n ≥ 3 replicates.

Caspase-3 activity

Caspase-3 activity was quantified with a colorimetric assay kit (BioVision, Mountain View, CA, USA) according to the manufacturer's protocol. Experiments were conducted in triplicate, and the average normalized optical density was reported.

β-Galactosidase staining

Staining was performed with the senescence β-galactosidase staining kit from Cell Signaling Inc. (Danvers, MA, USA) according to the manufacturer's protocol.

Flow cytometry

Cells were fixed with 4% PFA, blocked with 1% BSA, and incubated with fluorophore-conjugated primary antibodies in 1% BSA for 2 h (1:10, eBiosciences, San Diego, CA, USA). Antibodies included FITC-conjugated CD90 and PE-conjugated CD105. Tests were performed using a FACScalibur flow cytometer (Becton Dickson, Franklin Lakes, NJ, USA), with an unstained hMSC population containing both passage groups as a control. Data were analyzed with FlowJo (Tree Star, Ashland, OR, USA), and reported percentages were calculated by gating the desired cell population on the forward/side scatter plot and setting a fluorescence exclusion for >99% of the control population.

Reverse transcriptase–polymerase chain reaction (RT-PCR)

RNA was harvested with QIAshredder (Qiagen, Valencia, CA, USA) and RNeasy column (Qiagen) following the manufacturer's protocol. cDNA was synthesized with the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA, USA), and RT-PCR was performed using the iQ SYBR Green supermix (Bio-Rad) on a Bio-Rad CFX Real Time Instrument for glyceraldehyde 3-phosphate dehydrogenase (GAPDH), vascular endothelial growth factor (VEGF), HIF1A, c-MYC, matrix metalloproteinase-2 (MMP-2), membrane type I MMP (MMP-14), and tissue inhibitor of metalloproteinase-2 (TIMP-2). β-Actin was used as a reference gene. MMP sequences were obtained from a previous study (35). Primer sequences used were as follows: GAPDH, 5′-GCCTCAAGATCATCAGCAATG-3′ (sense) and 5′-CTTCCACGATACCAAAGTTGTC-3′ (antisense); VEGF, 5′-CAGAAGGAGGAGGGCAGAATC-3′ (sense) and 5′-CACACAGGATGGCTTGAAGATG-3′ (antisense); p53, 5′-GCGTGTGGAGTATTTGGATGAC-3′ (sense) and 5′-AGTGTGATGATGGTGAGGATGG-3′ (antisense); c-MYC, 5′-CGTCCTCGGATTCTCTGCTCTC-3′ (sense) and 5′-CCTCATCTTCTTGTTCCTCCTCAG-3′ (antisense); HIF1A, 5′-TACCCTAACTAGCCGAGGAAGAAC-3′ (sense) and 5′-ATCAGCACCAAGCAGGTCATAG-3′ (antisense); β-actin, 5′-ATCATGTTTGAGACCTTCAA-3′ (sense) and 5′-CATCTCTTGCTCGAAGTCCA-3′ (antisense). Data were analyzed with CFX Manager software (Bio-Rad), and results are presented as ΔΔC_t expression normalized to β-actin.

Osteogenic and adipogenic induction

Osteogenic and adipogenic induction media were purchased from Lonza. For adipogenesis, hMSCs were cycled between adipogenic induction and maintenance media with or without the NADPH oxidase inhibitor, diphenyleneiodinium chloride (DPI; 4 μM, Sigma) for a total of 15 d (3 d induction, 1 d maintenance, 3 d induction with or without DPI, 3 d maintenance, 3 d induction with or without DPI, and 2 d maintenance). Osteogenic induction medium was changed every 3–4 d for a total of 15 d, during which time DPI treatment occurred for a total of 6 d. To verify differentiation, alkaline phosphatase (ALP) activity and Nile red staining were performed for osteogenic and adipogenic differentiation, respectively. ALP activity was quantified using the p-nitrophenyl phosphate liquid substrate system (Sigma) according to the manufacturer's instructions (n=3). For Nile red staining, cells were fixed with 4% PFA and incubated with 100 ng/ml Nile red in 150 mM NaCl for 10 min. Cells were then washed, counterstained with Hoechst, and imaged using a Nikon Ti inverted microscope. The green fluorescence channel is shown due to improved specificity (36).

SDS-PAGE gelatin zymography

The proenzyme and active forms of MMP-2 and MMP-9 were detected using SDS-PAGE gelatin zymography in the presence of SDS in 10% polyacrylamide gels containing 0.1% (w/v) gelatin (37). Medium samples were collected from cells incubated in serum-free DMEM 24 h prior to collection. Total protein concentrations were determined by Bradford assay, and the samples were diluted 1:1 in sample buffer containing 0.14M Tris-HCl (pH 6.8), 20% (v/v) glycerol, 4.5% (w/v) SDS, and 0.003% bromphenol blue. Protein (20 μg) was loaded into each well. Electrophoresis was performed in nonreducing conditions in a Bio-Rad Mini Protean 3 system. Proteins were renatured by washing the gels in four 10-min incubations in 2.5% Triton X-100. For gelatinolytic activity, the gels were incubated at 37°C for 18 h in 50 mM Tris-HCl (pH 7.3; Sigma) containing 1.47 mg/ml CaCl₂ (Sigma) and 0.05% Brij-35 (Fisher Scientific, Pittsburgh, PA, USA). The proteins were fixed for 2 h with gentle agitation in 30% methanol and 10% acetic acid at room temperature, followed by staining for 18 h with 0.1% (w/v) Brilliant Blue R-250, 0.64 M ammonium sulfate (Sigma), and 25% methanol. The gels were destained in 25% methanol for 2 h.

In vivo experiments and immunohistochemistry (IHC)

Animal housing and experimentation were performed according to the Vanderbilt University Institutional Animal Care and Use Committee guidelines under an approved protocol. For experiments, 8 × 10⁵ hMSCs from each group were injected subcutaneously with 200 μl Matrigel (BD Biosciences, San Jose, CA, USA) into the scapular region of athymic Foxn1^nu/nu female mice (4–6 wk of age, n=3 mice/group, 12 mice total; Harlan Laboratories, Indianapolis, IN, USA) and allowed to grow for 1 mo. Each week, animals were anesthetized with 2.5% isoflurane under steady flow and weighed, and tumor nodules were measured with electronic calipers. After 1 mo, animals were sacrificed by cervical dislocation under anesthesia and tumor samples were placed in 4% PFA for 24 h before being submitted for embedding, sectioning, and IHC. All IHC was performed by the Translational Pathology Shared Resource Core at Vanderbilt University Medical Center. Slides were deparaffinized and antigen retrieval was performed using standard methods. Slides were stained for 30–60 min with primary antibodies against murine F4/80 (1:1000 dilution; Novus Biologicals, Littleton, CO, USA), Ki67 (1:100; Abcam), CD31 (1:100; Dianova, Hamburg, Germany), and human S100 (1:900; Dako, Carpinteria, CA, USA) followed by appropriate HRP-conjugated secondary antibodies and detection with 3,3′-diaminobenzidine (DAB) substrate according to standard protocols. IHC slides were imaged under a Nikon stereoscope.

Statistical analysis

In all experiments, results are presented as means ± sem. Results from each experiment were initially analyzed using single-factor analysis of variance (ANOVA), and comparisons between individual test groups were then performed using an unpaired Student's t test. Values of P < 0.05 were considered statistically significant.

Osteogenic and adipogenic differentiation potential

To verify differentiation capacity of both passage groups, the osteogenic and adipogenic differentiation potentials of P7–9 and P12+ hMSCs were first examined (Fig. 1). hMSCs from both passage groups underwent osteogenic differentiation, but ALP activity was significantly higher in P7–9 cells (Fig. 1A), suggesting that late-passage cells have a diminished capacity for osteogenic differentiation. Cells from both passage groups exhibited a strong capacity for adipogenic differentiation, as evidenced by the presence of lipid droplets (Fig. 1B). Because ROS are known to play a key role in modulating hMSC differentiation (38, 39) DPI was added to cells periodically to verify ROS responsiveness of both passage groups. Inhibition of ROS production resulted in enhanced osteogenic differentiation (Fig. 1A) but prevented adipogenic differentiation (Fig. 1B). These findings are in agreement with the literature (38, 39) and confirm that both passage groups exhibit ROS-sensitive, multilineage differentiation potential.

Effects of NiSO₄ and hypoxia

To establish which conditions best mimic a cancerous environment, the effects of hypoxia with or without NiSO₄ treatment were investigated (Fig. 2). To first investigate the effects of hypoxia and/or Ni on hMSCs, cells of passage 5–15 were used such that the effect of passage was not considered at this point. As seen in Fig. 2A, NiSO₄ treatment in normoxia (T, N) reduced cell number compared to the untreated normoxia (U, N) condition. However, the untreated hypoxic (U, H) condition increased cell number vs. normoxia alone, and NiSO₄ treatment in hypoxia (T, H) led to a further increase in cell number. Interestingly, the cell number in each condition correlated linearly with intracellular H₂O₂ content (R² = 0.9474), suggesting a role of H₂O₂ in regulating cell proliferation.

Expression of cancer-related proteins was evaluated in each condition (Fig. 2B). In untreated normoxia, expression levels for HIF1A and c-MYC were very low, and p53 expression was moderate. On treatment with NiSO₄ in normoxia, the expression of c-MYC and p53 proteins increased as compared to untreated normoxia, and this correlated with a drastic reduction in cell number (Fig. 2A), verifying that the interplay of p53 and c-MYC regulates proliferation and apoptosis. In untreated hypoxia, HIF1A and c-MYC protein expression were up-regulated as compared to untreated normoxia, which correlated with increased cell number and supports the role of these transcription factors in accelerating proliferation. P53 expression did not appear to change between these two untreated conditions. Finally, NiSO₄ treatment in hypoxia stimulated a strong increase in HIF1A expression as compared to untreated hypoxia, indicating the synergistic role of nickel and hypoxia in accumulating HIF1A. In addition, c-MYC expression remained strong in this condition, and p53 expression did not appear to change, as compared to the untreated, normoxic condition. Taken together, the NiSO₄-treated hypoxia condition most closely resembled a common cancer profile (i.e., strong expression of HIF1A and c-MYC with moderate p53 expression). We therefore identified 100 μM NiSO₄ treatment under 0.5% O₂ hypoxia as the most effective condition for transformation induction (TI). Control was defined as the normoxic untreated condition (0 M NiSO₄, 21% O₂). As seen in Table 1, the relative expression levels of the three cancer-related proteins were more rigorously investigated between the control and TI conditions (n≥3 independent experiments). In the TI condition, HIF1A and c-MYC expression were significantly up-regulated while p53 expression did not statistically change (Table 1), confirming that the hMSCs in the TI condition exhibit an expression profile that generally mimics that of cancer cells.

Passage-dependent response to the TI condition

The passage-dependent response of hMSCs to the TI condition was then characterized in terms of proliferation, apoptosis, target protein and mRNA expression, and immune phenotype. The TI condition reduced proliferation of P7–9 hMSCs but stimulated proliferation in P12+ cells as compared to their respective controls (Fig. 3A). Furthermore, activity of the apoptosis-related protein, caspase-3, was increased in the TI condition for P7–9 cells but decreased significantly for P12+ hMSCs (Fig. 3B).

The two passage groups also exhibited different expression patterns for cancer-related proteins (Fig. 3C). In the control condition, basal expression of c-MYC and HIF1A was higher for P12+ cells, but expression of p53 was much lower as compared to P7–9 hMSCs. Interestingly, P7–9 expression of c-MYC was virtually nonexistent in the control condition, but c-MYC was expressed strongly in P12+ control cells. In the TI condition, expression of all three proteins increased for both passage groups, as compared to their respective controls.

The two passage groups also exhibited substantial differences at the level of gene expression (Fig. 3D–H). Expression of the glycolytic enzyme GAPDH increased significantly in the TI condition for both passage groups (Fig. 3D), and these results agree with literature findings regarding an increased dependence on glycolysis by hMSCs in hypoxia (24). c-MYC gene expression also increased significantly for both passage groups in the TI condition, but p53 expression did not change (Fig. 3E, F). Interestingly, p53 gene expression was higher for both conditions in P12+ cells, as compared to P7–9 cells, but p53 protein expression was actually lower for P12+ cells (Fig. 3C), suggesting a drastic difference in post-transcriptional regulation between the two passage groups. In the TI condition, HIF1A gene expression decreased for both passage groups compared to control (Fig. 3G). However, gene expression of the HIF1A-regulated VEGF did not change in the TI condition for P7–9 cells, but increased significantly for P12+ cells (Fig. 3H), indicating that the hypoxia-responsive expression program was indeed activated.

We then sought to identify the status of replicative senescence in both passage groups as well as the role of the TI condition in modulating this behavior (Fig. 4A). In the control condition, both passage groups stained positive for β-galactosidase expression (Fig. 4A, blue cells), indicating replicative senescence and supporting previous findings in the literature (15). In the TI condition, hMSCs from both passage groups effectively escaped the senescent state, as indicated by a drastic reduction in the number of blue-stained cells.

The expression level of the positive hMSC markers CD90 and CD105 was then investigated by flow cytometric analysis (Fig. 4B, and Table 2C). Unstained cells are represented by the red, leftmost curve (Fig. 4B and Table 2). Control hMSCs from both passage groups strongly expressed both markers. Surprisingly, when quantified, P12+ cells showed stronger expression of these markers, indicating an effective maintenance of stemness into later passage (Table 2). In the TI condition, CD90 and CD105 expression decreased in both passage groups, but when quantified, P12+ cells were shown to retain their stemness to a higher degree than P7–9 cells. These data indicate that the TI condition stimulated differentiation, or significant down-regulation, of hMSC-positive markers in P7–9 cells that was not observed to the same degree in P12+ cells.

ROS levels and oxidative capacity

ROS are known to play crucial roles in both healthy tissue maintenance and cancer (13). The conversion of O₂^·− to H₂O₂ is critical for maintaining healthy redox homeostasis and accumulated O₂^·− correlates with hMSC senescence (14). Therefore, the oxidative status of the two passage groups of cells was next investigated (Fig. 5). In the TI condition, intracellular H₂O₂ levels decreased for P7–9 cells but increased significantly in P12+ hMSCs (Fig. 5A). In parallel, mROS were significantly decreased in the TI condition in P7–9 cells but were unaltered in P12+ cells (Fig. 5B), and intracellular O₂^·− levels were significantly higher for all P12+ cells, as compared to all P7–9 cells (Fig. 5C). These data reveal that O₂^·− is not efficiently converted to H₂O₂ in P12+ cells and suggest a loss in reduction potential in P12+ hMSCs as compared to P7–9 cells. As seen in Fig. 5D, P7–9 cells in the harsh TI condition maintained strong ability for reduction, whereas P12+ cells exhibited a drastic decline in reduction potential. These data indicate that ROS-handling mechanisms (i.e., antioxidative activity/reduction potential) become disrupted in late passage hMSCs, and dysfunctional redox potential follows as an explanation for increased susceptibility to the TI condition, which is not observed in lower passage, P7–9 cells.

MMP gene expression and gelatinolytic activity

ROS levels have been shown to influence the activities of MMPs, a class of proteases responsible for degradation and remodeling of the extracellular matrix (ECM; ref. 40). MMPs are known to regulate fundamental hMSC functions, including trafficking, differentiation, and invasive capacity (41, 42), and are necessary for cancer invasion and metastasis (43, 44). We therefore looked at the role of ROS-mediated MMP activity in the two passage groups of hMSCs in response to the control and TI conditions. Specifically, the expression of MMP-2, MMP-14, and TIMP-2 at the gene and protein levels were investigated, as well as the role of ROS in modulating their activities, by adding the NADPH oxidase inhibitor, DPI. These proteins were selected because cooperative MMP-14 and TIMP-2 activity are required to activate the proenzyme form of MMP-2 (45). TIMP-2 plays a dual role, as it can either localize with MMP-14 or bind directly with MMP-2 to positively or negatively regulate MMP-2 activation, respectively (45).

Zymography revealed that MMP-2 protein expression was altered in response to TI, DPI, and passage (Fig. 5E). Pro-MMP-9 expression was consistent in all P7–9 conditions but decreased slightly for P12+ cells in TI and TI+DPI conditions, as compared to control. For both passage groups, strong expression of both pro- and active MMP-2 was observed in the control condition. However, in the TI condition, the two passage groups exhibited different responses: in P7–9 cells, MMP-2 activation was drastically down-regulated, but in P12+ cells, the active MMP-2 level remained high. This difference can be attributed to up-regulated MMP-2 gene expression in the P12+ TI condition that was not observed for P7–9 cells (Fig. 5F). In the TI+DPI condition, MMP-2 activation was inhibited in both passage groups and pro-MMP-2 levels were significantly down-regulated, indicating a causative role of ROS in MMP-2 stability and activation. Interestingly, at the gene level, DPI treatment stimulated opposite responses in the two passage groups (Fig. 5F–H). In P7–9 cells, MMP-2 and TIMP-2 expression did not change significantly, but MMP-14 expression was up-regulated. In contrast, in P12+ cells, MMP-14 expression did not change, but MMP-2 and TIMP-2 expression were significantly up-regulated. These data suggest that the mechanism by which DPI attenuates the expression and stability of pro- and active-MMP-2 are not the same for the two passage groups. Specifically, P7–9 cells did not up-regulate MMP-2 at the gene level, suggesting that regulation occurred at the transcription level (14). In contrast, significant up-regulation of both MMP-2 and TIMP-2 gene expression was observed in P12+ cells, while MMP-14 expression remained unaltered, suggesting that post-transcriptional activities and, potentially, TIMP-2-mediated inactivation of MMP-2 are responsible for regulating MMP-2 protein expression in late passage cells.

In vivo consequences following hMSC implantation

When hMSCs were subcutaneously implanted into nude mice, tumor-like nodules formed in all animals within 1 mo, although no significant difference in tumor sizes was found among groups. On resection, nodules from the untreated control groups (P7–9 control, P12+ control), as well as the carcinogenic hypoxia-treated P7–9 TI group, were identical in gross morphology, exhibiting little to no vasculature and a homogeneous structure that resembled cartilaginous tissue. In contrast, nodules removed from all animals in which carcinogenic hypoxia-treated P12+ hMSCs were implanted contained small, white structures embedded within the cartilaginous tissue, which suggests a stark difference in composition on initial observation. Histological analyses, including hematoxylin and eosin (H&E) staining and immunodetection of CD31, Ki67, and F4/80, revealed specific differences among the groups (Fig. 6). In all groups, H&E-stained sections of the subcutaneous implants revealed well-circumscribed nodules of mesenchyme with cartilaginous differentiation (Fig. 6A), which was confirmed by positive nuclear staining by S100 antibody. Angiogenesis of the nodules was evaluated by a CD31 antibody (Fig. 6B). Although single cells were stained positively for CD31, with the strongest expression exhibited in the two P12+ groups, vessel lumen formation was not observed in any of the samples, suggesting that the nodules were not functionally vascularized.

The subcutaneous nodules were surrounded by host soft tissues consisting predominately of fat and fibrous stroma with intermixing inflammatory cells. The cartilaginous nodules from groups P7–9 control, P7–9 TI, and P12+ control were all comparable with sparse to minimal cellularity and a low proliferative index, as indicated by Ki-67 IHC (5–10% nuclear staining; Fig. 6C). The nodules from group P12+ TI were similar in overall appearance to the other groups but displayed distinct differences. Specifically, the P12+ TI nodules had a significantly higher cellularity with a markedly increased proliferative index (20–50% nuclear staining by Ki-67), suggesting that this group most closely resembled the formation of a classical tumor.

The nodules from all groups were also evaluated for the presence of macrophages by F4/80 (Fig. 6D, E). The macrophages in control groups were small and evenly distributed throughout both the nodules and surrounding stroma. In contrast, the macrophages in TI groups were decreased in number, predominately arranged in clusters within the nodules, and displayed significantly more cytoplasm. These data suggest that treatment under the TI condition stimulated an immunomodulatory effect beyond the intrinsic capability of hMSCs cultured under control conditions. Overall, P12+ TI hMSCs exhibited significantly increased cellularity, a higher proliferative index, and potential bone formation in the center of the nodules while minimizing macrophage infiltration, indicating that the in vivo result for this group could represent a neoplastic process that closely resembles that of a classical tumor, as compared with other groups.