Mesenchymal Stem Cells from Human Fat Engineered to Secrete BMP4 Are Nononcogenic, Suppress Brain Cancer, and Prolong Survival

Glioblastoma is the most common and aggressive malignant primary intracranial neoplasm in adults (1). The median survival for patients with glioblastoma is approximately 14.6 months, despite aggressive combinatorial treatment (1). The malignant nature of glioblastoma and its ability to resist multimodal treatments have been attributed to its highly proliferative and migratory ability as well as its heterogeneous cell composition (2–6). This heterogeneity is theorized to be because of a small population of stem-like progenitor cells called brain tumor initiating cells (BTIC; refs. 2–6). BTICs are highly resistant to chemotherapy and radiation therapy, and may underlie the high recurrence rate and treatment failure observed in patients with glioblastoma (2–6). Therefore, therapies directly targeting BTICs might be more effective than current therapies.

Mesenchymal stem cells (MSC) are clonally expansive with the capacity to differentiate into osteocytes, adipocytes, and chondrocytes under specific in vitro stimuli (7, 8). Commonly used types of MSCs are bone marrow–derived MSCs (BM-MSC) and human adipose–derived MSCs (hAMSC; refs. 7 and 9). MSC's intrinsic ability to home to tumors, ease of isolation from various tissues, and ability to readily expand in vitro make them attractive candidates to deliver specific, targeted cancer therapeutics (9–15). The effects of MSCs on tumor cells in vivo, however, remain incompletely characterized, and seem to depend heavily on cancer type and source of MSCs (14, 15). Unlike BM-MSCs, hAMSCs are easier to obtain, more genetically and morphologically stable in long-term culture, have a lower senescence ratio, and have a greater proliferative capacity (7, 9). Of the limited number of studies that have evaluated the effect of hAMSCs on commercial glioblastoma cell lines, some found these cells reduced tumor recurrence and had an overall tumor-suppressive effect (11, 16, 17). However, there have also been reports of MSCs transforming into tumor-associated fibroblasts (TAF), which can potentially support tumor growth and promote a malignant phenotype (18–20). Yet, no studies have evaluated the effects of hAMSCs on human BTICs in vivo with a primary cell line. Furthermore, no studies have reported the changes that may occur in hAMSCs after they interact with human BTICs.

Because of their capability to target glioblastoma cells, hAMSCs can be used to deliver therapeutic agents to glioblastoma (9, 21–23). Bone morphogenetic protein 4 (BMP4) is a potential therapeutic agent that has been shown to have an antiproliferative effect on neural progenitor cells (24–28), and, more recently, has been shown to significantly decrease the proliferation of stem-like, tumor-initiation precursors of glioblastomas as well as drive the differentiation of these cells toward a predominantly glial fate (29). These findings make BMP4 a promising treatment for glioblastoma, but no studies thus far have investigated its therapeutic potential or its ability to be delivered via stem cells (29). The goals of this study were to investigate the interaction between BTICs and hAMSCs-BMP4 and the reciprocal effects of each cell type on the other's proliferation, differentiation, and migration. Furthermore, we investigated the effect of hAMSCs-BMP4 on survival in a mouse model of glioblastoma. These interactions are paramount to understanding the utility of hAMSCs and BMP4 to treat glioblastoma in human clinical trials.

Early passage hAMSCs and BTIC cultures were used and authenticated by Johns Hopkins Genetic Resources Core Facility. hAMSCs (Invitrogen, R7788-115) were cultured in MesenPRO complete media (1% antibiotic/antimycotic (Invitrogen, 15240-062), 1% Glutamax (GIBCO, 35050-061), 1 vial of MesenPRO RS growth supplement (GIBCO, 12748-018), and MesenPRO RS basal media (GIBCO, 12747-010)). Human BTIC cultures (276 and 612) were obtained from intraoperative tissue (as approved by Johns Hopkins Institutional Review Board) and cultured in laminin-coated flasks (Sigma, L2020, 1 μg/cm²) with stem cell media (30). As previously validated and shown by our group, the human BTIC cultures are able to form oncospheres, are multipotential, and form tumors when implanted into animal models (30). To evaluate the tumorigenic capacity of BTICs in vivo, BTIC line 276 was injected intracranially in mice by our group resulting in the formation of solid tumors, whereas 612 formed diffuse tumors (30, 31). The molecular subtype of BTIC culture 276 is mesenchymal and 612 is proneural, which was determined using a metagene score based approach for subtype designation, assessing 4 mesenchymal and 2 proneural genes using a microfluidics based quantitative PCR assay (32). Commercial U87 cells (ATCC, HTB-14) were cultured in Dulbecco's modified Eagle medium media with 10% FBS.

To induce the expression of BMP4 in hAMSCs, an MFG-based retroviral vector system was combined with a BMP2/4 hybrid (33). To identify hAMSCs and BTICs in our in vitro coculture and in vivo mouse experiments, we transduced these cells with lentiviral vectors coding for GFP, td-tomato, or GFP/bioluminescent proteins. Viral vectors were packaged from HEK293 cells. After collection and concentration, hAMSCs (hAMSCs-Vector, hAMSCs-BMP4, GFP/bioluminescent-hAMSCs, and td-tomato-hAMSCs) and BTICs (GFP-276 and GFP-612) were infected and sorted by a MoFlo cytometer (Beckman Coulter).

To investigate the effect and the safety of coinjected hAMSCs on glioblastoma cell proliferation in vivo, 6- to 8-week NOD/SCID mice were stereotactically injected with 0.5 × 10⁶ GFP-276 (n = 5), 0.5 × 10⁶ td-tomato-hAMSCs (n = 5), 0.5 × 10⁶ GFP-276 mixed with 0.5 × 10⁶ td-tomato-hAMSCs (coinjection; n = 5), and 1.0 × 10⁶ GFP/bioluminescent-hAMSCs into the right striatum (L: 1.34 mm, A: 1.5 mm, D: 3.5 mm; n = 5). Following injection, mice in the GFP/bioluminescent-hAMSC group were imaged using an IVIS small animal imaging system (Perkin Elmer) at different time periods (7, 14, and 28 days postinjection). After 4 weeks, animals were euthanized and perfused with 4% PFA. Brains were extracted, cryo-sectioned, and immunostained for human nuclei (Millipore, MAB4383). To quantify tumor area, tumor mass was outlined based on 4′,6-diamidino-2-phenylindole (DAPI) staining and calculated using Image J. To quantify glioblastoma cell migration, the distance from the tumor margin, determined by DAPI staining, to each human nuclei⁺/DAPI⁺/td-tomato⁻ cell was quantified. A blinded observer performed all counting.

To determine the effect of hAMSCs-BMP4 on the malignancy of orthotopic glioblastoma tumors in vivo, 1 × 10⁶ BTIC 276 were suspended in 2 μL PBS and stereotactically injected into the right striatum (L: 1.34 mm, A: 1.5 mm, D: 3.5 mm) of immunosuppressed nude mice. Two weeks postinjection, 0.5 × 10⁶ GFP-hAMSCs-Vector (n = 7), GFP-hAMSCs-BMP4 (n = 5), or equal volume of PBS (n = 5, 100 μL) were systemically injected into the left cardiac ventricle (34). After 2 weeks, the mice brains were perfused, fully cryo-sectioned at a 10 μm thickness, and immunostained for GFP, BMP4 (Abcam, ab93939), Ki67 (Thermo, RM-9106-s1), Nestin (Abcam, ab5922), Tuj1, GFAP, TNF-α (Abcam, ab6671), VEGF (Abcam, ab46154), and human nuclei. Tumor mass was outlined and cell number of positive staining inside the tumor mass was counted and normalized relative to DAPI inside the tumor bulk. The ratio of Ki67⁺/DAPI was used to measure proliferation; Nestin⁺/DAPI, Tuj1⁺/DAPI, and GFAP⁺/DAPI to measure differentiation; and TNF-α⁺/DAPI and VEGF⁺/DAPI to measure tumor necrosis and angiogenesis. Immunostaining for Nestin was used as a BTIC marker, as this has been validated in several studies (35, 36). To quantify glioblastoma cell migration, human nuclei antibodies were used. Tumor mass was outlined by DAPI and the center of tumor mass was calculated using Image J. The distance from the center of tumor mass to each human nuclei⁺/DAPI⁺/GFP⁻ cell was quantified based on 110 to 250 cells per group. A blinded observer performed all counting. 41, 38, and 41 slides have been analyzed in PBS, hAMSCs-Vector, and hAMSCs-BMP4 groups, respectively. All in vivo procedures were approved by the Johns Hopkins University Animal Care and Use Committee.

To determine the effect of hAMSCs-BMP4 on the survival of orthotopic glioblastoma tumors-bearing mice in vivo, 0.5 × 10⁶ U87 were suspended in 2 μL PBS and stereotactically injected into the right striatum (L: 1.34 mm, A: 1.5 mm, D: 3.5 mm) of immunosuppressed nude mice. Ten days postinjection, 0.5 × 10⁶ hAMSCs-Vector (n = 7), GFP-hAMSCs-BMP4 (n = 5), or equal volume of PBS (n = 10, 100 μL) were systemically injected into the left cardiac ventricle. Mice were followed for 125 days to monitor survival. A Kaplan–Meier survival analysis was performed with results reported as the median and mean survival times with a 95% confidence interval. The statistical difference between the 3 conditions was determined by log-rank analysis.

Results are reported as mean ± SEM. Comparisons were done using two-way ANOVA for MTS assays, one-way ANOVA (Kruskal–Wallis test follow up an Dunns posttest) for transwell and nanopattern assays, A Kaplan–Meier survival analysis was performed with results reported as the median survival times with a 95% confidence interval. The statistical difference between the 3 conditions was determined by log-rank analysis, and t tests (Mann–Whitney) for other experiments using GraphPad Prism 5 software. Statistical significance was defined as P < 0.05.

Prolific migration of glioblastoma cells and BTICs greatly contributes to the mortality of the disease. Human MSCs have been shown to have intrinsic tumor suppressive affects in some models (37), and the effects of hAMSCs and BMP4 on glioblastoma and BTIC migratory capacity has not yet been evaluated. Thus, we investigated whether unmodified hAMSCs and BMP4 secreting-hAMSCs (hAMSCs-BMP4) could affect the migratory capabilities of BTICs in vitro.

hAMSCs-BMP4 cells were able to synthesize and release BMP4 as shown by Western blots performed with cell lysates and conditioned media from hAMSCs-BMP4 and hAMSCs-Vector. The majority of mature BMP4 is extracellular, and there is no visible mature BMP4 in the hAMSCs-Vector cells (Fig. 1A). As shown in Supplementary Fig. S1A, 1 × 10⁶ hAMSCs-BMP4 cells secrete approximately 74 ng after 3 days of culturing.

Using in vitro Boyden chamber transwell assays, the effect of hAMSCs-Vector, hAMSCs-BMP4, and an exogenous 50 ng/mL BMP4 dose on BTIC migration was assessed (Fig. 1C). Conditioned media from empty vector–infected hAMSCs [hAMSC-Vector-conditioned media (CM)], hAMSCs-BMP4 (hAMSC-BMP4-CM), and BMP4-supplemented media resulted in a 2-fold decrease in the number of migrating BTICs (Fig. 1C, P < 0.001). However, there were no significant differences between these 3 treatments (P > 0.05). Similar findings were seen when using a different BTIC line (BTIC 612; Supplementary Fig. S1B and S1D).

To assess the effects of hAMSCs and BMP4 on BTIC migration speed, a nanopattern chamber was used (Fig. 1D). BTIC migration speed decreased when cultured in hAMSC-Vector-CM and hAMSC-BMP4-CM, as well as in the BMP4 treatment group (50 ng/mL) by almost 2-fold for both 612 (Fig. 1E, P < 0.01) and 276 (Supplementary Fig. S1E and S1F, P < 0.05) BTICs. There was no significant difference in the ability of hAMSC-Vector-CM, hAMSC-BMP4-CM, and BMP4 to decrease the migration speed of BTICs (Fig. 1E, P > 0.05).

Another key feature underlying the malignant nature of glioblastoma is its capacity for unlimited and rapid proliferation. Previous studies have demonstrated the effects of BMP4 in reducing the proliferative capabilities of BTICs (29), but there are none that report the effects of hAMSCs on BTIC proliferation. Therefore, MTS (to test cell viability and proliferation as a function of time) and EdU assays (to examine cell proliferation at specific time points) were used to evaluate the effects of BMP4, hAMSC conditioned media, and hAMSC-BMP4 conditioned media on BTICs. Using the MTS assay, hAMSC-CM treatment resulted in no statistically significant difference in proliferation between the hAMSC-CM and the control groups during the first 10 days (P > 0.05; Fig. 2A). However, hAMSC-CM decreased proliferation of both 276 (Fig. 2A) and 612 BTICs (Supplementary Fig. S2A) significantly at day 13 (P < 0.05). In comparison, exogenous BMP4 (100 ng/mL) demonstrated a significant decrease in BTIC proliferation after day 7 for 276 (Fig. 2B) and at day 13 for 612 (Supplementary Fig. S2B, P < 0.01). To verify these results and further investigate whether cell–cell interactions between hAMSCs and BTICs (coculture EdU assay) could affect the proliferation of BTICs, we cocultured hAMSCs with BTICs and found the proliferation of 276 cells decreased significantly (Fig. 2C and Supplementary Fig. S2C, 3-fold decrease at both day 5 and day 13, P < 0.05). Furthermore, when treated with exogenous BMP4 (50 ng/mL) for 48 hours, proliferation of 276 cells decreased significantly (Fig. 2D, 30% decreased, P < 0.05). BTIC proliferation also decreased significantly when cocultured with hAMSCs-BMP4 (Fig. 2E; for 276, 2-fold decrease at 5 days, P < 0.001). Similar effects were seen using the 612 BTIC line (Supplementary Fig. S2D and S2E). These experiments demonstrate that BMP4 and hAMSCs-BMP4 decrease the proliferation of BTICs effectively in vitro, and that unaltered hAMSCs can also decrease the proliferation of BTICs via secreted proteins and cell–cell contact.

BTICs seem to underlie the ability for glioblastoma migration and proliferation; inducing the differentiation of BTICs may attenuate the malignant features of glioblastoma. Previous studies have shown that BMP4 can induce differentiation of BTICs (29). We verified this in our experimental system and further examined the effects of hAMSC-secreted molecules and hAMSCs-BMP4 on BTIC differentiation potential. The ability of hAMSCs and hAMSCs-BMP4 to induce differentiation of BTICs was examined. BTICs were cultured in different media conditions and immunofluorescent staining was used to identify different lineage markers (Fig. 3A). Compared with the negative control (undifferentiated BTICs, cultured in stem cell media), the percentages of Tuj1⁺/DAPI and GFAP⁺/DAPI were both significantly increased in hAMSC-Vector-CM, hAMSC-BMP4-CM, and BMP4 treatment groups as well as in the positive control group (differentiated BTICs, cultured in stem cell media + 10% FBS) for 276 (P < 0.001; Fig. 3B). Similar results were obtained using 612 (Supplementary Fig. S3). Although the BTICs express GFAP or Tuj1 proteins, we observed many double positive cells, which would indicate maturation of the cells. However, the BTICs have not begun to take on the classic neuronal and astrocytic morphologies, which could be because of the timing of our experiments.

To consider hAMSCs as delivery vehicles in glioblastoma treatment, they must retain certain intrinsic characteristics when exposed to glioblastoma or glioblastoma tumor environment. The exposure of hAMSCs to BTIC-CM did not alter their proliferation capacity via MTS assay (276-CM were used in Fig. 4A, P > 0.05). This result was replicated when the effects of BTIC–hAMSC cell–cell interaction was examined by EdU coculture assays (Fig. 4B). The percentage of proliferating hAMSCs when cultured alone was not significantly different compared with those cocultured with BTICs (32% vs. 26%, P > 0.05). These results were replicated using BTIC line 612 (Supplementary Fig. S4A and S4B).

The ability of hAMSCs to remain multipotent upon exposure to BTIC-secreted factors was examined using differentiation assays. Similar to the negative control (undifferentiated hAMSCs), hAMSCs incubated in 276 BTIC-CM stained negative for adipogenesis (first row of Fig. 4D) and osteogenesis (second row of Fig. 4D). Although there was presence of Masson's Trichrome staining in the 276-CM and negative control groups, the positive control group (differentiated hAMSCs) had appreciably more staining (third row of Fig. 4D). Concurrently, mRNA expression levels were used to quantify adipogenic (CEBPA and LPL), osteogenic (OP and ALPL), and chondrogenic (SOX9) differentiation. Normalized and compared with the negative control group (undifferentiated hAMSCs), mRNA levels of all markers in the conditioned media group were not significantly different (P > 0.05; Fig. 4C). Similar results were obtained using 612 conditioned media (data not shown).

In addition, the proliferation, differentiation, and migration of hAMSCs-Vector and hAMSCs-BMP4 were examined to evaluate if retroviral modification altered hAMSC characteristics. In regard to proliferation, the effect of BMP4 on hAMSC proliferation was examined as a positive control. BMP4 significantly reduced the proliferation of hAMSCs by approximately 10% in a time-dependent manner (Fig. 4E, P < 0.05). BMP4-secreting hAMSCs also showed a decrease in proliferation when compared to hAMSCs-Vector suggesting an auto- or paracrine effect (Fig. 4F, P < 0.05). In regard to differentiation, the differentiation potential of hAMSCs was not altered when treated with exogenous BMP4 or when they were induced to express BMP4 (hAMSCs-BMP4; Fig. 4G). In regard to migration, hAMSCs-Vector and hAMSCs-BMP4 cells migrated toward BTIC-CM more than to control media indicating their preserved glioblastoma tropism (276, Fig. 4H and I, P < 0.05; 612, data not shown, P < 0.05). Engineered hAMSCs respond to and migrate toward factors secreted by BTICs, with a more pronounced response seen in hAMSCs-BMP4 (2-fold increase in hAMSCs-BMP4 group, and 1.1-fold increase in hAMSCs-Vector group), suggesting that modified hAMSCs might result in a more targeted therapy. BMP4 transduction may also enhance the migration ability of hAMSCs in vivo.

The potential of hAMSCs to undergo malignant transformation into TAFs and undergo subsequent increased tumor growth and migration when exposed to BTIC secreted factors was examined. The expression of TAF markers (increased expression levels of vimentin and ACTA2) was evaluated by Western blot analysis and real-time RT-PCR. There were no significant differences in protein levels up to 2 weeks, and there was even less ACTA2 expression after 3 weeks of culturing in BTIC-CM (276, Fig. 5A; and 612, data not shown). Similarly, there were no significant differences in mRNA levels after 1 to 3 weeks of culturing in BTIC-CM (276, P > 0.05, shown in Fig. 5B; 612, P > 0.05, data not shown). Moreover, there were no differences in vimentin or ACTA2 immunofluorescence staining when exposed in BTIC-CM after 3 weeks (data not shown). hAMSCs endogenously express vimentin and ACTA2, and exposure to BTIC-CM did not increase expression of these markers.

The proliferative capacity and survival time of hAMSCs was subsequently examined in vivo after intracranial coinjection of BTICs and hAMSCs (Fig. 5C). The GFP/bioluminescent-hAMSCs (GFP-hAMSC) group was imaged from 7 to 28 days postinjection. The bioluminescent signal declined dramatically after 14 days and remained virtually nonexistent in vivo (Fig. 5E). There was no observable GFP signal (hAMSCs) in the GFP-hAMSCs group at 3 months in vivo (Supplementary Fig. S5). In addition, there was no observable td-tomato (hAMSCs) in the coinjection group (data not shown). Even after staining for GFP (data not shown) and human nuclei, colocalization of human nuclei and GFP was only found in the GFP-276 and coinjection groups and no signals in GFP-hAMSCs or PBS groups (data not shown). After 4 weeks, the coinjection group had a smaller mean tumor area of 135,700 μm² as compared with GFP-BTIC group, with a mean tumor area of 209,800 μm² (P = 0.0189; Fig. 5D). When measuring individual cell distance from the tumor margin, there was no significant difference between the BTICs-only and coinjection groups (P = 0.3442; Supplementary Fig. S5C). In addition, as shown in Fig. 5F, tumors in the BTIC group seem larger than the coinjection group, suggesting that the presence of additional hAMSCs does not contribute to rampant tumor progression in vivo.

To examine the effects of hAMSCs-BMP4 on glioblastoma cell-proliferative capacity and migratory ability, and on stem-ness of BTICs in vivo, a mouse model of glioblastoma was created as described previously by our group (38). In this model, hAMSCs-Vector (n = 7), hAMSCs-BMP4 (n = 5), or PBS (n = 5) were administered via cardiac injection after glioblastoma tumor formation (Fig. 6F). As shown in Fig. 6A, there was only BMP4 (human specific antibodies) seen in mice injected with hAMSCs-BMP4. Subsequently, GFP staining confirmed the homing of the hAMSCs-Vector and hAMSCs-BMP4 groups to the tumor bulk. As shown in Fig. 6A, C, and D, hAMSCs-Vector and hAMSCs-BMP4 migrated to the tumor bulk (defined by DAPI density). Interestingly, in only the hAMSCs-BMP4 group, GFP signals were found around migratory BTICs (defined by Nestin⁺ cells not part of the tumor bulk; Fig. 6B). Human-specific Ki67 staining was used to assess proliferation, and no Ki67⁺ cells were observed colocalizing with GFP⁺ hAMSCs (Fig. 6C). To quantify the effect of hAMSCs-Vector and hAMSCs-BMP4 on glioblastoma cell proliferation, Ki67⁺ cells were normalized with corresponding DAPI⁺ cells in the tumor mass. There was a 2-fold decrease in the ratio of Ki67⁺/DAPI in the hAMSCs-BMP4 group compared with both the PBS group and hAMSCs-Vector group (Fig. 6C, P < 0.05). In addition, immunofluorescence staining for TNF-α and VEGF were performed to investigate characteristics of malignant tumors (TNF-α is a marker for necrosis, and VEGF is a proangiogenic molecule; TNF-α and VEGF secretion are also known associated with TAFs). As seen in Supplementary Fig. S6, there were decreased TNF-α and VEGF staining in the hAMSCs-Vector and hAMSCs-BMP4 groups (2-fold of decrease, P < 0.05)

The ability of hAMSCs-Vector and hAMSCs-BMP4 to induce differentiation of BTICs in vivo were evaluated by staining cells for Nestin, GFAP, and Tuj1. There was an increased number of Nestin⁺ cells in the PBS group, and an increased number of GFAP⁺ and Tuj1⁺ cells in the hAMSCs-Vector and hAMSCs-BMP4 groups (P < 0.05; Fig. 6D). We also observed that hAMSCs-BMP4 can decrease the number of Nestin⁺ cells compared with the hAMSCs-Vector group.

In addition, to determine if hAMSCs-Vector and hAMSCs-BMP4 can affect the migration of glioblastoma cells in vivo, the tumor bulk was outlined utilizing DAPI staining. The average distance of glioblastoma cells (human nuclei⁺/DAPI⁺/GFP⁻ cells) that migrated from the center of tumor bulk was calculated based on human nuclei staining (Fig. 6E). As shown in Fig. 6E, hAMSCs-Vector and hAMSCs-BMP4 both inhibited the migratory ability of glioblastoma cells significantly (P < 0.001). In addition, as compared with the hAMSCs-Vector group, hAMSCs-BMP4 significantly decreased the migration of glioblastomas (Fig. 6E, P < 0.001).

To investigate if hAMSCs-Vector and hAMSCs-BMP4 can affect the survival of glioblastoma bearing mice, 0.5 × 10⁶ U87 cells were stereotactically injected into immunosuppressed nude mice. Ten days postinjection, 0.5 × 10⁶ hAMSCs-Vector (n = 7), GFP-hAMSCs-BMP4 (n = 5), or equal volume of PBS (100 μL, n = 10) were systemically injected into the left cardiac ventricle. Mice were followed for 125 days to monitor survival. As shown in Fig. 6G, the median survival of mice treated with hAMSCs-BMP4 (undefined) was significantly greater than that of mice treated with hAMSCs-Vector (P = 0.01; 76 days) and control mice (P = 0.002; 52 days). There was no significant difference between the PBS and hAMSCs-Vector groups (P = 0.09).

Glioblastoma is the most common and aggressive malignant primary intracranial neoplasm in adults, with a median survival of approximately 14.6 months despite combinatorial treatments of surgical resection, chemotherapy, and radiotherapy (39). Glioblastoma has heterogeneous genetic alterations in pathways associated with proliferation, survival, invasion, and angiogenesis. Glioblastoma cells are known to use white matter tracts and microvasculature basement membranes to migrate long distances, making complete surgical resection of the tumor difficult, almost inevitably leading to recurrence (40). The well-known glioblastoma molecular subtype classifications are proneural, neural, classical, and mesenchymal (41). The 2 primary glioblastoma cell lines used in this study, 276 and 612, belong to mesenchymal and proneural subtypes, respectively. hAMSCs-BMP4 treatment was able to attenuate malignant tumor characteristics of 2 different subtypes of glioblastoma in this study, reinforcing the therapeutic effect of hAMSCs-BMP4 in potential future clinical trials.

Although human MSCs have been manipulated to express a wide variety of anticancer therapeutic factors because of their tropism toward inflammation and tumor cells (17, 42–44), the effects of hAMSCs on glioblastoma and BTICs have not been fully described. This study was the first to find that hAMSCs inhibit proliferation and induce differentiation of BTICs, as well as confirm that hAMSCs decrease the migration of BTICs in vitro. Furthermore, hAMSCs can induce differentiation, reduce proliferation and migration, and may even diminish angiogenesis of glioblastoma in vivo when injected intracardially. However, when intracranially coinjected with BTICs, we did not observe a difference in the extent of cell migration in vivo, but we did find a reduction in tumor size. These results indicate that hAMSCs can have intrinsic antitumor effects and are promising for glioblastoma treatment. However, we found that unmodified hAMSCs have a limited effect on inhibiting glioblastoma cell proliferation and survival, and they can only track glioblastoma tumor bulk but lack the ability to home to migratory glioblastoma cells in vivo. Enhancement of the hAMSCs by engineering them to deliver specific agents may augment their anticancer effects (37).

BMPs are known to play a role in the differentiation of adult neural stem cells into different mature cell types (45, 46). Recently, BMP4 has been shown to reduce glioblastoma tumor burden in vivo and improve survival in a mouse model of glioblastoma by potentially reducing the frequency of symmetric cell divisions or by blocking proliferation and inducing differentiation of BTICs (29). One of the challenges to effective treatment of glioblastoma is the targeting of BTICs, which seem to underlie the ability of the tumor to recur (2). BMP4 is an ideal therapeutic candidate because of its affect on BTICs; however, optimizing its delivery is critical (47). Aside from local delivery with polyacrylic beads (29), there are no reports describing stem cell-based vehicles for BMP4 delivery. The goal of the in vitro experiments was to demonstrate the potential therapeutic effects hAMSCs-BMP4 have on glioblastoma. We found that BMP4 treatment, whether exogenously administered or released by genetically modified hAMSCs, can decrease the proliferation of BTICs and make BTICs commit to mature lineages in vitro and in vivo. Most of these effects are also observed with unmodified hAMSCs, but to a lesser extent in vivo. Although there were no significant differences between the hAMSCs-Vector and hAMSCs-BMP4 treatment groups in regard to migration, proliferation, and differentiations in vitro, differences were noted in vivo with regard to glioblastoma proliferation, nestin expression, and glioblastoma migration (Table 1). Interestingly, hAMSCs-BMP4 and exogenous BMP4 can reduce migration and migration speed of BTICs, unlike the effect of BMP4 on other types of cancers (24–28). In addition, hAMSCs-BMP4 not only display tropism toward glioblastoma tumor bulk, but can also home to migratory glioblastoma cells in vivo. Most importantly, a single, cardiac injection of 1 million hAMSCs-BMP4 significantly increases survival of glioblastoma-bearing mice compared with the hAMSCs-Vector and PBS treatments. Commercial glioblastoma cell line U87 was used for the survival study because it is well-established and commonly used in glioblastoma survival studies and has been shown to be extremely aggressive with a low survival rate, making it ideal to study survival in a murine glioblastoma model (48–50). Future studies would be interested to use different subtypes of patient-derived BTICs, including proneural, mesenchymal, classical, and neural, to perform survival studies and to investigate the molecular mechanisms behind the therapeutic effect of hAMSCs-BMP4. These studies will be a promising step toward personalized glioblastoma therapy.

The present literature raises several additional concerns. The effects of cancer cells on the proliferative capacity and malignant potential of human MSCs is a critical consideration for its potential utility in clinical trials (9–11, 14). It has also been proposed that cancer cells may be able to induce MSCs to form TAFs, which can then support and stimulate tumor growth and migration as well as promote a malignant phenotype (18–20). We found that hAMSC proliferation does not increase in response to BTIC-secreted factors or coculturing with BTICs in vitro. When cultured in conditioned media from BTICs in vitro, hAMSCs did not upregulate their expression of fibroblast markers, suggesting that the BTIC–hAMSC interaction does not foster the adoption of a TAF phenotype. In vivo, hAMSCs were not detectable 14 days after intracranial injection, suggesting that they are functional for only a short window of time. This has both advantages and disadvantages. The advantage of having this window of time is minimization of potential deleterious effects of cell therapy, including oncogenesis, tumor induction, and neovascularization. The disadvantages include need for repeated cell injections, similar to chemotherapy. Despite the short window of time, however, a single cardiac injection of 1 million hAMSCs-BMP4 was able to significantly prolong survival of glioblastoma-bearing mice. When injected alone and with BTICs, hAMSCs also did not demonstrate an increased proliferative capacity and no hAMSCs remained in the brain 3 months after injection. Furthermore, in our model of glioblastoma, hAMSCs delivered to established tumors did not demonstrate colocalization with Ki67 and markers of tumor growth (TNF-α and VEGF; refs. 51 and 52) were attenuated in the presence of hAMSCs-BMP4. These results suggest that hAMSCs are neither tumor-supportive nor tumorigenic. In addition, we demonstrated that BTICs and BMP4 do not alter the stem cell properties, tumor tropism, or induce differentiation of hAMSCs in our experiments. Moreover, these cells retained their stem cell–like characteristics and tumor tropism in vivo, which is fundamental to their utility as a vehicle for antitumor agents. Two recent studies found that hAMSCs promoted growth and angiogenesis of glioblastoma cells in vivo (53). However, we discovered that hAMSCs-BMP4 decrease the proliferation of glioblastoma cells, and hAMSCs-Vector and hAMSCs-BMP4 decrease the expression of the angiogenesis markers, VEGF and CD31 (data not shown) in vivo, although there was not sufficient quality and number of cells immunoreactive to CD31 to allow for quantification and statistical analyses. hAMSCs may therefore contribute to better outcomes in a multifactorial fashion, of which, angiogenesis is one component. Other components include tumor proliferation, migration, and differentiation, among others. Differences between our tumor models and those of Akimoto and colleagues in cell type, MSC source, injection site, and timing of MSC injections may account for these contrasting results. Notably, this prior study cotransplanted MSCs and glioblastoma cells subcutaneously. Our experiments with intracranial injections were meant to be a proof the principle that ensured delivery of hAMSCs to the tumor mass and evaluate the safety of intracranial injection. We also delivered MSCs systemically through a single, cardiac injection to an established intracranial glioblastoma tumor mass, which is thought to be a more accurate model for glioblastoma in human patients.

In conclusion, our results demonstrate the extraordinary ability of hAMSCs-BMP4 to decrease the proliferative and migratory capacity of glioblastoma cells, induce differentiation of BTICs in vitro and in vivo, and ultimately prolong survival glioblastoma-bearing mice with a single, cardiac injection of one million cells. In addition, our findings demonstrate the safety and efficacy of engineered hAMSCs in delivering targeted therapy in a mouse model of glioblastoma. Both unmodified hAMSCs and hAMSCs-BMP4 do not undergo malignant transformation when exposed to glioblastoma cells, and do not support tumor growth. Further advances with hAMSCs-BMP4 to create a more sophisticated delivery system may include engineering these cells to control the secretion of BMP4. TGF-β and other markers specific to glioblastoma cells within the brain (54, 55) may serve as molecular switches to induce the contextually specific release of BMP4. Based on our findings, we are optimistic that engineered hAMSC-based anticancer therapies will continue to demonstrate their promise in clinical trials for glioblastoma. In the future, we predict this stem cell–based approach will have wide-reaching potential, including autologous hAMSCs from adipose tissue and the treatment of other primary and secondary brain cancers.

No potential conflicts of interest were disclosed.

Conception and design: Q. Li, S.J. Salas, D.A. Chesler, C.L. Smith, H. Guerrero-Cazares, A. Quinones-Hinojosa

Development of methodology: S.J. Salas, M. Zhu, C. Aprhys, K.L. Chaichana, D.A. Chesler, C.L. Smith, A. Levchenko, A. Quinones-Hinojosa

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Q. Li, O. Wijesekera, S.J. Salas, J.Y. Wang, M. Zhu, K.L. Chaichana, H. Zhang, C.L. Smith, A. Levchenko, A. Quinones-Hinojosa

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Q. Li, O. Wijesekera, S.J. Salas, J.Y. Wang, M. Zhu, K.L. Chaichana, D.A. Chesler, C.L. Smith, H. Guerrero-Cazares, A. Levchenko, A. Quinones-Hinojosa

Writing, review, and/or revision of the manuscript: Q. Li, O. Wijesekera, J.Y. Wang, K.L. Chaichana, D.A. Chesler, C.L. Smith, H. Guerrero-Cazares, A. Levchenko, A. Quinones-Hinojosa

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Q. Li, S.J. Salas, M. Zhu, A. Levchenko, A. Quinones-Hinojosa

Study supervision: S.J. Salas, H. Guerrero-Cazares, A. Levchenko, A. Quinones-Hinojosa

We thank Dr. K.H. William Lau from Loma Linda University for the gift of the BMP2/4 DNA plasmid, Bryant Beltran for designing the MATLAB program to track cells using the nanopattern assay, and Dr. K. Aldape and Dr. K. Wani from the University of Texas MD Anderson Cancer Center for the glioblastoma cell classification.

This research was supported by the National Institutes of Health (RO1, NS070024) and the Maryland Stem Cell Research Fund (A. Quinones-Hinojosa).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.