Tumor Stroma Engraftment of Gene-Modified Mesenchymal Stem Cells as Anti-Tumor Therapy against Ovarian Cancer

IFNβ Inhibits Proliferation of Ovarian Cancer Cells in Vitro

Recombinant IFNβ protein and IFNβ-expressing MSC (MSC-IFNβ) inhibited the proliferation of ovarian carcinoma cells in vitro in a concentration-dependent manner (Fig. 1). Human OVCAR3 cells were most sensitive to human IFNβ (inhibitory concentration 50% [IC₅₀] = 5 IU/ml; Fig. 1A): these cells were approximately 20 times more sensitive than SKOV3 cells (IC₅₀ = 100 IU/ml, Fig. 1C) and approximately 200 times more sensitive than HEY cells (IC₅₀ = 1000 IU/ml; Fig. 1E). The murine ovarian carcinoma cell line (ID8-R), was also inhibited at an IC₅₀ of 50 IU/ml (Fig. 1G), utilizing murine IFNb. Both OVCAR3 and SKOV3 cells also showed evidence of apoptosis, as determined by propidium iodide staining, and inhibition of proliferation (data not shown). These in vitro results were consistent with the results obtained from coculture of human MSC-IFNβ with OVCAR3, SKOV3, and HEY cells or murine MSC-IFNβ with ID8R cells (Fig. 1B, 1D, 1F and 1H respectively).

IFNβ half-life and toxicity in vivo

Plasma IFNβ levels after IP injections of recombinant IFNβ (40,000 IU) confirmed its short biological half-life with peak plasma levels of IFNβ decreasing 2 hours post injection and undetectable after 24 hours. However, after IP injection of MSC-IFNβ, IFNβ was detectable systemically at >5 IU/ml for up to 6 days (Fig. 2A). These data demonstrate that a single IP injection of MSC-IFNβ can generate systemic IFNβ levels for extended periods in vivo. To ensure that 5 IU/ml of IFNβ was a non-toxic level of IFNβ to mice, we injected 30 IU of murine IFNβ for 10 consecutive days into 5 mice and grossly observed the mice. During this treatment, the mice exhibited no outward signs of toxicity (fur ruffling, hunched posture, anorexia, wasting; data not shown).

Mechanisms of IFNβ based killing

To determine a potential mechanism of how MSC-produced IFNβ controls tumor growth in ovarian cancer cells in vitro, OVCAR3 (Fig. 2B) or SKOV3 (Fig. 2C) cells were co-cultured with MSC-IFNβ, recombinant IFNβ protein or MSC alone in the presence or absence of a pan-caspase inhibitor. Apoptotic ovarian cancer cells are characterized as having decreased tetramethyl rhodamine methyl ester (TMRM) fluorescence, and increased Annexin V–FLUOS staining is displayed as a percent of positive control. TMRM fluorescence defined the change in mitochondrial dysfunction that accompanies cytochrome c release during apoptosis and detection of externalization of phosphatidyl serine by Annexin V defined early apoptosis. In OVCAR3 and SKOV3 cells co-cultured with MSC-IFNβ or treated with recombinant IFNβ, apoptosis increased over 48 hours (OVCAR3 45-80%; SKOV3 37-67%). In ovarian carcinoma cells cultured with MSC (control) or alone (without IFNβ), apoptosis was less than 10%. When OVCAR3 or SKOV3 were cultured with IFNβ and caspase inhibitors, apoptosis was blocked for 48 hours (<5%). These results suggest that caspase activation is an important part of apoptosis in ovarian cancer treated with IFNβ, in the absence of a significant immune component.

Ovarian carcinomas selectively engraft intraperitoneally circulating MSC

To investigate if established ovarian tumors would engraft gene-modified MSC, we injected mice IP with five once weekly dose of 5×10⁵ MSC-βgal (× 5 weeks), and their localization was traced histochemically utilizing X-gal staining. One group of mice had established abdominal OVCAR3 (n=5) or SKOV-3 (n=5) (Fig. S2a-d) tumors initiated 15 days prior. As a control, normal SCID mice without tumors (n=3) were injected with MSC-βgal. IHC was performed 14 days after the last dose of MSC-βgal was administered (49 days after the first MSC injection). As shown in Figs. S2e-h, representative sections from ovarian tumors display X-gal-positive colonies (1-3 colonies per field) that were incorporated into the tumor architecture. When MSC-βgal was injected IP into animals bearing OVCAR3 or SKOV3 tumors, there were no X-gal positive cells in the remaining tissues (spleen, kidney, muscle; Fig. S2i) Taken together, these results suggest that if MSC do engraft in normal tissues or organs, it is below our limit of detection, whereas engraftment of MSC into the tumor microenvironment was robust.

Fate of tumor resident MSC

To determine the fate of tumor engrafted MSC, we surveyed tumors from animals that received MSC. IHC staining for, human α-smooth muscle actin, human desmin, FSP, or FAP (Fig. 3) was performed on 82-day-old SKOV3 tumors of mice that had received five IP injections (1× per week) of 1 million MSC. Most CD105⁺ (undifferentiated) MSC were detected on the periphery of the tumor, forming an encapsulating network. In contrast, MSC within and throughout the tumor became positive for α-SMA and desmin, exhibiting characteristics of smooth muscle-like myofibroblast cells that participate in fibrovascular networks. FSP, and/or FAP expression also identifies activated fibroblasts within the tumor microenvironment that has been infiltrated by reactive stroma.

MSC-IFNβ Modulate Tumor Growth in Xenograft Mouse Model In Vivo

Once it was demonstrated that MSC would engraft in established ovarian tumors, we tested the efficacy of MSC-IFNβ in vivo, utilizing a SCID mouse xenograft model. Ovarian carcinomas were established by IP injection of OVCAR3 (5×10⁶) or SKOV3 (6×10⁶) cells) and 15 days later, 5×10⁵ MSC-IFNβ (n=10) were administered IP once per week for five weeks. Control animals received once weekly (×5) IP injections of 5×10⁵ MSC-βgal (n=5), or no injections (tumor only) (n=5). The animals were monitored until death, and the differences in overall survival were analyzed by log-rank test. In mice bearing established OVCAR3 xenografts, five once weekly treatments with MSC-IFNβ not only controlled tumor growth but significantly extended survival (p=0.03; Fig. 4A) as compared with OVCAR3 control mice (p=0.01). In fact, in 70% (7/10) of OVCAR3 mice, we were unable to detect any tumor development after MSC-IFNβ treatment. These mice remained tumor free for 169 days, at which point, the remaining 7 mice were sacrificed and evaluated for evidence of disease. Although no residual disease was detected, we cannot decisively claim that the tumor was completely eradicated. In mice bearing established SKOV3 xenografts, five once per week injections of MSC-IFNβ also inhibited tumor growth and significantly prolonged survival (p=0.001), as compared with control SKOV3 tumor-bearing animals alone (p=0.01) (Fig. 4B). These results correlate strongly with the in vitro sensitivity of OVCAR3 cells to IFN-β.

Detection of IFNβ secreted by MSC in ovarian tumors

Immunohistochemical staining for IFNβ was performed on OVCAR and SKOV3 (Fig. 4C) tumors 1 or 3 days after IP injection of MSC-IFNβ. Strongly positive staining in the tumors, just 1 day after MSC-IFNβ injection, was detected. Interestingly, a 20-24 -fold increase in the levels of IFNβ production (OVCAR3 2.6-63%; SKOV3 2.1-43%) on day 3 was observed and associated with intense staining throughout the entire tumor. This observation suggests an increase in localized MSC after 3 days leading to an increase in total IFNβ production within the tumor microenvironment.

Murine MSC-IFNβ Modulate Tumor Growth in syngenic Mouse Model In Vivo

Next, we tested whether murine ovarian tumors would selectively recruit circulating gene-modified murine MSC (mMSC) as stromal precursors, and whether these cells could also control tumor progression, similar to what we observed in the xenotransplant models. We therefore utilized a syngenic mouse model whereby immunocompetent C56Bl/6J mice bearing ID8-R, a murine ovarian carcinoma stably expressing renilla luciferase, were treated with either murine MSC expressing murine IFNβ (mMSC-mIFNβ; n=5) or mMSC expressing firefly luciferase (mMSC-ffLuc; control cells; n=5) or murine-derived mouse embryomic fibroblast expressing firefly luciferase (C56Bl/6J MEF; n=3). We labeled all mMSC with firefly luciferase to facilitate non-invasive detection of biodistribution and engraftment into the tumor. By utilizing selective substrates, one can visualize either ID8-R tumor (renilla luciferase) or mMSC (firefly luciferase) in co-localization or as disparate entities. As shown in Fig. 5, both established ID8-R tumors or mMSC were readily detected in mice, and treatment of ID8-R tumors with mMSC-mIFNβ resulted in a significant reduction (1 log) of tumor-associated bioluminescence at day 53 over that of ID8-R treated with mMSC-ffLuc as determined by IVIS (Fig. 5A and 5C). Importantly, bioluminescence from both ID8-R and mMSC co-localized in tumor-bearing mice, as visualized by bioluminescence imaging (BLI; Figs. 5C,5D, and 5F), but we observed a random distribution of MEFs throughout the abdomen, with no specific association with the tumor bioluminescence (Fig. 5F). These data were further validated by IHC staining of the ID-8R tumors for ffLuc confirming the presence of labeled mMSC within the tumor whereas labeled-MEFs were not selectively localized to the tumor (data not shown). The BLI photon counts were graphed over time to show the ID8R tumor progression by luminescence compared to the tumor progression of the mice receiving the MSC-βgal or MSC-IFNβ. Immediately following the onset of IP MSC-IFNβ injection, the BLI decrease significantly compared to the controls (Fig. 5E).

Taken together, these data suggest that both hMSC and mMSC preferentially engraft in their respective tumor microenvironments. Thus the MSC deliver species-specific mIFNβ locally to the tumor microenvironment, resulting in reduction of tumor size and prolonged survival.

Cell isolation and culture

Human MSC were isolated as previously described ²⁴. Human ovarian cancer OVCAR cells were a gift from Dr. Judith K. Wolf (Department of Gynecologic Oncology, The University of Texas M. D. Anderson Cancer Center, Houston, TX). The cells were maintained in Roswell Park Memorial Institute medium (RPMI-1640) supplemented with 10% fetal bovine serum (FBS), L-glutamine, and a penicillin-streptomycin mixture (Gibco/Invitrogen, Carlsbad, CA). Human ovarian cancer SKOV3 cells were obtained from the American Type Culture Collection (Manassas, VA) and cultured in minimal essential medium Earl’s salts with nonessential amino acids (supplemented with 10% FBS, L-glutamine, and a penicillin-streptomycin mixture). Human ovarian cancer HEY cells were also obtained from the American Type Culture Collection and cultured in RPMI-1640 supplemented with 10% FBS, L-glutamine, and a penicillin-streptomycin mixture.

Murine ID8 ovarian tumors: ID8, a cell line derived from spontaneous in vitro malignant transformation of C57BL/6J mouse ovarian surface epithelial cells were obtained from Dr. Katherine Roby⁴², and were maintained in Dulbecco’s modified Eagle’s medium (Invitrogen, Carlsbad, CA) supplemented with 4% fetal bovine serum, 100 U/ml penicillin, 100 g/ml streptomycin, 5 g/ml insulin, 5 g/ml transferrin, and 5 ng/ml sodium selenite (Roche, Indianapolis, IN) in a 5% CO2 atmosphere at 37° C. These cells were stably infected with a lentivirus expressing renilla luciferase (RENLuc) and GFP ³³. Post-transfection GFP+ ID8 (ID8-R) cells were enriched by FACS (98% GFP+), grown to confluence, and used in subsequent experiments.

Murine MSC: Male, C57BL/6J, mouse marrow stromal cells (mMSC) were obtained from Darwin Prockop (Tulane University, New Orleans, LA). mMSC were cultured in alpha MEM with 10% horse serum, 10% fetal calf serum and supplemented with 10% FBS, L-glutamine, and a penicillin-streptomycin mixture. Passages 5-8 were used throughout this experiment. C57BL/6J MEFs were isolated from mice embryos using standard methods.

Adenoviral vectors and MSC transduction

Ad-IFNβ was created as previously described ²⁴. Fiber modified adenoviruses expressing either murine IFNβ (AdmIFNβ-F/K21) or firefly luciferase (Ad-FF/luc-F/K21) were generated as described in Yotnda et al ³⁴. mMSC were incubated with adenoviruses at 50 viral particles/cell (based on OD reading) for 4 hours, after which fresh media was applied. The next day mMSC were assessed for transgene production. For Ad-IFNβ 50vp/cell resulted in the production of 4×10⁴ IU IFNβ per 5×10⁵ MSC after 24 hours. For Ad-mIFNβ, 50vp/cell resulted in production of 5 IU muIFNβ/ per 1×10⁶ MSC after 24 hours. Firefly luciferase (FFLuc) expression was determined by addition of 2ul of a 40mg/ml stock solution D-Luciferin (Xenogen Co, Temecula CA) in 1 ml of media. Cells were visualized and quantitated by the Xenogen 100 IVIS. β-gal expression was determined by histochemical staining, and >90% of MSC were X-gal positive.

In vitro proliferation assay

Cell monolayers were washed with PBS, harvested with trypsin EDTA, and resuspended in the appropriate medium as described previously. Cells were plated at a density of 1000 (HEY & ID8-R), 3000 (SKOV3) or 4000 (OVCAR3) cells per well (on the basis of their growth rate) in 200 μl of medium. Cells were allowed to adhere to the plate overnight, after which human IFNβ (Avonex, Biogen Inc.) or murine IFNβ (mIFNβ; ID8-R cells) (Biosource, Camarillo, CA) was added to the plate in different dilutions (range from 0-10,000 IU). One plate was read by MTS (Promega Inc., Madison, WI) assay at the time of initial addition of IFNβ, to serve as the initial control. Eight wells were used for each dilution of IFNβ. Medium was changed daily, and after five days, the assay was read using MTS. Absorbance was measured at 490nm. Results were calculated as follows: relative proliferation = (OD_IFNb / OD_CNT) where OD_CNT corresponds to A₄₉₀ of controls cells with no treatment, OD_IFNb corresponds to wells treated with different concentrations of IFNβ for 5 days.

In vitro co-culture of MSC with OVCAR3, SKOV3, and HEY

MSC were infected with an adenovirus carrying the IFNβ gene (MSC-IFNβ), to produce levels of 40,000 IU/5×10⁵ cells/ 24hours. Another flask of MSC was infected with an adenovirus carrying the beta-galactoside gene (MSC-ßgal), at levels to achieve >98% transfected cells. After 24 hours, cell monolayers were washed with PBS and removed using trypsin-EDTA; all cell lines were then resuspended in RPMI 1640 with 10% FBS. OVCAR3, SKOV3, or HEY cells were plated in 4 ml of medium either alone or mixed with MSC-IFNβ or MSC-βgal in a ratio of 1:1 or 10:1 respectively in six-well plates at a starting concentration of 4 × 10⁴ cells per well. After 5 days, cells were trypsinized, counted, and fixed with 70% ethanol. Cells were then labeled with PE (Sigma), and the cell DNA content was analyzed using the FACScan flow cytometer (Becton-Dickinson, San Jose, CA). The relative numbers of MSC (diploid cells) and ovarian carcinoma cells (aneuploid cells) were determined using ModFit software (Verity Software House Inc, ME).

Mechanisms of IFNβ based killing

OVCAR3 or SKOV3 cells were co-cultured with MSC, MSC-IFNβ (created as previously stated), recombinant IFNβ or recombinant IFNβ with a pan-caspase inhibitor (1μM Z-VAD-FMK; R&D Systems, Minneapolis, MN) in 6 well transwell plates (Corning Inc., Corning, NY). Cells were incubated with either 1000 IU/ml/24 hours recombinant IFNβ, or with the appropriate number of MSC-IFNβ, which produced approximately 1000 IU/ml/24 hours of IFNβ. Negative controls were tumor cells alone, and positive controls were tumor cells with 10,000 IU/ml/24 hours recombinant IFNβ. Apoptosis was evaluated by two-color flow cytometry after staining the cells with the fluorescent mitochondrial potentiometric probe tetramethyl-rhodamine methyl ester (TMRM), and FITC-conjugated Annexin V. After treatment, cells were collected by trypsinization, washed twice in PBS and then resuspended in 100 μl of Annexin binding buffer (140 mM NaCl, 10 mM KH₂PO₄, 5 mM CaCl₂, pH 7.4) containing 25 nM of the mitochondrial potentiometric probe TMRM (Invitrogen, Carlsbad CA) and 1:100 dilution of Annexin V-FLUOS (Roche Diagnostics, Indianapolis IN) and incubated at 37°C for 30 min. Cells were washed in Annexin binding buffer and analyzed by flow cytometry in a FACS Calibur flow cytometer (FACScan; Becton-Dickinson, San Jose CA) using a 488 nm argon excitation laser. In this assay apoptotic cells display decreased TMRM fluorescence and increased Annexin V–FLUOS staining. Data is displayed as a percent of positive control.

Animals, cell administration, non-invasive imaging of tumors and survival analysis

Female CB-17 SCID or C57BL/6J mice were purchased from Harlan (Indianapolis, IN) or Jackson Labs (Bar Harbor, ME) and used in accordance with institutional guidelines and under approved protocols. Tumors were created in the SCID or C57BL/6J mice by IP injection of either 5 × 10⁶ OVCAR3, 6 × 10⁶ SKOV3, or 5 × 10⁶ stably expressing renilla luciferase ID8 (ID8-R) cells as a suspension in 1 ml of PBS. Survival durations were measured from the day of tumor cell injection until the day of death. Tumor imaging: tumors were imaged via IVIS (Xenogen 200 system, Caliper Life sciences, Hopkington MA). Mice with detectable tumors were used, and on day 25, the first of 5 (weekly) injections of 1 × 10⁶ C57BL/6J mMSC expressed either FFLuc and mIFNβ (expressing ~ 5 IU/24h/million cells) (treatment group), or FFLuc alone (control group), or MEF expressing FFLuc (control group) IP. Bioluminescence (FFLuc detects mMSC or MEF and RENLuc detects tumor) activity was acquired and quantitated (1-2 minute collection time). Differences in survival were determined by a two-tailed log-rank test. Statistical analysis was performed with statistical software (Statistica; StatSoft, Tulsa, OK).

Measurement of IFNβ concentration in mouse plasma

Mice with established ovarian tumors were injected with either 5 × 10⁵ MSC-IFNβ IP or 40,000 IU of recombinant IFNβ IP. 200 μl of blood was collected in capillary tubes at appropriate time intervals. Mouse plasma was collected and utilized for determination of serum IFNβ levels (IFNβ-ELISA, Fujirebio Inc., Tokyo, Japan). The National Institutes of Health standard for IFNβ-1a was utilized to determine concentration.

Tissue processing and imaging studies

Tumors and other organs were fixed in Bouin’s solution or embedded in an ornithine transcarbamylase compound (OTC, Miles, Inc., Elkhart, IN), then snap-frozen in liquid nitrogen and stored at −80°C. Whole tumors from several animals were immediately examined for the presence of β-gal by using X-gal staining. Frozen tissue was sectioned (6–8 μm) and processed for hematoxylin-eosin or X-gal immunohistochemical staining. Photomicrographs were taken (Zeiss Axioplan2; Carl Zeiss, Inc., Thornwood, NY equipped with a charge-coupled device camera (Hamamatsu Corp., Bridgewater, NJ).

X-gal histochemical staining

Whole tumors were fixed in 0.5% glutaraldehyde for 10 minutes and washed with PBS. Tissue was then incubated with a 2% X-gal solution (Sigma) with 1M MgCl₂, 30 mM potassium ferricyanide, and 30 mM potassium ferrocyanide overnight and refixed in 10% neutral-buffered formalin. Slides from frozen tissues were fixed with cold acetone/ethanol (1:1) for 20 minutes, stained with X-gal overnight, and counterstained with nuclear fast red.

Immunohistochemical analyses

IFNβ, α-SMA or desmin was detected in frozen sections of the mouse tumors. Briefly, sections were fixed for 5 minutes in ph7 formalin, after endogenous peroxidase activity was quenched by incubating the sections in 0.3% hydrogen peroxide in methanol for 30 minutes. The sections were then treated with rabbit anti-IFNβ antibody (diluted 1:1000; Chemicon, Temecula, CA), mouse anti-human α-SMA (Biomeda Corp., Foster City, CA), or rabbit anti-human desmin (diluted 1:1000, Novus Biologicals, Littleton, CO) according to the procedures in the appropriate peroxidase kit (Vector Laboratories, Burlingame, CA). Peroxidase substrate was developed using the 3,3′-diaminobenzidine (DAB) or AEC (3-amino-9-ethylcarbazole) substrate kit (Vector Laboratories). Slides were then counterstained with hematoxylin (hematoxylin QS; Vector Laboratories), dehydrated, and either mounted with VectaMount Permanent Mounting Medium (Vector Laboratories) or mounted with a low-viscosity aqueous mounting medium (Scytek Laboratories, Logan, UT).