Tumor Resection Recruits Effector T Cells and Boosts Therapeutic Efficacy of Encapsulated Stem Cells Expressing IFNβ in Glioblastomas

Glioblastoma (GBM) is the most common primary malignant brain tumor in adults and is associated with a very poor prognosis (1). Current treatment for GBM consists of maximal surgical tumor resection followed by radiation and chemotherapy (2). Despite the current advances in therapeutic interventions, GBM almost always recurs and the associated patient mortality is nearly 100%. Given the overall survival benefit observed with immunotherapies in melanoma and prostate cancer patients (3–5), there is an earnest need for evaluating immunotherapies for GBM.

Recent advances in cancer immunology have led to an increased understanding of the concept of cancer immunosurveillance and immunoediting: Newly emerging tumor cells can potentially be recognized and eliminated by the immune system, but tumors escape eradication via the process of immunoediting, thereby tilting the immune balance toward the development of an immunosuppressive tumor microenvironment (6). Immune checkpoint blockade with mAbs targeting cytotoxic T lymphocyte–associated protein 4 (CTLA-4), programmed cell death protein 1 (PD-1) or PD-1 ligand (PD-L1) is a promising strategy to overcome immunosuppressive tumor microenvironments and has recently shown favorable results in the clinical therapy of multiple cancer types (7). However, despite the importance of tumor resection as a first-line treatment for most solid tumors, surprisingly, the specific influence of tumor resection on the tumor microenvironment and immunomodulatory therapy has been poorly explored. Immune response to the damage of healthy, nontumorous tissue was studied and is known to initiate activation of the innate immune response, followed by vigorous infiltration of T cells into the lesion resulting in anti-inflammatory response (8). We hypothesized that a first-line treatment of maximal surgical tumor resection would invoke an acute immune reaction, possibly enough to break the immune tolerance within the tumor microenvironment and that administration of immunomodulatory agents after tumor resection would have superior therapeutic efficacy in treating GBM. In this study, we therefore developed syngeneic mouse tumor models of GBM resection and characterized the immune response of intact and resected tumors. Our results indicate that the resection-induced immune reaction can be further modulated toward a tumor-specific immune response via local delivery of IFNβ, leading to significantly increased survival of mice. Modulating the nonspecific immune reaction after tumor debulking toward a tumor-specific immune response might be an ideal immunotherapy strategy in GBM treatment.

IFNβ belongs to type I IFNs that bind to the IFNα/β cell surface receptor complex (IFNAR; ref. 9) and induces the classical JAK–STAT pathway as well as the PI3K and p38 MAPK pathways (10). A number of preclinical studies have shown that IFNβ has direct antitumor activity on many tumor cell types (11–13). In addition, IFNβ acts as an immunostimulatory molecule, which is known to indirectly provoke an antitumor response via modulation of the immune system (14–16). However, despite these bifunctional modes of action, the clinical translation of IFNβ treatments for cancer so far has been restricted by its short half-life and systemic toxicity (17–19).

In our previous studies, we have shown that stem cells have the inherent ability to migrate toward malignant lesions and can be engineered to continuously express therapeutic agents, making them ideal pathotropic delivery agents (20–24). We therefore thought that the stem cell–based on-site delivery of therapeutic molecules can address the issues related to the short half-life of IFNβ, while achieving therapeutic concentrations locally without causing systemic toxicity (25). In this study, we created a highly secretable variant of IFNβ by replacing its endogenous signal sequence with that of Flt3 ligand to enhance the secretion from engineered stem cells and tested our hypothesis whether targeted delivery of IFNβ can simultaneously inhibit tumor cell proliferation and boost immune response in tumor microenvironment when combined with surgical resection of intracranial GBM in syngeneic mouse models. We show that mouse mesenchymal stem cells (MSC) secreting IFNβ, encapsulated in synthetic extracellular matrix (sECM), and placed in the tumor resection cavity increase the survival of mice by direct induction of cell-cycle arrest in tumor cells and significantly enhance CD8 T-cell infiltration into the resected peritumoral microenvironment. Moreover, IFNβ shows greater direct effect on human than mouse GBM cells by driving cells from cell-cycle arrest toward apoptosis, which has great potential for future clinical translation.

The following lentiviral (LV) and retroviral (RV) vectors were used in this study: LV-pico2-Fluc-mCherry expressing firefly luciferase (Fluc) and mCherry (FmC), a kind gift from Dr. Andrew Kung (Memorial Sloan Kettering Cancer Center, NYC, NY); RV-MGRi-GFP-Fluc, RV-MGRi-HSV-TK, and MFGi-GFP as described previously (26) To construct viral IFNβ vectors, the human and mouse IFNβ genes were amplified from human and mouse cDNA of IFNβ (Sino Biological Inc.) by PCR using the forward primer with XhoI site (5′cggcctcgagatgacagtgctggcgccagcctggagcccaacaacctatctcctcctgctgctgctgctgagcatgagctacaacttgcttgga3′) for a secretable variant of human IFNβ, in which endogenous signal sequence was replaced with that of Flt3L, (5′cggcctcgagatgaccaacaagtgtctcctc3′) for human IFN with its own signal sequence and the reverse primer with BamHI site (5′ggcggatcctcagtttgcgaggtaacctct3′). Secretable mouse IFNβ was amplified the same way with the forward primer with XhoI site (5′ cggcctcgaggctagcatgacagtgctggcgccagcctggagcccaacaacctatctcctcctgctgctgctgctgagcgaattcatcaactataagcagctccagc 3′), (5′cggcctcgagatgaacaacaggtggatcctc3′) and the reverse primer with BamHI site (5′ggcggatcctcagttttggaagtttctggtaag3′). The amplified cDNA of human and mouse IFNβ were cloned into retroviral MFGi-GFP vector with XhoI/BamHI digestion, respectively. Lentiviral vector cloning were performed similarly using primers with NheI and XhoI sites and ligating the PCR products into the pLV-CSC-IG bearing IRES-GFP (27).

Female C56/B6 or female SCID (both 8–9 weeks of age; Charles River Laboratories) 25 to 30 g in weight were immobilized on a stereotactic frame to create a cranial window 7 to 10 days prior to tumor cell implantation. Using a SZX stereo-microscope system (Olympus), a small circular portion of the skull covering the right cerebral hemisphere (∼7 mm in diameter) was removed. In 7 to 10 days, the mice were again immobilized on a stereotactic frame, the previously established cranial window was exposed, and CT2A-FmC or Gl261-FmC cells (1 × 10⁵ cells/mouse) were superficially implanted into the right cerebral cortex (2.5 mm lateral from bregma, 0.5–0.7 mm deep) of C57/B6 mice. GBM4-FmC cells were implanted in SCID mice the as described above. In the following weeks, tumor growth was noninvasively monitored via bioluminescence imaging (BLI) of the tumor cells Fluc signal activity. For fluorescence-guided tumor resection, mice were immobilized on a stereotactic frame, the skin was opened and the superficial tumor was exposed. A Leica surgical microscope with ×20 magnification was used for surgical resection to reduce the tumor volume up to the tumor–tissue interface, leaving margins of the dura intact. Finally, the wound was copiously irrigated and the skin closed with 4-0 Vicryl suture. GBM tumor burden was followed by Fluc imaging over time and mice were sacrificed when neurologic symptoms became apparent. All in vivo procedures were approved by the subcommittee on Research Animal Care at Massachusetts General Hospital (approval number: 2016N000054).

Mononuclear cells were isolated from mouse brains using a discontinuous percoll gradient as described previously (28). The following primary antibodies directly conjugated to fluorophores were used: Gr-1 (Ly6C and Ly6G)-BV421, CD11b-PECY7, CD11c-PE, CD45.2-APCCY7, CD4-BV711, CD8-PERCPCY5.5, IFN-γ-BV421, (BD Biosciences), granzyme B-eFlour660, CD38-FITC, and Egr2-APC (Thermo Fisher Scientific). Single-cell suspensions were stained with antibodies against surface molecules and counterstained after fixation and permeabilization with intracellular markers. For intracellular cytokine staining, cells were stimulated with phorbol 12-myristate 13-acetate (50 ng/mL), ionomycin (1 μg/mL), and brefeldin A (5 μg/mL; Sigma Aldrich) for 4 hours prior to staining with antibodies against surface proteins followed by fixation and permeabilization and staining with antibodies against IFNγ and granzyme B. Stained mononuclear cell preparations were analyzed by flow cytometry (BD Biosciences LSRFortessa).

Data were analyzed by Student t test when comparing two groups and expressed as mean ± SD for in vitro analysis and mean ± SEM for BLI imaging analysis of tumor volumes. Differences were considered significant at P < 0.05 (*) and P < 0.01 (**). Survival times of mouse groups were analyzed using Prism 5 software (GraphPad Software) and compared using a Mantel–Cox log-rank test.

To assess the immunologic impact of surgical resection in a clinically relevant context, syngeneic GBM resection models were established in immunocompetent C57BL/6 mice. Both CT2A and GL261 syngeneic mouse GBM cells were engineered to express fluorescent (mCherry) and bioluminescent (Fluc) proteins, herein referred to CT2A-FmC and GL261-FmC, and characterized with respect to their growth in vitro and in vivo (Fig. 1A and B; Supplementary Figs. S1 and S2A and S2B). Surgical resection of established intracranial tumors derived from engineered GBM lines resulted in a small but significant survival benefit compared with mice with unresected tumors (Fig. 1C and D; Supplementary Fig. S2C; P = 0.039 in CT2A- and GL261-FmC tumors).

We next investigated the influence of surgical tumor resection on T cell immune response in the CT2A mouse model. Mice were divided into tumor-free, tumor-free resected, CT2A, and CT2A resected groups (Fig. 1E). BLI confirmed the presence of Fluc signal in CT2A-FmC–bearing mice and reduction of signal in the resected group (Fig. 1E). H&E staining on histologic sections further confirmed the correlation of tumor size with Fluc signal (Fig. 1E). IHC with the T-cell marker CD3 indicated that brains of resected and unresected CT2A-bearing mice had significantly greater numbers of T cells as compared with tumor-free and tumor-free resected brains (Fig. 1E and F). To further analyze immune response after tumor resection over time, immune cells were isolated as mononuclear cells from brains harvested 2, 4, and 6 days postresection in tumor-free resected and CT2A resected groups (Supplementary Fig. S3A). The tumor-bearing group of mice showed a gradual, time-dependent increase in the number of mononuclear cells up to day 6 after tumor resection, compared with the tumor-free resected group (Supplementary Fig. S3B). Flow cytometry subanalysis on mononuclear cells further indicated that significant numbers of MDSCs resided within CT2A tumors, which were markedly reduced after tumor resection (Fig. 1G). Infiltration of dendritic cells peaked at day 4 after tumor resection (Figs. 1H). Infiltration of lymphocytes including CD4⁺ and CD8⁺ T cells gradually increased postresection in the tumor-free resected group, and the increase was further enhanced in CT2A resected group (Figs. 1I–K). Tumor resection enhanced recruitment of both CD4⁺ and CD8⁺ effector T cells into the resected area at day 5 after tumor resection (Figs. 1L and M; Supplementary Fig. S4). Together, these results demonstrate that surgical resection of CT2A tumors greatly decreases the number of tumor-associated MDSCs and simultaneously increases the number of effector T lymphocytes recruited into the remaining tumor area.

The observed immune reaction after tumor surgery raised the question whether immune response could be regulated via administration of immune modulators other than the currently used checkpoint inhibitors. We chose IFNβ as it is known to specifically enhance CD8⁺ cytotoxic T-cell recruitment (14, 29) and also directly induce apoptosis in cancer cells, thereby possibly exposing tumor-related antigens and potentiating antitumor efficacy (9, 10). Both mouse GBM lines (CT2A and GL261) expressed the Ifnar1 and Ifnar2 transcripts for IFNβ's cognate receptors IFNAR1 and IFNAR2 (Supplementary Fig. S5A) and displayed a reduced cell viability in the presence of recombinant mouse IFNβ (mIFNβ) in vitro (Supplementary Fig. S5B and S5C) confirming the direct antiproliferative effects of IFNβ (11–13). To validate the immunostimulatory effect of mIFNβ, syngeneic mice bearing CT2A tumors were resected and recombinant mIFNβ (60 ng/mice) or PBS were instilled into the resection cavity. Presence of recombinant mIFNβ increased CD8⁺ T-cell infiltration more than 2-fold, while the percentage of CD4⁺ helper T cells out of total mononuclear cells remained unchanged (Fig. 2A). M1 and M2 macrophages were gated on the basis of CD38 and Egr2 (30). Tumor resection led to an increase in M1/M2 macrophage ratio, and mIFNβ treatment from within resection cavity shifted the macrophage polarization toward a higher M1/M2 ratio with a reduction in Egr2⁺ M2 macrophages (Supplementary Fig. S6). This shift toward the antitumor macrophage population is consistent with an increased frequency of IFNγ-producing CD4⁺ T cells (Fig. 1L), which are known to polarize macrophages and enhances their proinflammatory and antitumor properties (31).

As prolonged systemic treatment of therapeutic dosages of IFNβ has been shown to result in considerable side effects (32), we next explored the feasibility of on-site delivery of IFNβ from within the tumor resection cavity. We generated lenti- and retroviral vectors bearing a highly secretable variant of human and mouse IFNβ by swapping the original signal sequence of IFNβs with that of the Flt3 ligand signal sequence (herein referred to as hIFNβ and mIFNβ, respectively) to maximize secretion of IFNβs (Supplementary Fig. S7A–S7C). Utilizing these vectors, mouse MSCs were engineered to express mIFNβ (MSC-mIFNβ; Fig. 2B). Treatment with conditioned medium (CM) from MSC-mIFNβ reduced the cell viability of both CT2A-FmC and GL261-FmC lines in a time-dependent manner similar to recombinant mIFNβ (Fig. 2B; Supplementary Fig. S5B and S5C). To test which cellular mechanisms are involved in cell viability reduction, CT2A-FmC cells were treated with mIFNβ CM and consecutively pulsed with bromodeoxyuridine (BrdUrd). Cell-cycle analysis revealed that CT2A GBM displayed a reduction in the proportion of cells in the S-phase, with a concomitant increase in G₀–G₁ and G₂–M phases (Fig. 2C and D). However, no evidence of apoptosis induction was observed in response to mIFNβ CM treatment (Fig. 2C and E; Supplementary Fig. S7D), contrary to the previously reported effect of IFNβ on human GBMs (33). mIFNβ further resulted in the phosphorylation of STAT1 and p38, indicating a conserved response to this classical JAK/STAT signaling pathway and alternative signaling pathways, respectively (Fig. 2F). CT2A did not show a change of phosphorylation status in Akt (S473) in response to mIFNβ (Fig. 2F). In conclusion, these data indicate that mIFNβ can directly suppress proliferation of mouse GBM via induction of cell-cycle arrest.

To dissect out the immunomodulatory effects from the direct anti-GBM response of mIFNβ in vivo, mouse MSCs expressing mIFNβ or GFP (MSC-mIFNβ/-GFP) were injected into established CT2A-FmC tumors in immunocompromised (SCID) or immunocompetent (C57/B6) mice (n = 5/group; Fig. 2G). In SCID mice, intratumoral injection of MSC-mIFNβ did not demonstrate significant inhibition of tumor growth but produced a small increase in survival compared with MSC-GFP–treated counterparts (18 days vs. 15 days of median survival; P < 0.01; Fig. 2H and J). However, in the context of an intact immune system, MSC-mIFNβ greatly suppressed tumor growth and significantly prolonged survival (46 days vs. 31 days of median survival; P < 0.01), with one mouse showing long-term survival exceeding 80 days (Fig. 2H–J). These results indicate that therapeutic MSC-mIFNβ can act directly on mouse GBM in vivo and that antitumor efficacy is enhanced when acting in conjunction with a fully functional immune system.

In light of the immune cell infiltration caused by surgical resection, we tested whether the therapeutic efficacy of MSC-mIFNβ after tumor resection results in an additional survival benefit by potentiating IFNβ's immunostimulatory response. Encapsulating MSCs in a synthetic extracellular matrix (sECM) was necessary for their effective retention in the brain (Fig. 3A). In culture, sECM-encapsulated MSC-mIFNβ were shown to reduce CT2A-Fmc cell viability (Fig. 3B). In vivo, syngeneic mice bearing CT2A-FmC tumors were resected, and the cavity was filled with sECM-encapsulated MSC-mIFNβ/-GFP (Figs. 3C). The growth of residual tumors was suppressed in mice implanted with MSC-mIFNβ, compared with the increasing tumor volumes in mice implanted with MSC-GFP (Fig. 3D). This translated into significantly prolonged survival of MSC-mIFNβ–treated mice compared with MSC-GFP controls, with half of the mice surviving beyond 80 days (vs. 23 days of median survival of control group; P = 0.0008; Fig. 3E). Nevertheless, one concern with using therapeutic stem cells in the brain is that they could aberrantly proliferate and/or differentiate. To address this concern, our MSC-mIFNβ were additionally engineered to express the suicide gene, herpes simplex virus thymidine kinase (HSV-TK) and its imageable variant HSV-TK-GFP-Fluc, enabling their selective eradication by treatment with the prodrug ganciclovir (GCV). In culture, MSC-mIFNβ-TK were readily killed in response to GCV in monolayers and encapsulated in sECM (Supplementary Fig. S8). In vivo, when MSC-mIFNβ-TK-Fluc were encapsulated and placed in the resection cavity, systemic treatment with GCV also effectively reduced their numbers, compared with PBS-treated controls in which MSC number increased (Fig. 3F). Indeed, sectioned brains confirmed a large reduction in GFP-positive MSC-TK-mIFNβ upon treatment with GCV (Fig. 3G). In conclusion, these data indicate that the immune reaction following surgical tumor resection can be regulated by placing sECM-encapsulated MSC-mIFNβ into the resection cavity and that prolonged local delivery of IFNβ results in a significant survival benefit. Furthermore, our suicide gene strategy to eliminate MSC-mIFNβ from the brain posttreatment provides a solution regarding the concerns of using therapeutic stem cells in clinical settings of cancer therapy.

To translate our findings into clinical settings of GBM therapy, we next set out to test hIFNβ's direct anticancer activity in human GBM lines. We first confirmed that IFNAR1/2 are universally expressed in transcriptional level with relatively low range of variations in human GBM patients according to TCGA database (n = 166), indicating IFNβ's broader applicability as an anticancer agent for GBM patients compared to EGFR and VEGFA, two well-known therapeutic targets for GBM treatment (Fig. 4A). Consistently, transcriptional expression of IFNAR1/2 was confirmed in all of the human GBM lines (Fig. 4B). A significant reduction in cell viability in two patient-derived primary GBM lines (GBM4 and BT74) was seen following 72-hour incubation with CM (100 ng/mL hIFNβ: obtained from 293T expressing a secretable variant of hIFNβ, Supplementary Fig. S7B and S7C; P < 0.01 vs. nontreated for all lines; Fig. 4C; Supplementary Fig. S9). GBM4 cells showed a significant increase of S-phase cells in cell cycle, and the emergence of an apoptotic population was coincident with reduction of total cell number in the S-phase following hIFNβ exposure as compared with controls (Fig. 4D; Supplementary Fig. S10). The increase of cells in S-phase in response to hIFNβ was not due to greater rate of cell proliferation (Fig. 4E). Cleaved PARP (c-PARP) was observed starting 48 hours post-hIFNβ treatment in GBM4 cells (Figs. 4F). These data suggest that hIFNβ can suppress cell viability of GBMs by inducing prolonged cell-cycle arrest followed by apoptosis.

To further characterize the direct hIFNβ-mediated antitumor effects on human GBM, IFNβ-induced signaling pathways were explored in more detail. A panel of GBM lines showed STAT1 activation after hIFNβ treatment (Fig. 4G) which lasted up to 72 hours in GBM4 cells (Fig. 4H). hIFNβ treatment also induced activation of the MAPK, p38 tumor suppressor and the DNA checkpoint protein Chk1, but not Akt in GBM4 cells (Fig. 4H). Chk1 phosphorylation was not observed in CT2A mouse GBM cells in response to mIFNβ (Supplementary Fig. S11). These data are consistent with the result showing an increase of GBM cells in S-phase arrest in response to hIFNβ (Fig. 4D; Supplementary Fig. S10), suggest that p38 and Chk1 are both involved in the IFNβ signaling that leads to cell-cycle arrest in the S-phase, and prolonged cell-cycle arrest may induce apoptosis in GBM4 cells.

Next, we wanted to test the direct antitumor effect of MSC secreted hIFNβ on patient-derived GBM cells in an orthotopic xenograft mouse model of GBM resection. Coculture of MSC-hIFNβ with GBM4-FmC cells demonstrated a significant reduction of cell viability, confirming functionality of the therapeutic stem cells expressing hIFNβ (Fig. 5A). Furthermore, in vitro antitumor efficacy was retained after sECM-encapsulation of MSC-hIFNβ, indicating that hIFNβ can readily diffuse out of the sECM and effect surrounding tumor cells in a time-dependent manner (Fig. 5B). In vivo, the patient-derived GBM4-FmC line admixed with MSC-hIFNβ and intracranially implanted into immunodeficient SCID mice did not grow in contrast to its continuous growth when admixed with MSC-GFP control (Fig. 5C). Having confirmed the antitumor effect of MSC-hIFNβ, we then resected fully established GBM4-FmC (Fig. 5D–H), followed by implantation of sECM-encapsulated MSC-hIFNβ or MSC-GFP into the tumor resection cavity. MSC-hIFNβ completely suppressed GBM4 growth up to 10 days posttreatment in contrast to immediate tumor progression observed in the control group (Fig. 5E and G). However, despite initial suppression of tumor regrowth in MSC-hIFNβ–treated mice, tumors started to reestablish their growth pattern after 10 days, overall resulting in a modest but significant survival benefit compared with the control group (50 days vs 40 days of median survival; P < 0.01; Fig. 5H). These results reveal that human IFNβ has a greater direct antitumor activity than mouse IFNβ due to combined effects of cell-cycle arrest and apoptosis. Given the lack of immune response in this xenograft model and the intact immune systems in human GBM patients, the use of MSC-hIFNβ offers great potential for future clinical translation.

In this study, we developed and characterized syngeneic orthotopic mouse models of GBM resection and show that surgical debulking of orthotopic GBM evokes a characteristic, time-dependent peritumoral immune reaction in the tumor microenvironment, which in combination with local delivery of immunomodulatory therapeutics can be utilized to significantly boost the hosts' anticancer immune response. Our syngeneic mouse tumor models of GBM resection have great value for the study of immune response to immunomodulatory agents and provide new insights that may help guide clinicians in the difficult transition from preclinical to clinical anti-GBM immunotherapies.

One of the cornerstones of clinical care for GBM patients is surgical tumor debulking (2), yet nearly all preclinical models focus on treating established solid tumors. Previously, we have shown the feasibility of resecting GBMs developed from human patient derived cell lines in orthotopic xenograft animal models (22). In an effort to understand the changes in immune response after tumor resection, we developed a syngeneic mouse GBM model by integrating fluorescent and bioluminescent markers and extensive optical imaging to simultaneously confirm the presence of established tumors, visualize the extent of tumor debulking and serially monitor tumor regrowth postresection. As tissue damage is known to result in activation of both innate and adaptive immune response (8), our tumor resection model may enhance three key elements crucial for successful anticancer immunotherapy: (i) removal of immunosuppressive cells such as MDSCs from the tumor microenvironment (34); (ii) enhancement of effector CD4/CD8 T-cell infiltration into tumor microenvironment (35); and (iii) reduction of tumor size (36). Although a temporal reduction of the number of MDSCs in the spleen after surgical resection of tumors was reported in the 4T1 murine mammary tumor model (37), our study shows that surgical tumor debulking of GBM significantly reduces MDSC in the tumor microenvironment. As such, our findings serve as an excellent preclinical immuno-oncology model to study how current and/or developing immunotherapeutic agents alter different immune-related effector elements in the clinical context of postresection tumor microenvironments.

The inherent risk associated with systemic immunomodulatory therapies to induce autoimmune reactions against healthy tissues or organs motivated us to seek out an alternative that would allow safe and efficient modulation of the observed postresection immune response. Strategically, a cytokine-induced establishment of favorable conditions for the host immune system postresection would be an ideal solution to allow tumor elimination via activation of a variety of stimulated immune effectors. However, short half-lives and toxicity associated with systemic cytokine treatments so far have prevented clinical translation of such approaches (38). IFNβ is an immunostimulatory molecule, which is known to indirectly provoke an antitumor response via various modulations of the immune system. These include potentiating MHC class I expression, initiating specific T-cell responses against tumor cells, activating the innate immune system, and inhibiting angiogenesis (14–16, 39). Our results also showed that tumor resection and IFNβ treatment shift macrophage polarization toward a higher M2/M2 ratio. It has been shown that M1 macrophages have antitumor function (40), further supporting therapeutic potential of our IFNβ therapy in a tumor resection model. Although, a number of preclinical studies have shown that IFNβ has potent direct antitumor activity in addition to its well-known immunomodulatory effects (11–13), clinical trials utilizing systemic administration of IFNβ have had only very limited success, mostly due to its short half-life of approximately 5 hours and systemic toxicity (17, 18). In the case of GBM, clinical trials have generally been disappointing. A study in which patients with recurrent GBM received intravenously administered recombinant IFNβ documented no evidence of treatment efficacy (41). In a separate phase I study, patients with recurrent GBM were administered IFNβ through a reservoir connected via a catheter to the tumor cavity (Ommaya/Rickham reservoir). In this study, side effects from IFNβ therapy as well as problems with the reservoir itself (blocking and infection) resulted in many study dropouts; however, 25% (3/12 patients) showed a partial response (42). These studies indicate that despite IFNβ demonstrating promising efficacy in preclinical cancer models, its successful clinical translation requires further investigation, especially in regard to better delivery methods. In an effort to circumvent the above issues, we therefore genetically engineered MSCs to release a highly secretable variant of IFNβ and subsequently utilized these engineered MSCs as a vehicle for prolonged on-site delivery to intracranial GBMs.

Despite extensive preclinical evidence demonstrating the potential of cell-based therapies for GBM, only very few studies have explored methods to introduce therapeutically “armed” stem cells in GBM resection cavities. Hereby, stem cell “wash-out” and rapid diffusion from the resection cavity by cerebrospinal fluid are potential threats to treatment success. On the basis of our previous studies (21, 22, 24), we therefore employed biodegradable sECM encapsulation as an attractive approach for retention of MSC-IFNβ within the resection cavity. The importance of sECM has been intensively examined in general cell therapy, especially in regenerative medicine in which the nature of sECM as porous hydrogels promotes nutrient input and waste outflow and enhances cell seeding, retention, and migration (43). In this study, we observed superior survival of sECM encapsulated mMSC-IFNβ in resection cavities in comparison with non-sECM–encapsulated mMSCs, and our study further reveals that sECM-encapsulated MSCs are effective in increasing local drug concentrations and exposure times. Together, these properties resulted in enhanced tumor cell killing and significantly increased survival of sECM-MSC-IFNβ–treated mice.

As discussed above, in addition to its immunomodulatory function, IFNβ has direct cytotoxic effects on tumor cells. However, the detailed mechanisms leading to cell death after IFNβ treatment were not fully understood. A recent study in human neuroblastoma cells demonstrated that IFNβ induces tumor cell killing by p38-mediated reactive oxygen species (ROS) production (44). IFNβ further promotes senescence in human fibroblasts mediated by ROS production and activation of the ATM/Chk2 DNA checkpoint, leading to activation of p53 and RB (45). Exploring the species-specific therapeutic effects of IFNβ in more detail, we demonstrate that IFNβ has direct antiproliferative activity in both human and mouse GBM types, however with noticeable differences in their response. STAT1 activation followed by p38 phosphorylation lasted much less in CT2A cells than in human GBM4 in response to IFNβ treatment (Figs. 2F and 4I). The wild-type status of p53 in CT2A may further favor accumulation of CT2A cells in the G₀–G₁ and G₂ cell-cycle phases (46). In contrast, we show that in human GBM4 cells, hIFNβ induces prolonged JAK/STAT activation independent of PI3K–Akt, activates p38 tumor suppressor and DNA checkpoint via Chk1, which ultimately leads to cell-cycle arrest in the S-phase. Homozygous deletion of CDKN2A&B (p15, CDK4 inhibitor) in GBM4 may also contribute to bypass cell-cycle arrest in G₁–S transition in response to IFNβ (47). In addition, mutations either in p53 or its downstream p14/p16 CDK inhibitors in most of the established GBM cell lines indicate that IFNβ may function similarly in other patient-derived GBM lines as shown in our studies with GBM4 (48).

The biggest safety concern with stem cell–based therapies for cancer is the survival and potential proliferation of remaining therapeutic stem cells after tumor therapy. In this study, we solved this safety issue with the prodrug converting system HSV-TK/GCV. Moreover, stem cells expressing HSV-TK provide additional advantages when translated into clinical GBM therapies, particularly the ability of the prodrug GCV to cross the blood–brain barrier and its dual function as a PET imaging agent as well as the bystander effect on surrounding tumor cells associated with GCV-induced stem cell elimination. Positron detection is the most sensitive imaging modality. PET ligands such as 18F-FHBG, 18F-FEAU, and 124I-FAIU have been used to provide 3D images of the concentration of cells in different locations within a tumor (49, 50). As the HSV-TK expression can be repetitively evaluated over time by PET scanning and does not fade with cell division, the HSV-TK/GCV system is an ideal strategy for the monitoring and safe termination of stem cell–based therapies in clinical settings.

In conclusion, our study shows the development of clinically relevant GBM tumor resection models to assess immunotherapeutic effects in the context of additional treatments with locally delivered immunomodulatory agents. Given that human IFNβ shows greater direct effect on human GBM cells than their mouse counterparts, this study has a great potential to be translated into GBM patients. We envision that after the neurosurgical removal of the main tumor mass, the patient's own reprogrammed cells or HLA (human leukocyte antigen)-compatible MSCs engineered to express IFNβ will be encapsulated in sECM and implanted into the resection cavity after tumor debulking. These cells would then result in simultaneous tumor cell killing as well as targeted immune response, which could be further potentiated via the systemic administration of immune checkpoint inhibitors, with the ultimate goal being improved patient outcome.

K. Shah owns equity in and is a member of the Board of Directors of, AMASA Technologies, a company developing stem cell based therapies for cancer. K. Shah's interests were reviewed and are managed by Brigham and Women's Hospital and Partners HealthCare in accordance with their conflict of interest policies. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K. Shah, S.H. Choi, D.W. Stuckey

Development of methodology: K. Shah, S.H. Choi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.H. Choi, S. Pignatta, C. Reinshagen, J.K. Khalsa, N. Roozendaal, J. Martinez-Quintanilla, K. Tamura, E. Keles, K. Shah

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K. Shah, S.H. Choi, D.W. Stuckey, S. Pignatta, J.K. Khalsa, N. Roozendaal, J. Martinez-Quintanilla

Writing, review, and/or revision of the manuscript: S.H. Choi, D.W. Stuckey, C. Reinshagen, J.K. Khalsa, N. Roozendaal, J. Martinez-Quintanilla, K. Shah

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.W. Stuckey, E. Keles, K. Shah

Study supervision: K. Shah, S.H. Choi, D.W. Stuckey

Other (provision of study material): K. Shah, S.H. Choi, D.W. Stuckey

We thank G. Prestwich (University of Utah) and T. Zarembinski (Biotime Inc.) for providing us with synthetic extracellular matrix gels and Ravi Mylvaganam and Christina Luo for their help with flow cytometry.

This work was supported by NIH grants R01CA138922 (K. Shah), R01CA173077 (K. Shah) and 1S10RR023440-01A1 (MGH Pathology CNY Flow Cytometry Core and sorting shared instrumentation grant).

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.