Neural stem cells secreting bispecific T cell engager to induce selective antiglioma activity

Significance

Glioblastoma (GBM) is the most aggressive and lethal brain tumor in adults. The GBM microenvironment is highly immunosuppressive, rendering T cells incapable of recognizing and eliminating malignant cells. We developed bispecific T cell engagers (BiTEs) that successfully engage patients' T cells to attack and kill GBM expressing interleukin 13 receptor alpha 2 (IL13Rα2). Neural stem cells (NSCs) engineered to secrete BiTE migrate readily within the brain tissue to established tumors and produce BiTE protein within the malignant tissue. Treatment of animals bearing IL13Rα2-expressing patient-derived GBM xenografts with therapeutic NSCs resulted in significant survival benefits in mice. These results show an exciting potential of NSCs as a delivery platform for BiTE therapy to improve outcomes for GBM patients.

Abstract

Glioblastoma (GBM) is the most lethal primary brain tumor in adults. No treatment provides durable relief for the vast majority of GBM patients. In this study, we've tested a bispecific antibody comprised of single-chain variable fragments (scFvs) against T cell CD3ε and GBM cell interleukin 13 receptor alpha 2 (IL13Rα2). We demonstrate that this bispecific T cell engager (BiTE) (BiTE^LLON) engages peripheral and tumor-infiltrating lymphocytes harvested from patients' tumors and, in so doing, exerts anti-GBM activity ex vivo. The interaction of BiTE^LLON with T cells and IL13Rα2-expressing GBM cells stimulates T cell proliferation and the production of proinflammatory cytokines interferon γ (IFNγ) and tumor necrosis factor α (TNFα). We have modified neural stem cells (NSCs) to produce and secrete the BiTE^LLON (NSC^LLON). When injected intracranially in mice with a brain tumor, NSC^LLON show tropism for tumor, secrete BiTE^LLON, and remain viable for over 7 d. When injected directly into the tumor, NSC^LLON provide a significant survival benefit to mice bearing various IL13Rα2⁺ GBMs. Our results support further investigation and development of this therapeutic for clinical translation.

Routine treatment of newly diagnosed glioblastoma (GBM) consists of surgical resection, chemotherapy, and radiation, which results in a median GBM patient survival of less than 2 y, with just 5% of patients surviving beyond 5 y (1). The blood–brain barrier (BBB) limits therapeutic access to the tumor (2). An immunosuppressive microenvironment and molecular heterogeneity of GBM present a unique set of challenges for developing effective therapies for this type of brain tumor (3–10).

The development of treatments for lessening the immunosuppressive effects of GBM represents an active area of preclinical and clinical neuro-oncology research. Many, if not all, approaches being tested involve increasing T cell cytotoxic antitumor activity (11–17). Large numbers of functional cytotoxic tumor-infiltrating lymphocytes (TILs) correlate with improved progression-free survival for GBM patients (18–21). However, the immunosuppressive milieu of GBM impairs T cell cytolytic function, altering the effectiveness of T cell-based therapies for treating GBM (22–26). Numerous lymphocyte-directed treatments are being investigated, including the use of bispecific T cell engagers (BiTEs) (17, 27, 28). BiTEs can be produced and used without patient-specific BiTE individualization and can, therefore, be considered “off-the-shelf” therapeutics (29–32). The use of BiTEs targeting tumor-associated antigens (TAAs) has been approved by the Food and Drug Administration (FDA) in treating liquid malignancies, and BiTE-associated treatments are currently being evaluated in multiple clinical studies for solid tumors (e.g., NCT03792841, NCT04117958, and NCT03319940) (33–36).

BiTEs consist of two single-chain variable fragments (scFvs) connected by a flexible linker (37). One of the scFvs is directed to a TAA and the other to CD3 epsilon (ε) that is expressed on T cells (38). BiTEs engage TILs and cancer cells in a major histocompatibility complex (MHC)-independent manner and are, therefore, unaffected by MHC down-regulation that occurs in GBM cells (37–40). The specificity of BiTE's tumor antigen-directed scFv is imperative to harness the full therapeutic potential of the recombinant molecule (41). BiTE anticancer activity requires BiTE binding with malignant and immune cells simultaneously; single-arm binding to a tumor antigen or CD3ε is therapeutically ineffective (42, 43).

The short half-life of BiTEs in plasma necessitates a frequent or continuous infusion into patient circulation to maintain therapeutic levels of BiTE (43–45). Several approaches have been proposed and tested to compensate for the rate of BiTE biologic life in treating peripheral cancers (45–47). These include the recombinant protein's sustained production by subcutaneous injection of mesenchymal stem cells (MSCs) seeded into a synthetic extracellular matrix scaffold, liver translation of BiTE messenger RNA (mRNA), and peritumoral delivery of MSCs secreting BiTEs. A recent study also explored the use of modified immune cell delivery of BiTE to GBM (48), and the reduction of tumor burden in treated animals has been observed. It remains to be determined if these bicistronic antiglioma treatments share chimeric antigen receptor (CAR) T cells' fate, which includes a low penetrance and short survival of CARs within the brain (27, 49, 50); both are limiting factors for sustained and efficient delivery of BiTEs by CAR T cells.

Neural stem cells (NSCs) have inherent advantages as a cellular carrier of antineoplastic agents to the site of GBM since they are native to the brain. NSCs have demonstrated tropism to brain tumors in several preclinical models. These cells can withstand a harsh oxygen-deprived environment of GBM. Here, we investigated NSCs as producers of BiTEs targeting the tumor-associated antigen interleukin 13 receptor alpha 2 (IL13Rα2) and their antitumor activity using in vitro and in vivo models of GBM. In vitro, BiTEs show significant antitumor activity when used in cocultures that include T cells harvested from patients’ blood and tumor tissue. In vivo, NSCs modified for BiTE synthesis migrate to a tumor in animal subjects’ brains while functioning as intra- and peritumoral BiTE producers. The following are details of the results from our experiments in characterizing NSCs that produce IL13Rα2-directed BiTEs. We interpret these findings as support for additional safety and efficacy analysis for their potential clinical translation in treating GBM patients.

Results

Development of Tumor-Targeting BiTEs.

BiTE-targeting IL13Rα2 was generated using the scFv of mAb47 against the IL13Rα2 that has been previously described by our group and scFv of the mAb OKT3 directed toward the invariable ε chain of CD 3 (CDε) (49, 51–54). ScFvs were connected using a flexible glycine/serine linker in the following orientation: mAbOKT3_VH-mAbOKT3_VL-mAb47_VL-mAb47_VH (Fig. 1A). Short (SL, 5 amino acids) and long (LL, 23 amino acids) linkers were used to construct BiTE^SLON and BiTE^LLON. BiTE^SLOFF and BiTE^LLOFF control molecules were generated by replacing the complementary determinant region 3 of the mAb47 light chain with the sequence of the mAb MOPC-21, which prevents IL13Rα2 binding (Fig. 1A) (38). A polyhistidine (6His) tag was added at the C terminus of BiTE constructs for BiTE purification and detection. All acronyms for BiTE molecules are listed in SI Appendix, Table S1. Lentiviral vectors (pLVX-IRES-ZsGreen1) encoding complementary DNA (cDNA) for each BiTE were constructed, and corresponding lentiviral particles were used to transduce HEK293T cells for the production of BiTE proteins. Recombinant BiTE proteins were purified from culture supernatants using HisPure resin. Purified BiTE integrity was verified by Western blotting using anti-His antibodies (Fig. 1B).

Fig. 1.

Design of BiTE targeting IL13Rα2-expressing gliomas. (A) BiTEs consist of the scFv fragments of mAbOKT3 against the CD3ε and scFv of mAb47 against IL13Rα2 connected by either short or long linkers named as BiTE^SLON and BiTE^LLON, respectively. The CDR3 of the light chain (VL/M) of the mA47 was replaced with a sequence of the nonspecific MOPC21 antibody to generate BiTE^LLOFF and BiTE^SLOFF to controls with abridged binding to IL13Rα2. (B) Western blotting showed a specific single band after affinity purification and detection with anti-His antibodies at ∼55 kDa. (C) Increasing concentration of BiTE molecules bound to human IL13Rα2 immobilized on ELISA plates (1 μg/mL) and detected using anti-His tag antibodies. EC₅₀ of BiTE^LLON is 6.767 μg/mL, and EC₅₀ of BiTE^SLON is 57.87 μg/mL. (D) BiTE^LLON but not BiTE^LLOFF engaged CDε-expressing cells from glioma patients in the killing of IL13Rα2-expressing glioma cells (target-to-effector [T:E] ratio 1:20, two-way ANOVA, n = 3–4, ****P < 0.0001). (E) Example of flow cytometry chromatogram for IL13Rα2-negative GBM39 cell line and IL13Rα2-positive cell lines, GBM6, and GBM12. (F) Expression of IL13Rα2 at the cell surface in GBM6, GBM12, and GBM39 cell lines (one-way ANOVA, n = 3, ****P < 0.0001). (G) Binding of BiTE^LLON and BiTE^LLOFF to GBM cell lines (one-way ANOVA, n ≥ 3, ****P < 0.0001). (H) Expression of CD3ε on the surface of GBM patients' PB lymphocytes and TILs (n ≥ 6). (I) CD3ε expression on the cell surface of CCRF-CEM, a T lymphoblastoid cell line (28), and Jurkat, a T lymphoblast cell line. (J) Binding of BiTE^LLON to CEM and Jurkat cell lines (n = 3). OD_450nm, optical density at 450 nm. MFI, median fluorescent intensity.

Characterization of BiTE Binding and Function.

Binding of purified BiTEs to IL13Rα2 and CD3ε epitopes was examined using an enzyme-linked immunosorbent assay (ELISA) and cell-binding assays. BiTE^LLON showed 8.5× higher affinity binding to immobilized IL13Rα2 than BiTE^SLON (effective concentration, 50% [EC₅₀] values of 6.767 µg/mL and 57.87 µg/mL, respectively) (Fig. 1C). Control BiTEs did not bind to IL13Rα2 (Fig. 1C).

A majority of gliomas express IL13Rα2, with GBM expressing it at the highest levels (SI Appendix, Fig. S1A) (55, 56). The ability of BiTEs to engage donor T cells in antiglioma activity was determined using cocultures of IL13Rα2-expressing GBM cells (SI Appendix, Fig. S1 B and C) with BiTEs and T cells. BiTE^LLON successfully engages T cells, as indicated by BiTE concentration-dependent cytotoxicity, with a maximal effect observed at 2.5 μg/mL (Fig. 1D and SI Appendix, Fig. S1D). BiTE^SLON did not induce T cell antitumor activity (Fig. 1D).

Patient-derived xenograft cell lines GBM6, GBM12, and GBM39 express different levels of IL13Rα2 (Fig. 1 E and F). BiTE^LLON binds to GBM6, and GBM12 cells, but not to GBM39 in which IL13Rα2 is either absent or expressed at a level beneath that required for flow cytometry detection. At a concentration of 2 μg/mL, BiTE^LLON saturates IL13Rα2 in GBM6 and -12 (Fig. 1G) and occupies 50% of available receptor in U251 cells (SI Appendix, Fig. S1E).

Flow cytometry analysis of GBM patient T cell CD3ε expression revealed substantial interpatient variability in peripheral blood (PB) and TIL (20) populations. CD3ε-associated fluorescence intensity ranged from 649 to 3,058 and from 706 to 1,753 for PB and TILs, respectively (Fig. 1H). We used human lymphocyte Jurkat and CEM cell lines to further test BiTE interaction with CD3 (Fig. 1I and SI Appendix, Fig. S1 F and G) and observed saturation binding as BiTE concentrations approached 1.5 μg/mL (Fig. 1J). BiTE CD3ε binding was also apparent when using IL13Rα2 binding defective BiTE^LLOFF (SI Appendix, Fig. S1H).

To examine BiTE-induced T cell antitumor activity in vitro, we used the chromium 51 (⁵¹Cr) release assay. BiTE^LLON engaged T cell killing of IL13Rα2-expressing GBM6 and GBM12 glioma cells in a target-to-effector (T:E) ratio-dependent fashion (Fig. 2 A and B). GBM39 cells, negative for IL13Rα2 expression, were not killed by the T cells, at any tested T:E ratio, in the presence of BiTE^LLON and BiTE^LLOFF (Fig. 2C). Similar killing activity was observed with T cells harvested from the blood of a patient with another type of brain tumor, colloidal meningioma, in coculture with IL13Rα2-positive GBMs in the presence of BiTE^LLON, but not IL13Rα2-negative cells or BiTE^LLOFF (SI Appendix, Fig. S2 A–C). At day 3 of coculturing, the majority of IL13Rα2-expressing GBM cells were killed by T cells in the presence of BiTE^LLON, but not in BiTE^LLOFF (Fig. 2D). GBM cell killing was associated with proliferation and activation of T cells, as exemplified by the expression of CD69 and CD25 T cell markers. There was no proliferation or expression of activation molecules for T cells cocultured with BiTE^LLOFF, or for T cells cultured with BiTE^LLON, as well as BiTE^LLOFF in the absence of IL13Rα2-expressing GBM cells (Fig. 2 E–G and SI Appendix, Fig. S2 D–F). T cell proliferation and activation were associated with significant increases in T cell secretion of IL2, interferon γ (IFNγ), and tumor necrosis factor α (TNFα) (Fig. 2 H–J and SI Appendix, Fig. G-I).

Fig. 2.

BiTE^LLON activates T cell and induces the killing of IL13Rα2⁺ gliomas. Chromium 51 (⁵¹Cr) release assay shows that (A) GBM6 and (B) GBM12 are killed by donor T cells in the presence of BiTE^LLON, but not BiTE^LLOFF, in all tested target-to-effector (T:E) ratios (two-way ANOVA, n = 4, ****P < 0.0001). (C) GBM39 gliomas are spared by donor T cells in the presence of either BiTE^LLON or BiTE^LLOFF. (D) Bright-field pictures of T cells with IL13Rα2-expressing GBMs after 3 d in coculture in the presence of BiTE^LLON(I) and BiTE^LLOFF(II) (Scale bar: 50 μm.). (E) Proliferation and expression of activation markers (F) CD69 and (G) CD25 after coculture of donor T cells for 3 d with BiTE^LLON, BiTE^LLOFF, stimulation with CD2CD3CD28 beads (Stim.) or absence of any form of stimulation (No stim.) with GBM12 gliomas (one-way ANOVA, n = 3, ****P < 0.0001). ELISA for the production of (H) IL2 (24 h), (I) IFNγ (48 h), and (J) TNFα (48 h) by T cells after coculture with GBM12 in the presence of BiTE^LLON, BiTE^LLOFF, stimulation with CD2CD3CD28 beads (Stim.), or absence of any form of stimulation (No stim.) (one-way ANOVA, n ≥ 4, ****P < 0.0001).

BiTE Engages GBM Patients' Lymphocytes in an Antiglioma Activity.

It has been shown that T cells in GBM patients exhibit an exhaustion phenotype characterized by the expression of PD1/TIM3/Lag3 (8, 57, 58). Our analysis of patients' peripheral lymphocytes and TILs indicates cells that are positive for all exhaustion markers (triple-positive cells) comprise 5.48 ± 4.8% of PB PD1⁺CD8s and 7.67 ± 3.87 of PD1⁺TILs (Fig. 3 A and B). Low-level expression of activation markers CD69 and CD25 was detected in CD8⁺ cells (Fig. 3 C and D). In addition, flow cytometry analysis showed that 35.68 ± 5.21% of PB CD8s and 28.99 ± 5.218% of TILs isolated from patients' tissues are positive for IFNγ (SI Appendix, Fig. S3A). The average number of TNFα⁺ cells within the PB CD8s and TILs was 1.83 ± 0.7% and 8.67 ± 6.17%, respectively (SI Appendix, Fig. S3B).

Fig. 3.

BiTE^LLON engages GBM patients' lymphocytes in an antiglioma activity. Basal levels of (A) PD-1, exhaustion markers (B) PD1, Lag3, Tim3, and activation markers (C) CD69, (D) CD25, in cells isolated from PB and TILs as evaluated by flow cytometry (n ≥ 7). GBM12 ⁵¹Cr killing assay determined the reengagement of PB and TILs from GBM (E) patient 1 and (F) patient 2 into glioma killing in the presence of BiTE^LLON and CD2CD3CD28 beads (Stim.), but not BiTE^LLOFF or not stimulation (No stim.) (one-way ANOVA, n ≥ 2, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). Flow cytometric analysis of (G) PD-1 and (H) PD-1, Tim3, and Lag3 after coculture for 3 d of PB lymphocytes or TILs with IL13Rα2-expressing gliomas (n = 3). Evaluation of (I) secreted (ELISA) and (J) intracellular (flow cytometry) levels of IFNγ in lymphocytes isolated from GBM patients’ PB (one-way ANOVA, n = 2–12, *P < 0.05, **P < 0.01, ****P < 0.0001). Secreted (K and L) intracellular (flow cytometry) levels of IFNγ in TILs isolated from GBM patients (one-way ANOVA, n = 2–3, *P < 0.05, **P < 0.01, ****P < 0.0001).

We processed blood and tissues from two GBM patients and found that PB and TILs could be stimulated for antiglioma activity when cocultured with IL13Rα2⁺ GBMs in the presence of BiTE^LLON, but not BiTE^LLOFF (Fig. 3 E and F). The potency of TILs cocultured with BiTE^LLON in killing GBM cells was variable between patients in relation to bead-stimulated cells.

PB and TIL cocultured with GBM cells showed activation and increased expression of PD-1 at the cell surface (Fig. 3G and SI Appendix, Fig. S3 C and D). There was also an increase of triple+ peripheral lymphocytes and TILs after BiTE^LLON and bead stimulation, but not when treated with BiTE^LLOFF or in unstimulated conditions (BiTE^LLON: PB CD8 triple+ 13.52 ± 6.38% and TILs triple+ 20.52 ± 9.98%; Stimulated: PB CD8 triple+ 28.03 ± 17.45% and TILs triple+ 35.15 ± 1.66%) (Fig. 3H).

BiTE^LLON induced significantly more IFNγ⁺ PB CD8 cells, as compared to BiTE^LLOFF or unstimulated cells (P < 0.0001) (Fig. 3 I and J). In stratifying cells according to exhaustion markers, we determined that contributors to the production of IFNγ in BiTE^LLON-stimulated cultures are distributed within all exhaustion marker subtypes, with PD1⁺Lag3⁺ cells comprising the largest subgroup (46.53 ± 1.7%) (SI Appendix, Table S2). In positive control PB coculture (activated with beads), PD-1⁺ cells were determined as the major producers of IFNγ (SI Appendix, Table S2). The profile of TNFα in cocultured PB lymphocytes was similar to the IFNγ (SI Appendix, Fig. S3 E and F).

Patient TILs secreted significantly more IFNγ when cocultured with BiTE^LLON than negative controls and bead-stimulated cells (P < 0.005) (Fig. 3K). Flow cytometry analysis also showed a higher expression of IFNγ⁺ in TILs when cultured with BiTE^LLON as compared to negative controls, but not to bead-activated cultures (Fig. 3L). When stratified by individual exhaustion phenotypes, triple+ TILs were the most significant contributors to IFNγ in BiTE^LLON- and bead-stimulated culture conditions (SI Appendix, Table S2). Flow cytometry analysis revealed that BiTE^LLON causes an increase in TIL intracellular TNFα (SI Appendix, Fig. S3G). The most significant contributors to the pool of TNFα⁺ cells were triple+ TILs (SI Appendix, Table S2).

Altogether, these data show that BiTE^LLON stimulates antiglioma activity in patient T cells and induces cytokine production in triple exhausted TILs.

Development of NSCs Secreting Functional BiTE.

BiTEs can cross the BBB, but the rate at which they are degraded and cleared from the body necessitates repeated systemic administration in cancer patients (35, 59). We and others have demonstrated that NSCs can migrate in the brain toward GBM cells/tissue and produce therapeutic antibodies for extended periods of time (60–62). For this study, we modified immortalized NSCs (63–65) for BiTE^LLON and BiTE^LLOFF synthesis and secretion (SI Appendix, Fig. S4A) (60). Modified NSCs were fluorescence-activated cell sorter (FACS) sorted by their expression of the ZsGreen1 reporter (Fig. 4A). Cytogenic studies of NSCs modified for synthesis and secretion of BiTE^LLON (NSC^LLON) showed the same karyotype as the parental NSCs (SI Appendix, Fig. S4B). Results from immunocytochemical analysis (Fig. 4 B and C) and Western blotting (Fig. 4D) showed that NSC^LLON and NSC^LLOFF produce BiTE protein. BiTEs secreted by NSC^LLON, but not NSC^LLOFF, showed strong binding to human recombinant IL13Rα2 (hrIL13Rα2) immobilized on ELISA plates (Fig. 4E). Quantitative analysis revealed that 1 × 10⁶ of NSC^LLON produced 1.6 ± 0.13 μg of BiTE within the first 24 h and 2.42 ± 0.26 μg after an additional day in culture (48 h) (Fig. 4F). NSC^LLON and NSC^LLOFF are tropic for GBM cells in vitro (Fig. 4G and SI Appendix, Fig. S4C).

Fig. 4.

NSCs produce and deliver BiTE^LLON protein to tumors in vivo. BiTE-transduced NSCs were sorted for the expression of (A) ZsGreen1 (sample histogram) protein and evaluated for the production of BiTE proteins using immunocytochemistry by (B) NSC^LLON and (C) NSC^LLOFF cells immobilized on cover glass (cells-DAPI-nucleus [I], NSC-ZsGreen1 [II], BiTE protein-α-His-Tag [III], and [IV] merged together. (Scale bars: 50 μm.) (D) Western blotting using anti-His Tag and anti–glyceraldehyde-3-phosphate dehydrogenase (anti-GAPDH) antibodies. (E) NSC^LLON but not NSC^LLOFF produced protein binding to IL13Rα2 protein immobilized on ELISA plates. (F) Production dynamics of BiTE^LLON by NSC^LLON after 24 and 48 h in culture (t test, n = 3, **P < 0.01). (G) Migratory capacity of parental (NSCs) and modified NSC^LLON and NSC^LLOFF toward conditioned media collected from human xenograft cell lines GBM39, GBM12, and GBM6 cultured in vitro was assessed in invasion chambers (one-way ANOVA, n = 3, *P < 0.05). (H) Immunocytochemical analysis of NSC^LLON labeled with superparamagnetic iron oxide contrast (SPIO) particles used for dynamic in vivo tracking of NSCs by MRI. Scale bar, 50 µm. (I) Labeled NSC^LLON injected into a contralateral hemisphere, robustly egressed from the injection site, and migrated toward the tumor. The change in the NSC^LLON movement within the brain is expressed as a change in R2*, a signal corresponding to labeled NSCs (P = 0.0005 as described in SI Appendix, Statistics and data analysis) and achieved (J) maximum tumor coverage within 2 d (P = 0.0063). Purple arrow, injection time point = 0 h. AU, arbitrary units.

Iron-labeled NSC^LLON (Fig. 4H and SI Appendix, Fig. S4 D and E) were followed by MRI for 5 d after injection into the hemisphere contralateral to the tumor-bearing brain of animal subjects (SI Appendix, Fig. S4F), as previously described (62). T2* (effective T2) weighted MRI signal showed NSC^LLON accumulation in the tumor-bearing hemisphere that reached a maximum level at 24 to 48 h after injection (Fig. 4I and SI Appendix, Fig. S4 G and H). The signal from the hemisphere injected with NSC^LLON (contralateral to the tumor-bearing hemisphere) was highest at 3 h after injection and progressively decreased after that (Fig. 4I). Quantitative analysis of R2* data (where R2* is 1/T2*), revealed maximal tumor coverage by NSC^LLON within 48 h from the intracranial injection of cells (P = 0.0063) (Fig. 4J and SI Appendix, Fig. S4G).

NSCs could be immunosuppressive or immunoreactive due to the presence of MHCI and PDL-1 at their cell surface (SI Appendix, Fig. S5A) (66, 67). NSCs express MHCI on the surface (SI Appendix, Fig. S5A). Profiling for checkpoint molecules showed that 20.02 ± 2.26% of NSC^LLON and 39.57 ± 3.9% of NSC^LLOFF expressed a low level of PD-1 ligand (PDL-1) at the cell surface (SI Appendix, Fig. S5 B–D). Modified NSCs do not express PDL-2 or cytotoxic T lymphocyte-associated protein 4 (CTLA4) (SI Appendix, Fig. S5 C and D). The presence of PDL-1 did not affect secreted BiTE^LLON from promoting a strong T cell antiglioma response (NSCs vs. NSC supernatant) (SI Appendix, Fig. S5E).

NSC^LLON, but not BiTEs from NSC^LLOFF, produced BiTE that potently stimulates donor’s lymphocytes, leading to the killing of IL13Rα2-positive GBM cells (Fig. 5 A and B). NSC^LLON and NSC^LLOFF did not induce lymphocyte killing of IL13Rα2-negative cells (Fig. 5C). GBM patient lymphocytes were also activated against GBM cells by NSC^LLON (Fig. 5D). NSC^LLON, but not NSC^LLOFF, up-regulated lymphocyte expression of granzyme B (GrzmB) and Lamp1, markers of degranulation, and mediators of the perforin-associated killing of IL13Rα2⁺ GBM cells. These effects were not observed in the absence of IL13Rα2 (Fig. 5E). BiTEs secreted by the NSC^LLON also induced lymphocyte proliferation (Fig. 5F), activation (Fig. 5 G and H), and production of type 1 cytokines: IL2, IFNγ, and TNFα (Fig. 5 I–K) in cocultures with IL13Rα2-positive GBM cells (Fig. 5 F–K and SI Appendix, Fig. S5 F–K).

Fig. 5.

BiTE^LLON secreted by NSCs is functional in vitro. NSC^LLON produce BiTE protein that induced a robust killing in two IL13Rα2⁺ cell lines, (A) GBM6 and (B) GBM12 (two-way ANOVA, n = 4, **P < 0.01, ****P < 0.0001). (C) BiTE secreting NSCs did not engage donor T cells to kill GBM39, an IL13Rα2-negative cell line. (D) NSC^LLON presence in a coculture, but not NSC^LLOFF, induced GBM patient's T cells' reengagement to kill GBM12 cells (two-way ANOVA, n = 3, ***P < 0.001). (E) Coculture of donor T cells with GBM12 in the presence of NSC^LLON and bead stimulation, but not NSC^LLOFF or absence of stimulation, induced expression of Granzyme B (GrzmB) and Lamp1 (markers of cytotoxic T cell activity), (F) proliferation and expression of activation markers (G) CD69 and (H) CD25 in T cells (one-way ANOVA, n ≥ 3, ****P < 0.0001). Activation of T cells was associated with the production of (I) IL2 (24 h), (J) IFNγ (48 h), and (K) TNFα (48 h) by T cells after coculture with GBM12 in the presence of BiTE^LLON or stimulation with CD2CD3CD28 beads (Stim.), but not in BiTE^LLOFF or absence of any form of stimulation (No stim.) (one-way ANOVA, n ≥ 4, ****P < 0.0001).

NSC^LLON Administration Extends the Survival of Glioma-Bearing Animals.

We injected NSC^LLON proximally (1.5 mm) to GBM12 intracranial tumors in NSG mice (SI Appendix, Fig. S6A). Immunohistochemical analysis of sections of brains harvested on days 3 and 7 following NSC^LLON injection showed robust infiltration of the tumor mass by stem cells and the production of BiTE protein within the tumor and at the tumor periphery (Fig. 6A and SI Appendix, Fig. S6 B and C). We also evaluated NSC^LLON for the duration of in vivo viability by histopathologic analysis of resected mouse brains from animal subjects euthanized at 90 d following NSC^LLON injection (SI Appendix, Fig. S7 A–C), with GBM12 serving as a positive control. There were no Ki67-positive (e.g., proliferating) cells found in the mouse brain injected with NSC^LLON, in contrast to the brain with GBM12 tumors (SI Appendix, Fig. S7B). Only a few NSC^LLON cells were detected near the third ventricle, but not in other brain regions (SI Appendix, Fig. S7C). Moreover, implantation of NSC^LLON in tumor-bearing mice in the absence of T cells did not affect the tumor progression compared to untreated control mice in GBM12- and GBM39-bearing mice (GBM12 median survival: not treated [NT], 22 d; NSC^LLON, 26 d; log-rank test, P = 0.269) (SI Appendix, Fig. S7D) (GBM39 median survival: NT, 84.5 d; NSC^LLON, 71 d; log-rank test, P = 0.095) (SI Appendix, Fig. S7E). Interestingly, the ex-vivo analysis of GBM6 and GBM12 glioma cells harvested from brain xenografts showed a much smaller fraction of glioma cells expressing IL13Rα2 (SI Appendix, Fig.S7 F and G) in comparison to GBM6 and GBM12 cells cultured in vitro (Fig. 1 E and F). Infiltration of NSC^LLON into the tumor mass with the accompanied secretion of BiTE^LLON could be detected in tumors 3 d after NSC^LLON injection (Fig. 6 A, II, III, and IV and SI Appendix, Fig. S6B). Both NSCs and BiTE were also detectable in the tumor at day 7 postinjection, albeit at lower levels (Fig. 6 A, II vs. 6 A, VI and Fig. 6 A, III vs. 6 A, VII).

Fig. 6.

NSC^LLON treatment significantly extends the survival of glioma-bearing animals. (A) NSC^LLON injected superior and 1.5 mm proximal to the tumor mass (for large magnification and orientation concerning NSC^LLON injection, please see SI Appendix, Fig. S6) persist for an extended time, infiltrate tumor mass (direction of infiltration is shown with red arrows), and secrete therapeutic proteins in situ (dotted line marks the tumor edge, DAPI cells, blue; NSCs, green; BiTE^LLON, magenta; I–IV, 3 d and V–VIII, 7 d after injection; NSCs and BiTE expression are shown with white arrows) (Scale bar: 50 μm). (B–D) Animals bearing IL13Rα2-expressing gliomas (1.0 × 10⁵ GBM6 or GBM12) were treated with a single intracranial (i.c.) injection of 1 × 10⁶ NSC^LLON and PB cells, either at the tumor mass i.c. (red, 1.5 × 10⁶ PB) or injected intravenously (i.v.) (blue, 7 × 10⁶ PB). (B) GBM12-bearing animals treated with NSC^LLON and PB i.c. survived significantly longer than not treated (NT) or treated with NSC^VC and PB (Kaplan–Meier survival curves were compared using log-rank tests, P < 0.0001, and adjusted for P values using the Holm–Sidak method: NT vs. NSC^VC P = 0.31, NSC^VC vs. NSC^LLON ***P = 0.0003, n = 8). (C) Similarly, treated mice transplanted with GBM6 cells survived significantly longer from the control NSC^VC or NT animals (Kaplan–Meier survival curves were compared using log-rank tests, P = 0.0002, and adjusted for P values using the Holm–Sidak method: NT vs. NSC^VC P = 0.54, NSC^VC vs. NSC^LLON ***P = 0.0009, n ≥ 8). (D) GBM6-bearing mice treated with an i.c. NSC^LLON and i.v. PB lived significantly longer than mice transplanted with control NSC^VC and i.v. PB animals (Kaplan–Meier survival curves were compared using log-rank tests, ***P < 0.001, n = 11).

We treated GBM12-bearing mice by intratumoral injection with either NSC^LLON or NSC vector control (NSC^VC) and PB lymphocytes (Fig. 6B). Treatment with NSCs secreting therapeutic BiTEs increased average animal survival by 67%. Animals bearing GBM6 tumors were treated using the same protocol as for GBM12-bearing animals (Fig. 6C). GBM6-bearing mice treated with NSC^LLON+ PB lived on average 63.3% longer than animals that were NT or those that received an injection of NSC^VC+ PB . In contrast to GBM6 and GBM12, treatment of mice bearing IL13Rα2-negative GBM39 tumors with NSC^LLON and PB did not extend animals' survival compared to the NT group (SI Appendix, Fig. S7E). Animals bearing GBM6 tumors and treated with intratumoral injection of NSC^LLON but an intravenous infusion of PB also survived significantly longer than mice treated with control NSC^VC (Fig. 6D). Histopathologic analysis of brains from the control and treated animals showed the absence of therapy-related toxicity, including demyelination, encephalomyelitis, or neuronal loss.

Altogether, these data show that NSCs secreting BiTE^LLON deliver a sustained therapeutic protein source for local engagement of CD3⁺ cells for an antiglioma activity.

Discussion

This study reports the development and functional analysis of NSCs secreting novel proteins that stimulate patient-derived T cell antitumor activity in vitro and in vivo. We show that IL13Rα2-targeted BiTEs, secreted by NSC^LLON, promote T cell killing of IL13Rα2⁺ tumor cells by engaging the tumor cell antigen target with T cell CD3ε. The genetically modified NSCs produce and secrete BiTE in vivo while infiltrating the tumor mass, with treated animals bearing intracranial tumors experiencing significant survival benefit.

Demonstration of target specificity is essential for limiting undesirable side effects from BiTE treatment of cancer patients (30, 32, 44, 68), and, for this reason, we have developed BiTE targeting IL13Rα2 (51, 52). We show BiTE^LLON binding specificity for PDX cells expressing IL13Rα2 and T cell CD3ε, with this dual binding specificity promoting T cell-mediated cell death to IL13Rα2-positive tumor cells. The absence of cell killing in T cell–tumor cell cocultures with nonbinding BiTE^LLOFF, and in cocultures where tumor cells lack the IL13Rα2 antigenic target, supports the specificity of BiTE^LLON. Maximal BiTE antitumor activity was observed at a concentration of 2.5 µg/mL and a T:E cell ratio (1:5), suggesting that relatively low amounts of recombinant protein are needed for antineoplastic effect (43).

Prolong T cells' stimulation results in the expression of activation markers PD-1, Tim3, Lag3, and T cell exhaustion (69, 70). GBMs are known to suppress lymphocyte activation processes and promote T cell exhaustion characterized by T cell expression of PD-1, Tim3, and Lag3 (57, 58). Our results show that BiTE^LLON activates patient-derived peripheral and tumor-infiltrating T cells to kill IL13Rα2⁺ GBMs, irrespective of T cell expression of these proteins.

It has been shown that activation of T cells by targeted immunotherapeutics can lead to aberrant cytokine production and cytokine release storm (CRS) in treated patients (44, 71). In our study, donor T cells stimulated directly with BiTE^LLON, or by BiTE^LLON secreted by NSCs, produced IFNγ and TNFα, but only in the presence of IL13Rα2-expressing tumor cells. Interestingly, we found that CD8⁺ T cells, positive for PD-1, Tim3, and Lag3, are major contributors to the pool of secreted cytokines when treated with BiTE^LLON, while in the presence of IL13Rα2-expressing cells. Notably, BiTE^LLON successfully activated both PB lymphocytes and TIL cells against IL13Rα2-expressing tumor in cases when T cells were harvested from patients receiving steroid treatment, which is known to suppress T cell activity (72, 73).

Unlike blood vessels of a healthy brain, GBM vasculature is considered to be disorganized and leaky, allowing larger molecules to pass from circulation to tumor tissue. Indeed, systemically delivered BiTEs, directed against the tumor-associated antigen epidermal growth factor variant III (EGFRvIII), were shown to reach the intracranial tumor in an amount that was sufficient to reduce tumor burden (74). Nonetheless, it is expected that the systemic administration of BiTE for treating GBM would require repeated treatments. Interventions such as focused ultrasound transiently opening the blood–tumor barrier (74) might be an option to improve the delivery of therapeutic protein to GBM. An approach that would potentially circumvent the need for repeated administrations is to deliver treatment locally and using a platform for sustaining therapeutic presence. To achieve this, we have investigated direct intracranial administration of BiTE-secreting NSCs. Therapeutic NSCs' migration toward and dissemination within brain tumors in mice, following systemic and intracranial NSCs' administration routes, have been previously demonstrated by our group and colleagues (67, 74, 75). In the current study, intracranially administered NSC^LLON show tropism to and dissemination within glioma xenografts. Our results also show that NSC^LLON persist and secrete BiTE for at least 7 d following administration, thereby extending the supply of BiTE within the tumor. Others have reported the persistence of engineered human NSCs in a mouse brain for up to 12 wk (74). Nevertheless, it is anticipated that there will be a decrease in the number of NSCs within the tumor over time, as indicated by MRI (Fig. 4 I and J and SI Appendix, Fig. S4G). This results in decreased levels of locally secreted BiTEs, contributing to this therapy’s limited antitumor efficacy. Sustained or repeated delivery of therapeutic NSCs to brain tumor patients, using already clinically available methods, such as an Ommaya reservoir placement, would presumably extend treatment benefit.

The drawback of many targeted therapies against tumors is antigen escape (27, 28, 49), necessitating the development of therapeutics to multiple targets. This is partly due to the heterogeneous expression of TAA, including IL13Rα2, in GBM tissue (28, 55, 75). In our experimental models, only a fraction of GBM12 and GBM6 cells express IL13Rα2 in vivo, recapitulating the pattern of antigen expression in GBM. Thus, treatment-associated selection against TAA-positive cells is another factor that would likely limit the therapeutic efficacy of any BiTE therapy directed to a single tumor antigen. Antigen escape in tumors treated with CAR T cells directed to a single target has been demonstrated in preclinical and clinical GBM studies (27, 28, 49). The CAR T cells secreting BiTEs have recently been investigated as a platform for targeting multiple tumor antigens (48). However, the short life of CAR T cells in the tumor environment is expected to translate into a similarly short presence of BiTEs produced by T cells. Stem cells engineering strategy or loading NSCs with vectors encoding for multispecific molecules could overcome the limitations mentioned above. The oncolytic virus expressing BiTE, cytokine, and checkpoint blockade has been generated and tested in combination with CAR T cells (76). A multifunctional vector delivered to the tumor by NSCs will ensure a robust intratumoral distribution of therapeutic molecules.

Tumor inhibition of T cell function through an aberrant expression of immune checkpoint proteins and tumor MHC down-regulation is a critical mechanism by which GBM escapes antitumor immune response (39, 57, 58). Based on this knowledge, immune-checkpoint inhibitor (3) therapy has been tested in clinical trials. However, the results have shown a lack of survival benefit for treated GBM patients (3, 77–79). Based on the results presented here, it would be interesting to compare the antitumor activity of local NSC^LLON administration in combination with systemically or locally delivered checkpoint inhibitors. Interestingly, a BiTE molecule engineered to interact with T cells and PD-L1 has been recently tested in preclinical models of PD-L1–positive tumors, with results showing extended survival of treated animal subjects (80). Based on these findings and results presented here, one can envision a scenario in which multiple types of modified NSCs, each producing BiTEs that target distinct tumor antigens, were used in treating GBM patients.

It is relevant to consider shortcomings associated with animal models that lack a fully functional immune system and, therefore, have an inherent deficiency in recapitulating the tumor microenvironment of a GBM. Future studies utilizing immunocompetent hosts will be necessary for revealing BiTE interactions with a fully functional host immune system (81). The use of humanized mouse models, in which mice have been modified for durable production of human immune cells, or the use of genetically engineered mice that express the human CD3 receptor, will also be of interest.

Thus, the current study represents a starting point for expanding the investigation of BiTEs produced by genetically modified NSCs, for treating GBM. The results we have in hand are encouraging and merit continued exploration of this therapeutic modality.

Materials and Methods

For a detailed description of cell lines, generation of BiTE molecules and NSCs secreting BiTEs, functional in vitro assays, flow cytometry, immunocytochemistry, animal tumor, and MRI studies, statistics, and data analysis, please see SI Appendix, Materials and Methods.

Data Availability

All study data are included in the article and/or SI Appendix.