Bi-specific T cell-engaging antibody triggers protective immune memory and glioma microenvironment remodeling in immune-competent preclinical models

Abstract

Background

Bispecific T cell-engagers (BTEs) are engineered antibodies that redirect T cells to target antigen-expressing tumors. BTEs targeting tumor-specific antigens such as interleukin 13 receptor alpha 2 (IL13RA2) and epidermal growth factor receptor variant III (EGFRvIII) have been developed for glioblastoma (GBM). However, there is limited mechanistic understanding of the action of BTE since prior studies were mostly conducted in immunocompromised animal models. To close this gap, the function of BTEs was assessed in the immunosuppressive tumor microenvironment (TME) of orthotopic and genetically engineered mouse models (GEMM) with intact immune systems.

Methods

A BTE that bridges CD3 epsilon on murine T cells to IL13RA2-positive GBM cells was developed, and the therapeutic mechanism was investigated in immunocompetent mouse models of GBM. Multicolor flow cytometry, single-cell RNA sequencing (scRNA-seq), multiplex immunofluorescence, and multiparametric MRI across multiple preclinical models of GBM were used to evaluate the mechanism of action and response.

Results

BTE-mediated interactions between murine T cells and GBM cells triggered T cell activation and antigen-dependent killing of GBM cells. BTE treatment significantly extended the survival of mice bearing IL13RA2-expressing orthotopic glioma and de novo forming GBM in the GEMM. Quantified parametric MRI validated the survival data, showing a reduction in glioma volume and decreased glioma viability. Flow cytometric and scRNA-seq analyses of the TME revealed robust increases in activated and memory T cells and decreases in immunosuppressive myeloid cells within the brains of mice following BTE treatment.

Conclusions

Our data demonstrate that the survival benefits of BTEs in preclinical models of glioma are due to the ability to engage the host immune system in direct killing, induction of immunological memory, and modulation of the TME. These findings provide a deeper insight into the mechanism of BTE actions in GBM.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Bi-specific T cell-engaging antibodies (BTEs) targeting interleukin 13 receptor alpha 2 and epidermal growth factor receptor variant III have been developed and shown to activate T cells that mediate killing of glioma cells in vitro and in vivo.

WHAT THIS STUDY ADDS

• This study demonstrates that BTEs elicit in situ tumor immune memory within the central nervous system in immunocompetent preclinical models of glioblastoma.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Insights into the mechanism of action of BTEs inform response biomarkers for inclusion in window-of-opportunity clinical trial assessments and provide a mechanistic basis for the rational selection of future combinatorial strategies.

Introduction

Glioblastoma (GBM) is an aggressive and incurable brain tumor with a poor prognosis.¹ The current standard of care for primary disease involves surgical resection, followed by radiation therapy and chemotherapy with the alkylating agent temozolomide.² Despite significant advancements in surgery, radiation, and systemic therapy, GBM progression is inevitable. Median overall survival is approximately 15 months, and the 5-year survival rate is less than 5%.^{3 4} The absence of an established standard of care at recurrence underscores the urgent need for novel therapies to improve outcomes in GBM. GBM is regarded as among the most challenging solid tumors to treat due to the blood-tumor barrier limiting the penetration of therapeutics into the tumor microenvironment (TME), immune suppressive TME, heterogeneous expression of tumor-associated antigens, and the limited T cell presence within the tumor.⁵

T cell-based immunotherapeutic treatments in hematopoietic malignancies have been increasingly successful. Among those treatments are bi-specific T cell engagers (BTEs), a class of engineered bispecific antibodies with potent anticancer properties. BTEs consist of two single-chain variable fragments (scFvs) derived from different monoclonal antibodies fused by a peptide linker. One of the scFv of the BTE recognizes a T cell stimulatory receptor, such as the CD28 or CD3 epsilon (CD3E) subunit of the T cell receptor complex. The other scFv is specific for a tumor antigen. In bridging T and tumor cells, BTEs activate T cells and direct T cell-mediated killing of malignant cells. Unfortunately, there are significant hurdles in the development of BTE therapy for solid tumors such as GBM. Our group has previously shown the efficacy of a BTE targeting the glioma-associated antigen interleukin 13 receptor subunit alpha 2 (IL13RA2) in preclinical patient-derived xenograft (PDX) models of GBM. Others have demonstrated a significant therapeutic effect of BTE targeting epidermal growth factor receptor variant III (EGFRvIII), a glioma-specific antigen, in PDX and murine glioma models⁶ and a clinical study evaluating the safety of BTEs is anticipated (NCT04903795). To overcome the issue of glioma antigen heterogeneity, a DNA-encoded tri-specific T cell engager has been developed to target IL12Rα2 and EGFRvIII.^{7 8} However, these preclinical studies were conducted in immunocompromised hosts lacking a fully functioning immune system, rendering them unable to recapitulate the highly immunosuppressive TME in human GBM.⁹

As a potential strategy to overcome the blood-tumor barrier,¹⁰ BTEs with varying target specificities continue to be developed.811,14 Gaining a deeper understanding of the mechanistic actions of BTEs in fully immunocompetent GBM models is crucial for informing future clinical study designs and the rational selection of combinatorial approaches. Therefore, this study was designed to investigate the therapeutic mechanisms of BTEs in orthotopic and genetically engineered mouse models (GEMM) of GBM. Our findings demonstrate that the activation of murine T cells depends on the binding of BTE to its target, IL13RA2, both in vitro and in vivo. Moreover, BTE treatment increases the proportion of activated T cells within glioma and leads to a higher frequency of memory T cells in the brains of long-term surviving (LTS) mice. Together, these findings highlight the ability of BTEs to activate host immunity, promote T cell memory, and sustain therapeutic efficacy of glioma.

Results

Generation and functional characterization of the CD3×IL13RA2 BTE

scFvs derived from the monoclonal antibody 2C11-145 binding murine CD3e¹⁵ and the monoclonal antibody against human IL13RA2 (clone 47)^{16 17} were used to generate a BTE protein suitable for studies in an immunocompetent host. The two scFvs were genetically linked with a 23-amino-acid GlyS linker.¹⁸ A negative control BTE (NC-BTE) was generated by aligning and replacing the unmatched amino acids in the CDR3 domain of variable domain of the light chain (VL) of scFv targeting IL13RA2 with those from the scFv p588 sequence—a previously established negative control antibody (MOPC1).¹⁸ Constructs were generated by subcloning the codon-optimized complementary DNA (cDNA) encoding BTE and NC-BTE in pLVX-IRES-ZsGreen1 lentiviral constructs (figure 1A). To produce BTE proteins, 293T cells were transduced with BTE and NC-BTE constructs and subsequently flow-sorted to enrich for the brightest ZsGreen1-positive 293T cells. A western blot of the purified BTE confirmed a protein with the anticipated molecular weight of approximately 55 kDa (figure 1B). The specificity of BTE to IL13RA2 was confirmed with a plate ELISA, where BTE, but not NC-BTE, bound specifically to recombinant IL13RA2, with no detected binding IL13RA1 (figure 1C).

Figure 1. Generation and functional characterization of the CD3×IL13RA2 BTE. (A) Schematic representation of the BTE consisting of the scFv targeting murine CD3E specific component, derived from 2C11-145, and the scFv targeting IL13RA2, derived from clone47, fused via a flexible GlyS (Gly4Ser) linker. NC-BTE differed from the BTE in the VH domain of IL13RA2-scFv, which was swapped with the VH domain of scFv p588 of the non-specific MOPC21 antibody. (B) A western blot of the affinity-purified BTE at various concentrations was detected using an anti-His HRP conjugated antibody. A single band ∼was observed at 55 kDa. (C) ELISA assay showing dose-dependent and specific binding of BTE to recombinant IL13RA2 (red circles), but not to IL13RA1 (blue triangles); NC-BTE (black circles) does not bind to IL13RA2. (D) Cr⁵¹ release assay of murine T cells and glioma cells treated with BTE (red), NC-BTE (white), or Dynabeads Mouse T-Activator CD3/CD28 (gray) showing BTE, but not NC-BTE, induced dose-dependent killing of IL13RA2-expressing SMA560 and GL261 murine glioma cells (E:T ratio=20:1; incubation time=24 hours; n=3/group). (E) Flow cytometric analysis of murine T cells co-cultured with IL13RA2-expressing SMA560 and GL261 murine glioma cells showing BTE, but not NC-BTE, induced robust expression of T cell activation markers CD69 and CD25 (E:T ratio=20:1; incubation time=24 hours; n=3/group). One-way ANOVA followed by Tukey’s HSD test for pairwise comparisons. ANOVA, analysis of variance; BTE, bispecific T cell-engager; Cr⁵¹, chromium-51; E:T, effector-to-target; HSD, Honest Significant Difference; HRP, horseradish peroxidase; IL13RA2, interleukin 13 receptor alpha 2; NC-BTE, negative control BTE; mCD3E, mouse CD3 epsilon chain; scFv, single-chain variable fragment; VL, variable domain of the light chain; VH-MUT, mutated variable domain of the heavy chain.

Next, the ability of BTE to activate T cells and direct T cell-mediated killing of murine glioma cells was evaluated in vitro. The murine glioma cell lines GL261, SMA560, and CT2A, previously modified to express IL13RA2,¹⁹ were used as target cells assessed via chromium-51 (Cr⁵¹)-release. T cells activated with CD3/CD28/CD2 immuno-beads served as a positive control. The BTE, but not NC-BTE, significantly increased the killing of SMA560-IL13RA2 and GL261-IL13RA2 cells in a dose-dependent manner (figure 1D); an effect not seen in the parental SMA.560 and GL261 cells (online supplemental figure 1A). The BTE also triggered significant expression of the early activation markers CD25 and CD69 on CD8+T cells on engagement with SMA560-IL13RA2 and GL261-IL13RA2 cell lines (figure 1E). This activation was further confirmed with live cell imaging, in which increased fluorescent intensity was observed in NOD.Nr4a1GFP/Cre mouse T cells following BTE treatment (online supplemental figure 1B and C). Collectively, these data confirm the creation of a functional BTE and corresponding negative control NC-BTE protein suitable for studies in immunocompetent syngeneic models of glioma.

BTE treatment in IL13RA2-expressing preclinical models enhances survival and expands T cell effector responses in the TME

To clarify if the BTE could be distributed into the brains of GL261-IL13RA2 glioma-bearing mice, the AF647-labeled BTE was administered intravenously into the tail vein, and longitudinal dynamic live fluorescent imaging was performed. The BTE was detected in the brain 3 hours after administration, with peak signal occurring between 3 and 6 hours and diminishing between 24 and 72 hours (figure 2A). To gain insight into the tumor penetration of the BTE, mice bearing GL261-IL13RA2 gliomas were injected intravenously with the Cy5-labeled BTE and the brains were collected at 3 hours post-injection. Confocal microscopy confirmed the heterogeneous accumulation of Cy-5-labeled BTE in the extravascular space of the tumor (figure 2B i-ii). Cy-5-labeled BTE was absent in the brain outside of the tumor (figure 2B i-iii). Furthermore, quantified Cy5-labeled BTE signal as a function of distance from lectin-positive blood vessels showed no preferential accumulation of BTE near tumor vasculature (n=5) (figure 2C, online supplemental figure 3A).

Figure 2. BTE enhances survival and modulates the tumor microenvironment in mice bearing IL13RA2-expressing gliomas. (A) Dynamic live fluorescent imaging of AF647-labeled BTE showing penetration and persistence in the brains of GL261-IL13RA2 tumor-bearing mice (fluorescent signal measured as radiance (p/s/cm²/sr); image acquisition time=3, 6, 24, 48, and 72 hours post AF647-BTE injection; n=3). (B) Confocal images of brain cryosections showing the accumulation of the Cy5-labeled BTE in the tumor (i,ii), but not in surrounding brain tissue (i,iii). A dashed line marks the tumor border. Green: blood vessel stained with Lectin-FITC; blue: Hoechst nuclear stain; red: Cy5-labeled BTE. Scale bar is 100 µm. (C) Line profile analysis showing Cy5-BTE fluorescence is concentrated in tumor regions and not specifically enriched near lectin-positive vessels. (D) Kaplan-Meier survival curve of SMA.560-IL13RA2 glioma-bearing mice treated with either BTE (200 µg/animal) (red), NC-BTE (200 µg/animal) (blue), or saline (black). Wald test from Cox proportional hazards regression analysis. (E) The experimental schema used for the flow cytometric analysis of the brains of SMA.560-IL13RA2 glioma-bearing mice following treatment with either BTE or NC-BTE (n=5/group). Flow cytometric analysis revealed (F) no change in CD4 helper T cell (CD45+CD4+), (G) increase in regulatory T cell (CD45+CD4+ FOXP3+), (H) increase in CD8 T cell (CD45+CD8+), (I) increase in memory-like CD8 T cell (CD45+CD8+CD127HiKLRG1−), (J) increase in tissue-resident memory CD8 T cells (CD45+CD8+ CD127 Hi KLRG1− CD103+), (K) no change in short-lived memory CD8 T cells (CD45+CD8+ CD127 Hi KLRG1+). One-way ANOVA followed by Tukey’s HSD test for pairwise comparisons. ANOVA, analysis of variance; BTE, bispecific T cell-engager; HSD, Honest Significant Difference; FITC, fluorescein isothiocyanate; FOXP3, forkhead box P3; IL13RA2, interleukin 13 receptor alpha 2; i.p., intraperitoneal; KLRG1, killer cell lectin-like receptor G1; MS, median survival; NS, non-significant; NC-BTE, negative control BTE; p.adj, adjusted p value.

Next, a dose titration of BTE at 50 µg/mouse or 200 µg/mouse was conducted in SMA.560-IL13RA2 glioma-bearing mice. These mice were treated with the BTE or saline on 7, 9, and 11 days post-glioma implantation and followed for survival (online supplemental figure 2A). Both concentrations of BTE (50 µg/animal: median survival (MS)=18, n=8, p=0.0298 relative to the control; 200 µg/animal: MS=27.5, n=8, p=0.0054 relative to the control) extended the survival of mice compared with saline (MS=15, n=8) (online supplemental figure 2A). No statistical difference in survival was noted between the 50 and 200 µg dose. To confirm the antigen specificity of the BTE, a separate survival experiment was conducted in the parental SMA.560 glioma-bearing mice, where no increase in survival was found between the BTE, NC-BTE, and saline groups (online supplemental figure 2B). Based on these results, a survival analysis was conducted in the SMA.560-IL13RA2 glioma-bearing mice treated at 200 µg/animal with either the BTE, NC-BTE, or saline on days 7, 9, and 11 post-glioma implantation (figure 2D). There was no difference in the MS between saline (MS=17, n=7) or the NC-BTE treated group (MS=19, n=8) (figure 2D). The BTE significantly extended the survival of mice (MS=33, n=7, p=0.0035) with 40% surviving over 150 days (figure 2D). A similar survival benefit of the BTE was seen in the CT2A-IL13RA2 (online supplemental figure 2C) and GL261-IL13RA2 murine glioma models (online supplemental figure 3B).

To determine how the BTE affects the immune composition within the TME, SMA.560-IL13RA2 glioma-bearing mice were euthanized 3 days after the last dose of either BTE (n=5) or NC-BTE (n=5) (figure 2E). The brains were processed into single-cell suspensions, the cells stained with fluorochrome-conjugated antibodies against immune cell markers and analyzed by multicolor flow cytometry. Although there was no difference in the frequency of CD4 T cells between treatment groups (figure 2F), there was an increased frequency of regulatory T cells (Tregs) (CD45⁺CD4⁺FOXP3⁺; adjusted p value (p.adj)=0.040) (figure 2H), CD8⁺ T cells (CD45⁺CD8⁺; p.adj=0.039) (figure 2H), memory-like CD8⁺ T cells (CD45⁺CD8⁺CD127^hi; p.adj=0.023) (figure 2I), tissue-resident memory CD8⁺ T cells (CD45⁺CD8⁺CD127^hiCD103⁺; p.adj=0.009) (figure 2J), and monocytic myeloid-derived suppressor cells (MDSC)(CD45⁺CD11B⁺CD11C⁻LY6G⁻LY6C⁺; p.adj=0.050) (online supplemental figure 2D) in the mice treated with BTE relative to NC-BTE. There was no statistical difference in short-lived memory CD8⁺ T cells (CD45⁺CD8⁺CD127^loKLRG1⁺; p.adj=0.102) (figure 2K), polymorphonuclear MDSC (CD45⁺CD11B⁺CD11C⁻LY6G⁺LY6C⁻; p.adj=0.080) (online supplemental figure 2E), or dendritic cells (CD45⁺CD11B⁺CD11C⁺; p.adj=0.159) (online supplemental figure 2F).

To understand if BTE treatment was dependent on CD8 T cells, we conducted a survival experiment in CD8 knockout mice bearing GL261-IL13RA2 murine glioma. No survival benefit was recorded in BTE-treated animals compared with controls, suggesting BTE efficacy is dependent on CD8 T cells (online supplemental figure 3C). Additionally, we performed an experiment to compare BTE administration using intravenous and subcutaneous (s.c.) route, and found that intravenous (MS=26.5, n=10, p=0.031), but not s.c. (MS=19.5, n=10, p=0.391) treatment, significantly extended survival of mice bearing GL261-IL13RA2 compared with control (MS=20.0, n=8) (online supplemental figure 3D).

To further characterize this T cell expansion, both the glioma and spleen from CT2A-IL13RA2-bearing mice were analyzed by flow cytometry (online supplemental figure 2G). Similar to the findings in the SMA.560-IL13RA2 model, there was a significant increase in the frequency of Tregs in the brain following BTE treatment (p.adj=0.039), but not in the spleen (online supplemental figure 2H). Markers of T cell cytotoxicity and activation were increased, including granzyme B (GZMB) (p.adj=0.006) (online supplemental figure 2I), CD69 (p.adj=0.015) (online supplemental figure 2J), and CD25 (p.adj=0.001) (online supplemental figure 2K) in brain CD8 T cells, but not in spleen. The brain CD8 T cells showed upregulated lymphocyte-activation gene 3 protein (Lag3) expression (p.adj=0.036) (online supplemental figure 2L), but not T-cell immunoglobulin and mucin domain 3 (Tim3) (online supplemental figure 2M) or programmed cell death protein 1 (PD1) (online supplemental figure 2N). Notably, there was an increased frequency of CD44⁺CD62L⁻CD8⁺ effector resident T cells in the brain (p.adj=0.001), but not the spleen, of BTE-treated animals (online supplemental figure 2OP).

BTE treatment is protective of glioma rechallenge and generates glioma resident immunological memory

To clarify if the BTE triggers protective immunological memory, LTS SMA.560-IL13RA2 mice from the prior experiments were rechallenged in the contralateral hemisphere with parental SMA.560 glioma cells (n=5) alongside age-matched control mice (n=5). While age-matched control mice died within 20 days, the SMA.560 rechallenged SMA.560-IL13RA2 mice survived over 210 days without neurological symptoms (figure 3A). These long-term survivors were then euthanized, and the brains were examined by histology (n=2/group) and flow cytometry (n=3/group). H&E staining revealed the presence of gliomas in control mice that died within 20 days, but the absence of gliomas in the LTS (figure 3B). Immunohistochemical staining of CD8 on serial tissue sections revealed a wide distribution of CD8-positive cells throughout the brain of rechallenged LTS mice (figure 3C). Quantitative analysis showed significantly more CD8+T cells in the brains of LTS compared with controls in hemispheres both ipsilateral (p.adj=2.22×10–16) and contralateral (p.adj=4.78×10–7) of the glioma implantation site (figure 3D). In flow cytometric analysis of LTS and control brains, there was a significant increase in the frequency of CD8 effector resident T cells (CD44⁺CD62L⁻) (p.adj=0.0035) (figure 3E) and tissue-resident memory CD8 T cells (CD44⁺ CD62L⁻CD103⁺) (p.adj=0.028) (figure 3F) in the brains of rechallenged LTS mice compared with control. Moreover, interferon (IFN)-γ and tumor necrosis factor α expression was increased in the CD8+T cells of the LTS compared with control (p.adj=0.093) (figure 3G). Furthermore, despite high-grade gliomas being highly infiltrated by myeloid cells, LTS mice exhibited significantly lower frequency of CD11b⁺ cells in the brain compared with control (p.adj=0.001) (figure 3H).

Figure 3. The rechallenge of LTS mice with gliomas demonstrated the generation of memory. (A) Kaplan-Meier survival curve of LTS (red) and age-matched control (black) mice following implantation of parental SMA.560 murine gliomas into the contralateral brain hemisphere to the primary tumor implantation site. Animals were followed for survival, and statistical significance was tested via the Wald test from Cox proportional hazards regression analysis. (B) The mice of LTS (210 days post-secondary injection) and age-matched controls (at endpoint) were sacrificed and stained using H&E to evaluate for the presence of tumors in the ipsilateral and contralateral hemispheres of the implantation site. Scale bar represents 2.5 mm. (C, D) Serial sections of the same brains stained for infiltrating CD8+T cells. (C) Representative image of one LTS-ipsilateral section with magnified images in the cerebrum (arrowheads) and the cerebellum (arrowheads). The scale bar represents 2.5 mm for the H&E image and 100 µm for the magnified anti-CD8 IHC. (D) Quantitative analysis of CD8+T cells in the ipsilateral and contralateral hemispheres of the brain of LTS, age-matched control, and non-tumor-bearing mice. One-way ANOVA. (E-H) Flow cytometric analysis of freshly dissected brains of LTS (red) and age-matched control (white); n=3/group. (E) A representative contour plot shows an increased frequency of CD8 resident T cells (CD44+CD62L⁻) in the brains of LTS (left) compared with the control (right). Quantified frequency of (F) tissue-resident memory CD8 T cells (CD8+CD44+ CD62 L⁻CD103+), (G) inflammatory cytokine-producing CD8 T cell (CD8+TNFα +IFNγ+), and (H) total population of myeloid cells (CD45+CD11B+). Differences among multiple groups were evaluated using one-way or two-way ANOVA with Tukey’s HSD test. ANOVA, analysis of variance; HSD, Honest Significant Difference; IFN, interferon; IHC, immunohistochemistry; LTS, long-term survival; MS, median survival; TNF, tumor necrosis factor; UD, undetermined.

BTEs have a therapeutic effect in high-grade GEMMs

To more closely model human GBM, the BTE was evaluated in a GEMM with de novo gliomagenesis in adult mice.²⁰ This GEMM generates IL13RA2-expressing gliomas with loss of p53 and phosphatase and tensin homolog (PTEN) and overexpression of platelet derived growth factor subunit B (PDGFB). Histological analysis of GEMM gliomas revealed key morphological characteristics of human GBM, including pseudo-palisading necrosis, microvascular proliferation, hemorrhage, heterogeneous expression of IL13RA2, infiltration of CD11b⁺ macrophages, and a low frequency of CD3⁺ T cells within the tumor bed (figure 4A). As an overexpression model, the levels of IL13RA2 expression in the GEMM were compared with those of two human PDX GBM cell lines, GBM6 and GBM38. All three cell lines showed similar expression of IL13RA2, presented as mean fluorescent intensity and percent of positive cells (online supplemental figure 3E). Treatment of the GEMM with the BTE (n=20, MS=50.5 days) significantly extended survival compared with control (n=19, MS=41 days) (p.adj=0.0045) (figure 4B). To further characterize the changes within TME induced by the BTE, cells from the brain and spleen were analyzed via flow cytometry 48 hours after the final treatment (figure 4C). In this model, there was no difference in Tregs in the brain between the BTE and control group (CD45⁺CD4⁺FOXP3⁺) (p.adj=0.613) (figure 4D). Similar to the findings in the implanted orthotopic gliomas, there was increased expression of the CD69 activation marker in the brain, but not the spleen, of the BTE-treated animals (p.adj=0.016) (figure 4E). In contrast, there was no difference in the GZMB cytotoxicity marker GZMB (p.adj=0.542) (figure 4F), possibly related to the kinetic differences of expression secondary to the time point of sampling between the models. There was an increase in effector resident CD8 T cells (CD44⁺ CD62L⁻) (figure 4G) and tissue-resident memory CD8 T cells (CD44⁺ CD62L⁻CD103⁺) (p.adj=0.022) (figure 4H). Lastly, multiplex seqIF visualization of the TME of GEMM treated with BTE and saline revealed that IL13RA2 (green) expression is retained in tumor tissue after 3 weeks of BTE treatment (figure 4I).

Figure 4. BTE treatment results in therapeutic benefits in genetically engineered mice bearing de novo glioma tumors. (A) Histopathological analysis of RCAS-P53fl/fl-PTENfl/fl-IL13RA2 GEMM glioma model with high magnification images showing morphological characteristics similar to human GBM. H&E (top-left) shows areas of pseudopalisading necrosis (i), microvascular proliferation (ii), and hyperplasia (iii). IHC staining revealed many infiltrating CD11B+macrophages (top-right), scarce infiltrating CD3+T cells (bottom-right), and heterogeneous IL13RA2 expression (bottom-left) within the tumor bed. (B) Kaplan-Meier survival curve of the de novo GEMM treated with BTE (red) (n=20) or saline (black) (n=19). Animals were followed for survival, and statistical significance was tested via the Wald test from Cox proportional hazards regression analysis. (C) An experimental setup was used in flow cytometric analysis of the brains and spleens of the GEMM following treatment with either BTE or saline (n=5/group). Flow cytometric analysis revealed (D) no change in regulatory T cell (CD45+CD4+ FOXP3+) and brain-specific increases in (E) activated CD8 T cell (CD45+CD8+ CD69+), (F) cytotoxic CD8 T cell (CD45+CD8+ GZMB+), (G) resident CD8 T cells (CD45+CD8+ CD44+ CD62L−), and (H) tissue-resident memory CD8 T cells (CD45+CD8+ CD44+ CD62L− CD103+). (I) Multiplex IF of saline (left) and the BTE (right) treated GEMM demonstrating that there was no difference in IL13RA2 expression. Visualization of tumor architecture was done via staining of the nuclei with DAPI, glioma cells with GFAP, and blood vessels with CD31. Scale bars equal to 50 µm. BTE, bispecific T cell-engager; DAPI, 4′,6-diamidino-2-phenylindole; DF-1, a continuous cell line derived from chicken embryonic fibroblasts; FOXP3, forkhead box P3; GFAP, glial fibrillary acidic protein; GBM, glioblastoma; GEMM, genetically engineered mouse model; GZMB, granzyme B; IF, immunofluorescence; IHC, immunohistochemistry; IL13RA2, interleukin 13 receptor alpha 2; i.p., intraperitoneal; MS, median survival; RCAS, replication-competent avian sarcoma virus; PDGFB, platelet derived growth factor subunit B; PTEN, phosphatase and tensin homolog; p.adj, adjusted p value.

BTE treatment remodels the TME to immune effector responses in de novo gliomas

To gain insights into how the BTE impacts the TME, scRNA-seq was performed in the GEMM model following BTE treatment. Similar to human GBM, scRNA-seq revealed a GEMM TME dominated by the resident and peripheral myeloid cells (figure 5A). Uniform manifold approximation and projection (UMAP) analysis as a function of treatment showed fewer peripheral myeloid cells in the BTE-treated animals compared with saline treatment (figure 5A), likely due to the reduced glioma size in the BTE-treated animals. Subclustering and subsequent dimensional reduction of T and natural killer cell populations (online supplemental figure 3A) revealed distinct treatment-dependent differences in cell distributions within UMAP projections (figure 5B). Notably, the relative frequency of these cells remained similar between treatment groups (online supplemental figure 3B). Gene ontology enrichment analysis of significant differentially expressed genes (DEGs) in the CD8 T cell cluster from the BTE-treated group mapped to pathways involved in cytokine production, T cell differentiation, T cell activation, and antigen processing and presentation (figure 5C). Evaluation of specific genes key to CD8 T cell effector function revealed elevated expression of genes involved in activation, cytotoxicity, and memory, and reduced expression of exhaustion genes (figure 5D). Memory and effector function were validated by multiplex immunofluorescence of the GEMM TME, which showed increased CD8 and CD4 T cells co-expressing T-cell factor 1 (TCF1 (figure 5E) and IFNγ (figure 5F) in BTE-treated animals compared with saline.

Figure 5. ScRNA-seq reveals remodeling of tumor microenvironment in de novo glioma tumors on BTE treatment. (A) UMAP visualization of single-cell RNA sequencing data from cells isolated from the tumors of de novo GEM glioma models. The data were derived from two independent experiments, each with two treatment groups (n=5/group): BTE-treated and saline-treated. In each experiment, isolated cells from each mouse were pooled together for each group, resulting in a total of four samples (two BTE-treated and two saline-treated) totaling 64,124 cells. CCA integration was used to combine data from the two experiments into a shared UMAP, with the left panel colored by cell cluster identity and the right panel colored by treatment group. Cells were enriched for CD45-positive populations at a CD45+:CD45- ratio of 8:2. (B) UMAP visualization of T and NK cell subclusters following CCA integration, displayed based on saline (top) or BTE treatment (bottom). Subclusters include CD8 T cells (red), CD4 T cells (blue), regulatory T cells (green), γδ T cells (brown), and NK cells (purple). (C) Gene ontology analysis of DEGs in CD8 T cells that were upregulated in BTE-treated samples compared with saline. (D) Heatmap of DEGs in CD8 T cell cluster across all samples related to activation-cytotoxicity, CD8 T cell memory, and exhaustion as a function of treatment. (E) Multiplex IF imaging of the gliomas in the GEMM model treated with saline (top) or the BTE (bottom). Visualization of tumor architecture was done via staining of the nuclei with DAPI, glioma cells with GFAP, and blood vessels with CD31. High magnification images show glioma infiltration with CD8 (green) and CD4 (yellow) T cells with intracellular TCF1 (magenta) following BTE treatment, denoted with single arrows; Scale bars: 20 µm. (F) Multiplex immunofluorescence imaging of the gliomas in the GEMM model treated with saline (top-left) or the BTE (top-right) Sections were stained for tumor architecture DAPI, GFAP, and CD31, and T cell subsets with CD8 (green) and CD4 (yellow). Below each image are high magnification images of glioma infiltrated CD8 (bottom-left) and CD4 (bottom-right) T cells with intracellular IFNγ (red). Scale bars for bigger panels: 20 µm, for smaller panels: 10 µm. BTE, bispecific T cell-engager; cDC, conventional dendritic cells; CCA, canonical correlation analysis; DAPI, 4′,6-diamidino-2-phenylindole; DEGs, differentially expressed genes; GEMM, genetically engineered mouse model; GFAP, glial fibrillary acidic protein; IF, immunofluorescence; IFN, interferon; MDSC, myeloid-derived suppressor cell; NK, natural killer; ODC, oligodendrocytes; pDC, plasmacytoid dendritic cells; p.adj, adjusted p value; scRNA-seq, single-cell RNA sequencing; TCF1, T-cell factor 1; UMAP, uniform manifold approximation and projection.

Volumetric MRI reveals tumor-specific radiographic responses and increased TME complexity after BTE treatment

To gain deeper insights into BTE effects on tumor progression, therapeutic response, and TME changes, advanced MRI techniques were utilized on the GEMM treated with BTE and saline. Representative two-dimensional (2D) MRI images of the glioma (circled in yellow) and the corresponding three-dimensional rendered glioma image generated following manual segmentation across the whole brain are shown in figure 6A. BTE treatment relative to saline reduced the glioma tumor volume (online supplemental figure 3F). In addition to measuring tumor size, T2 parametric tumor maps were generated from the MRI 2D images to quantify tumor viability and therapy response. The intratumor T2 values of magnetic resonance (MR) 2D images from BTE-treated and saline-treated animals were visualized by superimposing quantitative, color-coded parametric tumor maps, and maps were generated from multi-echo sequences using voxel-by-voxel least-squares fitting procedures (figure 6B, online supplemental figure 4G). Further color-coded (red to blue) parametric mapping of 2D MR images in multi-voxel regions revealed specific differences in radiomic-like features and patterns between BTE and saline-treated animals (figure 6C). Derivation of quantitative tumor response and tumor viability indexes from the combination of these in vivo indexes revealed significantly higher tumor response indexes in BTE-treated animals compared with saline controls (p.adj=0.003)(figure 6D).

Figure 6. MRI demonstrates responses of de novo tumors to BTE in GEMM. (A) A representative image set of two gray-scale MRI anatomical brain sections from saline control (left) and BTE-treated (right) mice was used. The whole-tumor 3D delineation was done to extract tumor size and to enable rendered 3D visualization of the overall tumor size. (B) A representative set of two quantitative T2 parametric colored maps for a saline control and BTE-treated mouse overlayed on a corresponding anatomical gray-scale image. (C) Comparative clustered T2 parametric image of each tumor (control saline on the left and BTE-treated on the right) with different colors depicting intratumor regions with different T2 values (BLUE/CYAN low values and YELLOW/RED high values). An H&E stain of a representative section for each mouse was selected to match the exact location of the MRI-derived image as closely as possible (using anatomical features). The clustered intra-tumor regions were then used to extract volumetric quantities for each T2 range, with the driving hypothesis (from our work and existing literature) that an overall decrease in T2 tumor values can be associated with therapeutic response. (D) Box chart plot of each Tumor Response Index value in each mouse, which agrees with survival analysis, flow cytometry, and RNAseq data.BTE, bispecific T cell-engager; GEMM, genetically engineered mouse model; p.adj, adjusted p value; 3D, three-dimensional.

Discussion

BTEs are recognized as potent targeted T cell-based immune therapeutics that function independently of major histocompatibility complex (MHC) -restricted antigen recognition and exhibit strong antitumor activity.^{21 22} This type of therapeutic strategy is highly relevant for solid cancers such as GBM, which downregulates MHC expression as an immune evasion strategy. Since the US Food and Drug Administration (FDA) approval of blinatumomab in 2014 for B-cell precursor acute lymphoblastic leukemia,²³ eight BTEs have been approved for use against hematopoietic malignancies. Driven by this clinical success, over 100 distinct BTEs have been evaluated in more than 250 clinical trials for a variety of applications in solid tumors.²⁴ However, the complex TME in solid tumors poses unique challenges for BTE therapies not encountered in liquid malignancies, and only two BTEs, tarlatamab (small cell lung cancer)²⁵ and tebentafusp (melanoma),²⁶ have been FDA approved. In GBM, the highly immunosuppressive, myeloid cell-dominated TME thwarts antitumor immune activity.27,29 Our group and others have demonstrated that BTEs targeting IL13RA2 and EGFRvIII elicit a powerful therapeutic effect in PDX mouse models of GBM.^{7 8 18 19} Yet, the immunodeficient mice used in these studies fail to recapitulate the hallmark immunosuppressive TME of GBM. Thus, investigating how BTEs behave in GBM models with intact immune TME is critical for understanding their clinically relevant therapeutic potential.

To generate a BTE suitable for studies in an immunocompetent host, we genetically fused a scFv145-2C11 recognizing murine CD3E with our anti-IL13RA2 scFv via a flexible linker based on our previously validated BTE configuration.¹⁸ In vitro, the resulting BTE activated and directed murine T cell-mediated killing of IL13RA2-positive murine glioma cells. In vivo, BTE accumulated in the brains of immunocompetent mice bearing IL13RA2-positive murine glioma after systemic injection. As such, we hypothesized that BTE accumulation in brain tumors following systemic delivery would enable engagement of host T cells with glioma cells and promote antitumor activity in immunocompetent GBM models. Consistent with our hypothesis, systemic delivery of BTE resulted in target-specific T cell activation confined to the brain tumor, with no activation observed in peripheral tissues. We demonstrated that the BTE exhibited significant antitumor activity, extending survival across four murine GBM models. The BTE penetration into tumor tissue and its therapeutic effects align with our recent study demonstrating uptake and antitumor activity of a humanized BTE in humanized PDX glioma models.³⁰ The therapeutic effect was target-dependent, as BTE-treated mice bearing parental gliomas lacking IL13RA2 showed no response, and CD8⁺ T cell dependent, as no therapeutic effect was observed in CD8 KO mice bearing IL13RA2-expressing gliomas. Unlike treatments with CAR T cells,^{31 32} no loss of IL13RA2 expression in tumor tissue was observed after three cycles of BTE treatments in GEMM. This specific on-cancer activation of T cells observed in this study is of importance. Using BTEs when the target is expressed in healthy tissues, even at low levels, can lead to significant damage to normal organs and neurotoxicity due to inadvertent T cell activation distant from the tumor. In fact, BTE-induced on-target, off-tumor toxicities are well-documented in solid tumors and warrant careful consideration.^{33 34} In that respect, it is well established that IL13RA2 is a glioma-specific antigen35,37 that is expressed in other peripheral cancers38,43 and brain metastases,⁴⁴ but not in normal tissues. The specific expression of IL13RA2 in glioma, but not in other tissues, provides reassurance regarding the safety of a humanized BTE protein to be evaluated in future toxicity studies.

The scarcity of antigens expressed in GBM makes IL13RA2 and EGFRvIII, two antigens expressed exclusively in glioma cells and not in normal brain tissue, unique and suitable for targeted applications. Not surprisingly, these two targets are the most tested in preclinical and clinical settings for GBM.⁴⁵ BTE targeting EGFRvIII has been shown to significantly improve the survival of mice bearing human orthotopic glioma tumors,⁶ and a clinical study evaluating the safety of this BTE is anticipated (NCT04903795). Our group has previously demonstrated that BTE-secreting neural stem cells improve the survival of mice bearing PDX gliomas expressing IL13RA2.¹⁸ Others have also reported the therapeutic effect of different BTE configurations in GBM as a gene therapy approach.⁴⁶ However, the immense inter- and intratumoral antigen heterogeneity in GBM presents major challenges for targeted therapies, as targeting a single antigen often leads to antigen escape.^{31 47} Park and coauthors attempted to address this limitation using a gene therapy approach. They co-delivered two DNA-encoded BTEs targeting EGFRvIII and HER2, demonstrating superior tumor growth suppression and clearance compared with BTE alone.^{7 8} Others adopted a different approach to overcome GBM heterogeneity by creating multispecific therapies, such as EGFRvIII-CAR T cells secreting anti-EGFR BTEs¹² or employing EGFR-IL13RA2-specific CAR T cells.¹¹ These studies showed superior efficacy compared with non-BTE-secreting CAR T cells in immunocompromised mice bearing orthotopic GBM. Two clinical trials (NCT05168423, NCT05660369) evaluating the safety and feasibility of CART-EGFR-IL13RA2 cells¹³ and an EGFRvIII CAR-T cell secreting an EGFR-specific BTE¹⁴ have been initiated.

The studies we conducted in immunocompetent hosts not only determined the ability of BTE to activate T cells in the immunosuppressive TME of GBM but also revealed additional mechanisms contributing to the therapeutic action of BTEs. For example, the analysis of tissues revealed a significant enrichment of tissue-resident memory (TRM) T cells in the brains of tumor-bearing mice treated with BTE across the GBM models evaluated. The increased frequency of these cells likely contributed to the glioma rejection in the LTS mice rechallenged with the parental lines lacking IL13RA2 expression, an indication of induced epitope spreading. Notably, the GEMM, which more closely recapitulates the morphology of human GBM, including a low frequency of T cells, also responded to BTE treatment and showed an increased frequency of TRM CD8 T cells. The scRNA sequencing in this GEMM showed that the CD8 T cells in the brains of BTE-treated animals had enhanced cytotoxic (CD69, GZMB, CD28, and IFNγ) and memory (Sell, CD44, Itgae, CxCr3, LY6C2, CCR5) profiles, and reduced exhaustion profiles (Lag3, Havcr2, Pdcd1, CC244a, and CD96) compared with control mice. The presence of Tcf1-positive T cells was greatly elevated in tumor tissue from BTE-treated mice compared with the control. Tcf1 has been shown to play an important role in T cell development and memory, including responses to cancer.⁴⁸ TRM T cells are known to stay localized in peripheral tissues, continuously monitor the surroundings, and remain primed to rapidly respond to distress signals. These traits position TRM T cells as key contributors to antitumor surveillance, and there is evidence supporting that the effectiveness of immunotherapy relies on the generation of TRM T cells.⁴⁹ In glioma, the presence of TRM T cells was reported in the context of T cells expressing fibrinogen-like 2 (FGL2)-specific single-chain variable fragments (T-αFGL2) cell treatment, peptide and virus treatments.50,52 Virus-specific memory T cells have been found to populate the mouse and human GBM TME and could be reactivated in the experimental setting. Moreover, a high presence of CD8⁺ TRM cells has been shown to correlate with better survival in glioma.⁵³ Evaluation of tumor tissue from patients with both lung⁵⁴ and melanoma cancer⁵⁵ treated with anti-programmed cell death protein 1 (PD-1) and responding to the treatment showed an increase in the frequency of TRM CD8 T cells. In patients who were treated with CAR-T cell therapy, enriched populations of TRM T cells were seen in responders to CD19-specific and B-cell maturation antigen (BCMA)-specific CAR T cells.⁵⁶ Given the established importance of TRM T cells to immunotherapeutic response and sustained anticancer immunity, the observation that TRM cells may have formed memory against glioma cells likely participated in the therapeutic response to rechallenge in the BTE-treated group.

Non-invasive monitoring of GBM, especially in the context of immunotherapy to differentiate between tumor progression and treatment-related neuroinflammation, has become a major focus of study in recent years.57,59 Advanced MRI imaging techniques were used in this study to better understand BTE therapeutic effects on tumor dynamics in the GEMM. Qualitative visualization of non-invasive MRI tumor images exhibited both size and T2 quantitative features and patterns consistent with a therapeutic effect of BTE. In addition to a clear reduction in average tumor size detected through MRI-based volumetrics, the BTE group also exhibited less heterogeneous T2 patterns with overall decreased T2 values relative to saline-treated tumors. It has been shown that heterogeneity and higher T2 values correspond to the more complex microenvironment associated with rapidly progressing tumors.60,62 It is possible that the changes we observed are linked to alterations in TME cell composition, as quantified by our flow cytometry and scRNA-seq analyses. Despite the limitations of the analysis, the multiparametric MRI analysis could provide a non-invasive response assessment in the future.

Studies in immunocompetent hosts provide needed insights into the behavior of T cells. However, there are well-documented differences between mouse and human T cells.⁶³ The extracellular domains of CD3E exhibit a low level amino acid sequence conservation (47%) between humans and mice.⁶⁴ The BTE we designed and tested in this study binds murine, but not human, T cells.⁶⁵ It has been shown that BTEs also exhibit a relatively short plasma half-life.⁶⁶ In that respect, several modifications of BTE, such as fusion with engineered Fc or human serum albumin, offer the opportunity to significantly expand the time of BTE in systemic circulation.67,70 In the context of brain tumors, the balance between the circulation time and the size of the protein should be carefully investigated to ensure sufficient penetration in GBM tissue. Thus, the design of reagents suitable for both human and mouse studies, along with modifications that enhance penetration into brain tumors and improve plasma stability, will be essential to overcome the limitations of preclinical testing for the clinical translation of BTEs.

In conclusion, our study demonstrates survival benefits of BTEs in several immune-competent models of GBM, along with target-specific immune engagement, induction of immunological memory, and modulation of the TME. This provides strong evidence of BTE-induced responses and engagement of the host immune system that reinforce the mechanism of action. These findings also encourage further investigation into other brain and peripheral cancers expressing IL13RA2. While combating GBM remains our primary focus, translating this immunotherapeutic strategy may offer therapeutic benefits in other solid tumors, warranting further exploration of BTE’s potential.

Methods

Cell culture

GL261-IL13RA2, CT2A-IL13RA2, and SMA-560-IL13Rv2 were modified with human IL13RA2 as previously described.¹⁹ Glioma cell lines were cultured in Dulbecco’s Modified Eagle Medium (Corning Life Sciences; Corning, New York, USA; Cat No. 10-013-CV) supplemented with 10% v/v fetal bovine serum (FBS) (R&D Systems; Minneapolis, Minnesota, USA; Cat. No. S11550) and 1% v/v Penicillin-Streptomycin (Corning Life Sciences; Cat No. 10-013-CV). Cell lines were screened monthly for mycoplasma via the MycoAlertTM Mycoplasma Detection Kit (Lonza, Walkersville, Maryland, USA; Cat No. LT07-318).

Generation of BTEs and binding analysis

The BTE molecule consists of two scFvs of antibodies, scFv145-2C11 (against murine CD3 epsilon subunit of T-cell receptor complex; CD3E)¹⁵ and scFv47 (against human IL13RA2),^{16 17} connected through a 23aa Gly₄S flexible linker, synthesized by GenScript (Piscataway, New Jersey, USA) and subcloned into a pLVX-IRES-zsGreen 1 (Takara Bio, Mountain View, California, USA; Cat. No. 632187) lentiviral expression cassette. The CDR3 region of the scFvIL13RA2 light chain was replaced with the CDR3 domain of a non-specific antibody MOPC-21 (designated NC-BTE) to generate a negative control protein unable to interact with IL13RA2. For the production of soluble BTE proteins, HEK293T cells were transduced with lentiviral particles. Proteins were isolated from the supernatants using His-Tag affinity chromatography as previously described¹⁸ and characterized for their size and binding specificity against IL13RA2 by western blot and ELISA.

Western blot and ELISA

Western and ELISA were carried out as previously described.¹⁸ A detailed description of the method can be found in the online supplemental materials.

Chromium-51 release assay

Chromium-51 (Cr⁵¹) release assays were used to assess BTE-induced T cell-mediated cytotoxicity against glioma cells. Briefly, splenic murine T cells were isolated from C57BL/6 mice using the EasySep Mouse T Cell Isolation Kit (STEMCELL Technologies; Cat. No. 19851). Glioma cells were labeled with Cr⁵¹ (0.05 mCi) (Revvity, Waltham, Massachusetts, USA; Cat No. NEZ030001MC). Murine T cells were then co-cultured with labeled glioma cells at a 20:1 effector-to-target (E:T) ratio and treated with various concentrations of BTE protein. The positive control consisted of Dynabead CD3/CD28-activated T cells (Gibco, Waltham, Massachusetts, USA; Cat11452D). After 18 hours, supernatants were carefully removed, placed into 96 well LumaPlate-96 (PerkinElmer; Wellesley; Massachusetts, USA; Cat No. 6006633), dried via evaporation, and Cr⁵¹ activity was measured in a gamma counter (PerkinElmer). Specific lysis was computed relative to spontaneous and maximum activity (measured via target cell incubation with 1% Triton X-100). Data was displayed as the SE of mean of triplicates. Statistical significance was denoted based on the p value from the student t-test.

Uptake of fluorescent-labeled BTE in GL261-IL13RA2 tumors after systemic delivery

To visualize the uptake of BTE in the brain of mice bearing intracranial GL261-IL13RA2 tumors were injected intravenously with 20 µg of AlexaFluor647-labeled BTE or with 50 µg of Cy5-labeled BTE for live animal imaging or for confocal microscopy, respectively. Blood vessels were visualized with fluorescein-labeled tomato lectin (Vector Labs). Cell nuclei were visualized with Hoechst (Thermo Fisher) as previously described.⁷¹ For confocal microscopy, brains were harvested at 3 hours after BTE injection and frozen immediately on dry ice. For high magnification imaging, snap-frozen brains were sliced into~1 mm-thick slices. The slices were placed on a glass slide and imaged using a Nikon Eclipse AR1HD inverted confocal microscope with 405 nm (Hoechst), 488 nm or 647 nm (fluorescein isothiocyanate (FITC)-lectin), and 640 nm (Cy5) excitation lasers and corresponding emission filters. Multiple random areas were acquired per section at 512×512 resolution. Signal intensity of BTE in proximity to tumor vasculature was measured as migration of BTE (Cy5) fluorescence as a function of distance from blood vessels using the Line Profile function in Fiji (ImageJ V.1.54p).

Orthotopic xenograft and genetically engineered mouse model

Orthotopic models used CD45.1, B6.129S2-Cd8a^tm1Mak/J (eg, CD8 knockout (KO)), and C57Bl/6 mice sourced from Jackson Laboratories. The VMDK mice were provided by Dr J Sampson (Duke University). The de novo GEMM glioma model employed nestin-Tva PTEN^fl/fl p53^fl/fl mice generously gifted by Dr O Becher (Mount Sinai, New York, USA). Gliomagenesis in the GEMM was induced using a continuous cell line of chicken embryo fibroblasts (DF1 cells) producing RCAS-cre and by inserting PDGFB and IL13RA2 using the RCAS-PDGFB-IL13RA2 virus as previously described.²⁰ In all models, gliomas were initiated via stereotactic injection of either murine glioma cells (orthotopic model) or DF-1 viral-producing cells (GEMM) into the cerebral cortex, approximately 1 mm posterior to the bregma and 1 mm to the right of the midline, as previously described.¹⁹ Mice were randomized by gender into treatment groups: BTE, NC-BTE, or saline. In syngeneic CT2A and GL261 models, treatments were administered intraperitoneally (i.p.) on days 7, 9, and 11 post-glioma implantations. In a separate experiment, intravascular and s.c. routes of BTE delivery were tested in the GL261 model. In the de novo GEM model, treatments were given i.p. 4 days per week starting day seven post-implantation for 3 weeks. Mice were monitored for survival with endpoints defined as either a 20% reduction in body weight or the onset of neurological symptoms.

Flow cytometry

Cells were blocked with Dulbecco’s Phosphate-Buffered Saline (DPBS) (Corning Life Sciences; Cat. No. 20-031-CV) supplemented with 2% FBS (R&D Systems) and TruStain FcX anti-mouse CD16/32 (1:200; BioLegend, San Diego, California, USA; Cat. No. 101319). Surface antigens were stained using fluorochrome-conjugated antibodies in 2% FBS-DPBS for 20 min on ice. Dead cells were labeled with eBioscience Fixable Viability Dye eFluor 780 (1:1000 in DPBS; Invitrogen, Waltham, Massachusetts, USA; Cat. No. 65-0865-18). Cells were fixed and permeabilized with the eBioscience forkhead box P3 (FOXP3)/Transcription Factor Staining Buffer Set (Invitrogen; Cat. No. 00-5523-00), followed by intracellular staining with fluorochrome-conjugated antibodies. Data acquisition was performed using the BD FACSymphony A5-Laser Analyzer (Becton Dickinson, Franklin Lakes, New Jersey, USA) at the Robert H. Lurie Comprehensive Cancer Center Flow Cytometry Core Facility. Flow cytometric analyses were conducted using FlowJo V.10.9.0 software (Becton Dickinson).

For in vitro T cell activation experiments, splenic murine T cells were freshly isolated from C57BL/6 mice using the EasySep Mouse T Cell Isolation Kit (STEMCELL Technologies, Vancouver, Canada; Cat. No. 19851). Murine T cells were co-cultured with glioma cells at a 20:1 E:T ratio in triplicate wells of a 96-well plate treated with various concentrations of BTE, NC-BTE, or Dynabeads Mouse T-Activator CD3/CD28 (Gibco, Carlsbad, California, USA; Cat. No. 11 456D) for 24 hours. Following incubation, cells were collected for flow cytometric staining. For the TME analysis, orthotopic glioma cells were implanted and randomized into the following treatment groups: BTE, NC-BTE, or saline. Treatments were administered i.p. starting 7 days post-glioma implantation/initiation. Brains and spleens were harvested 48 hours after the final treatment and mechanically dissociated into single-cell suspensions using a 70 µm cell strainer. The cells were counted using trypan blue exclusion and processed for flow cytometric staining.

Histology

The protocol for detecting immune cells in the murine brain of the GEMM has been previously described.²⁰ The following antibodies were used: IL13RA2 (R&D Systems, Cat No. AF146), Olig2 (Abcam, ab109186), Ki67 (Abcam, ab16667), CD3 (Abcam, ab16669), and CD11b (Abcam, ab133357). For the detection of CD8⁺ T cells in the brains of long-term surviving mice, 4% paraformaldehyde-fixed paraffin-embedded 4 µm thick-tissue section from the mouse brains were incubated with the CD8a monoclonal antibody (4SM16) (Invitrogen, 14-0195-82), followed by 3,3′-diaminobenzidine chromogen staining. The quantification of the stained areas was performed using the ImageJ software. The tissue was stained with H&E to highlight the morphological changes, and the image analysis was performed using the NDP.view V.2.8.24 software (Hamamatsu Photonics K.K.).

Magnetic resonance imaging

MRI of glioma-bearing mice was conducted using a Bruker Clinscan 7T system equipped with a dedicated multichannel brain coil for high sensitivity and spatial resolution of whole-brain images. The detailed description of the procedure, imaging sequences, parameters and analysis is described in online supplemental materials.

Single-cell RNA sequencing

scRNA-seq was performed on both control and BTE-treated gliomas from the Nestin-Tva-PTEN^fl/fl-P53^fl/fl mice from two independent experiments with identical study design. These mice were randomized evenly by gender into two groups, with five mice per group. 7 days post-glioma induction, mice were treated with either saline or BTE (10 mg/kg) via i.p. route four times per week for 3 weeks. 48 hours after the last treatment, mice were anesthetized and perfused intracardially with 10 mL of saline. The glioma tissue was excised and placed into single-cell suspension using the Adult Brain Dissociation Kit, Mouse and Rat (Miltenyi Biotec, Gaithersburg, Maryland, USA; Cat No. 130-107-677). Single-cell suspensions were then incubated with anti-CD16/32 (Fc Block) antibodies (BioLegend) and then subjected to anti-mouse CD45 bead-based separation (Miltenyi Biotec; Cat. No 130-052-301). The scRNA-seq was performed on the glioma-infiltrating immune cells enriched at a CD45+to CD45- cell ratio of 8:2 at a 4,000 live cells/µL concentration.

The PIPseq T2 3’ Single Cell RNA Kit v4.0 kit (Fluent Biosciences, Cat No. FBS-SCR-T20-4-V4.05) was used for scRNA-seq prep. PIPseq uses a microfluidics-free, templated emulsification technology to generate monodispersed cell droplets. 4×10⁴ cells per sample were added to particle-templated instant partitions, emulsified via rapid mechanical vortexing, lysed to capture messenger RNA on barcoded poly(T) decorated beads, transcribed into cDNA, amplified with PCR, and sample indexed with unique P7 and P5 Illumina-compatible indexes. The barcoded cDNA libraries were then subjected to pooled sequencing on an Illumina NovaSeq X Plus platform at a depth of 650 million reads per sample at Novogene (Sacramento, California, USA). Demultiplexed FASTq files were aligned to the Human Hg38 genome assembly (GRCh38.p13, 2022) using the PIPseeker (Fluent BioSciences) software. The resulting data matrix files were analyzed for each sample in R (V.4.2.3) using the Seurat (v5) package. Cells were filtered with>2.5% mitochondrial read counts with>500 expressed genes. The scDblFinder (V.1.18.0) was used to identify and remove doublets. In Seurat, the data was log-normalized using a scale factor of 10,000. The default parameters were used to identify highly variable genes. Principal component analysis was conducted on the top 2,000 variable genes. The top 17 principal components were selected for neighboring and clustering identification. To correct the batch effect between the two experiments, canonical correlation analysis (CCA) was performed. Subsequently, the JoinLayers function was employed to merge sample layers. The clusters were visualized in UMAP of the first 17 CCA reductions. FindAllMarkers was used to find the top differentially expressed genes between clusters and identify cell types based on canonical markers. Seurat workflow was used to subcluster immune cells, and the FindMarkers function was used to identify DEGs between treatment groups. The ClusterProfiler (V.4.12.0) was used to perform gene ontology enrichment analysis on significant DEGs. The R package ggplot2 (V.3.5.1) and ComplexHeatmap (V.2.20.0) was used to construct the plots and heat maps.

Automated multiplexed sequential immunofluorescence imaging

Automated hyperplex immunofluorescence staining and imaging were performed on formalin fixed paraffin embedded (FFPE) sections using the COMET system (Lunaphore Technologies). The following targets and the associated antibodies were used for this study: glial fibrillary acidic protein (GFAP) (Sigma, Clone GA5, MAB360, 1/3000 dilution); CD31 (Abcam, Clone EPR17259, AB225883, 1/1500 dilution); CD4 (Abcam, Clone EPR19514, Ab183685, 1/500 dilution); CD8 (Cell Signaling, Clone D4W2Z XP, 98941, dilution 1/500); IFN-γ (Bioss, BS-0480R, polyclonal, dilution 1/100); TCF1 (Cell Signaling, Clone C63D9, cat#2203, dilution 1/500); IL13RA2 (Cell Signaling, E7U7B, 85677, dilution 1/200). The detailed description of the protocol can be found in the online supplemental materials.

Statistical analyses

All statistical analysis was completed in R (V.4.2.3). Statistical analysis was performed by one-way analysis of variance, and p values were adjusted for multiple comparisons using Tukey’s test. The Kaplan-Meier method was used to compare survival between treatment groups and analyzed with the log-rank (Mantel-Cox) test. P values were adjusted for multiple comparisons using Bonferroni’s correction.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.