STING-induced blood-brain barrier opening in combination with radiation therapy potentiates anti-glioma immune and therapeutic activity monitored with [18F]-FLT PET

Radiation therapy (RT) is the standard of care for glioblastoma but is not curative. Triggering the cGAS/STING pathway with potent agonists such as 8803 exerts marked activity across high-grade glioma preclinical models through the induction of anti-tumor immune responses. To determine if the combination of 8803 with radiation therapy (RT) warrants consideration in the upfront treatment setting and to clarify the underlying mechanisms of therapeutic activity, C57BL/6 mice harboring intracerebral CT-2A or QPP8v high-grade gliomas were treated with fractionated RT (2 Gy daily for 5 days), intratumoral 8803, or both regimens. The treatment with 8803 with RT resulted in 80% long-term and undefined median survival (MS) relative to monotherapy (control MS: 25; 8803 MS: 36; RT MS: 38 days) in the high-grade CT-2A glioma model (combination vs. control p<0.0001), but not in the radiation-resistant QPP8v model. CRISPR gene editing was used to knock out (KO) the STING pathway in CT-2A glioma tumor cells to model the epigenetic silencing of STING expression in human glioblastoma. In this setting, the combinatorial therapeutic effect has been maintained, highlighting that the direct contribution of immune system in mediating survival benefit. The 8803-mediated STING activation was visualized longitudinally using [18F]-FLT uptake. In CT-2A-bearing mice, maximum peak of the [18F]-FLT uptake was observed at 72-96 hours after 8803 administration. Sc-RNA sequencing of the CT-2A glioma during the therapeutic window identified a nitric oxide synthase gene signature in the Inflam-TAM(2) population associated with combinatorial efficacy. However, the therapeutic effect of 8803, RT, or the combination of RT + 8803 was not ablated in the NOS2 background mice, indicating that this pathway is not a major contributor to therapeutic activity. scRNA sequencing revealed that 8803 treatment reprograms the blood-brain barrier (BBB) through changes in the PECAM and CD147 pathways in the endothelial cell. A single intracranial injection of 8803, but not RT monotherapy, in non-tumor bearing C57BL/6J mice, induced diffuse bi-hemispheric blood-brain barrier (BBB) opening, shown by fluorescent imaging up to 24 hours, but not in STING KO background mice. In summary, a 8803, combined with RT, triggers distinctive anti-glioma immune reactivity, facilitates brain immune infiltration through the opening of the BBB, and warrants consideration for clinical trials in the up-front setting for glioblastoma that can be monitored with [18F]-FLT PET imaging. Overall design: At designated time points, tumor-bearing mice were euthanized and brains collected following cardiac perfusion and chopped up using a scalpel, dissociated, and suspended using a mixture of IMDM 1X (Corning) containing 2% inactivated FBS (Millipore Sigma), collagenase (100 µg/mL) and DNase (20 units/mL). Samples were incubated for 35–40 minutes at 37°C with agitation. The tissue was filtered through a 70 µm nylon cell strainer (BD Biosciences) and then underwent centrifugation at 4°C. Half of the pellet representing the whole TME was resuspended in culture media, and the other half in a 20 mL mixture of 5.4 mL Percoll Plus (GE Healthcare) overlaid with 12 mL 1X PBS and 0.6 mL 10X PBS (Corning) for immune enrichment. The tube was centrifuged at 800g for 10 minutes at 4°C, with 9 acceleration and 0 deceleration. After centrifugation, the immune-enriched cell pellet was collected, washed, mixed with the whole TME samples at a ratio of 1:1, stained with Trypan blue dye (MilliporeSigma), and counted using a Countess II FL automated cell counter in a Countess cell-counting chamber (Invitrogen, Thermo Fisher Scientific). Post-library preparation cells were sequenced using the Illumina NovaSeq. Preprocessing of raw sequencing files included alignment to mm10 reference using Cell Ranger (v3.1.0). Next, ambient mRNA and doublets were removed using SoupX (v.1.3.0) and sDblFinder (v1.4.0), respectively. Then, a standard Seurat R Package workflow was applied (34). Cells were cleaned using a mitochondrial DNA threshold of 10% and a UMI range of 100 to 5000. In addition to Seurat functions, the Harmony algorithm was used for batch correction (35). Non-tumor cell clusters were annotated using three parallel methods: 1) comparison against known cell markers; 2) examination of DEGs against the Human Protein Atlas; and 3) singleR package, an automated cell assignment algorithm. Differential abundance testing was conducted using the MiloR package (36).