Head-to-head preclinical treatment design prioritizes promising therapies for Neurofibromatosis 1 optic glioma clinical translation

Authenticated preclinical brain tumor models provide unprecedented opportunities to evaluate next-generation treatments. However, many therapies that exhibit robust anti-tumor activity in mice fail to show efficacy in clinical trials. To prioritize therapeutic candidates for future clinical translation, we implemented a head-to-head preclinical strategy using a well-characterized murine model of NF1-optic pathway glioma (Nf1OPG). Nf1OPG mice were treated with standard of care (carboplatin), clinically evaluated (everolimus, mirdametinib), and investigational (pexidartinib, HBS-101, lamotrigine) drugs during the period of most rapid tumor growth (6-12 weeks of age). Ant-tumoral efficacy was assessed by proliferation (%Ki67+ cells) and optic nerve volume, while vision-related outcomes were measured using retinal nerve fiber layer (RNFL) thickness and retinal ganglion cell (RGC) determinations. Tumor microenvironment (TME) soluble mediator (Ccl2, Ccl3, Ccl4, Ccl5) and tumor cell marker (NeuN, Gpr17) RNA expression was quantitated by qRT-PCR. Outcomes were compared to carboplatin-treated Nf1OPG, untreated Nf1OPG, and Nf1+/- mice. While all agents restored normal tissue architecture, reduced optic nerve proliferation, and decreased TME soluble mediator and tumor cell marker RNA expression, only lamotrigine also reduced optic nerve volume. Everolimus, lamotrigine, and HBS-101 all restored RNFL thickness to wild-type levels, whereas carboplatin showed a trend towards normalization. This referential study design affords direct head-to-head comparisons of investigational therapies relative to standard-of-care-treatment using clinically meaningful outcomes (optic glioma growth and RNFL thickness). Using this strategy, lamotrigine emerged as the most promising therapy for limiting tumor progression and vision loss in Nf1-OPG mice relevant to clinical translation in children with NF1. Overall design: Raw sequencing data from Nf1OPG mouse samples were processed using the 10X Genomics Cell Ranger pipeline (version 7.2.0) to generate gene count matrices, which were aligned to the mm10-2020-A mouse reference genome. The aligned data were merged with previously published datasets (GSM7816061 and GSM7816062 from GSE244433), resulting in a total of four samples for analysis. Raw counts were used as input for the Seurat R package (version 5.2.1). For quality control, cells were excluded if they had fewer than 500 or more than 5,000 detected genes, more than 5% mitochondrial gene content, or fewer than 500 or more than 20,000 transcript counts. Normalization and variance stabilization was performed using the SCTransform function and 2,000 highly variable features of each dataset were identified and scaled using the FindVariableFeatures and ScaleData functions. Principal component analysis (PCA) was conducted using RunPCA. The data were integrated using the IntegrateLayers function with CCAIntegration. The integrated data was subjected to dimensional reduction and clustering using RunUMAP, FindNeighbors and FindClusters. Each cell type was identified using cell type-specific markers. *************************************************************** The table below lists GEO accessions reused/reanalyzed for this study. ***************************************************************