Machine-learning Directed Conversion of Glioblastoma to Dendritic Cell-like Antigen Presenting Cells as Cancer Immunotherapy

Here, we report the development and application of a machine-learning precision method to identify cell fate determinants (CFD) that specifically reprogram GBM into induced antigen-presenting cells with DC-like functions (iDC-APC). In murine GBM models, iDC-APCs acquired DC-like morphology, regulatory gene expression profile and functions, including phagocytosis, direct presentation of endogenous antigens, and, especially, vigorous cross presentation of exogenous antigens comparable to natural DCs to prime naïve CD8+ CTLs, a hallmark DC function critical for anti-tumor immunity. ORF expression clones were purchased from Genecopoeia. To create mF3, cDNAs encoding mouse IRF8, BATF3, and ID2 were synthesized, linked by P2A and T2A cleavage sequences, and cloned into the lentiviral vector pSF-simple (pSF-simple-mIRF8-P2A-mBATF3-T2A-mID2). For PU.1 constructs, cDNAs encoding mouse PU.1 (Spi1) and PU.1 delta the PEST domain (dP) (16) were cloned into the lentiviral response plasmid pSF-Lenti that contains PGK-puro to generate pSF-Lenti-PU.1 and pSF-lenti-detaPU.1. Lentiviral constructs were produced and used to reprogram KR158 cells. On day 5 after transduction, cells were collected and single cell RNAseq libraries were prepared using 10X single cell lib prep kit. Each sample contained populations from different groups (to reduce the sequencing cost). Each individual group was annotated in downstream analysis by characteristic gene markers.