iPSC-derived NF1-CDKN2A-PRC2 deficient neural crest cells mimic glial-to-neuro-mesenchymal transition and form MPNST-like tumors in vivo

Neurofibromatosis Type 1 (NF1) predisposes to peripheral nerve tumor development. Commonly, the progression from a benign plexiform neurofibroma (PNF) towards a deadly malignant peripheral nerve sheath tumor (MPNST) involves a poorly understood glial-to-mesenchymal transition and the sequential loss of NF1, CDKN2A, and polycomb repressive complex 2 (PRC2). Using an iPSC-derived neural crest (NC) model, we reproduced this malignant transformation through gene editing. NF1-CDKN2A double-knockout (2KO) NCs retained glial differentiation capacity and formed neurofibroma-like tumors in vivo, requiring inactivation of p14ARF and p16INK4a. Additional PRC2 loss (3KO) disrupted pluripotency and induced mesenchymal stem cell-like features in iPSCs an NCs. 3KO NCs suffered a global chromatin reprograming that silenced SOX10 preventing gliogenesis and activated neuro-mesenchymal programs. Gene signatures characterizing this glial-to-neuro-mesenchymal transition were recapitulated in human PNF-ANNUBP-MPNST tumors. 3KO NC spheres formed MPNST-like tumors in vivo upon nerve engraftment, genuinely mimicking an early-stage MPNST. We used the developed 3D NC spheroid models for the discovery of drugs targeting MPNSTs by performing a high-throughput screening of an epigenetic compound library. We found that poly(ADP-ribose) polymerase inhibitors (PARPi) exhibit selective efficacy in PRC2-deficient NC spheroids. We confirmed that Olaparib-Selumetinib combination in a MPNST PDX mouse model was well tolerated and significantly suppressed tumor growth.