Despite the myriad of cellular and animal models available for studying the pathogenesis of NF1, NF2 and schwannomatosis and for developing new treatments, none is perfect. Rather than reviewing the most commonly used tools, I will focus on genetic temporality, the tumor microenvironment and their crosstalk with tumor cells, as well as on the triggering and modifying factors that initiate and shape these complex diseases, factors that remain largely unaddressed in current models. I will then discuss how these dimensions can be integrated into therapeutic development, from the selection of readout parameters to the design of combination strategies. Finally, I will introduce the concept of the virtual tumor, a digital twin integrating multi-omic and imaging data, as a forward-looking framework for studying these diseases and for testing therapeutic hypotheses in silico before they reach the bench and the clinic.

Background: We identified three molecular subgroups of malignant peripheral nerve sheath tumors (MPNSTs) defined by specific combinations of tumor suppressor gene (TSG) inactivation (G1, G2 and G3) (1). To better understand the molecular mechanisms driving their initiation, we generated different induced pluripotent stem cell (iPSC)-derived neural crest (NC)-based models (2) carrying representative subgroup-specific TSG inactivation signatures.

Methods: Using CRISPR/Cas9 gene editing, we generated and validated three isogenic NC models harboring distinct TSG inactivation signatures: 4KO-G1 (NF1-/-, CDKN2A-/-, SUZ12-/-, TP53-/-), 3KO-G2 (NF1-/-, TP53-/-, EED-/-) and 3KO-G3 (NF1-/-, CDKN2A-/-, TP53-/-). These models were functionally characterized for NC and mesenchymal stem cell (MSC) marker expression, genomic instability, and proliferative capacity. Genomic, transcriptomic, and epigenomic analyses are ongoing. Tumorigenic potential is being evaluated using luciferase-labeled NC spheres engrafted into nude mice.

Results: We successfully established three NC cell-based MPNST models that recapitulate the TSG inactivation signatures of the corresponding NF1 MPNSTs groups identified. The 4KO-G1 model retained both NC and MSC features, whereas 3KO-G3 cells maintained a NC phenotype. Notably, both 4KO-G1 and 3KO-G3 models, in which TP53 gene was inactivated, displayed significantly increased proliferation compared to their parental counterparts. Luciferase-expressing 4KO-G1 cells were engrafted into nude mice, and tumor growth monitored for four months. Formed tumors are currently being characterized.

Conclusions: We developed NC-based models representing the major molecular subgroups of MPNST. These models provide a valuable preclinical platform to investigate early events in MPNST pathogenesis, define the tumorigenic impact of specific TSG inactivation combinations, and evaluate novel therapeutic possibilities.

1) doi.org/10.1101/2025.08.13.670072
2) doi.org/10.1038/s41467-026-73119-8

Background: Spinal ependymomas are the most frequent tumor of the spinal cord. NF2-related Schwannomatosis (NF2-SWN) is associated with spinal ependymomas. They are characterized by cystic development and peritumoral edema with morbidity. To explore the developmental and cellular mechanisms underlying these lesions, we developed a conditional mouse model with Nf2 inactivation in FoxJ1-expressing ependymal cells.
Methods: We generated FoxJ1Cre/Nf2flox/flox mice using a tamoxifen-inducible system and induced recombination Nf2 inactivation at different developmental stages (E16, P0–4, and P15). Mice were monitored up to 15 months with post-mortem MRI and detailed histopathological analysis.
Results: Conditional Nf2 inactivation in FoxJ1 expressing cells did not lead to ependymoma formation but resulted in frequent ependymal proliferations (87%) and cystic dilations of the central canal (54%), mimicking human pathological features. Proliferative lesions were composed of FoxJ1⁺ ependymal cells and FoxO3⁺/GFAP⁺/S100β⁻ populations resembling immature radial glial-like progenitors. These proliferations were often accompanied by hydrocephalus and periventricular expansion, particularly after late postnatal induction. Early induction (E16 and P0–P4) caused higher mortality and malformations, while the P15 cohort showed a reproducible phenotype with minimal lethality. No frank tumor formation was detected in any cohort. Post-mortem MRI showed cystic formations.
Conclusion: These findings indicate that Nf2 loss in differentiated ependymal cells induces disorganized ependymal proliferation and canal cystic dilations, mimicking early tumorigenic changes without ependymoma formation. This mouse model provides a valuable preclinical platform for studying the earliest stages of NF2-SWN spinal ependymoma pathogenesis and testing targeted anti-angiogenic or anti-proliferative therapies.

Background:
Neurofibromatosis type 1 and type 2 (NF1/NF2) are tumor predisposition syndromes in which tumor initiation and progression are strongly influenced by the peripheral nervous system (PNS) microenvironment. However, the mechanisms by which the human perineural niche regulates Schwann cell tumorigenesis remain poorly understood due to species-specific differences and the lack of tractable human multicellular models. We hypothesize that NF1/NF2-mutant Schwann cells engage in bidirectional interactions with neuronal, stromal, and immune populations that promote pro-tumorigenic remodeling. To address this, we are developing a human pluripotent stem cell (hPSC)-derived platform to model nerve-immune-tumor interactions in a human PNS context.

Methods:
We generated an isogenic series of CRISPR-edited hPSC lines with NF1 or NF2 disruption, including clones with partial or complete protein loss. hPSCs were differentiated into motor and sensory neurons, Schwann cell precursors, neural crest-derived fibroblasts, macrophages, and mast cells. Multicellular assembloids were generated by combining Schwann cell precursor–fibroblast spheroids with neuronal spheroids, followed by immune cell incorporation. NF1−/− hPSCs were further engineered with TP53 knockout to model malignant progression. Spatial transcriptomics of NF patient tumors will provide in vivo validation.

Results:
We established validated NF1- and NF2-edited hPSC lines capable of differentiating into major PNS microenvironmental cell types. First-generation multicellular PNS assembloids were successfully assembled, maintained, and are currently undergoing characterization.

Conclusion:
We generated a genetically controlled 3D platform that recapitulates the NF human perineural niche with defined cellular complexity. Together, this approach establishes a translationally relevant model to define how the human PNS microenvironment contributes to tumor initiation, adaptation and progression in NF.

Background: Legius syndrome is an autosomal dominant disorder caused by variants in SPRED1 and is characterized by café-au-lait macules, skin-fold freckling, macrocephaly, and neurodevelopmental deficits. A large percentage of affected individuals exhibit learning difficulties, Attention Deficit Hyperactivity Disorder (ADHD), developmental delay, and autism spectrum disorder (ASD), phenotypes that are also observed in Spred1 knockout mouse models. However, the underlying biological mechanisms driving these neurodevelopmental difficulties remain poorly understood.

Methods: Induced pluripotent stem cell (iPSC) lines were generated by reprogramming peripheral blood mononuclear cells from four unrelated patients with Legius syndrome, including two individuals with neurodevelopmental difficulties. CRISPR/Cas9 or prime editing were used to correct the pathogenic SPRED1 variants to establish isogenic control lines. All cell lines were extensively characterized for pluripotency and genomic integrity. iPSCs were differentiated into cortical neurons using a lentiviral NGN2-based differentiation protocol. Immunofluorescence, RNA sequencing, and multi-electrode arrays will be used to assess structural, transcriptional, and functional differences between patient-derived and isogenic control lines.

Results: We generated four Legius syndrome iPSC lines and gene-corrected isogenic control lines. All lines exhibited normal morphology and karyotypes, expressed pluripotency markers, and possessed trilineage differentiation potential. The iPSCs were successfully differentiated into cortical neurons, providing a platform to study key aspects of neuronal development, morphology, and function in the patient and isogenic iPSC lines.

Conclusion: This is the first established human iPSC model developed for Legius syndrome. It is a valuable resource for investigating the molecular and neurobiological aspects of the disorder in a human neuronal context.