Neural stem cells (NSCs) are controlled by diffusible factors. The transcription factor Sox2 is expressed by NSCs and Sox2 mutations in humans cause defects in the brain and, in particular, in the hippocampus. We deleted Sox2 in the mouse embryonic brain. At birth, the mice showed minor brain defects; shortly afterwards, however, NSCs and neurogenesis were completely lost in the hippocampus, leading to dentate gyrus hypoplasia. Deletion of Sox2 in adult mice also caused hippocampal neurogenesis loss. The hippocampal developmental defect resembles that caused by late sonic hedgehog (Shh) loss. In mutant mice, Shh and Wnt3a were absent from the hippocampal primordium. A SHH pharmacological agonist partially rescued the hippocampal defect. Chromatin immunoprecipitation identified Shh as a Sox2 target. Sox2-deleted NSCs did not express Shh in vitro and were rapidly lost. Their replication was partially rescued by the addition of SHH and was almost fully rescued by conditioned medium from normal cells. Thus, NSCs control their status, at least partly, through Sox2-dependent autocrine mechanisms.

The transcription factor Sox2 is active in neural stem cells, and Sox2 'knockdown' mice show defects in neural stem/progenitor cells in the hippocampus and eye, and possibly some neurons. In humans, heterozygous Sox2 deficiency is associated with eye abnormalities, hippocampal malformation and epilepsy. To better understand the role of Sox2, we performed in vitro differentiation studies on neural stem cells cultured from embryonic and adult brains of 'knockdown' mutants. Sox2 expression is high in undifferentiated cells, and declines with differentiation, but remains visible in at least some of the mature neurons. In mutant cells, neuronal, but not astroglial, differentiation was profoundly affected. ␤-Tubulin-positive cells were abundant, but most failed to progress to more mature neurons, and showed morphological abnormalities. Overexpression of Sox2 in neural cells at early, but not late, stages of differentiation, rescued the neuronal maturation defect. In addition, it suppressed GFAP expression in glial cells. Our results show an in vitro requirement for Sox2 in early differentiating neuronal lineage cells, for maturation and for suppression of alternative lineage markers. Finally, we examined newly generated neurons from Sox2 'knockdown' newborn and adult mice. GABAergic neurons were greatly diminished in number in newborn mouse cortex and in the adult olfactory bulb, and some showed abnormal morphology and migration properties. GABA deficiency represents a plausible explanation for the epilepsy observed in some of the knockdown mice, as well as in SOX2-deficient individuals.

In mouse and human neural progenitor and glioblastoma "stem-like" cells, we identified key targets of the Polycomb-group protein BMI1 by combining ChIP-seq with in vivo RNAi screening. We discovered that Bmi1 is important in the cellular response to the transforming growth factor-β/bone morphogenetic protein (TGF-β/BMP) and endoplasmic reticulum (ER) stress pathways, in part converging on the Atf3 transcriptional repressor. We show that Atf3 is a tumor-suppressor gene inactivated in human glioblastoma multiforme together with Cbx7 and a few other candidates. Acting downstream of the ER stress and BMP pathways, ATF3 binds to cell-type-specific accessible chromatin preloaded with AP1 and participates in the inhibition of critical oncogenic networks. Our data support the feasibility of combining ChIP-seq and RNAi screens in solid tumors and highlight multiple p16(INK4a)/p19(ARF)-independent functions for Bmi1 in development and cancer.

Polycomb group (PcG) proteins form transcriptional repressor complexes with well-established functions during cell-fate determination. Yet, the mechanisms underlying their regulation remain poorly understood. Here, we extend the role of Polycomb complexes in the temporal control of neural progenitor cell (NPC) commitment by demonstrating that the PcG protein Ezh2 is necessary to prevent the premature onset of gliogenesis. In addition, we identify the chromodomain helicase DNA-binding protein 4 (Chd4) as a critical interaction partner of Ezh2 required specifically for PcG-mediated suppression of the key astrogenic marker gene GFAP. Accordingly, in vivo depletion of Chd4 in the developing neocortex promotes astrogenesis. Collectively, these results demonstrate that PcG proteins operate in a highly dynamic, developmental stage-dependent fashion during neural differentiation and suggest that target gene-specific mechanisms regulate Polycomb function during sequential cell-fate decisions