Capicua regulates neural stem cell proliferation and lineage specification through control of Ets factors

Capicua (Cic) is a transcriptional repressor mutated in the brain cancer oligodendroglioma. Despite its cancer link, little is known of Cic’s function in the brain. Here, we investigated the relationship between Cic expression and cell type specification in the brain. Cic is strongly expressed in astrocytic and neuronal lineage cells but is more weakly expressed in stem cells and oligodendroglial lineage cells. Using a new conditional Cic knockout mouse, we show that forebrain-specific Cic deletion increases proliferation and self-renewal of neural stem cells. Furthermore, Cic loss biases neural stem cells toward glial lineage selection, expanding the pool of oligodendrocyte precursor cells (OPCs). These proliferation and lineage selection effects in the developing brain are dependent on de-repression of Ets transcription factors. In patient-derived oligodendroglioma cells, CIC reexpression or ETV5 blockade decreases lineage bias, proliferation, self-renewal and tumorigenicity. Our results identify Cic is an important regulator of cell fate in neurodevelopment and oligodendroglioma, and suggest that its loss contributes to oligodendroglioma by promoting proliferation and an OPC-like identity via Ets overactivity.


50
The identification of genes recurrently mutated in cancer often presents opportunities to uncover 51 previously unappreciated mechanisms regulating normal development, and vice versa. The 52 transcriptional repressor Capicua (CIC) has been identified as a likely tumour suppressor gene, as 53 recurrent mutations in CIC and/or reduced expression of CIC are found in several cancer types. In the 54 brain, CIC mutations are nearly exclusively found in oligodendrogliomas (ODGs) - glial tumors that are 55 composed of cells resembling oligodendrocyte precursor cells (1, 2). Indeed, concurrent IDH1/2 56 mutation, single-copy whole-arm losses of 1p and 19q, and mutation of the remaining copy of CIC on 57 As cells differentiated, Cic was increasingly localized to the nucleus, but with notable differences 111 between the levels detected in neuronal, astrocytic, and oligodendrocytic cell lineages. In the adult 112 (P56) cortex, NeuN+ neurons displayed the strongest nuclear Cic ( Figure  1I). The increase in Cic within 113 the neuronal lineage was detectable during the process of embryonic neurogenesis, with a modest 114 increase detected as cells transitioned from Sox2+ stem cells to Tbr2+ neuronal intermediate  and Gro-L domain (Figure  2A). We used these animals for in vivo studies as well as to generate cell 144 lines with which we could further study proliferation and lineage selection in a cell-autonomous manner.

146
By crossing Cic-floxed mice to Foxg1-Cre mice (in which Cre is expressed throughout the forebrain by 147 E10.5 (26)), we generated mice deficient in CIC in the embryonic telencephalon and subsequent 148 postnatal cerebrum (Figure  2A,B). Cic Fl/Fl ;;FoxG1 Cre/+ animals were born in approximate Mendelian ratios 149 and were grossly indistinguishable at birth. They failed to thrive postnatally, however, becoming visible 150 runts by postnatal day 7, and were uniformly lethal by P22 when they did not survive past weaning 151 (Suppl Figure  S2B). Gross and cursory light microscopic exam showed the presence of the expected 152 major anatomic structures in the brain, as well as the presence of laminated cortex, write matter tracts, 153 deep nuclei, and hippocampi (Suppl Figure S2A). The cerebra of Cic-null brains, however, were smaller 154 than littermate controls ( Figure 2C). The reason for lethality is unclear. We have not found major 155 anatomic defects detected in other organs and suspect that poor feeding secondary to impaired 156 neurologic function may be related to their progressive decline. This, however, remains speculative.

158
Interestingly, closer evaluation of P21 cortices revealed a shift in cell populations, with decreased 159 neurons and increased oligodendroglia cells and astrocytes in Cic-null cortices compared to controls 160 ( Fig 2D,D'

174
One possibility that could explain such a skewing of cell populations could be that there was increased 175 death in a particular population. However we found no evidence of increased apoptosis in the knockout 176 brains (Suppl Fig S3E). Other possibilities, were that Cic loss could be increasing proliferation and 177 affecting lineage selection in NSCs, or that Cic loss could have specific effects on OPC proliferation and 178 maturation -possibilities that are not mutually exclusive. In the following work, we focus our 179 investigations on Cic's role at the NSC stage, and examine its role in proliferation and lineage selection. Others recently reported proliferative gains upon loss of Cic (13), and we too found evidence of this. By 184 electroporating pCIG2-Cre (or pCIG2 empty vector control) into Cic Fl/Fl embryos, we could study the 185 effects of Cic loss in a discrete population of VZ cells and their subsequent progeny (schematic Figure   186 4A). 48 hours after Cre electroporation into VZ NSCs at E13 (early to mid-neurogenesis), EdU 187 incorporation was markedly increased, as were the numbers of Sox2 positive cells. Among GFP+ cells, 188 EdU+ cells were increased by 2.6-fold in cre- vs. control-electroporated brains (pCIG2-Cre 9.99±0.92% 189 Edu+, n=4, vs. pCIG2-Empty;; 3.77±0.81%, n=5, p<0.0001)( Figure  3A,A'). The fraction of GFP+ cells 190 expressing Sox2 was also increased (pCIG2-Cre 11.09±1.10% Sox2+ vs. pCIG2-Empty 5.48±1.01%;; 191 n=4 and n=5, respectively;; p<0.05) ( Figure 4B,B'). There was no change in activated Caspase-3,

192
indicating that the increased Sox2+ fraction in Cic-deleted cells was due to increased apoptosis in other 193 cells (Suppl Figure S3D,D'). Together, these findings supported a cell-autonomous increase in NSC 194 proliferation with CIC loss. Of note, there was also an increase in Edu+ cells among non-GFP cells 195 within the area of the electroporated patches, suggesting additional non-cell autonomous effects which 196 we did not pursue.

198
To confirm the cell autonomous gains in NSC proliferation with CIC ablation, we turned to cell culture.   Figure  S3C), supporting that the differences in viable cell numbers for Cic-null cells was due to their 209 higher proliferative rates, rather than reduced cell death. Together, the data show that Cic is a strong

220
To investigate whether Cic loss alters modes of cell division, we performed a paired cell assay. Cic null 221 and control NSCs were seeded at low density in adherent cultures for 24 hrs, then fixed and stained for     To directly test cell type specification capacity, we challenged Cic-null and control NSCs with exposure 283 to different lineage-promoting culture conditions. Neuronal differentiation was induced by culturing cells 284 with B27 and cAMP. Astrocytic differentiation was induced by culturing NSCs in 1% FBS and N2.

285
Oligodendroglial differentiation was induced by culturing cells in media with B27 and tri-iodo-thyronine.

286
After a 10 day exposure to these conditions, cultures were analyzed for cellular identity and morphology.

288
After 10 days in the neuronal condition, fewer Tuj1+ cells were generated in absolute numbers in Cic

306
We then asked what the Cic-null NSCs became, if not neurons and astrocytes, respectively. In the 307 neuronal condition, the Tuj1-population was predominantly comprised of Sox2+ cells, followed by

335
The results suggest to us that Cic-null cells are less sensitive to neuronal or astrocytic differentiation

477
There are some notable differences between the work reported by Yang,et al. (13), however, and our 478 studies. One difference is that the prior study used HOG cells that, despite their historic name, do not 479 carry the ODG-defining genetic features of 1p19q loss or CIC mutation;; similarly the Pdgfra-         Cocktail (Thermo). 20 μg protein lysate (50 μg in the case of CIC probing) was run on 4-12% Bis-Tris or 3-8% Tris-Acetate gels (Thermo). Protein was transferred to PVDF membranes in NuPAGE Transfer 632 Buffer (Thermo). Membranes were blocked in TBST with 5% powdered milk. Primary antibodies diluted 633 were applied for 1 hr at room temperature or overnight at 4°C. Horseradish peroxidase-coupled 634 secondary antibodies were applied for 1 hr at room temperature. Membranes were developed using 635 ECL Plus Western Blotting Reagent (GE Healthcare) and X-ray film.

639
For each of 3 biologic replicates per condition, 100 ng RNA was subjected to nanoString analysis on 640 nCounter system (NanoString Technologies) using a custom gene expression codeset containing 641 neurodevelopmental and brain cancer associated genes, and 3 housekeeping genes. Data were 642 analyzed using nSolver software and exported to PRISM software for further statistical analyses.

643
Counts for Etv1, Etv4, and Etv5 were normalized to the average of 3 house-keeping genes (GAPDH,

685
Data are represented as mean ± SD from at least 3 biologic replicates for experiments. Comparisons 686 between experimental and control samples were made using 2-tailed t-test or, when there were more 687 than 2 group, using ANOVA. Tukey's procedure was applied post-hoc to correct for multiple 688 comparisons during multiple pairwise analyses (e.g. differential Cic expression among cell types).

689
Bonferroni post-hoc correction was applied for statistical analysis of data from NanoString assays.