Neural stem cell self-renewal, neurogenesis and cell fate determination are three forces that control the generation of specific classes of neurons at the correct place and time during development. Much remains to be discovered of the cellular networks operating to control these processes. We report here that the transcription factor Pax6 directly regulates genes controlling the balance between neocortical stem cell maintenance, neurogenesis and the production of basal progenitor cells in a dosage-dependent fashion. While consistent with genetic loss-of-function studies 
, the direct nature of the control of these processes by Pax6 and their dosage sensitivity are unexpected. We propose that this dosage sensitivity reflects the need for a critical level of Pax6 within neocortical stem and progenitor cells.
Using Pax6 ChIP, we have identified a set of the promoters bound by Pax6 in neocortical stem cells in vivo at E12.5. This set of Pax6-bound genes defines the components of potential networks regulated by Pax6 in this tissue, and is noteworthy for the enrichment for genes involved in controlling cell cycle progression, transcription factors expressed in the three main cell types found in the early cortex (ventricular zone stem cells, basal progenitor cells and differentiating neurons) and also transcription factors expressed specifically in non-cortical neurons, including the ventral forebrain, spinal cord and retina, as well as in pancreatic islet cells.
These binding data are compatible with a number of models for Pax6 action. For example, Pax6 could repress expression of genes that are not normally expressed in the neocortex, or alternatively could simply bind to target sites in the promoters of genes normally expressed in other cell types in a Pax6-dependent manner, without driving their expression in the cortex. Similarly, for those genes expressed in cortex, Pax6 could either positively or negatively regulate cell cycle progression, and positively or negatively regulate basal progenitor cell genesis. Evidence for gene regulation by Pax6 is essential to resolve these questions. Therefore, we combined the binding data with transcriptome data from Pax6 gain- and loss-of-function experiments in the early developing cortex in order to identify those genes whose expression is dependent on Pax6 and the nature of that dependency, finding that 22% of Pax6-bound genes show evidence for regulation in vivo.
Transcriptome analyses of Pax6 gain and loss of function cortices identified a subset of genes that show reciprocal regulation, but also clearly demonstrate that the gain and loss of function gene expression changes are not simply complementary, in agreement with the reported anatomical phenotypes of Pax6 null and Pax6 over-expressing cortices 
. However, there is a striking overlap and complementarity in the changes of expression in basal progenitor genes observed in each mutant. We found that Pax6 directly promotes expression of a large set of genes specifically expressed in basal progenitor cells, including the key determinant of that cell type, the transcription factor Eomes/Tbr2: mutations in Eomes/Tbr2 lead to a loss of the cortical intermediate progenitor cell population, accompanied by a reduction in neurons in all cortical layers 
. Increasing Pax6 levels drives basal progenitor cell genesis from cortical stem cells, primarily by increasing Eomes/Tbr2 expression, along with the other basal progenitor cell genes such as Gadd45g, Neurod1, Sstr2 and Hes6 
. Basal progenitor cells undergo a limited number of mitotic divisions to generate neurons 
, thus the overall effect of shifting the stem cell population towards basal progenitor cells is to increase neurogenesis at the expense of neural stem cell maintenance in the early stages of cortical development, ultimately resulting in microcephaly, as observed here.
However, there are also changes in expression specific to either gain or loss of Pax6 function, with many more genes showing altered expression upon increased Pax6 levels. For example, from the combined Pax6 binding and regulation data we have found that Pax6 expression positively regulates stem cell self-renewal by promoting expression of the transcription factor Hmga2 and the G1 cyclin dependent kinase, Cdk4. Hmga2 promotes neural stem cell self-renewal by reducing expression of two negative regulators of the cell cycle, p16Ink4a
and p19Arf 
. Hmga2 reduces expression of p16Ink4a
indirectly via repression of JunB, a positive regulator of their expression, and we also found Pax6 binding to the promoter of JunB in cortical stem cells. p16Ink4a
slows cell cycle progression by inhibiting the G1 cyclin-dependent kinase Cdk4 
, and Cdk4 is bound and positively-regulated by Pax6 in cortical stem cells. In contrast, Pax6 also directly promotes expression of Pten and Fzr1/Cdh1, both of which reduce neural stem cell proliferation and self-renewal 
. Thus, under normal conditions in vivo
Pax6 has the potential to both promote and limit stem cell self-renewal. However, when Pax6 levels are increased, as in the D6-Pax6 transgenic cortex, the neurogenic functions of Pax6 are dominant over its ability to promote self-renewal.
We have also placed Pax6 in the context of other transcriptional regulators of self-renewal and neurogenesis, Hes1, Neurog2 and Ascl1/Mash1, in order to extend our coverage of the cellular networks controlling these processes. The marked overlap between those genes directly and indirectly regulated by Pax6 with the genes regulated by all three of the other factors provides strong evidence for the operation and architecture of the network regulating cortical neurogenesis, and the central importance of the basal progenitor population as a major output of that network. Pax6 and Neurog2 cooperate to promote neurogenesis, both directly and via the basal progenitor population, and this is opposed by the oscillating expression of Hes1 
. At the same time, Pax6 also promotes stem cell self-renewal in a manner that counterbalances its neurogenesis-promoting activity. However, when over-expressed, the promotion of neurogenesis and basal progenitor cell genesis by Pax6 is dominant over the promotion of self-renewal.
Therefore, we propose that there is an optimal level of Pax6 that determines the balance between neocortical stem cell self-renewal and neurogenesis: increasing that level drives stem cells to a neuronal or basal progenitor fate, whereas reducing the level leads to early cell cycle exit, manifest as increased early neurogenesis 
. In both cases, altering Pax6 levels leads to a depletion of the stem cell population by exiting to neurogenesis, but by different pathways and with different neuronal fates: cortical pyramidal cells when Pax6 is increased, inhibitory interneurons when Pax6 is absent. The sensitivity of cortical development to Pax6 levels underlines the importance of assessing subtle structural and functional anomalies in humans heterozygous for Pax6 mutations, as has been done for aniridia patients 
Finally, the findings of Pax6 function in neocortical stem and progenitor cells presented here have similarities with the functions of the ES cell pluripotency regulator Oct4 
. Oct4 shows marked dosage effects in ES cells in vitro
such that a reduction in Oct4 levels leads to trophectoderm differentiation and a two-fold increase in Oct4 levels leads to differentiation to primitive endoderm and mesoderm 
. Loss of Pax6 leads to a depletion of the cortical stem cell pool, via increased early neurogenesis secondary to a failure to self-renew, and also a switch in the fates of the neurons produced from glutamatergic cortical neurons to an inhibitory interneuron identity 
. Similarly, increased Pax6 expression also leads to depletion of the stem cell pool, but in this case by driving stem cells to a basal progenitor fate, leading to an overproduction of early-born, deep-layer cortical neurons. Thus the level of Pax6 controls whether neural stem cells will self-renew, generate cortical neurons or produce basal progenitor cells.