During forebrain development, radial glia generate neurons through the production of intermediate progenitor cells (IPCs). The production of IPCs is a central tenet underlying the generation of the appropriate number of cortical neurons, but the transcriptional logic underpinning this process remains poorly defined. Here, we examined IPC production using mice lacking the transcription factor nuclear factor I/X (Nfix). We show that Nfix deficiency delays IPC production and prolongs the neurogenic window, resulting in an increased number of neurons in the postnatal forebrain. Loss of additional Nfi alleles (Nfib) resulted in a severe delay in IPC generation while, conversely, overexpression of NFIX led to precocious IPC generation. Mechanistically, analyses of microarray and ChIP-seq datasets, coupled with the investigation of spindle orientation during radial glial cell division, revealed that NFIX promotes the generation of IPCs via the transcriptional upregulation of inscuteable (Insc). These data thereby provide novel insights into the mechanisms controlling the timely transition of radial glia into IPCs during forebrain development.
The coordinated proliferation and lineage-specific differentiation of neural progenitor cells plays an integral role in the formation of the mammalian cerebral cortex. The primary neural progenitor cells that generate the neurons of this structure are the radial glia, which develop from neuroepithelial cells around embryonic day (E) 10.5 in rodents (Anthony et al., 2004; Mori et al., 2005). Since the large number of neurons generated during development come from a relatively small initial population of progenitor cells, the radial glial cell pool is first amplified by undergoing symmetric proliferative divisions, also known as self-expanding divisions. Subsequently, radial glial cells undergo asymmetric divisions to give rise to either a neuron that migrates directly to the cortical plate (direct neurogenesis), or, more frequently, an intermediate progenitor cell (IPC) (indirect neurogenesis) (Götz and Huttner, 2005; Huttner and Kosodo, 2005). IPCs are morphologically different from radial glia in that they are delaminated from the adherens junctional belt at the ventricular surface of the brain (Noctor et al., 2004). The majority of cortical neurons arise through the production, expansion and differentiation of IPCs (Haubensak et al., 2004; Sessa et al., 2008).
The timely generation of IPCs is required for normal neuron number in the postnatal brain. Despite the importance of IPCs, our understanding of the mechanism by which asymmetric division of radial glia is coordinated to ensure timely IPC production is limited. In the classical model of neural stem cell division, inferred largely from work in Drosophila melanogaster, large changes in spindle orientation result in the asymmetric inheritance of the apical membrane into one daughter cell and an asymmetric cell fate (Knoblich, 2008). However, the vast majority of radial glial cell cleavage planes in the mammalian telencephalon are perpendicular to the ventricular surface, and deviate only slightly from this angle. As a result, the apical membrane typically segregates into both daughter cells (Konno et al., 2008; Asami et al., 2011; Shitamukai et al., 2011). Therefore, it is unlikely that unequal segregation of the apical membrane accounts for IPC- and neuron-generating divisions in the mammalian cortex. Rather, one proposed model is that small fluctuations in cleavage plane orientation (reviewed by Matsuzaki and Shitamukai, 2015) lead to changes in cell volume and intracellular organelle inheritance to promote IPC production (Wang et al., 2009).
The argument that small fluctuations in spindle orientation promote IPC production largely comes from loss- and gain-of-function studies of the mammalian homolog of the D. melanogaster adaptor protein Inscuteable (INSC) (Konno et al., 2008; Postiglione et al., 2011; Petros et al., 2015). INSC regulates the spindle orientation of radial glia and IPC production in a gene dose-dependent manner, whereby loss of Insc reduces oblique divisions and IPC number, whereas high levels of Insc expression increase oblique divisions and IPC number (Postiglione et al., 2011; Petros et al., 2015). Currently, it remains unclear whether INSC-dependent spindle orientation directly regulates IPC production or whether another INSC-dependent mechanism might regulate the development of IPCs. However, a pivotal question that arises from these data is how Insc expression itself is regulated during cortical development to facilitate IPC development.
Transcription factors of the Nuclear factor one (Nfi) family (Nfia, Nfib, Nfix) play a crucial role in astrogliogenesis. Mice lacking Nfix exhibit markedly reduced numbers of astrocytes throughout the embryonic cerebral cortex and cerebellum (Piper et al., 2011; Heng et al., 2014). In addition to promoting astrocyte lineage progression, individual Nfi knockout mice also exhibit elevated numbers of progenitor cells and delayed expression of neuronal markers within the ammonic neuroepithelium of the presumptive hippocampus during embryonic development (Piper et al., 2010, 2014; Heng et al., 2014). From these findings we posited that NFIs could play a previously unrecognized role in the production of IPCs. Here, we use the ammonic neuroepithelium of mice lacking Nfix and Nfib as a model to investigate this hypothesis. We demonstrate that NFIs are autonomously required by radial glia for timely IPC production, and that NFIs directly activate the expression of Insc, providing novel insight into the cellular processes governing the transition of radial glial cells to IPCs during hippocampal development.
NFIX is expressed by radial glia and IPCs during hippocampal development
To investigate whether NFIX promotes indirect neurogenesis and IPC production during the development of the cerebral cortex, we focused predominantly on the ammonic neuroepithelium of the hippocampus. This region was chosen for two reasons; first, NFIX expression within the ventricular zone (VZ) and subventricular zone (SVZ) of the cerebral cortex exhibits a gradient, such that expression is highest in the caudomedial cortex, particularly within the presumptive hippocampal primordium (Fig. 1A,B); second, newly generated neurons from the ammonic neuroepithelium (predominantly pyramidal neurons) migrate only a short distance radially before settling in the cornu ammonis (CA) region (Altman and Bayer, 1990), rendering it a simpler model of development than the six-layered neocortex.
Previous studies have demonstrated that neural progenitor cells in the hippocampal neuroepithelium express NFIX during development from at least E13.5 onwards (Heng et al., 2014); however, the cell type-specific identity of these cells has not been established. The transition of radial glia to IPCs during development is demarcated by the sequential expression of PAX6 then TBR2 (EOMES) (Englund et al., 2005). We classified radial glia as PAX6+ TBR2– rather than using PAX6 expression alone as a marker for these cells, thereby excluding newborn radial glia-derived IPCs that also express low to moderate levels of PAX6 (Arai et al., 2011). Co-immunofluorescence at E14.5 (Fig. 1C) revealed that radial glia express NFIX (209/209 cells), as did the vast majority of IPCs (TBR2+ nuclei; 208/209 cells). Moreover, the expression of NFIX was maintained in radial glia and IPCs throughout the stages of the cell cycle, including mitosis (Fig. S1). This same expression pattern was also observed at E13.5 and E15.5 (Fig. S1).
Delayed IPC development in Nfix−/− mice
We have previously reported elevated numbers of PAX6+ cells (indicative of increased numbers of radial glial cells) and delayed neuronal differentiation in the hippocampus of Nfix−/− mice at E16.5 (Heng et al., 2014). It remained unclear, however, when these developmental defects first become apparent and which stage of the lineage transition from a stem cell to a mature neuron is affected in these mice. Using PAX6 and TBR2 colabeling, we investigated how the numbers of radial glial cells in the ammonic neuroepithelium of Nfix−/− mice change from the onset of hippocampal neurogenesis at E13.5 to the peak of neurogenesis at E15.5. Relative to controls, we found significantly more radial glia (PAX6+ TBR2–) in Nfix−/− mice at E13.5, E14.5 and E15.5 (Fig. 2A-E). Moreover, the magnitude of the change was smallest at E13.5 (P=0.022) and progressively increased at E14.5 (P=0.0096) and E15.5 (P=0.014), from which we infer that the lack of Nfix culminates in an ongoing (rather than a temporally restricted) delay in the transition of radial glia into IPCs. In support of this, we found fewer IPCs at E13.5 in Nfix−/− mice than in controls (P=0.00013) (Fig. 2F), demonstrating that radial glia generate IPCs less efficiently in these mice. By E14.5 and E15.5, the total number of IPCs in Nfix−/− mice was comparable to that in wild type (Fig. 2F), probably because the increased size of the radial glial pool resulted in an increase in the absolute number of IPC-generating divisions. However, measured as a proportion of all cell types in the VZ/SVZ, the number of IPCs remained reduced at E15.5 in Nfix−/− mice (data not shown).
Collectively, these data suggest that Nfix-deficient radial glial cells undergo increased self-expanding divisions and reduced IPC-generating divisions during early hippocampal development. To examine this further, we performed birth-dating experiments by injecting time-mated dams with the DNA analog BrdU at E13.5 before sacrificing embryos 24 h later, thereby labeling all proliferating cells (predominantly radial glia) that were in S-phase at the time of injection, as well as their progeny. We found that 24 h later (at E14.5), there were substantially fewer IPCs generated by BrdU-labeled radial glia in Nfix−/− mice than in controls (Fig. 2G; P=0.0001).
The impaired IPC development observed in Nfix−/− mice could potentially occur as a result of increased direct neurogenesis from radial glia, similar to that observed in Elp3 conditional knockout mice (Laguesse et al., 2015). To determine whether this was the case we performed a 48 h BrdU chase experiment (labeling with BrdU from E13.5). There were significantly fewer BrdU+ cells that had exited the cell cycle (BrdU+ Ki67–) in mutant mice relative to controls at E15.5 (P=0.00016; Fig. 2H-J). This argues that the impaired IPC development in Nfix−/− mice is not offset by an increase in direct neurogenesis. Indeed, a significantly greater number of BrdU+ cells remained as PAX6+ TBR2− radial glia at E15.5 (P=2.2×10−5). These data, coupled with the expansion of the radial glial population from E13.5-E15.5, support the hypothesis that Nfix-deficient radial glia undergo more self-expanding divisions at the expense of IPC development during this stage of hippocampal development.
Nfix-deficient radial glia have a longer S-phase
We next posited that Nfix-deficient radial glial cells should display cellular characteristics associated with fewer IPC-generating divisions. Interestingly, a recent study revealed that radial glia committed to neurogenic divisions have a shorter S-phase (1.8 h) than radial glia during self-expanding divisions (8 h) (Arai et al., 2011). To further assess if NFIX promotes IPC production, we measured S-phase duration as a proxy for the frequency of neurogenic divisions. We hypothesized that the population of Nfix-deficient radial glia would exhibit a longer average S-phase due to the reduction in neurogenic divisions. To test this, we performed consecutive injections of the S-phase markers EdU and BrdU at E14.5 to determine the relative rate of progression through S-phase, as well as cell cycle duration (see Materials and Methods, Fig. 3A-C) (Martynoga et al., 2005). We identified radial glia as TBR2− nuclei in the VZ, defined as the region apical to the thick band of TBR2+ cells (Fig. 3C). Consistent with our hypothesis, Nfix−/− radial glia had a significantly longer S-phase than wild-type radial glia (P=0.025; Fig. 3D). Total cell cycle duration in Nfix−/− radial glia was not significantly different (P=0.051; Fig. 3E), nor was G1/G2/M phase length (P=0.11; Fig. 3F). Thus, the increased S-phase duration further indicates that Nfix−/− radial glia undergo proportionally fewer IPC-generating divisions during the period of peak neurogenesis.
Overexpression of NFIX in vivo promotes IPC and neuron generation
If loss of NFIX impairs IPC generation, then NFIX overexpression should result in an increased rate of IPC and neuronal differentiation. To investigate this, we used in utero electroporation to overexpress an HA-tagged mouse NFIX construct containing a bicistronic GFP reporter (NFIX pCAGIG) (Heng et al., 2014) or vector-only control (pCAGIG) in the presumptive hippocampus of wild-type CD1 mice at E12.5 (Fig. 4A,B). At E14.5 we found that NFIX overexpression led to a significantly higher percentage of electroporated cells becoming IPCs (P=0.037) and neurons (P=0.016), while fewer cells remained as radial glia (P=1.8×10−5), as compared with electroporated cells in the control condition (Fig. 4C-E). Importantly, co-staining for the neuronal marker TBR1 revealed that all neurons were located basal to the band of TBR2+ cells, demonstrating that NFIX overexpression did not induce gross neuronal migration errors (data not shown). Together, these data are supportive of a role for NFIX in promoting the commitment of radial glia into IPCs.
Loss of four Nfi alleles results in a more severe IPC phenotype
We have thus far demonstrated that Nfix-deficient radial glia generate IPCs with reduced efficiency. Despite this, Nfix−/− mice lack the gross morphological abnormalities often associated with mouse models in which radial glial populations are expanded and there is decreased neurogenesis (Chenn and Walsh, 2002; Farkas et al., 2008). For example, these phenotypes are usually associated with increased tangential length of the cortex and reduced radial thickness. We therefore questioned whether the loss of additional Nfi alleles would result in a more severe phenotype. We removed four Nfi alleles by generating conditional floxed Nfix and Nfib mice, which were then crossed to a mouse line expressing a tamoxifen-activated form of Cre recombinase under the control of the ubiquitous Rosa26 promoter. Time-mated dams were then injected with tamoxifen (two sequential injections, at E10.5 and the other at E11.5) to delete Nfix and Nfib, and embryos were analyzed at E15.5. We found that Cre-expressing Nfixfl/fl; Nfibfl/fl (hereafter Nfixfl/fl; Nfibfl/fl; cre) embryos were almost entirely devoid of NFIX and NFIB immunoreactivity (Fig. 5H,I), whereas cre-negative animals retained NFIX and NFIB expression (Fig. 5E,F). In Nfixfl/fl; Nfibfl/fl; cre animals, the tangential length to radial width ratio of the presumptive hippocampus was markedly increased (251%) compared with control animals (P<0.0001), as was that of the neocortex (P=0.025) (Fig. 5A-C). The length/width ratio of the Nfixfl/fl; Nfibfl/fl; cre hippocampus was also significantly larger than in Nfix−/− mice (P=0.0008), suggestive of a more severe phenotype than in Nfix−/− animals (Fig. 5C). Furthermore, cellular analysis revealed that, compared with controls, Nfixfl/fl; Nfibfl/fl; cre animals had a greater increase in radial glial cell number (PAX6+ TBR2−; P=0.0006; Fig. 5D) than Nfix−/− mice (Fig. 2E) at E15.5. Moreover, Nfixfl/fl; Nfibfl/fl; cre animals had a 50% reduction in IPC number at E15.5 (P=0.0002; Fig. 5D), whereas in Nfix−/− mice the IPC number did not differ significantly from that of wild type at this age (Fig. 2F). Together, these data demonstrate that the loss of additional Nfi alleles results in morphological and cellular phenotypes associated with reduced IPC production.
During development, the cell-cell interactions between different progenitor populations is important for regulating the balance between the self-expansion of radial glia or their commitment to IPC and neuronal differentiation (Namihira et al., 2009). Owing to the ubiquitous expression of NFIs by radial glia during cortical development, it is possible that NFIs promote the generation of IPCs by regulating the expression of extrinsic signaling molecules that affect the fate of neighboring radial glial cells (non-cell-autonomous effect), rather than the host cell itself (cell-autonomous effect). Distinguishing between these scenarios could provide important insight into the mechanisms through which NFIs drive the generation of IPCs from radial glia. To analyze these possibilities, we took advantage of the different DNA recombination efficiencies of the floxed Nfix and Nfib alleles (Fig. S2). The reduced recombination efficiency of the Nfib floxed allele resulted in a stochastic effect whereby, in random areas of the cortex, both Nfix alleles had recombined after administration of tamoxifen whereas both Nfib alleles had not. The result was that, at E15.5, Nfixfl/fl; Nfibfl/fl; cre cortices were sparsely patterned with NFIX– NFIB+ clones (Fig. 5E-K). We took advantage of this serendipitous occurrence to determine whether NFIs were cell-autonomously required by radial glia for IPC development. Were this the case, then NFIX– NFIB+ clones should show increased numbers of IPCs compared with the cells surrounding the clone. Conversely, a non-cell-autonomous requirement for NFI would predict that the number of IPCs within NFIX– NFIB+ clones would be similar to that in the surrounding cells. We stained Nfixfl/fl; Nfibfl/fl; cre hippocampi for NFIB, NFIX and TBR2, and analyzed the proportion of cells within NFIX– NFIB+ clones that were IPCs (TBR2+). Within these NFIX– NFIB+ clones there was a greater proportion of IPCs compared with immediately adjacent NFIX– NFIB– regions (P=0.0029; Fig. 5J-L). These data, together with our results from overexpressing NFIX through in utero electroporation (Fig. 4), support a model in which NFIs are autonomously required by radial glia for normal IPC development.
Insc is a target for transcriptional activation by NFIs during hippocampal development
What are the transcriptional mechanisms through which NFIs promote IPC production? To address this question, we analyzed existing microarray datasets that examined alterations in gene expression within the entire hippocampal primordium of E16.5 Nfix−/− (Heng et al., 2014), Nfib−/− (Piper et al., 2014) and Nfia−/− (Piper et al., 2010) mice. In these datasets there were 668, 1893 and 1099 misregulated genes relative to wild-type controls, respectively (±1.5-fold change, P<0.05). To focus our search, we further filtered these data using a number of stringent parameters. First, we refined our analysis to only include genes misregulated in all three microarray datasets. We took this approach because it is likely that NFIs promote IPC production through a common transcriptional mechanism; this is supported by previous data highlighting the functional similarity of NFIs in the development of other cortical phenotypes (Vidovic et al., 2015), the additive effects of deleting NFIX and NFIB on IPC development (Fig. 5), and our analysis of the microarray datasets, which revealed that a highly significant proportion of genes are commonly misregulated across each of the three datasets (Table S1). Next, we only included genes with a greater than ±2.5-fold change (Fig. 6A, Table S2). Finally, we used a recently derived chromatin immunoprecipitation and sequencing (ChIP-seq) dataset (performed using a pan-NFI antibody on embryonic stem cell-derived neural stem cell cultures) to identify which of these remaining genes had a promoter region bound by an NFI protein (Fig. 6B) (Mateo et al., 2015). After this final filter, only six candidate genes remained in the analysis pipeline (Fig. 6C). Crucially, one of these was Gfap, an established target of NFI transcriptional activation in vitro (Namihira et al., 2009) and in vivo (Heng et al., 2014), highlighting the effectiveness of our filters. Ncam1, which regulates neural cell adhesion and migration, was also misregulated, as were three genes of unknown function in central nervous system development (Rasd2, Entpd2, Il16).
Significantly, the final gene in the analysis pipeline was the mammalian homolog of Drosophila insc. Because Insc promotes IPC production during mouse cortical development (Postiglione et al., 2011), we inferred that NFIs might activate Insc expression, and that the downregulation of Insc in Nfi null mice (Fig. 6C) contributes to the IPC phenotype observed within these lines.
NFIs activate Insc promoter-driven transcriptional activity
To investigate whether NFIs activate transcription of Insc, we first validated the microarray results from Nfix−/− E16.5 hippocampi (Heng et al., 2014). Using qPCR we determined that there was a ∼50% reduction in hippocampal Insc mRNA compared with that in wild-type littermates at E16.5 (P=0.05; Fig. 6D). Importantly, Insc was also downregulated in the medial cortex (hippocampal primordium and medial neocortex) of Nfix−/− mice at E13.5, when the IPC phenotype of these mice first becomes apparent (P=0.044).
Next, we verified the capacity of NFI proteins to physically bind to the Insc promoter by performing an in vivo ChIP assay on tissue isolated from E14.5 forebrains using a pan-NFI antibody. The ChIP assay revealed enrichment of NFI protein binding at a region in the Insc promoter that corresponds to both the NFI dyad consensus site (P=0.05; Fig. 6E) and the NFI ChIP peak detected by Mateo et al. (2015) in vitro.
Finally, we asked whether NFIs could active Insc promoter-driven transcriptional activity. A region of the Insc promoter, which included the dyad consensus site, was cloned upstream of the luciferase gene (Fig. 6F,G). Co-transfection of the Insc luciferase construct with an NFIX expression plasmid in Neuro2a cells revealed that NFIX strongly activated Insc promoter-driven transcriptional activity (P=1.8×10−6; Fig. 6H). Likewise, NFIA (P=0.0001) and NFIB (P=0.0026) also enhanced Insc promoter-driven transcriptional activity (Fig. 6H).
NFI-deficient radial glia phenocopy the cleavage plane defects of Insc knockout mice
If the downregulation of Insc contributes to the impaired generation of IPCs in Nfix−/− mice, then we would anticipate that the spindle orientation and cleavage plane deficits seen within Insc conditional knockout mice (Postiglione et al., 2011) would be recapitulated within the radial glia of Nfix−/− mice. We assessed cleavage plane orientation of the dividing progenitor cell relative to the ventricular surface (Fig. 6I-L). At E14.5 in wild-type mice, 71% of radial glia divided with a vertical cleavage plane (the angle between the cleavage plane and the ventricular surface was between 90° and 60°), whereas 23.5% of radial glia divided with an oblique division plane (between 60° and 30°; Fig. 6M). Consistent with previous reports (Postiglione et al., 2011; Insolera et al., 2014), mitotic radial glia with a horizontal cleavage plane (between 30° and 0°) were rare, occurring in only 4.7% of cells (Fig. 6M). In Nfix−/− mice, however, the vast majority of radial glia divided with a vertical cleavage plane (89.9%), and the proportion of oblique divisions was greatly reduced (10.1%) (Fig. 6N). Moreover, no radial glial cell was observed to divide with a horizontal cleavage plane in Nfix−/− mice. Similarly, in Nfixfl/fl; Nfibfl/fl; cre radial glia at E15.5, 98.4% of radial glia divided with vertical cleavage plane, 1.6% with an oblique cleavage plane, and no cells were observed dividing horizontally (Fig. 6O). This phenotype is reminiscent of the cleavage plane phenotype of Insc conditional knockout mice, where there are substantially fewer oblique divisions, and horizontal cleavage planes are not seen (Insolera et al., 2014). Thus, the data support a model whereby downregulation of Insc in NFI-deficient mice results in fewer oblique divisions in radial glia.
In vivo rescue of IPC number in Nfix−/− radial glia through INSC overexpression
If NFIX promotes IPC development, in part or wholly through activating Insc expression, then overexpressing INSC in Nfix−/− radial glia should help restore IPC generation to wild-type levels. To test this hypothesis, we used the CRISPR/Cas system to delete Nfix from radial glia within the ammonic neuroepithelium of embryonic CD1 wild-type mice. After electroporation at E12.5 with plasmids encoding Nfix-CAS9 and pCAGIG (knockout condition) or lacZ-CAS9 and pCAGIG (control condition), embryos were assessed for NFIX expression. In the knockout condition, only 8% of electroporated (GFP+) cells retained NFIX immunoreactivity compared with 96% in the control (Fig. S3). We then assessed TBR2 expression, and found proportionally fewer TBR2+ GFP+ cells in the knockout compared with the control condition (P=0.03), reminiscent of our findings using Nfix−/− mice (Fig. 7A-D). Finally, for the rescue condition, we co-electroporated the Nfix-CAS9 plasmid with INSC pCAGIG (Fig. S3). In this rescue condition there was an increased proportion of TBR2+ GFP+ cells compared with the knockout condition (P=0.023). Moreover, the number of TBR2+ GFP+ cells in the rescue condition was not significantly different from that of the control condition (P=0.62) (Fig. 7A-D). Thus, Insc overexpression in Nfix−/− radial glia was sufficient to restore IPC numbers to wild-type levels, confirming that NFIX regulation of INSC contributes to IPC production during hippocampal development.
A prolonged neurogenic window results in increased neuron number in the hippocampus and neocortex of P20 Nfix−/− mice
What effects do the increased self-expanding divisions of radial glia and the delayed IPC generation have on the late embryonic phenotype of Nfix−/− mice? We had previously observed that the reduction in TBR2+ cells in the Nfix−/− hippocampus was not permanent, so that at E18.5 there were greater numbers of these cells than in wild-type mice (Heng et al., 2014). This suggested that the expansion of the radial glial population in the presumptive hippocampus of Nfix−/− mice during early development would sustain the pool of radial glia and prolong the period of neurogenesis, thereby resulting in an overall greater production of neurons.
We investigated this by immunostaining wild-type mouse hippocampal sections for PAX6 and TBR2 at E18.5. This revealed very few radial glia or IPCs in the ammonic neuroepithelium, signaling the end of hippocampal neurogenesis. By contrast, in Nfix−/− mice at E18.5 the radial glial (P=2.9×10−5) and IPC (P=1×10−5) populations were ∼2.5-fold and ∼3.5-fold larger than in the control, respectively (Fig. 8A-C). We then asked whether this increase in radial glia and IPC number culminated in the increased production of CA pyramidal neurons in the postnatal hippocampus. Counts of the mature neuronal marker NeuN (Rbfox3) at postnatal day (P) 20 revealed significantly more neurons within both the CA1 and CA3 regions of the mutant compared with the controls (P=0.0011; Fig. 8D-F). Moreover, we tracked the fate of E18.5 progenitor cells by injecting pregnant dams with BrdU and immunostaining for BrdU and NeuN in Nfix−/− and wild-type mice at P20. In wild-type mice there were very few neurons in the P20 CA regions that were born at E18.5 (BrdU+ NeuN+). Remarkably, however, the number of late-born neurons in Nfix−/− mice was ∼10-fold higher (P=0.0002; Fig. 8D-F), which is likely to have contributed to the overall greater number of CA neurons. Additionally, we found a significantly greater number of S100β+ astrocytes in the stratum oriens of postnatal Nfix−/− mice (P=0.0107; Fig. S4). Despite the increases in neuron and glial number in these regions, the overall size of the hippocampus was not significantly different in Nfix−/− mice compared with controls, probably because the dentate gyrus, as previously described (Heng et al., 2014), is much smaller in these mutant mice.
Given that postnatal Nfix−/− mice have a dorsoventral expansion of the cingulate cortex and neocortex (Campbell et al., 2008), we asked whether the increased neuronal generation observed in the postnatal hippocampus was evident within the broader dorsal telencephalon, and so may contribute towards this phenotype. At E18.5 we saw similar phenotypes within the neocortex as in the hippocampal ammonic neuroepithelium, with increased radial glial (P=0.031) and IPC (P=0.0002) numbers compared with controls, although the phenotype was substantially less severe than that observed in the hippocampus (Fig. S5A-C). BrdU labeling at E18.5 also revealed significantly more late-born neurons (BrdU+ NeuN+) in the upper layers of the neocortex (P=0.0167; Fig. S5D-F). Thus, the prolonged neurogenic period of Nfix−/− mice is likely to contribute to the increased dorsal-ventral size of these regions.
Collectively, these findings highlight the crucial role that NFIX plays in promoting indirect neurogenesis within the developing dorsal telencephalon, significantly enhancing our understanding of the transcriptional mechanism underpinning the development of this brain region.
Studies have begun to reveal the factors that drive the progression of radial glia into IPCs during development (Farkas et al., 2008; Sessa et al., 2008; Saffary and Xie, 2011). For example, loss of the transcription factor PAX6 from radial glia downregulates the expression of the microtubule-associated protein SPAG5 and, in turn, radial glia undergo more divisions with oblique cleavage planes and generate more basal mitoses (Götz et al., 1998; Asami et al., 2011). However, these basal mitoses retain the hallmarks of radial glial cells, indicating that PAX6 regulates the delamination of radial glia progenitors from adherens junction complexes, but does not regulate other key features of IPC differentiation. Likewise, HES transcription factors inhibit the neurogenic divisions of radial glia by repressing proneural genes (Mizutani et al., 2007; Pierfelice et al., 2011), but it is unclear how this repression directly affects the maintenance of apical-basal polarity, cell cycle progression, or other aspects of radial glial cell biology. In this study, we have uncovered a novel role for members of the NFI transcription factor family in orchestrating the transition of radial glia to IPCs. Although NFIs have previously been shown to be important for astrogliogenesis and stem cell maintenance in the neocortex and hippocampus, here we reveal that NFIX-mediated activation of Insc expression promotes the timely generation of IPCs.
The function of INSC has been extensively studied during D. melanogaster neurogenesis. INSC acts in D. melanogaster neuroblasts as an adaptor protein, linking two protein complexes that assemble at the apical cell cortex to control spindle orientation (Wodarz et al., 1999; Schaefer et al., 2000). The control of mitotic spindle orientation by INSC during neuroblast division leads to unequal inheritance of the cell fate determinants Prospero, Numb and Brat to the basal daughter cell and results in neuronal differentiation (Knoblich, 2008). There is a single insc homolog in mice (Katoh and Katoh, 2003). INSC is expressed at very low levels within the developing cerebral cortex, which is perhaps why changes in INSC expression result in only modest fluctuations in spindle orientation in mice compared with flies. These modest fluctuations in spindle orientation are, however, correlated with large increases in IPC number with increased INSC expression, and vice versa (Postiglione et al., 2011). Despite the clear relationship between the level of Insc expression and the rate of IPC production in the mouse cortex, the mechanism through which Insc expression is regulated within the cortex has, until now, remained unclear.
Although we found that the impaired IPC development and spindle defects of Nfix-deficient radial glia closely mirror those of Insc conditional knockout mice during early development (Postiglione et al., 2011), there are some important differences between these mutants. One of these differences is that the number of radial glial cells in Nfix-deficient mice is substantially increased during early development compared with controls. By contrast, although the radial glia in Insc conditional knockout mice were quantified using only PAX6 as a marker (thereby including newborn IPCs that retain PAX6 expression), there were no gross changes in the number of radial glia in these mice. As a corollary to this, we found that the expansion of radial glia in Nfix−/− mice sustained the pool of radial glia progenitors into late development, thereby ensuring an overall increase in the number of neurons in the hippocampal CA1/CA3 neuronal layers and neocortex, whereas the neocortical neuron number in Insc conditional knockout mice was reduced (Postiglione et al., 2011). These differences are likely to be explained by the diverse cohort of genes that are likely to be under NFI transcriptional control, in addition to Insc. For example, we have previously shown that NFIs can repress genes involved in stem cell maintenance during cortical development, such as Sox9, Ezh2 and Hes1 (Piper et al., 2010, 2014; Heng et al., 2014). Therefore, although the downregulation of Insc contributes to the impaired IPC generation in Nfix null mice, the misregulation of these additional factors probably explains the expansion of the radial glial pool beyond that observed following loss of Insc alone. NFIX is likely to have functions in cell types other than radial glia that could also account for differences between these mouse lines. For example, NFIX is expressed by IPCs. Does NFIX regulate the differentiation of IPCs to mature neurons, and if so, does this contribute to the accumulation of IPCs at E18.5? Future studies using a Tbr2- or Dcx-driven cre could address this question.
Our data further highlight the pleiotropic roles of NFIs during forebrain development. How NFI proteins promote the generation of IPCs by activating Insc promoter-driven transcription, while repressing genes involved in stem cell maintenance, and also being crucial in astrocytic lineage progression, is a fascinating and open question. Future studies aimed at identifying the genome-wide chromatin binding profile of NFIs in purified populations of neural stem cells, neurons and glia across developmental time will lead to a broader understanding of how these transcription factors fulfil these functions during development.
MATERIALS AND METHODS
The work performed in this study conformed to The University of Queensland's Animal Welfare Unit Guidelines for Animal Use in Research (AEC approval numbers QBI/353/13/NHMRC and QBI/355/13/NHMRC/BREED) and those of the IACUC at Roswell Park Cancer Institute. All experiments were performed in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes, and were carried out in accordance with The University of Queensland Institutional Biosafety Committee.
Nfix−/− and Nfix+/+ littermates used in this study have been described previously (Campbell et al., 2008). Conditional Nfix and Nfib alleles, each of which contains loxP sites flanking exon 2 of the respective gene were generated as described previously (Messina et al., 2010; Hsu et al., 2011). The conditional lines were then bred together to produce the double conditional strain (Nfixfl/fl; Nfibfl/fl), which was subsequently crossed to Rosa26-CreERT2 mice (#008463, The Jackson Laboratory). The Nfixfl/fl; Nfibfl/fl; cre nomenclature is used to reflect the fact that the Nfib allele did not recombine as efficiently. Animals of either sex were used.
In utero electroporation
In utero electroporation was performed as previously described (Suarez et al., 2014), with minor modifications whereby 0.5-1 µl of plasmid DNA was injected into one lateral ventricle and electroporated caudomedially into the presumptive hippocampus using 35 V. Plasmid expression constructs were pCAG IRES GFP (pCAGIG) or NFIX pCAG IRES GFP (NFIX pCAGIG). For the rescue experiment, gNfix-CAS9, glacZ-CAS9, INSC pCAGIG and pCAGIG constructs were used. For details of plasmid construction, see the supplementary Materials and Methods.
Measurement of cleavage plane orientation in Nfix−/− radial glia
To analyze cleavage plane orientation within radial glia, we stained sections for phospho-histone H3 (PHH3), α-tubulin and with DAPI. Sections were imaged using a 63× objective on a spinning disk confocal microscope through a depth of 10 µm (consecutive 1 µm z-steps). Hippocampal radial glia undergoing mitosis were identified as PHH3-positive cells located at the ventricular surface of the ammonic neuroepithelium. For each cell three angle measurements were taken from adjacent z-stacks and averaged. Because the metaphase plate of radial glia rocks extensively until anaphase (Haydar et al., 2003; Sanada and Tsai, 2005), only cells that were in anaphase or telophase were analyzed, as revealed by chromosomal (DAPI and PHH3 staining) and mitotic spindle arrangement (α-tubulin staining). We measured cleavage plane orientation by the angle created by the vector that runs parallel to the ventricular surface and the vector that runs through the cleavage plane of the dividing cell.
Immunofluorescence and immunohistochemistry
Embryos were immersion fixed at E14.5 or younger in 4% paraformaldehyde (PFA) or perfused transcardially (E15.5 and older) and sectioned for immunohistochemistry using the primary and fluorescently conjugated secondary antibodies described in detail in the supplementary Materials and Methods.
Nfix−/− mouse hippocampal cell counts
For each section (two sections per animal) cell counts were performed from two 100 µm sampling fields, spanning the width of the hippocampal primordium (E13.5-E18.5) or neocortex (E18.5) as described in the supplementary Materials and Methods.
Nfix−/− mouse birth-dating experiments
BrdU birth-dating experiments were performed using a single low dose (50 mg/kg) or high dose (200 mg/kg) injection. For the low dose, dams were injected at E13.5 or E18.5 and examined 24 h or 20 days later, respectively. For the high dose, dams were injected at E13.5 and examined 48 h later. Cell counts were then performed as described in the supplementary Materials and Methods.
Measurement of cell cycle kinetics in Nfix−/− radial glia
Mean cell cycle (Tc) and S-phase (Ts) durations were determined by a dual-pulse labeling method whereby EdU (50 mg/kg) and 1 h later BrdU (50 mg/kg) were injected into pregnant dams, before the animals were sacrificed 1.5 h post EdU injection. Counts of EdU+ and BrdU+ cells were then performed to determine Tc and Ts as described in the supplementary Materials and Methods and Fig. S6.
Nfixfl/fl; Nfibfl/fl; Rosa26-CreERT2 tamoxifen treatment and cell analysis
Animals were injected with 2 mg tamoxifen dissolved in corn oil at E10.5 and E11.5 before being immersion fixed in PFA at E15.5. Cell counts were performed as in Nfix−/− mice, except for the clonal analysis, where a minimum of two clones per animal were assessed regardless of their rostral-caudal or lateral position within the hippocampus. Cellular analysis of these clones is described in the supplementary Materials and Methods.
Quantitative real-time PCR (qPCR)
The medial cortex (hippocampal primordium and medial neocortex) of E13.5 embryos or the entire hippocampus of E16.5 embryos was dissected and snap frozen. After RNA extraction (RNeasy Micro Kit, Qiagen) and reverse transcription (Superscript III, Invitrogen), qPCR was performed using SYBR Green (Qiagen) using 500 nM of forward and reverse primers. Primer sequences and further technical details are provided in the supplementary Materials and Methods.
Reporter gene assays
DNA was transfected into Neuro2A cells (1×104 cells) in a 96-well plate using Lipofectamine 2000 (Invitrogen), and Cypridina luciferase was added to each transfection as an internal control. After 24 h, luciferase activity was measured using a dual-luciferase system (Switchgear Genomics), as described in the supplementary Materials and Methods.
A putative ChIP peak within the Insc promoter was identified by analysis of a published ChIP-seq dataset (supplementary table S3, NFI tab, in Mateo et al., 2015). E14.5 forebrains were dissociated and fixed in 1% formaldehyde for 8 minutes. Nuclei were lysed and chromatin sonicated using the Bioruptor Pico (Diagenode) to 100-500 bp. Immunoprecipitation was performed with 8 µg goat anti-NFI (sc-30918, Santa Cruz) or 8 µg goat IgG control (AB-108-C, R&D Systems) antibody coupled to 40 μl Protein G Dynabeads (10003D, Thermo Fisher Scientific). Enrichment of NFI at the Insc promoter was calculated relative to the IgG control using the delta CT method, and normalized to a negative control primer set. For primer sequences and further details see the supplementary Materials and Methods.
Student's t-tests were performed when comparing two groups (n>4) or a Mann-Whitney U-test (n=3). For comparisons of more than two groups, ANOVA was performed followed by multiple comparisons analysis (corrected using the Holm-Šídák method) using a pooled estimate of variance if appropriate (Prism 6.0, GraphPad). See the supplementary Materials and Methods for further details.
Microscopy was performed in the Queensland Brain Institute's Advanced Microscopy Facility. We thank Rowan Tweedale and Dominic Ng for critical analysis of the manuscript. For provision of reagents, we thank Juergen Knoblich, Nathan Godde and Patrick Humbert.
L.H. designed, carried out, and interpreted the experiments and wrote the manuscript. O.Z., I.G., H.M., J.O., T.J.H., A.E., I.G.-N., J.I.H. carried out experiments, interpreted data and edited the manuscript. D.V., T.H.B., L.J.R. interpreted data and edited the manuscript. R.M.G. generated mouse lines, interpreted data and edited the manuscript. M.P. provided reagents and mouse lines, designed experiments, interpreted data and wrote the manuscript.
This work was supported by National Health and Medical Research Council project grants [1003462, 1057751 and 1022308 to M.P.]; an Australian Research Council Discovery Project grant [DP160100368]; and by National Institutes of Health and NYSTEM grants [HL080624, C026714 and C026429 to R.M.G.]. M.P. was supported by an Australian Research Council Future Fellowship [FT120100170] and a Research Award from the Hydrocephalus Association. L.H., A.E. and D.V. were supported by Australian Postgraduate Awards. Deposited in PMC for release after 12 months.
The authors declare no competing or financial interests.