Tangential migration and proliferation of intermediate progenitors of GABAergic neurons in the mouse telencephalon

In the embryonic neocortex, neuronal precursors are generated in the ventricular zone (VZ) (Kriegstein and Alvarez-Buylla, 2009; Noctor et al., 2001; Miyata et al., 2001; Tamamaki et al., 2001). As the neocortex grows, the neurogenic intermediate progenitors (nIPs) (Kriegstein and Alvarez-Buylla, 2009) of principal excitatory neurons (Noctor et al., 2004; Miyata et al., 2004; Haubensak et al., 2004; Wu et al., 2005; Hansen et al., 2010) and GABAergic inhibitory neurons (Letinic et al., 2002; Inta et al., 2008) appear in the subventricular zone (SVZ). Unlike the neural precursors of the VZ, the nIPs are multipolar as well as bipolar, and they are not fixed to a given position as are, for example, radial glia, which are anchored by their radial processes. Therefore, nIPs may be able to migrate within the neocortex and the telencephalon (Tabata et al., 2009).

The nIPs of the principal excitatory neurons are characterized by their expression of proneural transcriptional factors, such as Ngn2 (Neurog2 – Mouse Genome Informatics) (Miyata et al., 2004), Nex/Math2/Neurod6 (Wu et al., 2005) and Tbr2 (Eomes – Mouse Genome Informatics) (Englund et al., 2005); i.e. they express some markers that are often used to identify neuronal precursors, yet they continue to proliferate. By contrast, the intermediate progenitors of GABAergic neurons (IPGNs) have been identified as proliferating cells, which are positive for Mash1/Ascl1 immunoreactivity and can be labeled by GFP driven by a neuron-specific promoter (Letinic et al., 2002; Inta et al., 2008). Most characteristics of IPGNs, including their sites of origin, are still unclear.

Here, we have used the co-expression of the GABA-synthesizing enzyme GAD67 (Gad1 – Mouse Genome Informatics), a very early marker of the GABAergic fate (Yun et al., 2002), and cell-cycle markers to identify IPGNs in the perinatal mouse brain. We then employed a variety of approaches to investigate their gene expression profiles, origin, migration, fat, and possible roles in the mouse neocortex.

All mice were treated in accordance with the Regulations for Animal Care of Kumamoto University. To generate the GAD67-knock-in Cre mice, we designed a targeting vector in which the Cre recombinase gene was inserted into the initial methionine site of the Gad1 gene in frame. A knock-in vector, pGAD67CreTV, contained a 2.9 kb fragment at the 5′ side, a Cre gene just behind the GAD67 translational start, a Pgk-neo-p(A) cassette flanked by two Flp recognition target (frt) sites, a 6.8 kb fragment at the 3′ side and a MC1 promoter-driven diphtheria toxin gene (see Fig. S1A in the supplementary material). Linearized pGAD67CreTV was introduced into C57BL/6 mouse ES cells and the G418-resistant clones were picked up. Positive clones were identified by Southern blot hybridization or PCR. The sequence of these primers is provided in Table S1 in the supplementary material. The TVA Cre-reporter TG mouse was generated by the pronuclear injection of linearized DNA containing CApromoter-loxP-stop-loxP-tvA-stop-polyA (see Fig. S2 in the supplementary material). All the embryos were obtained by timed mating as described in the text. The morning after mating was considered embryonic day 0 (E0). The day of birth was E19 or postnatal day 0 (P0).

Immunohistochemistry and immunocytochemistry were performed as described previously (Tamamaki et al., 2003a; Wu et al., 2005). The antibodies used in this experiment are anti-βIII-tubulin (mouse Tuj1, Babco 1/500), anti-BrdU (mouse, Becton Dickinson, 1/1000; rat, Abcam, 1/1000), anti-Cre (rabbit, Novagen, 1/5000), anti-cyclin D1 (rabbit, Anaspec, 1/200), anti-Dcx (guinea pig, Chemicon, 1/1000), anti-Delta4 (rabbit, Rockland, 20 μg/ml), anti-Dlx2 (rabbit, a gift from Dr J. L. R. Rubenstein, UCSF, CA, USA, 1/200), anti-GABA (rabbit, Sigma, 1/5000), anti-GAD67 (mouse, Chemicon, 1:500), anti-GFAP (rabbit, Dako, 1/2000), anti-GFP [rabbit and guinea pig (Tamamaki et al., 2000)], anti-Ki-67 (mouse, Becton Dickinson B56, 1/50; rabbit, Epitomics, 1/200), anti-Map2 (mouse, Sigma HM2, 1/250-1/500), anti-Mash1/Ascl1 (mouse, PharMingen, 1/500), anti-NCAM (rabbit, Chemicon, 1/200), anti-nestin (mouse, Becton Dickinson, 1/500), anti-Notch3 [goat, Santa Cruz (M-20), 1/50], anti-PCNA (mouse, Novocastra, 1/100), anti-phosphorylated H3 (rabbit, Upstate, 1/1000), anti-Sox2 (rabbit, Chemicon, 1/1000), anti-TVA (rabbit, a kind gift from Dr A. D. Leavitt, UCSF, CA, USA).

The methods for single-cell RT-PCR and single-cell microarray analysis have been described previously (Esumi et al., 2008). PCR primers were designed using the GenBank sequence database for each target gene and are listed in Table S1 in the supplementary material. Microarray data are available at Array Express (Experiment name: Gene expression profiles of intermediate progenitors of GABAergic neuron; ArrayExpress accession, E-MTAB-452)

The caudal two-thirds of the neocortex at E18 was dissected (Fig. 1A) and dissociated into single cells. The GAD67-GFP-positive cells were isolated by sorting using FACSVantage (Becton Dickinson). The sorted cells (5×10⁴-10⁵ cells) were cultured for 2-7 days. The culture medium was Neurobasal medium conditioned with E17-18 embryonic brain tissue, followed by the addition of 5% B27 supplement, 20 ng/ml bFGF, 20 ng/ml EGF and 5 μg/ml BrdU in the presence or absence of arabinosylcytosine (Ara-C) (0.5 μM). For dot-blots, DNA was purified from the cells and applied to a nitrocellulose membrane (5 ng/dot). After denaturing the DNA, the membrane was incubated with an anti-BrdU antibody. The BrdU-immunoreactivity was detected with an ECL kit (Amersham).

Coronal brain slices (100 μm) were placed in the conditioned medium used for the cell culture. In some slices, nuclear division was monitored by the red fluorescence of a histone-H2B-mCherry fusion protein (Kanda et al., 1998). The fluorescent images were captured at 5-minute intervals for 8 hours (Wu et al., 2005).

Plasmids, CApromoter(1.7 kb)-loxP-mRFP(0.7 kb)-stop-loxP-GFP(0.7 kb)-stop-poylA(0.6 kb), CApromoter-loxP-mRFP-stop-loxP-FucciGreen(1 kb; MBL, co LTD, Japan)-stop-poylA, CApromoter-loxP-mRFP-stop-loxP-CFP(0.7 kb)-stop-poylA or CApromoter-FucciGreen-stop-poylA (1 μg/μl) were colored with Indian ink and injected into the third ventricle of E14 GAD67-Cre knock-in mouse embryos or E14 Nkx2.1-Cre mouse embryos (Fig. 6A). Next, we placed a needle electrode at one side of the orbit of the eye and the other, flat electrode on the top of the head of the embryo. By passing the electric current (40 V, 5 Hz square wave, 2 seconds) from the orbit to the parietal brain, the DNA was introduced into VZ cells of the third ventricle before the colored DNA solution could seep into the lateral ventricle. Accordingly, one day after the transduction, mRFP red fluorescence was seen almost exclusively in the wall of the third ventricle. The labeling was seen occasionally in the medial ganglionic eminence and rarely in the lateral ganglionic eminence. No mRFP fluorescence was seen in the neocortical VZ 1 day after the electroporation. To transduce the neocortical VZ with the GFP Cre-reporter plasmid, we injected the DNA solution into the lateral ventricle, and placed a needle electrode on the surface of the neocortex and the other, flat electrode, on the opposite side of the head of the embryo, at the temporal surface. The current was injected from the needle electrode (n=10). Two or 4 days after the electroporation, the embryos were recovered and fixed in the buffered fixative containing 4% paraformaldehyde, and the head of each embryo was cut into serial sections using a cryostat. The GFP, mRFP, CFP, Fucci-Green and Hoechst fluorescence was observed under a fluorescence microscope. Some sections were autoclaved and processed for immunohistochemistry.

RCAS-CMV-GFP recombinant virus was grown in chick embryonic fibroblast cells, harvested and concentrated (1×10¹⁰) by sedimentation in an ultracentrifuge (50,000 g). The virus solution was colored with Indian ink. E15-E18 embryos and P0 newborn mice were obtained by mating the TVA Cre-reporter mice and GAD67- or Nkx2.1-Cre-mice. We penetrated the skulls of the embryos caudal to bregma (the frontal suture) with a glass capillary and maintained an angle tangential to the dorsal neocortical surface pointing rostrally. We then injected the virus solution. Four to 7 days later, the mouse brains were recovered, and the infected cells were visualized by GFP immunohistochemistry.

GFP-expressing cells in GAD67-GFP knock-in mice (Tamamaki et al., 2003a) were prevalent from the marginal zone (MZ) to the SVZ at E18 in the neocortex (Fig. 1A), and often formed clusters in the SVZ and intermediate zone (IZ). All of them were positive for GABA and GAD67 immunoreactivity (Fig. 2A-D). In addition, the vast majority of the GFP-positive cells in the VZ, SVZ and IZ were positive for MAP2, βIII-tubulin (Tuj1), Dlx2, Dcx and NCAM immunoreactivity (see Fig. S3 in the supplementary material). However, all of the GFP-positive cells in the E18 neocortex are not necessarily positive for these neuronal markers. Some of them may be precursors of GABAergic neurons.

We next used immunohistochemistry and immunocytochemistry to evaluate the colocalization of GFP with cell-cycle indicators. First, we detected cells in the DNA synthetic phase (S phase) using BrdU labeling. Thirty minutes after BrdU injection into the peritoneal cavity of pregnant mice, BrdU immunoreactivity was detected in the E18 mouse neocortex (Fig. 1B, Fig. 2E). (To ease the quantification of GFP/BrdU-double-labeled cells, we counterstained the nuclei with propidium iodide.) The cells that were double-labeled by GFP and BrdU accounted for only 1.5% (4/263 in the IZ and 7/495 in the SVZ) of the GFP-positive cells, but represented 5% of the BrdU-positive cells in the IZ, SVZ and VZ. We also identified GFP-positive cells in mitosis (M phase) in the SVZ by their chromosomal condensation (Fig. 1C).

Immunolabeling for GFP and the proliferation marker Ki-67 (Mki67 – Mouse Genome Informatics) revealed that double-positive cells (Fig. 1D, Fig. 2F) accounted for 22% of the GFP-positive cells in the IZ (75/337) and 29% in the SVZ (96/335) of the dorsal neocortex. Double labeling for GFP and the M-phase marker phosphorylated histone H3 (P-H3; Fig. 2G; see Fig. S4A in the supplementary material) was also seen, starting at E16. In sections immunolabeled for PCNA and GFP, all the nuclei in some clusters of GFP-positive cells were double-labeled (see Fig. S4B in the supplementary material). Some GFP-positive cells were positive for nestin (Fig. 2H). Thus, the cell-cycle markers Ki-67, PCNA, BrdU, P-H3 and a neurogenic marker (nestin) were colocalized with GABA and GAD67 in the GAD67-GFP mice (see Table S2 and Fig. S4B in the supplementary material). To verify that the co-localization of these markers was not unique to the GAD67-GFP knock-in mice, we also immunolabeled the wild-type murine neocortex, and found that the Ki-67 immunoreactivity and GAD67 immunoreactivity were also colocalized in dissociated cells (data not shown). The marker colocalization occurred only after E16.

We also considered whether these co-labeled cells might be neurons in apoptosis, which are known to re-enter the cell cycle, express cell-cycle markers and incorporate BrdU, although they never enter M-phase (Herrup and Yang, 2007). However, chromosomal condensation with P-H3 immunoreactivity is detectable only in M-phase, and, therefore, our immunolabeling results showing these two features indicated that at least some of the GAD67-GFP-positive cells in the neocortex at E18 are IPGNs.

To rule out the possibility that our immunolabeling experiments were detecting Ki-67 immunoreactivity from previous cell-cycle events, we further verified our results by using single-cell RT-PCR to detect Ki-67 mRNA in GFP-positive cells. GFP-positive cells in the SVZ of the E18 neocortex were dissected (Fig. 2A,B), dissociated into single cells, identified by their GFP fluorescence and picked up one by one by aspiration through a glass capillary pipette (Esumi et al., 2008). The total RNA obtained from a single cell was converted into cDNA and amplified by nested PCR. To detect only cDNA, the PCR primers were designed to span more than one intron (see Table S1 in the supplementary material). Almost all of the 54 β-actin-positive cells expressed Gad1 mRNA, and a significant fraction (10/54) also expressed Mki67 mRNA (Fig. 3A; see Table S2 in the supplementary material).

Single-cell RT-PCR is useful for detecting the mRNAs co-expressed in a single cell. However, the number of genes detectable by this method in a single cell is, at best, limited to 10. Therefore, we also profiled the gene expression by single GFP-positive cells using single-cell microarray analysis (Esumi et al., 2008). Although the array data included almost the same number of false-Absent calls as Present calls, we did not encounter any false-Present calls. Moreover, we increased the reliability of the array data by examining the expression of important genes by immunohistochemistry or single-cell RT-PCR using amplified cDNA samples as the template.

Of the neuronal markers, we noted that Tα1 tubulin (Tuba1 – Mouse Genome Informatics) had Present calls in 10 out of 10 cells, and doublecortin (Dcx) in nine cells (Fig. 3B). Gad1 cDNA was also detected in all 10 cells by single-cell RT-PCR and by single-cell quantitative RT-PCR (Esumi et al., 2008), confirming that these cells were neurons or neuronal precursors. Some had differentiated into a GABAergic neuronal subtype, as indicated by Present calls for Npy and Reln (Fig. 3B). We also noted that many genes related to DNA replication had Present calls in a number of cells (Fig. 3C; see Table S3 in the supplementary material). In particular, cell #2 had Present calls for Pcna, cyclin D2 (Ccnd2 – Mouse Genome Informatics) (Glickstein et al., 2009), cyclin E1 (Ccne1) and cyclin E2 (Ccne2), and cell #4 had Present calls for cyclin D1 (Ccnd1) and cyclin D2 (Fig. 3D). Cell #2 also had a Marginal call for Mki67, which was expressed predominantly in the G1-S phase. To confirm the appearance of cyclin D1 in the GAD67-GFP-positive cells, we performed immunohistochemistry (see Fig. S5 in the supplementary material). From the gene-expression profiling, we speculated that cells #2 and #4 were in S phase and G1 phase, respectively.

The single-cell gene-expression analyses also confirmed that the GAD67 and cell-cycle markers are co-expressed in the GFP-positive neuron precursors. In addition, we detected Dcx expression in cell #4 (Fig. 3B). Since Dcx is an essential gene expressed by precursors of excitatory neurons in the SVZ (Bai et al., 2003), nIPs migrating from the VZ to the SVZ may express Dcx, and maybe also IPGNs. Single-cell RT-PCR also showed that GFP-positive cells expressing Mki67 and Gad1 mRNAs often expressed the proneural gene Mash1/Ascl1 (see Fig. S6 in the supplementary material). All the listed cells in the array analysis had Absent calls for the major glial markers (Gfap, Cnp, Pdgfra, Plp1).

To confirm further that the cell-cycle phases represented cell division and not apoptosis, we examined the expression of E2f1, a suppressor of cell-cycle entry (Wang et al., 2007). E2f1 had a Present call in cell #10, which showed fewer Present calls for DNA-replication genes (Fig. 3D). In addition, we found no GFP-positive cells that co-expressed E2f1 and Mki67, whereas one GFP-positive cell co-expressed delta-like 1 (Dll1) and E2f1 (see Fig. S7 in the supplementary material), which may have been in the process of terminating cell division and differentiating into a neuron. In addition to Dll1, the array data indicated that genes related to asymmetric cell division, such as Notch3, delta-like 4 (Dll4) and Numb were also expressed by the GFP-positive cells. Immunohistochemistry also revealed Notch3 immunoreactivity and Dll4 immunoreactivity in one of a pair of GAD67-GFP-positive cells in close contact (Kageyama et al., 2009; Sprinzak et al., 2010) (see Figs S8, S9 in the supplementary material). The expression of these molecules is consistent with the idea that an IPGN divides in the neocortex to give rise a GABAergic neuron and to renew itself (see also Discussion).

To confirm that some of the GFP-positive cells in the GAD67-GFP mouse were IPGNs, we collected the GFP-positive cells with a fluorescence activated cell sorter (FACS) in culture medium (Fig. 4A-D). We tested several different culture conditions for growing the sorted cells and found that a medium for neural stem cells (Tropepe at al., 1999) did not induce proliferation (Fig. 4E). The IPGNs proliferated in a medium conditioned with embryo telencephalic tissue, which provided unknown growth factors. The ability of the cells to incorporate BrdU in culture medium was sensitive to AraC (Fig. 4E), and, after 2 days in culture, a small number of GFP-positive cells were positive for BrdU (Fig. 4F). Most of the BrdU-positive cells were found in pairs. After 5 days in culture, the number of BrdU-positive cells had not increased, but they extended Tuj1-positive dendrites and axons (Fig. 4G).

We also recorded the division of GAD67-GFP-positive cells in brain slices by video microscopy (Fig. 4H; see Movies 1, 2 in the supplementary material). The cells were observed for up to 8 hours, during which scattered GFP-positive cells divided. The two daughter cells remained GFP-positive during our observation period. We also monitored their nuclear division by observing the red fluorescence of a histone-H2B-mCherry fusion protein (Kanda et al., 1998). After every video recording, we performed immunohistochemistry and confirmed the presence of P-H3 immunoreactivity and BrdU in the GFP-positive cells, demonstrating that BrdU had been incorporated from the culture medium into the nucleus, that the GFP-positive cells had divided, and therefore that the IPGNs in the brain slices proliferated in vitro (Fig. 4I,J).

Next, to prove the proliferation of the IPGNs and to trace the fate of the progeny of IPGNs in vivo, we produced GAD67-Cre knock-in mice (see Fig. S1 in the supplementary material) and crossed them with TVA Cre-reporter mice [which express TVA, an avian sarcoma leukosis virus (ASLV) receptor] (see Fig. S2 in the supplementary material). We used the resulting GAD67-Cre/tv-A progeny in the following experiments, in which GAD67-expressing cells were transduced by infection of replication-incompetent GFP-expressing simple retrovirus, a recombinant ASLV (Hughes et al., 1987).

A small amount of ASLV solution was injected into the parenchyma of the neocortex of 20 E15-P0 GAD67-Cre/tv-A mice. Proliferating GAD67-positive cells (i.e. IPGNs) were susceptible to retroviral infection, and they (Fig. 5A) and/or their descendents could be identified by GFP immunohistochemistry in the injected brains, which were recovered 4-7 days later. Their descendents included non-pyramidal neurons (five cells) in the neocortex (Fig. 5B), many migrating cells in the rostral migratory stream (RMS) (124 cells) (Fig. 5C), and granule (141 cells) and periglomerular (24 cells) cells in the olfactory bulb (Fig. 5D,E). From the distribution of labeled IPGNs and migrating cells in the RMS, we could roughly speculate the injection site in the neocortex (Fig. 5A, inset).

When virus was injected into the lateral ventricle, no labeled cells were seen; nor were any cells labeled in the brain of wild-type mice. However, the level of GFP expression from the recombinant ASLV DNA was not high enough for us to perform a subtype analysis of the nonpyramidal neurons in the neocortex or of the labeled cells in the olfactory bulb. We tried repeating this experiment in five Nkx2.1-Cre mice (Kessaris et al., 2006) mated with TVA Cre-reporter mouse and found two nonpyramidal neurons in the neocortex, many migrating cells in the RMS (224 cells), as well as granule cells (120 cells) and periglomerular cells (35 cells) in the olfactory bulb.

Next, we examined the origin of the IPGNs in the embryonic neocortex using three different approaches. First, we introduced a GFP Cre-reporter plasmid with a floxed modified red fluorescent protein (mRFP)-stop into various parts of the telencephalic VZ of E14 GAD67-Cre mouse embryos by electroporation (Fig. 6A, plasmid 1). Depending on the placement of the electrodes, the transduced domain, which was identified by the mRFP fluorescence in the VZ, could be shifted along the ventral-to-dorsal axis (black arrows in Fig. 6B-D,G). When the electroporation site included the VZ of the ganglionic eminence (GE), the medial GE (MGE) could be distinguished from the rest of the structure, by its green fluorescence (Fig. 6B,C), whereas the lateral GE (LGE) emitted intense red fluorescence throughout the observation period. In fact, the transduced cells in the LGE were positive for both GFP and mRFP. However, the cells in the SVZ of the MGE may have a higher GAD67 promoter activity than those in the LGE (Tamamaki et al., 2003a), so the Cre recombinase in the MGE may have deleted more floxed mRFP cDNA than in the LGE. Thus, we assumed that the green cells (white arrows in Fig. 6C,D) originated in the MGE and the red cells in the LGE (Fig. 6D, inset). Four days after the electroporation, many green and red cells were observed in the dorsal neocortex and the hippocampus (Fig. 6D).

Next, we autoclaved the sections to eliminate the mRFP fluorescence and detected GFP and Ki-67 by immunohistochemistry. GFP-positive cells in the neocortical SVZ were often positive for Ki-67 immunoreactivity (7/69) (Fig. 6E,F). As these GFP-positive cells were in the GAD67-positive cell lineage, we regard these GFP/Ki-67 double-positive cells in the SVZ as IPGNs. This experiment indicates that IPGNs in the neocortical SVZ at P0 originate in the MGE and LGE. When we introduced the GFP Cre-reporter plasmid into the VZ of the dorsal neocortex at E14 and examined the brain at E18, no green cells were seen in the neocortex (Fig. 6G,H).

Next, we repeated the experiment described above, but used a plasmid in which fluorescent ubiquitylation-based cell cycle indicator-green (Fucci-G) cDNA (Sakaue-Sawano et al., 2008) was used instead of GFP (Fig. 6A, plasmid 2), and another in which the GFP cDNA was replaced with CFP cDNA (Fig. 6A, plasmid 3). Cre-positive cells in the GE delete the floxed mRFP-stop cassette and transcribe the Fucci-G and CFP cDNAs into mRNAs. As Fucci-G cDNA encodes a modified AzamiGreen (mAG) fused with a Geminin fragment, a cell in S-G2-M phase accumulates Fucci-G protein in the nuclei (Fig. 6J,K), but one in the G1 and G0 phases removes Fucci-G by degradation and appears as an mRFP- and CFP-double-positive cell (Fig. 6K). After the transduction of the VZ cells in the GE with the Fucci-G and CFP-Cre-reporter plasmids, cells triple-labeled for Fucci-G, mRFP and CFP were found in the GE, neocortex and hippocampus (see Fig. S10 in the supplementary material). mRFP-labeled cells lined up the cell migration stream in the LGE (arrows in K) (Fig. 6K). About 10-20% of the mRFP-positive cells were positive for Fucci-G at E18 (Fig. 6J). These Fucci-G-positive cells were also positive for Dcx immunoreactivity (Fig. 6L). As the Fucci-G-positive cells belonged to the GAD67-positive cell lineage and were in the S-G2-M phases, we regarded them as IPGNs.

When the plasmids for the CFP Cre-reporter and Fucci-G expression (Fig. 6A, plasmid 3 and 4) were introduced into the LGE and MGE in Nkx2.1-Cre mouse embryos at E14, we found cells that were triple-labeled for mRFP, Fucci-G and CFP, and cells that were double-labeled for mRFP and Fucci-G in the neocortical SVZ at E16-E18 (Fig. 6M,N). We speculated that the former cells expressed Cre and originated in the MGE. The latter did not express Cre but express Fucci-G in the LGE. As both of cells were positive for Fucci-G and Dcx immunoreactivity, we regarded them as IPGNs.

Third, we mapped the contributions of different VZ regions to the production of IPGNs in the neocortex by using a battery of transgenic mice that express Cre in defined neuroepithelial domains: Nkx2.1-Cre (MGE) (Kessaris et al., 2006) and Gsh2-Cre (MGE and LGE) (Kessaris et al., 2006), and Emx1-Cre (cerebral cortex) (Iwasato et al., 2000). These mice were crossed to a GFP Cre-reporter (Novak et al., 2000) to label the progeny originating from each Cre-expressing domain. Next, we immunostained sections for GABA immunoreactivity and Ki-67 immunoreactivity on the GFP-positive cells in the neocortex of all three genotypes of mice at P0. In the Nkx2.1-Cre/GFP and Gsh2-Cre/GFP mice, GABA immunoreactivity was frequently seen on GFP-positive cells in the neocortical SVZ and IZ. In the NKx2.1-Cre/GFP mice, the percentages of double-labeled cells among GFP-positive and GABA-positive cells in the SVZ and IZ were 24.4% and 16.4%, respectively (Fig. 7A,H). In the Gsh2-Cre/GFP mice, these percentages were 19.3% and 54.4%, respectively (Fig. 7B,G). In the dorsal neocortical SVZ of Emx1-Cre/GFP mice, we could not clearly detect any double-labeled cells in above-background fluorescence, although some may have migrated into the dorsal neocortex from the pallial-subpallial boundary (B) (Kohwi et al., 2007). As about 20% of the GABA-positive cells in the neocortical SVZ at P0 are IPGNs (see above), we examined whether Ki-67 immunoreactivity and GABA immunoreactivity were colocalized in GFP-positive cells in the Nkx2.1-Cre/GFP and Gsh2-Cre/GFP mouse neocortex. Our results revealed many triple-labeled cells (Fig. 7C-F). The results indicate that IPGNs originate in LGE or MGE, or both. Based on the data obtained from these three different experiments, we conclude that IPGNs in the neocortical SVZ originate in the MGE and LGE.

To follow the fate of the IPGNs in the neocortical SVZ, we monitored them for up to 6 weeks after birth in the GAD67-GFP knock-in, GAD67-Cre/GFP, Nkx2.1-Cre/GFP, Gsh2-Cre/GFP and Emx1-Cre/GFP mice. As the GAD67-GFP mice grew, the GFP-positive cells in the neocortical SVZ gathered and formed cell clusters, which appeared as cell chains in successive sections and were negative for NeuN immunoreactivity (Fig. 8A). The GFP-positive cells in the cell chains were also positive for Dcx (see Fig. S11 in the supplementary material), GABA and GAD67 immunoreactivity (Tamamaki et al., 2003a). In addition, they were surrounded by GFAP-positive processes, as shown previously in the SVZ of the lateral wall of the lateral ventricle (Alvarez-Buylla and Garcia-Verdugo, 2002) (Fig. 8B). These observations were reminiscent of migrating young neurons (A cells) in the RMS. However, some of the GFP-positive cells in the chains were positive for Ki-67 immunoreactivity (Fig. 8C), which corresponds to the IPGNs characterized here. Therefore, Ki-67 immunoreactivity is co-localized with Dcx immunoreactivity, even in the wild-type mouse brain (see Fig. S11 in the supplementary material). These Ki-67-positive SVZ cells in the adult neocortex were reminiscent of transit-amplifying cells (C cells) (Doetsch et al., 2002). In fact, a C-cell marker, Mash1/Ascl1 immunoreactivity (Adachi et al., 2007), was often in close contact with GFP immunoreactivity in the GAD67-GFP mouse (see Fig. S12 in the supplementary material) (see Discussion). Moreover, their mRNA colocalized in the IPGNs (see Fig. S6 in the supplementary material).

To reveal the origin of the cells in the RMS in the dorsal neocortex, we examined GFP-positive cells in the adult neocortex of the Nkx2.1-Cre/GFP, Gsh2-Cre/GFP, Emx1-Cre/GFP and GAD67-Cre/GFP mice. We labeled brain sections for Dcx immunoreactivity and counted the number of Dcx/GFP-double-labeled cells (A cell + IPGNs) in the cell chains above the hippocampus (Fig. 8D-F). We found no Dcx immunoreactivity on the GFP-positive cells in the neocortical SVZ of the Nkx2.1-Cre/GFP mouse, (0/343) (Fig. 8D), possibly because the IPGNs in the neocortical SVZ, which originated in the MGE and produced granule cells for the olfactory bulb at perinatal stage (Fig. 5), may be exhausted at 6 weeks of age. In the Gsh2-Cre/GFP mice, the GFP-positive cells in the neocortical SVZ were small bipolar and often aligned in cell chains. The GFP-positive cells made up 45% (146/324) of the Dcx-positive cell in the dorsal-neocortex (Fig. 8E). In the Emx1-Cre/GFP mice, GFP-positive cells in the SVZ made up 19% (40/210) of the Dcx-positive cells in the dorsal neocortex (Fig. 8F). Although we could not determine the origin of 34% of the Dcx-positive cells in the dorsal neocortex, it is clear that a significant number of neurogenic IPGNs arrive in the dorsal neocortex from the MGE, LGE and maybe also from pallial-subpallial boundary (Emx1 cell-lineage), as well as additional unknown areas of the telencephalon, to produce the young neurons populating the RMS (Kriegstein and Alvarez-Buylla, 2009; Kohwi et al., 2007; Seri et al., 2006; Young et al., 2007; Lledo et al., 2008).

Besides migrating in the telencephalon and producing neocortical GABAergic neurons and olfactory granule cells, IPGN-like GFP-positive cells were often found along the neocortical SVZ of GAD67-Cre mice (Fig. 8G; compare with Fig. 5A). These IPGN-like cells often showed SOX2 immunoreactivity (inset in Fig. 8G). These Sox2-positive cells (Pevny and Nicolis, 2010) may be IPGNs that have settled in the neocortex and are reserved for neurogenesis in the case of brain damage or other pathology (Ohira et al., 2010). A schematic diagram summarizes our findings on the origin, differentiation and migration of IPGNs, and their progenies (Fig. 8H).

It is well known that neocortical GABAergic neurons originate in the subpallium of rodents telencephalon (Anderson et al., 1997; Tamamaki et al., 1997), specifically in the MGE (Lavdas et al., 1999), LGE (Fogarty et al., 2007) and caudal GE (Nery et al., 2002), and are supplied by tangential migration to the pallium during embryonic stages. The supply of GABAergic neurons to the neocortex ends as the mouse grows and the distance between these structures increases. Under normal conditions, most of the brain regions may not require new neurons during the adult life of an animal. However, the ratio of GABAergic neurons in the adult neocortex is fixed at about 15-25%, depending on the brain area (Jones 1993; Tamamaki et al., 2003a), and a pathological change in the brain could cause disproportionate damage to the GABAergic versus excitatory neurons. If there were no way to add GABAergic neurons to the neocortex, brain function could be easily compromised. Therefore, although IPGNs may not be needed to populate the neocortex with GABAergic neurons during development, they may be required to adjust the ratio of GABAergic neurons by Notch-Delta-regulated proliferation in embryo brain and in adult brain under pathological conditions.

A few studies have reported the production of GABAergic neurons in the adult cerebral cortex (Liu et al., 2003; Ohira et al., 2010), but the origins of the progenitors for these neurons has been completely uncertain. The IPGNs described here are a likely source of GABAergic neuronal production in the adult cerebral cortex, even though they supply only a small number of GABAergic neurons to the neocortex of the perinatal brain, and primarily generate a large number of granule cells and periglomerular cells for the neonatal olfactory bulb. IPGNs in the adult brain have not been described, although labeling and tracing studies that should have marked them have been performed (Inta et al., 2008; Suzuki and Goldman, 2003). It is possible that their numbers are so low that IPGNs have not been recognized as a separate subpopulation or they may have been mis-identified as astrocytes, given their appearance (Fig. 5A).

Here, we have revealed that IPGNs in the neocortex originate in the subpallium (Figs 6 and 7), and we did not detect any GAD67-promoter-driven Cre expression in the VZ of the dorsal neocortex in E14-P0 mice (Fig. 6G,H). Our results support the idea that the VZ of the cerebral cortex cannot produce GABAergic neurons (Gorski et al., 2002). However, IPGNs also seemed to be produced at the pallial-subpallial boundary. Dcx-positive cells in the RMS of the dorsal neocortex comprised up to 19% of the GFP-positive cells in Emx1-Cre/GFP mouse and comprised completely GFP-positive cells in GAD67-GFP mouse (Fig. 8F; see Fig. S11 in the supplementary material). As the pallial-subpallial boundary is characterized by the expression of Dlx5/6 and Emx1 (Kohwi et al., 2007), IPGNs from the boundary may be in the Emx1 lineage but produce GABAergic young neurons for the olfactory bulb. Therefore, the IPGNs from the boundary may give rise the RMS cells in Emx1 lineage.

As IPGNs are not anchored by any radial fibers, IPGNs may be able to migrate randomly (Tabata et al., 2009) or one way when cued to do so and make a long journey even to the entire part of the cerebral cortex. The journey of the IPGNs from the pallium to the neocortex may be guided by the molecular mechanisms used by GABAergic young neurons (Marin et al., 2001; Tamamaki et al., 2003b).

Several cell migratory streams run in the embryonic brain and populate the various brain regions with many different types of neurons. Here, we examined a cell migratory stream extending from the GE to the neocortex and found IPGNs migrating in this stream. We speculate that the presence of neurogenic intermediate progenitors in this migratory stream is not unique, but just the first such example to be described.

In this study, we revealed characteristics of IPGNs that are different from those of neural stem cells. The generation of neurons from neural stem cells involves many differentiation steps. Previously, all of the differentiation steps were thought to occur in daughter cells arising from neural stem cells in the embryonic brain. However, nIPs (Kriegstein and Alvarez-Buylla, 2009) distribute the sequential differentiation steps among the cells in their lineage (Noctor et al., 2004; Miyata et al., 2004; Haubensak et al., 2004; Wu et al., 2005; Hansen et al., 2010; Letinic et al., 2002). The differentiation steps involve the expression of transcription factors for neuronal differentiation, of cytoskeletal and cell-adhesion molecules specific to neurons, and of molecules essential for neuron-specific function. During this process, some molecules are expressed in a phasic manner or gradually plateau. As none of the molecules show abrupt changes in expression, they are likely to overlap temporally and be colocalized in differentiating cells and asymmetrically dividing cells, even though their primary functions may be in distinct differentiation steps.

Recent studies show that the regulation of molecular expression during asymmetric cell division, such as Notch signaling, is more complex and dynamic than previously thought (Kageyama et al., 2009; Sprinzak et al., 2010), which helps explain why Notch and Delta are co-expressed in neuronal progenitors and daughter cells, and supports our observation that Notch3-positive cells in the neocortical SVZ at P0 also express Dll1 and Dll4 (Fig. 3E). Moreover, Mash1/Ascl1 immunoreactivity is used as a marker of C cells (Doetsch et al., 2002; Adachi et al., 2007). When a C cell divides and produces a C cell and an A cell for the olfactory bulb, and when markers of A cells are expressed in advance of the asymmetric cell division by leakage, the dividing C cells may contain the Mash1/Ascl1, Ki-67 and Gad1 mRNAs, as we detected by single-cell RT-PCR in the GFP-positive cells (see Fig. S6 in the supplementary material). Moreover, GABA may not be solely a transmitter stored in synaptic vesicles at the perinatal stages but might function as a regulator of cell proliferation (LoTurco et al., 1995; Haydar et al., 2000). In addition, when IPGNs preparing for asymmetric cell division are arrested in the S-G2-M phases for cell migration (Fig. 6J,K), cell-cycle markers and Dcx could accumulate in the same cell (Fig. 3B #4 cell).

The neural stem cells in the SVZ of the adult telencephalon (B cells) are the source of GABAergic granule cells and periglomerular cells in the olfactory bulb, and nIPs (C cells) assist to expand the neuron production. As the SVZ of the ventral wall of the lateral ventricle is completely occupied by B cells and C cells, some C cells may migrate from the subpallium to the dorsal neocortex to find an appropriate niche for neuron production. Thus, the IPGNs in the neocortex may be C cells that have accumulated GAD67 and Dcx during migration. Although we found IPGNs co-expressing Gad1, Dcx and cell-cycle markers, GAD67 and Dcx may not be unique; rather, different kinds of nIPs may accumulate different sets of molecules during migration (Luskin, 1998). In any case, we expect that further study of the IPGNs will show that neurogenesis in the embryonic stages and adult is, at least in part, a continuous event.

We thank Drs N. Kessaris, W. D. Richardson, T. Iwasato and C. G. Lobe for their donation of the Nkx2.1-Cre, Gsh2-Cre, Emx1-Cre and Z/G Cre-reporter mice. We also thank Drs T. Kanda, A. D. Leavitt, J. L. R. Rubenstein and R. Tsien for their donation of antibodies and cDNAs. This study was supported by a Grant-in-Aid for Scientific Research on Priority Area-Advanced Brain Science Project from the Ministry of Education, Culture, Sports, Science and Technology, Japan to N.T., and by a COE grant in IMEG in Kumamoto University and National Natural Science Foundation of China (No. 30970946) to S.W. The funding bodies had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.