308 nature neuroscience ��� volume 5 no 4 ��� april 2002 articles Cell diversity in the vertebrate nervous system is generated through a developmentally regulated cell fate restriction1,2. In the cerebral cortex, precursor cells are multipotent at early stages of development3 and become specified to generate a single cell type at later stages4,5. These distinct lineages have been observed under isolated in vitro conditions3,6���8 and show characteristic stereotyped lineage trees8, implying that cell-intrinsic determi- nants act as fate-restrictive cues. The transcription factor Pax6, which is expressed during neu- rogenesis in the cerebral cortex and is crucial for patterning the telencephalon9���11, is a possible intrinsic fate determinant. Pax6 is detected specifically in radial glial cells. In Small-eye (Sey) mice, which lack functional Pax6, these cells have a distorted mor- phology as well as altered gene expression patterns and cell cycle characteristics10,12. As radial glial cells of the cerebral cortex gen- erate neurons in vitro and in vivo13���16, we hypothesized that Pax6 might be involved in the neurogenic potential of radial glial cells. We used loss- and gain-of-function approaches to address this question. We show here that the neurogenic progeny of radial glial cells was reduced in the Pax6 mutant cortex, whereas Pax6 transduction enhanced the neuronal lineage and instructed even astrocytes towards neurogenesis. RESULTS Reduced neurogenic radial glia in Pax6 mutant To determine the progeny of radial glial cells in the Pax6 loss- of-function condition, we isolated radial glial cells by fluores- cence-activated cell sorting (FACS) using green fluorescent protein (GFP) expression driven by the human GFAP pro- Glial cells generate neurons: the role of the transcription factor Pax6 Nico Heins1, Paolo Malatesta1, Francesco Cecconi2, Masato Nakafuku3, Kerry Lee Tucker4, Michael A. Hack1, Prisca Chapouton1, Yves-Alain Barde2 and Magdalena G��tz1 1 Max-Planck Institute of Neurobiology, Am Klopferspitz 18a, 82152, Planegg-Martinsreid, Munich, Germany 2 Dipartimento di Biologia, Universita degli studi ���Tor Vergata���, Via della Ricerca Scientifica, 00133 Rome, Italy 3 University of Tokyo, Graduate School of Medicine, 7-3-1 Hongo, Bunkyoku, Tokyo, 113-0033 Japan 4 Friedrich Miescher Institute for Biomedical Research, Maulsbeerstr. 66, 4058 Basel, Switzerland P.M. and N.H. contributed equally to this work Correspondence should be addressed to M.G. (mgoetz@neuro.mpg.de) Published online: 18 March 2002, DOI: 10.1038/nn828 Radial glial cells, ubiquitous throughout the developing CNS, guide radially migrating neurons and are the precursors of astrocytes. Recent evidence indicates that radial glial cells also generate neurons in the developing cerebral cortex. Here we investigated the role of the transcription factor Pax6 expressed in cortical radial glia. We showed that radial glial cells isolated from the cortex of Pax6 mutant mice have a reduced neurogenic potential, whereas the neurogenic potential of non- radial glial precursors is not affected. Consistent with defects in only one neurogenic lineage, the number of neurons in the Pax6 mutant cortex in vivo is reduced by half. Conversely, retrovirally mediated Pax6 expression instructs neurogenesis even in astrocytes from postnatal cortex in vitro. These results demonstrated an important role of Pax6 as intrinsic fate determinant of the neurogenic potential of glial cells. moter13,17. The hGFAP���GFP transgene was crossed into the Pax6 mutant background using Sey mice that express a trun- cated, non-functional form of Pax6 (ref. 18). GFP-positive cells from wild-type and homozygous Pax6 mutant (Sey/Sey) lit- termates were analyzed by FACS (Fig. 1a���d). Consistent with previous sorting results, most GFP-positive cells sorted from the cortex of embryonic day (E)14 wild-type and Sey/Sey mice were precursors immunoreactive for nestin and the radial glial antigens RC2 and GLAST (Fig. 1c,d, gray). In wild-type cortex, sorted radial glial cells were also Pax6 immunoreactive (78 �� 4% of sorted RC2-positive cells, n = 159, data not shown). Besides radial glial cells, a relatively high proportion of postmitotic neu- rons labeled with neuron-specific antiserum against ��-tubulin- III was included in the GFP-positive fraction (Fig. 1c,d 2 hours). This most likely was due to the remaining GFP signal in the young neuronal descendants of the sorted radial glia13. How- ever, GFP-positive cells were reduced in the Pax6 mutant mice compared to their wild-type littermates (Fig. 1a,b p 0.005, 4 litters). Thus, GFP is targeted to the same populatiozn of GLAST-positive radial glia in Pax6 mutant mice, but their pro- portion is lower than in wild-type mice. The progeny of the GFP-positive cells sorted from wild-type and Sey/Sey cortex was examined after 1 week in vitro by double staining with cell- type specific antisera as described previously13 (see also Fig. 1 legend). Pax6 mutant radial glia generated a significantly small- er number of pure neuronal clones within 1 week in vitro as compared to wild-type littermates (Fig. 1c,d blue fraction for example, Fig. 1e,g). Correspondingly, more non-neuronal clones were generated by the GFP-positive precursors from �� 2002 Nature Publishing Group http://neurosci.nature.com

articles nature neuroscience ��� volume 5 no 4 ��� april 2002 309 included in the sorted fraction. This allowed us to reconstruct the entire progenitor pool of wild-type and Pax6 mutant cor- tex composed of hGFAP���GFP-positive radial glia and hGFAP���GFP-negative, non-radial glia precursors (Fig. 2). One- third of all precursors were GFP negative in E14 wild-type cor- tex, and they generated neurons only (Fig. 2a, gray, hatched). A further third of wild-type cortex precursors were GFP positive and neurogenic (green, hatched), and the last third were GFP- positive and non-neurogenic (green and green with squares, Fig. 2a). In the Pax6 mutant progenitor pool, the GFP-nega- tive neurogenic population was still of the same size as in wild- type cortex (Fig. 2a,b, gray, hatched), whereas the GFP-positive neurogenic population was reduced (green, hatched, Fig. 2b). Thus, the GFP-negative fraction does not compensate for the reduced neurogenesis by the GFP-positive fraction. In addi- tion, the GFP-negative precursors of the Pax6 mutant cortex contained a non-neurogenic population generating only imma- ture, nestin-positive precursors that is not detectable in the wild type (white, Fig. 2b). Thus, the loss of Pax6 leads to an Fig. 1. Progeny of hGFAP���GFP-positive cells from wild- type (WT) and Pax6 mutant cortex. Quantification of the green fluorescent radial glial cells by FACS of the cortex of WT and Sey/Sey mice at embryonic day (E) 14, both carrying the transgene GFP under the human GFAP promoter17. (a,b) Left, dot plots of cells in for- ward scatter and side scatter polygon indicates the gate selecting the healthy cells. Right, fluorescence intensity versus number of events. Gray, non-transgenic negative control color, GFP-positive cells gray squares, sorting gate. The proportion of GFP-positive cells in the Sey/Sey cortex (b) is one-third that in WT cortex (a). (c,d) Quantification of clusters (identified as shown in e���h) of sorted cells 2 h or 7 div (days in vitro) after sort- ing from E14 WT (c) and Sey/Sey (d) cortex. Cellular composition of clusters was assessed by immunostain- ing with cell type-specific antisera. GFP-positive cells from Sey/Sey cortex generate few neurons (d, blue), increasing from 14% to 35% of clusters from 2 h (left) to 7 div (right), whereas the majority of radial glial cells from WT cortex generate neurons (c), increasing from 26% to 70% of clusters between 2 h and 7 div. Number of clusters examined, 487 (2h WT), 318 (7div WT), 632 (2h Sey/Sey), 474 (7div Sey/Sey) number of GFAP- immunostained clones, 77 (WT), 195 (Sey/Sey). Clones containing GFAP-positive cells constitute the same proportion of the non-neuronal clones generated by WT (16% of 26% or 62% of non-neuronal clones) or Pax6 mutant (38% of 62% or 60% of non-neuronal clones). Statistically significant (p 0.005) differences between WT and Sey/Sey are indicated by **. (e���h) Examples of clusters considered as clones. Sorted cells were plated at low density on rat cortical cells, allow- ing them to be identified by the mouse neural-specific antibodies M2 and M6 on the immuno-negative rat cor- tical cells. Distinct clusters of M2/M6-immunoreactive cells separated by at least 200 ��m were considered as clones derived from individual sorted precursor cells after 1 week in vitro (for methodology, see ref. 13). Corresponding fluorescence micrographs (e���h) show- ing examples of clones generated by GFP-positive sorted cells from WT (neuronal clone: e,g) and Sey/Sey (glial clone: f,h) E14 cortex. ��-tubulin-III immunoreac- tivity (green) labels neurons (g) and GFAP immunoreac- tivity (blue) labels astrocytes (h). Scale bar, 50 ��m. Sey/Sey cortex (Fig. 1c,d, red), containing the same relative proportion of GFAP-positive astrocyte clones (hatched, red examples shown in Fig. 1f,h). As the population of the hGFAP���GFP-positive cells was smaller in the Pax6 mutant background, the neurogenic pop- ulation might have merely lost the GFP signal and therefore be contained in the GFP-negative fraction. Indeed, the hGFAP���GFP-negative fraction from Sey/Sey cortex contained more precursor cells than did that from wild-type (wild type, 11 �� 4% nestin-positive cells of all GFP-negative sorted cells, n = 161 Sey/Sey, 20 �� 5%, n = 136). Relatively few cells in the GFP-negative fraction contained Pax6 in wild-type cortex (47% of sorted GFP-negative precursors n = 173). Most cells from the cortex of hGFAP���GFP mice sorted by the lack of green flu- orescence were postmitotic neurons, however (NeuN or ��-tubulin-III positive). We therefore used the incorporation of bromodeoxyuridine (BrdU supplied continuously in the culture medium) to discriminate sorted cells that divide in vitro���that is, precursors���from the postmitotic neurons a b c d e f g h �� 2002 Nature Publishing Group http://neurosci.nature.com