Generation of GABAergic and dopaminergic interneurons from endogenous embryonic olfactory bulb precursor cells

In contrast to the local origin of the excitatory pyramidal and mitral/tufted neurons, the brain region in which cortical and olfactory bulb(OB) interneurons originate during the embryonic period is species-dependent. In humans, the main source of cortical GABAergic neurons appears to be the cortical ventricle zone (VZ) (Letinic et al., 2002); it is nonetheless thought that many embryonic rodent OB interneurons (as well as cortical interneurons) arise in the ganglionic eminences (GE) (Marín and Rubenstein, 2003). Dlx2/Gsh2/Er81-expressing precursor cells in the lateral ganglionic eminence (LGE) are a source of mouse OB interneurons(Stenman et al., 2003). Supporting evidence includes the in vivo observation that migratory cells that leave the GE reach the cortex and OB(Anderson et al., 1997; Anderson et al., 2001; de Carlos et al., 1996; Lavdas et al., 1999; Pencea and Luskin, 2003), and that GE precursor cells grafted in the walls of lateral ventricles migrate to the cortex and the OB (Wichterle et al.,1999; Wichterle et al.,2001). In addition, mice lacking Dlx1/2, Gsh1/2, Arx and Sp8 transcription factors, expressed in the GE, show markedly decreased expression of GABAergic and dopaminergic cell markers in the embryonic granule and periglomerular layers of the OB (Anderson et al., 1997; Bulfone et al.,1998; Corbin et al.,2000; Qiu et al.,1995; Stenman et al.,2003; Toresson and Campbell,2001; Waclaw et al.,2006; Yoshihara et al.,2005; Yun et al.,2003; Yun et al.,2001). Other data suggest that, in addition to this exogenous neuron source, interneurons can be generated intrinsically within the embryonic OB. These include observations that GABAergic markers are restored in the Gsh2 knockout mouse OB at late embryonic stages(Corbin et al., 2000; Toresson and Campbell, 2001; Yun et al., 2003), that the Gsh1/2 double null mutation reduces, but does not abolish, glutamic acid decarboxylase 67 (GAD67; Gad1 - Mouse Genome Informatics) and Er81 expression in the OB (Stenman et al.,2003; Yun et al.,2003), and that a considerable number of GABAergic neurons remains in the OB of Arx mice (Yoshihara et al.,2005). Finally, the Gsh2 null mutation does not affect OB neuron generation and differentiation from cultured LGE cells(Jensen et al., 2004).

Overall, these findings indicate that the LGE may be the primary source of OB interneurons during the embryonic period(Stenman et al., 2003; Wichterle et al., 1999) but do not rule out dopaminergic and GABAergic neuron generation from local OB precursor cells, particularly before migratory LGE cells reach the OB. According to previous studies (Pencea and Luskin, 2003), the rostral migratory stream (RMS) reaches the rat OB by embryonic day (E) 16.5; this embryonic age would correspond approximately to mouse E14.5. A recent study using Arx to label migratory cells in the mouse (Yoshihara et al.,2005) corroborates the idea that the first GE-derived cells are detected in the OB at E14.5.

Here we show efficient differentiation of local E13.5 OB precursor cells into GABAergic and dopaminergic interneurons, when transplanted into E13.5 OB slices or into P5-P7 OB in vivo, or when plated in short-term dissociated cultures. GABAergic neurons isolated from the rostral half of E13.5 OB and from E13.5 GE were distinguished by their distinct migration patterns following transplant into the OB, and by their different gene expression in short-term dissociated cultures. This suggests that the OB contains a distinct, endogenous pool of interneuron precursor cells. Our results also show that these precursors differentiate in vivo into morphologically mature neurons, suggesting that they could integrate into the OB circuit.

Tissue culture reagents were purchased from Gibco-Life Technologies and Sigma. Mouse care was in accordance with European Union guidelines. The day on which a vaginal plug was found was considered E0.5. E13.5 OB cells were prepared from the rostral half of CD1 mouse OB to avoid contamination of LGE extending into the caudal part of the OB and of the accessory olfactory bulb(Jiménez et al., 2002). OB cells and GE (both medial and lateral eminence) cells were mechanically dissociated, then cultured for 6 or 17 days essentially as described(Vicario-Abejón et al.,1998). Cells were also isolated and cultured from the OB of E13.5 embryos from pregnant mice that were injected intraperitoneally (i.p.) with 5′-bromo-2-deoxy-uridine (BrdU; 100 μg/g) 4 or 14 hours before embryos were removed.

Olfactory bulb stem cells (OBSC) were prepared from E13.5-14.5 mice. In some experiments, only the rostral half of the E13.5 OB was dissected. Cells were plated and expanded with fibroblast growth factor 2 (Fgf2; PeproTech) and epidermal growth factor (Egf; PeproTech). For cell differentiation assays,mitogens were removed and cells were cultured for 15-20 days in Dulbecco's modified Eagle medium (DMEM)/nutrient mixture F12 (F12)/insulin,apotransferrin, putrescine, progesterone, sodium selenite (N2) plus 5% fetal bovine serum (FBS). Some cultures were supplied with brain-derived neurotrophic factor (Bdnf; PeproTech). Clonal analysis experiments of E13.5-derived primary cell spheres were performed as reported(Vicario-Abejón et al.,2003).

Cells were prepared from the rostral half of the OB from E13.5 transgenic C57BL/6 mice expressing enhanced green fluorescent protein (EGFP)(Okabe et al., 1997). Cells were also prepared from E13.5 CD1 mouse OB, and subsequently infected with an EGFP-expressing lentiviral vector (ViraPower Lentiviral Expression System;Invitrogen) (provided by P. Tsoulfas, University of Miami). Both types of labeled cells were suspended in Hank's balanced salt solution (HBSS), then implanted into E13.5 OB slices. To prepare slices, E13.5 brains were embedded in 4% low-melting-point agarose, and 250 μm-thick coronal sections were cut on a vibratome. Sections were transferred to inserts containing a polycarbonate culture membrane (8 μm pore size; Nunc; Corning Costar). Using a pulled glass pipette tip with a 50-70 μm outer diameter connected to a Hamilton syringe (mounted on a stereotaxic apparatus), 1 μl of a suspension of 5×10⁴ cells/μl was injected. Slices were cultured for 2 days in DMEM/F12/N2 plus 10% FBS, then fixed with 4%paraformaldehyde (PFA). Slices were immunostained with antibodies to GFP(1:500; Molecular Probes), GABA (1:1000; Sigma), tyrosine hydroxylase (Th;1:100; Chemicon) or Tbr1 (1:100, provided by M. Sheng, MIT; or 1:2000,provided by R. Hevner, University of Washington).

Cells were suspended in HBSS, then implanted into the subependymal zone(SEZ) of postnatal day 5-7 (P5-P7) mouse OB, when hosts were at the peak of maximal interneuron formation (Hinds,1968a). Mice were anesthetized by placing them in ice for 1-3 minutes. Each mouse was then positioned in a miniaturized stereotaxic device(Cunningham neonatal rat adaptor; Stoelting) fixed to a standard stereotaxic apparatus (Kopf) (Vicario-Abejon et al.,1995). Stereotaxic coordinates for implants into P5 OB were anteroposterior to bregma (AP) +1.3 mm, lateral to midline (L) 0.6 mm, ventral to dura (V) 0.5 mm, and for P7 OB were AP +1.6 mm, L 1.0 mm, V 0.8 mm. Subsequently, 1.5-2 μl of a suspension of 10⁵ cells/μl was injected. Only round aggregates, possibly formed by macrophages and cell debris, were found in control mice transplanted with freeze-thaw killed cells(not shown).

At 1 to 10 weeks post-transplant, mice were anesthetized by i.p. ketamine/xylazine injection and perfused transcardially with 0.9% NaCl and 4%PFA. Brains were post-fixed, embedded in 3% agarose, and cut in serial 50μm vibratome sections. Sections containing the OB were collected and examined under a confocal fluorescent microscope. Some sections were immunostained with antibodies to GABA, GAD (1:75), Th, calbindin (1:100;Swant), parvalbumin (1:1500; Swant) or synaptophysin (1:4; Zymed).

Cells were incubated with antibodies to Dlx2 (1:2000; provided by D. Eisenstat, University of Manitoba), Gsh2 (1:2000; provided by K. Campbell,University of Cincinatti), Pax6 (1:300; Covance), GABA, Th, Tbr1,β-III-tubulin (TuJ1, 1:2000; Covance), synapsin I (1:750; provided by M. Kennedy, Caltech), GAD, MAP2ab (1:150; Sigma), BrdU (1:1000) or SV2 (1:50). GAD, BrdU (G3G4) and SV2 antibodies were developed by D. I. Gottlieb,Washington University School of Medicine, S. J. Kaufman, University of Illinois and K. M. Buckley, Harvard Medical School, respectively, and were obtained from the Developmental Studies of Hybridoma Bank (University of Iowa Department of Biological Sciences).

For histology, E12.5-13.5 embryo heads were fixed overnight in 4% PFA. Alternatively, E13.5 embryos were first perfused transcardially. Heads were immersed in 30% sucrose, then frozen at -70°C. Cryostat coronal or sagittal sections (14 μm) were incubated with the antibodies mentioned above. Postnatal mouse brain cryostat sections were double immunostained with a rabbit antibody to the tyrosine kinase domain of TrkB [1:50; provided by S. Feinstein and M. Radeke (University of California)(Vicario-Abejón et al.,1998)] and mouse anti-SV2.

To determine the number of cultured cells expressing a specific antigen,ten random fields were counted per coverslip using a 40× objective and fluorescence filters(Vicario-Abejón et al.,2003). In transplantation experiments, the percentage of GABA⁺ cells and Th⁺ cells were calculated relative to the number of E13.5 GFP⁺ cells found in the slices. The percentage of cells located in the different cell layers was calculated relative to transplanted E13.5 GFP-positive OB or LGE cells found in the postnatal OB. Morphological analysis was performed on 50 neurons derived from transplanted E13.5 GFP⁺ OB cells, and on 50 neurons derived from E13.5 GFP⁺ LGE cells. Statistical analysis was performed using Student's t-test.

It is proposed that transcription factor expression specifies the types of interneurons generated in specific brain regions(Butt et al., 2005; Yuste, 2005). Dlx/Gsh2/Er81-expressing precursors in the LGE are a source of mouse OB interneurons (Stenman et al.,2003); the newborn neurons then migrate to the OB. To determine whether Gsh2, Dlx2 and interneuron markers are expressed in mouse E13.5 OB,before the main wave of migration from the LGE to the OB, we immunostained sections with anti-Gsh2 (Stenman et al.,2003) and anti-Dlx2 antibodies(de Melo et al., 2005). Gsh2 and Dlx2 expression were abundant in the medial ganglionic eminence (MGE) and LGE (Fig. 1A-H), whereas E13.5 OB did not express Gsh2 and Dlx2 (Fig. 1I,J). These findings indicate that the RMS does not reach the mouse OB at E13.5 but later, reportedly at E14.5 and after(Pencea and Luskin, 2003; Yoshihara et al., 2005). By contrast, we found cells expressing interneuron markers such as GABA and Th in the OB (Fig. 1K-M). As predicted, we found cells expressing Tbr1, a marker of mitral/tufted neurons(Bulfone et al., 1998), in the OB (Fig. 1N). These data indicate the presence of interneurons within the OB before LGE migratory cells expressing Dlx2 or Gsh2 could be detected in the OB.

To study the developmental expression of Gsh2 and Dlx2 and the generation of interneurons in the OB independently of GE influence, we isolated cells from the rostral half of the E13.5 OB and from the GE, and plated them in short-term dissociated cultures for 6 days(Fig. 2). GAD⁺neurons were detected in both cultures as early as 6 hours post-plating (not shown). At 6 days, 22.3 and 75.9% of total neurons (TuJ1⁺ cells)were GAD⁺ in OB and GE cultures, respectively(Fig. 2M). Double labeling of GAD⁺ neurons with anti-Dlx2 and anti-Gsh2 antibodies revealed a different expression pattern for these transcription factors in OB and GE GABAergic cells. Of the total OB GAD⁺ neurons, none (0%) expressed Gsh2 (Fig. 2A-C,N) and 53.3%expressed Dlx2 (Fig. 2D-F,N),whereas in GE cultures, 16.1 and 70.4% of GAD⁺ neurons expressed Gsh2 (Fig. 2G-I,N) and Dlx2(Fig. 2J-L,N), respectively. Differences between OB and GE Dlx2⁺ and Gsh2⁺ cells were best observed when proportions were calculated of double-labeled Dlx2⁺GAD⁺ cells divided by the number of total Dlx2⁺ cells, and of Gsh2⁺GAD⁺ cells divided by total Gsh2⁺ cells (Fig. 2O). In OB cultures, 41.5% of the total Dlx2⁺ cells were GAD⁺, whereas no Gsh2⁺ cells were GAD⁺. By contrast, 75.5% of Dlx2⁺ cells and 49.1% of Gsh2⁺cells in GE cultures co-labeled with GAD(Fig. 2O). These results show a significantly closer relationship in GE than in the OB between acquisition of a GABAergic phenotype and Dlx2 and Gsh2 expression. The absence of double-labeled Gsh2⁺ and GAD⁺ cells in OB cultures strongly suggests that most GAD⁺ neurons derive from local OB and not from GE precursor cells.

To determine whether OB interneurons express pallial markers, we stained sections with anti-Pax6 (Puelles et al.,2000) (Fig. 3A-C). Pax6 labeled the entire OB neuroepithelial zone (NEZ) and cells distributed in the mantle zone (MZ). Cultures of E13.5 OB cells showed 52.0±1.1%(n=3) GAD⁺ neurons expressing Pax6(Fig. 3D-F). Many GABA⁺ cells located in the MZ were marked with anti-BrdU in sections prepared from mice that received BrdU to label dividing cells 4 or 14 hours before sacrifice (Fig. 3G-I). In the OB cultures, 95.4±0.9% (n=3)GABA⁺ neurons co-labeled with BrdU(Fig. 3J-L), indicating that the large majority of OB interneurons are descendants of E13.0-13.5 dividing precursors.

To analyze which neuron types are generated from E13.5 OB precursors within the E13.5 OB, we transplanted EGFP-expressing cells into E13.5 OB slices prepared from wild-type embryos. Precursor cells were transplanted immediately after extraction from EGFP transgenic embryos, or were prepared from wild-type embryos followed by immediate incubation with a lentiviral vector expressing EGFP, and then transplanted. Precursor cells were thus injected with no previous culture. Two days after injection, we found numerous GFP⁺cells expressing interneuron markers in the slices(Fig. 4). Of the total GFP⁺ cells, 31.7% were GABA⁺(Fig. 4A-C,G) and 12.1% were Th⁺ (Fig. 4D-G). Cells expressing Tbr1 were also found (not shown). These results indicate that interneurons can be generated endogenously within the OB.

Concurring with the considerable numbers of GAD⁺ cells found in OB dissociated cultures, and with the fact that the majority of OB dopaminergic neurons are also GABAergic(Carleton et al., 2003; Shipley and Ennis, 1996),GABA⁺ and Th⁺ cells with clear neuron morphology were abundant in these cultures (Fig. 5A,B). We then performed two types of experiments to determine whether OB neurons can differentiate extensively and display mature morphologies. E13.5 OB cell suspensions were cultured for 15-20 days to promote neuron maturation (Fig. 5C-H). Cultures were double-stained with antibodies to MAP2ab (a general neuron marker that labels dendrites specifically, in addition to cell bodies) and to GABA, Th or Tbr1. The cultures had many MAP2ab⁺neurons with large number of dendrites and dendritic branches(Fig. 5C,E,G). We detected GABAergic and dopaminergic interneurons(Fig. 5D,F), as well as mitral neurons (Fig. 5H), suggesting that the major OB neuron types differentiate to attain mature morphological features in this culture system.

We studied neuron differentiation and cell migration in vivo by transplanting suspensions of EGFP-expressing E13.5 OB cells (without prior culture) into P5-P7 OB SEZ (Fig. 6). For comparison, EGFP-expressing E13.5 LGE cells were also grafted. One week after surgery, we observed transplanted GFP⁺ OB cells with a migratory, relatively simple neuron morphology in the OB(Fig. 6A). Transplanted cells migrated away from the injection site (in an apparently radial pattern) to reach the different OB layers. By 2 weeks post-transplant, grafted OB cells showed more complex neuron morphology (Fig. 6B), and had reached the granule cell layer (GCL), the internal plexiform layer (IPL), the boundary between the IPL and the mitral cell layer(ML), and the glomerular layer (GL), where the cell bodies were located. OB cells migrated preferentially to the GCL (67.1% of total GFP⁺ cells found in the host OB) and to the ML boundary with the IPL and the external plexiform layer (EPL) (termed ML + PL) (15.5%)(Fig. 6B,C,K). Fewer grafted cells migrated to the GL (8.9%). At 4 weeks post-transplant, many OB cells differentiated into morphologically distinct granule neurons(Fig. 6C-E). Grafted OB cells gave rise to GABA-, GAD- and Th-expressing neurons, but not to neurons expressing calbindin or parvalbumin (Table 1, and not shown). Transplanted LGE cells(Fig. 6F-K) also migrated and differentiated within the host OB. In contrast to transplanted OB cells, LGE cells migrated preferentially to the GCL (78.9%) and the GL (17.7%) rather than to the ML + PL (4.7%). The tendency of LGE cells to differentiate into glomerular and periglomerular (PGC) neurons was corroborated by expression of calbindin and parvalbumin, two GC/PGC neuron markers(Table 1, and not shown). For morphological analysis (Fig. 6L), we counted the number of GFP⁺ neurons in the GCL and in the ML + PL possessing one apical dendrite(Fig. 6D,I), two apical dendrites (Fig. 6E) or more than two dendrites (Fig. 6J). The large majority of neurons found in the postnatal OB transplanted with OB cells (Fig. 6D,E,L), had one(51%) or two (43%) apical dendrites. Only 6% of neurons were multidendritic(Fig. 6L). In the OB transplanted with LGE cells, 46 and 19% of neurons had one(Fig. 6I) and two apical dendrites, respectively, whereas 25% were multidendritic(Fig. 6J,L). These results show distinct migration and differentiation patterns for E13.5 OB and E13.5 LGE cells within the early postnatal OB.

We next studied the maturation of grafted OB cells in animals that were allowed to grow for 1 to 2.5 months (Fig. 7). As mentioned, many granule neurons had a single, large apical dendrite that crossed the ML and extended dendritic branches with spines in the EPL; some branches reached the EPL/GL boundary(Fig. 7A). Nonetheless, granule neurons located in the IPL/ML boundary had two major dendrites, which also extended branches with dendritic spines into the EPL(Fig. 7A,C,D). Transplanted granule cells also had a few shorter dendrites oriented toward deeper parts of the GCL (Fig. 7A). Anti-synaptophysin antibody staining (Fig. 7E,F) showed close proximity of dendritic spines to synaptic boutons. These morphological characteristics resemble those reported for OB granule neurons (Carleton et al.,2003; Lledo and Saghatelyan,2005; Shipley and Ennis,1996). Fewer OB implanted cells migrated to the GL, where they differentiated into periglomerular neurons(Fig. 7B). Like host periglomerular neurons (Fig. 7Binset), some grafted cells oriented their cell bodies horizontally. GFP⁺-transplanted cells were labeled with anti-GAD and anti-GABA antibodies, which confirmed the GABAergic phenotype(Fig. 7G-O). Some transplanted GFP-expressing cells were Th⁺, indicating a dopaminergic phenotype(Fig. 7P-R). These results indicate that local E13.5 OB precursor cells can differentiate and mature in vivo into the main types of OB interneurons, and that the cell body of these neurons resides in the appropriate OB cell layers.

The previous results suggest the existence of an endogenous source of OB precursor cells that generate interneurons in addition to mitral/tufted neurons. Precursor cells with neural stem cell (NSC) features, i.e. self-renewal and the potential to produce neurons and glia, have been isolated(Sanai et al., 2004; Vicario-Abejón et al.,2003). To test whether OB precursor cells having NSC features are able to generate GABAergic, dopaminergic, and mitral/tufted neurons, we expanded E13.5 and 14.5 OBSC and induced them to differentiate. A characteristic of GABAergic neurons from various brain regions is that they respond to Bdnf (Vicario-Abejón et al., 2002). As this neurotrophin and its high-affinity TrkB receptor are expressed in the OB (Nef et al., 2001) (Fig. 10A-E), we tested the response of OBSC-derived neurons to Bdnf.

E13.5 OBSC that had differentiated for 15-20 days generated GAD⁺neurons expressing Dlx2, although GAD⁺ and Dlx2^- neurons were also found (Fig. 8A,B), as occurred in short-term dissociated OB precursor cells(Fig. 2). Double staining of control E13.5 and 14.5 OBSC-derived neurons with anti-MAP2ab and anti-GABA antibodies showed 64 and 57% GABA-positive neurons, respectively(Fig. 8C,D,G). Addition of Bdnf to the cultures produced no statistically significant changes in the number of GABAergic neurons (Fig. 8E-G). Some Bdnf-treated neurons showed more complex morphology than controls(Fig. 8C-F). To rule out possible detection of GABA⁺ neurons in E13.5 OB cultures derived from LGE extension into the caudal OB, we prepared cells from the rostral half of the E13.5 OB. In these conditions, 57.2±2.9% (n=4) of neurons were GABA⁺.

We used clonal analysis to confirm that GABAergic neurons in OBSC cultures derived from dividing stem cells and not from committed GABAergic progenitors(Fig. 8H-L). We plated E13.5-derived cells obtained from primary spheres; the following day, 31.4%(n=2 experiments) of the wells contained a single cell(Fig. 8H). After 7-10 days,72.2% of single cells formed spheres (Fig. 8I), which were then induced to differentiate(Fig. 8J-L). Of these clones,67% generated GABA⁺ neurons(Fig. 8K,L). Most clones also contained glial cells (not shown). The results suggest that the precursor cells in the OB neuroepithelium that show NSC characteristics in culture are a source of GABAergic cells.

We then analyzed whether OBSC can differentiate into dopaminergic neurons(Fig. 9A-E). Of the total MAP2ab⁺ neurons in E13.5 and 14.5 OBSC-derived cultures, 18 and 16%, respectively, were Th⁺(Fig. 9E). As observed in vivo,the majority of cultured Th⁺ neurons expressed GAD (not shown). In this assay, Bdnf had no significant effect on Th⁺ cell numbers(Fig. 9C-E).

The OB excitatory neurons, mitral and tufted cells, express the Tbr1 transcription factor. In the previous assays(Fig. 8G, Fig. 9E) we observed no major differences in neuron differentiation between E13.5- and E14.5-derived OBSC;mitral/tufted neuron generation was thus only tested in E13.5 OBSC cultures(Fig. 9F-J). Of the total MAP2ab⁺ neurons in control and Bdnf-treated cultures, 9.2 and 11.6%, respectively, were Tbr1⁺(Fig. 9F).

Unexpanded OB precursor cells generate morphologically mature GABAergic and dopaminergic interneurons in culture (Fig. 5) and in vivo (Figs 6, 7). To further assess the maturation stage of neurons arising from OB cells, we stained OBSC-derived neurons with antibodies to synapsin I and SV2(Fig. 10), two presynaptic proteins involved in synaptogenesis and synapse function(Vicario-Abejón et al.,2002). We tested the effects of Bdnf on synaptic marker expression by OB neurons, as staining with antibodies to TrkB revealed this neurotrophin receptor in mitral (Fig. 10A)and periglomerular neurons (Fig. 10B). Furthermore, double staining showed that TrkB was highly expressed at SV2⁺ synaptic terminals in the glomeruli(Fig. 10C-E). E13.5 OBSC differentiated into neurons (Fig. 10F-Q), including GABAergic and Th⁺ cells that expressed synapsin I and SV2 in a punctate pattern(Fig. 10H-Q) characteristic of mature neurons (Vicario-Abejón et al., 1998). Neurons cultured with Bdnf had a more complex neuron morphology, with more neurites and neuritic branches than control cells(Fig. 10F,G). In these conditions, Bdnf-treated neurons showed more synapsin I⁺ and SV2⁺ boutons than control neurons(Fig. 10; compare panels I with H; L,M with J,K; P,Q with N,O).

The excitatory neurons of the rodent, primate and human cortex (pyramidal neurons) and OB (mitral and tufted neurons) are produced in local germinal zones of the embryonic dorsal telencephalon(Bulfone et al., 1998; Hinds, 1968b; Nomura and Osumi, 2004; Rakic, 1971). Interneuron origin during the embryonic period nonetheless appears to be more diverse and species-dependent. In humans, the main source of cortical GABAergic neurons appears to be the cortical VZ (Letinic et al., 2002). By contrast, it is thought that many embryonic mouse OB interneurons (as well as cortical interneurons) are not generated locally in the OB or in the cortical VZ, but arise in the GE(Marín and Rubenstein,2003; Stenman et al.,2003), from which they migrate to their final destination(Anderson et al., 1997; de Carlos et al., 1996; Lavdas et al., 1999; Pencea and Luskin, 2003; Wichterle et al., 1999; Wichterle et al., 2001).

Using a combination of in situ analysis of OB development, cell transplant and cell culture approaches, we show the generation of GABAergic and dopaminergic interneurons from E13.5 OB precursor cells in vivo, in slices transplanted with precursor cells and in short-term dissociated cultures. This strongly suggests that, in addition to receiving interneurons from the LGE,local precursor cells produce interneurons within the OB. Unexpanded E13.5 OB precursor cells transplanted in E13.5 slices gave rise to 31% GABAergic neurons, whereas 22% of cells in dissociated cultures were GAD⁺. In postnatally grafted mice we found many neurons, especially granule cells. The three experimental systems thus indicated that local OB interneuron generation is a quantitatively significant and consistent event. Furthermore,transplanted cells migrated within the host OB and gave rise to morphologically mature granule neurons and periglomerular cells. Granule neurons had dendritic spines, which may be contacted by synaptophysin-positive boutons. These findings suggest that interneurons born within the OB could integrate into synaptic circuits, as do neurons generated in the neonatal and adult forebrain subventricular zone (SVZ)(Belluzzi et al., 2003; Carleton et al., 2003; Lemasson et al., 2005).

To confirm that interneurons are generated endogenously in the OB, we studied OB cell development independently of the GE using precursor cells isolated from the rostral half of E13.5 OB. Our data showed no Dlx2 or Gsh2 expression in the OB at this age, when these markers are expressed abundantly in the GE, concurring with recent reports that the first GE-derived cells are not detected in mouse OB until E14.5(Pencea and Luskin, 2003; Yoshihara et al., 2005). These results, and the complete absence of Gsh2⁺/GAD⁺ neurons in OB cultures, are consistent with the suggestion that the majority of interneurons found in these cultures, in postnatal mice and E13.5 slices grafted with E13.5 OB precursor cells, as well as in situ in the E13.5 OB,arise from endogenous OB and not from GE precursor cells. The BrdU data support the concept that most differentiated interneurons are born from dividing precursors in the E13-13.5 OB.

Dlx2 and Gsh2 proteins were not detected in E13.5 OB sections, but were expressed in E13.5 plated directly in short-term dissociated culture for 6 days. These findings suggest that these transcription factors are expressed by endogenous OB precursor cells and that detectable levels of protein expression are found later than E13.5. In support of this, Gsh2 mRNA was found in the E13.5 OB by RT-PCR (not shown). Precursor cells expressing these transcription factors appear to be distinct in the OB and the GE, based on GAD colocalization; 41.5% of Dlx2⁺ and 0% of Gsh2⁺ cells coexpressed GAD in OB cultures, whereas 75.5% of Dlx2⁺ and 49% of Gsh2⁺ cells colocalized with GAD in GE cultures. Furthermore,migration of transplanted OB and LGE precursors to the distinct postnatal OB layers was dependent on the region of cell origin; OB cells migrated preferentially to the GCL and the IPL-ML-EPL, whereas LGE cells were found predominantly in the GCL and GL. OB and LGE cells also generated different interneuron subtypes that can be distinguished by their morphology and expression of calcium-binding proteins. These data indicate that the embryonic OB contains a distinct local population of interneuron-generating precursor cells, and suggest the importance of the spatial origin of interneuron precursors in shaping their final phenotype, as proposed(Butt et al., 2005; Yuste, 2005). Specifically, a considerable proportion of GAD⁺ neurons express the telencephalic pallial domain marker Pax6 (Puelles et al., 2000). Although Pax6 is expressed in the dorsal LGE, which is also Gsh2⁺ (Kohwi et al.,2005; Yun et al.,2001), the absence of GAD⁺Gsh2⁺ cells in cultures prepared from the rostral half of the OB suggests that the GAD⁺Pax6⁺ neurons found in the OB cultures derived from pallial OB Pax6⁺ precursors. In accordance with our results,cortical NSC differentiated into Dlx⁺ and GAD⁺ neurons,suggesting an endogenous subpopulation of cortical interneurons(He et al., 2001; Parmar et al., 2002). This resembles data showing genesis of Dlx2⁺ cells(Nery et al., 2003) and of Dlx2⁺Mash1⁺-expressing GABAergic neurons in the dorsal human and rodent cortical VZ (Bellion et al., 2003; Letinic et al.,2002). These findings support the concept that interneurons originate in dorsal telencephalic VZ as well as in ventral telencephalic zones, and that the quantitative contribution of each germinative zone to total embryo interneuron number is species-dependent(Gotz and Sommer, 2005).

The capacity of local OB precursors to give rise to interneurons in the embryonic period appears to persist in the neonate(Lemasson et al., 2005) and is gradually lost in adulthood, although interneuron generation has been found in the adult rodent and in human OB (Gritti et al., 2002; Hoglinger et al., 2004; Liu and Martin,2003), in addition to the well-studied, predominant postnatal-adult SVZ source of OB interneurons(Kornack and Rakic, 2001; Lois and Alvarez-Buylla,1994).

Here we show that a previously characterized OBSC population(Vicario-Abejón et al.,2003) not only generates mitral/tufted neurons, as predicted, but that differentiation of OBSC gave rise to large numbers of GABAergic neurons. It is very unlikely that OBSC originate in the GE rather than in the OB neuroepithelium, as stem cell migration between the OB and the GE has not been reported (Reid et al.,1999).

To confirm that OB GABAergic neurons arise from NSC and not from committed GABA progenitor cells in OBSC cultures, we performed clonal analysis experiments, in which the majority of single cell-derived clones gave rise to GABA⁺ neurons (as well as to glial cells), indicating that many of these neurons are produced by multipotent cells with the ability to self-renew. As observed in the short-term OB dissociated cultures, a proportion of OBSC-derived GABAergic neurons were Dlx2⁺, suggesting that under these conditions, mitogen treatment that expands OBSC produces no obvious changes in the expression of genes that confer dorsoventral neural patterning. In support of this hypothesis, we also found neurons expressing Tbr1, a dorsal telencephalic marker in differentiating OBSC cultures, as well as Tbr-1 mRNA (not shown). GAD-negative/Gsh2⁺ cells were also found in OBSC differentiating cultures (not shown), as seen in the short-tem OB dissociated cultures.

E13.5 OB precursor cells plated in short-term dissociated cultures or transplanted into the neonatal OB developed a mature morphology, including dendritic spines in granule neurons. Consistent with these results,OBSC-derived neurons differentiated extensively and expressed presynaptic proteins in a punctate pattern characteristic of mature neurons. OBSC-derived GABAergic and dopaminergic neurons cultured with Bdnf show more synapsin I and SV2 expression than control cells, although this neurotrophin had no clear effect on neuron survival. Bdnf nonetheless promotes the generation and/or survival of new neurons in adult rostral SVZ and OB(Nanobashvili et al., 2005; Zigova et al., 1998). OBSC-derived neuron responses to Bdnf, the abundance of TrkB at synaptic sites, and the lack of Bdnf effects on the formation of new neurons from OBSC(Vicario-Abejón et al.,2003) suggest that Bdnf promotes morphological and synaptic maturation of OB neurons. These findings concur with the action reported for this neurotrophin in promoting synapse formation and stabilization(Hu et al., 2005; Martinez et al., 1998; Nanobashvili et al., 2005; Rico et al., 2002; Vicario-Abejón et al.,2002).

In summary, our results show efficient formation of mature GABAergic and dopaminergic neurons from endogenous embryonic OB precursor cells. These locally generated interneurons, along with the mitral and tufted neurons and the extrinsically generated OB interneurons, may contribute to the formation of the functional OB synaptic circuit.

We thank M. Sheng, K. Campbell, D. Eisenstat, R. Hevner, M. Kennedy, S. Feinstein, and M. Radeke for gifts of antibodies, P. Tsoulfas for a lentiviral vector, E. J. de la Rosa, J. de Carlos, and L. López-Mascaraque for helpful discussions, R. Varillas and T. de la Cueva for technical assistance,and C. Mark for editing the English text. E.V.-V. and F.deC. are grateful to the Dept. of Neurobiology, Hospital Ramón y Cajal, Madrid, for initial support. This work was funded by grants from Fundación La Caixa(NE03/72-02), from the MEC (SAF2004-05798), and the Comunidad de Madrid (CM)(GR/SAL/0835/2004) to C.V.-A., from the MEC (BMC 2004-0152) to F.deP., from the FIS (01/3136 and PI020768) to F.deC., and from the MEC(GEN2001-4856-C13-02) to A.B. F.deC. is a fellow of the Ramón y Cajal Program. E.V.-V. and M.J.Y.-B. were supported by the MEC and the CM,respectively. The Department of Immunology & Oncology was founded and is supported by the CSIC and by Pfizer.