Pallial origin of basal forebrain cholinergic neurons in the nucleus basalis of Meynert and horizontal limb of the diagonal band nucleus

Cholinergic neurons of the basal forebrain (ChBF) are located in three main areas: anteromedial [medial septal nucleus (MS), or Ch1], intermediate [vertical and horizontal limbs of the diagonal band nucleus (VDB and HDB), or Ch2/Ch3] and caudolateral [the basal nucleus of Meynert (MB), or Ch4] (Mesulam et al., 1984; Mesulam, 1990). Even though ChBF are functionally related to attention and cognition and their degeneration is implicated in Parkinson and Alzheimer's diseases (Bohnen and Albin, 2010), their origin and migration are still unclear. Previously, ChBF were reported to originate in the subpallium (Sussel et al., 1999; Marin et al., 2000; Marin and Rubenstein, 2001; Xu et al., 2008); however, we have observed a different origin and analyzed the cellular mechanisms involved in ChBF development.

The formation of heterogeneous neuronal regions in the brain depends on complex temporospatial patterns of gene expression, regulating the local production of neurons and cellular migrations from distant germinative areas. Thus, the vertebrate cortex from the pallial telencephalon (pallium) derives from the proliferation of local neurons (Molyneaux et al., 2007; Cobos et al., 2001b) and from the interneuron migration of ventral subpallial structures (Wichterle et al., 1999; Cobos et al., 2001a; Tuorto et al., 2003; Wonders and Anderson, 2006). Although clonal experiments in avian embryos have demonstrated a countercurrent migration of cells from the pallium to the subpallium (Striedter et al., 1998; Cobos et al., 2001a), this has yet to be analyzed in detail.

In this work, we analyzed pallio-subpallial neuronal migration in mouse embryos. Neuroepithelial transplants in organotypic cultures of GFP and control mouse brains were performed. We observed a tangential migration of cells positive for calretinin and reelin from the ventral pallium (VP) to the subpallium. Moreover, since these migratory cells also expressed the pallial gene Tbr1, we studied ChBF in Tbr1 mutant (Tbr1^–/–) mice, whose brains present a severe disruption of the VP (Bulfone et al., 1998; Hevner et al., 2001; Hevner et al., 2002). In Tbr1^–/– mice, no cholinergic cells were detected in the HDB, MB (the corticopetal ChBF) or the subpallial-derived amygdala. Finally, through in vitro assays and analysis of Fgf8 hypomorphic mice, we substantiate Fgf8 as a candidate attractive signal from the medioventral telencephalon.

Thus, our study describes a new, pallial origin of the corticopetal ChBF. The differences in ChBF ontogenesis shown here might explain their different connection patterns (cortex versus hippocampus) and functions in cognitive processes, as well as their specific responses to pathological stimuli. Conversely, the common ontogenetic origin of various ChBF nuclei and olfactory structures might explain the early comorbidity in neurodegenerative diseases.

All animal experiments were performed in compliance with the Spanish and European Union laws on animal care in experimentation (Council Directive 86/609/EEC), as well as being analyzed and approved by the Animal Experimentation Committee of our university.

Control mice (ICR) and CD1-GFP transgenic mice (Hadjantonakis et al., 2002) were used for the organotypic cultures and transplant experiments at E14.5. Tbr1^–/– mice (Bulfone et al., 1998) were analyzed at P0 and P21, using Tbr1^+/– mice as controls. Fgf8 mutant mice (Meyers et al., 1998) were studied at E14.5-16.5; mutant alleles were maintained on a mixed 129/CD1 Swiss genetic background. Fgf8^null/+ and Fgf8^neo/+ mice were crossed to produce Fgf8^neo/null embryos, and Fgf8^neo/+ and Fgf8^neo/+ to produce Fgf8^neo/neo embryos. PCR genotyping was performed as previously described (Hebert and McConnell, 2000). Fgf8^neo/+ embryos were used with wild-type embryos as controls.

Organotypic slice cultures of E14.5 and E15.5 mouse brains were prepared as previously described (Anderson et al., 1997). Sections including medial ganglionic eminence (MGE) and VP were incubated on floating polycarbonate membrane filters (8 μm pore size, Nunc). Then, the graft was performed as described by Marin et al. (Marin et al., 2001). Depending on the experiment, either part of or the whole VP was grafted homotopically using GFP-expressing mice as donors and control mice as hosts. Grafted sections were incubated for 24 hours, fixed with 4% PFA and mounted. In some cases, the sections were paraffin embedded, cut and processed immunohistochemically using: anti-GFP polyclonal antibodies (1:1000, Molecular Probes, Leiden, The Netherlands; 1:500, Aves Labs, Tigard, OR, USA), anti-reelin monoclonal antibody (1:500, Chemicon, Temecula, CA, USA), anti-Tbr1 polyclonal antibody (1:25, Abcam, Cambridge, UK), anti-choline acetyltransferase (ChAT) polyclonal antibody (1:100, Chemicon), anti-Nkx2.1 polyclonal antibody (1:1000, anti-TTF from Biopat Immunotechnologies, Italy) or anti-Tuj1 monoclonal antibody (1:1000, anti-neuronal class III beta-tubulin from Covance, Berkeley, CA, USA). For fluorescence microscopy, fluorescent secondary antibodies were used (Alexa series, Molecular Probes).

To quantify the migration distance in slice cultures, the longest distance covered by the pallial cells from the graft-host boundary was measured in each experimental paradigm: whole subpallium, lateral ganglionic eminence (LGE) or MGE grafts. Statistical analysis was performed with the Mann-Whitney sum test using SigmaStat software.

Co-cultures were performed as described by Marin et al. (Marin et al., 2003). E14.5 mice were processed as described above to obtain brain slices. Small pieces of MGE and VP were placed on top of the Matrigel drops (BD Biosciences, Franklin Lakes, NJ, USA) and covered with Matrigel diluted 1:1 in Neurobasal culture medium (Gibco Life Technologies). Co-cultures of MGE and VP were separated by 400 μm and incubated for 24 to 48 hours.

Heparin on acrylic beads (Sigma) or Affi-Gel Blue Gel beads (Bio-Rad) were preincubated overnight with recombinant mouse Fgf8b (1 μg/μl; R&D Systems) or with the Fgf8 inhibitor SU5402 (40 μM; Calbiochem, La Jolla, CA, USA), respectively. Controls employed beads with BSA (1 μg/μl; Sigma). The soaked beads were co-cultured for 48 hours with VP explants in Matrigel cultures as described above.

To quantify cell migration in cultures, the VP piece was subdivided into four sectors depending on the bead position, and then cells migrating from the sectors nearest to each bead were counted, as well as from the sectors on the opposite side, which were used as controls (see Fig. 7O). Statistical analysis was performed by Student's t-test using SigmaPlot software.

HEK293T cells were cultured in 10% fetal bovine serum (FBS)-supplemented Dulbecco's modified Eagle's medium (DMEM; Sigma), 100 units/ml penicillin-streptomycin (Sigma) and 2 mM l-glutamine (Sigma).

pCMV-eGFP-IRES1hyg was generated by subcloning eGFP cDNA as a BamHI-NotI fragment from pEGFP-N1 (Clontech, Cambridge, UK) into BamHI and NotI digested pIRES1hyg (Clontech, Cambridge, UK).

The mouse Fgf8b coding region without stop codon was subcloned into the EcoRI site of a pEGFP-N1 vector to obtain pCMV-Fgf8b-eGFP-an.

The mRFP1 coding region without stop codon and the cDNA fragment encoding the leader sequence and GPI signal of the human decay-accelerating factor (CD55 molecule, NM_000547) were subcloned into BamHI and NotI digested pIRES1hyg vector to generate pCMV-mRFP1-GPI-IRES1hyg.

One day before transfection (with Lipofectamine 2000, Invitrogen), the cells were seeded at a density of 0.5×10⁵ cells/cm² in multi-well (12- or 24-well) plates. The cells were incubated with DNA-lipid complexes for 4 hours (following the supplier's instructions), after which the lipofection mix was removed and replaced with fresh medium. Drug selection of stable transfectants was performed with 50-100 μg/ml hygromycin B (hyg; Calbiochem) and G418 disulfate salt solution (Sigma).

Tbr1 mutant brain paraffin sections were immunostained following the protocol described above. The antibodies used were: anti-Pax6 monoclonal (1:5, Developmental Studies Hybridoma Bank, Iowa City, IA, USA), anti-ChAT polyclonal (1:100, Chemicon), anti-reelin monoclonal (1:500, Chemicon) and anti-Nkx2.1 polyclonal (1:1000, Biopat Immunotechnologies). To quantify differences between wild type and Tbr1^–/– in the basal forebrain (BF), cells expressing Pax6, ChAT and reelin were counted at P0 and P21 (n=3 for each stage). Cell counts were performed using a parallel series of sections stained by these specific markers in three different control embryos and three mutant embryos, counting four consecutive sections of the BF in each series. Statistical analysis was performed by Student's t-test using SigmaPlot software.

Fgf8 mutant mice were analyzed at E14.5-16.5. Paraffin coronal sections were immunostained with anti-ChAT polyclonal antibody (1:100, Chemicon) and stained with Cresyl Violet.

In order to detect pallial-subpallial migration, we prepared sagittal explants from E14.5 control mice (ICR), where VP from GFP mice (GFP-VP) was homotopically grafted after the excision of equivalent host territories. In some experiments, GFP-VP was transplanted in place of the host anterior olfactory nucleus (AO) and, specifically, the olfactory cortex close to the pallio-subpallial boundary (Fig. 1A). After 24 hours in culture, a large number of GFP⁺ cells were detected throughout the superficial subpallium (n=7; Fig. 1A-C,I), whereas fewer GFP⁺ cells were detected scattered throughout the striatum (ST) (n=6; Fig. 1A,B,D). Immunohistochemical analysis of the explants confirmed that the GFP⁺ migrating cells in the olfactory tubercle (Tu) and superficial subpallium were positive for reelin (Fig. 1E-H), Tbr1 (Fig. 1I-L) and calretinin (CR; also known as calbindin 2 – Mouse Genome Informatics) (data not shown), but negative for Nkx2.1 (Fig. 1M-P). Moreover, to confirm that these migrating cells were young postmitotic neurons we demonstrated that the GFP⁺ migrating cells expressed class III beta-tubulin (Tubb3) in their trailing process, a neuron-specific protein that is recognized by Tuj1 monoclonal antibody (Fig. 1Q-S).

At E14.5, the presence of fusiform CR-positive neurons in the superficial subpallium suggests a tangential migration through the subpial regions of Tu and preoptic area (POA) (Fig. 1T). Moreover, these CR⁺ cells appeared oriented in opposite directions, away and towards the VP, sharing a common migratory substrate (Fig. 1U), and were also reelin⁺ (Fig. 1V). Pax6 and ChAT double-positive basal forebrain (BF) neurons were detected (Fig. 1W), indicating the maintenance of this pallial marker (Carney et al., 2009) in ChBF. Interestingly, Tbr1 is expressed in postmitotic cells of the developing cortex from E10 to adulthood (Bulfone et al., 1995). Also, Tbr1 can be detected in the olfactory bulb (OB) and AO, as well as in the POA (Bulfone et al., 1995; Bulfone et al., 1998). In the E14.5 mouse embryo, Tbr1-expressing cells were mapped within the superficial subpallium as a cellular stream connecting the VP to the BF region [Fig. 1X; from Eurexpress (Diez-Roux et al., 2011)], confirming that Tbr1 expression is a suitable marker for pallio-subpallial migration in mouse.

At E15.5 and P14, many cells at the surface of the Tu expressed Tbr1 (Fig. 2A-D), some of which could be cholinergic neurons (ChAT⁺) (Fig. 2D-G). In the BF, 40% of cholinergic neurons were shown to co-express ChAT and Tbr1 or Nkx2.1 (Fig. 2H,L). However, in the MB there were more pallial Tbr1⁺/ChAT⁺ neurons (Fig. 2I-K) than subpallial Nkx2.1⁺/ChAT⁺ neurons (Fig. 2L-O) (60% and 20%, respectively).

The pallial origin of ChBF was explored by examining ChAT expression in experimental VP migrating cells. To this end, we performed immunohistochemistry on grafted slices (n=20) and observed that many VP migrating neurons expressed ChAT when they reached the BF (30-50% of GFP⁺ cells in MB and HDB expressed ChAT; Fig. 3A-F).

In Tbr1-deficient mice, the primary olfactory cortex and claustrum (lateral pallium and VP), OB, AO and pallial amygdala (which are VP-derived areas) are disrupted (Bulfone et al., 1998; Hevner et al., 2001). Since VP is the source of pallio-subpallial migratory cells, we examined whether this migration is affected in Tbr1 mutant mice. It has been reported that the majority of the striatal and Tu interneurons derive from the medial ganglionic eminence (MGE) (Marin et al., 2000). We analyzed Nkx2.1 expression in Tbr1 mutant mice. Nkx2.1 is expressed in part of the septum (Se), pallidum (PA), substantia innominata (SI) and the bed nucleus of the stria terminalis (Puelles et al., 1999; Sussel et al., 1999; Shimamura et al., 1995). Nkx2.1 expression was normal in these nuclei in Tbr1^–/– brains (Fig. 4A-F), suggesting that the subpallium was not affected.

Tbr1^–/– mice were analyzed at P0 (n=3; Fig. 5A-H) and P21 (n=4; Fig. 5I-Q) for the presence of migrating pallial neurons in the subpallium. According to previous data, the targets for the migrating cells are the Tu, HDB and MB. Thus, three markers were used to identify VP migrating cells: reelin (D'Arcangelo et al., 1997; de Bergeyck et al., 1998), Pax6 (Stoykova et al., 1997; Chapouton et al., 1999; Toresson et al., 2000; Yun et al., 2001) and choline acetyltransferase (ChAT). Immunohistochemical analysis of paraffin-embedded sagittal and coronal sections was performed (Fig. 5) using Tbr1^+/– mice as controls. At P0, the controls presented an abundance of reelin⁺ cells in the Tu [22±1 cells/section in four consecutive sections (cs); n=3; Fig. 5A,B,D,R], whereas the mutant did not express this marker (n=3; Fig. 5C,E,R). Pax6 was detected in deeper strata of subpallial areas than reelin in the control mice (54.5±5 cells/section in four cs; n=3; Fig. 5A,B). In Tbr1^–/– embryos there were significantly fewer Pax6⁺ cells (31.7±1.7 cells/section in four cs; n=3; Fig. 5C,R), and these gradually decreased towards the subpallium, resulting in very few dispersed cells in the HDB. Lastly, ChAT⁺ cells were detected in the BF of control mice at P0, as well as in the ventral pallidum (PV) and ST (47±3 cells/section in four cs; n=3; Fig. 5F,G,R), whereas very few ChAT⁺ cells were observed in BF of Tbr1^–/– brains (22±1 cell/section in four cs; n=3; Fig. 5H,R).

At P21, the differences between control and mutant mice were even more evident (Fig. 5I-Q,S). Tbr1 mutants either lacked or expressed very low levels of reelin (Fig. 5K,S), Pax6 (Fig. 5S) and ChAT (44±3 versus 11±1 cells/section in four cs of WT versus Tbr1^–/–, respectively; Fig. 5Q,S) in MB and HDB. Analysis of this region by Cresyl Violet staining further confirmed BF structural anomalies (Fig. 5L-N,P).

A subpallial origin of the cholinergic neurons in the ST and BF, including Tu, has been proposed previously (Sussel et al., 1999; Marin et al., 2000; Marin and Rubenstein, 2001), as Nkx2.1 mutants lack cholinergic neurons in the subpallium, including the striatal interneurons and BF projecting neurons. These defects have also been associated with a loss of Lhx8 expression in the developing ventral telencephalon (Zhao et al., 1999; Zhao et al., 2003). Alternatively, subpallium disruption might also eliminate an attractive signal for the pallial migrating cells, resulting in a lack of both subpallial and pallial ChBF.

To analyze whether the medial subpallium (MGE and Se) contains attractive signals for pallio-subpallial migration, explants were prepared from E14.5 GFP mice. Different subpallial areas from control embryos were transplanted into GFP slices. First, we homotopically transplanted the whole subpallium from control to GFP explants (n=16). After 24 hours, a large number of GFP⁺ cells migrated ventrally from the VP (GFP⁺ Tbr1⁺) to the grafted subpallium (Fig. 6A-F). There were numerous GFP⁺ cells in the mantle layer of the lateral ganglionic eminence (LGE) that migrated further along the superficial areas of the grafted subpallium (Fig. 6B,C,M,N). When only the LGE was transplanted (n=16; Fig. 6G-I), there was a substantial reduction in the number and migratory distance of GFP⁺ cells migrating into the grafted LGE (Fig. 6H,I,M,N). In these cases, GFP⁺ pallial-derived cells were detected clustered near the border of the graft. These results are in accordance with the existence of medial attractive signals that act upon pallial-derived neurons. When the medial area of the subpallium was grafted without the LGE, the number of GFP⁺ cells migrating into the MGE was also reduced (n=15; Fig. 6J-N). In this case, only a few pallial-derived cells were detected invading the grafted MGE (Fig. 6K-N). This observation might indicate that the MGE represents a non-permissive substrate for the pallial cells. This would fit with the superficial localization of the migratory cells in the subpallium, with little invasion into deep LGE domains, and no migration into MGE derivates.

Thus, our results confirm that, in E14.5 mouse embryos, VP neurons migrate tangentially through superficial regions of the subpallium and populate the BF. Also, we show that the medial subpallium has an attractive role for VP cells.

In order to further confirm the chemo-attractive effect of the medial telencephalon, small portions of wild-type (WT) MGE or LGE and GFP-VP were co-cultured in a Matrigel matrix (n=26; Fig. 6O,P). After 36 hours, MGE cells were detected migrating towards the neocortex, as previously seen (Marin et al., 2003), and in over half of the explants (14 of 26) GFP⁺ pallial cells were observed moving towards the MGE explant (Fig. 6O). Parallel experiments using LGE never showed that VP-derived cells migrate preferentially towards LGE explants (n=12; Fig. 6P). These results corroborate the existence of diffusible signals deriving from the MGE that stimulate the migration of VP cells.

Fgf8 is produced by the anterior neural ridge (ANR) and is involved in cell survival and telencephalic patterning (Crossley and Martin, 1995; Shimamura and Rubenstein, 1997; Crossley et al., 2001; Kawauchi et al., 2005; Storm et al., 2003; Storm et al., 2006; Bailey et al., 2006). Recently, it has been shown that Fgf8 is required for rostral pallium specification and for the generation of CR⁺ Cajal-Retzius cells (Zimmer et al., 2010). Moreover, its early expression in the rostral tip of the neural tube persists at later stages in the commissural plate of the telencephalon midline (Fig. 7A) (Crossley et al., 2001), suggesting that Fgf8 might also play a role in pallio-subpallial tangential migration.

Although a strong telencephalic phenotype in Fgf8 hypomorphic mutants has previously been reported (Storm et al., 2006), the BF was not carefully analyzed. First, we confirmed the progressive reduction of Fgf8 expression in the anterior telencephalon in Fgf8^neo/neo and Fgf8^neo/null mice (Fig. 7A-C). Then, we studied the distribution of ChBF in these progressive Fgf8 hypomorphic mice. Coronal sections of E16.5 brains were immunostained with anti-ChAT antibody and counterstained with Cresyl Violet. Whereas ChAT was detected in MB and HDB cells in WT brains (Fig. 7D,G), this was strongly reduced in Fgf ^neo/neo (Fig. 7E,H) and disappeared altogether in the Fgf8^neo/null mouse (Fig. 7F,I). Cresyl Violet showed that the BF region and medial subpallium were structurally preserved in the mutants (Fig. 7E,H). These results suggested that the lack of ChAT expression might be a consequence of migratory alterations resulting from the reduction of Fgf8 signaling. In order to experimentally demonstrate the attractive role of Fgf8, we grafted Fgf8-expressing cells to cultured slides where medial pallium was excised, and this recovered the pallio-subpallial cellular migration (n=9/12; Fig. 7J-M). Experiments with control cells did not show any increase in cell movements (n=5/5; Fig. 7N). Using VP explants in Matrigel cultures we demonstrated the diffusion of Fgf8 signal in the matrix and the negative effects of factors that block Fgf8 activity (SU5402) in pallial migrating cells (Fig. 7O,P). Altogether, these results implicate Fgf8 in the guidance of cells migrating from the VP towards the superficial area of the subpallium.

Tangential migration is defined as the free movement of cells unrelated to radial glia cells (Rakic, 1972). In the vertebrate telencephalon, tangential migration generally consists of cell movements from the subpallium to the pallium of GABAergic interneurons (De Carlos et al., 1996; Anderson et al., 1997; Lavdas et al., 1999; Marin and Rubenstein, 2003; Marin and Rubenstein, 2001; Cobos et al., 2001a; Olivier et al., 2001; Tuorto et al., 2003; Metin et al., 2007). Here, we describe a countercurrent pallio-subpallial tangential migration in the developing mouse telencephalon. Although this migratory pathway was previously observed in chick and mouse embryos (Fishell et al., 1993; Chapouton et al., 1999; Puelles et al., 2000), no detailed analysis was performed. Recently, it has been shown that Emx1-positive progenitors migrate into the dorsal ST and amygdala to generate GABAergic medium spiny and excitatory neurons, respectively (Cocas et al., 2009). Garcia-Moreno et al. reported that cells from the rostromedial telencephalon move ventrally, following a ventromedial route, and populate Broca's diagonal nuclei and Tu (Garcia-Moreno et al., 2007). These ventromedial migrating cells express reelin and possibly represent medial components of our pallio-subpallial migration. However, the cholinergic differentiation of migrating neurons from VP-derived amygdala (basal and lateral amygdaline complex) in the subpallial amygdala (central complex) (Medina et al., 2004) has not been reported previously. The functional importance of these cholinergic amygdaline cells is clear because stress reactions are regulated by changes in microRNAs acting through the cholinergic function in the central amygdala (Meerson et al., 2010), as well as by nicotine dependence mechanisms (for a review, see Markou, 2008).

Cells expressing Tbr1 and Pax6 were previously shown to migrate from the VP to superficial areas of the subpallium (Puelles et al., 2000; Carney et al., 2009). Also, we have demonstrated that Pax6⁺ cells migrate from the VP into the superficial layers of the BF in mouse embryos. These cells also express reelin, a key protein for neuronal migration and cortical layering (D'Arcangelo et al., 1995; Ogawa et al., 1995; Pearlman et al., 1998). However, reelin is widely expressed in chick embryos, suggesting that its function is not exclusively to control laminar development (Bernier et al., 2000). For example, reelin-activated cell adhesion mechanisms (Hoffarth et al., 1995) could mediate cellular interactions along the superficial migratory stream in the subpallium (Fig. 3I,J). In agreement with these data, a significant alteration in the ChBF has been reported in heterozygous reeler mice (Sigala et al., 2007).

Previous results have identified the existence of a lateral migratory stream through the pallio-subpallial boundary towards the olfactory cortex in mouse (De Carlos et al., 1996; Medina et al., 2004; Carney et al., 2009). Although VP migrating cells may partially follow this stream on their way to the BF, the pallio-subpallial migration must include more dorsal pallial regions because in most of our subpallial-grafted explants (Figs 6 and 7) we have excised the boundary territory from the donor to expose the VP to the grafted subpallium (abundant GFP-Tbr1⁺ cells and cortical histology at the host edge confirmed the VP origin of migrating cells).

Taken together, our results demonstrate that, in addition to the ventrodorsal tangential migration of interneurons and oligodendrocytes in the developing telencephalon of sauropsids, a dorsoventral migration of cholinergic neurons contributes to the cellular complexity of the basal prosencephalon (Fig. 8A).

The transcription factor Tbr1 is expressed in cortical glutamatergic neurons of the preplate and cortical layer 6 (Bulfone et al., 1995; Hevner et al., 2001). In this work, we analyzed the pallio-subpallial migration pathway in Tbr1^–/– mice. Reelin, Pax6 and ChAT expression were absent or strongly decreased in the mutant subpallium. Previous studies have reported that ChAT-positive cells in the subpallium originate from the MGE (Marin et al., 2000; Xu et al., 2008). As such, ChAT expression in the BF should not be affected in Tbr1^–/– mice as both the MGE and Nkx2.1 expression in the subpallium were normal. Furthermore, the fate map analysis of Nkx2.1 expression reported by Xu et al. demonstrated that Nkx2.1 lineage-related cells mainly form MS and VDB nuclei, whereas the HDB nuclei are practically devoid of this transcription factor [see figure 9D in Xu et al. (Xu et al., 2008)]. This observation, in accordance with our expression analysis, suggests a dual origin of ChBF: corticopetal cholinergic neurons of the HDB and MB from the VP (Tbr1⁺), and hippocampus projecting cholinergic neurons in VDB and MS nuclei from the medial subpallium (Nkx2.1⁺).

Finally, the LIM-homeobox gene Lhx8 is required for the development of cholinergic neurons in the mouse BF (Marin et al., 2000; Asbreuk et al., 2002; Zhao et al., 2003). Lhx8 mutant mice present alterations of the cholinergic system in the dorsal ST; however, neurons in the ventral ST and Tu are unaffected, possibly owing to their pallial origin (Zhao et al., 2003) (this work). Elshatory and Gan (Elshatory and Gan, 2008) reported that most of the cholinergic neurons in the MB express Islet1 and disappear when Islet1 is conditionally inactivated with Six3-cre, suggesting a subpallial origin of these cells. The results from Elshatory and Gan's study do not rule out the possibility of a VP origin of ChBF. First, as they did not perform independent clonal experiments, one cannot eliminate the possibility of alternative interpretations, such as the activation of Six3 in VP-derived cells or the involvement of compensatory mechanisms. Second, they report that some cholinergic neurons remained in the transgenic MB, which is in agreement with our results.

Our results indicate that pallial cells migrate due to signals released from the medial subpallium (MGE and Se). It was previously shown that the tangential ventrodorsal migration of interneurons is regulated by the simultaneous expression of chemo-repulsive and chemo-attractive signals in the POA and cortex, respectively (Marin et al., 2003; Wichterle et al., 2003). We believe that a similar mechanism could control pallio-subpallial migration, and our experiments indicate that Fgf8 signaling from the telencephalic midline is a key signal for this process. The morphogenetic activity of Fgf8 during neural tube development, both in the isthmic and anterior prosencephalic organizers (Crossley et al., 1996; Martinez et al., 1999; Storm et al., 2006), suggests a long-distance effect and supports an attractive role for this signal. Since the transcriptional regulation of Fgf8 expression in these signaling centers is relatively uncharacterized, further analysis of the possible control of Fgf8 and Fgf receptor (Fgfr) genes by transcription factors, such as Tbr1, in the BF needs to be undertaken. The reduction of cell migration by SU5402 in Matrigel explants proved the involvement of Fgfrs. Fgfrs are required for the normal morphogenetic activity of Fgf8 signaling during dorsal prosencephalic development (Paek et al., 2009; Stevens et al., 2010). Some cells in the developing BF express Fgfr, but because they have been related to the oligodendroglial lineage (Bansal et al., 2003; Furusho et al., 2011) a more precise analysis is required to explore their co-existence with pallio-subpallial migrating neurons.

Once they reach their destination in the BF, the VP cells seem to differentiate into cholinergic neurons. As the neurons in the claustrum, ChBF establish cortical connections and are involved in complex brain functions, such as cognition (reviewed by Sarter et al., 2005), as well as in the physiopathology of dementia in neurodegenerative diseases (Alzheimer's and Parkinson) and schizophrenia (reviewed by Bohnen and Albin, 2010). Amygdaline cholinergic cells are functionally involved in addiction and the stress response (Markou, 2008; Meerson et al., 2010).

The cholinergic hypothesis of geriatric memory dysfunction (Bartus et al., 1982) indicates a link between ChBF degeneration and cognitive disorders. Furthermore, the degeneration of ChBF is considered an early symptom in neurodegenerative diseases (for a review, see Hampel et al., 2008). Similarly, olfactory alterations have been identified as a risk factor and an early marker of neurodegenerative diseases (Tabert et al., 2005; Arnold et al., 2010). The early degeneration of granular neurons in the OB (Wesson et al., 2010) and of cholinergic neurons of the MB (Whitehouse et al., 1982) occur as two independent processes. Our present results demonstrate that these neurons have a common origin in ventral pallial progenitors and that they undergo an extensive migration from their origin to their final destination. Investigating the role of the amyloid precursor protein (APP) in this neuronal migration (Young-Pearse et al., 2010) and the specific trophic dependencies related to these common ontogenetic processes in the ChBF (Stewart and Appel, 1988) could represent new directions to further our knowledge of neurodegenerative-related cognitive disorder pathophysiology.

We thank C. Redondo, A. Torregrosa and O. Bahamonde for technical assistance, Dr Jonathan Jones for help in revising the manuscript, and Dr Arancha Botella for help with statistical analysis.