Patterning of the basal telencephalon and hypothalamus is essential for guidance of cortical projections

Axon growth and pathfinding is governed by the expression of guidance molecules in stereotypic patterns during development (Tessier-Lavigne and Goodman, 1996). Understanding the mechanisms that control wiring of the brain therefore not only requires the identification of guidance molecules and the ways they influence axon pathfinding, but also the factors that ensure the precise expression of these cues in the developing brain. In that respect, it has been suggested that the patterned expression of axon guidance molecules may be controlled by distinct sets of regulatory genes whose expression is defined by specific morphogenetic cues during a particular window of time (Wilson et al., 1993; Wilson et al., 1997). In agreement with this notion, recent studies have provided evidence supporting the role of several transcription factors in governing the patterned expression of axon guidance molecules in the developing brain and spinal cord of vertebrates (Logan et al., 1996; Shigetani et al., 1997; Ba-Charvet et al., 1998; Lauderdale et al., 1998; Matise et al., 1999; Tuttle et al., 1999).

The neocortex (isocortex) is the origin of two of the major axon tracts in the mammalian forebrain: (1) the corticothalamic projection, which originates primarily from neurons located in layer 6; and (2) the pyramidal tract, which originates from neurons in layer 5 (Jones, 1984). This latter tract comprises axons that course through the cerebral peduncle and innervate multiple subcortical targets, such as the superior colliculus, pons and spinal cord. The neocortex, on the other hand, receives most of its afferent projections from neurons of the dorsal thalamus, via the thalamocortical pathway. Because these major axonal pathways are essential for the complex integration of sensory-motor information that underlies vertebrate behaviors, understanding the mechanisms that control the development of these connections is central to the study of brain development.

Previous studies have identified pioneering axonal populations and specific cell domains that are candidates for guiding corticothalamic connections (de Carlos and O’Leary, 1992; Mitrofanis and Baker, 1993; Molnár and Blakemore, 1995; Métin and Godement, 1996; Molnár et al., 1998; Braisted et al., 1999; Molnár and Cordery, 1999; Tuttle et al., 1999; Auladell et al., 2000). By contrast, little is known about the cues that govern the guidance of major axon tracts in the cerebral peduncle as they course through the forebrain, although several molecules have been identified that cooperate in the final arrangement of corticospinal axons (Dahme et al., 1997; Cohen et al., 1998; Dottori et al., 1998; Castellani et al., 2000; Kullander et al., 2001a; Kullander et al., 2001b; Leighton et al., 2001; Yokoyama et al., 2001).

One forebrain region that appears to play a major role in the guidance of corticofugal and thalamocortical axons is the subpallium. Neuronal populations located in the subpallium send pioneering projections to the cortex and thalamus, and it has been suggested that these axons may form a scaffold system that guides corticofugal and thalamocortical axons, respectively (Métin and Godement, 1996; Molnár et al., 1998; Braisted et al., 1999; Tuttle et al., 1999). Moreover, several guidance molecules expressed in the ventral telencephalon have been shown to affect the development of corticothalamic and thalamocortical projections (Métin et al., 1997; Richards et al., 1997; Bagnard et al., 1998; Polleux et al., 1998; Braisted et al., 2000).

Nkx2-1 (Titf1 – Mouse Genome Informatics, MGI), a homeobox-containing gene member of the vertebrate Nkx family, is expressed in proliferative and postmitotic zones of the forebrain that give rise to a large region of the subpallium and to the hypothalamus (Price et al., 1992; Shimamura et al., 1995; Kohtz et al., 1998; Sussel et al., 1999). Homozygous mice containing a targeted disruption of the Nkx2-1 gene die at birth and have severe morphological abnormalities in the subpallium and hypothalamus (Kimura et al., 1996; Takuma et al., 1998; Sussel et al., 1999). In the telencephalon, loss of Nkx2-1 function causes a molecular transformation of the medial ganglionic eminence (MGE), which instead of producing its normal derivatives (e.g. globus pallidus), gives rise to LGE-like derivatives (e.g. striatum) (Sussel et al., 1999). It is conceivable that the abnormal patterning of the basal telencephalon in Nkx2-1-deficient mice may cause a change in the balance of guidance cues normally expressed in this region and, consequently, corresponding defects in the guidance of cortical projections. Thus, the selective nature of the basal telencephalon in Nkx2-1 mutants provides a unique opportunity to examine the requirement of this region in providing cues that influence axon guidance.

We present evidence that, in Nkx2-1 mutants, layer 5 cortical axons follow an abnormal course as they exit the basal telencephalon and most of them are unable to reach the cerebral peduncle. By contrast, guidance of corticothalamic and thalamocortical axons is largely normal in the absence of Nkx2-1 function, indicating that axonal projections arising from different cortical layers are guided through at least partially independent mechanisms. In addition, when some pyramidal tract axons leave the telencephalon in Nkx2-1 mutants, they follow an aberrant course towards the chiasmatic region, indicating that the region connecting the telencephalon and the diencephalon also constitutes a crucial decision point in the pathway of pyramidal tract axons. In vitro experiments suggest that the basal telencephalon and the rostrobasal hypothalamus contain an activity that inhibits the growth of corticofugal axons. Analysis of the expression of candidate guidance molecules in Nkx2-1 mutants suggests that Slit2 appears to be at least a component of this repulsive activity.

A mouse mutant strain with a null allele of Nkx2-1 (a gift from S. Kimura) was used in this study. This mouse strain was maintained by backcrossing to C57BL/J6 mice for more than ten generations. In addition, green fluorescent protein (GFP)-expressing transgenic mice maintained in a CD1 background (a gift of A. Nagy) were used for explant cultures experiments. For staging of embryos, midday of the vaginal plug was considered as embryonic day 0.5 (E0.5). PCR was performed as described by Sussel et al. (Sussel et al., 1999) to genotype offspring resulting from Nkx2-1 heterozygous matings. Mouse colonies were maintained at UCSF in accordance with NIH and UCSF guidelines.

Embryos were obtained by dissection of pregnant mice at various stages of development. Embryos younger that E15.5 were anesthetized by cooling and fixed overnight in 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS). Embryos were then cryoprotected in 30% sucrose in PBS, embedded in OCT compound, and cut frozen in the transverse or sagittal plane on a cryostat at 10 μm. E18.5 embryos were anesthetized by cooling and perfused with 4% PFA. Brains were removed and postfixed for 3 hours, cryoprotected in 30% sucrose in PBS, and cut frozen in the transverse plane on a sliding microtome at 40-50 μm. Immunohistochemistry in cryostat sections or in free-floating sections was performed as described before (Marín et al., 2000). The following primary rabbit polyclonal antibodies were used: anti-NKX2-1 (Biopat Immunotechnologies, Caserta, Italy), diluted 1:2000; anti-calbindin (Swant, Bellinzona, Switzerland), diluted 1:5000; anti-GABA (Sigma), diluted 1:2000; and anti-tyrosine hydroxylase (Pel-Freeze), diluted 1:1000.

In situ hybridization experiments were performed using ³⁵S-riboprobes on 10 μm frozen sections as described previously (Bulfone et al., 1993). The cDNA probes used in this study were for: Dlx2 and Dlx5 (Liu et al., 1997); Nkx2-1 (Shimamura et al., 1995); Nkx2-4, kindly provided by R. Harvey; Pax6, kindly provided by P. Gruss; Sim1, kindly provided by C. Fan; SF1, kindly provided by H. Ingraham; EphA4, kindly provided by A. Nieto; EphA7, kindly provided by A. Wanaka; EphB1, EphB2 and EphB3, kindly provided by M. Henkemeyer; ephrinA2, kindly provided by D. Wilkinson; ephrinA5, kindly provided by U. Drescher; Sema6a, kindly provided by W. D. Snyder; and Sema3a, Sema3c, Sema3f, Slit1, Slit2, Robo1, Robo2 and netrin 1, kindly provided by M. Tessier-Lavigne.

E13.5, E14.5 and E18.5 Nkx2-1 homozygous and wild-type littermate embryos were perfused with 4% PFA and their brains removed and kept in fixative. Crystals of the axonal tracer 1,1′-dioctodecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) (Molecular Probes, Eugene, OR) were placed in the neocortex to label anterogradely corticofugal axons. In other cases, DiI crystals were placed in the dorsal thalamus or in the basal telencephalon to label retrogradely corticofugal neurons. Crystals of similar shape and size (100-200 μm in diameter) were used in homozygous and wild-type littermate embryos. Brains were kept in 4% PFA at room temperature for 4-8 weeks to allow the DiI to diffuse. Subsequently, brains were embedded in 4% low melting point agarose and 100 μm transversal sections were cut in a vibratome. Sections were counterstained with 50 μg/ml Hoechst 33342 (Molecular Probes).

Embryos (E12.5-14.5) were removed by Cesarean section from timed-pregnant CD1 wild-type mice mated by GFP-expressing transgenic mice. Brains were removed in ice-cold Krebs buffer as described before (Marín et al., 2000). To isolate the cortex, brains were dissected in half along the midline and the telencephalon separated from the rest of the forebrain. Pieces of dorsal cortex were cut with a fine tungsten needle. To isolate basal telencephalon and hypothalamus, brains were embedded in 4% low melting point agarose in PBS and 250 μm transversal sections were cut with a vibratome. Sections at the appropriate levels were collected and the preoptic area and hypothalamus isolated from them and dissected into small pieces.

Collagen was prepared from adult rat-tail, as described before (Lumsden and Davies, 1986). To set up the co-cultures, 25 μl of collagen solution was pipetted onto the bottom of four-well dishes (Nunc, Roskilde, Denmark) and allowed to gel for about 45 minutes. Explants were then placed onto this base and 25 μl of collagen were added on top. In most cases, collagen co-cultures consisted of three pieces (wild-type basal telencephalon or hypothalamus, GFP-expressing dorsal cortex and wild-type dorsal cortex) arranged sequentially from left to right and separated approximately by 300 μm. After a period of 45 minutes, to allow the collagen to gel, Neurobasal/B-27 medium (Gibco BRL, Life Technologies, Gaithersburg, MD) was added. Explants were cultured in a sterile incubator (37°C, 5% CO₂). In collagen cultures with isolated pieces of dorsal cortex, growth of cortical axons was similar from all sides of the explant.

The telencephalon consists of pallial (cortical) and subpallial (subcortical) subdivisions. In this paper, we also refer to the subpallium as the ‘ventral telencephalon’ because it expresses molecular markers that are characteristic of ventral parts of the central nervous system. It should be noted, however, that topologically the subpallium is in fact rostral to the pallium (Cobos et al., 2001). The term ‘basal telencephalon’ is used to designate only the most ventral aspect of the subpallium, which includes derivatives from the anterior entopeduncular region and the preoptic area.

Nkx2-1 is expressed in the developing mouse forebrain from approximately embryonic day 8.5 (E8.5) (Price et al., 1992; Shimamura et al., 1995; Kohtz et al., 1998; Sussel et al., 1999). After neural tube closure, expression of NKX2-1 is restricted to two domains in the forebrain: ventral regions of the subpallium and hypothalamus (Fig. 1A). NKX2-1 is not found in cortical progenitors or in postmitotic cells derived from the cortex (Fig. 1A-C). From E12.5, corticofugal axons directed towards subcortical targets leave the cortex and transverse the developing striatum on their route to the diencephalon (de Carlos and O’Leary, 1992; Auladell et al., 2000). At this stage, NKX2-1 is expressed in both proliferative and postmitotic cells of the subpallium in a domain that includes the anlage of the globus pallidus [medial ganglionic eminence (MGE)], part of the septum, anterior entopeduncular region (AEP) and preoptic area (POa) (Fig. 1B; data not shown). In their trajectory, corticofugal axons grow through the striatum and leave the telencephalon in close proximity to the developing globus pallidus (Métin and Godement, 1996), which expresses NKX2-1. By contrast, corticofugal axons avoid entering the most ventral regions of the telencephalon, which are defined by the combined expression of NKX2-1 and Shh (Sussel et al., 1999; Olivier et al., 2001). At around E14, corticofugal axons form a tight bundle of fibers as they exit the telencephalon through the internal capsule, which is surrounded by regions of NKX2-1 expression (Fig. 1D,E). From here, cortical axons destined to the thalamus make a sharp turn caudalwards and follow straight across the ventral thalamus into the dorsal thalamus. A reciprocal route is followed by most thalamocortical axons (Métin and Godement, 1996; Molnár et al., 1998; Braisted et al., 1999; Auladell et al., 2000). By contrast, cortical axons directed towards other subcortical structures, such as the superior colliculus, pons and spinal cord, maintain their ventral course as they enter the diencephalon and later bend caudally to form the cerebral peduncle. These axons course caudally following a route parallel to the alar/basal boundary, avoiding the basal hypothalamic region (de Carlos and O’Leary, 1992), which is defined by NKX2-1 expression (Fig. 1A,C,F,G). Thus, developing corticofugal projections extend through the forebrain in close relation to regions defined by the mantle expression of NKX2-1 (Fig. 1H).

Our previous work has provided evidence that disruption of Nkx2-1 function results in re-specification of the ventral telencephalon to a more dorsal fate (Sussel et al., 1999). Although previous studies have shown hypothalamic defects in Nkx2-1 mutants (Kimura et al., 1996; Takuma et al., 1998), it has not been established whether these defects are a consequence of abnormal dorsoventral patterning. Thus, we studied the hypothalamic expression of genes that define distinct dorsoventral positions in the hypothalamus.

At E11.5, a structure that morphologically resembles a small hypothalamus is found in Nkx2-1 mutants (Fig. 2A-H; data not shown). To determine whether the mutant hypothalamus has molecular characteristics similar to those of the wild-type hypothalamus, we studied the expression of molecular markers whose forebrain expression is restricted to the Nkx2-1 domain, such as the homeobox gene Nkx2-4, the orphan nuclear receptor SF1 (Nr5a1 – Mouse Genome Informatics) and Shh. With the exception of its most ventral aspect, expression of these factors is not detectable in the mutant hypothalamus (Fig. 2A,E,D,H; data not shown). We next analyzed the expression of Pax6, a paired-homeobox gene that has been shown to play an opposing role to Nkx genes in dorsoventral patterning (Ericson et al., 1997; Briscoe et al., 1999; Sussel et al., 1999; Briscoe et al., 2000; Sander et al., 2000). Pax6 expression is restricted to alar domains in the forebrain of wild-type embryos, and therefore it is normally excluded from the hypothalamus (Fig. 2B,D) (Stoykova et al., 1996). In Nkx2-1 mutants, Pax6 expression is expanded ventrally into the hypothalamus, more prominently in its posterior aspect (Fig. 2F,H). In addition, the expression of the bHLH-PAS transcription factor Sim1 is also altered in Nkx2-1 mutants. Whereas Sim1 expression is restricted to the alar domain of the rostral diencephalon in wild-type embryos, Sim1 expression is expanded ventrally in Nkx2-1 mutants (Fig. 2C,D,G,H). Nevertheless, neither Pax6 nor Sim1 are expressed in the most ventral aspect of the mutant hypothalamus (data not shown), suggesting that only the most ventral domain of the hypothalamus maintains some normal molecular properties in the absence of Nkx2-1 function. Thus, the hypothalamus shows a ventral-to-dorsal shift in its molecular properties in the absence of Nkx2-1 function.

Histological analysis at E18.5 supports the evidence for the loss of ventral structures and expansion of dorsal components in the hypothalamus of Nkx2-1 mutants (Fig. 2I-L). For example, calbindin immunohistochemistry revealed that the mammillary hypothalamus is greatly reduced or absent in Nkx2-1 mutants, whereas the tuberomammillary hypothalamus has an increased density of calbindin-positive cells (Fig. 2J,L). In addition, the proportion of GABAergic neurons found in the hypothalamus of Nkx2-1 mutants is increased in relation to the non-GABAergic cell populations (Fig. 2I,K). This observation is consistent with the finding that expression of Dlx genes is preserved in the hypothalamus of Nkx2-1 mutants (data not shown). In sum, as in the ventral telencephalon, loss of Nkx2-1 function in the hypothalamus does not result in the complete loss of this territory, but rather in a transformation of its molecular and morphological characteristics.

Analysis of patterning defects present in Nkx2-1 mutants suggests that the ventral telencephalon and the hypothalamus of these mice have acquired molecular characteristics that are different from the equivalent wild-type territories. As corticofugal and thalamocortical projections grow in close relation to the territories patterned by NKX2-1 (Fig. 1), we wondered whether the transformation of these domains could affect the course of these tracts in Nkx2-1 mutants.

To study cortical projections in Nkx2-1 mutants, DiI crystals were inserted into the parietal cortex of fixed brains from E14.5 mice. In wild-type brains, corticofugal axons coursed through the internal capsule and arrived at the junction between the telencephalon and the diencephalon (n=8, Fig. 3A). Some of the leading axons at this stage continued into the dorsal thalamus, whereas the remaining joined the developing cerebral peduncle (data not shown). DiI deposits in the dorsal cortex of Nkx2-1 mutants consistently labeled a large number of corticofugal axons (n=8). In Nkx2-1 mutants, cortical axons grew normally into the internal capsule (data not shown), suggesting that guidance of corticofugal axons is normal to this point. At the telencephalic-diencephalic boundary, however, the majority of corticofugal axons did not maintain their original course, but instead extended towards the pial surface of the telencephalon (Fig. 3E, arrow). The remaining labeled axons entered the diencephalon normally (Fig. 3E, arrowhead) and grew into the dorsal thalamus (data not shown).

Analysis of DiI deposits in the parietal cortex of Nkx2-1 mutant mice at E18.5 revealed that the guidance defect of corticofugal projections is even more prominent at this stage. As in earlier stages, corticofugal axons reached the internal capsule normally in Nkx2-1 mutants. Moreover, corticothalamic axons followed a normal route at the telencephalic-diencephalic boundary, turning caudally after leaving the telencephalon and growing longitudinally through the ventral thalamus to reach the dorsal thalamus (n=12; Fig. 3B-D,F,H,I). By contrast, cortical axons directed to the cerebral peduncle failed to enter the diencephalon and instead extended toward the pial surface (Fig. 3B-D,F,H,I). In most cases, labeled axons in this ectopic peduncle stopped at the surface of the telencephalon, while a few labeled axons extended inside the telencephalic stalk or caudally into the amygdala (data not shown). By contrast, DiI crystal deposits in the cortex of Nkx2-1 mutants revealed a normal pattern of retrogradely labeled cells in the dorsal thalamus (Fig. 3B,C,H), suggesting that thalamocortical connections develop normally in Nkx2-1 mutants. Immunohistochemical detection of 5-HT and calretinin, two markers of thalamocortical fibers at this stage, confirmed this finding (data not shown).

To confirm that the ectopic bundle consists of axons originally directed towards the cerebral peduncle (pyramidal tract axons, which originate primarily in cortical layer 5) (Jones, 1984) and not of corticothalamic axons (which originate in cortical layer 6) (Jones, 1984), we used retrograde DiI labeling to identify the neurons that extend their axons into this ectopic projection. We inserted small DiI crystals in the surface of the telencephalic stalk of Nkx2-1 mutants, in the region of the ectopic peduncle, and found that retrogradely labeled cells were primarily confined to cortical layer 5 (n=3) (Fig. 3G). DiI placements in the dorsal thalamus, however, confirmed that corticothalamic projections are essentially normal in Nkx2-1 mutants and that they originate from cortical layer 6, as in control mice (n=5) (Fig. 4). Thus, corticothalamic and thalamocortical projections develop normally in the absence of Nkx2-1 function, whereas pathfinding of pyramidal tract axons is severely disrupted in the telencephalon of Nkx2-1 mutants.

Despite the prominent defect found in the guidance of layer 5 projections at the level of the telencephalic stalk, DiI deposits in the parietal cortex labeled a small bundle of axons in the cerebral peduncle of some Nkx2-1 mutants (n=7/12; Fig. 3H). These results suggest that, at least in some cases, some cortical axons are able to enter the cerebral peduncle normally. In wild-type brains, corticofugal axons in the cerebral peduncle grew ventrally through the peduncular region of the hypothalamus until they reached the diencephalic alar/basal boundary, where they turned caudalwards towards the mesencephalon (Fig. 5A,C; see also Fig. 1H). By making this 90° turn, pyramidal tract axons avoid entering the chiasmatic region of the hypothalamus. In Nkx2-1 mutants, however, most pyramidal tract cortical axons failed to turn caudally and, instead, grew into the chiasmatic region (Fig. 5D,F). Compared with wild-type brains, very few cortical axons reached the mesencephalon and rhombencephalon in the mutants (Fig. 5B,E). Thus, in Nkx2-1 mutant mice, most layer 5 cortical projections either fail to leave the telencephalon (Fig. 3) or fail to turn caudally once they reach the hypothalamus (Fig. 5).

The previous experiments suggest that abnormal guidance of layer 5 cortical projections occurs in Nkx2-1 mutants because patterning defects in the basal telencephalon and hypothalamus lead to alterations in the signals that direct this axon tract. In Nkx2-1 mutants, cortical axons invade the telencephalic stalk (the most ventral aspect of the subpallium, which includes derivatives from the AEP and POa) and the chiasmatic region, two regions that cortical axons normally do not enter. These observations suggest that the basal telencephalon and the hypothalamus may contain a repulsive or non-permissive activity for cortical axons, and that the loss of this activity in Nkx2-1 mutants may divert these axons from their normal course. To test this hypothesis, we carried out collagen co-culture assays using explants from E13.5-E14.5 mice, ages at which cortical axons are extending though the subpallium and diencephalon in vivo. Wild-type explants of the dorsal cortex obtained from GFP-expressing transgenic mice were co-cultured adjacent to explants of the dorsal cortex and explants of the basal telencephalon, both obtained from littermates (Fig. 6A). In most cases, cortical axons grew into the adjacent cortical explant, but avoided growing into the basal telencephalic explant (n=26/30; Fig. 6B). Co-cultures using explants from the hypothalamus (Fig. 6C), instead of the basal telencephalon, showed similar results (n=30/38; Fig. 6D). Most cortical explants cultured alone in collagen as a control showed symmetrical axonal outgrowth (n=10/13; data not shown). Therefore, both the basal telencephalon and the hypothalamus contain an activity that prevents the ingrowth of cortical axons.

As the territories patterned by Nkx2-1 contain an activity that inhibits the growth of cortical axons, it is conceivable that loss of this chemorepellent activity in Nkx2-1 mutants may account for the defects observed in the guidance of layer 5 cortical axons. We employed a candidate molecule approach to identify genes encoding axon guidance proteins whose expression is altered in the basal telencephalon of Nkx2-1 mutants. First, we analyzed the expression different members of the semaphorin family, which has been previously described to affect the guidance of cortical and thalamic axons (Bagnard et al., 1998; Polleux et al., 1998; Leighton et al., 2001). Expression of Sema3a, Sema3c, Sema3f and Sema6a was maintained in the basal telencephalon of Nkx2-1 mutants (Table 1). Next, we examined the expression of netrin 1, because this molecule has been implicated in the guidance of both corticofugal and thalamocortical axons (Métin et al., 1997; Richards et al., 1997; Braisted et al., 2000). In situ hybridization analysis showed that expression of netrin 1 is also preserved in the subpallium of Nkx2-1 mutants (Table 1). Then, we analyzed the expression of Eph tyrosine kinase receptors and their ligands, the ephrins, as previous studies have demonstrated a role for these molecules in the guidance of cortical axons (Henkemeyer et al., 1996; Orioli et al., 1996; Dottori et al., 1998; Kullander et al., 2001a; Kullander et al., 2001b; Leighton et al., 2001; Yokoyama et al., 2001). Interestingly, expression of EphB2, EphB3 and ephrinA5 was reduced in the forebrain of Nkx2-1 mutants (Table 1). Nevertheless, analysis of layer 5 cortical projections in EphB2/EphB3 double mutants (n=8; O. M., M. Henkemeyer and J. L. R. R., unpublished) and in ephrinA5 mutants (n=4; O. M., J. Holmberg, J. Frisen and J. L. R. R., unpublished) revealed no obvious defects. Thus, loss of EphB2/EphB3 or ephrinA5 alone does not seem to recapitulate the defects found in Nkx2-1 mutant mice.

Slit proteins constitute another family of extracellular matrix molecules that possess chemorepulsive activity for growing axons in a variety of systems (Brose and Tessier-Lavigne, 2000). Like semaphorins, Eph/ephrins and netrin, Slits are also expressed in the forebrain in a pattern that may affect the growth of cortical axons (Yuan et al., 1999; Marillat et al., 2001) (Bagri et al., 2002). Recently, it has been reported that loss of Slit2 function in mice produces a dramatic defect in the guidance of cortical axons (Bagri et al., 2002). Interestingly, this defect is strikingly similar to that found in Nkx2-1 mutants, as in the absence of Slit2 most cortical projections destined to the cerebral peduncle course abnormally to the ventral surface of the telencephalon. Interestingly, we found that expression of Slit1 and Slit2 in the forebrain of Nkx2-1 mutants was significantly altered. The most remarkable finding was that Slit2 expression in the AEP and POa domains of the ventral telencephalon was almost completely absent in the mutant (Fig. 7B,F). The alteration in Slit1 expression was less unexpected because, in line with other markers that expand ventrally in the telencephalon (see Fig. 2) (Sussel et al., 1999), Slit1 expression was found throughout the entire VZ of the subpallium in Nkx2-1 mutants, including the POa, a region that normally lacks Slit1 expression (Fig. 7A,E). Moreover, the number of postmitotic Slit1-expressing cells in the mantle of the POa was increased compared with wild-type mice (Fig. 7A,E).

Analysis of the expression of Slit1 and Slit2 in the hypothalamus showed similar results; expression of Slit1 was increased, whereas expression of Slit2 was severely reduced compared with wild-type littermates (Fig. 7C,D,G,H). Thus, the telencephalic stalk region and the hypothalamus of Nkx2-1 mutants lack expression of Slit2, but have increased Slit1 expression. In situ hybridization analysis revealed that there are no obvious differences in the expression of the Slit receptors Robo1 and Robo2 in the cortex of wild-type and Nkx2-1 mutant mice (data not shown), showing that cortical neurons should be competent to respond to Slit in Nkx2-1 mutants.

Our previous experiments show that efferent forebrain fibers tha contribute to the cerebral peduncle (e.g. layer 5 cortical projections) are misrouted in the basal telencephalon and hypothalamus of Nkx2-1 mutants, and that loss of Slit2 expression is correlated with this phenotype. We reasoned that afferent forebrain fibers could also be affected, particularly as Slit2 function is required for the guidance of rostral-growing fibers in the hypothalamus, such as the mesotelencephalic dopaminergic projections (Bagri et al., 2002). Thus, we studied the distribution of dopaminergic fibers using antibodies against tyrosine hydroxylase (TH), the rate-limiting enzyme for catecholamine synthesis. In both control and Nkx2-1 mutant mice, dopaminergic fibers from the substantia nigra/ventral tegmental area coursed longitudinally in the medial forebrain bundle, which lies ventromedial to the cerebral peduncle. In normal mice, dopaminergic fibers turned dorsally as they approach the telencephalic stalk, and then entered the subpallium. In Nkx2-1 mutants, however, a large number of TH-positive fibers failed to enter the telencephalon and instead coursed directly into the chiasmatic region, a region that they normally avoid (Fig. 8). This is the same error that the efferent corticofugal fibers make (Fig. 5D, Fig. 9B). Although many axons take an aberrant path, many are successful at entering the telencephalon and innervating the striatum in Nkx2-1 mutants (data not shown) (Sussel et al., 1999). Thus, Nkx2-1 function is required for the proper trajectories of major efferent and afferent fibers of the cerebral peduncle.

In this study, we provide evidence that appropriate patterning of the basal telencephalon and hypothalamus is required for the establishment of some of the major efferent and afferent projections that interconnect the telencephalon with the rest of the brain (Fig. 9A,B). Loss of Nkx2-1 homeobox gene function causes a molecular transformation of the basal telencephalon and hypothalamus, which results in the enlargement of the LGE and the alar prethalamic diencephalon, respectively (present study) (Sussel et al., 1999). Consequently, most layer 5 cortical projections fail to project out of the telencephalon, and instead course abnormally to the surface of the telencephalic stalk region (AEP). Moreover, most of those corticofugal axons that are able to enter the hypothalamus grow towards the chiasmatic region, instead of caudally (Fig. 9A,B). Likewise, many afferent dopaminergic fibers show similar pathfinding errors in the hypothalamus (Fig. 9A,B).

In contrast to the cerebral peduncle defects, corticothalamic and thalamocortical projections develop normally in the absence of Nkx2-1 function. Together with our in vitro observations that both the basal telencephalon and the hypothalamus contain an activity that prevents the growth of cortical axons, these findings suggest that loss of a chemorepellent activity in Nkx2-1 mutants may account for the guidance defects of layer 5 cortical axons. This notion is supported by the fact that the basal telencephalon and hypothalamus of Nkx2-1 mutants lack expression of Slit2, an extracellular matrix molecule that has been recently demonstrated to play a key role in the guidance of cortical projections (Bagri et al., 2002). Insights from these experiments are discussed below.

Previous studies have analyzed the early development of corticofugal and thalamocortical projections, and have identified pioneering axonal populations, potential intermediate targets and critical guidance decision points for these axons (Blakemore and Molnár, 1990; Ghosh et al., 1990; Shatz et al., 1990; de Carlos and O’Leary, 1992; Miller et al., 1993; Mitrofanis and Baker, 1993; Molnár and Blakemore, 1995; Métin and Godement, 1996; Molnár et al., 1998; Braisted et al., 1999; Molnár and Cordery, 1999; Tuttle et al., 1999; Auladell et al., 2000; Braisted et al., 2000). One general insight derived from these studies is that multiple cell domains distributed along the length of each of these axonal pathways are involved in the guidance of the axons to their target.

It has been previously shown that the subpallium plays a prominent role in the guidance of corticofugal and thalamocortical axons. Specifically, the mantle regions of the LGE and MGE act as intermediate targets for corticofugal and thalamocortical axons, respectively (Métin and Godement, 1996). Moreover, the initial trajectories of corticofugal and thalamocortical axons into the subpallium appear to be pioneered by transient projections from the ganglionic eminences. For example, projections from the mantle of the MGE to the dorsal thalamus seem to constitute a substrate for the growth of thalamocortical axons into the telencephalon (Mitrofanis and Baker, 1993; Métin and Godement, 1996; Molnár et al., 1998; Braisted et al., 1999; Tuttle et al., 1999). In addition, a projection from the mantle of the LGE to the cortex may be used by corticofugal fibers as a guiding axonal scaffold in their growth toward the internal capsule (Métin and Godement, 1996). Thus, the mantle regions of the LGE and the MGE are permissive territories for the growth of corticofugal and thalamocortical axons.

In the present study, we have identified a subpallial telencephalic subdomain as an important territory in the guidance of cortical axons destined to the cerebral peduncle (pyramidal tract axons). This telencephalic domain is located just outside the mantle of the MGE [the globus pallidus; VTel1 domain of Tuttle et al. (Tuttle et al., 1999)], and consists of the mantle regions of the AEP and POa. Our in vitro experiments demonstrated that this region constitutes a non-permissive substrate for cortical axons. Thus, the subpallium consists of two different domains that are essential for the guidance of corticofugal and thalamocortical projections: (1) a permissive territory that these axons can traverse (the mantle zone of the LGE and MGE); and (2) a non-permissive territory that these axons normally avoid (the mantle zone of the AEP and POa) (Fig. 9D).

Another novel finding derived from the analysis of Nlx2-1 mutants is that corticofugal axons derived from different cortical layers respond differently to cues present in the basal telencephalon. Thus, whereas a large number of layer 5 axons fail to reach the diencephalon in Nkx2-1 mutants, layer 6 axons follow a normal path to the dorsal thalamus. This suggests that different mechanisms regulate the guidance of layer 5 and layer 6 axons as they course through the ventral telencephalon. Alternatively, layer 6 axons may be equally sensitive to cues present in the telencephalic stalk area, but additional mechanisms may compensate for their loss in Nkx2-1 mutants. For example, it has been hypothesized that corticothalamic and thalamocortical axons use each other as guidance scaffolds (Molnár et al., 1998). As thalamocortical projections develop normally in Nkx2-1 mutants, their presence in the internal capsule may be sufficient to ensure the guidance of cortical projections to the thalamus.

It has been previously suggested that the mantle of the MGE contains a cell domain that constitutes a permissive substrate for the growth of thalamocortical axons [VTel1 domain (Braisted et al., 1999; Tuttle et al., 1999); perireticular nucleus (Mitrofanis and Baker, 1993; Métin and Godement, 1996; Molnár et al., 1998)]. This domain is defined by the expression of Nkx2-1 and netrin 1 (Tuttle et al., 1999), and seems to facilitate the extension of thalamic axons into the telencephalon by two different mechanisms: (1) by providing a growth-promoting environment to thalamic axons, which may consist at least in part of netrin 1 activity (Braisted et al., 2000); and (2) by providing an axonal scaffold through its transient projections to the dorsal thalamus (Mitrofanis and Baker, 1993; Métin and Godement, 1996; Molnár et al., 1998; Braisted et al., 1999; Tuttle et al., 1999). As this domain is located in the region of the basal telencephalon patterned by Nkx2-1, it was anticipated that these neurons would not be present in Nkx2-1 mutants. However, a similar cell domain is present in Nkx2-1 mutants, as identified by location, netrin 1 expression and projections to the dorsal thalamus (O. M., J. B., L. P. and J. L. R. R., unpublished), suggesting that patterning of this cell domain is not dependent on Nkx2-1 function. The correlation between the presence of these cells and normal pattern of thalamocortical projections in Nkx2-1 mutants reinforces the notion that this cell domain may play an important role in the guidance of thalamocortical axons (Braisted et al., 1999; Tuttle et al., 1999; Braisted et al., 2000).

Analysis of Nkx2-1 mutants shows that the hypothalamus is also a crucial region for the guidance of pyramidal tract axons. The hypothalamus, which contains a non-permissive activity for the growth of cortical axons, is normally avoided by pyramidal tract axons in their route toward the mesencephalon. The presence of this repulsive activity may be required to ensure that pyramidal tract axons turn appropriately into a longitudinal path roughly parallel to the alar/basal boundary, avoiding entering the basal hypothalamus. Interestingly, in vitro experiments suggested that the hypothalamus is also repulsive for thalamocortical axons, preventing them from growing excessively ventrally in their route towards the telencephalon (Braisted et al., 1999). Transformation of the hypothalamus in Nkx2-1 mutants, however, does not impair thalamocortical projections, perhaps owing to the endurance of other repellent cues for thalamocortical axons in the mutant hypothalamus.

As discussed above, corticofugal axons extend along several telencephalic and diencephalic domains in their route towards the midbrain, approaching some territories and avoiding others. These observations suggest that diffusible or contact-mediated guidance cues exist in several potential guidance decision points within the forebrain to control the pathfinding of cortical axons. The molecular nature of these signals is only beginning to be revealed.

The early steps in the guidance of neocortical axons, including both corticothalamic and pyramidal tract axons, appear to be controlled by common mechanisms. For example, two class 3 semaphorin proteins, Sema3a and Sema3c, regulate the initial extension of cortical axons toward the cortical white matter. Semaphorins appear to determine the initial direction of cortical axons through a complex mechanism involving repulsion from the marginal zone and attraction from the subventricular zone (Bagnard et al., 1998; Polleux et al., 1998). Corticofugal axons are subsequently attracted laterally towards the internal capsule by a mechanism that involves netrin 1 (Métin et al., 1997; Richards et al., 1997), which is prominently expressed in the mantle of the LGE and MGE (Serafini et al., 1996; Livesey and Hunt, 1997; Métin et al., 1997; Richards et al., 1997).

As shown here, a crucial decision point in the guidance of corticofugal fibers in located at the telencephalic/diencephalic boundary. Corticofugal fibers enter the peduncle systematically at the same location, and subsequently split into the corticothalamic and pyramidal tract projections. Recently, genetic analysis of mice carrying loss-of-function alleles for Slit1, Slit2 or both genes has demonstrated that these proteins are involved in guiding the exit of corticofugal fibers from the telencephalon by preventing their entry into the basal telencephalon (Bagri et al., 2002). Loss of Slit1 function does not result in a profound abnormality in the guidance of corticofugal axons, whereas loss of Slit2 produces a phenotype that is rather similar to that found in Nkx2-1 mutants. Thus, in Slit2 mutant mice, corticofugal axons fail to enter the peduncle normally, and instead follow an abnormal course toward the surface of the telencephalon (Bagri et al., 2002). Of note, although Slit2 is expressed in the ventricular zone, analysis of Slit2 mutant mice suggests that Slit2 protein is likely to be localized in the mantle zone, where it can exert an effect on the guidance of cortical projections (Bagri et al., 2002). It has been proposed that Slit2 protein may become localized to the mantle zone by virtue of being expressed by radial glial cells, whose long processes may deliver Slit2 protein to more superficial regions (Bagri et al., 2002).

Two important differences exist between the phenotypes observed in Slit2 and Nkx2-1 mutant mice. First, a population of neocortical axons aberrantly crosses the midline at the level of the anterior commissure in Slit2 mutants, indicating that Slit proteins are also involved in guidance across the midline (Bagri et al., 2002). This phenotype has not been observed in Nkx2-1 mutant mice, where the misguided corticofugal axons stop ipsilaterally close to the midline but do not cross it. Second, although to a minor extent, corticothalamic projections also appear to follow an abnormal course in Slit2 mutant mice (Bagri et al., 2002), they are normal in Nkx2-1 mutants. Although further genetic experiments may be required to clarify these differences, the ventral expansion of Slit1 expression in the telencephalon of Nkx2-1 mutants could explain them. Thus, ectopic Slit1 expression at the midline of the basal telencephalon may prevent cortical axons from approaching the midline. Moreover, the increased levels of Slit1 expression in the basal telencephalon may help corticothalamic axons to follow their normal path, but may not prevent pyramidal axons from taking an abnormal course. If this hypothesis were correct, these results would indicate that corticothalamic and pyramidal tract axons respond differentially to Slit1 and Slit2 or that both Slit proteins have different biological properties in vivo (e.g. different diffusion capabilities). In agreement with this notion, it has been suggested that different Slit cleavage fragments may have different cell association characteristics and therefore different extents of diffusion and binding properties in vivo (Brose et al., 1999). Accordingly, recent in vitro experiments have demonstrated that different Slit2 proteolytic fragments elicit different responses in different types of axons (Ba-Charvet et al., 2001; Battye et al., 2001; Chen et al., 2001).

Additional evidence suggesting that the guidance abnormalities found in Nkx2-1 mutants are at least partially due to loss of Slit2 expression arises from the comparison of hypothalamic defects in Nkx2-1 and Slit2 mutants. As in the telencephalon, the hypothalamus of Nkx2-1 mutants lacks Slit2 and has an expanded Slit1 expression domain. In both Slit2 and Nkx2-1 mutants, pyramidal tract fibers follow abnormal paths as they approach the hypothalamus (this study) (Bagri et al., 2002). Similarly, ascending dopaminergic projections make similar pathfinding errors as they course through the diencephalon in both Slit2 and Nkx2-1 mutants (this study) (Bagri et al., 2002). By contrast, thalamocortical fibers course ventrally into the hypothalamus of Slit2 mutants (Bagri et al., 2002), a defect that was not observed in Nkx2-1 mutants. It is possible that this abnormality may be compensated in Nkx2-1 mutants by the increased levels of Slit1 expression. In conclusion, the results obtained from the analysis of Nkx2-1 and Slit mutants suggest that although Slit1 and Slit2 may be functionally redundant proteins in vitro, they most likely play different biological roles in the guidance of cortical and thalamic axons in vivo.

Although lack of Slit2 function in the basal telencephalon and hypothalamus of Nkx2-1 mutants may be the primary cause of the abnormal guidance of pyramidal tract axons, additional factors may also contribute to this phenotype. Thus, a normal balance of attractive/permissive and repulsive/non-permissive cues may be required to determine the normal exit-point of pyramidal tract axons from the telencephalon and to ensure their normal transition through the forebrain (Fig. 9).

Previous studies on mice containing a targeted disruption of the Nkx2-1 gene have demonstrated that loss of Nkx2-1 function causes prominent morphological abnormalities in the basal telencephalon and hypothalamus (Kimura et al., 1996; Takuma et al., 1998). More recently, the molecular mechanisms that underlie the telencephalic defects observed in Nkx2-1 mutant mice were determined (Sussel et al., 1999). Specifically, loss of Nkx2-1 function results in the molecular repatterning of the MGE into a LGE-like tissue. Consequently, derivatives from the telencephalic region patterned by Nkx2-1 do not develop normally and, in turn, there is an expansion of the territories derived from the LGE, such as the striatum (Sussel et al., 1999). We have extended these observations and found that Nkx2-1 plays a similar role in patterning the hypothalamus. Thus, Nkx2-1 expression in the hypothalamic anlage is required to maintain molecular characteristics of the developing hypothalamus (e.g. Nkx2-4 and SF1 are reduced to a small region near the ventral midline) and to repress molecular characteristics of dorsal alar fates (e.g. Pax6 and Sim1 are expanded ventrally). The role of Nkx2-1 in the specification of ventral fates in the hypothalamus is further supported by the described functions of other Nkx genes in the vertebrate central nervous system. For example, Nkx2-2 or Nkx6-1 mutant mice also have ventral to dorsal transformations (Briscoe et al., 1999; Sander et al., 2000). Moreover, the Drosophila homolog of Nkx2-1, ventral nervous system defective (vnd), is required to specify ventral column identity (Chu et al., 1998; McDonald et al., 1998; Weiss et al., 1998). Interestingly, although other genes of the Nkx family are expressed in the hypothalamus in a pattern almost identical to Nkx2-1, such as Nkx2-4 (compare Fig. 1A with Fig. 2A), loss of Nkx2-1 function is enough to alter the dorsoventral patterning of this region.

The internal capsule traverses the NKX2-1-positive regions of the MGE, demonstrating that Nkx2-1 expression alone does not define the tissues that are non-permissive for these axons. Instead, a subdomain of the Nkx2-1 region contains the repellent agent(s) (e.g. Slit2). Thus, it is probable that Nkx2-1 alone does not directly regulate Slit2, but rather, that it contributes to the regulation of another transcription factor(s) that restrict Slit2 expression to the AEP/POa region.

In the present study, we have identified two forebrain regions, the basal telencephalon and the hypothalamus, that play a crucial role in patterning one of the most prominent tracts in the brain, the pyramidal tract. The mechanisms that underlie the guidance of cortical axons through these two regions seem to be similar, suggesting that a single patterning event, mediated by the homeobox gene Nkx2-1, is necessary to establish the appropriate exit of pyramidal tract axons from the telencephalon, as well as their normal path through the extra-telencephalic forebrain. Furthermore, Nkx2-1 patterning appears to be essential for the development of other fiber tracts that follow a course similar to the pyramidal tract axons, although in an opposite direction, such as the ascending dopaminergic mesotelencephalic projections. Thus, Nkx2-1 function is required to ensure the precise expression of signals that control the guidance of some of the major afferent and efferent projections of the telencephalon.

We are very grateful to M. Henkemeyer and J. Frisen for providing the EphB2/EphB3 and ephrinA5 mutant mice, respectively, to A. Bagri for his help with the collagen co-culture experiments, and to M. Tessier-Lavigne, S. Garel, A. Bagri, A. Plump and S. Pleasure for comments and discussions. This work was supported by the research grants to J. L. R. R. from Nina Ireland, NARSAD, NIDA (R01DA12462) and NIMH (RO1 MH49428-01, RO1 MH51561-01A1 and K02 MH01046-01). O. M. is a NARSAD Young Investigator Award recipient and a UC Davis MIND Institute Scholar.