Pax6 is required for the normal development of the forebrain axonal connections

The embryonic forebrain is comprised of distinct domains, prosomeres, each with specific morphological features and gene expression patterns(Puelles and Rubenstein,1993). Their borders often coincide with changes in the trajectories of axonal pathways (Wilson et al., 1997). TCAs and CFAs have to travel through numerous newly formed subdivisions of the embryonic brain. These critical boundaries outlined by distinct molecular properties (Puelles et al., 2000) include the diencephalon-telencephalon junction and the PSPB. Recent data indicate that some transcription factors are expressed(Pax6, Tbr1, Ngn2, Emx2 and Otx2) or escape (Emx1)the region of the ventral pallium (seePuelles et al., 2000), thus having a modulatory potential for the molecular patterning at the PSPB and the axonal pathfinding.

Early-generated, largely transient neuronal populations, have been described to extend pioneering axonal projections through these critical boundaries, thus providing scaffolds or temporary targets, `guide-post cells',for the developing axonal projections(McConnell et al., 1989;Mitrofanis and Guillery,1993). It has been suggested that the early outgrowth of TCAs from the diencephalon might be governed by pioneering projections from cells within the internal capsule (Métin and Godement, 1996), while the crossing of the PSPB by the TCAs might be dependent on selective fasciculation with these corticofugal projections(Molnár and Blakemore,1995). Recent studies in homeobox gene mutants seem to support these suggestions. In Mash1 (Ascl1 — Mouse Genome Informatics) homozygous mice, the internal capsule cells with thalamic projections are missing and TCAs fail to enter the internal capsule(Tuttle et al., 1999). InEmx2 homozygous mice, the early internal capsule projections take an aberrant ventral route at the diencephalon-telencephalon junction, and some of the TCAs follow them as they traverse that region(López-Bendito et al.,2002). In Tbr1, Gbx2 and Pax6 mutants, both TCAs and CFAs become derailed at the PSPB(Miyashita-Lin et al., 1999;Hevner et al., 2001;Hevner et al., 2002), but the nature of these defects is not fully understood.

Pax6 has been found to have an essential role in brain morphogenesis (Walter and Gruss,1991; Schmahl et al.,1993; Stoykova et al.,1996; Stoykova et al.,1997; Caric et al.,1997; Warren et al.,1999), and has been implicated in early axonal pathfinding(Mastick et al., 1997;Kawano et al., 1999;Pratt et al., 2000). APax6-positive group of cells was observed to extend along the PSBP,from the ventricular/subventricular zones (VZ/SVZ) towards the mantle of the basal telencephalon (Stoykova et al.,1997; Fernandez et al.,1998; Puelles et al.,2000). Lack of Pax6 function in thePax6^Sey/Sey mutant brains, results in a prominent ventralization of the molecular characteristics of the region of the ventral pallium that leads to severe morphological malformation in the basolateral cortex (Stoykova et al., 2000;Toresson et al., 2000;Yun et al., 2001).

Using a Pax6/lacZ knockout mouse(St-Onge et al., 1997), we studied the consequences of the lack of Pax6 on the TCA and CFA pathfinding with special interest to the axonal navigation within the PSPB. We found that in Pax6-deficient brains, both TCAs and CFAs showed severe malformations of their pathfinding when reaching the PSPB domain and these correlated with ultrastructural defects of this region and a failure of the expression of axonal guidance molecules, such as the attractants Sema3c and Sema5a.

In the Pax6/lacZ knockout, the start codon and the entire paired domain of Pax6 is replaced with β-galactosidase sequence, thus placing the activity of this reporter gene under the control of the endogenousPax6 promoter (St-Onge et al.,1997). Mouse embryos were derived from intercrosses of heterozygous males and females on a mixed background (NMRI and C57BL/6J×DBA/2J) and the day of vaginal plug was considered as E0.5. Brains were isolated from matched heterozygous and homozygous embryos at stages E14.5, E15.5 and E18.5. This knockout shows phenotype similar as thePax6^Sey/Sey mutation(St-Onge et al., 1997;Hill et al., 1991). The care and handling of the animals prior to and during the experimental procedures followed European Union regulations. Fetuses were removed by Caesarean section under pentobarbital anaesthesia (100 mg/kg i.p., to the pregnant female). Embryos were anaesthetized by hypothermia prior decapitation. Embryonic brains were fixed overnight in freshly prepared 4% paraformaldehyde in 0.1 M phosphate buffer (PB, pH 7.4).

The expression pattern of Pax6 in the isolated brains from heterozygous and homozygous embryos was analysed by wholemount-β-gal staining (Gossler and Zachgo,1993; Stoykova et al., 1988). Stained brains were embedded in 4%agarose and coronal sections were cut with a Vibroslicer (Leica, VT1000S) at 50 μm. Sections were mounted on slides in PBS/glycerol, coverslipped and examined with a light microscope. In situ hybridization on wholemounts and in sections was performed with Pax6 in situ probe (Walther and Gruss,1991) as described elsewhere (Stoykova and Gruss, 1994).

E14.5 and E18.5 wild-type, heterozygote and homozygous brains were used(see numbers at Table 1 for each time point). After β-galactosidase staining of the isolated brains,tissue was embedded in 4% agarose and sectioned at 40- 100 μm coronally with a Vibroslicer. Alternatively, after immersion of brains in 30% sucrose solution overnight at 4°C, tissue was sectioned coronally at 20 μm with a microtome. Immunohistochemistry was carried out with rat anti-L1(Boehringer, 1:100) and mouse monoclonal anti-TAG1 (Developmental Studies Hybridoma Bank, Iowa; 1:100) antibodies. Sections were incubated overnight with primary antibodies at 4°C. Primary antibodies were detected by a biotinylated secondary antibody (1:100) and the avidin-biotin complex (ABC,dilution 1:100) method (Vector). The reaction was visualized using 0.05%diaminobenzidine solution (Sigma). Sections were rinsed, dehydrated and mounted in Eukitt mounting media. To study cytoarchitecture, sections were treated with 0.5% Cresyl Violet solution. Sections were visualized by light microscopy and documented with a Leica DC500 digital camera.

Brains at E14.5, E15.5 and E18.5 were fixed overnight in 4%paraformaldehyde (see Table 1). To expose the dorsal thalamus, a coronal cut was made at the caudal edge of the diencephalon. The zona limitans intrathalamica provided the landmark for orientation. Single crystals of DiI (1, 1- dioctadecyl -3, 3,3′,3′-tetramethyl-indocarbocyanine perchlorate; Molecular Probes),and of DiA [4-(4-(dihexadecylamino)styryl)-N-methylpyridium iodide; Molecular Probes], were placed with a fine tungsten wire into the dorsal thalamus or dorsal cortex under a dissecting microscope. Brains were incubated in 4%paraformaldehyde at room temperature (RT) for 3 weeks. The brains were then embedded in 4% agarose and 100 μm coronal Vibroslicer sections were cut. Tissue was counterstained with 0.002% bisbenzimide. Sections were coverslipped in 1:1 mixture of PB and glycerol and analysed with a fluorescent microscope(Leica DMR) or a laser scanning confocal microscope (Leica TCS).

Four E15.5 mice and four E18.5 mice (two Pax6^-/- and two Pax6^+/+ brains for each age) were used. Brains were immersed overnight in ice-cold fixative containing 2% paraformaldehyde and 2%glutaraldehyde in 0.1 M phosphate buffer (PB, pH 7.2-7.4). Tissue blocks were dissected and washed in 0.1 M PB. Coronal 60 μm sections of the telencephalon were cut with a Vibroslicer. After several washes in 0.1 M PB,selected sections were treated with OsO₄ (1% in 0.1 M PB),block-stained with uranyl acetate, dehydrated in a graded series of ethanols and flat-embedded on glass slides in Durcupan (Fluka) resin. Ultrathin sections (60 to 80 nm) from selected areas of the telencephalon were picked up on coated copper grids. The grids were counterstained with saturated aqueous uranyl acetate followed by lead citrate. The sections were examined with a Joel (JEM 1010) transmission electron microscope.

In situ hybridization experiments were performed using³⁵Sriboprobes on 10 μm frozen sections from E14.5 and E18.5 brains (see Table 1) according to standard protocols described previously(Stoykova et al., 2000). The cDNA probes were for Sema3a, Sema3c, Sema5A, Sema5b, Sema6a and netrin 1(gifts from A. Puschel and M. Tessier-Lavigne).

The expression pattern seen in E12.5 wild-type embryos after whole-mount in situ hybridization with Pax6 probe(Fig. 1A) and the pattern seen in Pax6 homozygous embryos with β-galactosidase staining(Fig. 1B) were very similar. This indicated that the β-galactosidase activity in the transgenic embryos correctly reflected the well-characterized expression pattern ofPax6 in developing mouse CNS(St-Onge et al., 1997). However, we detected a wider β-gal staining pattern in the pallium of the transgenic embryos that included expression not only in cortical progenitors,but also in postmitotic cells of the cortical plate. Such a discrepancy between the normal distribution of mRNA and the β-gal staining pattern in transgenic animals was accounted for by the longer half-life of theβ-galactosidase protein when compared with the transcript (seeScardigli et al., 2001). We examined the endogenous expression of Pax6 using the Pax6 in situ probe in wild-type brain sections and the lacZ in situ probe in sections from heterozygous and homozygous embryonic brains to investigate this further. In sections of E14.5 wild-type(Fig. 1C), heterozygous(Fig. 1D) and homozygousPax6^-/- (Fig. 1E) brain, the hybridization signals for Pax6(Fig. 1C) and lacZ(Fig. 1D,E) were confined only to the germinal neuroepithelium (VZ/SVZ) of the pallium. In accordance with the previously reported analysis for homozygousPax6^Sey/Sey brains(Stoykova et al., 1997;Stoykova et al., 2000), the prominent expression of lacZ at the PSPB faded in the transgenic homozygous brains (Fig. 1E). Furthermore, a very similar expression was detected by the Pax6 andlacZ in situ probes in differentiating nuclei within the ventral thalamus (VT) of the wild-type (Fig. 1F) and heterozygous (Fig. 1G) brains. With these experiments, we excluded the possibility that, in the transgenic embryos, some regulatory sequences that are responsible for the regionalized expression of Pax6 in the pallial VZ/SVZ were missing. We concluded, therefore, that the lacZexpression pattern in the Pax6^-/- mice correctly represented the endogenous Pax6 expression pattern in the developing telencephalon.

In order to study axonal pathway formation in thePax6^-/-, we used carbocyanine dyes to trace TCAs in fixed brains (see Table 1). DiI crystal placement into dorsal thalamus from E14.5 to E18.5 in wild-type brains labelled TCAs extending to the primitive internal capsule having a normal fasciculation pattern (see Fig. 3A,B). This was also seen in heterozygous brains(Fig. 2D). DiI crystal placements in the dorsal thalamus revealed a group of cells within the internal capsule, which were situated medial to the β-galactosidase stained stripe of cells at the PSPB (Fig. 2G-I). It is known that some cells of the internal capsule project to dorsal thalamus (Métin and Godement, 1996; Molnár et al., 1998). However, in the homozygous brains, only a few backlabelled cells were seen at the ventral telencephalon along the predicted trajectory of the TCA pathway (Fig. 3I,J), but many were observed located the hypothalamus(Fig. 3F-H).

As it has been suggested that TCA development is delayed in the lack ofPax6 (Pratt et al.,2000), we examined TCA development in thePax6^-/- at E15.5 and E18.5. In the E15.5 brains, TCAs projected from the dorsal thalamus to the hypothalamus(Fig. 3E-G) and by E18.5, some TCAs reached the ventral telencephalon(Fig. 3K,L) from where they ascended dorsally towards the ganglionic eminences, following an aberrant trajectory. Although they could be followed towards the PSPB, they were not able to cross it to enter into the cortex(Fig. 4Q).

We then examined whether the ability of CFAs to navigate across the PSPB was affected in the Pax6^-/- brains. In wild-type and heterozygotes mice at E14.5 and E15.5, CFAs extended through the intermediate zone to reach the PSPB, displaying a normal fasciculation pattern(Fig. 4A,B). In thePax6^-/- brains, CFA tracing studies revealed several abnormalities. At E15.5, the CFA pathway started to descend towards the PSPB,but it was displaced laterally (Fig. 4C,D). At later stages, large numbers of axons were seen to course ventrolaterally through the developing lateral pallium, extending towards the base of the ventral pallium (the amygdaloid region)(Fig. 4H). The majority of CFAs were unable to change their trajectories to cross the PSPB(Fig. 4J), nevertheless some entered into the subpallium fasciculating abnormally in large bundles at more caudal levels (Fig. 4K-M). We observed more numerous backlabelled cells in the marginal zone distant from the cortical crystal placement sites in the mutants, suggesting that they developed long range projections (Fig. 4N). The callosal projections were present in the mutants, but their fasciculation pattern was also abnormal(Fig. 4O). Cortical placement of DiI crystals in wild-type brains at E15.5 revealed continuous axonal bundles (Fig. 4F), which extended to groups of backlabelled cells in the dorsal thalamus (data not shown) (see Molnár et al.,1998). As carbocyanine dyes travel in both anterograde and retrograde directions in fixed tissue, and thalamic fibres arrive to the cortex by E15.5, the bundles labelled at E18.5 contained both sets of fibres. In the Pax6^-/- brains, however, we did not observe backlabelled groups of cells in the dorsal thalamus after cortical crystal placements, and we could not detect any interaction between CFAs and TCAs when we used double labelling (Fig. 4Q,S,U). This confirmed our observation obtained with anterograde tracing from the thalamus, and indicated that, although thalamic axons approached the PSPB, they did not reach the cortex in the mutants.

To examine the gross pattern of TCA growth, we performed L1 immunohistochemistry. Interestingly, we observed that L1 expression was not limited only to TCAs at stage E14.5 and E18.5 as previously thought(Fukuda et al., 1997), but L1 immunoreactive fibres were present in dorsal cortex at E14.5(Fig. 5A,B). At this stage,TCAs normally extend only to the level of the PSPB and do not invade the intermediate zone of the cortex (De Carlos and O'Leary, 1992; Erzurumulu and Jhaveri, 1992). Therefore, the detected immunoreactivity for L1 in the cortex at this age cannot be due the presence of TCAs, suggesting that the CFAs were also labelled as they extend in the intermediate zone of the dorsal cortex towards the PSPB.

TCAs and CFAs are known to reach the PSPB synchronously(De Carlos and O'Leary, 1992)and the two sets of axons intermingle(Molnár et al., 1998). By E18.5 in wild-type brains, TCAs have reached to the cortex through the intermediate zone (Catalano et al.,1991). L1 immunolabelling in wild-type sections(Fig. 5E,I) revealed the gross organization of the fascicles in the striatum, PSPB and in the intermediate zone. In Pax6^-/-, L1 immunoreactivity was detected on both CFAs and TCAs at the two studied ages (Fig. 5C,G) and at E14.5, TCAs were seen emerging from the basal ventral telencephalon and had turned dorsomedially towards the subventricular zone of the ganglionic eminence, but did not cross the PSPB(Fig. 5C,D). At E18.5, inPax6^-/- brains, L1 immunolabelled fibres reached the PSPB,but they could not be followed crossing this boundary(Fig. 5F,G,J,K). The L1-positive fibre trajectory of CFAs in the cortex was displaced ventrolaterally and axons continued their path to the ventral pallium as if they were unable to cross the PSPB (Fig. 5D). As described above with carbocyanine dyes(Fig. 4Q), using immunohistochemistry for L1 we described similar results which point out that TCAs and CFAs failed to meet each other at the PSPB.

CFAs selectively express the cell adhesion molecule TAG1(Denaxa et al., 2001). The immunostaining for TAG1, however, seemed to reveal only the initial part of this pathway, as TAG1 immunoreactivity was drastically reduced at the level of PSPB (see López-Bendito et al.,2002). In the wild-type forebrain at E14.5, TAG1 immunoreactive CFAs appeared to extend through the intermediate zone of the developing pallium to reach the PSPB (Fig. 5M,N). At E18.5, TAG1 immunoreactivity was expressed primarily by cells of the cortical plate and marginal zone in the hippocampus and corpus callosum (not shown). In the Pax6^-/- brains, TAG1 immunoreactive fibres followed aberrant trajectories, being displaced laterally in thick bundles (Fig. 5O,P). In addition, TAG1 expression in cortical plate and marginal zone was reduced, preserved mostly within the medial pallium. These findings were consistent with the abnormalities observed in CFA development inPax6^-/- with DiI tracing.

We further examined the cellular organization of the PSPB domain in relation to the β-galactosidase staining in the developing forebrain of E14.5 and E18.5 heterozygous and homozygous embryos. In E14.5Pax6^-/- brains, the telencephalic neuroepithelium at the PSPB appeared broader and more irregular compared with normal, and consisted of aggregated cells which were heavily stained with Cresyl Violet(Fig. 6D-F). By E18.5, a considerable cell mass was prominent, which often occluded most of the lateral ventricle at rostral level (Fig. 5F,G). The extended region consisted of cells that were heavily stained for Nissl but negative for β-galactosidase stain. These cells were densely packed together and appeared to be continuous with the cortical ventricular zone. The thickened pallial germinal neuroepithelium of the E18.5Pax6^-/- was also β-galactosidase negative(Fig. 4G,Fig. 5F,G,K). In thePax6^-/-, the ventricular cellular mass and the atypical PSPB did not show immunoreactivity for the glial cell markers GFAP or vimentin, and did not express calbindin or GABA, but were immunoreactive for the neuronal marker, β-tubulin class III(Menezes and Luskin, 1994)(data not shown), suggesting that they comprised undifferentiated neurones. InPax6^-/- brains, β-galactosidase staining of the dorsal cortex continued to the ventral pallium at the lateral surface(Fig. 4G), whereas in heterozygotes the unstained lateral cortex occupied this zone and theβ-galactosidase staining formed a narrow stripe linking stained cells of the amygdala to the cortex (Fig. 4E). In addition, Nissl staining revealed that the cortical plate was drastically reduced in the ventral-most lateral pallium in thePax6^-/- brains (Fig. 6F).

In order to characterize further the morphological abnormalities seen in the Pax6^-/- mouse, which cause severe disruption in the development of CFAs and TCAs at the PSPB, we performed electron microscopic analysis at that boundary in wild-type and homozygous brains at two different stages, E15.5 and E18.5. In wild-type brains, heterogeneous cell populations within the region of this boundary were observed. Axons crossing the PSPB were seen at E15.5 as well at E18.5 (Fig. 7E,O). However in the Pax6^-/- brains, the cells within the PSBP domain appeared to be more densely packed than in wild-type brains. This abnormality was more prominent at later stages(Fig. 7H,S). Cells located in this area were seen tightly attached to neighbouring cells(Fig. 7I). In accordance with previous data (Stoykova et al.,1997) and our tracing experiments described above, a few axonal tracts were observed at electron microscopy level traversing the PSPB in thePax6^-/- at both ages(Fig. 7H). At the lateral pallium of the Pax6^-/- brains, bundles of axons were observed extending superficially (Fig. 7J), possibly corresponding to the CFAs that in the mutant were seen to follow an aberrant trajectory along this area (seeFig. 4J,K).

In an attempt to look for possible molecular determinants for the observed aberrant growth of the CFAs and TCAs in the Pax6^-/- brain,we examined the expression of several semaphorins (Sema3a, Sema3c, Sema5a,Sema5b and Sema6a) and netrin 1, which have previously been implicated in thalamocortical and/or corticofugal pathfinding(Bagnard et al., 1998;Métin et al., 1997). A prominent defect in the expression pattern of Sema3c andSema5b was detected within the embryonic pallium inPax6^-/- (Fig. 8). Sema3c is assumed to act as attractive axon guidance signal secreted by the cells of the SVZ(Bagnard et al., 1998). In agreement with Bagnard et al. (Bagnard et al., 1998), the expression of Sema3c in E14.5 wild-type cortex was confined to the SVZ and intermediate zone of the entire pallium,thus including the dorsal, lateral and ventral pallial domains(Fig. 8A). A similar pattern was seen in the heterozygous brains (Fig. 8B). However, in the Pax6^-/- brains the expression of Sema3c was completely abolished in the region of the ventral and lateral pallium, while in the dorsal pallium the expression was shifted into the most superficial zone of the cortical plate(Fig. 8C). In sections of E18.5 wild-type brains, the signal was fainter, but still detectable all along the extending CFAs towards the basolateral telencephalon (data not shown). Similarly as at stage E14.5, the expression of Sema3c was not detectable within the ventral and lateral pallium in thePax6^-/- mice.

In wild-type brains, Sema5a had a strong expression in the VZ of the rostral pallium with a lateral (high) to medial (low) expression gradient(Fig. 8D,E). In thePax6^-/- brains, the lower expression level in the medial pallium was preserved, while the very prominent hybridization signal seen within the VZ at the PSPB was not visible any more (compareFig. 8G withFig. 8E, arrows). At E14.5,netrin 1 was expressed in the basal telencephalon including the region of the primitive internal capsule (Métin et al., 1997). In the Pax6^-/- brains, the netrin 1 expression is present and might even appear stronger and more widespread than in wild-type (data not shown), extending to the region where the aberrant thalamic axons enter to the ventral telencephalon. These results suggest that the altered expression pattern of Sema3c, Sema5b and perhaps netrin 1 in the Pax6 mutant may be responsible for the altered growth of the CFAs and TCAs.

In this study, we show that Pax6 is involved in pathfinding not only of the thalamocortical, but also of the corticofugal axonal connections(Fig. 9). The abnormalities are complex and numerous factors might be responsible for the failure of both CFAs and TCAs to navigate to their targets. We demonstrate that inPax6^-/- brains, cells with early dorsal thalamic projection at the internal capsule are more dispersed. In addition,alterations in the structure of the PSPB were observed, together with the failure of expression of axon attractive molecules Sema3c, Sema5a and netrin 1. The results suggest that the prominent ventralization of the molecular patterning at the PSPB as shown previously(Stoykova et al., 2000;Toresson et al., 2000;Yun et al., 2001), might affect environmental cues and differentiation of guide-post cells within the internal capsule and at the PSPB regions, both of which are essential for the TCAs to reach the cortex.

We have previously suggested in our `handshake hypothesis', that axons from the thalamus and from the early-born cortical preplate cells meet and intermingle in the basal telencephalon, so that thalamic axons grow over the scaffold of preplate axons and become `captured' for the waiting period in the subplate (Blakemore and Molnár,1990; Molnár and Blakemore, 1995). The use of the Pax6^-/- mice(St-Onge et al., 1997) allowed us to examine the relationship between Pax6 expression at the PSBP and the cortical connectivity in normal and in Pax6 loss-of-function conditions. In agreement with previous results(Kawano et al., 1999;Pratt et al., 2000;Hevner et al., 2002), we found that the lack of PAX6 resulted in a failure of TCA growth into the telencephalon, which is accompanied by a misrouting of the majority of the TCAs into the hypothalamus. Remarkably, DiI labelling of the mutant cortex revealed that, in addition to TCAs, the CFAs also follow an aberrant trajectory with an abnormal fasciculation pattern. Although perinatally(E18.5) the thalamic axons reach the PSPB, they are not able to cross this region. Double DiI/DiA labelling experiments clearly indicated that no interactions occur between TCAs and CFAs within the internal capsule, and these two sets of axons remained separated by the PSPB. These observations were further supported by the L1 and TAG1 immunohistochemistry results. The aberrant development of these tracts may be partly explained by abnormal axonal fasciculation because CFAs and TCAs would be unable to interact and use each other as a scaffold to extend towards their targets. Accordingly, mice with mutations in transcription factor genes expressed in the pallium(Tbr1) and dorsal thalamus (Gbx2), show TCA and CFA pathfinding errors in the region of the internal capsule(Miyashita-Lin et al., 1999;Hevner et al., 2001;Hevner et al., 2002). Tracing experiments in these knockout mice have shown that lack of early CFA is associated with abnormal development of TCAs. It is conceivable therefore that the development of both fibre systems depends on the same factors requiring the intact morphology of the PSPB and/or interactions between the two fibre systems.

The lack of Pax6 in the Pax6^Sey/Sey causes a complex phenotype reflected in defects of the molecular regionalization and boundary formation in developing telencephalon(Stoykova et al., 1996;Stoykova et al., 1997;Stoykova et al., 2000;Warren and Price, 1997;Toresson et al., 2000;Yun et al., 2001), axonal pathfinding (Mastick et al.,1997; Pratt et al.,2000), adhesive properties(Stoykova et al., 1997;Meech et al., 1999), cortical progenitor proliferation (Warren et al.,1999), cell migration (Caric et al., 1997; Brunjes et al.,1998; Chapouton et al.,1999), radial glia differentiation(Götz et al., 1998) and neurogenesis (Heins et al.,2002). This phenotype is established early in embryonic development, and becomes progressively more complex at later stages, which seriously hampers the determination of the primary effect of Pax6 in the developing brain.

The choreography and timing of developmental events seem to be important for the normal development of telencephalic connectivity. We identified several defects of TCA development in the Pax6^-/- brains. It is conceivable that changes in the location of forebrain domains, and their boundaries, may displace transient `guide-post' cells and their pioneering projections, which are thought to be crucial for TCA guidance(Métin and Godement,1996; Molnár et al.,1998). Cells with similar function have been described within the internal capsule and ventral thalamus in numerous species(Molnár, 1998). In lack of Pax6 in the Pax6^Sey/Sey, the differentiation of the ventral thalamic nuclei is completely abolished(Stoykova et al., 1996;Warren and Price, 1997)(Fig. 1H), including the reticular nucleus, which has an essential role in the control of the thalamocortical circuits (Mitrofanis and Guillery, 1993). We found that only a few thalamic projecting cells were labelled within the PSBP after a thalamic placement of the DiI crystals and these cells were displaced towards the hypothalamus. In the developing telencephalon, Pax6 expression is confined to the germinal neuroepithelium (VZ/SVZ) with a rostrolateral (high) to caudomedial (low)gradient, with a most prominent expression within the lateral and ventral pallium (Stoykova et al.,1997; Smith-Fernandez et al.,1998; Puelles et al.,2000). In the Pax6^-/- at the pallial VZ, there is an intensive ectopic expression of ventral markers in the lateral and dorsal pallium, which seems to include direct transcriptional regulation between other genes, e.g. Pax6 and Ngn2; Pax6 andGbx2 (Stoykova et al.,2000; Toresson et al.,2000; Yun et al.,2001). These pattern abnormalities seem to affect the progenitor identity of the ventrolateral telencephalon, leading to a failure of the lateral cortical plate formation and a dysgenesis of basolateral structures,claustrum, endopiriform nucleus and piriform cortex(Stoykova et al., 2000). It is conceivable therefore that the described defects in the generation and/or functional properties of the TCAs, guide-post cells within the internal capsule, ventral thalamus and perhaps hypothalamus might be a further consequence of the ventralization of the telencephalic neuroepithelium at the PSBP border in Pax6^-/- mice.

Semaphorins are thought to be involved in cortical and thalamic axon guidance (Puschel, 1999;Skaliora et al., 1998). In thePax6^-/- brains, the expression of Sema3c was completely abolished from the SVZ and IZ of the lateral and ventral pallium,thus including the region where the CFA defects were observed. In the dorsal pallium, however, the expression was shifted into the most superficial zone of the cortical plate. Sema3c (previously known as SemE) is assumed to act as attractive axon guidance signal secreted mostly by the cells of the SVZ(Bagnard et al., 1998). Thus,the abnormal expression pattern of Sema3c in thePax6^-/- brain correlates well with both, the CFA pathfinding defects within the PSBP and with the aberrant orientation of the cortical axons towards the cortical surface and might be responsible for them.

It has been suggested that secreted chemoattractants, e.g. netrin 1, might influence the growth of navigating axons, especially when they change their direction of growth (Métin et al.,1997; Braisted et al.,2000; Tessier-Lavigne and Goodman, 1996). As reported in this work, the expression of netrin 1 in the basal telencephalon of the Pax6^-/- embryos(including the region of the internal capsule) was present and appeared wider and stronger, when compared with the wild-type brain, which might additionally hamper the proper growth of the early TCA axons through the altered environment in the mutant. This possible difference in netrin 1 expression pattern shall need further quantitative analysis.

Previous evidence indicated that the expanded VZ/SVZ in thePax6^Sey/Sey pallium is a consequence of accumulation of only partially differentiated (Stoykova et al., 2000; Heinz et al., 2002) or prematurely differentiated(Warren et al., 1999)neuroblasts. Furthermore, in this mutant, the adhesiveness of pallial cells is increased, accompanied by a severe downregulation of the expression of the homophilic cell-adhesion molecule R-cadherin and of tenascin (Stoykova et al.,1977; Warren et al., 1999). We have described similar abnormalities at the PSPB in thePax6^-/-. The densely packed adhering cells especially apparent at the anterior part of this boundary might not be penetrable for TCAs or CFA, thus making an additional obstacle for the TCAs and CFA to traverse the IC, the gateway between thalamus and cortex.

There is a continuing debate as to how the developing pallium influences TCA development, and in turn whether TCAs affect development of the cortex. It has been demonstrated that there is a shift in the areal identity in the cerebral cortex of Emx2 and Pax6^Sey/Sey mutant mice, which is matched with altered distribution of TCAs in theEmx2^-/- brains (Bishop et al., 2000; Mallamaci et al., 2000). In the Pax6^-/- brains, no thalamic projections reach the cortex, suggesting that the shift observed in cortical representation in the Pax6^-/- cortex might be in fact independent from TCA targeting and do not rely on the presence or absence of thalamic projections. Our tracing studies demonstrated that in bothEmx2^-/-(López-Bendito et al.,2002) and Pax6^-/- mice (present study) a large proportion or the entire early TCA population were misrouted at the border between the diencephalon and telencephalon. In addition, in thePax6^-/- an additional defect prevented the TCAs from invading the cortex through the PSPB. Although, both of these early TCA targeting abnormalities are associated with the altered cortical Emx2and Pax6 expression patterns they are related to earlier guidance defects, most probably independent from the cortical alterations.

In summary, we have found that axonal pathfinding and fasciculation of both CFA and TCA are dependent upon the normal expression of Pax6 during early forebrain development. The pathfinding defects may be secondary to displacement and dispersion of guidepost cell populations within the ventral thalamus and internal capsule and to abnormal barrier formation at the PSPB as a result of defects in molecular patterning. These cellular changes are also associated with altered expression of proteins (including Sema3c, Sema5a and netrin 1) required for normal axonal guidance. Our study suggests thatPax6 has numerous roles in thalamocortical guidance, and these seem to be independent from the cortical regionalization.

The excellent technical support of Ms Silke Eckert and Mr David Modoux is gratefully acknowledged. We are very grateful to Professor John Parnavelas and Dr Richard Branton for their thoughtful comments on the manuscript. We are grateful to Andreas Puschel and Marc Tessier-Lavigne for their help with the probes. L. J. held a BBSRC studentship. This work was supported by Grants from The European Community (QLRT-1999-30158, QLG3-CT-2000-00158), The Wellcome Trust (063974/B/01/Z), The Royal Society, Swiss National Science Foundation(3100-56032.98), Human Frontier Science Program (RGP0107/2001) and St John's College, Oxford.