A role for netrin-1 in the guidance of cortical efferents

Growing axons are guided towards their targets along specific pathways by multiple cues present in the environment of their growth cones, to which they respond with complex growth behaviors (Tosney and Landmesser, 1985; Caudy and Bentley, 1986; Bovolenta and Mason, 1987; Godement et al., 1990; Godement, 1994). Molecular cues include both membrane- and extracellular matrix (ECM)-bound molecules, as well as diffusible factors secreted by intermediate or final targets. Some diffusible guidance molecules are thought to be capable of establishing concentration gradients, which might eventually be stabilized by binding to cell membranes or the ECM (Kennedy et al., 1994; Fan and Raper, 1995; Wadsworth et al., 1996). These long-range guidance cues appear to be capable of acting over distances of a few hundred micrometers (reviewed in Tessier-Lavigne and Placzek, 1991). Short- and long-range guidance cues can affect growth cone extension and orientation by eliciting either repulsive/inhibitory or attractive/permissive (growthpromoting) responses (for review, see Tessier-Lavigne and Goodman, 1996).

The first diffusible chemoattractant molecules identified in the vertebrate central nervous system, the netrins, (Serafini et al., 1994; Kennedy et al., 1994), are expressed in the ventral spinal cord and ventral structures of the brain (Kennedy et al., 1994; Serafini et al., 1996; Livesey and Hunt, 1997). Netrin-1 is a bifunctional molecule which can promote the outgrowth of commissural axons and is implicated in attracting their growth cones toward the ventral midline in spinal cord, hindbrain and telencephalon (Kennedy et al., 1994; Shirasaki et al., 1995; Serafini et al., 1996). Netrin-1 is also implicated in repelling the growth cones of trochlear motoneurons and branchiomo- toneurons away from the ventral midline (Colamarino and Tessier-Lavigne, 1995a; Varela-Echevarria et al., 1997). This laminin-related molecule has homologs in Caenorhabditis elegans and Drosophila that show a remarkable degree of conservation in sequence and apparent function (Hedgecock et al., 1990; Ishii et al., 1992; Mitchell et al., 1996; Harris et al., 1996). Recent studies in C. elegans show that the nematode netrin UNC-6 is also involved in the guidance of noncommis- sural pioneer axons in the cephalic end of the worm (Wadsworth et al., 1996).

Reciprocal and topographic projections between different cortical areas and thalamic nuclei in mammals provide an example of ipsilateral, non-commissural projections in the rostral brain of vertebrates. During the first stages of their growth, the thalamocortical and corticothalamic projections independently converge ventrally towards the basal telencephalon (De Carlos and O’Leary, 1992; Erzurumlu and Jhaveri, 1992; Miller et al., 1993; for review, see Molnar and Blakemore, 1995). The ganglionic eminence of the basal telencephalon (embryonic basal ganglia) appears to be an intermediate target for these axons during the early stages of their growth (Métin and Godement, 1996). After meeting in the ganglionic eminence (GE), early thalamic and cortical projections might be guided by interactions with each other (Allendoerfer and Shatz, 1994), at least over distal portions of their trajectory (Molnar and Blakemore, 1995). The mechanisms that guide these axons before they encounter one another are, however, poorly understood. In a previous study (Métin and Godement, 1996), we suggested that both diffusible and contact-mediated guidance cues might together control the pathfinding of cortical fibers in the GE because oriented growth of cortical axons was observed toward GE subregions in coculture experiments, and clear axocellular appositions were observed between growing cortical axons and cell bodies in the lateral ridge of the GE in E12.5 and E13.5 mice embryos. These earlier studies did not, however, make it possible to determine the relative contribution of the two mechanisms.

In the present paper, we describe coculture experiments that were designed to further elucidate the potential role of diffusible factors in guiding corticofugal axons extending toward the GE. In particular, we wished to assess the potential role of netrin-1, which is expressed in the basal ganglia of the forebrain (Serafini et al., 1996), as a candidate chemoattractant for corticofugal fibers within the GE. Thus, we first analyzed in detail the distribution of netrin-1 mRNA in the mouse basal telencephalon and then addressed its potential involvement in guiding cortical efferents using functional assays in vitro. Our results support a role for netrin-1 in guiding corticofugal axons. A recent study by Richards et al. (1997) has provided similar evidence that netrin- 1 may guide cortical axons toward the internal capsule in the rat.

C57 Bl/6 mouse embryos were bred in the laboratory. Embryonic day E0 is the day of the vaginal plug.

Telencephalic vesicles of E12.5 or E13.5 mouse embryos were dissected in cold F12-DMEM medium (pH 7.2, Gibco) and the pia removed. Cortical explants, dissected with thin tungsten needles, consisted of either thin rostrocaudal bands whose length varied from 400 μm to 1200 μm and mean width was close to 150 μm (±30 μm; see Fig. 1A), or large squares about 1000 μm × 1000 pm extending ventrally as far as the ventricular angle of the telencephalon (see Fig. 3A). Cortical explants from E12.5 embryos comprised the whole thickness of the cortical wall, whereas thin bands from E13.5 cortical wall only comprised the cortical plate and the superficial part of the intermediate zone.

Explants from the GE originated from one of the following areas (Fig. 1A). (1) The lateral ridge of the ganglionic eminence (LGE) from which we dissected rostrocaudal bands (approx. 250 μm × 500 μm) in the region located below the parietal area. Explants comprised the whole thickness of the vesicle at E12.5, and the ventricular zone with the adjacent mantle zone at E13.5. In some experiments, the mantle zone and ventricular zone of E13.5 explants were tested separately. (2) The sulcus between the two ridges of the GE, the intermediate sub- pallial sulcus (ISS), which forms a deep furrow in the rostral telencephalon of E12.5 embryos, and ends before the caudal pole of the GE, in a region where LGE and MGE fuse. Explants from the ISS mostly consisted of the proliferative neuroepithelium of the ventricular zone (mean dimensions, 300 μm × 400 μm). (3) The medial ridge of the GE (MGE). Pieces (approx. 350 μm × 350 μm) comprising the ventricular zone and the adjacent mantle zone were dissected either close to the telodiencephalic sulcus, or in more caudal and more rostral regions.

After dissecting, explants were transferred into a drop of collagen (Cellon, Luxembourg) on a coverslip, using glass micropipettes. The thin bands of embryonic cortex were oriented in collagen with their pial surface against the coverslip. In some cases, they curved and turned before gel polymerisation; the pial surface was then orthogonal to the coverslip (see Fig. 2B). The large square cortical explants were oriented with their ventricular surface on the coverslip. GE explants were placed either at a distance from long cortical strips, or along the lateral edges of large square explants. After polymerization, gels were covered with F12-DMEM supplemented with 10% heat- inactivated fetal calf serum (Gibco), and 20 U/ml penicillin/strepto- mycin and were incubated at 37°C in a 5% CO₂ atmosphere.

Cortical axons began to develop in the collagen gel after 10 hours in E12.5 cocultures, and extended quickly in the following 4–6 hours. The outgrowth of cortical axons started after 20 hours in E13.5 cocultures, and axons appeared fully developed after 36 hours in culture. The axonal growth within large E12.5 cortical explants was analysed after 2 days in culture.

Cocultures were fixed in 4% paraformaldehyde, 0.33 M sucrose, 1×PBS. In cocultures of thin bands of cortex and GE explants, the outgrowth of cortical axons into the collagen gel was analyzed and quantified as explained in Fig. 2D, using camera lucida drawings and/or microphotographs of cocultures. Significance was assessed using Student’s t-test. In cocultures of large cortical explants and GE explants, cortical axons were labeled by placing small crystals of DiI at the pial surface of fixed cortical explants. Cortical axons were labeled over their entire length a few days later, as indicated by labelling of growth cones at their tip.

Brains sections from E12.5 and E13.5 embryos were prepared for netrin-1 mRNA in situ hybridization as follows. Heads of E12.5 embryos and brains of E13.5 embryos were dissected in cold 4% paraformaldehyde in PBS, fixed overnight in the same fixative, cryoprotected in 10%, 20% and 30% sucrose in PBS, embedded in gelatin (7.5% gelatin, 15% sucrose in PBS) and frozen in − 50°C isopentane. Cryostat sections (10 μm) were made either in the frontal or horizontal planes. In situ hybridization using digoxigenin-labeled probes was performed as described by Schaeren-Wiemers and Gerfin-Moser (1993). A 2.7 kb netrin-1 probe (consisting of part of the 3’UTR sequence) and a control sense probe were used. After linearization and labeling (using a Stratagene kit), probes were purified on a Sephadex column (Pharmacia Biotech). Probes larger than 900 bp were basehydrolyzed and used at a final concentration of 500 ng/ml. In addition, netrin-1 expression was examined on flattened telencephalic vesicles of E12.5 and E13.5 mouse embryos. Brains were dissected in cold F12-DMEM medium (Gibco), flattened on a filter and fixed in 4% paraformaldehyde in PBS at 4°C for 12 hours. After pretreatment with proteinase K (10 p g/ml in Tris-EDTA, 15 minutes), in situ hybridization was performed as described above. After detection, vesicles were dehydrated in graded alcohols and kept in a benzyl benzoate/benzyl alcohol (2/1) mixture.

Small squares (approx. 150 μm × 150 μm) or bands (approx. 120 μm × 500 μm) of cortical tissue were dissected from the caudal half of the cortex of E12.5 mouse brains. In each experiment, either squares or bands were used. To quantify axonal growth from explants, we placed four square explants or two long explants in each gel and tested the same concentration of netiin-1 in two or three cultures per experimental series. Purified recombinant chick netrin-1 protein produced in human embryonic kidney 293 cells (Shirasaki et al., 1996) was directly diluted into culture medium. Explants were cultured for 18 hours, fixed with 4% paraformaldehyde, 0.33 M sucrose in PBS, and observed with Nomarski optics on an inverted microscope (Zeiss Axiovert). Axonal growth in collagen gels was quantified using a CCD camera and an image processing program (Optimas). In each experimental series, we recorded pictures from at least 5 distinct explants for each concentration. For each explant, the number and length of fibers or fascicles leaving the cortical explant were counted in at least two zones showing low fiber outgrowth and two zones showing high fiber outgrowth. Measurements were performed using NIH Image. Mean and standard deviation values were calculated for each netrin-1 concentration in each experimental set.

In these experiments we used a netrin-1-secreting 293-EBNA cell line (Shirasaki et al., 1996). The parental 293-EBNA cell line (InVitrogen) was used as a control. Aggregates of 293-EBNA cells were produced as described by Kennedy et al. (1994). Thin rostrocaudal bands (approx. 120 μm × 500 μm) of E12.5 cortex and aggregates of netrin-1/293- EBNA cells (255 ± 52 pm in diameter), or aggregates of 293-EBNA cells (269±56 pm in diameter) were cocultured at a distance in collagen gels. Each gel contained two to four cocultures. The outgrowth of cortical axons into the collagen gel was analyzed on fixed cocultures by counting the number of cortical fibers and fascicles growing opposite netrin-1-secreting cells or control cells, and by normalizing these results to a 100 pm long segment of cortical explant. In another set of experiments, large explants (approx. 1000 μm × 1000 μm) of E12.5 cortex were cocultured in collagen gels with aggregates of netrin-1/293-EBNA cells or aggregates of parental 293 EBNA cells placed along their rostral and caudal edges. The trajectories of cortical axons were studied after labeling with DiI in fixed cocultures (see above).

Affinity-purified rabbit antibodies raised against domains VI and V of chick netrin-1 protein (T. E. Kennedy et al., unpublished), and control affinity purified non-immune rabbit IgG, were added to cortex/LGE and cortex/ISS cocultures. Antibodies were added to the culture medium at a concentration of 10 μg/ml. Cocultures were prepared as described above, using explants from E12.5 embryonic brains. We took care to perform cocultures with GE explants that were as close as possible in size. In cocultures considered for quantification, the mean diameter of the ISS explants was 222 ±31 pm in control cocultures, 221 ± 38 pm in cocultures with anti-netrin-1 antibodies, and 215 ±37 pm in cocultures with control IgG. To quantify results, we counted the total number of bundles growing toward the GE explant under the different experimental conditions.

In the present study we performed cocultures of cortical explants and explants from subregions of the ganglionic eminence, or GE (the lateral ridge of the GE [LGE]), the medial ridge of the GE [MGE], or the sulcus in between [ISS]; see Fig. 1A) to determine whether diffusible factor(s) are involved in the growth of corticofugal axons toward and within the GE. The cocultures were performed in three-dimensional collagen gels, which can stabilize gradients of diffusible guidance molecules secreted by intermediate target tissues such as the floor plate (Lumsden and Davies, 1983; Tessier-Lavigne et al., 1988). Explants were prepared from E12.5 and E13.5 embryos.

In E12.5 mouse embryos, cortical axons have started to extend into the intermediate zone of the cortical wall in a developmental gradient and along trajectories similar to those described in E11 hamster embryos (Métin and Godement, 1996) and E14–E16 rat embryos (Richards et al., 1997). Most axons do not exit the cortical wall at this stage, but axons arising from lateral cortical areas have just entered the mantle zone of the LGE. At E13.5, a larger number of cortical axons have reached the LGE, in particular those arising from cell bodies located rather medially; only a very small proportion of DiI labeled cortical axons are seen to reach the MGE, and the great majority stop within the LGE (Métin and Godement, 1996 and Fig. 8A).

In a first set of experiments, GE explants were cultured at a distance from rostrocaudal strips of cortex in collagen gels (see Figs 1B- F and 2C).

In cortex/LGE cocultures, with explants separated by less than 250–300 μm, rather straight bundles of axons developed from the edge of the cortical explant facing the LGE explant (Fig. 1B,E). Isochronic cocultures of E12.5 and E13.5 explants gave comparable results (Fig. 2E), although LGE explants from E12.5 mouse embryos appeared slightly more effective than those from E13.5 embryos in promoting outgrowth of cortical axons. The amount of axon growth elicited by LGE explants depended on their distance from the cortical explants. Indeed, large numbers of thick bundles were only observed in the region opposite LGE explants that were positioned less than 200 μm away. Axonal outgrowth, measured as the mean number of cortical bundles, was significantly higher in the region facing the LGE explant than in more lateral regions (Fig. 2C,E; P=0.0001). When the shortest distance between cortical and LGE explants was more than 250 μm, differences in axonal growth between the two explants and outside the intervening zone of the coculture could not be reliably detected. In control cultures of cortex, as in cocultures of explants separated by more than 300 pm, a few isolated cortical axons extended into the collagen gel. Their length did not appear to differ from that of cortical axons growing close to LGE explants but their trajectories appeared much more sinuous. At both embryonic stages studied, the effect of LGE explants on cortical explants varied even when distances between explants were similar (see standard deviation in Fig. 2E). This could reflect a variation in explant size or, perhaps more likely, a functional heterogeneity in the pieces of LGE used in this study. However, we did not observe a significant difference when testing E13.5 LGE explants that only consisted of either the ventricular zone or the mantle zone.

In cortex/ISS cocultures, thick, straight bundles of cortical axons developed between both explants (Fig. 1C). The axon outgrowth resembled that observed in cortex/LGE cocultures. (i) The mean number of cortical bundles growing between cortex and ISS explants separated by less than 200 p m (5.86±3.31 bundles per 100 p m of cortical explant, n=19) did not differ statistically from that observed in similar cortex/LGE cocultures (5.51±2.68 bundles per 100 μm of cortical explant, n=18; i-test, P=0.48). (ii) Outgrowth was not stimulated when the distance between the cortex and ISS explants exceeded 250 μm, and in cocultures with explants less than 200 μm apart, the outgrowth of cortical bundles was significantly higher in the region facing the ISS explant than in more lateral regions (Fig. 2E; P=0.0001). (iii) Among cocultures of explants separated by roughly the same distance, large variations were observed in the number of cortical bundles extending toward ISS explants (see standard deviation in Fig. 2E). Similar results were obtained with ISS explants from both caudal and rostral origins (data not shown). As observed with LGE explants, the mean length of cortical axons did not vary significantly with the distance to the ISS explant.

Cortex/MGE cocultures gave a strikingly different result. Explants from the rostral, caudal, lateral and medial MGE were tested in cocultures. None was seen to elicit significant outgrowth of cortical axons in cocultures of either E12.5 or E13.5 tissues (Fig. 1D,F). The mean number of cortical bundles growing from the region opposite MGE explants less than 200 μm away (1.28±1.38 per 100 μm, n=25) was much lower than the mean number of bundles growing opposite LGE or ISS explants at the same distances (see above; P=0.0001 in both cases). Thin fascicles of cortical axons were occasionally observed between cortical and MGE explants. However, the outgrowth of cortical axons close to MGE explants (explants less than 200 μm away) did not differ statistically from that observed from explants placed at a greater distance from MGE explants (Fig. 2E; f-test, P=0.1692).

Thus, explants from the lateral ridge of the GE and from the sulcus between the two ridges of the GE, but not explants from the medial ridge of the GE, promote the outgrowth of axons from cortical explants located up to 200 μm away.

To test whether LGE or ISS explants not only promote axonal outgrowth from cortical explants but can also orient the growth of extending axons, we used the coculture paradigm illustrated in Fig. 3E (see Methods).

In explants from E12.5 mouse embryos, DiI-labeled cortical axons followed roughly dorsoventral trajectories in parietal explants (Fig. 3B), and rostrally directed trajectories in occipital explants (Fig. 3C). After 2 days in culture, a large proportion of cortical axons had grown as far as the ventral edge of the explant (compare D with B and C in Fig. 3), though in a few cases a small proportion of axons deviated toward one of the lateral borders of the explant (data not shown). We therefore established cocultures with one GE explant flanking the rostral edge and another GE explant flanking the caudal edge of the cortical explant (similar results were obtained with LGE and ISS explants). If the GE explants can attract axons, we would expect the two GE explants to produce divergent trajectories of axons within the same cortical explant, a situation which was never observed within control cortical explants.

In these cocultures, DiI-labeled cortical axons were oriented toward the ectopic GE explants both at the rostral and caudal ends of the cortical explant (Fig. 3E and Table 1). In each coculture, several (at least two) DiI injections were performed along a rostrocaudal line. Cortical axons that were directed toward a GE explant were always labeled from an injection site less than 200 to 400 μm from this explant. Because both GE explants could equally influence axonal growth in the center of the cortical explant and abolish turns in this region, we also established cocultures with only one GE explant at one of the lateral edges of the cortical explant. In these cocultures, cortical axons that deviated toward the GE explant were labeled from cortical regions at similar distances (200–400 μm) from the GE explant. Axons labeled from the center of the cortical explant actually reached the ventral edge of the cortical explant, but were shorter than cortical axons growing toward the LGE or ISS explants (see Fig. 3E). This observation could not be quantified owing to the size of the injection sites.

In contrast, we did not observe diverging trajectories of cortical axons in cocultures of cortical and MGE explants (except in one case, see Table 1); most DiI-labeled cortical axons reached the ventral end of the cortical explant (Fig. 3F), as observed in control cortical explants (Fig. 3D).

netrin-1 is expressed in the basal telencephalon of mice and rats at early developmental stages (Serafini et al., 1996; Livesey and Hunt, 1997). To assess its potential involvement in cortical axon guidance toward the GE, we performed a detailed in situ hybridization study of the expression of netrin- 1 in the GE at developmental stages where in vitro experiments showed effects of LGE and ISS explants on cortical axons.

At E12.5, netrin-1 was expressed across the entire medio-lateral and rostrocaudal extents of the ventricular zone of the GE (except for a small approx. 100 μm wide rostrocaudal band along the ventricular angle) and in a few cells of the mantle zone (Fig. 4A–D). There were clear differences in signal intensity in the ventricular zone both along the rostrocaudal axis and along the mediolateral axis of the telencephalon, as observed in flattened whole-mounted vesicles (Fig. 4E). The expression level appeared higher in the lateral ridge and in the sulcus between the two ridges than in the medial ridge of the GE; otherwise, it decreased smoothly from the rostral pole toward the caudal pole of the GE. Cells of the mantle zone exhibited the same gradient of netrin-1 mRNA, being more strongly labeled in the rostral and lateral part of the GE.

At E13.5, the pattern of expression of netrin-1 still exhibited the same spatial characteristics along the rostrocaudal and mediolateral axes as at E12.5 (Fig. 4F-H). The most important change was observed along the radial dimension of the brain: the number of cells expressing netrin-1 in the thicker mantle zone at this age had considerably increased, and the ventricular zone still expressed netrin-1 but was thinner than the day before. We also observed that the gradient of netrin-1 expression along the rostrocaudal axis was steeper than at E12.5.

Netrin-1 transcripts were thus strongly expressed in the two regions of the GE that promote axonal outgrowth from cortical explants and induce turning of cortical axons in vitro. Their level of expression was lower in the MGE, which showed neither a growth-promoting effect nor a turning effect on cortical axons in vitro. This correlation between netrin-1 expression and outgrowth-promoting activity prompted us to test the potential involvement of netrin-1 in mediating the effects of the LGE and ISS.

Recombinant netrin-1 protein, when added to the culture medium, had a strong outgrowth-promoting effect on cortical explants cultured in collagen gels. In control cortical explants cultured without recombinant netrin-1, only sparse outgrowth was observed (Fig. 5A). When netrin-1 was added to the medium at a concentration of 100 ng/ml or higher, the number of axons leaving the cortical explants increased dramatically (Fig. 5C). At low netrin-1 concentrations (approx. 100 ng/ml), axons formed thin fascicles. The number and width of fascicles increased with netrin-1 concentration up to an optimal value (near 400 ng/ml), and decreased thereafter (Fig. 5A,C). The response to higher concentrations varied between experiments. In some series, higher concentrations were less efficient than optimal concentrations, whereas in other series (e.g. Fig. 5C), the difference was not significant. Because results were quantified without distinguishing between fibers and fascicles, we underestimated the efficacy of optimal netrin-1 concentrations. The mean length of cortical fibers or fascicles did not change markedly with netrin-1 concentration (Fig. 5A,B). It was slightly increased in the optimal range of netrin-1 concentration (near 400 ng/ml). However, the ratios between the maximum and minimum length values were close to 1.5, much less than the ratios between the maximum and minimum numbers of fibers and/or fascicles (from 3.6 to 5.4; see Fig. 5B,C). These experiments show that in vitro, early cortical neurons do respond to the netrin-1 protein, and that netrin-1 promotes the growth of axons from cortical explants in a dosedependent manner.

Aggregates of 293-EBNA cells secreting netrin-1 stimulated axonal outgrowth from cortical explants in collagen gel cocultures (Fig. 6A,C). In contrast, in cultures with aggregates of parental 293-EBNA cells only sparse and isolated cortical axons were observed around the cortical explant. In most cases, the density of axons growing around cortical explants cocultured with parental 293-EBNA cells was even lower than the density of axons growing around cortical explants cultured alone (compare with control cultures in Fig. 5A), suggesting a mild inhibitory effect of the parental cell line on cortical neurons.

The outgrowth elicited by netrin-1-secreting cells on cortical explants exhibited two characteristics that distinguished it from the outgrowth observed in cortex/LGE or cortex/ISS cocultures: (i) Unlike in cortex/LGE or cortex/ISS cocultures, aggregates of netrin-1-secreting cells promoted axonal outgrowth in regions of the cortical explants located more than 200 μm from the aggregates (Fig. 6A). (ii) Cortical axons growing toward the netrin-1-secreting cells were less fasciculated than cortical axons growing opposite LGE or ISS explants (compare Figs 1 and 7A with 6A), except in regions of the cortical explants that were further away from the cell aggregates, where axon outgrowth was also fasciculated. These two differences between outgrowth elicited by cell aggregates and by GE explants could be accounted for if the netrin-1-secreting cell aggregates secrete higher levels of netrin-1 than do the explants, and if high concentrations of netrin-1 elicit less fasciculated outgrowth than do lower concentrations.

When aggregates of cells secreting netrin-1 were placed at the rostral and caudal edges of large cortical explants, a large proportion of cortical axons deviated from their normal dorsoventral trajectory, and were directed toward the closest aggregate of netrin-1-secreting cells. Such deviations were never observed within explants cocultured with aggregates of control cells (Fig. 6D,E and Table 1). The proportion of turning axons and the distance over which such deviations were observed in cocultures of cortical explants with aggregates of netrin-1-secreting cells were never as great as those observed in cocultures of LGE and ISS tissue.

When cortical explants were cultured with LGE explants or ISS explants in collagen gels in the presence of 10 μg/ml of a function blocking anti-netrin-1 antiserum, outgrowth of cortical axons normally elicited by the GE explant was completely abolished in most cases, or, where still observed, significantly reduced compared to control cultures (Fig. 7 and data not shown). In contrast to the large bundles of axons that grow towards GE explants in control cultures, in the presence of the antiserum only sparse axon outgrowth from the cortical explants as well as from the GE explants was observed, showing that the antiserum did not abolish all axon outgrowth in a non-specific way (Fig. 7B1,B2). The control non-immune rabbit serum, when added at the same concentration, did not have a statistically significant effect on the outgrowth elicited by the GE explants (Fig. 7C,D,E).

These results are consistent with the hypothesis that a netrin (likely netrin-1) is largely or completely responsible for the axonal outgrowth-promoting activity of LGE and ISS explants.

We have shown that LGE and ISS, but not MGE explants promote axonal outgrowth from early (E12.5-E13.5) cortical explants placed at a distance in a three dimensional matrix. Furthermore, LGE and ISS explants are able to deflect corticofugal fibers from their normal trajectories. We also present several converging lines of evidence suggesting that netrin-1 is largely or entirely responsible for the growth promoting action of LGE and ISS explants on cortical explants, and likely contributes to its chemotropic activity. (1) These GE subregions express netrin-1 transcripts in vivo. (2) Soluble recombinant netrin-1 protein elicits outgrowth of axons from cortical explants in vitro in a dose-dependent manner and with a time delay similar to that elicited by LGE and ISS explants. (3) The outgrowth of cortical axons is oriented toward a restricted source of netrin-1 protein in a three dimensional culture matrix. (4) A function-blocking anti-netrin antiserum, but not a non-immune rabbit antiserum, largely suppresses the outgrowth-promoting effect of LGE and ISS explants on cortical explants.

Our results agree with those of Richards et al. (1977) who demonstrated a directional effect of nascent internal capsule (i.e. the mantle zone of the GE) and cells secreting netrin-1 on axons from embryonic rat cortical explants. They also extend those results (1) by showing that GE tissue and netrin-1 can promote outgrowth of cortical axons into collagen gels under appropriate experimental conditions (see discussion below), (2) by providing direct evidence, through the use of the function-blocking antiserum, that netrin-1 contributes to the effect of GE tissue, and (3) by defining more precisely the regions of the nascent internal capsule that express netrin-1 and have these activities on cortical efferents (the LGE and ISS).

The lateral ridge and sulcus of the GE produce a diffusible factor(s) capable of inducing rapid (10-12 hours) and extensive axon outgrowth from cortical explants, when compared with the faint spontaneous outgrowth observed in cortical explants cultured alone. This outgrowth-promoting effect occurs over a distance of about 200 pm around the GE explants, not unlike the maximal distance over which explants of floor plate tissue can induce outgrowth of spinal commissural axons in vitro (Tessier-Lavigne et al., 1988; Placzek et al., 1990). In E13.5 cocultures, the same effect is also observed with explants from the mantle zone of the LGE, a region where the internal capsule will differentiate.

LGE and ISS explants do not simply promote outgrowth of cortical axons into the collagen matrix, but can also actually attract these axons. This was demonstrated by coculturing LGE/ISS explants with large cortical explants. In these cocultures, ectopic LGE or ISS explants were seen to attract cortical axons located more than 200 pm away, which turned towards the target explants within the cortical explant. In accordance with the ventrodorsal (i.e. lateromedial) gradient of axogen- esis found in vivo in the cortical wall (Métin and Godement, 1996; Richards et al., 1997), axons labeled from ventral injection sites exhibited trajectory reorientations during the coculture time, whereas axons labeled from dorsal sites that developed later displayed trajectories that were directed from the outset toward ectopic LGE/ISS explants. In the same cortical explant, axons directed toward ectopic GE explants were longer than axons reaching the ventral edge, suggesting that axonal elongation might also have been enhanced. Our experiments did not allow us to determine whether LGE and ISS explants also trigger the de novo formation of neurites by cortical neurons. Studies on cultures of dissociated cortical neurons will be necessary to answer this question.

Although turning occurred within cortical explants, turning of cortical axons towards LGE and ISS explants was not observed once the axons had exited the explant and were growing within the collagen matrix. This observation parallels similar observations on the effects of floor plate cells on commissural axons, since floor plate cells can cause turning of commissural axons within explants of dorsal spinal cord but not within collagen matrices (Tessier-Lavigne et al., 1988; Placzek et al., 1990). Thus, the collagen matrix assay does not seem well adapted to demonstrate a turn of cortical fibers (this study and Richards et al., 1997), perhaps because it does not provide a favorable growth environment. Indeed, cortical axons extending within the collagen matrix towards LGE or ISS explants are highly fasciculated (see Fig. 7A1), suggesting that axons prefer to grow on one another rather than in direct contact with the collagen gel. Mechanical constraints might then make turns of fasciculated fibers much more difficult to initiate than turns of unfascicu- lated fibers within cortical explants.

The finding of a strong outgrowth-promoting effect of LGE and ISS tissue in our assay might appear slightly at odds with the results of Richards et al. (1997) who found that the presumptive internal capsule (i.e., LGE) or 293T cells expressing netrin-1 did not increase the total number of axon bundles emerging from small pieces of cortex, but caused a redistribution of these bundles such that more were directed towards the internal capsule or netrin-secreting cells, and fewer away. The apparent discrepancy can, however, be explained if we assume that the collagen used in their experiments was more permissive for cortical axon growth than that used here. Consistent with this assumption, the amount of background outgrowth they observed from cortical explants cultured alone was much greater than in our experiments. Under their conditions, the permissive action of netrin-1 (or the LGE) might not be as easily detected as in our experiments, where the collagen was relatively non-permissive for growth. However, floor plate explants, which might have provided a better source of netrin- 1 protein, did increase the total number of axon bundles in their experiments (Richards et al., 1997), suggesting that their collagen did not provide a perfectly permissive substrate for axon growth, and that a high enough netrin-1 concentration could overcome its residual non-permissiveness.

We have not identified the types of cortical cells responding to LGE and ISS explants. Cortical explants from E12.5 and E13.5 mouse embryos comprise neuronal and glial progenitors, precocious neurons and radial glial cells (Angevine and Sidman, 1961; Luskin et al., 1988; Gadisseux et al., 1990). The earliest efferent cortical neurons are neurons in the subplate and in the deep cortical plate-presumptive layer VI cells (Marin-Padilla, 1988). DiI injections into large cortical explants labelled both axons that had reoriented growth toward ectopic GE explants and axons with normal ventrally directed trajectories, suggesting that not all cortical axons respond equally to the diffusible cue(s) secreted by the LGE and ISS explants. The netrin-1 receptor DCC appears to be expressed on Cajal-Retzius cells (Keino-Masu et al., 1996), which are association neurons in the embryonic mammalian cortex (Marin-Padilla, 1988). However, further studies will be required to determine whether other classes of cortical cells express DCC or other netrin receptors, such as neogenin (Keino-Masu et al., 1996) or vertebrate UNC-5 homologues (Leonardo et al., 1997; Ackerman et al., 1997).

The regions of the GE that promote cortical axon outgrowth and induce cortical axon turning in coculture experiments are the same as those that strongly express netrin-1. In early mouse embryos, netrin-1 is expressed by a continuum of cells in the ventricular zone of the ganglionic eminence, with a peak of expression in the medial part of the LGE that ends in the sulcus separating the two ridges of the GE. Although we have not examined the distribution of netrin-1 protein, it seems likely that its extracellular concentration might peak in the medial part of the LGE and around the ISS.

In older embryos, netrin-1 expression becomes reduced in the ventricular zone, whereas individual cells in the mantle zone strongly expressing netrin-1 increase in number. As observed in the cortical wall (Rakic, 1972; Gadisseux et al., 1990), postmitotic cells in the GE migrate from the ventricular zone towards the mantle zone of the GE (embryonic basal ganglia) along radial glial fibers (Halliday and Cepko, 1992; Misson et al., 1988; Liu and Graybiel, 1992). Whether progenitors expressing netrin-1 transcripts in the ventricular zone give rise to postmitotic cells in the mantle zone of the GE that express netrin-1 remains to be determined.

The ventricular zone contains both the cell bodies of radial glial cells and progenitor cells. The pattern of netrin-1 expression in the mouse basal telencephalon might therefore follow a temporal evolution showing similarities with that observed in the cephalic region of the nematode C. elegans. In early worm embryos, the netrin UNC-6 is first expressed in neuroglial cells, the ventral epidermoblasts and cephalic sheaths. It becomes restricted thereafter to discrete cellular entities along axonal pathways (Wadsworth et al., 1996). Similarly, the first cells to express netrin-1 in the GE might be radial glial cells, whose processes span the entire radial extent of the GE and are extremely dense in both the LGE and the MGE (Misson et al., 1988; Liu and Graybiel, 1992; Halliday and Cepko, 1992; C. Métin and P. Godement, unpublished data). The expression of netrin-1 mRNA might thereafter shift toward cell bodies in the mantle zone of the GE, where differentiated postmitotic neurons are located (Porteus et al., 1994).

Corticofugal axons, which can be visualized in frontal sections of E12.5 and E13.5 mouse embryos, are dorsoventrally oriented within the intermediate zone of the cortical wall, and make a turn at the ventricular angle suggestive of a change in guidance cues at this level (Métin and Godement, 1996 and Fig. 8A).

Before they reach the ventricular angle, cortical axons might be guided by cues that are intrinsic to the cortex or that diffuse there from extrinsic sources. Extrinsic diffusible cues secreted by the basal telencephalon, including netrin-1, could direct the growth of cortical axons within the cortical wall, as was recently suggested by Richards et al. (1997). However, this would imply that the netrin-1 protein diffuses in vivo for a greater distance than in collagen gels. Within large cortical explants cultured for several days, the growing axons followed trajectories that are oriented roughly ventrally, as observed in vivo. This would argue in favor of cues intrinsic to the cortex (either diffusible or involving contact interactions) that direct the growth of cortical axons within the cortical wall, rather than in favor of a preponderant role of netrin-1 secreted by an extrinsic source.

Instead, the pattern of netrin-1 transcripts and the in vitro range of action of the netrin-1 protein correlate well with an involvement of netrin-1 in the guidance of corticofugal axons at the ventricular angle and within the basal telencephalon in early mouse embryos. If we assume that the in vivo range of action of netrin-1 protein is similar to that observed in coculture experiments (approx. 200 - 400 pm), the growth cones of corticofugal axons, after having reached the ventricular angle, might be attracted toward the LGE and ISS regions by an increasing gradient of netrin-1 (scheme Fig. 8B). The netrin-1 protein secreted in the GE could, moreover, stimulate the growth of cortical axons within the LGE.

Netrin-1 may only account for part of the turning activity of GE and ISS tissues since in collagen gel cocultures, LGE-ISS explants are more potent than netrin-1-secreting 293 cells at inducing turning of cortical axons. These tissues might thus secrete a second chemoattractant for corticofugal axons, distinct from netrin-1, as has also been described for floor plate cells (Serafini et al., 1996). Both floor plate cells and cells secreting netrin-1 can promote outgrowth of commissural axons from explants of dorsal spinal cord and can cause turning of commissural axons within these explants (Tessier-Lavigne et al., 1988; Placzek et al., 1990; Serafini et al., 1994; Kennedy et al., 1994), but analysis of a netrin-1-deficient mouse indicates that, while netrin-1 appears to account for all of the outgrowth-promoting activity of floor plate cells, it accounts for only part of the turning activity of these cells (Serafini et al., 1996). It is possible that the same is true for the LGE and ISS.

In vivo, corticofugal axons wait for about 24 hours in the LGE before pursuing their route into the MGE. A change in the slope of the netrin-1 gradient, as well as transient local stop or repulsive signals could in principle cause cortical axons to pause before the MGE. They might be waiting for interactions with other axons such as reciprocal thalamocortical axons, locally change their characteristics (Dodd et al., 1988), or be modified by local cellular or molecular factors (Campbell and Peterson, 1993), thus allowing them to extend across the MGE compartment even though it expresses low levels of netrin-1 transcripts, and appears non-permissive for cortical axons in millicell cocultures (Métin and Godement, 1996).

As in the ventral spinal cord, where commissural axons do not approach the ventricular zone (Colamarino and Tessier- Lavigne, 1995b; Serafini et al., 1996), corticofugal fibers extending into the LGE never enter the ventricular neuroepithelium (Métin and Godement, 1996), even though it strongly expresses netrin-1. This is also seen in vitro, where cortical axons growing in collagen gels towards LGE or ISS explants constituting the ventricular zone stop abruptly just before reaching the attractive explant; in two dimensional cocultures in “millicell” chambers, cortical axons form large stumps on the ventricular face of the LGE explants (C. M., unpublished observations). This suggests that the ventricular neuroepithelium is an unfavorable environment for cortical axon outgrowth, and that its inhibitory action cannot be overridden by netrin-1.

Taken together, the results of the present study strengthen our previous hypothesis that the longitudinal sulcus in the GE constitutes a landmark separating two functionally distinct compartments in the developing basal telencephalon (Métin and Godement, 1996). This functional partition together with the netrin-1 expression pattern suggest an early patterning of the telencephalon along its longitudinal axis, which fits with early gene expression domains in the forebrain (Bulfone et al., 1993) and which might be superimposed on the transverse segmentation of the forebrain (Puelles and Rubenstein, 1993). This study and that of Richards et al. (1997) also provide evidence for a role of netrin-1 in the guidance of corticofugal axons across the basal telencephalon. Corticofugal projections thus provide an example of non-commissural projections in the forebrain of mammals whose early stages of development are likely to be controlled by chemoattractant(s), as was alsso observed for retinal ganglion cell axons (Deiner et al., 1997). An examination of the projections of these early efferent fibers in netrin-1-deficient mice (Serafini et al., 1996) should make it possible to determine the extent to which this netrin-mediated guidance is required for normal pathfinding in vivo.

We thank Dr Y von Boxberg for critically reading the manuscript, Dr P Godement for supporting this work, and S. Faynboym for providing the netrin-1 protein. Supported by grants from the International Spinal Research Trust, Human Frontiers Science Program and American Paralysis Association (M.T.-L.) and by fellowships from the American Cancer Society (T. S.) and the Spinal Cord Research Foundation (T. E. K.). M. T.-L. is an Investigator of the Howard Hughes Medical Institute.