Orientation of commissural axons in vitro in response to a floor plate-derived chemoattractant

Axonal growth cones are guided to their targets in the developing nervous system by cues in their local environment. Many of these cues take the form of molecules that are expressed in the extracellular matrix or on the surface of cells along the pathways taken by developing axons. Some molecules, such as N-CAM and N-cadherin, appear to provide adhesive substrates that are favorable for growth cone extension but may not play a primary role in determining the direction of axonal growth. Other molecules that promote axon extension, such as laminin, are present in much more restricted regions of the developing nervous system and may influence the pathway choice of axons (Jessell, 1988; Takeichi, 1988; Doherty et al. 1989; Tomaselli and Reichardt, 1989).

In addition to cues derived from cells along axonal pathways, there is increasing evidence that growth cones can orient in response to gradients of chemoattractant molecules that are selectively released by the intermediate or final targets of developing axons (Ramon y Cajal, 1909). Although chemotropic guidance of axons has been invoked on repeated occasions, the evidence in favor of this mechanism is still fragmentary. A point source of the trophic molecule nerve growth factor (NGF) has been shown to attract sensory axons in vitro (Letourneau, 1978; Gundersen and Barrett, 1979) and sympathetic axons in vivo (Menesini-Chen et al. 1978), but it is still unclear whether NGF normally exerts a tropic action in the guidance of axons within the developing embryo (see Davies, 1987). Certain classes of sensory axons may be guided by chemotropic factors other than NGF. Trigeminal sensory neurons extend axons in vitro in response to a local signal secreted by one of their normal target tissues, the maxillary epithelium (Lumsden and Davies, 1983, 1986). Similar in vitro studies have suggested that the axons of corticospinal projection neurons extend collateral branches into one of their final targets, the basilar pons, in response to a diffusible factor secreted by the pontine tissue (Heffner et al. 1990).

We have been examining the mechanisms that guide the axons of commissural neurons in the embryonic rat spinal cord. Soon after commissural neurons differentiate, their axons extend ventrally, close to the lateral edge of the spinal cord (Fig. 1A; Holley, 1982; Altman and Bayer, 1984; Wentworth, 1984; Dodd et al. 1988). When the axons reach a point just dorsal to the motor column, they alter their trajectory and project through the motor column towards the floor plate, a specialized group of epithelial cells located at the ventral midline of the spinal cord. The directed growth of commissural axons towards the ventral midline raises the possibility that the axons are guided to the ventral midline by a chemoattractant molecule secreted by the floor plate. In support of this idea, we have shown that the floor plate secretes a diffusible factor (or factors) that promotes the outgrowth of commissural axons from spinal cord explants in vitro and can deflect axons from their initial trajectory towards a local cellular source of the factor (Tessier-Lavigne et al. 1988).

In this study, we extend our previous observations by providing a quantitative assessment of the effectiveness of the floor plate factor in reorienting axon growth. We establish that the factor can diffuse through neural epithelium over considerable distances, and that the factor can override intrinsic polarity cues within the epithelium to reorient essentially all commissural axons. Previously, we found that the presence of a floor plate did not affect the density and length of axons within dorsal spinal cord explants (Tessier-Lavigne et al. 1988). Here, we extend those observations by showing that the orienting effect of the floor plate occurs in the absence of any detectable effect on the number of surviving commissural neurons, or on the rate of commissural axon growth within neural epithelium or on two-dimensional substrates. These findings provide further evidence that the floor plate guides commissural axons to the ventral midline in the developing embryo.

Two segment pieces of the spinal cords of 24 – 27 somite rat embryos (E11 – E11.5) were dissected and embedded in threedimensional collagen gels as described previously (Fig. 1B and Tessier-Lavigne et al. 1988). The culture medium was Ham’s F12 (Gibco) supplemented with the following: glucose (8mg ml^-1), glutamine (2 MM), MEM vitamins (Gibco, 1 %), penicillin (100 units ml^-1), streptomycin (100 μg ml^-1), transferrin (0.2 mg ml^-1), insulin (10 μg ml^-1), bovine serum albumin (10 μg ml^-1), triiodothyronine (20ngml^-1), sodium selenite (10ngml^-1), putrescine dihydrochloride (32 μg ml^-1), progesterone (13 ng ml^-1) and corticosterone (40 ng ml^-1 ) (Ham’s F12S). Explants were cultured in Ham’s F12S containing 5 % horse serum (Gibco) at 37 °C in a 5 % CO₂ environment.

Floor plates dissected from E11 and E13 rat embryos, and the remainder of the E11 spinal cord, were collected in L15 medium. E11 floor plate, or tissue from the remainder of the spinal cord, was cultured in 4-well dishes (Nunc) treated with poly-D-lysine (Sigma, 20 μg ml^-1) and laminin (Collaborative Research, 20 μg ml^-1). E13 floor plates were cultured in 35 mm dishes (Nunc), with or without prior coating with poly-D-lysine/laminin. After overnight incubation in Ham’s F12S containing 5% horse serum, tissues were rinsed twice in serum-free Ham’s F12S, and incubated for a further 48–60 h. Floor plates or the remainder of the spinal cord dissected from 100 E11 embryos, or floor plates from 40 E13 embryos were used to condition 1ml medium. Conditioned medium was aliquoted and frozen at – 20 °C for periods up to a month without detectable loss of activity. Before use, conditioned medium was diluted with an equal volume of fresh Ham’s F12S medium containing 10% horse serum.

The top collagen layer was removed and explants were floated off the bottom collagen cushion, then incubated in enzyme-free dissociation medium (Speciality Media; New Jersey) for 20 min. The tissue was then rinsed and triturated to a single cell suspension and plated into 35 mm tissue culture dishes (Nunc) coated with poly-D-lysine and laminin in Ham’s F12S medium containing 5 % horse serum.

The dorsal third of the E11 spinal cord was incubated in enzyme-free dissociation medium for 20min. The tissue was then rinsed and triturated to give a single cell suspension. Cells were plated onto 35 mm 4-well tissue culture dishes (Sterilin) treated with poly-D-lysine (20 μg ml^-1), collagen or poly-D-lysine and laminin (see above), and incubated in Ham’s F12S with 5 % horse serum or in E13 floor plate-conditioned medium.

Explants cultured in collagen gels were processed for immunocytochemistry as previously described (Tessier-Lavigne et al. 1988). TAG-1 was detected using monoclonal antibody 4D7 (Yamamoto et al. 1986) diluted at 1:1 or rabbit antiTAG-1 antibodies (Dodd et al. 1988) diluted at 1:2000 in phosphate-buffered saline, pH7.4 (PBS) containing 0.1% Triton X-100 and 1% heat-inactivated goat serum (FUGS). Monoclonal antibody 2H3 (available from Developmental Studies Hybridoma Bank) was used as a general axonal marker and monoclonal antibody 5A5, which detects the polysialylated form of N-CAM, was used as a neuronal marker. Both were diluted 1:1 with PBS containing 0.1% Triton X-100. Peroxidase- and FITC-conjugated secondary antibodies (TAGO and Boehringer Mannheim) were used at a 1:100 dilution in PBS with 1 % HIGS.

For surface labelling, cultures were washed with L15 supplemented with 8mg ml^-1 glucose and 0.1 % BSA (L15⁺) and then incubated with primary antibodies for 30 min at 22°C. Cultures were washed twice in L15⁺ with 1% HIGS, and incubated with FlTC-conjugated secondary antibody diluted 1:100 in the same buffer for 30 min at 22 °C. Cultures were then washed twice and fixed in 4 % paraformaldehyde in 0.12 M phosphate buffer (pH 7.4) for 15 min, washed twice in L15⁺ and coverslipped in 0.04% paraphenylenediamine (Sigma) in 0.1M sodium carbonate (pH9.0), 50% glycerol. Cells were counted and measurements of axon lengths were performed on a Zeiss Axioplan microscope under phase and epifluorescence optics using a videomicroscope (Imaging Technology Inc.).

Retrograde or anterograde labelling of commissural neurons in dorsal spinal cord explants was performed with Di-I (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate, Molecular Probes; 2.5mgml^-1 in dimethylformamide; see Honig and Hume, 1986, and Godement et al. 1987). Explants were fixed in 4% paraformaldehyde in 0.12 M phosphate buffer (pH 7.4) for 24 – 48 h and washed extensively in PBS. The top layer of collagen above the region to be injected was removed and a fine incision made in the tissue using a drawn-out glass microcapillary. Between 5 and 30 nl of Di-I was injected into the tissue using a similar microcapillary attached to a mouth pipette. Dil was allowed to diffuse at 4 °C for 1 – 5 days and explants were then examined under a Zeiss Axioplan with epifluorescence optics.

Commissural neurons, which can be identified by expression of the TAG-1 glycoprotein (Dodd et al. 1988; Furley et al. 1990), begin to extend axons around embryonic day (E) 11. When explants of E11 dorsal spinal cord are cultured in a collagen gel matrix, commissural axons grow along a dorsoventral trajectory similar to that taken in vivo (Fig. 1A) until they reach the cut ventral edge of the explant (Tessier-Lavigne et al. 1988). Commissural axons project into the collagen matrix in response to a diffusible factor (or factors) secreted by the floor plate (Tessier-Lavigne et al. 1988). Different patterns of outgrowth are evoked by placing a floor plate explant opposite or to one side of the ventral edge of the dorsal explant (Fig. 2). The direction of commissural axon growth is biased towards the floor plate. This results at least in part from the reorientation of axons within the neural epithelium before they emerge into the collagen gel (Tessier-Lavigne et al. 1988).

To provide information on the orientation of commissural axons in relation to the source of floor plate factor, we examined the angle of emergence of axons from the ventral edge of dorsal explants under two conditions: (i) with a uniform concentration of the factor (in the form of conditioned medium); and (ii) with a local source of the factor (achieved by placing a floor plate explant opposite or lateral to the ventral edge).

In the presence of conditioned medium, axons emerged from both the lateral and ventral edges of the explant and, to a lesser degree, from the dorsal edge of the explant (Fig. 3). As shown in Fig. 4A, axons emerged from the ventral edge of these explants with a mean angle of 5.5°±28.6° (S.D.; n=271 bundles from 10 explants), where 0° indicates an angle of emergence perpendicular to the cut edge of the explant. The range of angles was broad but exhibited an approximately Gaussian distribution, centered around the perpendicular. The clustering of emerging axons around the perpendicular probably reflects the existence of intrinsic cues that direct axons along a dorsoventral trajectory within the explant.

We next examined the extent to which the angle of axon emergence is altered when outgrowth is evoked by a floor plate explant. With a floor plate placed opposite the ventral edge of the dorsal explant (Fig. 4B), the mean angle of emergence was 0.2°±18.0° (S.D.; n=274). The angles of emergence were clustered more sharply around the perpendicular, with about half (50.4%) of the bundles emerging within ±10° of the perpendicular, compared to 27.5 % in the presence of conditioned medium. Thus, although cues intrinsic to the dorsal explant direct commissural axons along a dorsoventral trajectory within the explant, a localized source of floor plate factor with its high point opposite the ventral edge of the explant can sharpen the distribution of angles at which the axons emerge.

The ability of the floor plate to orient commissural axons was further assessed in experiments in which the floor plate explant was positioned 200 – 400 μm from one of the lateral edges of the dorsal explant (Fig. 4C). Most (91.2%) of the bundles that emerged were oriented towards the floor plate (Fig. 4C), with a mean angle of emergence of 26.5°±21.4° (S.D.; n=113). Under these conditions therefore the floor plate is capable of orienting the growth of virtually all emerging commissural axons.

Once in the collagen gel, axons did not markedly reorient their growth (data not shown) which may be a consequence of their pronounced fasciculation after they emerge from the dorsal explant (Figs 2 and 3). To avoid problems of interpretation caused by the fasciculation of axons within the artificial environment of the collagen gel, we examined the effect of the floor plate on commissural axonal trajectories within the dorsal neural epithelium. To approach this, it was first necessary to determine the trajectory of different populations of axons within the neural epithelium in the absence of floor plate.

Within dorsal explants cultured alone, TAG-1⁺ commissural axons grow dorsoventrally until they reach the cut ventral edge of the explant (Tessier-Lavigne et al. 1988; Fig. 5A). At the time of isolation almost no commissural axons have extended within the explants (data not shown). Thus in vitro these axons grow along a dorsoventral trajectory in response to polarity cues that are intrinsic to the dorsal neural epithelium.

In addition to the TAG-1⁺ axons oriented along the dorsoventral axis, there was a population of axons that grew parallel to the ventral edge of the explant (along the original rostro-caudal axis; Fig. 5B). These axons did not express TAG-1, but could be labelled with MAb 2H3, which recognizes an antigen present in all developing axons. TAG-1^-/2H3⁺ axons presumably derive from association neurons, which are the only other differentiated neurons in the dorsal spinal cord at this age. Consistent with this, their route was similar to that taken by association axons in vivo (Fig. 1A; Altman and Bayer, 1984; Dodd et al. 1988). In double-labelling experiments, we did not observe any TAG-1^-/2H3 ⁺ axons oriented along the dorsoventral axis (data not shown), although a small population of axons could have been obscured by the high density of TAG-1⁺ commissural axons that are oriented along this axis. These observations indicate that most, if not all, dorso-ventrally oriented axons derive from commissural neurons.

The effect of the floor plate on the complete trajectory of individual axons within the neural epithelium was initially assessed by dye labelling. Since dorsal explants contain both commissural and association axons, it was first necessary to show that dye injections can label commissural axons selectively.

Injection of Dil into the region near the roof plate of dorsal explants cultured for 40 h selectively labelled dorso ventrally oriented, presumed commissural axons (Fig. 5C). Retrograde dye labelling of the commissural axons that had emerged from dorsal explants cultured for 40 h with a floor plate revealed that these axons were also oriented along the dorsoventral axis within the explant, but derived from neurons located some distance below the roof plate (Fig. 5D). Association axons oriented parallel to the ventral edge were observed only when dye was injected more ventrally within the explant (not shown). It appears, therefore, that commissural axons can be anterogradely labelled by injecting Dil into the roof plate region of dorsal explants and retrogradely labelled by Dil injection near to the tips of axons that have emerged into the collagen gel.

With this background information, we examined the orienting effect of the floor plate on axonal trajectories within dorsal explants. We wished to determine whether the factor is effective at orienting axons in this environment and the distance over which the factor can diffuse through neural epithelium. We therefore modified the experiment of Fig. 2C, as illustrated in Fig. 6A. E11 floor plate explants were apposed to one of the lateral edges of a long (700 – 1000 μm) piece of dorsal spinal cord. After 40h in culture (Fig. 6B), we assessed the trajectory of commissural axons by injecting Dil into the region of the neural epithelium in and adjacent to the roof plate.

The initial growth of Dil-labelled axons originating from this region was along the dorsoventral axis (Fig. 6C). Axons located far from the floor plate continued along this trajectory to the cut ventral edge of the explant. However, axons located closer to the floor plate deviated away from the dorsoventral trajectory and were reoriented towards the floor plate (Fig. 6C). The distance from the floor plate over which axons turned ranged from 155 μm to 375 μm for different explants, with a mean of 243±54 μm (S.D.; n = 15 explants). Within this distance all labelled axons reoriented their growth such that they were directed towards the floor plate explant. At greater distances, axons failed to reorient towards the floor plate and continued to project along the dorsoventral axis.

This experiment did not reveal whether the floor plate affects axons that originate from neurons located more ventrally than the site of dye injection. It was not possible to determine this by injecting dye more ventrally, since this also resulted in the labelling of association axons. Several explants were therefore serially sectioned and labelled with anti-TAG-1 antibodies. The pattern of growth of TAG-1⁺ axons throughout the neural epithelium deduced from thin sections (Fig. 6D,E) was consistent with that of Dil-labelled axons (Fig. 6C). TAG-1⁺ axons within 195±47 μm (S.D.; n=14) reoriented their growth. (This distance was slightly shorter than that estimated from Dillabelling experiments because the explants shrank by ∼7 % when processed for sectioning and because the distance was underestimated in cases where the plane of section obscured the boundary between responsive and unresponsive axons.) Thus, the floor plate can override the intrinsic dorsoventral polarity cues within the dorsal neural epithelium and reorient all commissural axons.

It is possible that the factor reorients axon growth within the neural epithelium only over a short distance, with more distant axons reorienting their growth by fasciculating on axons closer to the floor plate, resulting in a contact-dependent cascade of fasciculation. This seemed unlikely, since TAG-1⁺ axons appeared unfasciculated (Fig. 6D,E; see also Holley et al. 1982; Holley and Silver, 1987). However, we addressed this issue directly by interposing a piece of E11 ventral spinal cord between the floor plate and the dorsal spinal cord. After 40h in culture, these explants were sectioned and labelled with antibodies to TAG-1. The pattern of growth of motor axons in the ventral explants is not influenced by the presence of the floor plate (Tessier-Lavigne et al. 1988). In contrast, commissural axons were reoriented towards the floor plate over a distance of 310±40 μm, under conditions in which the nearest axons were separated from the floor plate by an interposed piece of ventral spinal cord 150 – 220 μm in length (n=7; not shown). These results suggest that the floor plate factor can diffuse long distances through ventral neural epithelium to reorient the growth of commissural axons. This is pertinent since ventral neural epithelium is the normal environment through which the factor must diffuse in the developing embryo.

In these experiments, there was no axon outgrowth into the collagen matrix (compare with Figs 2 and 3). Within the limit of diffusion of the floor plate factor, commissural axons reorient their growth and project through neural epithelium to the floor plate. Beyond the limit of diffusion, the floor plate does not appear to influence commissural axons at all, either by reorienting them within the neural epithelium or by promoting their growth into collagen.

The operation of a tropic mechanism in vivo requires that a gradient of the chemotropic factor is established. We therefore examined the specificity of expression of the floor plate-derived chemoattractant. Previously, we found that the ability to promote growth into collagen was restricted to the floor plate at E11 and largely restricted at E13, although E13 ventral spinal cord also showed a small amount of activity (Tessier-Lavigne et al. 1988). Since we do not know whether the orienting and outgrowth-promoting actions of the floor plate are mediated by the same molecule(s), we tested the ability of different regions of the spinal cord to reorient commissural axon growth within neural epithelium.

Regions of the spinal cord taken from brachial levels were cultured directly apposed to the lateral edge of a 700 – 1000 μm piece of dorsal explant, as shown in Fig. 6A. After 40 h in culture, the projection pattern of commissural axons within the long dorsal explant was examined in serial cryostat sections immunolabelled with anti-TAG-1 antibodies. Commissural axons turned towards floor plate obtained from E11 and E13 spinal cord (not shown). In contrast, axons did not deviate from their dorsoventral trajectory towards E11 dorsal or ventral explants. Thus, at E11, only the floor plate appears to release the chemoattractant. Reorientation was not observed towards E13 dorsal spinal cord. However, reoriented TAG-1⁺ axons were seen with E13 ventral spinal cord from which the floor plate had been removed, although the reorientation occurred only within about 10 μm of the ventral tissue and not within the remainder of the dorsal neural epithelium (not shown). Thus, within the E11 – E13 spinal cord, the regional expression of orienting activity parallels that of the outgrowth promoting activity.

The results described above demonstrate that a localized source of factor, in the form of a tissue explant, can compete effectively with the intrinsic polarity cues within the dorsal explant. With conditioned medium, axons extended into the collagen gel from the lateral and dorsal edges of the explant (Fig. 3), yet the pattern of TAG-1⁺ axons within the explants (Fig. 7) was not detectably different from that in explants cultured alone (Fig. 5A). Thus, a uniform concentration of factor (in the form of conditioned medium) does not prevent most axons from responding to the polarity cues within the neural epithelium. It can, however, override these cues to the extent of directing the growth into the collagen gel of axons located near the dorsal or lateral edges of the explant.

The orienting effect of the floor plate appears to reflect the action of a chemotropic factor. It is possible that the floor plate also has a trophic effect, increasing the total number of commissural neurons surviving in these explants, or affecting the rate of commissural axon growth. This could affect the number of axons growing into the collagen gel in the presence of floor plate or the timing of emergence of these axons.

To determine whether the floor plate has a trophic action, we examined whether the total number of neurons in dorsal explants is altered by the presence of a floor plate explant. Dorsal explants were cultured alone or with a ventrally located floor plate explant. After 40 h, when the pattern of axon outgrowth under these two conditions differed most markedly (Fig. 2A,B), the dorsal explants were removed from the collagen and dissociated into a single cell suspension. Cells were plated on laminin-coated dishes, maintained for 8 – 12 h and then labelled with monoclonal antibody 5A5, which recognizes commissural and association neurons in vivo and in vitro (Dodd et al. 1988; unpublished observations). We did not use TAG-1 as a marker because its expression is transient on commissural neurons in vivo and in vitro and therefore does not provide an accurate indication of the total number of commissural neurons (Dodd et al. 1988; D. Karagogeos et al., in preparation). The number of 5A5⁺ neurons obtained from explants grown in the presence of a floor plate was not significantly different from that of explants grown alone (Table 1), suggesting that the floor plate does not affect the survival or differentiation of neurons in dorsal spinal cord explants. These experiments do not exclude the possibility that the floor plate influences the relative proportions of commissural and association neurons without affecting their total num-ber, but an analysis of commissural axons in dorsal explants makes this unlikely (see below).

We next examined the effect of the floor plate on the rate of extension of TAG-1⁺ axons within dorsal explants. Dorsal explants were cultured for 20 h alone or in the presence of a floor plate placed opposite the ventral-most edge of the dorsal explant. At 20 h, TAG-1⁺ axons are still growing within the explant and are unfasciculated. Explants were fixed, serially sectioned and labelled with anti-TAG-1 antibodies. The pattern of TAG-1⁺ axons in these explants appeared similar in the presence or absence of floor plate (see Tessier-Lavigne et al. 1988). The lengths of all TAG-1⁺ neurite segments in each section were measured and then summed. Even though the number and lengths of individual neurite segments in any one section are dependent on the plane of section, their sum provides a measure of the combined lengths of TAG-1+ axons in the explant. As shown in Table 2, the combined lengths of TAG-1⁺ axons in dorsal explants was not significantly different in the presence or absence of a floor plate explant.

It seemed possible that the factor(s) was released by the floor plate only after a delay or diffused slowly and did not reach an effective concentration in the dorsal explant during the first 20 h of culture. We therefore examined the effect of floor plate-conditioned medium, added at the time of explantation, on the overall pattern and combined lengths of TAG-1⁺ axons in dorsal explants. As with floor plate explants, the pattern of TAG-1⁺ axons in dorsal explants cultured for 20 h with conditioned medium appeared similar to that of explants cultured alone (data not shown). The combined lengths of TAG-1⁺ axons in these explants was also unaffected by floor plate-conditioned medium (Table 2). These results rule out the possibility that a delay in secretion or diffusion of the factor underlies the lack of influence of the floor plate on the rate of axon growth.

Dorsal explants may provide a favourable environment for commissural axon growth, such that axons are already extending at a maximal rate which cannot be increased by the floor plate. We therefore examined whether floor plate-conditioned medium affects the extension of axons by commissural neurons dissociated from E11 dorsal spinal cord and grown on less favorable substrates. Because the floor plate affects the growth of commissural axons from neural epithelium into a threedimensional collagen matrix, we examined the growth of commissural axons on a two-dimensional collagen substrate, and also on poly-D-lysine (PDL). Both collagen and PDL alone were less favorable substrates for commissural axon outgrowth than PDL with laminin (not shown). Cells were cultured for 40 h in the presence or absence of conditioned medium then labelled with anti-TAG-1 antibodies. On a collagen substrate, cells clumped and failed to extend neurites, whether grown in control medium or with floor plate-conditioned medium (not shown). Neurite extension was observed on PDL, but the number of cells extending TAG-1⁺ neurites and the length of these neurites was not increased in the presence of floor plate-conditioned medium (Fig. 8).

Collectively, these observations indicate that the floor plate orients the growth of commissural axons by releasing a chemotropic factor without affecting the number of differentiated commissural neurons or the rate of commissural axon extension.

The movement of many cell types is directed by gradients of soluble molecules emanating from a point source. For example, cellular slime moulds orient their growth by chemotaxis up a gradient of cyclic AMP, and polymorphonuclear leukocytes are attracted to complement peptides (see Trinkaus, 1985). The directed growth of axons and the ability of axons to reach their targets via aberrant routes raises the possibility that growth cones can be guided by gradients of chemoattractant molecules diffusing from restricted cell populations within their targets.

A chemotropic factor is expected to have three major characteristics. First, it should be secreted by cells that form the target of growing axons. Second, it should diffuse through the environment of growing axons, setting up a concentration gradient that can reorient axons at a distance from its target. Third, it should cause reoriention by a direct action on the growth cones or axons of responsive neurons. Our experiments show that the floor plate secretes a factor that is capable of diffusing through the neural epithelium and which can effectively reorient the growth of commissural axons.

The stereotyped growth pattern of commissural axons within E11 dorsal explants in vitro provides a basis for analyzing their behaviour when confronted with the floor plate-derived chemoattractant. Two sets of observations provide evidence that the floor plate-derived factor can diffuse through different environments and orient developing commissural axons in vitro, (i) the orientation of commissural axons growing into the collagen matrix in the presence of the floor plate factor, and (ii) the orientation of commissural axons within the neural epithelium under the influence of a floor plate explant.

In dorsal explants grown alone, axons remain within the neural epithelium, but in the presence of the factor (presented either as a floor plate explant or as floor plate-conditioned medium), they grow into the collagen matrix (Tessier-Lavigne et al. 1988). The different patterns of emerging axons observed in the presence of a localized source or a uniform distribution of the factor provide evidence that growth cones orient up a gradient of the factor. In a uniform concentration of the factor, axons emerge from all sides of the explant. With a localized source either opposite the ventral edge or to one side of the explant, there is a marked orientation of axons towards the presumed high point of the factor (Fig. 4). Once axons emerge into the collagen matrix, however, they do not appear to reorient their growth further. This contrasts with the behavior of individual trigeminal sensory and corticospinal axons, which reorient as they grow through a similar matrix in response to chemoattractants (Lumsden and Davies, 1983; Heffner et al. 1990). The reason why commissural axons fail to reorient in the collagen gel is unclear, but could result from their growth in thick fascicles. Because axons emerge in fascicles, these experiments on their own do not exclude the possibility that only a subpopulation of commissural axons respond to the floor plate and that others follow these pioneers.

The reorientation of commissural axons within the neural epithelium, however, both provides information on the extent of diffusion of the floor plate-derived factor and demonstrates that all commissural axons are responsive to it. The factor can diffuse about 250 pm through both dorsal and ventral neural epithelium and is effective at reorienting virtually all commissural axons within this range. Since we always examined the effect of the floor plate after 40 h in culture, we do not know whether this distance represents the maximum range of action of the factor or simply the distance over which the factor can diffuse within this time. The point at which commissural axons break away from their initial trajectory in the dorsal spinal cord in vivo and grow to the floor plate occurs about 100 – 150 μm from the ventral midline, within the range of diffusion of the factor observed in vitro.

It is possible that commissural axons respond only to a gradient of the chemotropic factor, in which case a uniform concentration might be expected to have no effect on axonal growth. However, axons emerge from all edges of the explant in the presence of floor plate-conditioned medium. This raises the question whether the axons that emerge do so in response to a uniform concentration of the floor plate factor or to a gradient of the factor established at the border between the dorsal explant and the collagen gel. The factor may have growth-promoting activity even when it is not presented in a concentration gradient with the consequence that axons respond to a uniform concentration of the factor by growing into collagen. Alternatively, the factor may be degraded or otherwise inactivated by neural tissue, creating a local concentration gradient, which may be sufficient to direct axon outgrowth from the explant into the collagen gel. At present we cannot exclude that the outgrowth promoting and orienting effects of the floor plate are mediated by distinct molecules. However, the specificity of expression reported here indicates that, if there are distinct factors, they must both be restricted to the floor plate in E11 embryos.

Chemotropic activity can be detected in the floor plate at the time that commissural axons first begin to extend axons (∼E11). By E13, the ventral region of the spinal cord also exhibits low levels of activity. If the deviation in trajectory observed as commissural axons reach the dorsal aspect of the motor column indicates the point at which commissural axons first respond to the chemotropic factor, then it is to be expected that the ventral spinal cord would contain detectable activity. The activity found in ventral spinal cord may result from the synthesis of low levels of the factor by ventral spinal cord cells or from the diffusion of factor from the floor plate into the adjacent spinal cord. In any event, the chemotropic activity of the floor plate always greatly exceeds that of adjacent ventral tissue. Thus it is likely that at all relevant developmental stages there is a concentration gradient of the factor with its high point in or near the floor plate.

Our studies demonstrate the specificity and range of action of the floor plate-derived chemoattractant in vitro. The contribution of the chemoattractant to the guidance of commissural axons in vivo remains to be established. At one extreme, the presence of the factor might be an absolute requirement for the maintained ventral growth of commissural axons after they have reached the motor column. Alternatively, commissural axons may be capable of growing ventrally beyond the motor column even in the absence of the factor. The role of the factor might then be to provide a cue that ensures that commissural axons project directly through the motor column to the ventral midline, rather than fasciculate with motor axons extending out of the spinal cord.

We thank Eric Hubel and Ira Schieren for assistance with figures, Karen Liebert for preparing the manuscript, and Lorna Role for helpful discussions. We are also grateful to Ron McKay for raising the possibility that a cascade of fasciculation underlies the reorientation of axons observed in the dorsal spinal cord. Supported by grants to J.D. from the NIH (No. NS22993), the McKnight Endowment Fund for Neuroscience, the Esther A. and Joseph P. Klingenstein Foundation and the Irma T. Hirschl Foundation. M.P. was supported by a fellowship from the European Molecular Biology Organization. M. T.-L. is a Lucille P. Markey Scholar and was supported in part by a grant from the Lucille P. Markey Charitable Trust. T.J. is an Investigator of the Howard Hughes Medical Institute.