Survival in vitro of motoneurons identified or purified by novel antibodybased methods is selectively enhanced by muscle-derived factors

In terms of its function in the adult animal, the spinal motoneuron is one of the best-studied of all the cells of the nervous system. Furthermore, the timing and pattern of its early development in the embryo are well described in several species. It was using this system, for example, that Hamburger and colleagues (reviewed in Hamburger, 1977) first provided evidence for the existence of muscle-derived motoneuron growth factors necessary for motoneuron survival at early stages, and probably involved in regulation of many subsequent key events in embryonic and neonatal motoneuron development (Henderson, 1988; Oppenheim and Haverkamp, 1988). That these factors remain still to be identified (Barde, 1989) reflects, at least in part, the cellular complexity of the spinal cord and hence of cultures prepared from it. Until quite recently, it was not possible to be sure that a putative trophic substance actually enhanced survival or neurite outgrowth of motoneurons as opposed to other neurons in these cultures, and it remains impossible to ascertain whether such a factor affects only motoneurons.

Faced with these difficulties, investigators have adopted two complementary approaches: (i) identification of motoneurons in mixed cultures; (ii) enrichment or purification of motoneurons before culturing. For example, antibodies against choline acetyltransferase have been used to identify cholinergic neurons in cultures of spinal cord (Smith et al. 1986). However, such antibodies are difficult to use with early embryonic cells and are not completely specific, since other cholinergic neurons appear at early stages in the neural tube. The most widely used technique for identification has involved back-filling by retrograde transport of fluorescent or enzyme tracers injected in the limb bud (Eagleson and Bennett, 1983; Tanaka and Obata, 1983; Calof and Reichardt, 1984; O’Brien and Fischbach, 1986; Schaffner et al. 1987; Smith et a!. 1986; Tanaka, 1987; Martinou et al. 1989, 1990). Providing that appropriate precautions are taken to prevent diffusion of tracer from the site of injection, such methods provide unambiguous identification of motoneurons projecting to the limb. However, they present several disadvantages. First, only motoneurons that are sufficiently developed to transport the tracer can be labelled. Second, not all the motoneurons in a given segment of spinal cord are labelled: unlabelled populations can thus comprise both motoneuronal and non-motoneuronal cells. Third, the tracer itself may affect development in culture, or be lost quite rapidly. Such phenomena probably help to explain the wide variation in motoneuron abundance reported by investigators using apparently very similar culture systems (for a detailed comparison, see Henderson, 1988).

The major advantage of retrograde labelling using fluorescent tracers has been to allow the preparation of highly enriched motoneuronal populations by use of the fluorescence-activated cell sorter. Indeed, since the physical characteristics of the fluorescent cells have been well-defined, it has proved possible to sort without prior retrograde labelling (Martinou et al. 1989). This approach has provided new insights into factors directly affecting motoneuron development in vitro but is costly and labour-intensive. For this reason, other workers have routinely enriched cultures for motoneurons: (a) by dissection at early stages, at which motoneurons are relatively abundant (Masuko et al. 1979; Berg and Fischbach, 1978; Henderson et al. 1981, 1983; Longo et al. 1982); (b) by culture of the ventral regions of the neural tube (Smith et al. 1985); or (c) by use of density gradient fractions in which choline acetyltransferase activity is concentrated (Schnaar and Schaffner, 1981; Flanigan et al. 1985; Dohrmann et al. 1987; Martinou et al. 1989). It is clear, however, that such techniques cannot provide purification of motoneurons to the same extent as cell sorting.

In 1984, Tanaka and Obata described a monoclonal antibody named SCI which, in transverse sections of embryonic chicken spinal cord, labelled motoneurons, ventral epithelium (floor-plate cells) and sensory fibers in the dorsal funiculus. The labelling of motoneurons disappeared by 7 to 8 days of incubation and was highly specific to chick. We show now that this antibody can be used to provide unambiguous identification of motoneurons in vitro. Using appropriate embryos and culture conditions, motoneurons can represent up to 30% of the total population, and all of them, but not other spinal cord cells, need exogenous factors if they are to survive in culture. Furthermore, the SCI antibody can also be used to purify motoneurons by the recent technique of panning (Barres et al. 1988). Muscle extracts, but not any known growth factor tested, can keep such purified cells alive for days.

Spinal cords were dissected from 4.5-day Leghorn chicken embryos (Hamburger-Hamilton stage 24·25, incubation at 37.6°C), treated with trypsin and dissociated as described elsewhere (Henderson et al. 1984). They were cultured in Ham’s F12 medium supplemented with penicillin (100i.u. ml ^-1), streptomycin (100μgml ^-1), glutamine (2HIM), insulin (10ggml ^-1) and glucose (10mM) on 12-mm glass coverslips. These had been coated with polyornithine (1.5μgml ^-1; 30000 M_r, Sigma) for 30min at room temperature, and then incubated with laminin from EHS sarcoma at 3μgml ^-1 in F12 medium for at least 2 h in the CO2 incubator. Supplements, muscle extract etc. were added to each well (Costar 24-well plates; 600 μl per well) and these were then seeded with 30000 spinal cord cells. Cultures were incubated at 37.2°C in 5% CO₂/95% air and saturating humidity. Cultures of dorsal and ventral spinal cord and of other brain regions were prepared in the same way (Taguchi et al. 1986). Basic FGF (Boehringer) and TGF-beta (R & D Systems) were used according to suppliers’ instructions. Ciliary neurotrophic factor (CN IF; a generous gift of K. Wewetzer and K. Unsicker) was purified as described by Barbin et al. (1984), except that in one case the gradient centrifugation was replaced with chromatography on heparin-Sepharose. Both preparations of CNTF showed a single band after SDS-polyacrylamide gel electrophoresis.

At indicated times of culture, neurons were fixed with acetone at -20°C, either following removal of all culture medium or by adding an excess of acetone in the presence of culture medium. They were rinsed with PBS, incubated with supernatant from hybridoma SCI (dilution 1:10) containing 15 % fetal calf serum for 30 min at 37°C, washed again in PBS and fixed with 10 % formaldehyde for at least 10 min at room temperature. After five washes with PBS containing 2 % BSA (w/v), the cells were incubated with biotinylated goat antimouse IgG (dilution 1:100; Amersham) for 30min at 37°C, washed, and incubated with phycoerythrin-streptavidin complex (dilution 1:50; Amersham), or with fluorescein-strepta-vidin (dilution 1:100; GIBCO-BRL) following the supplier’s instructions. After stabilization of the phycoerythrin, samples were mounted in either the mountant provided for phycoerythrin, or with Citifluor mountant for fluorescein. They were observed using standard fluorescence optics; no fading was observed even over long periods. No staining was observed when antibody SCI was replaced with an irrelevant mouse IgG, or when the biotinylated antibody was replaced by biotinylated goat anti-rabbit antibody (Amersham) or omitted. Similar staining patterns were obtained when SCI binding was revealed using a TRITC-coupled goat anti-mouse antibody (Cappel, not shown).

After trypsinization and dissociation, SCI supernatant was added to the suspension of total spinal cord cells at a final dilution of 1:10 and incubated at 37°C for 30 min in the CO₂ incubator. Cells were then washed on a 20% metrizamide cushion in the presence of DNAase (0.1 mg ml ^-1) by centrifugation for 15 min at 1200 revs min ^- . The cells were then resuspended in 1 ml culture medium in the presence of biotinylated secondary antibody (diluted 100-fold) for 30 min at 37°C. After further washing, streptavidin-fluorescein (1:100) was added at 37°C for 30min. The cells were then washed and mounted after fixation, or observed directly.

Extracts of innervated neonatal chick muscle were prepared as described (Henderson et al. 1983). Briefly, muscles were homogenized in the presence of a cocktail of protease inhibitors and centrifuged. Aliquots of the high-speed supernatant were stored at —20°C. Unless otherwise indicated, neonatal muscle extract was used at a protein concentration of 6μgml ^-1. Laminin was prepared from homogenates of the mouse EHS sarcoma (generously provided by Marc Vigny).

Polystyrene Petri dishes (100 mm) were coated with 100 gg of secondary antibody (affinity-purified goat anti-mouse IgG, Cappel) in 10 ml Tris buffer pH 9.5 for 12 h at 4°C. Dishes were washed 3 times with PBS and SCI hybridoma supernatant diluted 1:5 in PBS was incubated in the dishes for 1 h at room temperature. Dishes were then washed twice with PBS and incubated with BSA (0.2 % w/v in F12 medium) for 20 min at room temperature in order to avoid non-specific binding.

Suspensions of spinal neurons were added to the panning dish (3–5 cord equivalents per plate in 12ml complete medium). After I h incubation at room temperature, each plate was washed 8 times with PBS with gentle swirling to remove nonadherent cells.

Elution of cells from the dishes was achieved by competition with an excess of SCI (2–3 ml of undiluted supernatant) and slow shaking (about 20min). Dishes were rinsed with 2×1 ml culture medium and eluted cells combined with the first fraction. Any cells remaining attached were discarded. SC1positive cells (approx. 20000 per cord) were washed by centrifugation at 1000 revs min ^-1 for 10 min, and seeded on polyornithine and laminin (PORN-laminin)-coated coverslips at a density of 25 000 per 16 mm culture well.

Spinal cords dissected from 4.5-day embryos were divided using micro-scissors along a line dorsolateral to the limits of the floor-plate region stained by SCI antibody. The portion containing the right-hand anterior horn still attached to the floor plate was discarded, only the left anterior horn being retained for subsequent panning.

Before panning, spinal cord cells were centrifuged for 15 min at 4000 revs min* ^r on a 6.8% metrizamide cushion in F12 medium containing O.lmgml ^-1 DNAase. Only those cells retained by the metrizamide cushion were used in the panning procedure.

We first confirmed the specificity of the SCI antibody using the phycoerythrin-streptavidin complex as fluorophore (Fig. 1). As described by Tanaka and Obata (1984), the only cells labelled within the 5-day embryonic chick spinal cord were the motoneurons and floor-plate cells. Ventral roots and dorsal root ganglia were also intensely stained; however, the latter are completely removed during dissection of the spinal cord.

Cells from different regions of the central nervous system of 4.5-day chicken embryos were cultured on glass coverslips coated with polyornithine and laminin (PORN-laminin), in serum-free F12 medium supplemented with neonatal muscle extract (see Materials and methods). After 2 days in culture, coverslips were fixed with acetone, incubated with SCI antibody and processed for indirect immunofluorescence. No labelling was observed in cultures of embryonic telencephalon or mesencephalon, or of cells from the dorsal half of the spinal cord (Table 1). In cultures of ventral or total spinal cord, however, two morphologically distinct cell types were labelled by SCI antibody: large, multipolar neurons (Fig. 2A) and flat cells with a fibroblastic morphology (Fig. 2B). The latter represented between 10% and 30% of the total SCl-labelled population in 2-day cultures of total spinal cord. When anterior horns were freed of floor plate by microdissection, this value fell to <5 %. The fibroblast-like cells must therefore be floor-plate cells from the ventral epithelium.

At later stages (Tanaka et al. 1989), SCI also labels some neurons in the column of Terni. In order to exclude the possibility that the SCl-positive neurons observed in vitro were not motoneurons but preganglionic neurons that had not yet migrated, separate cultures were prepared from thoracic, and from lumbar and brachial regions of the spinal cord (not shown). The abundance of SCl-labelled neurons after culture in the presence of muscle extract was at least as great in limbsegment cultures, as expected for motoneurons (Oppenheim et al. 1989). Residual survival of SCl-labelled neurons in basal medium was identical in cultures from different segments. It remains possible that a small number of labelled cells in cultures of total cord are preganglionic neurons, but these can be excluded by using only limb segment cultures (Oppenheim et al. 1989). We therefore consider that SCI antibody unambiguously identifies motoneurons in culture.

The effects of muscle-derived factors on identified motoneuron survival were apparent when total spinal neurons were cultured for 2 days on PORN-laminin in the presence and in the absence of neonatal muscle extract (Fig. 3). In its presence, labelled motoneurons were abundant and well-developed (Fig. 3B). In the absence of muscle-derived factors, whereas other spinal cells survived normally and were visible by their background fluorescence (Fig. 3A), only rare motoneurons were present, together with SCl-positive cell debris. Many floor-plate cells survived in the absence of muscle extract (not shown). In certain experiments, survival even of other cells was enhanced by muscle extract; we attributed this to small variations in basal conditions and not, since this requirement was inconsistently expressed, to a physiologically significant mechanism. Since the mounting medium used for preservation of phycoerythrin fluorescence did not allow cells to be visualised using Nomarski optics, all motoneuron counts presented here are expressed relative to total cells estimated by background labelling. Any cell showing non-neuronal morphology was excluded from counts of either fluorescent or total cells: floor-plate cells were thus not taken into account.

Cultures were fixed after different times in culture, in the presence or absence of muscle extract (Fig. 4). In the initial cell suspension, labelled cells were only weakly fluorescent, as a result of degradation of SCI antigen by trypsin during cell dissociation. Intense SCI labelling was initially seen in intracellular perinuclear vesicles, presumably corresponding to the Golgi apparatus (not shown), and was subsequently expressed at the cell surface on cell body and neurites. When muscle extract was present, full expression of SCI antigen was obtained after 2 to 3 days in vitro on a maximum of 25–30% of total neurons (Fig. 4). This percentage varied from culture to culture in the range 5–30% (5–50 % in cultures of ventral spinal cord), perhaps as a function of the exact developmental stage of the embryos, but more likely because of slight variations in the conditions of mechanical dissociation, to which motoneurons are extremely sensitive. From 3 days in culture onwards, the SCI labelling became rapidly more diffuse on individual cells (which were thus impossible to quantify precisely) and was barely detectable after 5 days in culture, although many neurons survived (not shown). This disappearance of labelling, corresponding to a theoretical age of 7.5 to 9.5 days in ovo, most probably reflects the developmentally regulated pattern of expression already reported in vivo (Tanaka and Obata, 1984).

In the absence of muscle extract, only low levels (<5 % of total cells) of SCl-positive cells were observed, at all stages of culture (Fig. 4). This difference could be explained by a survival-promoting effect of muscle extract, but equally well if muscle-derived factors were simply acting to increase expression of SCI antigen on cells whose survival is muscle-independent. In order to differentiate between these possibilities, parallel cultures were maintained for 1 day with or without muscle extract. At this point, all cultures were changed to medium containing muscle extract. High percentages of SCl-positive cells were once again only seen in those cultures maintained continuously in the presence of muscle extract (Fig. 5), demonstrating that deprivation of muscle-derived factors for 24 h was sufficient to initiate cell death in the motoneuron population.

The survival-promoting effect of muscle extract on identified motoneurons in mixed cultures on PORN-laminin was dose-dependent (Fig. 6). The concentration of protein required for half-maximal survival in the experiment shown was 4.7 μ gm l _-l”

Although high levels of SCI expression were only detected after 2 days in culture (Fig. 4), the use of biotinylated secondary antibody and streptavidin fluorescein allowed us to detect relatively faint SCI staining on the membrane of living freshly dissociated spinal cord cells. Immediately after treatment with trypsin, cells were dissociated and incubated with SCI in suspension. After washing by centrifugation, 30 % of cells present showed SCI immunofluorescence on at least part of their membrane. We used these SC1-binding sites to develop a method of motoneuron purification based on the panning procedure recently applied to retinal ganglion cells by Barres et al. (1988).

Polystyrene Petri dishes were coated with affinity-purified goat anti-mouse IgG antibodies at pH 9.5 (see Materials and methods), and then incubated with SCI hybridoma supernatant. After saturation of nonspecific protein-binding sites with BSA, suspensions of spinal cord cells were incubated on the panning plate. One hour later, non-adherent cells were removed by washing and then tightly bound cells (Fig. 7B) were eluted with an excess of SCI supernatant. The panned cells were plated on coverslips coated with PORN-laminin and cultured for 1 day or more in the presence of muscle extract, before being fixed and stained using SCI. All (>99%) of the cells were SCl-positive, and most of them rapidly developed quite complex neuronal morphology (Fig. 8A,B,C,E).

The adhesion of cells to the panning dish was the result of a specific interaction of SCI antibody with the SCI antigen. When SCI was omitted, or replaced by a hybridoma supernatant containing anti-myosin antibodies, no cells attached. Similarly negative results were obtained when neurons from the dorsal part of the spinal cord were incubated on dishes coated with SCI. The SCI staining seen on panned cells was not due to non-specific binding of the high concentration of antibody used to elute the cells; when SCl-negative cells from the dorsal spinal cord were taken through the elution procedure, they showed only background levels of fluorescence (Fig. 8D). Further evidence for the specificity of interaction was the observation that motoneurons often developed neurites within 3 h when cultured on the polystyrene panning dish, especially when SCI alone was used for coating (not shown). Their survival in these conditions was muscle-dependent but quantitatively variable, making it unreliable to perform bioassays using motoneurons directly attached to the dish.

Elution from the panning dish was therefore performed using SCI supernatant to compete for binding sites on the cells and on the dish. When serumcontaining hybridoma medium was used instead of SCI, approximately 2-fold less cells were eluted, all of which, however, were SCl-positive. Both trypsin and Ca ²⁺- and Mg ²⁺-free phosphate-buffered saline removed all cells from the dish, but viability of these cells in culture was extremely poor. Even using SCI supernatant, approximately 5 % of attached cells remained bound to the dish. When these were detached by a more vigorous stream of medium, they too were found to contain >95 % SCl-positive cells. It thus seems that the major determinant of specificity in the panning procedure is the binding step. However, elution with specific antibody provides the best compromise between efficiency of elution and preservation of motoneuron viability.

One surprising observation was that only about 2% of total spinal cord cells (i.e. approximately 7– 15% of SCl-positive cells) adhered to the panning plate. Many SCl-positive neurons remained in suspension and when cultured on PORN-laminin proved to be perfectly viable, and indistinguishable on morphological grounds from those that bound to the dish. These unbound motoneurons represented ca. 10% of total unbound neurons. This was unlikely to reflect a limited number of binding sites on the dish, as using different cell densities for panning (from 2×10 ⁵ to 10’ cells per dish), the percentage of cells bound was approximately constant. Furthermore, when non-adherent cells were incubated on a fresh panning dish, none attached. Since further increasing the cell density resulted in clumping of cells, routinely 3 spinal cord equivalents were panned per 100mm polystyrene dish, with a yield of approximately 10 ⁵ cells per dish. Similar results were obtained when panning was performed on limb segment dissociates (not shown).

In different experiments, populations of cells panned on SCI contained variable proportions (from 10% to 50%) of cells with non-neuronal morphology, corresponding to SCl-positive floor-plate cells. Although these did not hinder counting of motoneurons, confirmation of the direct neurotrophic effect of muscle extract in this preparation made it necessary to develop a system in which reproducibly only motoneurons were present. No cell-surface marker for chicken floor-plate cells is known; we therefore subdissected anterior horns to free them of floor-plate cells before panning. More than 98% of the panned cells showed neuronal morphology in culture. After 1.5 days, almost no cells survived in basal medium whereas 90% of those initially attached survived in the presence of muscle extract (Fig. 9). In an alternative approach, more suitable to large-scale preparations, we took advantage of the observation that on continuous 3–20 % metrizamide density gradients, SCl-positive cells of nonneuronal morphology were found near the bottom of the tube (not shown). Total spinal cord cells were therefore centrifuged over a 6.8 % metrizamide cushion (Dohrmann et al. 1987) and only the band of cells retained by the cushion was subjected to panning. This preparation (Fig. 6A) was more homogeneous in size than that resulting from direct panning of subdissected anterior horns (Fig. 7B) and had a similar requirement for muscle-derived trophic support (not shown). However, in our hands, subdissection gave more reproduc-ibly viable preparations.

The survival of the purified motoneurons, counted using phase-contrast optics, was enhanced by neonatal muscle extract in a dose-dependent manner (Fig. 6). This did not result from differential adhesion, as 3h after seeding in a typical experiment, 12.7 and 15.3 viable cells per field were counted in the presence and absence of muscle extract, respectively, representing a plating efficiency of 40–50 %. Approximately 3.2μ gml ^-1 of muscle protein were required for half-maximal survival after 2 days on PORN-laminin, a value indistinguishable from that required for half-maximal motoneuron survival in mixed cultures using the same extract (Fig. 6).

We tested the ability of some other known growth factors to reproduce this survival-promoting effect. Murine nerve growth factor (NGF), recombinant human basic fibroblast growth factor (bFGF), ciliary neurotrophic factor (CNTF) from chicken embryo eye extract and human transforming growth factor-beta (TGFβ₁) had no significant effect on motoneuron survival (Table 2; Fig. 10), although each sample used was shown in parallel to be biologically active in other cell culture systems, involving chicken neurons in all cases except for TGF/3 (Unsicker et al. 1987; Dreyer al. 1989; B. Buisson and A. Dupuy-Dangeac, unpublished). Chicken TGFβshows 100% sequence identity to the human factor (Jakowlew et al. 1988). For each growth factor, two samples from different preparations or suppliers were tested. Parallel experiments were conducted using the same growth factors in cultures of total spinal cord. Concentrations in the ranges indicated in Table 2 were tested for their ability to increase motoneuron survival above basal levels in the absence of muscle extract, or to enhance the survival-promoting activity of a suboptimal concentration of muscle extract. Counts of SCl-positive neurons in the fluorescence microscope at 1, 3 and 4 days of culture showed that none of these factors had motoneuron survivalpromoting activity, either in the presence of muscle extract or in its absence (not shown).

The results reported here demonstrate that the SCI monoclonal antibody can be used to identify spinal motoneurons in culture. Several arguments support the hypothesis that the in vitro labelling pattern faithfully reflects the in vivo specificity. First, these large multipolar neurons closely resembled both motoneurons silver-stained in situ and motoneurons purified by cell sorting and cultured in similar conditions. Second, SCI labelling in the central nervous system was limited to cultures of the ventral region of the spinal cord and SCl-positive cells were abundant in lumbar and brachial segments, from which preganglionic neurons are virtually absent. Third, SCI labelling grew weaker with time in culture, with a time course closely resembling that reported for motoneurons (but not other SCl-positive structures) in vivo (Tanaka and Obata, 1984). As expected for motoneurons, the survival of the SCl-positive population was critically dependent on muscle-derived factors in conditions in which other spinal cord cells seemed not to share this requirement.

One major advantage of SCI over other means of identifying motoneurons in culture, apart from convenience, is that it probably labels all motoneurons present rather than a selected subpopulation. The reasons to believe this are, first, that SCI in vivo labels apparently all cells in the anterior homs at all rostrocaudal levels tested, in contrast for instance with antibodies to calcitonin gene-related peptide, which label only approximately half of the motoneurons present (Fontaine et al. 1986; New and Mudge, 1986). Second, the maximal abundance (nearly 30%) of SCl-positive neurons as a. percentage of total spinal cord cells seems unlikely to be an underestimate. Indeed, this value is considerably greater than any other previously reported using retrograde labelling (see Henderson, 1988), the closest being the estimate (14%) of Schaffner et al. (1987) using 13-day mouse embryos. It is striking that the figure of 30 % putative motoneurons corresponds exactly to the fraction of spinal neurons that we previously reported to put out neurites in the presence of muscle-derived fractions when cultured on tissue-culture plastic (e.g. Henderson et al. 1984).

This approach, which labels all motoneurons, permits a distinction to be made between factors affecting motoneurons as well as other cells, and those affecting only motoneurons. In appropriate conditions, neurons from the ventral but not the dorsal part of the spinal cord, when cultured on tissue-culture plastic, will put out neurites in response to muscle-derived factors, whereas >80% of the neurons from either region develop neurites on a PORN-laminin substratum (C. E. Henderson, unpublished results). The use of SCI antibody has allowed us to show that both motoneurons and other spinal neurons are responsive to laminin, but that in these conditions motoneurons are the only major population in the spinal cord whose survival in vitro is muscle-dependent. A second advantage of this technique is that motoneurons are not subjected to centrifugation or chemical labelling before their in vitro behaviour is observed; identification is always a posteriori. Especially when phenomena such as shortterm survival are being studied, minor insults to motoneurons could create artefactual dependence on exogenous factors, leading to ‘false positives’ in the identification of motoneuron growth factors. This, and the absence of astrocytes in our system, may help to explain why we found no evidence for the neurotrophic effect of TGF/31 described by Martinou et al. (1990) or that of bFGF described by Unsicker et al. (1987).

Potential disadvantages of the use of SCI antibody as a tool for identifying motoneurons in mixed cultures are that labelling is lost in a developmentally regulated fashion after 4 – 5 days of culture, making motoneurons unidentifiable at later stages, and that motoneurons are not separated from other cell types in the culture, which might therefore themselves produce neurotrophic activity under the influence of muscle extract. In order to overcome these problems, we developed a panning method for isolation of motoneurons. As presented here, this method provides a reliable bench-top method for production of reasonable quantities (5xl(r) of pure motoneurons. The low percentage of total spinal cord cells that adhered to the panning dish was unexpected. Our results suggest that only a subpopulation of SCl-positive cells in spinal cord dissociates expressed the antigen in an appropriate manner for interaction with the dish to occur. It is possible that the non-adherent SCl-positive cells were too far degraded by the trypsin used in dissociation; we have not yet found an enzymatic treatment that will dissociate the spinal cord without degrading the SCI antigen (other enzymes tried were collagenase, elastase and dispase at standard concentrations). Another possibility is that there are steric constraints for the interaction of immobilized SCI with the cell surface, and that, at low overall levels of antigen, these conditions are only met in a statistically infrequent subpopulation. Whatever the explanation, the panned motoneurons had properties closely resembling those of the total motoneuron population, both in terms of their morphology and in the concentration dependence of their survival response to neonatal muscle extract (Fig. 7). It is probable therefore that, even in mixed cultures, muscle effects on motoneuron survival are direct, not mediated by other cells of the spinal cord.

Motoneurons separated from other spinal cord cells by panning grew as well, and often better, than those in mixed cultures. In contrast, others using density gradients or cell sorting have frequently reported that it was necessary to use an astrocyte feeder layer for survival of enriched fractions (Schaffner et al. 1987; Martinou et al. 1989; H. El M’hamdi and C. E. Henderson, unpublished data) or that development of purified motoneurons was less complete than in the presence of interneurons (O’Brien and Fischbach, 1986). Immature motoneurons are probably not selected by these methods, owing to their smaller size and potentially lower capacity to transport tracers from the limb. SCI antigen, on the other hand, is present from 3 days in ovo (Tanaka and Obata, 1984). It is possible therefore that in addition to being more gentle than other methods, panning allows purification of immature motoneurons, better adapted to survival in vitro, and more appropriate for studies of factors involved in early neuromuscular contact formation. In support of this hypothesis, the cells bound to the panning plate showed considerable size heterogeneity, even when floor-plate cells had been removed by prior dissection (Fig. 7B), and the more homogeneous cultures prepared by centrifugation on metrizamide prior to panning were less reproducibly viable on PORN-laminin alone.

Two purified polypeptides have been reported to prevent or reduce motoneuron death when administered in vivo, although neither of them has yet been demonstrated to affect survival of isolated motoneurons. Ciliary neurotrophic factor (CNTF) protected motoneurons in newborn rat facial nuclei against axotomy-induced degeneration, suggesting that the early postnatal period of sensitivity to axotomy can be correlated with the low levels of CNTF expressed in the sciatic nerve at this stage compared to the adult (Sendtner et al. 1990). Our finding that otherwise biologically active preparations of CNTF from chick eye do not affect survival of identified embryonic motoneurons (Figs 9,10) suggests either that the in vivo effect of CNTF is mediated by a cell type or a co-factor not present in our cultures, or that sensitivity to CNTF is only acquired at later embryonic stages. However, it is also possible that the ‘CNTF’ molecules purified from sciatic nerve and eye may have different biological properties. It was recently shown that treatment of chick embryos with purified cholinergic development factor (CDF), a 22xlO ³M_r polypeptide (pl4.8) of unknown sequence, rescued about 30% of the spinal motoneurons that would normally have been lost during the period of naturally occurring cell death (McManaman et al. 1990). However, CDF did not affect total cell survival in cultures of motoneurons that were enriched but not purified, raising the possibility that its action might be indirect. Motoneuron survival quantified using SCI should provide an appropriate parameter by which to evaluate the neurotrophic action of this and other known molecules and on which to base a strategy for identification of the active molecule in muscle extracts.

This work was made possible by financial assistance from the Association Française contre les Myopathies (A.F.M.) and the Institut pour la Recherche sur la Moëlle Epinière (I.R.M.E.). F.K.X. was the recipient of a grant from the French Foreign Ministry. We are particularly grateful to Jean-Pierre Changeux, in whose laboratory the first part of this work was carried out, for constant support and disinterested advice, and to Claudine David for help with cultures and immunofluorescence labelling, Antoine Triller and Ned Lamb for advice on fluorescence microscopy, Nicole Basset for use of the cryostat, and Yves-Alain Barde for helpful suggestions. Konstantin Wewetzer and Klaus Unsicker (Marburg, FRG) generously provided CN I F, NGF was kindly given by Philippe Brachet (Angers) and anti-myosin antibodies were donated by Jocelyne Léger. We are indebted to Clément Mettling, Alain Prochiantz, Andrew Spence, Danièle Thierry-Mieg and Jean Valmier for critical reading of the manuscript.