Developmental expression of the axonal glycoprotein TAG-1: differential regulation by central and peripheral neurons in vitro

During the development of the nervous system, axons project to their targets along complex but stereotyped pathways. The growth and pathfinding of axons is thought to depend on interactions between molecules located on the surface and in the local environment of the growth cone (Dodd and Jessell, 1988; Harrelson and Goodman, 1988). Several axonal glycoproteins have been implicated in growth cone extension and many of these fall into three major structural classes: proteins with immunoglobulin and fibronectin type III domains (Jessell, 1988; Rutishauser, 1989), Ca²⁺-dependent adhesion proteins termed cadherins (Takeichi, 1990) and heterodimeric proteins termed integrins (Reichardt et al. 1990). Regulation of the structure and expression of these glycoproteins during development may affect the ability of neurons to extend axons. For example, the neural cell adhesion molecule N-CAM exhibits a striking change in polysialic acid (PSA) content during development (Edelman, 1986; Rutishauser, 1989) and removal of PSA from N-CAM on the surface of retinal ganglion neurons reduces axon outgrowth on N-CAM substrates (Doherty et al. 1990). During development retinal ganglion neurons also lose the ability to extend axons on a laminin substrate (Cohen et al. 1987). This appears to result from a developmental, loss of laminin receptors (Cohen et al. 1989), probably of the integrin family (Reichardt et al. 1991).

The axonal glycoprotein TAG-1 has been shown to exhibit both spatial and temporal restrictions in its expression during early neural development (Yamamoto et al. 1986; Dodd et al. 1988; Yamamoto et al. 1990). Within the spinal cord, TAG-1 is expressed transiently during the initial outgrowth of motor, commissural and dorsal root ganglion (DRG) axons (Dodd et al. 1988). The amino acid sequence of TAG-1 deduced from cDNA clones shows it to be a member of the immunoglobulin superfamily (Furley et al. 1990) with structural features in common with other vertebrate axonal glycoproteins, in particular N-CAM, LI and contactin/F3/Fll (Cunningham et al. 1987; Moos et al. 1988; Ranscht, 1988; Gennarini et al. 1989,a;b; Brummendorf et al. 1989). Moreover, neurons grown in vitro extend long processes on a substrate of TAG-1 (Furley et al. 1990), suggesting that the protein is involved in the growth of axons in vivo.

The transient expression of TAG-1 in vivo distinguishes it from that of other neuronal immunoglobulin family members, in particular N-CAM, LI and contactin, which can be detected on the surface of adult axons (Edelman, 1986; Rathjen and Rutishauser, 1984; Ranscht, 1988). To provide information on the mechanisms that control the expression of TAG-1 on developing axons, we have examined the regulation of this glycoprotein in vitro on embryonic CNS neurons obtained from the rat spinal cord and on peripheral neurons of the dorsal root ganglion (DRG). We show here that spinal cord and DRG neurons express both a glycosyl phosphatidylinositol-anchored surface form and a released form of TAG-1 and that the surface form of TAG-1 is regulated differentially by central and peripheral neurons in a cell-type-specific manner.

Spinal cords were dissected from embryonic day (E) 13 rats and placed into L15 medium at 4°C. The dorsal and ventral regions of the spinal cord were dissected and incubated separately with 0.05% trypsin (Gibco) for 20 min in Ca²⁺/ Mg²⁺-free modified essential medium (S-MEM) (Gibco) supplemented with 8 mg ml⁻¹ glucose. The tissue was then washed with S-MEM and triturated to give a single cell suspension Spinal cord cells were usually plated into 35 mm tissue culture dishes containing a monolayer of neonatal cortical astrocytes and grown in Ham’s F12 medium (Gibco) supplemented with N3 additives (F12-N3) (Romijn et al. 1982) at a density of 1–1 2×10⁴cellscm⁻² in a 5% CO₂ humidified incubator at 37°C. In some experiments, spinal cord neurons were plated directly onto substrates of poly-D-lysine (20μgμl⁻¹)/laminin (20μgml⁻¹) or hydrated rat tail collagen (∼100μgml⁻¹).

Dorsal root ganglia (DRG) were dissected from embryonic or neonatal rats and placed into L15 medium, trypsinized and treated as described above. Neurons were plated into 35 mm dishes (Nunc) treated with poly-D-lysine/laminin (20μgml^-1 each; Sigma, Collaborative Research) in F12-N3 containing 5 % horse serum (Gibco) and 100 ng NGF. Non-neuronal cells were eliminated from the cultures by treatment soon after plating with 10⁻⁵M cytosine arabinoside for 24–48 h. DRG from 21- to 90-day-old rats were isolated, dissociated and prepared for cell culture as described (Lindsay, 1988) Briefly, DRG were isolated and placed in L15 medium and incubated for 2×1.5h periods with collagenase (0 125%; Boehringer Mannheim) in F12 medium containing 5 % horse serum at 37 °C. DRG were then washed in F12 with 5 % horse serum and triturated to a give single cell suspension Culture conditions were as described for DRG obtained from younger rats.

Cells dissociated from embryonic spinal cord were plated onto a cortical astrocyte monolayer, grown overnight and incubated for 10 h with [³H]thymidine (NEN) (1 mCi ml⁻¹) (Kriegstein and Dichter, 1984) [³H]thymidine was added in RPMI-1640 medium with 5 % horse serum; 2 mM 1-glutamine; 1001.u. ml¹ penicillin/streptomycm and MEM vitamins (Gibco). The cultures were maintained for 14 h to 5 days after labelling. Cultures were fixed after various times by exposure to 1 % glutaraldehyde in 0.12M phosphate buffer (PB) for 20 – 45min, washed with 20mM PB, 0.9% saline (PBS), followed by distilled water and stored desiccated overnight. The dishes were then coated with NTB2 emulsion diluted IT with water and exposed for 40 – 48 h at 4°C Autoradiographs were processed in Ilford ID11 developer, washed three times in water, covershpped and viewed under phase optics with a Zeiss Axioplan microscope.

TAG-1 was detected with monoclonal antibodies (mAbs) 4D7 (IgM, Yamamoto et al 1986), and 1C12 (IgGl; Dodd et al. 1988) or with rabbit antibodies to immunoaffinity-purified TAG-1 (Dodd et al. 1988) LI was detected by mAbs 69A1 (IgG; Pigott and Davies, 1987) or ASCS4 (IgG; Sweadner, 1983) and by polyclonal antisera raised against mouse LI (Rathjen and Schachner, 1984). The polysialylated form of NCAM was identified by mAb 5A5 (IgM; Dodd et al. 1988). Fluoresceinated isotype-specific second antibodies were purchased from TAGO (goat anti-mouse IgM) and Boehringer Mannheim Biochermcals (goat anti-mouse IgG) and used at dilution of 1:100 to 1:200.

For immunofluorescence labelling (Dodd and Jessell, 1985), cultures were washed once at 22 °C with L15 supplemented with 8 mg ml⁻¹ glucose and 0.1% BSA (Gibco) and then incubated with primary antibodies (diluted in L15 containing 8mgml⁻¹ glucose and 0.1% BSA) for 30 min at 22 ¹¹C. Cultures were then washed twice m L15 – 1 % normal goat serum (NGS) and incubated with secondary FITC-conjugated isotype-specific antibodies diluted in L15-l% NGS for 30 min at 22 °C. Cultures were washed twice and fixed in 4% paraformaldehyde in 0.12 M phosphate buffer (PB) (pH7.4) for 15 –20mm, rinsed in 0.12M PB and covershpped in 0.04% paraphenylenediamine (Sigma) in 0 2M sodium carbonate (pH9 0): glycerol (11) Cultures were viewed on a Zeiss Axioplan microscope under epifluorescence optics. 100 – 300 neurons were analyzed for each data point.

For labelling of tissue sections, the spinal cord and DRG were dissected from rats fixed by immersion or perfusion with 4 % paraformaldehyde, washed in PBS and immersed in 30 % sucrose in 0.1 M PB overnight at 4°C. Tissue was mounted in OCT compound (Miles) and 10 – 15 μm cryostat sections were collected onto gelatin-subbed slides Sections were washed in PBS, incubated with primary antibodies overnight at 4°C and then washed in PBS/1% heat-inactivated NGS. Processing with peroxidase-conjugated second antibodies (Boehringer Mannheim) was performed as described (Dodd et al. 1988)

Embryonic spinal cord and DRG neurons were plated at a density of 1 – 1.2 × 10⁴cells cm⁻² in 35 mm or 60 mm diameter culture dishes. NP-40 cell lysates were prepared from embryonic spinal cord and DRG neurons maintained in vitro for 1 to 5 days. Cell lysates derived from 60 mm diameter dishes were centrifuged at 3000 g and the supernatant fraction separated by SDS-PAGE electrophoresis under reducing conditions, transferred onto nitrocellulose and reacted with rabbit antisera to TAG-1 diluted 1:50 000 (Dodd et al. 1988). Conditioned medium was centrifuged at 3000 g and the supernatant concentrated on a Centricon 10 (Amicon) and mixed with sample buffer prior to analysis by SDS-PAGE.

Western blotting experiments were also performed with rabbit anti-Ll antibodies and with two monoclonal antibodies against N-CAM, mAbs 5B8 and 5A5. mAb 5B8 recognizes an intracellular epitope on the 140 and 180 ×10³ iso forms of N-CAM mAb 5A5 recognizes an α·-2,8 linked polysiahc acid epitope on the polysialyl side chain of N-CAM (Dodd et al. 1988). In western blots of extracts of cultured spinal cord cells, mAb 5A5 reacts with a band in excess of 200 ×10³ indicating a high degree of N-CAM sialylation. In extracts of E13-15 spinal cord treated with endoneuramindase, the 140 × 10³M_r and 180 × 10³M_r isoforms of N-CAM are detected at high levels with mAb 5B8. Moreover, treatment of hving spinal cord neurons with PI-PLC did not significantly reduce the intensity of mAb 5A5 labelling, indicating that there is little or no expression of the 120×10³ GPl-linked isoform of N-CAM at this developmental stage.

Cultured spinal cord or DRG cells were incubated for 30 min at 37°C in 50mM Tris-HCl (pH 7.2) buffer with or without 0.7 unit ml⁻¹ phosphatidylinositol-specific phospholipase C (isolated from B. thuringiensis) (Low and Saltiel, 1988) Cells were processed for immunofluorescence or western blotting as described above.

To measure incorporation of ethanolamine into TAG-1, DRG cultures were incubated in [³H]ethanolamine (03mCi, 30 riiM) m F12-N3 medium. The medium was removed from the cultures, centrifuged and stored at 4°C. Cultures were then treated with 0.7 U phosphatidylinositol-specific phospholipase C and the released proteins and cell pellet collected separately. The cell pellet was solubilized in Triton X – 100. Supernatants and cell pellets were then concentrated and mcubated with rabbit anti-TAG-1 antiserum (1·1000) in a final volume of 20 μl. Antibody-antigen complexes were separated by addition of protein A-affigel beads (BioRad) Pelleted beads were washed extensively then boiled in SDS containing buffer and solubilized proteins run on a 10 % SDS-PAGE gel under reducing conditions (Dodd et al. 1988) Gels were dried and autoradiographs exposed for 4-16 weeks. As control, rabbit anti-Thy-1 antibodies or normal rabbit serum was used.

Retrograde labelling of motor neurons with Dil (1,1’-dioctadecyl-3,3.3’,3’-tetramethylindocarbocyanine perchlorate; 2.5 mg ml⁻¹ in DMSO; Molecular Probes) was performed using a modification of the method of Homg and Hume (1986). Dil (5nl) was injected in the hind limbs of E15 rat embryos with a glass micropipette. After a 16–20 h in vuro incubation in F12-N3 media at 37°C and aeration with 95 % O₂-5 % CO₂, embryos were processed either for histology or for tissue culture. For histology, embryos were fixed in 4% paraformaldehyde in 0.1 M PB for 2–3 h and embedded in 2.8% agarose (Sigma, type 1, low EEO). 200μm transverse sections of dye-injected embryos and were cut with a Vibratome. For tissue culture, the ventral spinal cord from injected embryos was isolated, dissociated with trypsin to give a single cell suspension and plated on astrocyte membranes obtained by freeze-thaw of a confluent monolayer of cortical astrocytes, because identification of labelled neurons was difficult with living astrocytes, due to the transfer of Dil from neurons to astrocytes.

Photographs were taken on a Zeiss Axioplan microscope equipped with a 35 mm camera using Ilford HP5 film. A 40×Plan-Neufluor objective was used and exposure times ranged from 10 to 60 s.

TAG-1 was characterized originally as an axonal glycoprotein that appears transiently on the surface of a subset of central and peripheral neurons (Yamamoto et al. 1986; Dodd et al. 1988; Furley et al. 1990; Yamamoto et al. 1990). The hydrophobicity of the carboxy terminal amino acid sequence of TAG-1 deduced from cDNA clones suggests that the protein is attached to the neuronal surface membrane by a glycosyl phosphatidyl-inositol (GPI) anchor (Furley et al. 1990). In support of this, biochemical studies have shown that TAG-1 is released from brain membranes and from the surface of neuroblastoma cells transfected with TAG-1 cDNA by bacterial phosphatidylinositol-specific phospholipase-C (PI-PLC) (Furley et al. 1990). To determine whether TAG-1 is attached to the surface of primary neurons via a GPI linkage, E13 dorsal spinal cord or E15 DRG neurons were grown in vitro for 1-3 days and then incubated for 30 min with control buffer or with PI-PLC. Expression of TAG-1 was analyzed by immuno-fluorescence histochemistry and by western blotting.

TAG-1 was detected on the surface of spinal cord and DRG neurons in cultures treated with control buffer (Figs 1A,B; 2A,B). After treatment with PI-PLC, TAG-1 disappeared from the surface of spinal cord (Fig. 1C,D) and DRG neurons (Fig. 2C,D), whereas surface expression of the related immunoglobulin family members N-CAM and LI was not affected by identical PI-PLC treatment (Figs 1E-H; 2E-H). LI and the major neuronal isoforms of N-CAM are transmem-brane proteins (Cunningham et al. 1987; Moos et al. 1988), thus PI-PLC treatment appeared selectively to remove GPI-linked proteins. When assayed by immu-noblotting, virtually all the TAG-1 in untreated spinal cord or DRG cultures was associated with the cell pellet, although a very small amount of TAG-1 was detected in the medium after a 30 min incubation (Fig. 3A, lanes 1,3; Fig. 3B, lane 6). Treatment of spinal cord and DRG cultures with PI-PLC for 30 min resulted in the appearance of large amounts of TAG-1 in the medium (Fig. 3A, lane 4; Fig. 3B, lane 8), although low amounts of TAG-1 remained associated with neurons (Fig. 3A, lane 2; Fig. 3B, lane 7). The residual TAG-1 may represent an intracellular store of the protein. These results indicate that all of the TAG-1 on the surface of dorsal spinal cord and DRG neurons is attached via a GPI linkage.

To examine whether neurons release TAG-1 in the absence of PI-PLC treatment, E13 dorsal spinal cord cells were grown on a monolayer of cortical astrocytes for 12–24 h and then incubated in serum-free medium for an additional 12 h. A large amount of TAG-1 was detected in medium conditioned by spinal cord neurons (Fig. 3A, lane 5). Similarly, E15 DRG neurons grown on a laminin substrate released large amounts of TAG-1 into the culture medium in the absence of PI-PLC treatment (Fig. 3B, lane 9). In control experiments, the transmembrane glycoproteins LI and N-CAM were not detectable in culture medium exposed to DRG neurons in the absence of PI-PLC treatment (not shown). Thus, TAG-1 is expressed on the surface of dorsal spinal cord and DRG neurons and is also released into the medium. These results extend other studies showing that neuroblastoma cells transfected with TAG-1 cDNA release TAG-1 into the culture medium in the absence of PI-PLC treatment (Furley et al. 1990).

To examine the relationship between surface and released forms of TAG-1, E15 DRG neurons in culture were incubated with [³H]ethanolamine. GPI-linked proteins incorporate [³H]ethanolamine into the lipid anchor and cleavage of the anchor by PI-PLC results in the release of proteins that retain [³H]ethanolamine group(s), (Ferguson and Williams, 1988; Doering et al. 1990). In the present experiments, incubation times were adjusted such that approximately equal amounts of immunoreactive TAG-1 were obtained by cleavage of surface TAG-1 with PI-PLC and by collection of medium exposed to DRG neurons in the absence of PI-PLC (Fig. 3C, lanes 10,11). TAG-1 released from DRG neurons by PI-PLC treatment had incorporated large amounts of [³H]ethanolamine as assessed by the intensity of the autoradiographic band (Fig. 3D, lane 12). However, the·intensity of the [³H]ethanolamine-labelled band was much lower in the form of TAG-1 that was released by DRG neurons in the absence of PI-PLC treatment (Fig. 3D, lane 13). Thus, it appears that a large fraction of released TAG-1 derived from a precursor form that had not incorporated ethanolamine or from a processed form from which the ethanolamine-containing anchor had been removed.

Immunocytochemical studies have shown that TAG-1 is expressed transiently on spinal cord neurons in situ, appearing on commissural neurons in the dorsal spinal cord over the period Ell to E15 and on motor neurons in the ventral spinal cord over the period E10.5 to E12 (Dodd et al. 1988). We therefore examined the time course of expression of TAG-1 on the surface of spinal cord neurons in vitro.

Approximately 45 % of E13 cultured dorsal spinal cord neurons expressed TAG-1 by 24 h after plating (Figs 4A,B; 5A). The percentage of TAG-l-labelled neurons decreased progressively with time in vitro and by 4–5 days after plating virtually no neurons expressed TAG-1 on their surface (Fig. 4C,D; 5A). TAG-1 immunoreactivity was not detectable in permeabilized dorsal spinal cord neurons that had been maintained in vitro for 4 days (not shown) suggesting that the inability to detect TAG-1 on the surface of dorsal spinal cord neurons in vitro did not result from retention of the protein intracellularly. Similar results were obtained when dorsal spinal cord neurons isolated from Ell embryos (the time at which the first commissural neurons begin to extend axons) were maintained in dissociated cell culture. About 60% of Ell neurons expressed TAG-1 on their surface after 24 h in culture. By 4 days after plating, the number of TAG-l-labelled neurons had decreased to 15 % and by 6 days in vitro no TAG-l-labelled neurons were detected (not shown). We also examined the time course of expression of TAG-1 on neurons isolated from ventral spinal cord of Ell rat embryos. By 24 h after plating, between 55 and 60% of Ell ventral spinal cord neurons expressed TAG-1 (Fig. 5B). There was a marked decrease in the number of TAG-1 labelled neurons over the next three days and by 5 days, no TAG-1 immunoreactive neurons were detectable (Fig. 5B). Thus, in vitro TAG-1 disappears from spinal neurons with a time course that reflects, approximately, the loss of the protein from the same neurons in situ.

The structurally related axonal glycoprotein LI has an expression pattern in embryonic spinal cord that is distinct from that of TAG-1. LI appears on neurons later than TAG-1 and is restricted primarily to fasciculated fiber tracts (Stallcup et al. 1984; Dodd et al. 1988; Holley and Schachner, cited in Schachner et al. 1990). The temporal expression of LI on cultured spinal cord neurons differed markedly from that of TAG-1. A low percentage of Ll-labelled neurons (10-30 % depending on whether monoclonal or polyclonal antibodies were used) was detected 24 h after plating (Figs 4E,F; 5A), with a progressive increase in the percentage of Ll-labelled dorsal neurons detected over the first few days in culture (Figs 4G,H; 5A). As described previously (Pigott and Davies, 1987), LI was expressed primarily on the neurites of dorsal spinal cord neurons (Fig. 4H), which made it difficult to determine the exact percentage of Ll-labelled neurons after the third day in vitro, by which time a complex network of neurites was visible. The number of Ll-labelled neurons in cultures of Ell ventral spinal cord also increased progressively with time in vitro, from about 30% on day 1 to at least 70% by day 5 (Fig. 5B). A third immunoglobulin family member, N-CAM, is expressed soon after neural differentiation and persists on axons throughout embryonic and postnatal development (Edelman, 1986; Dodd et al. 1988). Consistent with this, N-CAM was detected on virtually all spinal cord neurons within 24 h after plating and expression persisted for up to 8 days in vitro, the longest period examined (Figs 4I-L; 5A).

We also used western blotting to determine TAG-1 levels in E13 dorsal spinal cord cultures that had been maintained in vitro for 1 or 5 days. A prominent 135×10³M_r TAG-1 band was detected in neurons maintained in vitro for 1 day (Fig. 6, lane 1), whereas after 5 days the amount of TAG-1 was markedly decreased (Fig. 6, lane 2). In contrast, similar amounts of TAG-1 were detected in medium exposed for 12 h to E13 dorsal spinal cord neurons grown in vitro for 1 and 5 days (Fig. 6; lanes 3,4). Thus, TAG-1 appears to be released by spinal neurons at approximately constant levels after the surface form of TAG-1 has disappeared.

The progressive decrease in the number of TAG-l-labelled cells does not appear to result from neuronal death. The total number of neurons that survived in vitro, identified by their morphology and by expression of the neuronal antigens 3A10 and N-CAM (Dodd et al. 1988; Furley et al. 1990), did not change significantly over the first 4 days in vitro (Fig. 7A). By day 8 in vitro, the number of surviving neurons had fallen to about 75% of the number originally plated (Fig. 7A); however, this decrease occurred after the disappearance of TAG-l-labelled neurons. Dorsal spinal cord neurons have been shown to differentiate over an extended period, up to about E14-E15 in vivo (Altman and Bayer, 1984); thus it is possible that a constant number of neurons is maintained, despite neuronal death, by the differentiation of precursor cells. To determine whether cultures contained dividing precursors capable of differentiating into neurons in vitro, we dissociated the dorsal spinal cord of E13 rat embryos, plated them on astrocyte monolayers for 6 – 12 h and then incubated the cultures in [³H]thymidine for 10 h.

The cultures were grown for a further 14 h – 5 days and then processed for autoradiography. Silver grains were detected in the nuclei of many flat, non-neuronal cells; however, none of 1414 cells that were identified morphologically as neurons exhibited labelled nuclei (Fig. 7B). These observations indicate that proliferating neuronal precursor cells do not generate significant numbers of neurons in vitro. Thus, the loss of TAG-l-labelled neurons in vitro appears to result primarily from the disappearance of the glycoprotein from the cell surface rather than from cell death. These results do not, however, resolve whether the released form of TAG-1 derives from the same neurons that transiently express the surface form, although this seems likely.

The disappearance of the surface form of TAG-1 from spinal neurons could be controlled by environmental factors or by events that occur autonomously in neurons that synthesize TAG-1. To distinguish between these two possibilities, we first tested whether expression of TAG-1 was dependent on the substratum upon which the neurons were grown. The transient expression of TAG-1 by E13 dorsal spinal cord neurons in vitro was not a consequence of the use of an astrocyte monolayer as substratum, since the number of TAG-1 neurons decreased over a similar time course when E13 dorsal spinal cord cells were plated on substrata of poly-L-lysine/laminin or collagen (not shown). The loss of TAG-1 expression also occurred over a similar time course when E13 dorsal spinal cord neurons were grown on astrocyte monolayers in serum-free medium (not shown). Thus, transient expression of TAG-1 on the surface of dorsal spinal cord neurons was not critically dependent on substratum or serum factors.

To determine whether the loss of the surface form of TAG-1 with time in vitro is controlled by signals derived from other spinal cord cells, we compared the time course of disappearance of TAG-1 from the surface of E13 dorsal spinal cord neurons grown on cortical astrocytes alone or on astrocytes on to which E13 dorsal spinal cord cells had been plated 4 days previously. TAG-1 had completely disappeared from the surface of the older neurons at the time of addition of the new set of E13 spinal cord ceils (Fig. 8A; closed circles). No consistent difference was observed in the rate at which TAG-1 disappeared from the surface of E13 dorsal spinal cord neurons plated on astrocytes alone (Fig. 8A; open diamonds) when compared with sibling cultures in which an aliquot of the same cell preparation had been plated on astrocytes in the presence of E13 dorsal spinal cord cells aged in vitro (Fig. 8B). The time course of disappearance of the surface form of TAG-1 therefore appears to be dependent on the age of the neurons that express TAG-1 and not on the age of the surrounding neuronal or non-neuronal cells. Although these results suggest that the cell surface form of TAG-1 is regulated by events that occur autonomously in this subset of neurons, we cannot exclude that other changes in the environment can affect TAG-1 expression.

TAG-1 is not detectable within DRG in situ after P7 (Dodd et al. 1988). To examine whether TAG-1 is expressed transiently on the surface of embryonic or postnatal DRG neurons in vitro, we isolated DRG neurons from E15, E17, P3 and P5 rats and grew them on a substrate of poly-D-lysine/lammin for periods of up to 21 days. TAG-1 was expressed on the surface of the majority (80-90%) of DRG neurons within 24h after plating (Fig. 9A); however, in contrast to results obtained with spinal cord neurons, expression on DRG neurons persisted for the entire period in vitro. This result was obtained independent of the age at which neurons were isolated. Surface expression of TAG-1 also persisted when DRG neurons were grown in defined medium, on cortical astrocyte monolayers or on astrocytes with E13 dorsal spinal cord cells aged in vitro as described above (not shown). LI and N-CAM were expressed by the majority of DRG neurons at all times in vitro (not shown).

In these experiments, cultures were treated with cytosine arabinoside (10”‘M) for 36–48 h to suppress the proliferation of ganglionic non-neuronal cells. To examine whether the absence of non-neuronal cells was responsible for the persistent expression of TAG-1, we delayed addition of cytoside arabinoside until day 4 in vitro, thus permitting extensive proliferation of non-neuronal cells. Under these conditions, TAG-1 expression still persisted for 21 days in vitro (not shown). To ensure that neurons remained in direct contact with ganglionic non-neuronal cells, E15 and P5 DRG were also grown as explant cultures. Again, TAG-1 expression on DRG neurites persisted for at least 28 days in vitro (Fig. 9B). These results suggest that the persistent expression of TAG-1 on the surface of DRG neurons in vitro is not affected by the presence of central and peripheral glial cells.

In the embryonic spinal cord, TAG-1 does not reappear on the surface of neurons after the early phase of expression has terminated (Dodd et al. 1988). We therefore examined whether neurons isolated from the spinal cord at times when TAG-1 is no longer present on the cell surface can re-express TAG-1 on regenerating axons in vitro. Neurons were isolated from the dorsal spinal cord of E19, E20, P0 and P2 rats and plated on astrocyte monolayers. Although the recovery of viable neurons from these older ages was low, many dorsal spinal cord neurons regenerated neurites within 12 h of plating and survived for periods of up to 3 days in vitro (not shown). These neurons expressed LI (Fig. 10A,B) and N-CAM (Fig. 10C,D) on neuronal processes within 24 h after plating and expression of both proteins persisted for 3 days in vitro. In contrast, TAG-1 was not expressed on the neurite surface (Fig. 10E,F). These findings suggest that commissural neurons isolated from late gestational stage embryos do not re-express TAG-1 when they regenerate neurites in vitro. The low yield of neurons obtained from older spinal cords prevented us from examining whether neurons that failed to reexpress the surface form of TAG-1 continued to synthesize the released form.

In view of the low yield of late embryonic neurons, it is possible that the surviving neurons represent a population of neurons distinct from those that originally expressed TAG-1. To determine whether an identified class of spinal neurons re-expresses the surface form of TAG-1 during neurite regeneration after the normal period of expression of the protein in situ, we monitored TAG-1 expression on spinal motor neurons. Motor neurons were labelled by retrograde accumulation of the lipophilic carbocyanine dye, Dil (Honig and Hume, 1986). Histological studies established that within the ventral spinal cord Dil was restricted to the area occupied by motor neuron cell bodies, their dendrites and axon collaterals (Fig. 11 A), although Dil also accumulated in the cell bodies of primary sensory neurons in the DRG (Fig 11A). Motor neurons that regenerated neurites in vitro were identified on the basis of the accumulation of Dil. Of 147 Dil-labelled neurons examined, none expressed cell surface TAG-1 (Fig. 11B,C). In contrast, N-CAM was expressed on most Dil-labelled neurons (not shown), indicating that the presence of Dil does not interfere with the detection of glycoproteins on the cell surface. To examine whether the accumulation of Dil by motor neurons prevented them from expressing TAG-1 on the surface of their neurites, we labeled spinal motor neurons in Ell embryos by injecting Dil into the anterior half of somites at a time when many TAG-1 labelled motor axons are passing through the somite (Keynes and Stem, 1985; Dodd et al. 1988). Dil-labelled motor neurons isolated at Ell and maintained in vitro for 24 h were found to express TAG-1 on their surface (Fig. 11D-F). These results suggest that TAG-1 can be expressed over the period that neurite growth normally occurs and at a stage when motor axons still express TAG-1 in situ (Dodd et al. 1988).

In contrast to the results described above with spinal cord neurons, DRG neurons isolated at times when TAG-1 is no longer detected immunohistochemically in situ (Fig. 12A,B; Dodd et al. 1988) express the protein on regenerating axons in vitro. DRG neurons were removed from P6, P10, P15, P26-P28 and P60-P90 rats, dissociated and plated on a substrate of poly-L-lysine/lammin or on cortical astrocytes in the presence or absence of NGF (100 ng ml⁻¹). Regardless of the age of animals from which DRG were removed, over 90 % of the neurons expressed TAG-1 on their cell bodies and neurites by 24 h of plating (Fig. 12C,D). The surface expression of TAG-1 persisted for 3 days in vitro, the longest time period that we examined. LI (Fig. 12E,F) and N-CAM (not shown) were also expressed at high levels on adult DRG neurons in vitro. Postnatal DRG neurons also released large amounts of TAG-1 into the medium (not shown). Thus postnatal and adult DRG neurons grown in dissociated cell culture re-express the surface form of TAG-1.

Embryonic neurons express a variety of surface molecules that are thought to contribute to the extension and pathfinding of developing axons (Dodd and Jessell, 1988; Harrelson and Goodman, 1988; Takeichi, 1990; Reichardt et al. 1990). One of these molecules, TAG-1, is a 135×10³M_r member of the immunoglobulin family (Furley et al. 1990), which differs from many other vertebrate glycoproteins implicated in neuronal recognition and adhesion in that it is transiently expressed on a restricted subset of central and peripheral neurons in situ (Yamamoto et al. 1986; Dodd et al. 1988; Yamamoto et al. 1990). We show here TAG-1 is also expressed transiently on the surface of a subpopulation of spinal neurons in vitro and in addition is released from these neurons. TAG-1 disappears from the surface of cultured spinal neurons with a time course that parallels the loss of the protein from the same neurons in the developing embryo; however, neurons continue to release TAG-1 in vitro. The expression of TAG-1 on the cell surface appears to be regulated by mechanisms intrinsic to the neurons that synthesize TAG-1, and neurons isolated after surface expression of TAG-1 expression has ceased in vivo do not re-express surface TAG-1. In contrast to these results with spinal neurons, TAG-1 persists on the surface of DRG neurons in vitro well beyond the normal period of TAG-1 expression on these neurons in situ. Moreover, adult DRG neurons, which do not express detectable TAG-1 tn situ, re-express the protein on their surface when placed in vitro, suggesting that the cell surface form of TAG-1 is regulated differently by central and peripheral neurons.

TAG-1 has a hydrophobic carboxy terminal domain suggesting that the protein is attached to the neuronal surface membrane by a GPI anchor (Furley et al. 1990). In support of this possibility, TAG-1 can be released from brain membranes and from the surface of cell lines transfected with TAG-1 cDNA by treatment with PI-PLC (Furley et al. 1990). The present studies provide evidence that the predominant or exclusive mode of covalent attachment of TAG-1 to the surface of primary neurons is via a GPI anchor. In addition, embryonic central and peripheral neurons release large amounts of TAG-1.

Both the surface and released forms of TAG-1 can be detected in cell lines transfected with TAG-1 cDNA clones (Furley et al. 1990), which suggests that differential RNA splicing is not required for the generation of the two different forms of the protein. However, the biosynthetic processing that gives rise to the surface and released forms of TAG-1 in primary neurons is not established. The detection of [³H]ethanolamine in a fraction of the TAG-1 released in the absence of PI-PLC treatment suggests that released protein may derive, in part, from cleavage of the GPI-anchored surface form by endogenous phospholipases or proteases. A serum phospholipase with specificity for the GPI linkages of surface glycoproteins has recently been characterized (Huang et al. 1990). The amount of [³H]ethanolamine incorporated into the released form appears to be much lower than that incorporated into the PI-PLC cleaved surface form, which suggests that a large fraction of TAG-1 released from DRG neurons is not derived from the GPI-anchored cell surface form. The GPI anchor is added to proteins in the endoplasmic reticulum soon after synthesis of the nascent polypeptide chain. Cleavage of the carboxy terminal hydro-phobic region of the protein is required for attachment of the phosphoethanolamine residue to the new C-terminal residue (Doering et al. 1990). The released form of TAG-1, which lacks an ethanolamine group, could therefore be generated in the endoplasmic reticulum by failure to cleave the carboxy terminal domain of TAG-1 or by the failure to attach the GPI anchor after carboxy terminal peptide cleavage. Additionally, the released form of TAG-1 could be generated by proteolytic cleavage of the surface form.

TAG-1 shares many biochemical and functional properties with axonin-1, a protein shown to be synthesized by developing chick neurons in cell surface and released forms (Stoeckli et al. 1989; Ruegg et al. 1989, 1990) and it is possible that axonin-1 is the chick homologue of TAG-1. Analysis of the synthesis of axonin-1 by chick DRG neurons has provided evidence that, like TAG-1, the released form of axonin-1 does not derive primarily from the surface form (Ruegg et al. 1990).

TAG-1 is expressed on the surface of a subset of embryonic spinal neurons in culture. These neurons are likely to correspond to commissural and motor neurons, the two major classes of neurons in the spinal cord to express TAG-1 in vivo at early embryonic stages (Dodd et al. 1988), although other classes of spinal neurons express TAG-1 later in embryonic development (Furley et al. 1990; Vaughn et al. 1990). Over the first few days in vitro, spinal cord neurons lose TAG-1 from their surface, whereas N-CAM expression persists and the levels of LI are initially low but increase with time in vitro. Several lines of evidence indicate that under the present in vitro assay conditions, the disappearance of TAG-1 from the surface of dorsal spinal cord neurons appears to be regulated by events intrinsic to the neurons that synthesize TAG-1 rather than by factors in the environment of these neurons. First, TAG-1 is expressed in vitro on Ell dorsal spinal cord for a somewhat longer period than on E13 dorsal spinal cord neurons. Second, the loss of surface TAG-1 is not dependent on the substratum upon which dorsal spinal cord neurons are grown or on serum factors. Third, adding E13 dorsal spinal cord neurons to older spinal cord cultures does not accelerate the time course of disappearance of TAG-1. In contrast to these results with TAG-1, changes in substratum conditions have been shown to affect the surface expression of a 130 – 140 × 10³M_r axonal glycoprotein, Bravo, which appears to be a related immunoglobulin family member (de la Rosa et al. 1990). It remains possible that in vivo expression of TAG-1 is regulated, in part, by environmental signals.

Previous studies have revealed that TAG-1 is expressed on the axons of commissural and motor neurons only over the initial phase of their growth (Dodd et al. 1988). TAG-1 expression on commissural neurons ceases as axons reach the ventral midline of the spinal cord and cross the floor plate. These observations raised the possibility that the floor plate contributes to the loss of the TAG-1 from the surface of commissural axons (Dodd et al. 1988). The present in vitro experiments provide evidence that transient expression of TAG-1 on commissural neurons can occur independently of contact with the floor plate Similarly, the transient expression of TAG-1 by motor neurons is likely to be regulated by an intrinsic timing mechanism, rather than by interactions with their immediate or final targets. This does not preclude the possibility that the floor plate has a role in refining the timing or spatial expression of TAG-1 on commissural axons in vivo.

The transient expression of TAG-1 on the surface of spinal neurons in vitro could result from a developmental increase in the activity of endogenous phospholipases or proteases that cleave the membrane anchored form of the protein as it is inserted into the membrane. Alternatively, there could be a developmental loss of factors required for transfer of the GPI anchor to the protein backbone, leading by default to a released form of the protein. Neurons that synthesize TAG-1 are likely also to express other immunoglobulin superfamily members that are linked to the cell surface via a GPI linkage, for example Thy-1 (Ferguson and Williams, 1988) and F11/F3 (Brummendorf et al. 1989; Gennarini et al. 1989a). Thy-1 is expressed on the surface of cultured spinal neurons at a time when surface expression of TAG-1 has ceased (Brown et al. 1984; Xue and Morris, 1990), suggesting that neurons may have mechanisms for regulating, differentially, the expression of individual GPI-anchored proteins.

The finding of a persistent, released form of TAG-1 may provide an explanation for apparent differences in the expression of TAG-1 in vivo as assessed by different detection methods. Immunocytochemical studies with monoclonal and rabbit antibodies detect TAG-1 only transiently during development (Yamamoto et al. 1986; Dodd et al. 1988). However, biochemical studies show that TAG-1 mRNA and protein can be detected in postnatal and adult brain at times when there is no protein detectable by immunocytochemistry (Furley et al. 1990). It is possible in situ that antibodies detect only the cell surface form of TAG-1 and that the apparent disappearance of the protein reflects the loss of the cell surface form. The released form of TAG-1 may not be detected in vivo because it is bound to cell surface or extracellular molecules that mask the reactive epitopes of the protein. At present, the functions of the surface and released forms of TAG-1 are not clear. One possibility is that TAG-1 released into the local environment of axons may be capable of regulating axonal growth by interacting homophilically with TAG-1 or heterophilically with other proteins on the axonal surface.

The regulation of TAG-1 expression by DRG neurons appears to differ markedly from that of spinal neurons. TAG-1 is not detectable in rat DRG in situ after P6, whereas TAG-1 persists well beyond this time on the surface of DRG neurons in vitro. The persistent expression of TAG-1 is not caused by absence of non-neuronal cells in vitro, because TAG-1 expression persisted when DRG were grown as explant cultures. However, it is possible that the properties of non-neuronal cells change in vitro, leading to a deregulation of TAG-1 expression. Alternatively, TAG-1 expression could be contained by the formation of contacts between sensory axons and their peripheral targets which were not present in vitro. There is evidence that other biochemical properties of DRG neurons are regulated by contact with the peripheral targets of sensory neurons (Philippe et al. 1988; Marusich et al. 1986; Marusich and Weston, 1988).

Central and peripheral neurons also differ in their ability to re-express TAG-1 on regenerating axons in vitro. Commissural and motor neurons isolated from late embryonic or postnatal rats did not re-express TAG-1 when they regenerated axons in viuo. The mechanisms that direct expression of TAG-1 to the surface of early embryonic spinal neurons therefore do not appear to be reactivated when these neurons regenerate axons at later times. In contrast to spinal cord neurons, DRG neurons re-express TAG-1 on their surface when they are placed in culture at times when TAG-1 expression is no longer detected in situ. As discussed above, the absence of detectable TAG-1 on the surface of adult sensory neurons in situ may reflect the presence of environmental signals that repress TAG-1 expression. Removal of DRG neurons from the influence of these signals may permit the re-expression of TAG-1.

Central and peripheral neurons have markedly differing abilities to regenerate axons after damage Neuronal regeneration is thought to be regulated, in part, by the local environment of damaged axons (Bray et al. 1987; Schwab, 1990). However, the disappearance of neuronal cell surface glycoproteins during development also could contribute to the failure of axonal regeneration in the central nervous system. For example, some central and peripheral neurons lose the ability to extend axons on laminin substrata during development (Cohen et al. 1987; Hall et al. 1987). This appears to result from the inactivation of integrins or other laminin receptors present on the axonal surface that is triggered by contact of axons with their tectal targets (Cohen et al. 1989; Reichardt et al. 1990). The present observations provide additional evidence that glycoproteins implicated in cell adhesion and axonal growth are lost from the neuronal surface during development. In contrast to laminin receptors, however, the temporal expression of TAG-1 on spinal neurons may be regulated, in part, by a cell intrinsic timing mechanism that occurs independent of contact of the axon with its target.

We are grateful to Paul Patterson, Fritz Rathjen and Rod Piggott for providing antibodies to LI, to Martin Low for supplies of PI-PLC and to L Reichardt and E. de la Rosa for sending preprints of unpublished work. Caroline Kopek provided assistance in dissections. We also thank Dan Felsenfeld, Andrew Furley, Marc Tessier-Lavigne and Mary Hynes for helpful comments, and Karen Liebert and Vicki Leon for assistance in preparing the manuscript. This work was supported by the Howard Hughes Medical Institute (T.M.J.) and by grants from the Irma T. Hirschi Foundation and National Science Foundation (to J D ). T.M.J. is an Investigator of the Howard Hughes Medical Institute