Within the mammalian CNS, astrocytes appear to be a heterogeneous class of cells. To assay the number of distinct types of astrocytes in the rat spinal cord, cell lineage and phenotypic analyses were carried out on cultures from newborn rat spinal cord and five distinct types of astrocytes were observed. Proliferating precur- sors for each class of astrocyte were isolated by low density culture and shown to give rise to 5 distinct and morphologically homogeneous clusters of GFAP+ astro- cytes. Immunocytochemical analysis with antibodies A2B5 and Ran-2, which identify different glial lineages in optic nerve cultures, demonstrated that many clusters included both A2B5+ and A2B5− cells. Similarly, many clusters also possessed a mixture of Ran-2 + and Ran-2− cells, suggesting that in spinal cord cultures, in contrast to optic nerve cultures, expression of these antigens is regulated by individual cells rather than by cell lineage.

Single-cell cloning studies, revealed that the abun- dance and proliferative capacity of individual astrocyte precursors differed depending on the type of astrocyte. To assay the effects of a complex cellular environment on the composition of astrocyte clones, lineage analysis was performed in complete spinal cord cultures using a replication deficient retrovirus. Although similar mor- phologically homogeneous clones of cells to those seen with single-cell clones were observed, the proliferative capacity and relative abundance of the distinct astrocyte precursors differed from that seen in single-cell cloning studies. Together these observations suggest that in spinal cord, gliogenesis is considerably more complex than in the optic nerve and that cultures of newborn rat spinal cord contain multiple, distinct populations of astrocytes.

Two major classes of macroglial cells have been identified in the mammalian central nervous system (CNS): oligodendrocytes and astrocytes. While oligo- dendrocytes are a comparatively homogeneous class of cells, astrocytes appear to be a more heterogeneous class of cells. Classical studies defined two astrocyte classes in the mammalian CNS, fibrous and proto- plasmic (Cajal, 1909). The majority of fibrous astrocytes were located in CNS white matter and had a small cell body with cylindrical processes containing many inter- mediate filaments. By contrast, protoplasmic astrocytes were found mainly in grey matter and had broader processes containing fewer intermediate filaments (Peters et al. 1991). The distinctive bundles of inter- mediate filaments in astrocyte processes are composed, in large part, of glial fibrillary acidic protein (GFAP) (Bignami and Dahl, 1974a; Schachner et al. Y)TT). Clearly, not all GFAP-immunoreactive cells in the mammalian CNS can be classified simply into fibrous or protoplasmic astrocytes. For example, radial glial cells in the developing cerebral cortex (Levitt and Rakic, 1980), Bergmann glia of the cerebellum (Bignami and Dahl, 1974b; Schachner et al. 1977; Wilkin and Levi, 1986) and Müller cells of the retina (Newman, 1986) appear to be specialized forms of astrocyte-like cells. A major characteristic however, shared by all classes of astrocytes, is the expression of GFAP immunoreactivity at some stage in their development (Fedoroff, 1990), and since among CNS neural cells, GFAP expression is restricted to astrocytes (Bignami et al. 1972), it can be used to identify astrocytes unambigously.

The classification of astrocytes into fibrous and protoplasmic astrocytes by their location in the CNS may not be an accurate reflection of the extent of diversity within this family of cells. For example, cultures of rat optic nerve, an almost purely white matter region of the CNS, contains two classes of GFAP-immunoreactive astrocytes termed type-1 and type-2 astrocytes (Raff et al. 1983a). These two types of astrocytes differ in morphology, antigenic phenotype, lineage and proliferative capacity (Miller et al. 1989; Raff, 1989). The majority of type-1 astrocytes have a flat fibroblastic morphology, proliferate in response to serum, and label with the antibody Ran-2 (Bartlett et al. 1981) but not with the antibody A2B5 (Eisenbarth et al. 1979). By contrast, type-2 astrocytes have a process- bearing morphology and do not label with Ran-2 but do label with A2B5 (Raff et al. 1983a). These two types of astrocytes develop at different times and from different cell lineages at least in vitro (Raff, 1989). Type-1 astrocytes begin to differentiate during embryonic development from a committed precursor cell, while type-2 astrocytes begin to differentiate during the second postnatal week from a bipotential precursor cell that can also give rise to oligodendrocytes (Raff et al. 1983b). This bipotential precursor cell has been termed an oligodendrocyte-type-2-astrocyte (O-2A) progenitor cell. Cells with characteristics similar to optic nerve type-1 and type-2 astrocytes as well as 0-2A progenitor cells have been described in cultures from other regions of the CNS including the cerebellum (Levi et al. 1986a,b) and cerebral cortex (Ingraham and McCarthy, 1989) suggesting that they may be a common feature of the mammalian CNS.

The spinal cord is a comparatively complicated region of the CNS, containing a number of different axon tracts in the peripheral white matter, as well as neuron cell bodies in centrally located grey matter (Altman and Bayer, 1984), and the degree of astrocyte diversity in spinal cord is unclear. Morphological analysis suggested that, as in the optic nerve, spinal cord white matter contains two morphologically distinct types of astrocytes in addition to oligodendrocytes (Liuzzi and Miller, 1987), but little is known of additional astrocyte diversity in both spinal cord grey and white matter. Recent studies on spinal cord astrocytes and oligodendrocytes suggest that the devel- oping spinal cord is considerably more complex in terms of gliogenesis than the developing optic nerve. For example, early in development, the spinal cord has a ventrally located source of oligodendrocytes which is absent from the optic nerve (Warf et al. 1991) and later contains a distinct astrocyte precursor which is absent or rare in the optic nerve (Fok-Seang and Miller, 1991).

To determine the extent of astrocyte diversity in the rat spinal cord, we have used three distinct approaches: (1) immunohistochemical analysis with antibodies A2B5 and Ran-2, previously used in examining glial lineages in the optic nerve, (2) single-cell clonal analysis of newborn rat spinal cord cells, and (3) retrovirus- mediated gene transfer to study glial cell lineages in complete spinal cord cultures. In other systems includ- ing the peripheral nervous system (Barrofio et al. 1988; Dupin et al. 1990), optic nerve (Temple and Raff, 1986) and the developing CNS (Temple, 1989), single-cell cloning analysis has been utilized to define the intrinsic properties of specific neural precursor cells. This approach, however, has some limitations. For example, individual precursor cells may require specific cell–cell interactions to differentiate into mature cell types. In clonal culture, such cell–cell signals would be absent and the maturation process inhibited. Furthermore, only cells that survive and proliferate at low density can be assayed. An alternative approach to understanding precursor/progeny relationships is retroviral lineage analysis (Sanes et al. 1986; Turner and Cepko, 1987; Luskin et al. 1988). This technique provides a method by which the progeny of individual cells can be traced in a complex cellular environment and has been effec- tively used to study cell lineages both in situ (Sanes et al. 1986; Price et al. 1987; Turner and Cepko, 1987; Luskin et al. 1988; Grey et al. 1988) and in vitro (Price et al. 1987; Luskin et al. 1988; Vayasse and Goldman, 1990).

In the present study, we show that cultures of neonatal rat spinal cord contain five morphologically distinct classes of astrocytes. Although individual astrocyte precursors gave rise to morphologically homogeneous astrocytes, in general, both A2B5 and Ran-2 immunoreactivity varied among the progeny of individual cells. The relative abundance and proliferat- ive capacity of each distinct astrocyte precursor varied depending on the cell type. These studies suggest that the neonatal rat spinal cord contains multiple classes of astrocytes which develop from distinct precursor cells.

Dissociated cell culture

Dissociated spinal cord cultures were prepared as described previously (Warf et al. 1991). Briefly, spinal cords from newborn Sprague-Dawley rats were dissected and the meninges removed. The tissue was chopped finely and incubated in calcium–magnesium-free MEM (CMF-MEM) medium containing 0.05% trypsin for 30min at 37°C. A 1:1 dilution of 0.25 % EDTA in CMF-MEM was then added and the tissue incubated for a further 7–10min. 10 ml of DMEM- F12+10 % fetal bovine serum (FBS) and 250 μ1 of 0.5 mg ml−1 DNAase was then added and the cells dissociated by trituration through a Pasteur pipet (5×) followed by filtration through 30 μm nylon mesh to remove tissue clumps. Spinal cord cells were plated at densities of either 25×104 (high) or 0.5–1 ×102 (low) viable cells/12mm poly-L-lysine-coated coverslip. Cultures were grown in medium containing 50% immature cortical astrocyte conditioned medium and 15 % FBS in DMEM-F12, at 37°C in 5%CO2. Immature cortical astrocyte conditioned medium was included with the medium to enhance survival and proliferation of astrocyte precursors. Conditioned medium was prepared from confluent cultures of purified type-1 cortical astrocytes following a conditioning period of 3 days, as previously described (Smith et al. 1990).

Under these conditions, after 7 days, cultures of neonatal spinal cord comprised more than 90 % GFAP-immuno- reactive astrocytes. The remaining 10% of cells in these cultures were a combination of GC+ oligodendrocytes, meningeal fibroblasts, microglial cells and endothelial cells. Although neurons were present in the initial cell suspension, no neurofilament immunoreactive cells remained after 7 days. Single-cell-analysis indicated that the majority of the neurons survived for only 2–3 days in such cultures.

Immunohistochemistry

All the antibodies used in this study have been previously described. Cell surface labeling was performed on live cells in the presence of 50 % normal goat serum in DMEM for 30 min at room temperature. Antibodies were used at the following concentrations: A2B5 (Eisenbarth et al. 1979) ascites fluid 1:100; Ran-2 (Bartlett et al. 1981) hybridoma supernatant 1:5; 04 (Sommer and Schachner, 1981) hybridoma supernatant 1:8 and anti-Galactocerebroside (Ranscht et al. 1982) ascites fluid 1:100. Binding of these mouse monoclonal antibodies was visualized by subsequent labeling with a goat-anti-mouse Ig conjugated to rhodamine (Cappel 1:50) for 30 min under the same conditions. Intracellular labeling was performed following fixation and permeabilization of the cells in 5 % acetic acid in methanol for 12 min at −20°C. Incubation conditions were the same as for surface labeling. Anti-GFAP (Accurate) was used at 1:100 and visualized by labeling with goat anti-rabbit Ig conjugated to fluorescein (Cappel 1:50). In control studies in which primary antibodies were eliminated or replaced with non-specific serum of the same species, no specific staining was seen. In double-label studies, species- specific second antibodies were used from which all cross- reacting antibodies had previously been removed. All preparations were viewed on a Nikon Optiphot microscope equipped with epifluorescent optics. Photographs were taken on Tmax film (Eastman-Kodak) at 400 ASA.

Single-cell cloning

To grow spinal cord cells at clonal densities, cells were dissociated as above and the resultant supernatant diluted to yield 1 cell 20 rzl−1. 20 μl of cell suspension was then placed in each well of a Terasaki plate (Nuclon) and incubated for 2 h to allow for cell attachment. Each well was then subsequently screened for the presence of cells using a Nikon inverted microscope and those containing more than a single cell eliminated from further analysis. Single-cell clones were maintained at 37 °C as above and 50% of the medium replaced daily. Under these conditions, approximately one in three wells contained a viable single cell. The morphological analysis of the resultant clones was performed using a Zeiss IM inverted microscope with attached camera, following fixation of the whole plate in 3 % glutaraldehyde at 37 °C for 30min and staining with Coomassie blue, 7 % acetic acid and 35% methanol for 5 min at room temperature. Single-cell cloning analysis was performed from 4 different dissociations of two neonatal rat spinal cords in each case. A total of more than 300 individual clones were examined and the data from all experiments pooled.

Retroviral mediated gene transfer studies

To examine the morphology of the progeny of single glial precursors in a complex cellular environment, retroviral labeling techniques (Sanes et al. 1986) were used. Dissociated cell suspensions of newborn rat spinal cord were plated onto 12 mm poly-L-lysine-coated coverslips at a density of 1.5 ×105 viable cells/well in 50 μl of medium. After cell attachment (2–4 h), the BAG2 replication-incompetent retrovirus carry- ing the E. coli LacZ gene (PSI 2BAG Alpha ATCC no. 9560; Price et al. 1987) was added in 950 μ1 of DMEM-F12 medium +10% FBS. On the following day, the medium was replaced with one containing 50% astrocyte-conditioned medium described above. After 7 days, the cells were fixed in 2% paraformaldehyde with 0.4% glutaraldehyde in phos- phate-buffered saline (PBS) for 10min, washed and cells carrying the LacZ product visualized histochemically using X-gal as a substrate (1mg ml−1 X-gal, 4mM potassium ferrrocyanide, 4mM potassium ferricyanide, 2 mM MgCl2 in PBS pH 7.2; Sanes et al. 1986). To ensure that genetic recombination events resulting in infective virions had not occurred, all batches of supernatant used in these studies were tested on cultured NIH3T3 cells as previously described (Sanes et al. 1986), and no infective virus detected.

To confirm that infected cells represented the progeny of only a single precursor, a small volume of virus supernatant (4μ1), which resulted in not more than three discrete clones/well, was added to each well in a 24-well plate. Under these conditions, labeled clones were clearly spatially distinct from each other even when considerable migration of the progeny had occurred. In control studies, to establish the clonal nature of the labeled cells, cultures were infected with increasing quantities of the same virus supernatant, and the total number of clones per well assayed after 7 days. With increasing amounts of added virus, the number of individual clones increased linearly up to a volume of 25μ1 of virus. However, the number of cells in a specific type of clone remained constant (data not shown). These controls indicate that the clones observed were the product of a virus infection of a single precursor cell. Each retrovirus experiment was repeated on cultures established from at least three different spinal cord dissociations and a total of more than 100 individual coverslips were examined. The cells constituting clones were characterized on the basis of their morphology, and that morphology was then compared with the mor- phology of cells seen in single-cell cloning assays and low density cultures.

Spinal cord cultures contain astrocytes with distinct morphologies

A broad range of morphologies was apparent among neonatal spinal cord astrocytes after one week in culture (Fig. 1). For example, cells with an elongated cell body and fine processes, cells with a flattened angular-shaped cell body, as well as process-bearing and small round pancake-shaped astrocytes were commonly seen. In low density cultures, astrocytes sharing a similar morphology were frequently found in close proximity to each other, raising the possibility that cells with similar morphologies were derived from the same precursor. In cultures grown at 1×102 cells/well for 7 days, the majority of astrocytes were found in distinct clusters and, in most cases, all the astrocytes within an individual cluster had a similar morphology. Clusters of five morphologically distinct types of astrocyte were reproducibly seen in all preparations (Fig. 2). In one type, astrocytes had a characteristic elongated cell body and long fine processes (Fig. 2A). In the second type, the cells had a more flattened and angular morphology (Fig. 2B). In the third type, the cells had a smaller cell body and a process-bearing morphology (Fig. 2C). Two other types of astrocytes were round pancake-shaped cells (Fig. 2D,E). How- ever, one class of pancake-shaped cells was consider- ably smaller than the others. These observations suggested that neonatal rat spinal cord contains multiple types of GFAP-immunoreactive astrocytes that differ in their morphology when cultured under identical conditions.

Fig. 1.

Glial cultures from neonatal rat spinal cord contained astrocytes with multiple morphologies. Phase-contrast (A) and anti-GFAP (B) labeling of a culture of neonatal rat spinal cord after 7 days in culture. The majority of the cells in the culture were astrocytes as demonstrated by their anti-GFAP immunoreactivity. Spinal cord astrocytes exhibited a number of different morphologies including process-bearing cells, cells with an elongated cell body and fine processes as well as cells with more flattened and rounded morphologies. Bar=100μm.

Fig. 1.

Glial cultures from neonatal rat spinal cord contained astrocytes with multiple morphologies. Phase-contrast (A) and anti-GFAP (B) labeling of a culture of neonatal rat spinal cord after 7 days in culture. The majority of the cells in the culture were astrocytes as demonstrated by their anti-GFAP immunoreactivity. Spinal cord astrocytes exhibited a number of different morphologies including process-bearing cells, cells with an elongated cell body and fine processes as well as cells with more flattened and rounded morphologies. Bar=100μm.

Fig. 2.

At very low density, GFAP-immunoreactive astrocytes developed as isolated clusters of cells after 7 days in culture. In general, within individual clusters, all the astrocytes had a similar morphology. Five morphologically distinct astrocytes were seen. (A) Elongated cell body and fine processes. (B) Flattened cell body and angular morphology. (C) Small cell body and process-bearing morphology, (D) Small rounded ‘pancake’-shaped morphology and (E) Large rounded ‘pancake’-shaped morphology. The number of cells in the different types of cluster varied with cells of morphologies A, B and D containing the most cells. Bar=25 μm in all panels.

Fig. 2.

At very low density, GFAP-immunoreactive astrocytes developed as isolated clusters of cells after 7 days in culture. In general, within individual clusters, all the astrocytes had a similar morphology. Five morphologically distinct astrocytes were seen. (A) Elongated cell body and fine processes. (B) Flattened cell body and angular morphology. (C) Small cell body and process-bearing morphology, (D) Small rounded ‘pancake’-shaped morphology and (E) Large rounded ‘pancake’-shaped morphology. The number of cells in the different types of cluster varied with cells of morphologies A, B and D containing the most cells. Bar=25 μm in all panels.

Antigenic phenotype of spinal cord astrocytes

In optic nerve (Raff 1989) and cortical cultures (Behar et al. 1988), the expression of specific antigenic phenotypes has proved useful to distinguish between cells of the type-1 and O-2A lineages. To determine if astrocytes in spinal cord cultures with distinct mor- phologies also had distinct antigenic phenotypes, high density spinal cord cultures were labeled by indirect immunofluorescence with the cell surface antibodies; A2B5 to identify putative O-2A lineage cells, Ran-2 to identify putative type-1 astrocyte lineage cells, 04 to identify immature oligodendrocytes and anti-GC to identify more mature oligodendrocytes, in combination with anti-GFAP (Table 1). Few GC+ or O4+ cells were seen in any cultures. Most 04 immunoreactive cells were process-bearing and some of these cells were not GFAP-immunoreactive. By contrast, considerable numbers’ of A2B5-immunoreactive astrocytes were seen. The A2B5+ astrocytes were mainly of three distinct morphologies; process-bearing (as in Fig. 2C), small pancake-shaped (as in Fig. 2D) or large pancake- shaped (as in Fig. 2E). However, not all the astrocytes of any particular morphology were A2B5 immuno- reactive. Many spinal cord astrocytes were also Ran-2 immunoreactive. In general, astrocytes with a flattened morphology were more frequently Ran-2 labeled than were process-bearing astrocytes, although all the morphological classes of astrocytes included some labeled cells.

Table 1.

Antigenic phenotype of morphologically distinct astrocytes in spinal cord cultures

Antigenic phenotype of morphologically distinct astrocytes in spinal cord cultures
Antigenic phenotype of morphologically distinct astrocytes in spinal cord cultures

The diverse morphology of A2B5- and Ran-2- immunoreactive astrocytes has two possible expla- nations. First, as in optic nerve cultures, labeling with either A2B5 or Ran-2 may distinguish two distinct astrocyte lineages, and both lineages may give rise to individual astrocytes with multiple morphologies. Alternatively, the continuous expression of A2B5 or Ran-2 immunoreactivity may not reflect distinct astro- cyte cell lineages, but rather may depend on other factors such as the maturation state of individual cells. To distinguish between these two possible explanations, low density cultures of spinal cord astrocytes were labeled with either A2B5 or Ran-2 in combination with anti-GFAP. In these cultures astrocytes grew in distinct clusters and more than 75 % of the clusters containing A2B5+ cells also contained some A2B5− cells (Fig. 3). Similarly, many of the clusters containing Ran-2+ cells also contained some Ran-2− cells, even though individual cells within the great majority of all clusters had a similar morphology (Fig. 3). These observations clearly indicated that a single spinal cord astrocyte precursor had the capacity to give rise to both A2B5+ and A2B5− or Ran-2+ and Ran-2− progeny. There- fore, it seemed likely that in short-term rat spinal cord cultures, in contrast to short-term optic nerve cultures the continued expression of A2B5 or the expression of Ran-2 immunoreactivity was not a distinctive feature of a particular astrocyte lineage, but rather reflected some other characterisitic such as the maturation state of the individual cells.

Fig. 3.

Very low density cultures of neonatal rat spinal cord cells double labeled with anti-GFAP (B and E) and A2B5 (C) or Ran-2 (F) antibodies. Note that while both the process-bearing cells in the upper panels labeled with anti-GFAP only one of the cells was A2B5 immunoreactive. In the lower panels all the cells in the cluster had a similar morphology and labeled with anti-GFAP; however, only two of the cells labeled with Ran-2. (A and D) phase-contrast micrographs. Bar=25μm.

Fig. 3.

Very low density cultures of neonatal rat spinal cord cells double labeled with anti-GFAP (B and E) and A2B5 (C) or Ran-2 (F) antibodies. Note that while both the process-bearing cells in the upper panels labeled with anti-GFAP only one of the cells was A2B5 immunoreactive. In the lower panels all the cells in the cluster had a similar morphology and labeled with anti-GFAP; however, only two of the cells labeled with Ran-2. (A and D) phase-contrast micrographs. Bar=25μm.

Single-cell cloning revealed six morphologically distinct classes of glial cells

Because individual astrocyte precursors gave rise to cells with a constant morphology but variable antigenic phenotype, we used morphological characteristics to determine the abundance and proliferative capacity of the distinct astrocyte cell types in single-cell clonal cultures. To identify unambiguously the progeny of large numbers of single glial precursors in newborn rat spinal cord cultures, cells were grown at clonal densities in microwell cultures and the morphology of the progeny assessed after Coomassie blue staining.

This morphological clonal analysis reproducibly revealed six distinct types of cells (Fig. 4). Cells in clone type (4A) had an elongated cell body, and fine processes. Cells in clone type (4B) had a flattened cell body and a more angular morphology. Cells in clone type (4C) had a small cell body and process-bearing morphology. Cells in clone type (4D) had a small pancake-shaped cell body. Cells in clone type (4E) were very large and frequently pancake-shaped, while cells in clone type (4F) were small, bipolar cells. These data suggest there are at least six morphologically distinct types of non-neural cells in the developing rat spinal cord which continue to proliferate postnatally in culture. Comparison of the morphologies of these clonal derived cells with the anti-GFAP-labeled astro- cyte clusters (compare cell morphologies in Figs 2 and 4) suggests that at least five of these six populations of cells are astrocytes. The non-GFAP-immunoreactive cell type may represent either an immature glial precursor or a non-neural cell type such as meningeal or endothelial cells.

Fig. 4.

Single-cell cloning reproducibly revealed six morphologically distinct types of glial cells after 7 days in culture. Cells in panel A had an elongated cell body and fine processes. Cells in panel B had a more flattened morphology and few processes. Cells in panel C had a small cell body and process-bearing morphology. Cells in panel D had a small pancaked-shaped morphology. The cell in panel E were very large and pancaked- shaped while the cells in panel F were small and bipolar. Bar=50μm.

Fig. 4.

Single-cell cloning reproducibly revealed six morphologically distinct types of glial cells after 7 days in culture. Cells in panel A had an elongated cell body and fine processes. Cells in panel B had a more flattened morphology and few processes. Cells in panel C had a small cell body and process-bearing morphology. Cells in panel D had a small pancaked-shaped morphology. The cell in panel E were very large and pancaked- shaped while the cells in panel F were small and bipolar. Bar=50μm.

The abundance of the clonally isolated precursor cells differed between the individual cell types (Table 2). After 7 days in isolated culture, the most common type of clone (29%) was that containing cells with a flattened angular morphology as shown in Fig. 4B. 20 % of the clones contained cells with an elongated cell body and fine processes (4A), while 15 % of the clones contained cells with either process-bearing or pancake- shaped morphologies (4C,D,E). The least common clones were those containing small bipolar cells (4F) which constituted only 4% of the remaining clones.

Table 2.

Relative abundance and proliferative capacity of different precursors in clonal cultures

Relative abundance and proliferative capacity of different precursors in clonal cultures
Relative abundance and proliferative capacity of different precursors in clonal cultures

Not only did the abundance of the different clones vary, but the mean number of cells within the distinct types of clones also differed, suggesting that the individual precursors had distinct proliferative capaci- ties. Clones comprising cells with either an elongated cell body (4A) or a flattened angular morphology (4B) contained the largest number of progeny (Table 2) indicating significant proliferation of their precursors. By contrast, clones comprising large pancake-shaped cells and small bipolar cells (4E,F) contained few cells (Table 2), indicating limited proliferation of their precursors (Table 2). Similar relative levels of prolifer- ation were seen when clonal cultures were examined after 14 days in culture (Table 2). Therefore, under the specific culture conditions, the precursors of the distinct glial cell classes had different proliferative capacities.

A small proportion (7%) of clones examined were composed of cells with more than a single morphologi- cal phenotype. In general, such mixed clones contained cells of only two distinct morphologies, and the most common mixed clones contained cells with mor- phologies corresponding to both cell types 4A and B, as well as cell types 4C and D. Although such mixed clones could have been derived initially from more than one cell, it is also possible that a small proportion of spinal cord glial precursors have the capacity to give rise to more than one morphologically distinct cell type. Since the same cell types were frequently seen associated with each other, it may be that certain distinct types of astrocytes share a common precursor.

Glial precursors give rise to morphologically homogeneous populations of cells in complete spinal cord cultures

To determine if individual precursors gave rise to morphologically homogeneous clones of glial cells in the presence of other cell types, high density spinal cord cultures were established, and at the time of plating were infected with a replication incompetent retrovirus carrying the E. coli lacZ reporter gene. After 7 days, the cultures were fixed and the morphological charac- teristics of the progeny of individual cells determined through the expression of the reporter gene product.

Although similar cell types were seen using both retrovirus-mediated gene transfer studies (Fig. 5) and single-cell cloning assays (Fig. 4), quantitative analysis of the relative proportions and size of the distinct types of clones showed significant differences depending on the analytical approach used (Table 3). Most striking was the virtual absence of retrovirus-labeled clones of the small bipolar cells (Fig. 4F) and the comparatively low abundance of clones containing cells of types 5 D and E (Table 3). While together, clones containing cells of these morphologies constituted 33 % of the clones seen in single-cell cloning studies (Table 2), by retro- viral analysis they constituted only 5% (Table 3). The relatively low abundance of these three cell types in retroviral studies could result from a lower efficiency of virus infection in these cells. Alternatively, it may reflect either the selective development of other cell types, or the inhibition of development of these cells in higher density cultures. The proportion of clones containing cells with morphologies such as those seen in 5B and C increased significantly in the retrovirus- labeling studies as compared with the single-cell cloning analysis. This increase in clones of particular cell types may reflect a higher efficiency of virus infection, enhanced survival, or the morphological conversion between distinct cell types possibly as a result of cell –cell interactions.

Table 3.

Relative abundance and proliferative capacities of individual precursors in complete spinal cord cultures

Relative abundance and proliferative capacities of individual precursors in complete spinal cord cultures
Relative abundance and proliferative capacities of individual precursors in complete spinal cord cultures
Fig. 5.

Retroviral analysis of the progeny of single cells in complete spinal cord cultures reproducibly demonstrated 5 morphologically distinct types of glial clones after 7 days in culture. The labeled cells in panel A had elongated ceil bodies and fine processes. The labeled cells in B had a more flattened morphology. The labeled cells in C had a process-bearing morphology and small cell body while the labeled cells in both D and E were ‘pancake’ shaped although the cells in D were considerably smaller than those in E. Bar=50μm.

Fig. 5.

Retroviral analysis of the progeny of single cells in complete spinal cord cultures reproducibly demonstrated 5 morphologically distinct types of glial clones after 7 days in culture. The labeled cells in panel A had elongated ceil bodies and fine processes. The labeled cells in B had a more flattened morphology. The labeled cells in C had a process-bearing morphology and small cell body while the labeled cells in both D and E were ‘pancake’ shaped although the cells in D were considerably smaller than those in E. Bar=50μm.

Comparison of the number of cells within individual clones between the single-cell and retroviral studies indicated that isolated single cells gave rise to slightly larger clones than those seen in retroviral analysis (Compare Table 2 and 3). For example, in the two cell types that underwent the most extensive proliferation in clonal cultures (Figs 4 and 5A,B) the average number of cells/clone after 7 days in culture was 5.2 and 5.7, respectively, and the largest clones contained 14 cells and 26 cells. In retrovirus-labeling analysis, the average number of cells/clone of the same cell types was 3.7 and 3.4 respectively (Table 3) and the largest clones contained 6 and 17 cells. A reduction in the size of clones was also apparent with small pancake-shaped cells (Figs 4D and 5D) where the average number of cells/clone was 3.0 in single-cell cloning analysis, but only 1.8 in retrovirus analysis. The smaller sized clones in these studies may be explained by either the incomplete expression of the reporter gene in all the cells of a clone, the selection of the most prolific precursors within the population by single-cell cloning, or the inhibition of glial proliferation through cell-cell interactions with other cells in complete spinal cord cultures.

A small proportion (6%) of the retroviral-labeled clones contained cells with mixed morphologies. These mixed clones are likely to reflect the differentiative capacity of individual precursors, and not the overlap of two unrelated clones, since the addition of virions was titrated to yield no more than three clones/coverslip and the number of cells in these mixed clones was not significantly larger than in clones of a single type. Again, as in single-cell clonal analysis, the most commonly observed mixed clones were those composed of cells shown in 5A and B, as well as 5C and D (Table 3). These observations provide further support for the hypothesis that some of the distinct classes of astrocytes in cultures of newborn rat spinal cord share a common precursor cell.

Cultures of neonatal rat spinal cord contained several morphologically distinct classes of GFAP-immuno- reactive astrocytes. Single-cell clonal analysis, and retrovirus infections of complete spinal cord cultures revealed that the majority of astrocyte precursors in neonatal spinal cord gave rise to morphologically homogeneous astrocytes, while the relative abundance and proliferative capacity of the individual astrocyte precursors varied depending on the cell type. In addition, immunohistochemical analysis indicated that in general, while single glial precursors gave rise to clusters of morphologically homogeneous astrocytes, the cell surface properties of individual astrocytes within a cluster varied. These observations suggest that the spinal cord is considerably more complex in terms of astrocyte diversity than the structurally simpler optic nerve where only two distinct types of astrocytes have been described (Raff, 1989).

The variable antigenic phenotype of clonally related astrocytes in spinal cord cultures appeared to result from alterations in A2B5 and Ran-2 immunoreactivity on the surface of some progeny of particular spinal cord astrocyte precursors, and made it difficult to utilize these criteria for lineage analysis in spinal cord cultures. The variable expression of antibody labeling on spinal cord astrocytes differs from that seen with optic nerve astrocytes. In neonatal optic nerve, after 7 days in culture, the majority of type-1 astrocyte lineage cells were labeled with Ran-2, while the majority of O-2A lineage cells were labeled with A2B5 (Raff et al. 1983a, Lillien and Raff, 1990). In optic nerve cultures, however, A2B5 immunoreactivity was lost from cells of the O-2A lineage during differentiation into oligoden- drocytes (Raff et al. 1983a), and in long-term cultures, from type-2 astrocytes (Lillien and Raff, 1990), suggesting that even in this case A2B5 immunoreac- tivity was not a permanent marker for a particular cell lineage. It seems likely that initially in spinal cord cultures the expression of A2B5 immunoreactivity may characterize distinct populations of glial precursors, but that the stage in astrocyte maturation at which A2B5 expression is lost from the surface of spinal cord astrocytes differs from that seen in optic nerve cultures. Indeed, following A2B5-mediated complement lysis in newborn spinal cord cultures, no A2B5-immuno- reactive cells developed after a further 7 days in culture although in sister cultures treated with complement alone, large numbers of A2B5+ astrocytes developed (Fok-Seang and Miller, 1991). These observations indicated that in spinal cord cultures, as in optic nerve cultures, A2B5+ astrocytes are derived from A2B5+ cells and not from A2B5− cells. The regulation of A2B5 immunoreactivity on spinal cord astrocytes is unknown. If the loss of immunoreactivity is correlated with the maturation of individual cells, as appears to be the case with oligodendrocytes (Raff et al. 1983b) and type-2 astrocytes in optic nerve cultures (Lillien et al. 1990), then it is possible that in the spinal cord cultures, astrocytes are less synchronous in their developmental profile than they are in a simpler part of the nervous system such as the optic nerve.

Ran-2 immunoreactivity also appeared to character- ize initially distinct subpopulations of astrocyte precur- sors in spinal cord cultures, and at least some Ran-2 immunoreactive cells represent non-overlapping popu- lations with A2B5 immunoreactive cells. For example, following A2B5-mediated complement lysis of rat spinal cord cultures some Ran-2-immunoreactive cells remain, even though no A2B5-immunoreactive cells were present in such cultures (Fok-Seang and Miller, unpublished observation). However, the regulation of Ran-2 immunoreactivity during astrocyte maturation appears complex since in optic nerve cultures some type-2 astrocytes, which were not initially Ran-2 immunoreactive, acquire reactivity with extended cul- ture (Lillien and Raff, 1990). Thus, in spinal cord, the finding that a single astrocyte precursor has the capacity to give rise to both Ran-2 + and Ran-2− progeny may reflect either the development or the loss of immuno- reactivity by some progeny.

In single-cell clonal analysis six morphologically distinct classes of glial clones were isolated from neonatal rat spinal cord, suggesting that there are multiple glial precursors in the starting population. However, one morphological class of cell, the small bipolar cells were not seen in GFAP-labeling exper- iments and rarely in retrovirus analysis of whole cultures. Therefore, it seems likely that this particular type of cell is not a mature astrocyte, but rather may be a precursor cell which, in the presence of other spinal cord cells, differentiates and changes morphology. Although such bipolar cells could be meningeal or endothelial in origin, they may also represent the spinal cord equivalent of the optic nerve O-2A progenitor cells, which have a similar morphology, and in mixed culture differentiate into oligodendrocytes or process- bearing type-2 astrocytes depending on the culture conditions (Raff et al. 1983b).

Since some glial precursor cells, such as O-2A progenitors undergo morphological transformation during differentiation, it is probable that the six morphologically different types of clones seen in single- cell analysis do not reflect six completely distinct glial cell lineages, but rather that at least some of these cells represent discrete stages in the maturation of fewer cell lineages. This hypothesis is consistent with the finding that some clones contained a mixture of more than a single type of astrocyte, and that specific types of cells (e.g. Figs 4 and 5C,D) were frequently seen in association with each other, suggesting they are derived from the same precursor. Furthermore, we recently described a A2B5-immunoreactive astrocyte precursor from spinal cord cultures which first gave rise to process-bearing astrocytes (4C) and then in clonal culture underwent maturation to small pancake-shaped cells (4D), indicating that these two morphologically distinct cell types were derived from a common precursor cell (Fok-Seang and Miller, 1991).

The vast majority of astrocyte precursors in neonatal rat spinal cord cultures however, gave rise to cells with a homogeneous morphology, at least in short-term cultures. For example, the finding that in retroviral lineage analysis the majority of clones were composed of cells with similar morphology suggests that few glial precursors in neonatal rat spinal cord gave rise to multiple cell types at this age. It is likely that at earlier developmental stages, more glial precursor cells will have the capacity to give rise to multiple different types of astrocytes, as well as other neural cell types. Indeed, recent studies in the developing chick spinal cord have provided strong evidence for precursor cells that give rise to both motor neurons and glial cells during early development (Leber et al. 1990), while in cultures of embryonic rat cerebral cortex, clones containing both flat fibroblastic and process-bearing astrocytes have been observed (Price et al. 1987).

Comparison of the data obtained from single-cell cloning assays and retroviral analysis indicated a number of unexpected differences. Not only did the relative abundance of the different cell types vary between single-cell cloning and retroviral analysis, but the number of cells in individual clones was also different. In most cases, the number of cells in a clone was greater in the single-cell cloning studies than in the retroviral analysis and, as a result, this phenomenon was most noticeable in those cells that underwent the most proliferation. The reason for the differences in relative abundance and proliferative capacity is unclear. It may be that some glial precursors are less susceptible than others to virus infection and are, therfore, under- represented in the retroviral analysis, while the more susceptible precursors are over represented. Further- more, if the retrovirus incorporated into only one daughter cell during the initial cell division, then the apparent clone size in retroviral analysis would be smaller than that seen in single-cell cloning assays. It is possible, however, that the proliferation of some glial precursors is inhibited by the presence of other types of neural cells through a specific cell-cell interaction such as contact inhibition (Abercrombie and Heaysman, 1954). For example, in vitro cerebellar astrocyte proliferation is inhibited by coculture with granule cell neurons (Hatten, 1985) and such an inhibition of proliferation would also contribute to smaller sized clones in complete spinal cord cultures.

In conclusion, it appears that the neonatal rat spinal cord contains multiple morphologically distinct types of astrocytes, although the number of actual cell lineages generating this degree of astrocyte diversity is un- known. Cultures of rat optic nerve, which has a far simpler composition than the spinal cord, contain two classes of astrocytes derived from two distinct cell lineages (Raff, 1989). Which spinal cord astrocytes correspond to those seen in optic nerve cultures is unclear; however, it seems likely that most regions of the CNS will share some common classes of astrocytes, while more specialized classes of astrocytes will have a more restricted distribution.

We thank A. Hall, E. Noll, H. Zhang and J. Fok-Seang for helpful discussions on this work, and A. Hall and S. Landis for comments on the manuscript. This study was supported by Grant NIH-25597-03. R.H.M. is an Alfred P. Sloan fellow.

Abercrombie
,
M.
and
Heaysman
,
J. E. M.
(
1954
).
Observations on the social behaviour of cells in tissue culture 11. ‘Monolayering’ of fibroblasts
.
Expl Cell Res
.
13
,
276
91
.
Altman
,
J.
and
Bayer
,
S. A.
(
1984
).
The development of the rat spinal cord
.
Advances in Anal., Embryo/, and Cell Biol
.
85
,
Springer
Verlag Berlin
.
Barroffio
,
A.
,
Dupin
,
E.
and
Le Douarin
,
N. M.
(
1988
).
Clone forming ability and differentiation potential of migratory neural crest cells
.
Proc. natn. Acad. Sci. U.S.A
.
85
,
5325
5329
.
Bartlett
,
P. F.
,
Noble
,
M.
,
Pruss
,
R. M.
,
Raff
,
M. c.
,
Rattray
,
s.
and
Williams
,
C. A.
(
1981
).
Rat neural antigen-2 (Ran-2): a cell surface antigen on astrocytes, ependymal cells, Muller cells and leptomeninges defined by a monoclonal antibody
.
Brain Res
.
204
,
339
351
.
Behar
,
T.
,
Mcmorris
,
F. A.
,
Novotny
,
E. A.
,
Barker
,
J. L.
and
Dubois-Dalacq
,
M.
(
1988
).
Growth and differentiation properties of 0-2A progenitors purified from rat cerebral hemispheres
.
J. Neurosci. Res
.
21
,
168
180
.
Bignami
,
A.
and
Dahl
,
D.
(
1974a
).
Astrocyte specific protein and radial glia in the cerebral cortex of new born rat
.
Nature
252
,
55
56
.
Bignami
,
A.
and
Dahl
,
D.
(
1974b
).
The development of Bergmann glia in mutant mice with cerebellar malformations: reeler, staggerer and weaver. Immunofluorescence study with antibodies to glial fibrillary acidic protein
.
J. comp. Neural
.
155
,
219
230
.
Bignami
,
A.
,
Eng
,
L. F.
,
Dahl
,
D.
and
Uyeda
,
C. T.
(
1972
).
Localization of the glial fibrillary acidic protein in astrocytes by immunofluorescence
.
Brain Res
.
43
,
429
435
.
Cajal
,
S. R.
(
1909
).
Histologie du Systeme Nerveux de l’Homme et des Vertebres
. pp.
230
252
.
Paris
:
Maloine
.
Dupin
,
E.
,
Baroffio
,
A.
,
Dulac
,
C.
,
Cameron-Curry
,
P.
and
Le Douarin
,
N. M.
(
1990
).
Schwann cell differentiation in clonal cultures of the neural crest as evidenced by the anti-Schwann cell myelin protein monoclonal antibody
.
Proc. natn. Acad. Sci
.
87
,
1119
1123
.
Eisenbarth
,
G. S.
,
Walsh
,
F. S.
and
Nirenberg
,
M.
(
1979
).
Monoclonal antibody to a plasma membrane antigen of neurons
.
Proc. natn. Acad. Sci. U.S.A
.
76
,
4913
4917
.
Fedoroff
,
S.
(
1990
).
Astrocyte reaction in injury and regeneration
.
In Advances in Neural Regeneration Research
. pp.
161
170
. (
F.
Seil
, ed).
Wiley-Liss
,
New York
.
Fok-Seang
,
J.
and
Miller
,
R. H.
(
1991
).
Astrocyte precursors in neonatal rat spinal cord cultures
.
Submitted J. Neurosci
.
Grey
,
G. E.
,
Glover
,
J. c.
,
Majors
,
J.
and
Sanes
,
J. s.
(
1988
).
Radial arrangement of clonally related cells in the chicken optic tectum: Lineage analysis with a recombinant retrovirus
.
Proc. natn. Acad. Sci
.
85
,
7356
7360
.
Hatten
,
M.E.
(
1985
)
Neuronal regulation of astroglial morphology and proliferation in vitro
J. Cell Biol
.
100
,
384
396
.
Ingraham
,
C. A.
and
Mccarthy
,
K. D.
(
1989
).
Plasticity of process bearing glial cell cultures from neonatal rat cerebral cortical tissue
.
J. Neurosci
.
9
,
63
69
.
Leber
,
s. M.
,
Breedlove
,
s. M.
and
Sanes
,
J. R.
(
1990
)
Lineage arrangement and death of clonally related motorneurons in chick spinal cord
.
J. Neurosci
.
10
,
2451
2462
.
Levi
,
G.
,
Gallo
,
V.
and
Ciotti
,
M. T.
(
1986a
).
Bipotential precursors of putative fibrous astrocytes and oligodendrocytes in rat cerebellar cultures express distinct surface features and ‘neuron-like’-aminobutyric acid transport
.
Proc. natn. Acad. Sci. U.S.A
.
83
,
1504
1508
.
Levi
,
G.
,
Gallo
,
v.
,
Wilkin
,
G. P.
and
Cohen
,
J.
(
1986b
).
Astrocyte subpopulations and glial precursors in rat cerebellar cell cultures
.
Adv. in Biosci
.
61
,
21
30
.
Levitt
,
P.
and
Rakic
,
P.
(
1980
).
Immunoperoxidase localization of glial fibrillary acidic protein in radial glial cells and astrocytes of the developing rhesus monkey brain
.
J. comp. Neural
.
193
,
815
840
.
Lillien
,
L. E.
and
Raff
,
M. C.
(
1990
).
Analysis of the cell-cell interactions that control Type-2 astrocyte development in vitro
.
Neuron
4
,
525
534
.
Liuzzi
,
F. J.
and
Miller
,
R. H.
(
1987
)
Radially oriented astrocytes in the normal adult rat spinal cord
.
Brain Res
.
403
,
385
388
.
Luskin
,
M. B.
,
Pearlman
,
A. L.
and
Sanes
,
J. R.
(
1988
).
Cell lineage in the cerebral cortex of the mouse studied in vivo and in vitro with recombinant retrovirus
.
Neuron
1
,
635
647
.
Miller
,
R. H.
,
Ffrench-Constant
,
C.
and
Raff
,
M. C.
(
1989
).
The macroglial cells of the rat optic nerve
.
A. Rev. Neurosci
.
12
,
517
534
.
Newman
,
E. A.
(
1986
).
The Millier cell
.
In Astrocytes
Vol.
1
. pp.
149
166
. (Eds.
Fedoroff
,
S.
and
Vernadakis
,
A.
)
New York
:
Academic Press Inc
.
Peters
,
A.
,
Palay
,
s. L.
and
Def. Webster
,
H.
(
1991
).
The Fine Structure of the Nervous System: Neurons and their Supporting Cells
.
Oxford University Press
.
New York
.
Price
,
J.
,
Turner
,
D.
and
Cepko
,
C.
(
1987
).
Lineage analysis in the vertebrate nervous system by retrovirus-mediated gene transfer
.
Proc. natn. Acad. Sci. U.S.A
.
84
,
156
160
.
Raff
,
M. C.
(
1989
).
Glial cell diversification in the optic nerve
.
Science
243
,
1450
1455
.
Raff
,
M. C.
,
Miller
,
R.H.
and
Noble
,
M.
(
1983b
).
A glial progenitor cell that develops in vitro into an astrocyte or an oligodendrocyte depending on the culture medium
.
Nature
303
,
390
396
.
Raff
,
M. c.
,
Mirsky
,
R.
,
Fields
,
K. L.
,
Lisak
,
R. P.
,
Dorfman
,
S. H.
,
Silberberg
,
D. H.
,
Gregson
,
N. A.
,
Liebowitz
,
S.
and
Kennedy
,
M. C.
(
1978
).
Galactocerebroside is a specific cell surface antigenic marker for oligodendrocytes in culture
.
Nature
274
,
813
816
.
Ranscht
,
B.
,
Clapshaw
,
P.A.
,
Price
,
J.
,
Noble
,
M.
and
Seifert
,
W.
(
1982
).
Development of oligodendrocytes and Schwann cells studied with a monoclonal antibody against galactocerebroside
.
Proc. natn. Acad. Sci. U.S.A
.
79
,
2709
2713
.
Sanes
,
J. R.
,
Rubinstein
,
J. L. R.
and
Nicolas
,
J. F.
(
1986
).
Use of a recombinant retrovirus to study postimplantation cell lineage in mouse embryos
.
EMBO J
.
5
,
3133
3142
.
Scttachner
,
M.
,
Hedley-Whyte
,
E.T.
,
Hsu
,
D. W.
,
Schoonmaker
,
G.
and
Bignami
,
A.
(
1977
).
Ultrastructural localization of glial fibrillary acidic protein in mouse cerebellum by immunoperoxidase labeling
.
J. Cell Biol
.
75
,
67
73
.
Smith
,
G. M.
,
Rutishauser
,
u.
,
Silver
,
J.
and
Miller
,
R. H.
(
1990
).
Maturation of astrocytes in vitro alters the extent and molecular basis of neurite outgrowth
.
Devi Biol
.
138
,
377
390
.
Sommer
,
I.
and
Schachner
,
M.
(
1981
).
Monoclonal antibodies (01-04) to oligodendrocyte cell surfaces: an immunocytological study in the central nervous system
.
Devi Biol
.
83
,
311
327
.
Temple
,
S.
(
1989
).
Division and differentiation of isolated CNS blast cells in microculture
.
Nature
340
,
471
473
.
Temple
,
S.
and
Raff
,
M. C.
(
1986
).
Clonal analysis of oligodendrocyte development in culture: evidence for a developmental clock that counts cell divisions
.
Cell
44
,
773
779
.
Turner
,
D. L.
and
Cepko
,
C. L.
(
1987
)
A common progenitor for neurons and glia persists in rat retina late in development
.
Nature
328
,
131
136
Vayasse
,
P. J.
and
Goldman
,
J.E.
(
1990
).
Clonal analysis of glial lineage in neonatal forebrain development in vitro
.
Neuron
5
,
227
235
.
Warf
,
B. c.
,
Fok-Seang
,
J.
and
Millerr
.
H.
(
1991
).
Evidence for the ventral origin of oligoendrocyte precursors in the developing rat spinal cord
.
J. Neurosci. In press
.
Wilkin
,
G. P.
and
Levi
,
G.
(
1986
).
Cerebellar astrocytes In Astrocytes
Vol.
1
. pp.
245
268
. (Eds).
Fedoroff
,
S.
and
Vernadakis
,
A.
New York
:
Academic Press Inc
.