Development 102. 409-425 (1988) 409 Printed in Great Britain © The Company of Biologists Limited 1988

Development of macroglial cells in rat II. An in situ immunohistochemical study of oligodendroglial lineage from precursor to mature myelinating cell

RICHARD REYNOLDS and GRAHAM P. WILKIN Department of Biochemistry, Imperial College of Science and Technology, London SW7 2AZ, UK

Summafy

Using immunofluorescence with a panel of antibodies (WM) tracts of the folia, and then away from this that recognize antigens expressed by oligodendroglia, central into the and Purkinje the myelin-producing cells of the CNS, at different cell layers, and finally the molecular layer. The time stages of differentiation from precursor to mature cell, delay between the expression of GC, CNP and MBP we have investigated the development of cells of this was the same for oligodendroglia in all of these layers, lineage in cryostat sections of rat cerebellum. Our suggesting the presence of an intrinsic clock control- results are consistent with the view that glial precur- ling the initial expression of these myelin components. sors, identified by their expression of the ganglioside The early appearance of CNP in oligodendroglia GD3, arise in the subependymal layers of the 4th ventricle and migrate to their final position in the suggests a role for this protein in the early stages of cerebellum via the superior medullary velum, and to myelinogenesis. some extent the peduncles. As the cells reach their final destination they make the transition to recogniz- able galactocerebroside (GC)-expressing oligoden- + + Abbreviations: GC, galactocerebroside; CNP, 2',3'-cyclic droglia, via a GD3 /GC intermediate. The myelin- nucleotide 3'-phosphohydrolase; MBP, myelin basic associated protein 2',3'-cyclic nucleotide 3'-phospho- protein; WM, white matter; CNS, central nervous hydrolase (CNP) appears at the same time as GC, system; SMV, superior medullary velum; GFAP, glial whereas myelin basic protein (MBP) is expressed 2-3 fibrillary acidic protein. days after GC and CNP, immediately prior to myelin formation. A very clear progression of oligodendro- Key words: oligodendroglia, astroglia, myelination, glial differentiation was observed from the SMV into cerebellum, glial cells, immunohistochemistry, rat, the base of the cerebellum, up into the white matter macroglia] cell.

Introduction the oligodendroglial precursor cell into the mature myelin-producing oligodendroglia is crucial to the Although much is known about the morphological understanding of the relative failure of remyelination changes that occur during oligodendroglial differen- following the extensive demyelination seen in mul- tiation from earlier studies, both at the light and tiple sclerosis. Considerable debate still exists con- electron microscopic level (for review see Wood & Bunge, 1984), relatively little is known about the cerning the cell of origin and the mode of differen- sequential biochemical events that occur during the tiation of oligodendroglia in the developing CNS. The acquisition of oligodendroglial characteristics by morphological studies of Skoff et al. (1976) and immature cells, and during the assembly of the Imamoto et al. (1978) suggested that the immediate various myelin components. A knowledge of the precursors of astroglia and oligodendroglia are differ- sequence of events leading to the transformation of entiated enough to be quite distinct from each other. 410 R. Reynolds and G. P. Wilkin

More recent studies have used dissociated CNS cell antibody was prepared and characterized as described cultures and immunocytochemistry with cell-specific previously (Reynolds & Herschkowitz, 1986). Rabbit anti- markers to extend these earlier morphological obser- CNP antibodies (provided by Dr E. M. Carey, Sheffield) vations. Using these techniques two groups have and anti-MBP were produced and characterized as de- independently reached the conclusion that a bipoten- scribed elsewhere (Reynolds et al. 1987b). Mouse mono- clonal antibody against MBP (clone 2C3) was provided by tial progenitor glial cell exists in both the optic nerve Dr R. Thompson (Cambridge) and has been characterized and cerebellum that is capable of differentiation into elsewhere (Elfman et al. 1986). Rabbit polyclonal anti- an oligodendroglia or a fibrous astroglia (Raff et al. bodies against the astrocytic marker glial fibrillary acidic 1983; Levi et al. 1986). The time period in early CNS protein (GFAP) were purchased from DAKO Ltd. development during which these multipotential gliob- lasts, identified in dissociated cell culture, can be found in vivo is not known. Even with the advent of Immunofluorescence antibodies as cell-specific markers the major difficulty The indirect immunofluorescent staining technique with in all of these studies remains the identification of frozen sections was used. Sprague-Dawley rats ranging dividing and immature glial cells and the extrapol- from immediately postnatal to adult were anaesthetized ation from development in culture to development in with sodium pentobarbitone (Sagatal, May and Baker) and vivo. In a recent study, we used a monoclonal perfused through the left ventricle for 2min with PBS followed by 5 min with 4 % paraformaldehyde in PBS. The antibody LBj, which binds to G ganglioside (Curtis D3 cerebella were removed and placed in the same fixative for et al. 1987), to show that GD3 ganglioside expression a further 4-5 h and then in 30 % sucrose in PBS overnight. is characteristic of immature neuroectodermal cells in The tissue was frozen in isopentane cooled in solid CO2 and postnatal rat cerebellar cultures and, in particular, a 3-5/im parasagittal sections cut on a cryostat (Reichert- glial progenitor cell similar to that found in the rat Jung) at — 25°C. The sections were mounted on gelatin- optic nerve (Raff et al. 1983). GD3 was also shown to coated slides and placed in a humid box for 1-2 h before be expressed by cells of the germinal subventricular use. Sections were incubated twice with sodium boro- zones of the IVth ventricle, in agreement with Gold- hydride in PBS (0-5mgmr') and then washed extensively. man et al. (1984), and the immediate precursors of For single immunofluorescent staining with anti-CNP and anti-MBP antibodies, the sections were incubated in meth- oligodendrocytes in the rat cerebellum in vivo (Curtis anol (-20°C) for 5 min to improve antibody accessibility to et al. 1987). Thus it has proved possible to investigate the myelin antigens. the early stages of oligodendroglial differentiation in For single immunofluorescent staining of cryostat sec- vivo. tions, incubation with primary antibody was carried out During the period of rapid CNS myelination, a overnight followed by two rinses in PBS and incubation for number of specific proteins and lipids are synthesized 1 h with the appropriate fluorochrome-conjugated immuno- by developing oligodendroglia and, although the in globulins. For double-label immunofluorescence, the sec- vivo distribution and development of a number of tions were incubated overnight with anti-GD3 (ascites fluid these have been individually investigated (Stern- diluted 1:200-1:1000), monoclonal anti-GC (ascites fluid berger et al. 1978; Zalc et al. 1981), very little is diluted 1:200), or rabbit anti-CNP (diluted 1:200) anti- known about the sequential appearance of these bodies, washed twice with PBS and then incubated for 1 h with TRITC-conjugated sheep anti-mouse IgM (Serotec, oligodendroglial components. In the present study, 1:100 for GD3), FTTC-conjugated goat anti-mouse IgG3 we have investigated the development of cells of the (Nordic, 1:100 for GC) or FITC-conjugated swine anti- oligodendroglial lineage in the rat cerebellum using rabbit IgG (Nordic, 1:200 for CNP). After two 10 min PBS immunofluorescence with antibodies that recognize washes, the sections were incubated for 3h with mono- antigens expressed by these cells at different stages of clonal anti-GC antibodies (for GD3/GC colabelling), differentiation, from bipotential precursor to the monoclonal anti-MBP antibodies (ascites diluted 1:500, for mature myelinating cell. CNP/MBP colabelling) or with rabbit anti-CNP, anti- MBP, anti-GFAP or anti-GC antibodies (all diluted 1:200), followed by FITC-conjugated goat anti-mouse IgG3 (for Materials and methods GD3/GC colabelling), TRITC-conjugated goat anti-mouse IgG (for CNP/MBP colabelling), FITC-conjugated swine Antibodies anti-rabbit IgG (Nordic, 1:200, for GD3/CNP, GD3/MBP and G /GFAP colabelling), or TRITC-conjugated goat Mouse monoclonal antibody LB which binds to G D3 1; D3 anti-rabbit IgG (Cappel Labs, 1:400, for GC/CNP, ganglioside, has been described previously (Curtis et al. GC/MBP and GC/GFAP colabelling). Two 10 min PBS 1987) and was provided by Dr J. Cohen (University washes followed each incubation step. In some exper- College, London). Monoclonal antibodies, O3 and O4, iments, to allow anti-CNP and anti-MBP antibodies greater against sulphatide (Singh & Pfeiffer, 1985) were a gift from access to antigens in the myelin sheath, sections were Professor M. Schachner (Heidelberg). Monoclonal anti- incubated for 2 min in methanol at -20°C after the first galactocerebroside (anti-GC) antibody was provided by Dr primary antibody and fluorescence conjugate incubations. B. Ranscht (London; Ranscht et al. 1982). Rabbit anti-GC Oligodendroglial lineage in the cerebellum 411

Control stainings were carried out with either fluorescence- 1985). Although the monoclonal antibodies against conjugated antibodies alone or with preimmune sera. sulphatide (O3 and 04) did stain GC-positive (GC+) Sections were mounted in glycerol: PBS (9:1), containing oligodendroglia and myelin, using the fixation sched- 2-5% 1,4-diazobicyclooctane (Aldrich) to retard fluoro- ule for GC, they also labelled the filaments of radial chrome fading, and were viewed in a Reichert Polyvar microscope equipped with epifluorescence optics. glia and were therefore considered inappropriate as Toluidine-blue-stained sections were prepared by incu- markers for oligodendroglia in the present study. bating sections for lOmin at room temperature in a 01 % solution in 1 % disodium tetraborate (pH9-0). Sections were then washed in distilled water, air dried and mounted Oligodendroglial developmental pattern - birth to in glycerol. 24h (Fig. 1A,B) + Immediately postnatal GD3 cells were found in most of the subventricular zones lining the Results and also scattered over the developing inferior col- liculus (Fig. 1A). The developing inferior colliculus is Fixation and staining characteristics continuous with the base of the cerebellum via the Our previous investigations revealed the difficulty in + SMV and a small number of G cells was found in successfully staining cells expressing the glycolipids D3 the SMV and at the root of the SMV where it enters G and GC (Curtis et al. 1987) in tissue sections. In D3 the base of the cerebellum (Fig. 2A), and also a small several previous studies, it had proved impossible to + localize GC in sections using the monoclonal anti- number in the primitive folia. A few scattered GC body used in the present study (Miller et al. 1985; cells were already found at the base of the cerebellum and in the SMV (Fig. 2B) and, although these cells Monge et al. 1986). Therefore, we report here on the + + conditions necessary for successful immunofluor- were not GD3 , they were CNP . At this stage, there escent staining of these antigens. For both GC and was no MBP staining in the cerebellum or the SMV. GD3 localization it was essential that the cryostat Another possible route for migration of glial precur- sections were kept humid throughout the whole sors into the cerebellum would be via the developing + procedure and were not allowed to dry out at any peduncles. Small numbers of GD3 cells were indeed stage. Only perfusion of the animals followed by seen in the peduncles at early ages but their numbers 4-5 h immersion of the tissue in 4 % paraformal- were small compared with those seen in the SMV. dehyde gave satisfactory morphological preservation Brightly stained GFAP+ Bergmann glial fibres were together with bright staining. Although both GC and already present in two developing folia and positive GD3 staining were observed when tissue was fixed by cells and fibres were present at the base of the perfusion only, the positive cells were frequently cerebellum/SMV. extensively damaged. Postfixation of the paraformal- By 24h postnatal (Fig. IB), there was a large dehyde-fixed tissue sections with methanol, acetone, + number of GD3 cells at the base of the cerebellum ethanol/acetic acid (all for 5 min at — 20°C) or Triton and in the SMV (Fig. 3A,C) and intermingled X-100 (0-25%, 2min), resulted in total, or almost amongst these were large numbers of GC+ (Fig. 3B) total, loss of the staining pattern due to either + and CNP oligodendroglia (Fig. 3D). Only very extraction of the lipid antigen or loss of antigenicity. occasionally were G +/GC+ cells found but no Fixation of cryostat sections that were cut from D3 G +/CNP+ cells. G + cells were also scattered unfixed brains, as used previously for such studies D3 D3 throughout the developing folia (Fig. 4A) whilst GC+ (Zalc et al. 1981; Goldman et al. 1984), resulted in + both unacceptably poor morphological preservation and CNP cells were beginning to be found in the and also very low staining intensity. deep cerebellar regions and approaching the mouths of several folia (Fig. 4B). All the GC+ cells that had a The above fixation schedule was also optimal for + clearly stained cell body were also CNP but oc- CNP and MBP immunofluorescence; however, post- casionally cells with a weak punctate GC staining fixation of sections with methanol (—20°C, 2-5 min) were found that were not yet CNP+. No CNP+/GC" produced a much clearer and more intense myelin staining for both CNP and MBP. Labelling of cells were observed. MBP immunostaining was pres- oligodendroglial cell bodies with anti-CNP antibodies ent in small amounts in the SMV, at the site of the was relatively intense at all ages irrespective of developing , and also in a few weakly whether the sections were postfixed or not, whereas stained cells at the base of the cerebellum/SMV cell body staining for MBP was observed in cells only (Fig. 4C). Weak GD3 staining was also seen in the immediately before and just after myelin formation, outer external granule cell layer (Fig. 4A) identifying and this was always irrespective of fixation technique, neuroblasts, as observed previously (Goldman et al. in contrast to previous reports (Sternberger et al. 1984; Curtis et al. 1987). 412 R. Reynolds and G. P. Wilkin

»»»*

SMV

SMV

SMV Oligodendroglial lineage in the cerebellum 413

+ + 2-5 days postnatal (Fig. 1C,D) a few GD3 /GC cells were seen in the developing At 2 days postnatal (Fig. 1C), there was still a large WM, both antigens being only weakly labelled. GC+ + + number of GD3 cells at the base of the cerebellum and CNP oligodendroglia, although predominantly + but none in the SMV itself. GD3 cells were present localized in the putative WM (Fig. 6B) were also in all the folia in moderate numbers with no obvious found away from this central tract, in several folia, accumulation in the putative white matter tracts. within the developing granule cell layer and occasion- GC+/CNP+ oligodendroglia were found in increas- ally the layer (Fig. 6C,D). At 7 days, the ing numbers in the deep cerebellar regions and time lag between the expression of GC/CNP and approaching the folia but no increase in the amount MBP was clearly apparent. Short MBP+ myelin of MBP staining was noted. Between day 2 and 5 segments and a few MBP+ cell bodies were found + (Fig. ID) there was a decrease in the number of GD3 towards the ends of several folia in positions where cells at the base of the cerebellum and these cells were GC+/CNP+ oligodendroglia were seen 2-3 days becoming predominantly localized in the developing earlier. WM tracts and at the distal ends of the folia, although After 7 days postnatal, there was a progressive + + + + + no GD3 /GC cells were found. GC /CNP oligo- decrease in the number of GD3 cells in the cerebel- dendroglia were increasingly found in the central lum with only small numbers of positive cells present developing WM of the folia (Fig. 5A,C) and, by 5 in the developing WM amongst the more numerous days, positive cells were already found at the distal GC+ oligodendroglia (Fig. 7A,B). Between 7 and 10 ends of several folia, whilst other folia were still days GC+/CNP+ cells were found increasingly in the devoid of recognizable oligodendroglia. In contrast to internal granule cell layer and Purkinje cell layer, CNP, double immunofluorescence with monoclonal together with an increased number in the developing antibodies to MBP indicated that MBP was still WM tracts, producing a thickening of these tracts + + primarily localized to the deep cerebellar regions (Fig. 7C,D). Greater numbers of GD3 /GC cells (Fig. 5B), with little or no staining present in the folia were found between 9 and 11 days, predominantly (Fig. 5D). At this stage, the SMV and deep cerebellar located in the developing internal granule cell layer areas were strongly GC+, CNP+ and MBP+, with and Purkinje cell layer (Fig. 8), although both labels individual MBP+ myelin sheaths clearly visible. were weak compared to single labelled cells. All clearly labelled GC+ cells were also CNP+ 7-10 days postnatal (Fig. 1E,F) (Fig. 9A,B); however, the absence of G +/CNP+ + + D3 Between day 5 and 7, GC /CNP oligodendroglia cells suggests that, within the limits of detection of were found progressively towards the ends of the folia GC and CNP, GC appears just prior to CNP. By 10 but restricted to the central developing WM tracts. At days (Fig. IF), myelin, identified by MBP staining, 7 days postnatal (Fig. IE), there was an apparent was found at the ends of all but a few folia but was still + increase in the number of GD3 cells in the folia restricted to the central developing WM of the folia. (Fig. 6A), predominantly localized in the developing All of the GC+ oligodendroglia passed through a WM and fanning out at the tips of the folia. The transient stage of weak cell body MBP staining + + majority of GD3 cells were clearly not GC although (Fig. 9C,D) but this probably lasts only 1-2 days. MBP+ cell bodies were also found in the internal Fig. 1. Schematic representation of oligodendroglial granule cell layer at 10 days. development in the rat cerebellum from birth until postnatal day 10, as determined by 12 days postnatal - adult immunohistochemistry. (A) Birth, (B) 24 h, (C) 48h, + (D) 5 days, (E) 7 days, (F) 10 days. The diagrams were GD3 cells were absent from the cerebellum by 12 drawn from photographs of toluidine-blue-stained days postnatal, apart from a transient expression of parasagittal cryostat sections and the relative density of GD3 by Purkinje cell dendrites (R. Reynolds & G. P. stained cells drawn on from the fluorescence microscope Wilkin, unpublished data) between 12 and 20 days. + + + using the symbols below: x, GD3 cells; O, GC /CNP Oligodendroglial development after day 12 is charac- + + + + cells; A, GC /CNP /MBP cell bodies; = , short MBP terized by an increasing number of GC+/CNP+ cells myelin segments; EB, densely myelinated areas. The + in both the central WM of the folia and the grey presence of GD3 cells in the subventricular areas and the matter of the internal granule cell layer and Purkinje inferior colliculus has been added to give an indication of cell layer. MBP+ and CNP+ myelin was now seen a possible origin of the glial precursors entering the cerebellum. Where areas of densely packed myelin coursing into the internal granule cell layer and sheaths are present, stained oligodendroglial cell bodies between the Purkinje cells, representing the myeli- have not been drawn in for clarity. 4V, fourth ventricle; nated Purkinje cell axons and recurrent axon col- Aq, ; IC, inferior colliculus; laterals respectively, and there was a thickening of SMV, superior medullary velum. Bar, 350fim, for all the central WM. At 15 days, GC+ and CNP+ oligo- ages. dendroglia were now found in the molecular layer 414 R. Reynolds and G. P. Wilkin

Fig. 2. Double immunofluorescent labelling of rat cerebellar sections immediately postnatal with antibodies to GD3 (A) and GC (B). Sections show the superior medullary velum (boundary indicated by small arrows) entering the base of the + cerebellum. GD3 cells can be seen in the subventricular areas (small arrowheads) and at the base of the cerebellum (large arrowhead), with only few in the SMV itself. GC+ cells can be seen in the SMV and also entering the cerebellum (large + + arrow). No clearly GD3 /GC cells are seen. Inset shows the location of the areas photographed. Bar, 35 /im.

(Fig. 10A,B) in addition to the internal granule cell staining of the myelin sheath and also the oligoden- layer and Purkinje cell layer. By 20 days, the cells in droglial cell bodies became diffuse and finally could the molecular layer were greater in number and were be seen only as a weak dotted staining over the WM now occasionally also MBP+ and, whereas those in (Fig. 11B). The use of polyclonal anti-GC antiserum the inferior molecular layer were sometimes seen gives a slightly stronger staining of the myelin but less attached to myelin segments, those in the superior staining of the cell bodies and does not represent a molecular layer appeared not to be attached to significant improvement. Whereas MBP brightly myelin, even in the adult (R. Reynolds & G. P. stained myelin after 15 days postnatal but was absent Wilkin, unpublished data). from cell bodies, CNP staining was still present, in No GD3 staining could be seen in the WM or grey both myelin and cell bodies (Fig. 11C,D), in the adult matter in the 20-day or adult cerebellum (Fig. 11A) cerebellum. At no stage during development of the + + + + or in the subventricular areas of the 4th ventricle. As cerebellum were GC /GFAP or GD3 /GFAP myelin is produced and becomes compacted, GC cells found. Oligodendroglial lineage in the cerebellum 415

Fig. 3. Double immunofluorescent labelling of rat cerebellar sections 24 h postnatal with antibodies to GD3 and GC + (A,B) and GD3 and CNP (C,D). Intermingled amongst large numbers of GD3 glial precursors at the base of the cerebellum where the SMV enters it (A,C) are large numbers of GC+ (B) and CNP+ (D) oligodendroglia. No clearly + + + + GD3 /GC or GD3 /CNP cells are seen. Inset shows the location of the areas photographed. Bar, 35/im. 416 R. Reynolds and G. P. Wilkin

Discussion level in the rat cerebellum. As the antigens being studied could be adequately localized in situ by Using a range of antibodies that identify cells of the immunofluorescence on cryostat sections, it was not oligodendroglial lineage from progenitor to mature necessary to dissociate the tissue into single cell cell, we have been able to investigate the origins and suspensions prior to staining, as has been necessary in development of these cells at the light microscopic previous studies (Miller et al. 1985; Monge et al.

Fig. 4. Single immunofluorescent labelling of sections 24 h postnatal with antibodies to GD3 (A), CNP (B) and MBP (C). (A) Large numbers of + strongly GD3 glial precursors were present in the developing folia at 24 h postnatal. Note the weak labelling of neuroblasts in the external granule cell layer (arrows). (B) CNP+ oligodendroglia were already present away from the base of the cerebellum towards the mouth of several folia (large arrows). (C) MBP+ cell bodies and processes (arrowheads) were still restricted to where the SMV (boundary indicated by small arrows) enters the cerebellum. Inset shows the location of the areas photographed. Bar, 35 ^m. Oligodendroglial lineage in the cerebellum 417

Fig. 5. Double immunofluorescent labelling of rat cerebellar sections 5 days postnatal with antibodies to CNP (A,C) and MBP (B,D). Whereas large numbers of CNP+ oligodendroglial cell bodies are present in deep cerebellar regions (A), MBP staining is restricted to small areas of myelinated fibres (B; arrowheads). CNP+ cell bodies present in the WM of the folia (C) are MBP~ (D). Bar, 35 fim. 418 R. Reynolds and G. P. Wilkin

Fig. 6. Double immunofluorescent labelling of 7-day postnatal rat cerebellar sections with antibodies to GD3 and GC + (A,B) and single labelling with antibodies to GC (C) and CNP (D). Whilst numerous GD3 cells were present in both the developing white and (A), GC+ cells were mainly restricted to the central developing WM. In a few folia, GC+ cells were found in the internal granule cell layer (IGL) (arrowheads) at the ends of the folia (C) and CNP+ cells were seen in the IGL away from the central tract (D; arrowheads). Inset shows the location of the areas photographed. Bar, 35 /im. Oligodendroglial lineage in the cerebellum 419

Fig. 7. Double immunofluorescent labelling of 9-day postnatal rat cerebellar sections with antibodies to GD3 and GC + (A,B) and single labelling with antibodies to GC (C) and CNP (D). By 9 days a few GD3 cells only are now present in the developing WM of the folia (A) amongst the more numerous GC+ oligodendroglia (B). Many brightly GC+ (C) and CNP+ (D) cells are now seen in the internal granule cell layer (small arrows), often attached to short myelin segments. Occasionally oligodendroglia are now seen in the Purkinje cell layer (D; arrowhead). Bar, 35 fim. 420 R. Reynolds and G. P. Wilkin

Fig. 8. Double immunofluorescent Labelling of 10- (A,B) and 11- (C,D) day postnatal rat cerebellar sections with + + antibodies to GD3 (A,C) and GC (B,D). GD3 /GC cells (arrows), making the transition from glial precursors to recognizable oligodendroglia, can be seen here at both 10 and 11 days postnatal in the grey matter of a folium. Bar, 22/im. Oligodendroglial lineage in the cerebellum 421

1986). Thus, we have presented a comprehensive, hard to distinguish at early stages (Imamoto et al. truly in situ, immunohistochemical study of oligoden- 1978; Paterson et al. 1973; Altman, 1966), but sugges- droglial development. Earlier morphological and ted that the subependymal layers give rise to imma- ultrastructural studies showed that glial lineages are ture glial cells, which migrate to various regions of the

Fig. 9. GC/CNP and GC/MBP double immunolabelling of 10-day postnatal rat cerebellar sections. Whereas all GC+ cells (A) are also CNP+ (B) not all GC+ cells (C; arrows) are MBP+ (D). The myelin seen in C and D is both GC+ and MBP+ (arrowheads). Bar, 35^m (A,B) and 22urn (C,D). 422 R. Reynolds and G. P. Wilkin

Fig. 10. Immunofluorescent labelling of oligodendroglia in the molecular layer of 15-day postnatal rat cerebellar sections with antibodies to GC (A) and CNP (B). Small-cell-bodied oligodendroglia (arrows) first appear in the molecular layer at 15 days. Stained cells are also seen in the Purkinje cell layer (arrowheads). Bar, 28^m (A) and 22 urn (B). brain where they transform into mature cells. mouse brain transplanted into shiverer mice, which However, not much was known of the steps involved are unable to make MBP (Gumpel et al. 1983; in the formation of glial cells from immature precur- Gansmuller et al. 1986). + sors. Thus, the GD3 precursors make the transition to + + + Our results show a very clear progression of oligo- GC oligodendroglia via a GD3 /GC intermediate, dendroglial differentiation from the SMV into the as we have shown in culture (Curtis et al. 1987; Levi et base of the cerebellum, up into the putative WM al. 1987). GC+ cells then rapidly become GC+/CNP+ tracts of the folia and then away from these tracts into and after a further 2-3 days become GC+/ the granule cell and Purkinje cell layers. Finally, the CNP+/MBP+, followed by myelin formation. The last oligodendroglia to differentiate were found in the delay between the expression of GC, CNP and MBP molecular layer although their number was never was the same for oligodendroglia in all layers of the great. The folia, however, did not all differentiate at cerebellum, suggesting the presence within this cell of the same time; most folia had considerable myelin at an intrinsic developmental clock controlling the in- 10 days postnatal whilst a few folia still had no myelin itial expression of these myelin-associated antigens, at this stage. Although we have not demonstrated in addition to the intrinsic clock proposed for the migration per se, our results are consistent with the differentiation of bipotential progenitors into oligo- idea that the subependymal cells of the 4th ventricle dendroglia (Raff et al. 1985). This developmental + give rise to GD3 glial precursors which then migrate progression is in agreement with a recent study in to their final position in the cerebellum via the SMV culture (Knapp et al. 1987a,b), although the and possibly also the cerebellar peduncles. As soon as CNP+/GC~ cells seen in culture were never observed the cells reach their final destination their expression by us in situ. of GD3 appears to diminish as they begin to express The postulated role of gangliosides in cell-to-cell GC and extend multiple processes. This is in agree- recognition and communication phenomena (Ando, ment with a recent study in culture which showed that 1983), and our observation that GD3 is expressed by the glial progenitor cell is extremely motile until it glial precursors until they begin to differentiate finally differentiates into a recognizable multipolar oligo- into oligodendroglia, both in situ and in culture dendrocyte (Small et al. 1987), and with the obser- (Curtis et al. 1987), raises the question of whether vation that MBP-containing myelin could be found GD3 is involved in the process of migration. The great distances away from grafts of normal neonatal combination of thymidine-labelling studies and Oligodendroglial lineage in the cerebellum 423

Fig. 11. GD3, GC, CNP and MBP single immunostaining of 20-day postnatal rat cerebellar sections. No GD3 staining of either grey or white matter is seen (A). Apart from a few oligodendroglial cell bodies labelled with antibodies to GC (B) in the internal granule cell layer (arrowheads), the staining is of a diffuse nature over both the WM (arrows) and the grey matter (including the majority of the cell bodies). In contrast, by 20 days postnatal, antibodies to CNP (C) brightly labelled both interfascicular (small arrows) and grey matter oligodendroglial cell bodies (note the absence of labelling of the nuclei) and also the myelin. Antibodies to MBP brightly labelled only the myelin sheaths (D) with cell bodies only rarely labelled (arrow). Bar, 35 fim (A,B,D); 22/im (C). 424 R. Reynolds and G. P. Wilkin immunocytochemistry in situ should demonstrate This work was supported by the Multiple Sclerosis conclusively whether GD3 precursors are able to Society of Great Britain and Northern Ireland. We wish to migrate, whether these cells represent exclusively the thank E. Carey (Sheffield), J. Cohen (London), R. Thomp- migrating population, and at which stage in develop- son (Cambridge) and M. Schachner (Heidelberg) for gifts ment, in vivo, cells of the oligodendroglial lineage are of antibodies. capable of division. Seyfried etal. (1983) showed that the sharpest decline in GD3 concentration in the References mouse cerebellum was 9-12 days, which they sugges- ted correlated with the period of most-rapid granule ALTMAN, J. (1966). Proliferation and migration of cell differentiation. However, our results showing undifferentiated precursor cells in the rat during that glial precursors represented the predominant postnatal gliogenesis. Expl Neurol. 16, 263-278. + GD3 population suggest that this decline in GD3 may ANDO, S. (1983). Gangliosides in the nervous system. also be due to the cessation of migration by large Neurochem. Int. 5, 507-537. numbers of glial precursors. CURTIS, R., COHEN, J., FOK-SEANG, J., HANLEY, M. R., GREGSON, N. A., REYNOLDS, R. & WILKIN, G. P. Using freshly dissociated cell suspensions of mouse (1987). Development of macroglial cells in rat cerebellum, Monge et al. (1986) demonstrated a cerebellum. I. Use of antibodies to follow early in vivo similar order of appearance of oligodendroglial anti- development and migration of oligodendrocytes. J. gens; GC followed by Wolfgram protein 1 (Wl; CNP Neurocytol. (In press). is thought to be a major component of the Wolfgram ELFMAN, L., KYNOCH, P. A. M., SIDDLE, K. & protein doublet of myelin [Sprinkle et al. 1980]), and THOMPSON, R. J. (1986). Rat and mouse monoclonal then by MBP 4 days later. However, they found that, antibodies to human myelin basic protein. J. although GC was the first antigen to be expressed in Neurochem. 46, 509-515. oligodendroglia, it was the last to be incorporated GANSMULLER, A., LACHAPELLE, F., BARON-VON EVERCOOREN, A., HAUW, J. J., BAUMANN, N. & into the myelin sheath, which is in total contrast to GUMPEL, M. (1986). Transplantation of newborn our results which show GC present in myelin seg- fragments into the brain of shiverer mutant mice: ments as they are being produced (Fig. 6C). Our extensive myelination by transplanted results on the localization of MBP in developing oligodendrocytes. II. Electron microscopic study. Devi oligodendroglia are similar to those of Sternberger et Neurosci. 8, 197-207. al. (1978) who also found that MBP could be detected GOLDMAN, J. E., HIRANO, M., YU, R. K. & SEYFRIED, T. in these cells just before myelin sheath formation but N. (1984). GD3 ganglioside is a glycolipid characteristic decreased as the cells began to myelinate axons. In of immature neuroectodermal cells. J. Neuroimmunol. contrast to a later study of Sternberger et al. (1985), 7, 179-192. GUMPEL, M., BAUMANN, N., RAOUL, M. & JACQUE, C. MBP could not be detected in oligodendroglial cell (1983). Survival and differentiation of oligodendrocytes bodies of adult CNS tissue, even when a similar from neural tissue transplanted into newborn mouse fixation schedule was used. In contrast to MBP, CNP brain. Neurosci. Lett. 37, 307-311. immunostaining was still intense in cell bodies in the IMAMOTO, K., PATERSON, J. & LEBLOND, C. (1978). adult CNS, and this together with its early appear- Radioautographic investigation of gliogenesis in the ance in immature oligodendroglia, as early as GC, corpus callosum of young rats. I. Sequential changes of makes this protein a most useful tool for identifying oligodendrocytes. /. comp. Neurol. 180, 115-138. oligodendroglia at the light microscopic level at all KNAPP, P. E., BARTLETT, W. P. & SKOFF, R. P. (1987). stages of CNS development. Whereas myelin sheaths Cultured oligodendrocytes mimic in vivo phenotypic characteristics: cell shape, expression of myelin-specific were intensely stained by antibodies to MBP, staining antigens, and membrane production. Devi Biol. 120, of the myelin with antibodies to CNP was much 356-365. weaker and, at the light microscopic level, it is not KNAPP, P. E., SKOFF, R. P. & SPRINKLE, T. 1. (1987). possible to say whether CNP is present throughout Oligodendrocyte development in vitro: double label the myelin sheath. Clearly, further immunohisto- immunostaining for galactocerebroside and 2'3'-cyclic chemical studies at the electron microscopic level are nucleotide 3'-phosphohydrolase. J. Neurochem. 48 necessary to localize this protein within the myelin (Suppl.), S30. sheath. This early expression and persistence in LEVI, G., ALOISI, F. & WILKIN, G. P. (1987). mature cells, together with our evidence in tissue Differentiation of cerebellar bipotential glial precursors into oligodendrocytes in primary culture: culture showing CNP to be present in a putative developmental profile of surface antigens and mitotic growth region of the myelin membrane (Reynolds et activity. J. Neurosci. Res. (In press). al. 1987a), suggests a role for this protein in the very LEVI, G., GALLO, V. & Ciorn, M. T. (1986). Bipotential early stages of myelin formation and also in its precursors of putative fibrous and maintenance. oligodendrocytes in rat cerebellar cultures express Oligodendroglial lineage in the cerebellum 425

distinct surface features and "neuron-like" gamma- oligodendrocytes and astrocytes in primary dissociated aminobutyric acid transport. Proc. natn. Acad. Sci. culture: a possible role for oligodendrocytes in U.S.A. 83, 1504-1508. neurotransmitter metabolism. Brain Res. 371, 253-266. MILLER, R. H., DAVID, S., PATEL, R., ABNEY, E. R. & SEYFRIED, T. N., MIYAZAWA, N. & Yu, R. K. (1983). RAFF, M. C. (1985). A quantitative Cellular localization of gangliosides in the developing immunohistochemical study of macroglial cell mouse cerebellum: analysis using the weaver mutant. J. development in the rat optic nerve: in vivo evidence for Neurochem. 41, 491-505. two distinct lineages. Devi Biol. Ill, 35-41. SINGH, H. & PFEIFFER, S. E. (1985). Myelin-associated MONGE, M., KADIISKI, D., JACQUE, C. & ZALC, B. (1986). galactolipids in primary cultures from dissociated fetal Oligodendroglial expression and deposition of four rat brain: biosynthesis, accumulation and cell surface major myelin constituents in the myelin sheath during expression. J. Neurochem. 45, 1371-1381. development. An in vivo study. Devi Neurosci. 8, SKOFF, R., PRICE, D. & STOCKS, A. (1976). Electron 222-235. microscopic autoradiographic studies of gliogenesis in PATERSON, J. A., PRIVAT, A., LING, E. A. & LEBLOND, C. rat optic nerve. II. Time of origin. J. comp. Neurol. P. (1973). Investigation of glial cells in semithin 179, 313-333. sections. III. Transformation of subependymal cells SMALL, R. K., RIDDLE, P. & NOBLE, M. (1987). Evidence into glial cells, as shown by radioautography after 3H- for migration of oligodendrocyte-type 2 astrocyte thymidine injection into the lateral ventricle of the progenitor cells into the developing rat optic nerve. brain of young rats. /. comp. Neurol. 149, 83-102. Nature, Lond. 328, 155-157. RAFF, M. C, ABNEY, E. R. & FOK-SEANG, J. (1985). SPRINKLE, T. J., WELLS, M. R., GARVER, F. A. & SMITH, Reconstitution of a developmental clock in vitro: a D. B. (1980). Studies on the Wolfgram high molecular critical role for astrocytes in the timing of weight CNS myelin proteins: relationship to 2'3'-cyclic oligodendrocyte regeneration. Cell 42, 61-69. nucleotide 3'-phosphodiesterase. J. Neurochem. 35, RAFF, M. C, MILLER, R. H. & NOBLE, M. (1983). A glial 1200-1208. progenitor cell that develops in vitro into an astrocyte STERNBERGER, N. H., CERRO, C, KIES, M. W. & or an oligodendrocyte depending on culture medium. HERNDON, R. M. (1985). Immunocytochemistry of Nature, Lond. 303, 390-396. myelin basic proteins in adult rat oligodendroglia. J. RANSCHT, B., CLAPSHAW, P. A., PRICE, J., NOBLE, M. & Neuroimmunol. 7, 355-363. SIEFERT, W. (1982). Development of oligodendrocytes STERNBERGER, N. H., ITOYAMA, Y., KIES, M. W. & and Schwann cells studied with a monoclonal antibody WEBSTER, H. DEF. (1978). Myelin basic protein against galactocerebroside. Proc. natn. Acad. Sci. demonstrated immunocytochemically in oligodendroglia U.S.A. 79, 2709-2713. prior to myelin sheath formation. Proc. natn. Acad. REYNOLDS, R., CAREY, E. M. & HERSCHKOWITZ, N. Sci. U.S.A. 75, 2521-2524. (1987a). Differential localization of MBP and CNP in a WOOD, P. & BUNGE, R. P. (1984). The biology of the myelin-like membrane produced by oligodendrocytes in oligodendrocyte. In Oligodendroglia (ed. W. T. culture. J. Neurochem. 48 (Suppl.), S179. Norton), pp. 1-46. New York: Plenum Press. REYNOLDS, R., CAREY, E. M. & HERSCHKOWITZ, N. ZALC, B., MONGE, M., DUPOUEY, P., HAUW, J. J. & (19786). Immunohistochemical localization of myelin BAUMANN, N. A. (1981). Immunohistochemical basic protein and 2'3'-cyclic nucleotide 3'- localization of galactosyl and sulfogalactosyl ceramide phosphohydrolase in a myelin-like membrane produced in the brain of the 30-day old mouse. Brain Res. 211, by cultured oligodendrocytes. Neuroscience (In press). 341-354. REYNOLDS, R. & HERSCHKOWITZ, N. (1986). Selective uptake of neuroactive amino acids by both {Accepted 9 November 1987)