sign in sign up
<<
Home , Astrocyte, Central nervous system, Glia, Radial glial cell

Proc. Natl. Acad. Sci. USA Vol. 92, pp. 2061-2065, March 1995 Neurobiology

Radial glial transformation to is bidirectional: Regulation by a diffusible factor in embryonic KIM E. HUNTER AND MARY E. HATrEN Laboratory of Developmental Neurobiology, The Rockefeller University, 1230 York Avenue, New York, NY 10021 Communicated by Torsten N. Wiesel, The Rockefeller University, New York, NY, October 26, 1994

ABSTRACT During development of mammalian cerebral been to be a partially committed cell, with further cortex, two classes of glial cells are thought to underlie the differentiation toward an cell fate occurring in a establishment of cell patterning. In the embryonic period, progressive manner. migration ofyoung is supported by a system of radial To develop an in vitro model system for analysis of signals glial cells spanning the thickness of the cortical wall. In the that regulate the astroglial developmental pathway, we marked neonatal period, neuronal function is assisted by the physio- radial glial cells in situ so that they could subsequently be logical support of a second class of astroglial cell, the astro- identified in vitro, and we examined the control of process cyte. Here, we show that expression of embryonic radial glial formation and antigen marker expression by labeled cells. We identity requires extrinsic soluble signals present in embry- show that expression of a identity in mammalian onic forebrain. Moreover, astrocytes reexpress features of forebrain is determined by the availability of diffusible induc- radial in vitro in the presence of the embryonic cortical ing signals. These signals are temporally regulated during signals and in vivo after transplantation into and are found in embryonic but not mature . These findings suggest that the transformation of forebrain. Furthermore, we show that these signals act to radial glia cells into astrocytes is regulated by availability of transform mature astrocytes to a radial glial phenotype, inducing signals rather than by changes in cell potential. indicating that the transformation from radial glial cell to astrocyte is reversible. Preliminary biochemical characteriza- In the developing , astroglial cells are among the first cells tion indicates that the inducing signals are protein in nature to differentiate, with bipolar radial glia being predominant in and may represent one or more previously uncharacterized the embryonic period. Described over a century ago by neural growth regulators. These findings provide support for Kolliker (1), Retzius (2), and Ramon y Cajal (3), radial glia the role of extrinsic signals in determining and maintaining a have been proposed to function in the initial steps of brain radial glial identity. histogenesis, including support of neuronal migration and laminar patterning (3-7). During the perinatal period, changes in astroglial form occur, with bipolar radial glia disappearing MATERIALS AND METHODS and being replaced by multipolar astrocytes. Differential Embryonic Cell Cultures. Embryonic day 14 (E14) mouse expression of cellular antigen markers by these two general were removed, placed in Earle's balanced salt classes of astroglia in the perinatal mammalian forebrain solution (EBSS; GIBCO), labeled pially with 1,1'-dioctadecyl- (8-10), and direct observation of successive changes in glial 3,3,3',3'-tetramethylindocarbocyanine perchlorate (DiI; Mo- form during later phases of histogenesis, has provided support lecular Probes; 3 mg/ml in EtOH/dimethyl sulfoxide), and for the general view that radial glial cells transform into then incubated in EBSS for 3 hr in 95% 02 at 32°C. Neocor- astrocytes (8-11). tices were dissociated by treatment with trypsin (17) and plated A general model for cell specification in the , on poly(L-lysine) (0.01 mg/ml; Sigma) at densities of 105-106 termed , has emerged from studies of the neural cells per ml. For ceilings, 106 cortical cells from animals aged crest lineage, in which soluble factors control the expansion of E14 to P6 (postnatal day 6) were plated into membrane filter multipotent populations and restrict cell fate via a culture inserts (Nunc). Inserts were overlaid onto cul- progressive restriction of gene expression (for reviews, see refs. tures of 105 DiI-labeled neocortical cells in DMEM-F12 12-14). Support for the existence of multipotent central nervous (GIBCO) supplemented with penicillin-streptomycin and system (CNS) neural precursor cells in vivo has been gained 0.4% ReduSer II (GIBCO) and maintained at 35°C in 5% through in vivo lineage tracing methods (15, 16). Within CNS C02/95% air. After 24 hr, cultures were fixed (2% paraformal- neural populations, bipotential cells are thought to dehyde in phosphate-buffered saline), mounted in Slowfade generate neurons and glia, with the emergence of the radial glial (Molecular Probes), and then immunostained [mouse anti- lineage occurring early in cortical development (7). RC2 (gift of M. Yamamoto, University of Tsukuba, Japan) A critical question concerning the role of extrinsic signals in followed by fluorescein isothiocyanate (FITC)-conjugated neuropoiesis is whether differentiation factors are instructive goat anti-mouse IgM (Sigma)]. Cells were viewed by epifluo- or permissive. Instructive signals would commit a precursor rescence on a Zeiss Axiophot microscope. Coverslips were cell to a particular cell lineage at the expense of production of scored for the percentage of all Dil-positive cells bearing other classes of cells, while in a permissive system the gener- processes; 100 cells on each coverslip were scored in each case. ation of different classes of cells would follow a stochastic Preparation of Astrocyte Cultures and Assay of E14 Activ- mechanism, with differentiation factors permitting the sur- ity. Astrocytes were purified from P6 murine vival of particular, partially committed progenitor cells (12). (17). Purified astrocytes were plated onto coverslips coated in Both cases imply a series of steps that progressively restrict the poly(L-lysine) within 24-well plates at a density of 104 cells per fate of precursor cells, with terminal differentiation of a ml. Astrocytes were cultured with or without an insert (pre- particular cell type. In classical studies, the radial glial cell has pared as described above) of 106 E14 neocortical cells for 48

The publication costs of this article were defrayed in part by page charge Abbreviations: CNS, ; E14, embryonic day 14; payment. This article must therefore be hereby marked "advertisement" in P6, postnatal day 6; DiI, 1,1'-dioctadecyl-3,3,3',3'-tetramethyl- accordance with 18 U.S.C. §1734 solely to indicate this fact. indocarbocyanine perchlorate; GFAP, glial fibrillary acidic protein. 2061 Downloaded by guest on September 27, 2021 2062 Neurobiology: Hunter and Hatten Proc. NatL Acad ScL USA 92 (1995) hr and then fixed in 2% paraformaldehyde permeabilized in DMEM-F12 supplemented with penicillin-streptomycin. After ethanol, and immunostained for glial fibrillary acidic protein 30 min, collection of 50 2-ml samples began. Astrocytes were (GFAP) [rabbit anti-cow GFAP (Dako) followed by rhoda- cultured with 1 ml of each fraction for 2 days before immu- mine anti-rabbit IgG (Sigma), or RC2 followed by FITC- nostaining for GFAP and scoring for astrocyte morphology, as conjugated anti-mouse IgM (Sigma)] and then mounted in described above. Slowfade. Approximately 200 cells on each coverslip were In Utero Transplantations. Pregnant mice of 15 days ges- scored according to phenotype as normal or radial; the tation were anesthetized (375 jig of tribromoethyl per percentage of radial cells on each coverslip was calculated. g of body weight in 2.5% tertiary amyl alcohol) (Sigma). Western Blot Analysis of GFAP Levels. Astrocytes were Astrocytes harvested from P6 cortexwere labeled with PKH26 plated at subconfluent density into two T75 tissue culture as described (20) and resuspended at 20 x 106 cells per ml. flasks (Nunc) precoated with poly(L-lysine) in DMEM-F12 After laparotomy, the uterus was transilluminated to visualize with 10% fetal calf serum. On the following day, cells were , and astrocytes (0.5 ,ul) were injected into the lateral washed twice in DMEM-F12 supplemented with ReduSer and ventricle of each embryonic forebrain with a 1-,ul Hamilton then aspirated; one flask was given 10 ml of serum-free DMEM-F12; the other was given 10 ml of E14 neocortical cell syringe, after which the injection site was sealed with cyano- conditioned medium. After 2 days, astrocytes were rinsed acrylic adhesive, the embryos were replaced in the abdominal twice in and magnesium-free phosphate-buffered cavity, and the abdominal wall was sutured. Two days later, saline (GIBCO) and harvested with a cell scraper. Cells were embryos were perfused with 4% paraformaldehyde, and the pelleted and resuspended in 500 ,ul of protein sample buffer cortex was removed and postfixed for 24 hr. Cortices were (18), and then protein concentrations were determined (pro- embedded in 3% agarose; 100-,um coronal Vibratome sections tein detection kit; Bio-Rad). Samples were boiled and 5 or 20 were cut and mounted in Slowfade. Dye-labeled cells within ,ug was loaded onto a 12% acrylamide gel alongside Kaleido- Vibratome sections of transplant tissue were imaged with a scope molecular weight markers (Bio-Rad). Once run, gels Zeiss Axiovert microscope (20). were blotted onto nitrocellulose paper by using a Bio-Rad semi-dry blotting system. Western blots were analyzed as RESULTS described (19). Anti-GFAP antibody was used at 1:2500; alkaline phosphatase secondary antibody (Vector Laborato- Radial Glial Cell Differentiation Requires Diffusible Sig- ries) was used at 1:7500. Bound secondary antibody was nals. To selectively label cortical radial glia in situ, the li- detected by bromochloroindolyl phosphate/nitroblue tetrazo- pophilic dye DiI was applied to the pial surface of the E14 lium development. murine neocortex, a technique previously shown to mark radial Fast Protein Liquid Chromatography (FPLC) Gel Filtra- glia (9). By microscopy, labeled cells expressed the highly tion. Five milliliters of E14 cortical cell conditioned medium elongated form characteristic of radial glial cells (1-3, 21), was injected (1 ml/min) onto a Superdex 200 16/60 gel- spanning the thickness ofthe anlage and forming endfeet along filtration column (Pharmacia) using a running buffer of the pial and ventricular surfaces.

E D 70 90 - _ ce T V: 80- 60- a 70- ou =a~; C) 50 - 00 40 - ° 50- -0

T t 40- 0Cs,, 302 T -- 30- T +.cx, 20 -; * 20 i 204 -5 1 a 10- n- O. >. 0. E14 E18 P1 P6 P10 E14 cortical cell plated, Developmental age x 10-6 of cortical cell ceiling FIG. 1. Expression of radial glia identity in vitro. DiI-labeled radial glia within cultures of dissociated E14 neocortical cells. In high density cultures (106 cells per ml), 80% of radial glia extend processes (A), but, at low density (105 cells per ml), they remain rounded (B). Percentage of differentiated (process bearing) radial glia is proportional to the total number of neocortical cells plated (D). Individual radial glia in low density cultures differentiate when a ceiling of 106 E14 neocortical cells is provided (C). However, ceilings of cells derived from older developmental stages induce progressively fewer radial glia to differentiate, with cells of P6 and older failing to induce any differentiation (E). Data in D and E are means of two separate experiments; SEM is shown. (Bar = 30 t.m.) Downloaded by guest on September 27, 2021 Neurobiology: Hunter and Hatten Proc. NatL Acad Sci USA 92 (1995) 2063 To provide an in vitro model for radial glial differentiation, Radial Glial Inducing Factors Are Temporally Regulated we plated cells labeled in situ in microcultures at high cell During Cortical Development. As the number of radial glial density (106 cells per ml). In , Dil-labeled radial cells falls precipitously during the late embryonic period in vivo glial cells expressed the highly elongated form seen in vivo and (22), we examined whether the inducing activity detected in extended processes of up to 300 ,um (Fig. 1A). Labeled cells embryonic cortex also declined at later developmental stages. could be deemed viable, since dead Dil-positive cells were When labeled E14 radial glia were plated in microcultures shrunken and pyknotic. To confirm the identity of Dil-labeled under Millipore filter inserts with cortical cells harvested from cells, we immunostained the microcultures to examine expres- developmental ages E14 to P10, the amount of activity de- sion of the radial glial cell marker antigen RC2. In the clined with increasing age of the cortical ceiling cells (Fig. 1E). microculture system, all Dil-labeled cells expressed RC2, The time course of the decrease observed with this in vitro demonstrating that the pial labeling technique selectively assay system correlates with the schedule of disappearance of marks radial glial cells. However, these radial glia were found radial glia from the developing murine neocortex; in vivo, not to express GFAP (data not shown). radial glia are most fully differentiated during the period The extent of process formation by Dil-labeled radial glia E14-E18, after which a steady decline is notable, with the declined sharply as the number of cells plated decreased (Fig. majority ofradial glia having disappeared by the end of the first 1D). Negligible glial cell differentiation was seen when cells postnatal week (22, 23). This suggested that the transformation were plated at very low density (<105 cells per ml) where there of radial glial cells into astrocytes was related to the availability was a low probability of contact between cells. This suggested of a soluble factor(s) during cortical development. that extrinsic signals, present in the embryonic cortical popu- To test whether the factor(s) can induce GFAP-positive lation supported the expression of a radial glial identity in vitro. astrocytes to reexpress the features of radial glial cells, we To determine whether the differentiation signals we assayed examined the effect of an embryonic factor(s) on a later stage were soluble or membrane bound, we plated labeled radial of astroglial development. Astrocytes harvested from P6 cor- glial cells at very low density (105 cells per ml) beneath a dense tex were cocultured with a Millipore insert with 106 E14 (106 cells per ml) population of unlabeled E14 cortical cells neocortical cells. P6 astrocytes normally have a flattened plated on a Millipore filter insert. In coculture with E14 cells, phenotype in vitro and express GFAP (Fig. 2A). However, in -60% of radial glia elaborated processes (Fig. 1 C and E). coculture with an E14 cell ceiling, P6 astrocytes show a Since process formation occurred in the absence of cell dramatic change in form. The vast majority ('80%) of astro- contact, these results suggest that diffusible signals released by cytes became highly elongated. Among these cells, many E14 cortical cells promote elaboration of radial glial processes expressed a bipolar morphology; others elaborated branched in vitro. processes, some tipped with glial growth cone-like structures

D Control CM treated

_ ~ ~ ~~ .88-~~~~~~~~~~71

41.8 - 5 20 5 20

FIG. 2. Regulation of astroglial identity by a diffusible factor from embryonic cortical cells. P6 harvested astrocytes immunostained for GFAP after 48 hr in the absence B or presence (C) of a ceiling of 106 E14 neocortical cells. Astrocytes are normally flattened in vitro (A), but in the presence of an embryonic ceiling, >80% become highly elongated and thin (B). While many astrocytes assumed a simple bipolar form (b), some have branched processes (p), while others exhibit glial growth cone-like structures at the process termini (g). Elongated astroglial forms express the radial glial marker RC2 (C), absent from control astrocytes (data not shown). Probing Western blots of-astrocyte proteins with anti-GFAP antibody showed that this protein is downregulated in astrocytes after exposure to embryonic signals (D). Numbers 5 and 20 indicate micrograms of total astrocyte protein loaded in gel lane. Numbers 71.8 and 41.8 indicate positions of molecular mass standards relative to the 50-kDa GFAP. (Bars = 60 gm.) Downloaded by guest on September 27, 2021 2064 Neurobiology: Hunter and Hatten Proc. Natl. Acad ScL USA 92 (1995)

.60- bioactivity that can induce a radial astroglial phenotype elutes from the column within a range of a few fractions, estimated 50- to be 50-60 kDa (Fig. 3). In Utero Transplantations. To test whether astrocytes are

0 - competent to reexpress a radial glia identity in vivo, dye- 30- labeled P6 astrocytes were transplanted into the E15 neocor- tex. In agreement with our in vitro data, a number of astrocytes integrated into the cortical wall and elaborated the long processes characteristic of radial glia (Fig. 4). Some of the D 20- transplanted cells formed endfeet with the ventricular surface (Fig. 4A), while others had migrated toward the pial surface 10 (Fig. 4B and C). Many astroglia assumed a radial orientation within the wall of the embryonic cortex (Fig. 4 B and C). In 10 20 30 40 50 contrast, when neonatal astrocytes were transplanted into neonatal (P0 and P3) forebrain, none of the astrocytes as- Fraction Number sumed a radial phenotype (data not shown), instead remaining FIG. 3. Partial purification of embryonic inducing activity. Em- rounded within the cortical wall. Furthermore, previous stud- bryonic cortical cell conditioned medium can be FPLC size fraction- ies have shown that, when neonatal astrocytes are transplanted ated by gel filtration. Radial glia inducing bioactivity elutes in a peak, approximated to be in the range 50-60 kDa. Results of a single to adult rodent brain, they can migrate but retain throughout bioassay are shown. an epithelioid astrocytic phenotype (28, 29). (Fig. 2B). These elongated astroglial forms were reminiscent DISCUSSION of embryonic radial glia. By immunostaining and Western blot analysis, respectively, in the presence of the E14 ceiling astro- The divergence of astroglial and neuronal cell precursors from cytes expressed RC2 (Fig. 2C) and downregulated GFAP a common precursor is thought to occur early (7), with the (Fig. 2D). development of astroglial cells constituting a sublineage that Molecular Characterization of Radial Glial Differentiation generates radial glial cells and astrocytes (7-10). Although cell Factors. To determine whether any previously described culture experiments have provided supporting evidence for the growth factor was responsible for the effects of E14 cortical transformation of radial glial cells to astrocytes (10) seen in cells on radial glial cell development, we used the astroglial in vivo (8, 9), the present experiments analyze the factors that vitro system to assay platelet-derived growth factor, epidermal control this program of development within the astroglial growth factor, basic , insulin-like lineage. Our findings suggest that expression of radial glial cell growth factor, , and members of the type 1B trans- identity is regulated by the availability of a soluble factor in forming growth factor family, all previously identified as embryonic brain and that the transformation from radial glia astroglial growth regulators (17); stem cell factor, retinoic to astrocyte is bidirectional, with reexpression of radial glial acid, and the recently characterized neu differentiation factor, phenotype upon exposure to embryonic signals (Fig. 5). Our all of which are known to affect the development and differ- findings are consistent with the general view that the devel- entiation of a number of cell types (24-26); and dibutyryl opment of specific cell types in the CNS arises via the action cAMP, reported to affect astrocyte morphology (27). None of of soluble factors that restrict the potential of multipotential these factors induced astrocytes to express a radial glial precursor cells. identity in vitro, suggesting that the activity we have identified Recent experiments on the establishment of glial fate in the may be a previously uncharacterized growth factor. FPLC cell population implicate a role for glial growth fractionation of embryonic cortical cell conditioned medium factor in the commitment of multipotential neural precursor on a size-exclusion gel-filtration column and subsequent cells to a glial lineage (30). The present results on glial screening of fractions on cultured P6 astrocytes shows that the differentiation within one sublineage, the astroglial lineage,

FIG. 4. Expression of radial glial identity by astrocytes after transplantation into E15 neocortex. Pseudocolored confocal images show that P6 astrocytes (dye labeled) integrate into E15 neocortex after transplantation. Some transplanted astrocytes remained at the ventricular surface (V) (A); others were apparently migrating through the cortex, bearing a long trailing process (B and C), while some astrocytes had reached the superficial layers of the neocortex near the pial surface (P) (C). Many of the integrated astroglial forms assumed the orientation of embryonic radial glia (B and C). (Bar = 30 ,um.) Downloaded by guest on September 27, 2021 Neurobiology: Hunter and Hatten Proc. NatL Acad Sci USA 92 (1995) 2065 appropriate epigenetic signals. The developmentally regulated expression of these signals may therefore play a major role in determining astroglial identity. Thus, glial differentiation may fit into an overall program of neural development where complex signaling systems between neurons and glia underlie the establishment of the neuronal layers in the mammalian cortex. We thank N. Heintz and G. Fishell for advice and discussion and K. Precursor Transitional Stellate Zimmerman and S. Temple for critically reading the manuscript. This Cell Astroglial Forms Astrocyte work was supported by National Institutes of Health Grant 15429 (M.E.H.) and a Revson-Winston Postdoctoral Fellowship (K.E.H.). 1. Kolliker, A. V. (1890) Z. wiss. Zool. 51. Inducing Factor 2. Retzius, G. (1894) Biologische Untersuchungen 6, 1-24. 3. Ramon y Cajal, S. (1911) ofthe Nervous System (Maloine, Radial Glial Cell Paris); reprinted in trans. (1995); trans. Swanson, N. & Swanson, FIG. 5. Bidirectional transformation of radial glial cells to astro- L. W. (Oxford Univ. Press, Oxford). cytes. 4. Angevine, J. B., Jr., & Sidman, R. L. (1961) Nature (London) 192, 766-768. support the general conclusion that soluble factors promote 5. Caviness, V. S., Jr., & Sidman, R. L. (1973) J. Comp. Neurol. 148, glial differentiation in early phases of development 141-151. (30). Our finding that soluble factors in the embryonic brain 6. Rakic, P. (1972) J. Comp. Neurol. 145, 61-84. 7. P. P. are required for radial glial differentiation suggests the exis- Levitt, & Rakic, (1980) J. Comp. Neurol. 193, 815-840. 8. Schmechel, D. E. & Rakic, P. (1979) Anat. EmbryoL. 156, 115- tence of a differentiation factor(s) that induces expression of 152. this particular astroglial phenotype. An aspect of our studies, 9. Voigt, T. (1989) J. Comp. Neurol. 289, 74-88. not explored in depth for cells of the neural crest population, 10. Culican, S. M., Baumrind, N. L., Yamamoto, M. & Pearlman, is the reversibility of the program of CNS glial cell develop- A. L. (1990) J. Neurosci. 10, 684-692. ment, as terminally differentiated astrocytes can be induced to 11. Raff, M. C. (1989) Science 243, 1450-1455. reexpress markers of radial cells both in vitro and in vivo. 12. Anderson, D. J. (1989) 3, 1-12. Three general classes of mechanisms are consistent with our 13. Jessell, T. M. & Melton, D. A. (1992) Cell 68, 257-270. results. In the first, the radial glial differentiation factors would 14. Green, J. B. A. (1994) Cell 77, 317-320. induce expression of phenotypic markers for radial glia. In a 15. Gray, G. E. & Sanes, J. R. (1992) Development (Cambridge, U.K) second model, although our factor may specify expression of 114, 271-283. 16. Bronner-Fraser, M. (1993) Curr. Opin. Genet. Dev. 3, 641-647. the embryonic phenotype, as development proceeds expres- 17. Hunter, K. E., Sporn, M. B. & Davies, A. M. (1993) Glia 7, sion of an antagonist of this factor would permit expression of 203-211. the astroglial phenotype. Third, our factors could provide an 18. Laemmli, E. K. (1970) Nature (London) 227, 680-685. inhibitory signal, which prevents astroglial maturation, like the 19. Towbin, H., Staehelin, T. & Gordon, J. (1979) Proc. Natl. Acad. role played by activin in repressing neural induction in early Sci. USA 76, 4350-4354. Xenopus embryos (31), with the presence of this (inhibitory) 20. Gao, W. Q. & Hatten, M. E. (1993) Science 260, 367-369. factor maintaining the embryonic radial phenotype for the 21. Misson, J.-P., Takahashi, T. & Caviness, V. S., Jr., (1991) Glia 4, duration of development and neuronal layer formation. 138-148. The present evidence suggests that the activity we identified 22. Cameron, R. S. & Rakic, P. (1991) Glia 4, 124-137. is not mimicked by any of a range of characterized growth 23. Gressens, P., Richelme, C., Kadhim, H. J., Gadisseux, J.-F. & Evrard, P. (1992) Biol. Neonate 61, 4-24. factors already recognized as glial growth regulators (for 24. Witte, 0. N. (1990) Cell 63, 5-6. reviews see refs. 13 and 14). Preliminary biochemical analysis 25. Roberts, A. B. & Sporn, M. (1984) in The Retinoids, eds. Sporn, indicates that the embryonic signals are protein and in the M., Roberts, A. B. & Goodman, M. (Academic, Orlando, FL), molecular size range 50-60 kDa. Molecular characterization pp. 209-286. and cloning of the inducing activity will be required to further 26. Mudge, A. W. (1993) Curr. Biol. 3, 361-364. examine its role(s) in the developing CNS. 27. Shain, W., Forman, D. S., Madelian, V. & Turner, J. N. (1987) J. In conclusion, our findings are consistent with a general role Cell Biol. 105, 2307-2314. for extrinsic signals inducing CNS neural differentiation and 28. Emmett, C. J., Lawrence, J. M., Raisman, G. & Seeley, P. J. with a regulatory role for diffusible signals in glial develop- (1991) J. Comp. Neurol. 310, 330-341. ment (11, 32). However, they call into question traditional 29. Lindsay, R. M. & Raisman, G. (1984) 12, 513-530. 30. Shah, N. M., Marchionni, M. A., Isaacs, I., Stroobant, P. & views that astroglial maturation follows a progressive restric- Anderson, D. J. (1994) Cell 77, 349-360. tion of cell potential (reviewed in refs. 21 and 33). We have 31. Hemmati-Brivanlou, A. & Melton, D. A. (1994) Cell 77,273-281. shown that astroglial transformation, previously viewed as 32. Jessell, T. M. & Schacher, S. (1991) in Principles of Neural being irreversible, is in fact a bidirectional pathway, with Science, eds. Kandel, E. R., Schwartz, J. H. & Jessell, T. M. astroglia being able to express either the mature phenotype or (Appleton & Lange, Norwalk, CT), pp. 887-907. the embryonic phenotype depending on the availability of 33. Goldman, J. E. & Vaysse, P. J.-J. (1991) Glia 4, 149-156. Downloaded by guest on September 27, 2021