Identification of an adult-specific glial progenitor cell

The development of a multicellular organism is in many ways dependent upon the rapid division of tissue specific precursor cells which, through differentiation and sometimes through further cell division, generate the specialized cell types that make up the various tissues of the body. In contrast, in most tissues of adult animals, generation of new differentiated cells from precursor cells only occurs as a part of maintaining homeostasis, e.g. to compensate for naturally occurring cell death, or in response to injury. The mechanisms that control proliferation of precursor cells in developing and adult animals may therefore be very different. Moreover, although similar cell types may be produced in adult animals and during embryogenesis, fetal and adult precursor cells and their differentiated progeny may differ in at least some properties. For example, the generation of fetal versus adult forms of haemoglobin by erythrocytes appears to be determined at the level of the haematopoetic stem cells (Wood et al. 1985, 1988), suggesting that fetal and adult haematopoetic stem cells differ from each other and are specialized in ways that reflect the differing physiological requirements of fetal and adult animals.

Despite the obvious importance of understanding the properties of the precursor cells that function during development and in the adult animal, relatively little is known about these cells or about the mechanisms that control their division and differentiation. Attempts to understand the biological properties of precursor cells in general, and of adult-specific precursor cells in particular, have been greatly hampered by the inability to identify unambiguously most precursor cells or to grow these cells in conditions that would allow a detailed analysis of their properties. Thus, it has not yet even been possible to distinguish directly and unambiguously adult from perinatal precursor cells either during development or in vitro.

One of the few cell types in which the properties of fetal and adult cells with precursor function can be directly compared is the oligodendrocyte-type-2 astro cyte (O–2A) progenitor cell of the rat optic nerve (Raff et al. 1983b). These bipotential progenitor cells give rise to two of the major cell types of the optic nerve: oligodendrocytes, the myelin-forming cells of the cen tral nervous system (CNS) and type-2 astrocytes, which enwrap bare axons in astrocytic processes at the nodal gaps between adjacent myelin sheaths (ffrench-Constant & Raff, 1986b). O–2A progenitor cells can be readily identified and grown in tissue culture, where they can be induced to divide and differentiate in response to either specific cell-cell interactions or defined mitogens or differentiation-inducing factors (Raff et al. 1983b; Noble & Murray, 1984; Hughes & Raff, 1987; Noble et al. 1988; Raff et al. 1988; Richardson et al. 1988; Hughes et al. 1988; Lillien et al. 1988). Although it is the 0–2A progenitor cells from the optic nerves of perinatal rats which have been the most extensively examined, studies in our laboratory (G. Wolswijk, unpublished observations), and indepen dently by french-Constant & Raff (1986a), have demonstrated that 0–2A progenitor cells can be isolated from the optic n_erves of adult rats.

In the course of our studies on O–2A progenitor cells of the optic nerve of adult rats, it has become clear that these cells are fundamentally different from O–2A progenitor cells isolated from optic nerves of perinatal animals. In this paper, we present evidence that 0–2A progenitors isolated from the optic nerves of adult rats (O–2Aadult progenitor cells) differ from their perinatal counterparts (O–2Aperinatal progenitor cells) in antigenic phenotype, morphology, rate of migration, cell cycle time and time course of differentiation into either oligodendrocytes or type-2 astrocytes in vitro. The properties of the O-2^Aadutt progenitor cells may make these cells ideally suited for the needs of the adult CNS.

Monolayers of purified type-1 astrocytes were prepared by modifications of previously described methods (Noble et al. 1984; Noble & Murray, 1984). Dissociated cells from the cerebral cortex of newborn or 1-day-old Sprague Dawley rats were plated at a density of 2 brains per poly-L-lysine-coated (PLL; Sigma: 175000Mr, 20 μg ml-) NUNC tissue culture flask (85 cm² surface area) and grown in Dulbecco’s modified Eagle’s medium containing 4·5 gl–¹ glucose (DMEM; Imperial Laboratories) and supplemented with 10 % heat-inacti vated fetal calf serum (FCS; Imperial Laboratories), 2 mm glutamine (Sigma) and 25 μg ml-gentamicin (Flow Labora tories) (DMEM + 10%FCS). Once a confluent monolayer of flat cells had been formed (usually after 7 –10 days), the layer of process-bearing cells growing on top of the flat cells was removed by shaking the flasks overnight on a rotary platform (100 revs min–¹) at 37°C. Cultures were allowed to recover for 1 day and then treated with 20 μM-cytosine arabinoside (AraC; Sigma) for two 48 h periods to eliminate most of the faster-dividing non-astrocytes. Cells were then removed from the flask by washing with Ca²+-, Mg²+-free DMEM (DMEMCMF) followed by an incubation in 0·54mm-EDTA (Sigma) in DMEM-CMF for 10–15 min and addition of trypsin (bovine pancreas, 7pe III; Sigma) to a final concentration of 300 i.u. ml- . After the cells had detached from the surface of the flask, the trypsinization was stopped by adding 1 ml per 10ml of EDTA/trypsin solution of SBTI-DNAse (5200i.u. ml–¹ soybean trypsin inhibitor (Sigma), 74i.u. m1–¹ bovine pancreas DNAse I (Sigma) and 3·0mgm1–¹ bovine serum albumin (BSA, fraction V; Sigma) in DMEM]. Cells were then gently triturated through a 10 ml blow-out pipette, spun for 10 min at 500 g, resuspended and treated in suspension with agarose-absorbed rabbit complement (Buxted Ltd; diluted 1: 10) plus monoclonal antibody A2B5 (ascites, 1: 1000; Eisenbarth et al. 1979), monoclonal antibody O4 (concentrated hybridoma supernatant, 1:100; Sommer & Schachner, 1981) and anti-galactocerebroside (GalC) mono clonal antibody (hybridoma supernatant, 1:10; Ranscht et al. 1982). Treatment with these antibodies and complement eliminated A2B5+ neurones, oligodendrocyte-type-2 astro cyte progenitor cells (A2B5+ /04+), oligodendrocytes (A2B5+ /04+/Gale+) and type-2 astrocytes (A2B5+) {Abney et al. 1983; Raff et al. 1983a,b; Sommer et al. in preparation}. The remaining cells were plated onto PLL coated glass coverslips (Chance No. 0, 13 mm diameter) at a density of 20 000 cells per coverslip or into PLL-coated 85 cm² Nunc tissue culture flasks at a density of 1 x10⁶ cells per flask and fed with 50 % DMEM + 10 %FCS that had been conditioned by purified cultures of type-1 astrocytes for 1-2 days and 50 % fresh DMEM + 10 %FCS (as in Noble & Murray, 1984). Cultures were kept in a humidified incubator at 37°C under 92·5 % air and 7·5 % CO₂ atmosphere. After 3-4 days, cells were irradiated (20 Gray) and maintained in the presence of D MEM + 10 % FCS prior to use. At this stage, over 95 % of the cells in these cultures had the antigenic phenotype of type-1 astrocytes (GFAP+ A2B5-) as determined by indirect immunofluorescence (Raff et al. 1983a). The major contaminants of the cultures were GFAP-, fibronectin+, fibroblast like cells.

Optic nerve cells of newborn to 7-day-old and adult Sprague-Dawley rats were isolated using modifications described elsewhere (Raff et al. 1983a, Noble & Murray, 1984). The majority of the adult animals that were used were older than 1 year and none were younger than 8 months. Optic nerves were dissected from just behind the eyes to the chiasm, minced with a fresh scalpel and incubated in Leibowitz L-15 medium (Flow Laboratories) containing 333 i. u. ml–¹ col lageqase (Cooper Biomedical). After 60–120 min at 37°C, an equal volume of 30000i.u. ml–¹ trypsin in DMEM-CMF was added and incubation at 37°C was continued for another 20 min. The suspension was centrifuged (5 min, 500 g), the supernatant removed and the tissue was then resuspended and further incubated in a solution of 15000i.u. m1–¹ trypsin and 0·27mm-EDTA in DMEM-CMF for 20min. The digestion with trypsin was terminated by addition of an equal volume of SBTI-DNAse, followed by an incubation at 37°C for 10–20 min and a short centrifugation (500 g, 5 min). The supernatant was replaced with the medium appropriate for any given experiment, followed by trituration of the tissue through a 5 ml blow-out pipette and 25 G and 27 G hypodermic needles. 100 μl of the resulting cell suspension was plated onto either PLL-coated coverslips or coverslips coated with monolayers of type-1 astrocytes (as in Noble & Murray, 1984). The cells derived from 1 pair of adult optic nerves were usually plated onto 8–15 coverslips, while optic nerve cells derived from newborn to 7-day-old animals were plated at a density of 3000-4000 cells per coverslip. When the adult optic nerve cells were plated onto PLL-coated coverslips, cells were kept in DMEM+ 10%FCS for 3-4h, rinsed with L-15 medium (to remove debris and myelin] and placed in Falcon 6-well trays (3–4 coverslips per well). Cells were either fed with defined medium (DMEM containing 4·5 g1–¹ glucose and supplemented with 25μgm1–¹ gentamicin, 2mm-gluta mine, 0·234i.u. ml–¹ bovine pancreas insulin (Sigma), lOO μg ml-¹ human transferrin (Sigma), 0·0286% (v/v) BSA pathocyte (Miles Laboratories, Inc), 0·2 μM-progesterone (Sigma), 0·10 μM-putrescine (Sigma), 0·45 μM-L-thyroxine (Sigma), 0·224 μM-selenium (Sigma) and 0·49 μM 3,3’,5-triiodo-L-thyronine (Sigma); modified from Bottenstein & Sato, 1979] (DMEM-BS), or fed with 50 % fresh and 50 % type-I astrocyte-conditioned DMEM-BS supplemented with 0–5 %FCS (DMEM-BS/0·5 %FCS) or fed with DMEM + 10 %FCS. Optic nerve cells that were plated directly onto type-1 astrocyte monolayers were kept overnight on raised platforms before being rinsed with L-15 medium and placed in 50 % conditioned and 50 % fresh DMEM-BS/ 0·5 %FCS. In some experiments, adult optic nerve cells were plated into 85 cm² Nunc tissue culture flasks that had been coated previously with 1 × 10⁶ purified type-1 astrocytes and fed with 50 % fresh and 50 % type-1 astrocyte-conditioned DMEM-BS/0·5 %FCS. Two thirds of the culture medium was changed 2–3 times a week. Live cells were photographed through a Leitz inverted microscope using Kodak Tri-X-pan 125 ASA films.

To identify the cell types of interest, cultures were immunolabelled with various antibodies. Cells were incubated at room temperature in 40–50 μ1 of each antibody solution for 20–30 min, followed by an incubation in the appropriate rhodamine- or fluorescein-conjugated second antibody solution. Antibodies were diluted in Hanks’ balanced salt solution (Imperial Laboratories) containing 5 % heat-inactivated bovine donor calf serum (Imperial Laboratories) and 0-02 mm-Hepes (Sigma) (HBSS-5 %DCS). After every incu bation, cells were washed several times with HBSS-5 %DCS. Surface antigens (A2B5, 04, NSP-4 and GalC) were visualized on living cells, intracellular antigens [i.e. fibronectin and all classes of intermediate filaments including glial fibrillary acidic protein (GFAP) and vimentin] after fixation in methanol at -20°C for 10 min. Cultures were immunolabelled with mouse IgM monoclonal antibody A2B5 (ascites, 1: 1000), mouse lgM monoclonals 01 and 04 (concentrated hybridoma supernatants, 1:100; Sommer & Schachner, 1981), mouse IgM monoclonal antibody NSP-4 (hybridoma supernatant, 1:1; Rougon et al. 1983), mouse IgG3 anti-GaIC monoclonal antibody (hybridoma supernatant, 1:10), rabbit anti-fibronec tin antiserum (1:1000; Price & Hynes, 1985), rabbit anti GFAP antiserum (1:1000; Pruss, 1979), mouse IgGl anti vimentin intermediate filament monoclonal antibody (used at a concentration of 4 μg ml-¹; Boehringer Mannheim), and the mouse IgGl monoclonal antibody called intermediate filament antigen (IFA; fresh undiluted hybridoma supernatant; Pruss et al. 1981). All rhodamine- and fluorescein-conjugated class-specific anti-mouse immunoglobulin (lg) second antibodies were purchased from Southern Biotechnology Associ ates, Inc., USA and diluted 1:100 prior to use. The binding of A2B5, NSP-4, 01 and 04 antibodies was detected with either fluorescein- or rhodamine-conjugated goat anti-mouse lgM (anti-IgM-FI/Rd), anti-GalC antibody with goat anti-mouse IgG3-Fl or goat anti-mouse IgG3-Rd, anti-vimentin and IFA monoclonal antibodies with goat anti-mouse lgGl-Fl and the binding of the rabbit anti-GFAP and anti-fibronectin anti bodies with sheep anti-rabbit lg-Fl (Wellcome, 1: 100). In stainings involving both A2B5 and 04 {both mouse IgM monoclonals), cells were sequentially labelled with A2B5 and anti-IgM-Rd, followed by 04 and anti-IgM-Fl or vice versa. The first staining method revealed cells that expressed 04 and not A2B5 [i.e. mature oligodendrocytes (Sommer et al. in preparation)], the latter method, cells that were A2B5+O4−_ In some experiments, cultures growing on PLL-coated cover slips were incubated simultaneously with the A2B5 and the anti-GalC antibodies, followed by anti-IgM-Rd and anti IgG3-Fl. Cells were then fixed and labelled sequentially with anti-GFAP antibodies and anti-rabbit lg-FI. In such experiments, cells that were surface fluorescein+ (i.e. were Gale+) were easily distinguishable from cells that were internally labelled by the fluorescein conjugate (i.e. were GFAP+). After the immunolabelling, coverslips were washed, mounted in a drop of glycerol containing 22 mm-1,4-diazobicyclo [2,2,2] octane (Sigma) to prevent fading (Johnson et al. 1982; Davidson & Goodwin, 1983), sealed with transparent nail varnish and examined in a Zeiss Universal microscope equipped with phase contrast, epifluorescence, and rhodamine and fluorescein optics and equipped for photography. All immunolabelled cells in cultures of adult optic nerve cells were counted; in cultures of newborn optic nerves cells, at least 200 immunolabelled cells were scored per coverslip in each experiment. Immunolabelled cells were photographed using Kodak-Tri-X-pan 400 ASA films, while their corresponding phase-contrast image was recorded using llford XPl 400 films.

For in vitro autoradiography experiments, dissociated optic nerve cells were plated onto monolayers of type-1 astrocytes, grown for 2 days in DMEM-BS/0·5 %FCS and exposed to 2 μCi m1-¹ [3H]thymidine (Amersham, 2 μCi mmol–¹) for 20h. After the radiolabelling period, cells were immunola belled, fixed, dehydrated in ethanol, air-dried, and mounted face up on microscope slides (with Gurr fluoromount mountant, BDH Chemicals Ltd). Slides were then coated with Ilford L4 autoradiographic emulsion, which had been diluted with an equal volume of distilled water and kept at 45°C, and exposed for 2 days in a light-tight box at 4°C. After develop ment with Ilford contrast FF and fixation, cultures were covered with a coverslip and sealed with nail varnish. Cells with more than 5 silver grains above their nucleus were scored as cells that had taken up [3H]thymidine during the pulse.

After 1 day in culture, coverslips with adult optic nerve cells growing on monolayers of type-1 astrocytes were placed onto raised platforms and incubated in 100 μl culture medium containing agarose-absorbed rabbit complement plus either A2B5, O4 and anti-GalC monoclonal or O4 and anti-GaIC antibodies. After 30 min at 37°C, cultures were washed with L-15 and placed back into the original culture medium. Effectiveness of the complement kill was checked by immunolabelling with appropriate antibodies directly following the incubation with complement and antibodies.

Dissociated cells derived from one pair of adult optic nerves were plated into 85 cm² Nunc tissue culture flask coated with l×10 purified type-1 astrocytes and were grown in DMEM BS/0·5 %FCS. From day 2 onwards, cultures were screened for well-isolated colonies of O-2A progenitor-like cells. When colonies were found, their location was marked and their expansion was then followed daily. As the flask contained fewer than 20 colonies of O-2A progenitor-like cells, it is likely that each colony was derived from a single cell. The appearance of multipolar oligodendrocyte-like cells in these colonies confirmed that the colonies that were followed contained O-2A progenitor cells.

For migration studies, dissociated optic nerve cells of adult rats were plated in the centre of PLL-coated 6cm Nunc Petri dishes of which the edges were coated with 250 000 purified type-I astrocytes, as described previously (Small et al. 1987; Noble et al. 1988). After 2h, cultures were rinsed with L-15 medium and fed with 50 % fresh, and 50 % type-1 astrocyte-conditioned, DMEM-BS/0·5 %PCS. 3 or 6 days later, fields were selected that contained 1 or 2 cells with an O-2A progenitor-like morphology. Cultures were filmed on Olympus inverted microscopes adapted for time-lapse microcine matography. The stage was maintained at 37°C by means of a fan heater. Photographs were recorded every 300 s; 16 mm Kodak Infocapture AHU microfilm 1454 films were used. Migration distances were determined with an Hewlett Pack ard 9874 A digitizer. For cell cycle studies, dissociated adult optic nerve cells were plated into 85 cm² Nunc tissue culture flask precoated with type-1 astrocytes as described in the preceding section. A colony of adult optic-nerve-derived O-2A progenitor-like cells that had been followed initially on a daily basis for 11 days, as described in the preceding section, was followed for a further 10 days using time-lapse microcine matography.

Cultures prepared from embryonic and postnatal rat optic nerve contain members from the oligodendrocyte-type-2 astrocyte (O-2A) lineage and the type-1 astro cyte lineage, the two glial lineages of the rat optic nerve (see Fig. 1 for summary).

O-2A progenitor cells, the precursors to both oligo-dendrocytes and type-2 astrocytes, have a characteristic bipolar morphology (Temple & Raff, 1986; Small et al. 1987) and have been identified antigenically as cells that are labelled with the monoclonal antibodies A2B5 and NSP-4, but that express neither galactocerebroside [GalC: a specific marker for oligodendrocytes (Raff et al. 1978)]·nor glial fibrillary acidic protein [GFAP: a protein specifically expressed by astrocytes (Bignami et al. 1972)] (Raff et al. 1983b; Raff et al. 1985; ffrench Constant & Raff, 1986b). In addition, O-2A progeni tors derived from the optic nerves of perinatal rats contain vimentin intermediate filaments (IFs) (Raff et al. 1984b). O-2A progenitor cells also start expressing an antigen recognized by a monoclonal antibody, called 04 (Sommer & Schachner, 1981), shortly before birth; 7 days after birth more than 95 % of the O-2A progenitor cells are O4⁺in vivo (Sommer et al. in preparation). When O-2A progenitors differentiate into oligodendro-cytes, they acquire a multipolar morphology, express GalC and gradually lose vimentin filaments and the ability to bind the monoclonal antibody A2B5. In contrast, O-2A progenitor cells differentiating into type-2 astrocytes remain A2B5+NSP-4⁺ and acquire GFAP filaments. Type-2 astrocytes, like oligodendro-cytes, also have a multipolar morphology.

Type-1 astrocytes have a fibroblast-like morphology and label with the Ran-2 monoclonal antibody (Bartlett et al. 1981), express GFAP and do not label with the A2B5 and NSP-4 antibodies. Their precursor cells, which are also flat cells, are Ran-2⁺GFAP⁻ A2B5⁻ NSP-4⁻ (Abney et al. 1981, ffrench-eonstant & Raff, 1986b).

The first question we addressed was whether we could use antibodies to distinguish adult O-2A progenitor cells from their counterparts in perinatal optic nerve cultures. Although these two populations express many similar antigens, we found differences in the expression of intermediate filament proteins and in the expression of a cell surface antigen recognized by the 04 mono clonal antibody (Sommer & Schachner, 1981).

O-2A progenitor cells isolated from the optic nerves of adult rats, like their perinatal counterparts, are A2B5+ and do not express GalC or GFAP unless induced to differentiate into oligodendrocytes or type-2 astrocytes, respectively (ffrench-constant & Raff, 1986a; G. Wolswijk, unpublished observations). Per pair of adult nerves, we obtained between 300 and 700 cells (average: 505 ± 117) that were labelled with the A2B5 monoclonal antibody, although not so heavily as O-2A progenitor cells derived from perinatal optic nerves, but which were not labelled with anti-GalC and anti-GFAP antibodies. Immunolabelling of such cells with the NSP-4 antibody revealed that these cells, like their counterparts in perinatal optic nerve cultures (ffrench-eonstant & Raff, 1986b), were also NSP-4⁺ (not shown). Approximately 11 % of the O-2A lineage cells isolated from the optic nerves of adult rats ex pressed GalC and none were A2B5+GFAP+ type-2 astrocytes after 1 day of in vitro growth.

When cultures of adult optic nerve cells growing on type-1 astrocyte monolayers were immunolabelled with the 04 monoclonal antibody and with anti-GalC anti bodies, the number of O4⁺ Gale-GFAP-cells was very similar to that of the number of A2B5⁺GalC^- GFAP-cells after 1 day in culture, suggesting that many O-2A progenitor cells were also O4⁺. Direct O-2A progenitors found after 1 day of in vitro growth were O4⁺ (Table 1; Fig. 2).

In contrast to the virtually ubiquitous expression of vimentin IFs in both O4⁻ and O4⁺ O-2A progenitor cells isolated from perinatal rats (Wolswijk & Noble, in preparation), none of the O4⁺ cells in cultures of adult optic nerve cells, growing on glass coverslips and immunolabelled 1 day after plating, contained vimentin IFs (Fig. 2; Table 1). Again unlike perinatal O-2A progenitor cells, the few A2B5⁺O4^-cells derived from double-labelling analysis confirmed that >97 % of the adult optic nerve were vimentin. In addition, the adult optic nerve-derived O-2A progenitor cells could not be labelled with the anti-intermediate filament antigen (IFA) monoclonal antibody that recognizes a determinant shared by all classes of intermediate filaments (Pruss et al. 1981), suggesting that the adult O-2A progenitor cells express no intermediate filament proteins (Fig. 2). In contrast, virtually all O-2A progenitors from newborn to postnatal day 7 (P7) optic nerves were IFA⁺in vitro (Fig. 3).

In addition to revealing antigenic differences between adult and perinatal O-2A progenitors, the immuno labelling experiments also indicated that these cells differed morphologically from each other.

As reported previously, O-2A progenitor cells derived from the optic nerves of perinatal rats have a characteristic bipolar morphology when grown in vitro (Temple & Raff, 1986; Small et al. 1987; see also Fig. 3 and Fig. 4). Because of their morphology, they are easily distinguishable from oligodendrocytes and type-2 astrocytes, which both have a multipolar morphology, and from type-1 astrocytes, which have a fibroblast-like morphology (Raff et al. l983a,b). Virtually all of the bipolar cells found in cultures of newborn and P7 optic nerve cells growing on type-1 astrocytes or in type-1 astrocyte-conditioned medium (Astro-CM) appear to be O-2A pro_genitor cells (Temple & Raff, 1986; Small et al. 1987). Most of these bipolar cells are O4⁻ and all are A2B5⁺in vitro (Sommer et al. in preparation). The perinatal O-2A progenitors that are O4⁺ when grown in similar conditions are multipolar with several thin processes arranged in a radially symmetric manner (Sommer et al. in preparation).

In contrast to the bipolar morphology of A2B5⁺04- O-2A progenitor cells found in cultures of perinatal optic nerve cells, over 65 % of the O-2A progenitor cells isolated from the optic nerves of adult animals had one major branched process plus several smaller thinner processes when grown on type-1 astrocytes or in Astra-CM for 3 days and could be described as unipolar cells (Fig. 4). Those adult-derived O-2A progenitor cells that were not unipolar were generally multipolar and resembled the multipolar O4⁺ O-2A progenitor cells found in cultures of perinatal rat optic nerves.

In previous studies, the ability to identify unambiguously O-2A progenitor cells derived from perinatal animals allowed us to determine that proliferation of these cells is stimulated in vitro by factor(s) secreted by type-1 astrocytes (Noble & Murray, 1984; Raff et al. 1985). When O-2A progenitor cells derived from perinatal optic nerves were grown on monolayers of type-1 astrocytes or in the presence of Astro-CM, their numbers doubled approximately every day, corresponding with an average cell cycle time of <24 h (Noble & Murray, 1984; Temple & Raff, 1986; Noble et al. 1988).

When adult optic nerve cells were grown on type-1 astrocytes in a chemically defined medium sup plemented with 0·5 % fetal calf serum (DMEM-BS/ 0·5 %FCS), the number of O-2A lineage cells increased much more slowly than has been seen for populations of perinatal O-2A lineage cells. In a representative experiment with adult optic-nerve-derived cells shown in Fig. 5, the number of O-2A lineage cells (mostly O-2A progenitor cells and oligodendrocytes) increased just over fourfold between day 1 and day 7, corresponding to a doubling time for the total O-2A lineage population of 70 h (average for 3 separate experiments: 64 ± 5 h).

In addition, only about 40 % of adult O-2A progenitor cells took up radiolabelled thymidine in a 20 h pulse (applied after 2 days in vitro) as compared with 75 % of perinatal O-2A progenitors in newborn optic nerve cultures (Table 2).

Two lines of evidence indicated that the lower labelling index of adult optic-nerve-derived O-2A progeni tors was due to a long cycle time. (1) The vast majority of the O-2A progenitor cells proliferated when grown on monolayers of type-1 astrocytes. When we determined the proportion of the adult O-2A progenitor cells present on day 1 that had formed colonies of 2 or more cells after 14 days, we found that 65–85 % of the O-2A progenitors had proliferated and produced colonies ranging in size from 2 to 36 cells. (2) When the expansion of individual colonies of O-2A progenitorlike cells was followed on a daily basis, the colonies initially expanded with an average doubling time of 62 ± 10h (n = 14) (Fig. 6). With the appearance of multipolar cells (after 7-10 days in vitro), which most likely were oligodendrocytes, the doubling time of a colony decreased gradually. Time-lapse microcinematography experiments involving one of these colonies (colony N in Fig. 6) showed that cells with the unipolar morphology of adult optic-nerve-derived O-2A pro genitor cells divided every 65 ± 18 h (n = 24) when followed from day 11 to day 21. In contrast, bipolar O-2A progenitor cells present in cultures of perinatal optic nerves divided every 18 ± 4 h when grown under similar conditions (Noble et al. 1988).

A further distinction between perinatal and O-2A progenitor cells is in the antigens expressed by dividing O-2A progenitors. While the majority of the proliferating perinatal O-2A progenitor cells were A2B5⁺O4-in vitro, the majority of the adult O-2A progenitors expressed 04 labelling. In the autoradiography exper iments described in an earlier section, we found that about 92 % of the O-2A progenitor cells that took up (3H]thymidine in optic nerve cell cultures of newborn rats were A2B5⁺O4^-, while about 97 % of the prolifer ating O-2A progenitor cells in cultures of adult optic nerve were O4⁺ (Table 2). Furthermore, in contrast to the widespread uptake of [3H]thymidine by O4⁺ O-2A progenitor cells derived from adult optic nerves, less than 10 % of the O4⁺ O-2A progenitors cells had done so in cultures derived from the optic nerves of newborn rats.

As reported previously in studies on perinatal optic nerve cells (Noble & Murray, 1984), Gale⁺ oligoden drocytes present in both the newborn and the adult optic nerve cell cultures did not synthesize DNA in response to factor(s) produced by type-1 astrocytes (Table 2).

In our previous studies, we found that perinatal O-2A progenitor cells were very motile cells, capable of migrating in vitro at speeds of up to 100 μm⁻h ¹. The motile behaviour of the perinatal O-2A progenitors seems to be associated with a requirement for migration of this lineage into the optic nerve during development (Small et al. 1987). However, such rapid migration might not be necessary in adult tissues, and we were therefore interested whether adult and perinatal O-2A progenitors also differed in their capacity for migration.

Time-lapse microcinematography studies on adult O-2A progenitor cells demonstrated that cells with the morphology of adult O-2A progenitors were motile cells which migrated with an average speed of 4·3±0·7μmh⁻¹ (n=21) when grown on PLL-coated coverslips in Astra-CM, less than 20 % of the 21·4±1·6μmh⁻¹ speed of perinatal O-2A progenitor cells observed under similar conditions. When adult O-2A progenitor cells were grown on type-1 astrocyte monolayers, these cells migrated at a speed of only 2·3 ± 1·0 μm h⁻¹(n= 61) The unipolar adult O-2A progenitor cells migrated in the direction of their large process and, following a division, the progeny cells moved away in opposite directions. Adoption of a multipolar oligodendrocyte-like morphology by such cells coincided with the loss of their migratory behaviour, as has been found in previous studies on O-2A progenitor cells derived from the optic nerves of perina tal rats (Small et al. 1987).

Over 95 % of the bipotential progenitor cells of P7 optic nerves differentiate into oligodendrocytes or type-2 astrocytes within 3 days of plating in either chemically defined medium (DMEM-BS) or medium containing 10 % FCS (DMEM ⁺ 10 %FCS), respectively (Raff et al. 1983b). In contrast, in cultures of adult optic nerve cells plated onto PLL-coated coverslips and kept in either DMEM-BS (to allow differentiation along the oligodendrocyte pathway of development) or DMEM⁺ 10 %FCS (to induce differentiation along the type-2 astrocyte pathway), more than 65 % of the O-2A progenitor cells were still GalC^- and GFAP^- after 3 days (Fig. 7). With increasing length of culture time in vitro, the number of A2B5*GalC⁻GFAP⁻ O-2A progenitor genitor cells in the cultures decreased and the number of GalC⁺ oligodendrocytes or A2B5⁺GFAP⁺ type-2 astrocytes increased when the adult optic nerve cells were grown in, respectively, DMEM-BS or DMEM ⁺ 10 %FeS (Fig. 7). During the culture period, the number of O-2A lineage cells kept in DMEM-BS remained constant, while the number of such cells grown in DMEM ⁺ 10 %FeS increased significantly, suggesting that adult O-2A progenitor cells were stimu lated to divide in response to growth factor(s) present in FeS (data not shown).

It took 4-5 days for 50 % of the O-2A progenitor cells in cultures of adult optic nerve cells to become GalC⁺ oligodendrocytes or to become A2B5⁺GFAP⁺ type-2 astrocytes (Fig. 7). In contrast, 50 % of O-2A progenitor cells from the optic nerves of perinatal rats differentiate in less than 2 days (Raff et al. 1983b).

There were several other aspects in which differen tiation of adult O-2A progenitor cells was dissimilar from differentiation of perinatal O-2A progenitors. Although oligodendrocytes generated by adult O-2A progenitor cells, like oligodendrocytes generated by their perinatal counterparts (Raff et al. 1983b), lose A2B5 labelling with further maturation, these cells do so more slowly as compared with their counterparts in perinatal optic nerve cultures (G. Wolswijk, unpublished observations). In addition, when adult optic nerve cells were grown on coverslips in DMEM-BS for 7 days and immunolabelled with a different anti-GalC antibody, called 01 [an IgM antibody, used to allow double labelling with anti-vimentin antibodies (which were IgGl antibodies); Sommer & Schachner, 1981], and anti-vimentin antibodies, none of the 01⁺ cells (i.e. GalC⁺ cells) contained vimentin IFs, suggesting that the oligodendrocytes generated by differentiating O-2A progenitor cells isolated from adult optic nerves do not acquire vimentin IFs in vitro. In contrast, oligodendrocytes generated by O-2A progenitor cells in cultures of perinatal optic nerve cells, and probably also in vivo, initially contain vimentin IFs, which are lost with further maturation (Raff et al. 1984b).

There were also some fundamental differences between the differentiation of perinatal and adult O-2A progenitors into type-2 astrocytes. Unlike perinatal O-2A progenitor cells, which are vimentin⁺, adult O-2A progenitors acquire vimentin whilst differentiating into type-2 astrocytes. After 7 days of growth in DMEM⁺10%FeS, a stage at which over 95% of the A2B5⁺ cells expressed GFAP (Fig. 7), we found that over 65 % of such cells were also labelled with anti vimentin antibodies. Fes thus seems to induce the expression of both GFAP and vimentin in adult O-2A progenitor cells, and it seems likely that these cells had acquired first GF AP filaments and then vimentin IFs. Unlike the typical stellate morphology of type-2 astrocytes found in cultures of perinatal optic nerve cells grown in similar culture conditions, newly generated type-2 astrocytes in cultures of adult optic nerve cells generally had a more flattened morphology (data not shown).

Although a small proportion (<3 %) of the O-2A progenitor cells found in cultures of adult optic nerve were A2B5⁺O4-, and in this respect similar to perina tal O-2A progenitor cells, these 04-adult O-2A progenitors appeared to be more like adult progenitor cells than perinatal progenitors. (1) The adult A2B5⁺O4-adult optic nerve cells lacked vimentin IFs (Table 1), like O4⁺ adult progenitor cells but unlike perinatal progenitor cells. (2) The adult A2B5⁺O4-O-2A progenitors only generated small colonies of O-2A lineage cells over time courses where perinatal optic nerve-derived O-2A progenitor cells frequently genererated large colonies, suggesting that these cells divided at a rate similar to that of the O4⁺ O-2A progenitor cells in these adult-derived cultures. In these experiments, we first treated adult optic nerve cultures with 04 antibodies and complement to eliminate the O4⁺ adult O-2A progenitors. When the remaining cells were grown on monolayers of type-1 astrocytes, small colonies of O-2A lineage cells developed, but only at a frequency of <5 % of that seen when O4⁺ adult O-2A progenitor cells had not been depleted. The colonies that developed from the A2B5⁺O4-adult O-2A pro-genitors were of a size consistent with the slow division of adult O-2A progenitors. For example, the largest colony found after 7 days consisted of ten O4⁺Galc-O-2A progenitors and four O4⁺Galc⁺ oligodendro cytes, as compared with an expected maximum size for 7 days of growth of ⩽64 cells for a colony derived from a perinatal O-2A progenitor cell (Temple & Raff, 1986). All other colonies contained less than 14 cells (average for 7 colonies= 6 cells). Indeed, it is noteworthy that we have never seen large colonies of O-2A lineage cells developing in cultures of adult optic. Thus, we find no indication that adult optic nerves contain any perinatal type O-2A progenitor cells.

When cultures of adult optic nerve cells growing on type-1 astrocytes were incubated with A2B5, 04 and anti-GalC antibodies and complement, no new O-2A lineage cells appeared in the cultures, even after 14 days of in vitro growth, indicating that, as with perinatal O-2A lineage cells (Noble & Murray, 1984; Raff et al. 1984a), O-2A lineage cells in the adult optic nerve cultures did not arise from A2B5-cells.

We have found that O-2A progenitor cells isolated from the optic nerves of adult rats differ from their perinatal counterparts in a number of aspects (see Table 3 for summary). O-2A progenitor cells isolated from the optic nerves of perinatal rats can be labelled with the A2B5 and NSP-4 monoclonal antibodies (Raff et al. 1983b, 1985; ffrench-Constant & Raff, 1986b) and with anti-vimentin antibodies (Raff et al. 1984b). O-2A progenitor cells derived from perinatal rats first start expressing an antigen recognized by a monoclonal antibody, called 04 (Sommer & Schachner, 1981), shortly before birth (Sommer et al. in preparation). In vitro, O-2A progenitor cells derived from perinatal optic nerves are A2B5⁺04-, have a bipolar mor phology (Temple & Raff, 1986), are motile cells with an average rate of migration of 21·4 μm h-¹ (Small et al. 1987), have a cell cycle of about 18h (Noble & Murray,1984; Temple & Raff, 1986; Noble et al. 1988), and can be induced to differentiate into either oligodendrocytes or type-2 astrocytes in a time span of less than 3 days (Raff et al. 1983b, 1985). In contrast, O-2A progenitor cells derived from the optic nerves of adult rats lack intermediate filaments of any known class [like mature oligodendrocytes (Kachar et al. 1986)], are A2B5⁺, NSP-4⁺ and O4⁺, and have a unipolar morphology. In addition, the O-2A progenitor cells isolated from adult optic nerves have an average cell cycle of 65 h, migrate much more slowly (at an average rate of 4·3 μm h-¹) and require between 2 and 2·5 times longer to differen tiate into either oligodendrocytes or type-2 astrocytes in vitro.

The differences between the O-2A progenitor cells derived from the optic nerves of adult rats and their perinatal counterparts are striking enough that we propose the terminology ‘O-2^Aadutt progenitor cell’ to represent the cell type isolated from the optic nerves of adult rats and ‘O–2^Aperinatal progenitor cell’ to represent the cell type derived from the optic nerves of perinatal rats. Our description of the O-2^Aadutt progenitor cell extends significantly the earlier report of ffrench-Constant & Raff (1986a), who also found O-2A progenitor cells in the optic nerves of adult rats but did not detect the multiple differences that we observed between these cells and their perinatal counterparts.

There is suggestive evidence that other tissues also contain populations of precursor cells with differing properties in the developing and mature organism. For example, when haematopoietic stem cells derived from the embryonic yolk sac or from adult bone marrow were injected into irradiated adult mice, the embryonic cells consistently generated larger spleen colonies than the adult-derived cells (Metcalf & Moore, 1971). In ad dition, spleen colonies generated by the adult haematopoietic stem cells contained fewer differentiated haema topoietic cell types as compared with colonies generated by fetal stem cells. A third difference between fetal and adult haematopoietic stem cells is that the erythroid cells generated by fetal stem cells produce the fetal form of haemoglobin, while erythroid cells generated by adult stem cells produce the adult form of haemoglobin (Wood et al. 1985, 1988).

Embryonic myoblasts and adult muscle satellite cells also appear to be significantly different from each other in their biological properties, although both can function as precursor cells to multinucleated skeletal muscle fibres. For example, adult muscle satellite cells express acetylcholine receptors at all stages of their development, but such receptors appear in embryonic myoblasts only at the onset of differentiation (Cossu et al. 1987). Furthermore, phorbol ester tumour promoters block differentiation of embryonic myoblasts, but have no effect on adult muscle satellite cells (Cossu et al. 1983). There are also further indications that myoblasts derived from early and late embryos differ in their abilities to generate fast, slow and mixed fast/slow muscle types (Miller & Stockdale, 1986; Schafer et al. 1987).

It is not known whether different properties of perinatal and adult precursors represent intrinsic properties of these cells or are the result of exposure to dissimilar microenvironments. In the case of O-2^Aadult and O-^2Aperinatal progenitor cells, we have separately seen that these cells express the properties described in the present paper even when growing in the same culture (Wolswijk & Noble, in preparation). Our studies thus suggest that the differences between perinatal and adult O-2A progenitor cells represent intrinsic properties of these cells.

It seems likely that the differences between fetal and adult progenitors are related to differing requirements of fetal and mature organisms. For example, the switch from fetal to adult haemoglobin is of clear functional importance. As fetal haemoglobin has a higher affinity for oxygen than adult haemoglobin, it allows the fetal haemoglobin to be oxygenated at the expense of the maternal adult haemoglobin in the placenta (Stryer, 1981). In contrast, adult haemoglobin has the advantage that it allows a more efficient exchange of carbon dioxide and oxygen in the lungs as compared to fetal haemoglobin (Stryer, 1981). Thus, the fetal and adult forms of haemoglobin appear to have properties that suit the needs of their environment. However, it is difficult to imagine the functional value of the differences between fetal and adult muscle precursor cells, or of the reduced abilities to generate spleen colonies and the restricted potential for differentiation of adult haematopoietic stem cells.

At least some of the differences between O-2^Apermatal and O-2A^adult progenitor cells appear to be clearly related to the differing needs of their particular en vironments. During perinatal development it is necess ary to generate rapidly large numbers of oligodendrocytes (and type-2 astrocytes), and the 18 h cell-cycle time of the ^{O–2Aperinatal} progenitor cells would be important in this process. In addition, it appears that ^{O–2Aperinatal} progenitors populate the optic nerve as a result of migration from a germinal zone in or near the optic chiasm (Small et al. 1987), and the capacity of ^O-2Apermatal progenitors for rapid migration would play a critical role in this migratory process. These features of the ^{O–2Aperinatal} progenitor cells would, however, be inappropriate in the adult nervous system where only small numbers of new oligodendrocytes and type-2 astrocytes are needed as part of the normal glial turnover (Kaplan & Hinds, 1980), where there is insufficient space to accommodate continually the num ber of cells required during embryogenesis and where extensive migration may not normally be necessary.

The slow rates of migration, division and time course of differentiation of O-2A^adult progenitor cells suggest that these cells may be less effective than perinatal progenitors in carrying out oligodendrocyte replacement in a demyelinating lesion. It is particulary interest ing in this regard that the recovery of children suffering from optic neuritis, a demyelinating disease of the optic nerve, usually tends to be more complete than in adults suffering from this condition (McDonald, 1983). More over, Wolf et al. (1986) have recently reported that, although cells can migrate out of both perinatal and adult optic nerve explants into explants of chemically demyelinated cerebella, where migrating cells produce myelin, the cells derived from the adult optic nerve explants (a) took a longer time to myelinate the cerebellar axons than their perinatal counterparts, and (b) did not migrate as far as cells migrating out of the perinatal optic nerve explants. The results of these in vivo and organ culture studies are thus consistent with, and are perhaps due to, the properties of the O-2^Aadutt progenitor cells.

In all tissues in which there are differences between perinatal and adult progenitor cells the initial times of appearance and developmental origins of the adult progenitors, and their relationship to their perinatal counterparts, are unknown. The inability to identify and distinguish unambiguously between perinatal and adult progenitors in the haematopoietic system, or between embryonic and adult muscle precursors, makes it unlikely that the development of adult-specific pro genitor cells in these tissues will be easily understood. In contrast, the differences between O-2^Aadutt and O–2^Aperinatal progenitor cells of the rat optic nerve are so distinct that it should prove possible to determine the time of first appearance of O-2^Aadutt progenitor cells during development and to determine eventually the origin of this new cell type.

We wish to thank Peter Riddle, Carol Gomm and Liz Parsons for help with the time-lapse microcinematography studies, Rhona Mirsky for the anti-IFA hybridoma cells, Barbara Hyams for the assistance with Fig. 1 and Damian Wren, Ian McDonald and Martin Raff for helpful discussions and critical comments on the manuscript. This work was supported by grants from the Multiple Sclerosis Society and Medical Research Council of Great Britain.