A novel role for thyroid hormone, glucocorticoids and retinoic acid in timing oligodendrocyte development

The thyroid hormones, thyroxine (T4) and triiodothyronine (T3), have profound effects on the growth and development of many vertebrate tissues (Schwartz, 1983; Legrand, 1986; Oppenheimer, 1991). The effects of thyroid hormones on the development of oligodendrocytes, which make myelin in the mammalian central nervous system (CNS), typify those occurring in many other developing tissues. Hyperthyroidism accelerates the deposition of myelin whereas hypothyroidism delays it (Walters and Morell, 1981; Legrand, 1986; Dussault and Ruel, 1987). Thyroid hormones increase both the number of oligodendrocytes that develop in culture, as well as the amount of myelin that they produce (Almazan et al., 1985; Koper et al., 1986; Warringa et al., 1987). Although thyroid hormones have been shown to promote the synthesis of myelin-specific proteins by a direct action on oligodendro cytes, which are known to express thyroid hormone receptors (Koper et al., 1986; Dussault and Ruel, 1987; Yusta et al., 1988; Sarlieve et al., 1989; Puymirat, 1992), it is not known how these hormones accelerate myelination or increase the number of oligodendrocytes produced in culture.

Oligodendrocytes are postmitotic cells, but the precursor cells (also termed O-2A progenitor cells) that give rise to them proliferate extensively (Gard and Pfeiffer, 1990; Hardy and Reynolds, 1991). Although oligodendrocyte precursor cells will proliferate extensively in culture in response to signals from astrocytes (Noble and Murray, 1984) or to a combination of platelet-derived growth factor (PDGF), neurotrophin-3 (NT3) and insulin-like growth factor I (IGF-1) (Barres et al., 1994), they will not divide indefinitely: they have an intrinsic timing mechanism, or clock, that limits their number of divisions to about 8, after which the cells become unresponsive to mitogens, withdraw from the cell cycle and differentiate more or less synchronously into oligodendrocytes (Raff et al., 1985; Temple and Raff, 1986; Barres et al., 1994). Although, in the presence of mitogens in vitro, the progeny of an individual precursor cell tend to stop dividing and differentiate at around the same time, the population of precursor cells isolated from a 1-week-old optic nerve is heterogeneous in terms of prolif erative capacity: some cells divide 8 times before differentiat ing, while others divide fewer times or not at all (Temple and Raff, 1986), probably reflecting differences in the number of times the cells divided in vivo prior to their isolation.

This clock mechanism ensures that oligodendrocytes are generated even in the presence of plateau levels of mitogens and is thought to help regulate the timing of appearance of oligodendrocytes during development (Raff et al., 1988). In the absence of mitogens, oligodendrocyte precursor cells in culture differentiate prematurely (Raff et al., 1983). In either the presence or absence of mitogens, the differentiation of the precursor cells into oligodendrocytes has been thought to occur constitutively and not to require specific extracellular signals (Raff et al., 1983).

We show here that, in the presence of mitogens, purified oligodendrocyte precursor cells growing in serum-free medium fail to differentiate into oligodendrocytes unless there is thyroid hormone, glucocorticoid, or retinoic acid in the medium. These signalling molecules are not required for the survival, proliferation or differentiation of oligodendrocyte lineage cells. Instead, they induce oligodendrocyte precursor cells to stop dividing and, as a consequence, differentiate: in their absence, the precursor cells seem to divide indefinitely in response to mitogens. Experimentally induced hyperthy roidism accelerates the appearance of oligodendrocytes in the developing rat optic nerve, suggesting that thyroid hormone, and possibly glucocorticoids and retinoic acid as well, may control the timing of oligodendrocyte differentiation in vivo.

Sprague/Dawley (SID) rats were obtained from the breeding colony of University College, London. Oligodendrocyte precursor cells from the optic nerves of postnatal day 8 (P8) rats were purified to greater than 99.95% homogeneity by sequential immunopanning as described pre viously (Barres et al., 1992, 1993b). The cells were plated at a clonal density onto poly-D-lysine (PDL)-coated tissue culture dishes (60 mm, Falcon) in 2.5 ml of serum-free Dulbecco’s Modified Eagle’s medium (DMEM) containing bovine insulin (5 μ g/ml), human transferrin (100 μ g/ml), bovine serum albumin (100 μ g/ml), progesterone (60 ng/ml), putrescine (16 μ g/ml), sodium selenite (40 ng/ml), triiodothyronine (T3, 30 ng/ml) and PDGF (IO ng/ml), NT-3 (I ng/ml), high insulin (5 μ g/ml), ciliary neurotrophic factor (CNTF; 10 ng/ml) and forskolin (10 μ M). In some experiments, as described in the text, the triiodothyro nine was omitted. Recombinant human PDGF-AA was obtained from Peprotech (NJ). Recombinant mouse NT-3 and rat CNTF were gifts from Y. Barde and M. Sendtner, respectively. The other reagents were purchased from Sigma. All of the factors were used at concentrations that were on the plateau of their dose-response curves (Barres et al., 1993b). CNTF and forskolin were included in the medium to promote the long-term survival of oligodendrocytes and their precursors but did not promote proliferation or differentiation (Barres et al., 1993a,b; Louis et al., 1993; and our unpublished observations).

The numbers of cells plated were titrated to achieve about 50 clones per dish. The cultures were fed with fresh medium and growth factors (50% volume replacement) every 4 days and the number of cells per clone and their identity were tabulated. Cells were identified as either oligodendrocytes or precursor cells by their characteristic morpholo gies (Fig. 1), as previously described (Temple and Raff, 1985, 1986). The absence of precursor clones that contained a single cell suggests that there was little net migration between clones.

To label cells in S phase in vitro, BrdU (10 μ M; Boehringer Mannheim), which is incorporated into replicating DNA, was added to the cultures 1 to 24 hours prior to staining. After fixation with 4% paraformaldehyde for 90 seconds at room temperature and a 15 minute incubation in 50% goat serum containing I% BSA and 100 mM L-lysine to block non-specific binding, cells were surface-stained either with monoclonal anti-GC antibody (Ranscht et al., 1982; super natant used at 1:1) followed by fluorescein-coupled goat anti-mouse IgG3 (1:100) or with A2B5 antibody (Eisenbarth et al., 1979; super natant diluted I: I) followed by fluorescein-coupled goat anti-mouse IgM (u-chain specific). Cells were postfixed in 70% ethanol at -20°C for 10 minutes, incubated in 2 N HCI for 10 minutes to denature the nuclear DNA, followed by 0.1 M sodium borate, pH 8.5 for 10 minutes. The cells were then incubated in 50% goat serum contain ing 0.4% Triton X-100 for 30 minutes and labelled with monoclonal anti-BrdU antibody (ascites, 1:100; Magaud et al., 1988), followed by rhodamine-coupled goat anti-mouse IgG1 (I: 100). In some experi ments, cells were stained with rabbit anti-GFAP antiserum (diluted 1:100; Dako); in this case, the cells were fixed with acid-alcohol for IO minutes at -20°C and blocked with goat serum as above. The anti GFAP antibodies were detected with fluorescein-coupled goat anti rabbit IgG (H+L chain specific). The coverslips were mounted in Citifluor Mounting medium (City University, London) on glass slides, sealed with nail varnish and examined with a Zeiss Universal fluo rescence microscope.

Glial cell types were identified by their characteristic morphologies and antigenic phenotypes: astrocytes are labeled by anti-GFAP antiserum, oligodendrocyte precursor cells by A2B5 antibody (Raff et al., 1983) and oligodendrocytes by anti-GC antibody (Ranscht et al., 1982).

Thyroid hormone receptors were labelled with rabbit antisera raised to peptide fragments specific to each receptor subtype (Leehan et al., 1993). A rabbit antiserum that specifically detects the a-1 thyroid hormone receptor was obtained from Affiniti Reagents.

Polyclonal rabbit antisera were raised in rabbits by immunization with peptides corresponding to TR-specific sequences conjugated to keyhole limpet hemocyanin. One TRP2-specific antiserum, raised against amino acids (aa) 132-148 of rat TRP2 (Hodin et al., 1989), has been previ ously described (Leehan et al., 1993). A second TRP2-specific antiserum, raised against aa 97-114, was also used. Other anti-TR peptide sera used included the following: pan-P, raised against a portion of the hinge region common to TRPl and TRP2 (aa 248-265 ofTRP2); pan-a, raised against a portion of the amino terminus common to TRal and a2 (aa 14-34 of both; Lazar et al., 1988); and al-specific, raised against amino acids 424-444 of the unique carboxyl terminus of TRa2 (Lazar et al., 1988). The specificity of each antiserum was confirmed by testing its ability to appropriately supershift TR-DNA complexes in gel shift assays (not shown) and to immunoprecipitate TRs synthesized in vitro (Leehan et al., 1993). The addition of about 1000 molar excess of the peptide to which the antiserum was raised blocked the immuno histochemical staining in all cases.

The MTT survival assay was performed as described by Mosmann (1983). Approximately 5,000 purified oligodendrocytes or their precursor cells were plated in triplicate in 96-well Falcon plates in 100 μ I of the serum-free medium described above. MTT (3-(4,5dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide; Sigma) was dissolved in phosphate-buffered saline (PBS) at 5 mg/ml and steril ized by passage through a Millipore filter (0.22 μ m). This stock solution was added to the cultures (I :9) at 37°C for 2 hours. Viable cells with active mitochondria cleave the tetrazolium ring into a visible dark blue formazan reaction product. The viable and dead cells in each we11 were counted in an inverted microscope using bright field phase microscopy.

Newborn rat pups were treated daily with 3 μ g thyroxine (T4), or vehicle only, delivered subcutaneously as previously described (Patel et al., 1979; Walters and Morell, 1981). Injec tions were given on PO, PI and P2, and the animals were killed by decapitation on P3.

Oligodendrocyte precursor cells were purified from P8 postnatal rat optic nerve cell suspensions to greater than 99.95% purity by sequential immunopanning (Barres et al., 1992). The purified cells were cultured at clonal density in a serum-free medium that contained transferrin, progesterone, putrescine, selenium, tri-iodothyronine (T3), albumin and plateau concentrations of three growth factors (see Materials and Methods), PDGF, NT-3 and insulin (at a concentration suf ficient to activate IGF-1 receptors) (Barres et al., 1992, 1993b; Materials and Methods). The number of cells per clone and the predominant cell type in each clone were scored every 4 days. Oligodendrocytes and their precursor cells were easily distinguished by their characteristic morphology (Fig. I; Raff et al., 1983; Temple and Raff, 1986).

Under these culture conditions, individual precursor cells proliferated to form clones in which the cells more or less syn chronously differentiated into oligodendrocytes (Barres et al., 1994). After 4 to I 6 days of culture, some clones consisted primarily of precursor cells, while others were primarily oligo dendrocytes (Figs I, 2B,D). When both T3 and T4 thyroid hormones were omitted from the medium, however, few cells with the morphology of oligodendrocytes developed (Fig. 2A,C); the addition of either T3 (30 ng/ml) or T4 (30 ng/ml) promoted the development of oligodendrocytes (not shown). Similarly, cells that expressed galactocerebroside (GC), a gly colipid specifically expressed by oligodendrocytes but not their precursor cells, failed to develop in cultures grown at normal density (5,000 cells per 12 mm coverslip) in the absence of T3 or T4 (Fig. 3). The concentration of T3 that half-maximally induced oligodendrocytes (EDso) was 300 pg/ml (about 450 pM; Fig. 3).

To determine whether thyroid hormones act as survival factors for oligodendrocytes, we purified oligodendrocytes from P8 optic nerves and cultured them for 4 days in the presence or absence of T3 (30 ng/ml), prior to assessing their survival by the MTT assay (Mosman, 1983; Barres et al., 1992; Materials and Methods). The percentage of cells that survived was not decreased when thyroid hormones were omitted: no T3: 86.3±3.4; T3 (30 ng/ml): 84.6±5.1 (means ± s.e.m., n=3). To determine whether thyroid hormones are required for the survival of newly formed oligodendrocytes, we assessed their survival after 4 days of culture at clonal density in the presence of the survival factors high insulin, CNTF and forskolin, but in the absence of mitogens to force the precursor cells to dif ferentiate prematurely. The percentage of dead cells did not differ in the presence (23.2±1.7) or absence (24.3±1.5) of T3 (means ± s.e.m., n=3).

Thyroid hormones were clearly not required for cell division, as single precursor cells underwent clonal expansion in the absence of T3 and T4 (Fig. 2A,C). The cells appeared to divide indefinitely in the absence of thyroid hormone: by 16 days in vitro, clones as large as 4000 precursor cells were observed (Fig. 2C). By 16 days of culture, the average clone size was 31.2±5.1 (n=I0I) in the presence of T3 but was in the absence of mitogens to force the precursor cells to differentiate prematurely. The percentage of dead cells did not differ in the presence (23.2±1.7) or absence (24.3±1.5) of T3 (means ± s.e.m., n=3).

Thyroid hormones were clearly not required for cell division, as single precursor cells underwent clonal expansion in the absence of T3 and T4 (Fig. 2A,C). The cells appeared to divide indefinitely in the absence of thyroid hormone: by 16 days in vitro, clones as large as 4000 precursor cells were observed (Fig. 2C). By 16 days of culture, the average clone size was 31.2±5.1 (n=I0I) in the presence of T3 but was

501.9±74 (n=108) in the absence ofT3. Moreover, after 3 days in normal density cultures (5,000 cells per 12 mm coverslip), a higher percentage of cells synthesized DNA in the absence of thyroid hormone than in its presence: the percentage of cells that incorporated BrdU over 9 hours was 37.6±1.5 in the absence of T3 and 20.4±1.4 in the presence of T3 (30 ng/ml) (means ± s.e.m., n=3). The decreased rate of DNA synthesis in the presence of T3 was accounted for both by the ability of T3 to induce oligodendrocyte differentiation and by its ability to slow the cell cycle: during the first 4 days of culture the average cell cycle time was 24.7±0.6 (n=61) hours in the absence of T3 and 29.0±0.9 (n=53) hours in the presence of T3 (means ± s.e.m.).

To check whether thyroid hormone was necessary for the precursor cells to differentiate into oligodendrocytes, we cultured the cells at normal density in the absence of both mitogens and thyroid hormones: more than 98% of the cells differentiated into GC+ oligodendrocytes within 48 hours. Similarly, at clonal density, nearly all of the precursor cells dif ferentiated into cells with the morphology of oligodendrocytes within 48 hours, in the absence of both mitogens and thyroid hormones. There was no obvious difference in the morphology or rate of differentiation in the presence or absence of thyroid hormone when mitogens were not present.

Thus thyroid hormone is required for a precursor cell to dif ferentiate into an oligodendrocyte in the presence of mitogens but not in their absence, suggesting that thyroid hormone allows the timing mechanism in the precursor cell to stop pro liferation, which in tum initiates differentiation.

To determine whether thyroid hormone is required for the intrinsic timing mechanism to run, we cultured purified precursor cells in the absence of thyroid hormone for 8 days and then added T3 (30 ng/ml) for 4 days. As shown in Table 1, the percentage of oligodendrocyte clones in the cultures treated with T3 only from day 9 to 12 was about 95% of the percentage of oligodendrocyte clones in the cultures treated with T3 for the entire 12 days.

This finding suggests that the timing mechanism consists of at least two components: a counting component that is T3-inde pendent and an effector component that is T3-dependent. The T3-dependence of the effector mechanism, taken together with the observation that clones of precursor cells synchronously dif ferentiate into oligodendrocytes, raises the possibility that the counter may operate by increasing a cell’s sensitivity to thyroid hormone with time or successive cell divisions. If this is the case, then the extracellular concentration of T3 should control the time when the clock is triggered, thereby controlling how many divisions a precursor cell can make. As shown in Fig. 4 and Table 2, the average oligodendrocyte clone size was a function of the T3 concentration: after 12 days of culture the average oligodendrocyte clone size was about 9 cells per clone when the T3 was 3 ng/ml but increased by 5-fold to about 46 cells per clone at a T3 concentration of 0.03 ng/ml (Table 2). Similarly, as the T3 concentration was decreased, the percent age of clones that were still composed of precursor cells after 12 days of culture increased (Fig. 4). The EDso for these effects was about 300 pg/ml, similar to the EDso for the induction of oligodendrocyte differentiation (see Fig. 3).

To examine further whether T3 sensitivity increases with increasing number of divisions, we compared precursor cells purified from Pl4 optic nerve with precursor cells purified from Pl optic nerves, as Pl4 cells would be expected to have divided on average more times than the Pl cells. The cells were cultured at normal density in the presence or absence of T3 and after 1 day DNA synthesis was assessed by BrdU incorpora tion. As shown in Table 3, T3 inhibited the proliferation of P14 cells by about 3-fold but had little inhibiting effect on the Pl cells. Even in the absence of T3, more than twice as many Pl than Pl4 cells incorporated BrdU. Thus, PJ4 cells are more sensitive to thyroid hormone than Pl cells and, even in its absence, they seem to divide more slowly than Pl cells.

In vertebrates, two genes code for homologous thyroid hormone receptor proteins, which are called T3R-a and T3Rβ. Alternative RNA splicing produces two variants of each of these to generate four types of receptors: α-1, α-2, β-1 and β-2 (Hodin et al., 1990; Brent et al., 1991; Chin, 1991). All four forms are expressed in the brain (Lazar et al., 1988; Strait et al., 1990; Forrest et al., 1991; Bradley et al., 1992; Puymirat, 1992; Leehan et al., 1992). Each has a DNA-binding region and, except for a-2, a thyroid-hormone-binding region; a-2 does not bind thyroid hormone and may act as a dominant negative suppressor of the other thyroid hormone receptors (Koenig et al., 1989; Katz and Lazar, 1993). Thyroid hormone receptors have previously been shown to be expressed by oligodendrocytes (op. cit.), but it is not known whether they are expressed by oligodendrocyte precursor cells.

We analyzed purified oligodendrocytes and their precursor cells by indirect immunofluorescence, using a battery of antisera against T3Rs (see Materials and Methods). The nucleus in all oligodendrocytes and precursor cells was stained by the pan-P antiserum (Fig. 5), suggesting that all of these cells express βT3Rs. As oligodendrocytes were not stained by the β-2-specific antiserum, they presumably express mainly P l receptors. By contrast, some precursor cells were stained by the β-2-specific antiserum; some precursor clones were brightly stained while others were only lightly stained or unstained, but all cells within a clone stained similarly (Fig. SC). The proportion of precursor clones that were brightly stained by the β-2-specific antiserum was several fold higher at P14 than at PI, consistent with the possibility that the con centration of β-2 receptors in precursor cells increases with time or successive cell divisions, which in principle could account for the increasing sensitivity of the cells to thyroid hormone. The nucleus of some precursor cells was also stained by an α-2-specific antiserum; in this case the intensity of staining varied within clones, with larger, G2-sized nuclei generally staining most intensely. Two different antibodies specific for the aI receptor did not label the nuclei of precursor cells but, in contrast to the other three antisera, they also failed to label the nuclei of GH3 cells, which are known to express α-1 receptor mRNA, although at lower levels than other thyroid receptors (Lazar et al., 1988).

Oligodendrocytes first begin to appear in small numbers in the optic nerve during the first few postnatal days, but are not produced at an appreciable rate until about PS (Miller et al., 1985; Barres et al., 1992), around the time that thyroid hormone levels rise in the developing brain (Same), 1968; Dussault and Ruel, 1987; Larsen, 1989; Puymirat, 1992). To test the possi bility that the levels of thyroid hormone normally regulate the number of oligodendrocytes that develop in the optic nerve, we treated neonatal rats for three days with thyroid hormone (see Methods) and examined the optic nerves at P3. Whereas the average total number of cells per nerve and the average per centage of oligodendrocyte precursor cells or astrocytes per nerve were not changed, the number of oligodendrocytes per P3 nerve was increased more than five-fold (Table 4).

Retinoic acid (both 9-cis and all-trans, 10 nM) and dexam ethasone ( I μ M) each closely mimicked the effects of thyroid hormone on oligodendrocyte precursor cell division and differ entiation in clonal density cultures (Fig. 6), whereas vitamin D3 (100 nM), which activates a homologous nuclear receptor protein, had no effect. Like thyroid hormone, both retinoic acid and dexamethasone slowed the average cell-cycle time and allowed oligodendrocyte development (Fig. 7). The average cell cycle time of P8 precursor cells over the first 4 days in culture was 26.7±0.7 hours (n=77) in the ab ence of T3, retinoic acid and dexamethasone, but wa 36.0± 1.8 hour (n=55) in all-trans retinoic acid (10 nM) and 33.4±1.9 hours (n=62) in dexametha one ( 1 μ M) (means ± s.e.m.).

Retinoic acid receptors, glucocorticoid receptor and thyroid hormone receptors hare the ability to inhibit the activity of the transcription activator API (see Discussion). If they induce oligodendrocyte differentiation by inhibiting API, then the phorbol ester TPA, which stimulates API activity (Karin, 1990; Angel and Karin, 1991), should antagonize this effect. To test this possibility, we studied the ability of T3 to induce oligodendrocytes in the presence ofTPA. Control experiments showed that in the absence of mitogens TPA did not antago nize oligodendrocyte differentiation, either in the pre ence or absence of T3. In the presence of mitogens, however, TPA completely antagonized the ability of T3 to induce oligoden drocytes, as mea ured either by morphology or by GC expression (Fig. 7; Table 5).

Thyroid hormones, glucocorticoids and retinoic acid are hydrophobic signalling molecule, all of which have profound effects on the growth and development of many vertebrate tissue (Schwartz, 1983; Legrand, 1986; Oppenheimer, 1991; Hackney et al., 1970; DeLuca, 1991; Ragsdale and Brocke, 1991). The receptors for these molecules belong to a homolo gous family of ligand-activated transcription factors (Jensen, 1991). Although each type of receptor regulate different ets of genes, they all share the ability, when bound to their cognate ligands, to inhibit the activity of the transcription activator AP I (Yang-Yen et al., 1990; Danielsen, 1991; Diamond et al., 1990; Nicholson et al., 1990; Brent et al., 1991; Schule and Evans, 1991; Zhang et al., 1991; Oppenheimer, 1991; Lazar, 1993; Kerppola et al., 1993; Saatcioglu et al., 1993), but it is not known whether this property is important in their role as regulators of normal development.

In principle, the e hydrophobic signalling molecules could act on either po tmitotic cells or precursor cell to promote differentiation. Many previous studies have demonstrated that they promote differentiation of postrnitotic cells by regulating gene expression. In the ca e of the oligodendrocyte lineage, for example, thyroid hormones enhance the synthesi of myelin specific proteins by a direct action on the oligodendrocytes themselves (Dussault and Ruel, 1987; Puymirat, 1992). Our findings demonstrate that oligodendrocyte precursor cells are also an important target of action, as thyroid hormone, gluco corticoids and retinoic acid each promote the differentiation of purified oligodendrocyte precursor cells into oligodendrocyte . As expected, these precursor cells express receptors for thyroid hormone (this study), glucocorticoids (Bohn et al., 1991) and retinoic acid (8. A. Barre, M. Raff and P. Chambon, unpub lished observations).

The differentiation of oligodendrocyte precursor cells into oligodendrocytes is thought to be constitutive (Raff et al., 1983) and its timing controlled by the action of some mechanism in the precursor cells that limits the time or number of times that the cell divides (Raff et al., 1985; Temple and Raff, 1986; Raff et al., 1988). Our findings suggest that the timing mechanism can be separated into two components: a counting component and an effector component that stops cell proliferation. Whereas the counting component has been shown to be driven by mitogens that bind to cell-surface receptors, the effector component depends on hydrophobic sig nalling molecules, such as thyroid hormone, glucocorticoids and retinoic acid, that bind to intracellular receptors. In the absence of mitogens, oligodendrocyte precursor cells stop dividing and differentiate prematurely into oligodendrocytes. In the absence of the hydrophobic signals, the precursor cells seem to divide indefinitely and fail to differentiate into oligo dendrocytes; the counting mechanism continues to operate, however, as when thyroid hormone is added to cultures that have been deprived of hydrophobic signals for 8 days, the cells rapidly differentiate and the percentage of oligodendrocyte clones nearly catches up to that in cultures where thyroid hormone was present for the entire time. Thus even in the absence of the hydrophobic signals, it seems that the proper ties of a precursor cell change with time or with successive divisions.

Oligodendrocyte differentiation is associated with withdrawal from the cell cycle (Raff et al., 1983), but the relationship between the two processes has been uncertain (Temple and 1986). The timing mechanism might primarily control the onset of oligodendrocyte differentiation, with the cessation of proliferation following as a consequence. Alternatively, it might primarily control the cessation of cell proliferation, with oligodendrocyte differentiation following as a consequence. The second possibility has been favored as it would most simply explain why oligodendrocyte precursor cells differen tiate prematurely when deprived of mitogens (Temple and Raff, 1986). Our observations suggest that the primary function of the timing mechanism is to stop cell proliferation: a hydrophobic signalling molecule is required in the presence of mitogens for oligodendrocyte precursor cells to stop dividing, but is not required in the absence of mitogens for the precursor cells to differentiate into oligodendrocytes. Thus, the hydrophobic signal appears to be required to activate the effector component of the timing mechanism, which forces the cells to withdraw from the cell cycle and differentiation then follows as a consequence.

Our findings raise the possibility that the timing of oligoden drocyte differentiation in vivo may depend on the timing of appearance of the extracellular signals that are required for the oligodendrocyte precursor cells to stop dividing. Thyroid hormone is a strong candidate for such a signal. It is not available to the developing rat foetus until the thyroid gland becomes active around birth (Samel, 1968; Dussault and Ruel, 1987; Puymirat, 1992), which is when oligodendrocytes first appear in the rat CNS (Abney et al., 1981; Miller et al., 1985). Similarly, the human thyroid gland becomes active at about the 12th week of gestation and CNS myelination is already present by the 14th week (Friede, 1989). Moreover, hyper thyroidism accelerates myelination and hypothyroidism delays it (Walters and Morrell, 1981; Dussault and Ruel, 1987; Almazan et al., 1985; Tosic et al., 1992) and we have shown here that hyperthyroidism greatly accelerates the appearance of oligodendrocytes in the developing rat optic nerve. It thus seems likely that limiting amounts of thyroid hormone normally regulate the timing of oligodendrocyte development.

Glucocorticoids and retinoic acid also enable mitogen-stim ulated oligodendrocyte precursor cells to stop dividing in vitro. Do they normally help the cells to do so in vivo? Glu cocorticoids enhance myelination in vitro (Dawson and Kernes, 1979; Stephens and Pieringer, 1984; Almazan et al., 1986; Kumar et al., 1989), and adrenalectomy and glucocor ticoid treatment of postnatal rats alter myelination in vivo, although the interpretation of the latter results is complicated by the ability of glucocorticoids to decrease thyroid hormone and growth hormone levels (Bohn and Friedrich, 1982; Preston and McMorris, 1984; Meyer and Fairman, 1985). It is not yet known whether physiological levels of corticosteroids can activate the effector component of the timing mechanism in oligodendrocyte precursor cells. Cultured astrocytes_and Muller glial cells, but not neurons, can synthesize retinoic acid from retinol (Wuarin et al., 1991; Edwards et al., 1992; Maden and Holder, 1992), raising the possibility that retinoic acid could also play a part in the timing of oligodendrocyte differ entiation in vivo.

How does the effector mechanism induce oligodendrocyte precursor cells to withdraw from the cell cycle? Our findings that retinoic acid, glucocorticoids and thyroid hormone can activate the effector mechanism may provide an important clue to the answer: the receptors for these hormones share the ability to inhibit the activity of AP-1 transcription factors, which are formed by the dimerization of members of the Jun and Fos families of proteins and help mediate the proliferative response to growth factors in many cell types (Ransone and Verma, 1990; Angel and Karin, 1991; Riabowol et al., 1992; Johnson et al., 1992, 1993; Wang et al., 1992). It is possible, therefore, that the effector mechanism induces mitogenic unre sponsiveness by inhibiting AP-1 activity. The finding that the phorbol ester TPA, which enhances the activity of AP-1 (Karin, 1990; Angel and Karin, 1991), antagonizes the ability of thyroid hormone to inhibit oligodendrocyte precursor cell proliferation is consistent with this possibility. Furthermore, recent studies have shown that thyroid hormone, retinoic acid and glucocorticoids can inhibit AP-1 activity under physio logical conditions (Salbert et al., 1993).

How does the oligodendrocyte precursor cell count or keep time? It has previously been suggested that it might count cell divisions by the dilution with each division of a stable molecule required for proliferation in response to mitogens (Temple and Raff, 1986). A variant of this model, suggested by our present findings, is that AP-1 activity may decrease with successive divisions or with time, either because the concen trations of c-Fos or c-Jun themselves decrease, or because of changes in factors that heterodimerize with c-Fos or c-Jun or post-translationally modify them. A progressive decrease in AP-1 activity would account for the progressive increase in the sensitivity of oligodendrocyte precursor cells to the hydropho bic signals. If AP-1 activity regulates the rate of cell cycle pro gression (e.g. Schilthuis et al., 1993), this model could also account for the slowing of the cell cycle that we observe, both in the presence of T3 and with increasing number of cell divisions in the absence of T3 (Fig. 2 and our unpublished observations).

A simple hypothetical model of how the timing mechanism might work is presented in Fig. 8. We propose that the cell counts by decreasing AP-1 activity with time or with every cell division. When the activity falls low enough, the inhibitory hydrophobic signals can then reduce it below the level required to keep the cell dividing and division halts. In the absence of these signals, but in the presence of mitogens, the residual AP1 activity is sufficient to keep the cell dividing indefinitely. In the absence of mitogens, AP-1 activity drops below threshold and the cell stops dividing, even in the absence of hydropho bic signals.

Thyroid hormones play a crucial role in coordinating the development of mammalian tissues, including brain, lung, muscle, bone and erythroid cells (Legrand, 1986; Oppenheimer, 1991), much as they coordinate metamorphosis in tadpoles (Gilbert and Frieden, 1981). Myoblasts and erythroid progenitor cells appear to use counting mechanisms similar to oligodendrocyte precursor cells (Dainak et al., 1978; Allen and Dexter, 1982; Quinn et al., 1985). The present findings thus raise the possi bility that timing mechanisms in precursor cells in various tissues may depend on thyroid hormones, which may thereby coordinate differentiation in the developing animal.

A thyroid-hormone-triggered timing mechanism may operate in neuroblasts in the external granule layer (EGL) of the cerebellum. Hyperthyroidism causes early disappearance of the EGL, while hypothyroidism causes the EGL to persist (Legrand, 1986). Such observations led Hamburgh et al. (1971) to propose that physiological levels of thyroid hormone normally act as a’time clock’ in the developing cerebellum, not by triggering differentiation in an all-or-none fashion, but by controlling the rate at which cells complete a fixed number of cell cycles (also see Lauder, 1977).

Many mammalian cells, such as fibroblasts, can divide a limited number of times before they’senesce’. The operation of the timing mechanism in oligodendrocyte precursor cells may share at least three features in common with fibroblast senescence: (1) fibroblasts divide more slowly with time and successive cell divisions; (2) extracellular factors such as fetal calf serum promote fibroblast senescence (Loo et al., 1987; Cristofalo and Pignolo, 1993); and (3) AP-1 activity and c-fos mRNA levels decrease by as much as 20-fold in senescent cells (Campisi and Seshadri, 1990; Irving et al., 1992; Sikora et al., 1992; Riabowol, 1992; Riabowol et al., 1992; Cristofalo and Pignolo, 1993; Dice, 1993). Thus the counting mechanism that times the differentiation of oligodendrocytes and possibly many other cell types as well, may be usefully viewed as a rapid form of cell senescence.

We thank Jeremy Brockes for much helpful advice and guidance, Yves Barde for generously providing NT-3 and for a helpful sugges tion and Mark Shipman for his expert assistance with confocal microscopy. We also thank the British Multiple Sclerosis Society for their generous financial support.