Growth factors and transcription factors in oligodendrocyte development

Oligodendrocytes, the myelinating cells of the central nervous system (CNS), arise during development from glial progenitor cells known as O-2A progenitors (Raff et al. 1983). In vivo, the time course of oligodendrocyte production is precisely controlled. In the embryonic rat optic nerve, 0-2A progenitors proliferate and do not differentiate; then, starting on the day of birth, some 0-2 A progenitors stop dividing and differentiate into oligodendrocytes while the others continue to divide. Proliferation and oligodendrocyte differentiation continue together in the nerve for several weeks after birth (Miller et al. 1985). However, when embryonic or postnatal optic nerve cells are dissociated and cultured in medium containing low concentrations (0.5 %) of fetal calf serum (FCS), nearly all the O-2A progenitors stop dividing straightaway and differentiate prematurely into oligodendrocytes (Raff et al. 1983; Raff et al. 1985). Thus, there must be environmental signals present in vivo, but missing in vitro, that are required to keep O-2A progenitors dividing and delay the onset of oligodendrocyte differentiation. We are trying to identify these environmental factors and understand how they regulate the timing of oligodendrocyte differentiation during development. Platelet-derived growth factor (PDGF), in concert with other polypeptide growth factors, probably plays an important role in this process (reviewed by Richardson et al. 1990; see below).

The earliest committed O-2A lineage cells probably originate in the germinal zones around the ventricles (subventricular zones) and subsequently migrate away from there, maturing as they go, to populate the rest of the CNS with oligodendrocytes (Small et al. 1987; Levine and Goldman, 1988a,b;Reynolds and Wilkin, 1988). To identify environmental factors that might influence the step-wise development of O-2A lineage cells in vivo and relate this to studies in vitro, we need to establish what polypeptide growth factors are available, and which of these the O-2A progenitors can respond to, at each stage of their progress from immature neuroepithelial cell to mature oligodendrocyte’. With this aim in mind, we have started to map the distribution of PDGF and its receptors in the developing CNS by in situ hybridization and immunohistochemistry; our early progress and conclusions are summarized below.

There is now strong evidence that platelet-derived growth factor (PDGF) plays an important role in controlling the proliferation and differentiation of O-2A progenitor cells during development. O-2A progenitor cells freshly isolated from newborn rat optic nerves possess cell-surface receptors for PDGF (Hart et al. 1989a; McKinnon et al. 1990), and PDGF is a potent mitogen for O-2A progenitors in vitro (Noble et al. 1988; Richardson et al. 1988). Furthermore, PDGF can recreate the normal timing of oligodendrocyte development in vitro (Raff et al. 1988). For example, if optic nerve cells from E17 rats are cultured in chemically defined medium containing PDGF, the O-2A progenitors continue to divide for some time, and oligodendrocytes do not appear until the fourth day in culture, corresponding to the day of birth in vivo (E21). If E18 optic nerve cells are cultured in the same way, the first oligodendrocytes appear on the third day in vitro, and so on. As PDGF mRNA (Richardson et al. 1988; Pringle et al. 1989) and protein (H. Mudhar and W. Richardson, unpublished) is present in the perinatal brain and optic nerve, and PDGF-like mitogenic activity can be detected in extracts of optic nerves (Raff et al. 1988), it seems likely that PDGF may be important for normal development of the O-2A lineage in vivo.

PDGF is a disulfide-linked dimer of A and B chains, with the structure AA, AB or BB, depending on its source (for a review of PDGF, see Heldin and Westermark, 1989). For example, PDGF from human platelets is a mixture of all three dimeric isoforms, of which PDGF-AB is the major species, while some human tumor cell lines synthesize mainly PDGF-AA. PDGF elicits its biological effects by binding to transmembrane receptors with extracellular ligand-binding domains and intracellular tyrosine kinase domains. The unoccupied receptors are monomeric and inactive, but PDGF-binding induces dimerization and activates their tyrosine kinase activity. There are two types of PDGF receptor subunits with different ligand specificities: the a-receptor subunit (PDGF-α R) binds both A and B chains of PDGF, while the /Preceptor subunit (PDGF-β R) binds only B chains. Thus, the response of a cell to PDGF depends on the relative numbers of α - and β-receptors that it expresses as well as the PDGF isoform(s) that it encounters (see Fig. 1). O-2A progenitor cells express predominantly PDGF-α R (Hart et al. 1989a; McKinnon et al. 1990), and consequently respond to all three dimeric isoforms of PDGF, although PDGF-AA is effective at lower concentrations than PDGF-BB because PDGF-α R has higher affinity for PDGF A chains than for B chains (Heldin et al. 1988). Consistent with this, we could detect mRNA encoding the PDGF A chain, but not the B chain, in the rat optic nerve (Pringle et al. 1989), suggesting that PDGF-AA might be the predominant PDGF isoform in the nerve.

If PDGF is responsible for stimulating O-2A progenitor proliferation during development, what causes the O-2A progenitors eventually to stop dividing and differentiate into oligodendrocytes, and what dictates the timing of this decision? An O-2A progenitor cell differentiates into an oligodendrocyte when it is cultured on its own in a microwell in defined medium (Temple and Raff, 1985), so oligodendrocyte differentiation seems to proceed by default when an O-2A progenitor is deprived of PDGF and other exogenous signals. It is unlikely, however, that oligodendrocyte differentiation is triggered in vivo by PDGF withdrawal. First, PDGF A chain mRNA is present in the brain and optic nerve from before birth into adulthood (Richardson et al. 1988; Pringle et al. 1989), suggesting that PDGF may be continuously available throughout life. Second, O-2A progenitors in optic nerve cell cultures do not proliferate indefinitely in vitro; they eventually stop dividing and differentiate into oligodendrocytes, no matter how much or how often PDGF is added to the culture medium (Raff et al. 1988). Moreover, in these cultures, dividing and differentiating O-2A progenitors coexist in the same dish. This suggests that the decision to differentiate is cell-autonomous and does not depend on timed signals from other cells. Thus, in order to understand how oligodendrocyte differentiation is timed, we need to know why proliferating O-2A progenitor cells eventually lose their ability to divide in response to PDGF. In principle, this loss of responsiveness to PDGF could be caused by a loss of PDGF receptors at the cell surface, a block in one of the intracellular second messenger systems that transduces the mitogenic signal, or a deficiency in part of the cell’s replication machinery. PDGF receptor loss does not seem to be the key, because newly-formed, post-mitotic oligodendrocytes seem to have numbers of PDGF receptors similar to the proliferating O-2A progenitors that gave rise to them (Hart et al. 1989a). PDGF receptors are eventually lost from maturing oligodendrocytes, but this is a consequence of differentiation, not the cause. The PDGF receptors on newly-formed oligodendrocytes are functional, and linked into at least part of the signal transduction apparatus, because PDGF stimulation of young oligodendrocytes in vitro leads to elevation of cytosolic Ca²⁺ (Hart et al. 1989b), and activation of the proto-oncogene products c-Jun and c-Fos in the nucleus (Hart et al. 1990 and unpublished data). Thus, it appears that the mitotic block that develops in an O-2A progenitor cell, causing it to drop out of division and differentiate into an oligodendrocyte, lies downstream of nuclear protooncogene activation or, alternatively, in an unidentified, parallel signalling pathway.

Progress in understanding the molecular events that initiate oligodendrocyte differentiation has been hampered by the difficulty in obtaining O-2A progenitors in sufficient numbers or purity for biochemical analysis. However, recent technical developments now allow us to to purify 0-2A progenitors by immunoselection (B. Barres, unpublished) and keep them proliferating in vitro with a combination of basic fibroblast growth factor (bFGF) and PDGF (Bôgler et al. 1990; see below). This has enabled us to examine the expression of transcription factors that might be involved in controlling the switch between proliferation and differentiation of O-2A progenitors in vitro. For example, we have found that mRNA encoding the POU transcription factor SCIP (Monuki et al. 1989), also known as Tst-1 (He et al. 1989) or Oct-6 (Suzuki et al. 1990), is highly expressed in O-2A progenitors that are kept proliferating in vitro with a combination of bFGF and PDGF, but declines to background levels within 6 h after oligodendrocyte differentiation is initiated by growth factor withdrawal (E. Collarini, R. Kuhn, E. Monuki, G. Lemke and W. Richardson, unpublished data; see Fig. 2). This down-regulation of SCIP mRNA is followed by a decline in SCIP protein over the next 24 h, and the subsequent appearance of myelin-specific products (Fig. 2). SCIP was previously shown to be expressed in proliferating Schwann cells, and to be down-regulated when they start to express myelin-specific genes (Monuki et al. 1989). There is in vitro evidence that SCIP is a repressor of myelin gene expression in proliferating Schwann cells (Monuki et al. 1990; He et al. 1991), and the same may be true in proliferating O-2A progenitors. It is not yet known whether a high level of SCIP is necessary or sufficient to keep O-2A progenitors dividing and inhibit oligodendrocyte differentiation; there is some evidence, however, that POU transcription factors may be associated with cell proliferation in other systems (see Scholer, 1991 for a review).

It has been known for some time that there is mitogenic activity for Schwann cells on the surfaces of PNS axons, although it is not known whether this activity is synthesized by the neurons themselves or deposited by other cells (see below). Nevertheless, it is an attractive idea that neurons in both the PNS and CNS might produce mitogens for the glial cells with which they physically interact. The possibility that axons might influence the development of the oligodendrocyte lineage has been tested in vivo (Privât et al. 1981; David et al. 1984). Newborn rat optic nerves, which carry axons from ganglion neurons in the retina to the brain, were transected just behind the eye, causing the axons in the nerve to degenerate rapidly. The number of oligodendrocytes and their progenitors present in the nerve stump one week later was reduced more than eight-fold compared to untransected nerves, whereas the number of type-1 astrocytes (which belong to a different cell lineage) was reduced less than two-fold (David et al. 1984). Surprisingly, despite the large decrease in the population of oligodendrocyte lineage cells, the mitotic index of these cells did not differ significantly between transected and untransected nerves. This was interpreted to mean that O-2A progenitors depend on axons, not for proliferation per se, but rather for survival. These experiments suggested that intact axons are not obligatory for mitogenic stimulation of O-2A progenitors in the optic nerve, suggesting instead that glial cells in the nerve might be the source of the mitogen(s). Support for this latter view came from the finding that cultured astrocytes from rat cerebral cortex, which resemble type-1 astrocytes in optic nerve cell cultures, secrete a mitogen for O-2A progenitors (Noble and Murray, 1984). This astrocyte-derived mitogen was subsequently shown to be a form of PDGF, possibly PDGF-AA (Richardson et al. 1988; Pringle et al. 1989). In support of a glial origin for PDGF in the optic nerve, we have recently been able to demonstrate immunostaining of glial processes in postnatal mouse and rat optic nerves, with antisera raised against the mouse PDGF A chain (H. Mudhar and W. Richardson, unpublished).

Recently it was reported that many neurons in the CNS and PNS also synthesize PDGF A and/or B chains (Yeh et al. 1991; Sasahara et al. 1991; see Fig. 3), so we should not neglect the possibility that neurons might also be an important source of PDGF (or other growth factors) for O-2A progenitors in the developing CNS. There is some experimental evidence in support of this idea; O-2A progenitor cells from developing rat cerebellum or optic nerve are stimulated to divide in vitro in medium conditioned by cultures of young cerebellar interneurons, and about half of this mitogenic activity can be neutralized by antibodies against PDGF (Levine, 1989). Retinal ganglion neurons contain mRNA encoding the PDGF A chain, and immunostain with anti-PDGF sera (H. Mudhar and W. Richardson, unpublished; Fig. 3), so these neurons could conceivably supply PDGF to the optic nerve, although this would probably require that PDGF be anterogradely transported along, and secreted from, the axons of retinal ganglion neurons. It remains to be seen whether this can occur. It may be that neuron-derived PDGF is not involved in neuron-glial cell interactions, but instead has some autocrine function or mediates interactions between neurons and their targets. There is a recent report that many neurons in the developing rat brain possess PDGF-/3R, and that PDGF-BB enhances survival and neurite outgrowth of cultured cerebellar neurons (Smits et al. 1991).

For a proper understanding of oligodendrocyte development in vivo, we need to know the locations of the germinal zones where the first committed precursor cells arise and the routes along which they migrate. We also need to know the sequence of maturation events and where these occur, in order to determine what influence the local environment might exert at each stage of development of the O-2A progenitor cells. Therefore, studies of O-2A lineage development in situ are called for. There are several specific markers for the later stages of oligodendrocyte differentiation, such as (in order of appearance) the 04 antigen (Sommer and Schachner, 1981), galactocerebroside (Raff et al. 1978), 2’,3’-cyclic nucleotide 3’-phosphodiesterase (Trapp et al. 1988), and the myelin structural proteins (Dubois-Dalcq et al. 1986). However, there are no reliable markers for the earlier progenitor cells, because the antibodies that are often used to identify these cells in vitro, such as monoclonal A2B5 (Eisenbarth et al. 1979) and antibodies against ganglioside GD3 (Goldman et al. 1984), also label some immature or mature neurons in tissue sections. There is therefore a pressing need for additional O-2A lineage markers.

We wondered whether PDGF-α R might be restricted to O-2A lineage cells in the CNS, because these cells seem to comprise the vast majority of PDGF-α R⁺ cells in the rat optic nerve (Hart et al. 1989a). Therefore, we performed in situ hybridization experiments on sections of rat CNS, using a probe corresponding to the extracellular domain of the rat PDGF-α R. The results we obtained (N. Pringle, H. Mudhar, E. Collarini and W. Richardson, in preparation) suggest to us that expression of PDGF-α R may indeed be restricted to the O-2A lineage. The key observations are as follows. (1) PDGF-α R⁺ cells are

present in the optic nerve, which contains glial cells but no neurons, and these first appear at the chiasmal end of the nerve, spreading up to the retinal end in the first few days after birth just as O-2A progenitor cells do (Small et al. 1987). (2) In the cerebellum, PDGF-α R⁺ cells first appear in small numbers around the day of birth and increase rapidly during the next few days, becoming increasingly localized to the foliar white matter tracts. In the adult, some PDGF-α R⁺ cells remain, mainly in the molecular layer. This fits the pattern described for G_D3⁺ cells (putative O-2A progenitors) in the developing cerebellum (Reynolds and Wilkin, 1988). (3) The spatio-temporal distribution of PDGF-α R⁺ cells in the cerebral cortex is again suggestive of O-2A lineage cells. PDGF-oα R⁺ first appear at the lateral and medial aspects of the subcortical white matter at about E18, and spread into all parts of the cortex by the day of birth (see Fig. 4). They continue to increase in number until at least PIO, especially in the developing subcortical white matter and corpus callosum, before declining again in the adult. This is similar to the behaviour of a class of presumptive O-2A lineage cells described by Levine and Goldman (1988a), who studied oligodendrocyte development in the rat forebrain using antibodies against GDS and carbonic anhydrase (CA). In contrast, most cortical neurons are formed before birth. Moreover, cortical neurons in the postnatal cortex are arranged in a series of discrete lamellae parallel to the pial surface, whereas PDGF-α R⁺ cells never show any signs of stratification, reinforcing the view that PDGF-α R is expressed on glial cells, not neurons. (4) The PDGF-α R hybridization signal is invariably associated with cells that possess small, round, densely-staining nuclei that are characteristic of glial cells. Conversely, where neurons can be unambiguously identified by virtue of their position and/or nuclear morphology (e.g. retinal ganglion neurons, cerebellar Purkinje neurons, hippocampal neurons), these cells are always negative for PDGF-α R. (5) ¹²⁵I-PDGF binding studies in vitro indicate that neither cortical astrocytes nor microglia possess PDGF-α R (N. Pringle and W. Richardson, unpublished).

If it is true that PDGF-α R is restricted to the 0-2A lineage, then what might we deduce about the development of the 0-2A lineage from our in situ studies? We first see small numbers of PDGF-α R⁺ cells in the anterior forebrain at E14-E16. These cells first appear outside the subventicular germinal zones, suggesting that PDGF-α R is not expressed by the earliest committed 0-2A lineage cells but is up-regulated after, or shortly before they leave the germinal zones and migrate away towards their final destinations.

We also considered the possibility that SCIP might specifically mark early stages of the 0-2A lineage in the CNS, since SCIP is expressed highly in proliferating 0-2A progenitors in vitro, but is strongly down-regulated when these cells are induced to differentiate by withdrawing growth factors (see above). However, in situ hybridization of developing rat CNS tissue (Fig. 4) gives no indication that SCIP is restricted to the 0-2A lineage; on the contrary, SCIP seems to be expressed in many developing neurons and even some mature neurons. For example, SCIP is expressed in a broad band of cells in the developing retina at E16 (not shown); these cells cannot belong to the O-2A lineage because the rat retina is unmyelinated. In the anterior forebrain, SCIP mRNA is present at E16 in a broad band of cells in the subventricular zones inferior and lateral to the lateral ventricles. With increasing age, the zone of SCIP expression expands into the subventricular zone superior to the ventricles (Fig. 4), and then moves outwards towards the pial surface. By PO, the zone of SCIP expression lies just beneath the pial surface, superior to the developing subcortical white matter (see Fig. 4). This pattern of expression suggests that SCIP is expressed in a population of migrating neuronal progenitors in the cortex. By PIO, SCIP mRNA is no longer detected in the cortex, but a strong signal appears in hippocampal neurons in regions CAI and CA2. An intense SCIP signal is also observed at P5 in the trigeminal nerve, up to but not beyond the point where it enters the CNS. This presumably represents the point where Schwann cells give way to oligodendrocytes. Surprisingly, we have been unable to detect SCIP⁺ cells in the P5 optic nerve, even though SCIP mRNA can be detected in the nerve at this age on Northern blots (Monuki et al. 1989). We conclude that several, perhaps many cell lineages in the PNS and CNS express SCIP at some point in their development, perhaps associated with the transition from proliferation to differentiation. 0-2A progenitors presumably do express SCIP in vivo but at a level that is below the detection limit of our in situ procedure, at least in outlying regions of the CNS such as the optic nerve. Clearly, SCIP is not a useful marker for the 0-2A lineage in situ.

FGF is reported to be mitogenic for 0-2A lineage cells in culture (Eccleston and Silberberg, 1985; Saneto and deVellis, 1985; McKinnon et al. 1990), although there is at least one other report that contradicts this (Hunter and Bottenstein, 1990). This disagreement may result from differences in the purity and source of the O-2A progenitor cells and/or the particular preparation of FGF. Recently, it was found that the combination of bFGF and PDGF has a striking cooperative effect, stimulating prolonged proliferation of O-2A progenitors in the apparent absence of oligodendrocyte differentiation (Bôgler et al. 1990). bFGF is present in the developing and mature CNS (Gospodar-owicz, 1984; Gonzalez et al. 1990); by immunohistochemistry it appears to be present in the cell bodies of neurons (Janet et al. 1988). However, it is not known whether bFGF can be released from neurons, or any living cell, since its polypeptide precursor lacks a recognizable signal for entry into the constitutive secretory pathway. In any event, it seems unlikely that bFGF and PDGF act together on O-2A progenitors in the postnatal optic nerve, or other developing white matter tracts, otherwise oligodendrocyte differentiation would presumably be inhibited. Perhaps bFGF is released only from dying cells as a response to CNS injury. Alternatively, FGF may be released from living cells by an unconventional mechanism, but its biological effects restricted to a compart-ment(s) of the CNS where oligodendrocytes are not required. For example, receptors for bFGF (bFGF-R) might normally be expressed on O-2A lineage cells only at an early stage in their development, perhaps on a population of self-renewing cells near the ventricles whose function is to generate a steady stream of migrating O-2A progenitors or pre-progenitors. The migrating cells might normally lose bFGF-R as they move away from the subventricular layer, perhaps gaining PDGF-aR in their place. This hypothesis would fit with the observations that cells expressing bFGF-R are found predominantly in the subventricular zones of developing rat and chicken brain (Heuer et al. 1990; Wanaka, Johnson and Milbrandt, 1990), whereas putative O-2A lineage cells expressing PDGF-aR are located mainly outside of the subventricular zone (our results, see above). To explain why O-2A progenitors from the optic nerve respond to bFGF in vitro, we would also have to speculate that bFGF-R is inappropriately up-regulated when the cells are dissociated and placed in culture. More needs to be learned about the FGF receptors on O-2A lineage cells in vitro and in vivo, and the availability of FGF in the CNS, before this and other possibilities can be fully explored.

We thank the other members of our laboratories, past and present, for their contributions to the work described and for stimulating and enjoyable interactions. The work described here was in large part supported by the Medical Research Council, the Multiple Sclerosis Society of Great Britain and Northern Ireland, and the Wellcome Trust.