Oligodendrocyte development
William D Richardson
Wolfson Institute for Biomedical Research and Department of Biology,
University College London, Gower Street, London WC1E 6BT, UK
[in: Glial Cell Development. KR Jessen and WD Richardson, editors. Oxford University Press. In press, 2000)
Correspondence: Bill Richardson / tel +44 (0)20 7679 6729 / fax +44 (0)20 7209 0470 / w.richardson@ucl.ac.uk
I Specification of the oligodendrocyte lineage
Spinal oligodendrocytes are derived from ventral neuroepithelium
Fig.1 Appearance and spread of PDGFRa + oligodendrocyte progenitors in the embryonic mouse spinal cord and forebrain
Fig. 2 Oligodendrocyte lineage markers in the E13.5 mouse spinal cord
Spinal oligodendrocytes originate in the same part of the neuroepithelium as motor neurons
Fig. 3 Regionalization of the ventral spinal cord VZ
Fig. 4 Possible lineages connecting motor neurons, O-2A progenitors and astrocytes in the ventral spinal cord
Role of Sonic hedgehog in oligodendrogenesis
Negative influences on oligodendrocyte lineage specification
Oligodendrogenesis in the brain: analogies between forebrain and spinal cord
One or more oligodendrocyte lineages in the forebrain?
Where do astrocytes come from? Glial-restricted precursors and pre-progenitors
II Control of cell proliferation and survival in the oligodendrocyte lineage
Platelet-derived growth factor (PDGF)
Fibroblast growth factor (FGF)
Insulin and insulin-like growth factors
III Controls on oligodendrocyte differentiation
A cell division limiter in oligodendrocyte progenitors: the role of thyroid hormones
Cell-intrinsic and cell-extrinsic controls on progenitor cell proliferation and population growth
Conclusion and outstanding questions
Fig. 5 Summary diagram of oligodendrocyte lineage progression.
___________________________________________________________________________________________
Introduction
Oligodendrocytes, like other neural cells in the central nervous system (CNS), develop from the neuroepithelial precursors that line the lumen of the spinal cord and the ventricles of the brain the so-called ventricular zone (VZ). Although myelinating oligodendrocytes are one of the last cell types to differentiate in the CNS accumulating mainly after birth in rodents - their progenitors are present in the VZ much earlier than that. In the rat spinal cord, for example, the first overt oligodendrocyte precursors appear in the VZ around embryonic day 14 (E14) or even earlier (conception is E0, birth ~E21). Soon after this, they undergo a morphological transition, detach from the epithelium and migrate way from the VZ through the developing CNS (Reynolds and Wilkin, 1988; Levine and Goldman, 1988). These migratory cells are referred to as O-2A progenitor cells, to acknowledge the fact that they can differentiate into either oligodendrocytes or "type-2" astrocytes in vitro, depending on the composition of the culture medium (Raff et al., 1983). O-2A progenitors were first identified in rat optic nerve cell cultures (Raff et al., 1983); similar (though not necessarily identical) cells are found in cultures derived from many other parts of the CNS. O-2A glial progenitors continue to divide after they leave the VZ, unlike neuronal progenitors. Before they exit the cell cycle they undergo a transition to pro-oligodendrocytes. This is characterized by a change in morphology from a simple, often bipolar form to a more complex form with multiple processes, accompanied by a reduction in motility, a change in cell surface antigens (notably acquisition of the O4 antigen in rats) and altered growth factor responsiveness (Bansal and Pfeiffer, 1992). The pro-oligodendrocytes are thought to undergo a late burst of cell divisions near their site of terminal differentiation in fibre tracts (Pfeiffer et al., 1994). They then leave the cell cycle, associate with axons, express myelin gene products and mature into fully-fledged oligodendrocytes. There are controls at all stages of lineage development including specification, proliferation, differentiation and long-term survival.
I Specification of the oligodendrocyte lineage
Spinal oligodendrocytes are derived from ventral neuroepithelium
Oligodendrocytes are distributed widely through the adult mammalian central nervous system (CNS) with no obvious preference for either dorsal-ventral or anterior-posterior position. It has therefore been something of a surprise to discover that the precursors of oligodendrocytes arise in highly restricted parts of the VZ of the brain and spinal cord. It is now clear that oligodendrocyte progenitors in the spinal cord and brainstem are generated in a sub-domain of the ventral VZ near the floor plate (Yu et al., 1994; Ono et al., 1995; Timsit et al., 1995). Oligodendrocytes in the optic nerve develop from progenitors that originate in a specialized part of the VZ in the ventral diencephalon and migrate from there into the nerve via the optic chiasm (Ono et al., 1997). Therefore, in these parts of the neural tube at least, oligodendrocytes can be classified as a ventral cell type.
The evidence for a ventral origin of oligodendrocytes in the spinal cord has been reviewed before (Miller, 1996; Hardy, 1997; Rogister et al., 1999; Richardson et al., 2000; Thomas et al., 2000). There are three main types of evidence:
1) Dorsal versus ventral spinal cord cultures. When embryonic day 14 (E14) rat spinal cords are bisected longitudinally into dorsal and ventral halves and the cells from each half are cultured separately, oligodendrocytes develop only in the ventral cultures (Warf et al., 1991). The same is true for cultures of E6 chick or quail spinal cord (Pringle et al., 1996; Orentas and Miller, 1996; Poncet et al., 1996; Pringle et al., 1998). At later ages (E16 rat or E8 chick), both dorsal and ventral cultures generate oligodendrocytes. This is consistent with the idea that oligodendrocyte progenitors first arise in the ventral half of the cord but later proliferate and spread into the dorsal half.
2) Lineage markers in the spinal cord. Several early markers of the oligodendrocyte lineage are expressed in a restricted region of the ventral VZ at early embryonic times (e.g. E14 rat, E12.5 mouse, E6 chick) and later spread throughout the cord in a manner suggesting migration of individual progenitor cells (see Fig. 1). The time course of
______________________________________________________________________________
Figure 1 Appearance and spread of PDGFRa +
oligodendrocyte progenitors in the embryonic mouse spinal cord
and forebrain, visualized by in situ hybridization. In the spinal
cord the first progenitors appear in the ventral VZ on E13
(arrow), then they proliferate and migrate away from the midline.
Within a day-and-a-half (E14.5) they spread through most of the
cord. The equivalent ages in rat are E14 to E15.5 and in chick,
E7 to E9. In the forebrain (lower left), a cluster of PDGFRa +cells
appears in the ventral diencephalon (anterior hypothalamic
neuroepithelium) before E13 (arrow). These cells subsequently
proliferate and migrate into the dorsal forebrain including the
developing cortex (Cx) before birth (not shown). At higher
magnification (lower right) it can be seen that the migratory
cells are intensely labelled PDGFRa +
cells (arrows) that develop within the initial cluster of less
strongly labelled neuroepithelial cells (asterisk). MGE, medial
ganglionic eminence; LGE, lateral ganglionic eminence. back to index
_____________________________________________________________________________________________
initial appearance and subsequent spread into dorsal regions matches that predicted from the culture experiments quoted above. Informative lineage markers include the platelet-derived growth factor receptor-alpha (PDGFRa) (Pringle and Richardson, 1993; Nishiyama et al., 1996), antigens recognized by monoclonal antibody O4 (Ono et al., 1995), the NG-2 proteoglycan (Stallcup and Beasley, 1987; Levine and Stallcup, 1987; Nishiyama et al., 1996), transcripts encoding the myelin proteins cyclic nucleotide phosphodiesterase (CNP) (Yu et al., 1994) and myelin proteolipid protein (PLP/DM-20) (Timsit et al., 1995) and the transcription factors Sox10, Olig-1 and Olig-2 (Kuhlbrodt et al., 1998; Lu et al., 2000; Zhou et al., 2000) (more about these later). Lineage markers have obvious shortcomings few individual markers are truly lineage-specific but the overlap among these markers is persuasive (e.g. Fig. 2). PDGFRa+ cells in the late embryonic rat spinal cord are known to be O-2A progenitors because if these cells are immunoselected with an anti-PDGFRa antibody and cultured under appropriate conditions, they all differentiate into oligodendrocytes or type-2 astrocytes (Hall et al., 1996). Conversely, if PDGFRa+ cells are selectively removed from cultures of rat spinal cord cells by antibody-mediated complement lysis, then oligodendrocyte production is strongly inhibited Hall et al., 1996). The same goes for O4+ cells from embryonic chick spinal cord (Ono et al., 1995)
_______________________________________________________________________________
Figure 2 Oligodendrocyte lineage
markers in the E13.5 mouse spinal cord. Olig2, sox10
and PDGFRa transcripts co-localize in the
ventral neuroepithelium. Note that olig2 is
restricted to the CNS, sox10 is in both CNS and PNS (in
Schwann cells), while PDGFRa is
in mesodermal and ectodermal derivatives as well as the CNS.
Horizontal lines mark the ventral and dorsal limits of the
central canal. back to
index
________________________________________________________________________________
3) Chick-quail chimeras. When dorsal or ventral spinal cord neuroepithelium from a quail donor embryo is grafted into the equivalent position of a chick host at the same stage of development (homotypic, homochronic grafts), then donor-derived oligodendrocytes develop only from ventral grafts (Pringle et al., 1998). This is despite an earlier, misleading report to the contrary (Cameron-Curry and Le Douarin, 1995).
Taken together, the evidence suggests strongly that most or all oligodendrocytes in the spinal cord develop from progenitor cells that originate from the ventral neuroepithelium.
Spinal oligodendrocytes originate in the same part of the neuroepithelium as motor neurons
The localized origin of oligodendrocyte progenitors in the ventral spinal cord naturally raises the question of what makes the neuroepithelial precursors at that site differ from their dorsal and ventral neighbours. Fortunately, quite a lot is known already about how ventral neurons are specified and some of this information is directly applicable to oligodendrocytes.
Specification of ventral neurons motor neurons and several classes of ventral interneurons is known to depend on signal(s) from the notochord and floor plate at the ventral midline (Tanabe and Jessell, 1996). An important component of the signal is provided by Sonic hedgehog (Shh) protein, a vertebrate homologue of the product of the Drosophila segment-polarity gene hedgehog (hh). Shh from the notochord first induces formation of the floor plate in the adjacent neural tube, then the floor plate becomes a secondary source of Shh which acts as a graded morphogen, diffusing dorsally to establish a ventral-to-dorsal (high-low) concentration gradient in the ventral neural tube. Shh, acting through its cell-surface receptors Patched (Ptc) and Smoothened (Smo), represses expression of the homeodomain transcription factor Pax6, thus excluding Pax6 from the neuroepithelial domain immediately abutting the floor plate and establishing a ventral-to-dorsal (low-high) gradient of Pax6 in the remainder of the ventral neuroepithelium (Ericson et al., 1997) (Fig. 3). Pax6 regulates downstream genes, setting up discrete domains of gene expression within the neuroepithelium. For example, Pax6 represses Nkx2.2 expression, which is therefore restricted to the Pax6-negative domain abutting the floor plate (Ericson et al., 1997). Other transcription factors that are presumably regulated by Pax6 (and each other) include homeodomain proteins Nkx6.1 and Dbx1 (Briscoe and Ericson, 1999), the high-mobility-group (HMG) protein Sox10 (Kuhlbrodt et al., 1998) and the recently described basic helix-loop-helix (bHLH) proteins Olig-1 and Olig-2 (Lu et al., 2000; Zhou et al., 2000). Downstream of these is an additional set of transcription factors, including the homeodomain proteins Isl-1, Isl-2, Lim-3 Chx-10 and MNR2, which are closely involved in the differentiation of post-mitotic neurons after they have left the VZ (Briscoe and Ericson, 1999). Neurogenesis therefore depends on a hierarchy of protein-gene interactions initiated by a morphogenic gradient of [Shh]. This is analogous to the way the segmented structure of the Drosophila embryo becomes established along its anterior-posterior axis (though in this example the primary morphogen is Bicoid protein, not Hedgehog).
Figure 3
Regionalization of the ventral spinal cord VZ. On the left,
separate blocks of neuroepithelium give rise to different cell
types. The block nearest the floor plate generates visceral motor
neurons (vMNs), then successively more dorsal blocks give rise to
somatic motor neurons (sMNs) and ventral interneurons V2, V1 and
V0. These neuroepithelial domains are marked by their distinctive
patterns of gene expression. The vMN domain is defined by nkx2.2
while more dorsal regions express one or more of nkx6.1, pax6,
irx3 to name a few (Briscoe and Ericson, 1999). This
pattern is established before the onset of neurogenesis about
E10. Graded Shh is thought to induce an inverse gradient of Pax6
which in turn controls downstream genes in a
concentration-dependent manner. Of particular note, the olig1,
olig2, sox10 domain corresponds to that part of the VZ
that gives rise to somatic MNs. Later (~E13) this domain gives
rise to migratory PDGFRa + oligodendrocyte progenitors
(OLPs). PLP/DM20 is expressed weakly in the ventral VZ from
before E11 (W.-P. Yu and WDR, unpublished); the ventral limit of
expression is close to the floor plate but the dorsal limit is
not well defined. back
to index
__________________________________________________________________________
Therefore, under the action of Shh the neuroepithelium is subdivided along the dorsal-ventral axis, each division generating a different subset of cell types (Fig. 3). The domain nearest the floor plate (Pax6-negative, Nkx2.2-positive,) generates visceral (autonomic) motor neurons, which innervate muscles of the heart or diaphragm, for example. The next more dorsal domain (Pax6-low, Nkx2.2-negative) generates somatic motor neurons. In the brainstem these innervate craniofacial muscles such as those of the tongue (hypoglossal motor neurons), while in the cervical spinal cord they innervate skeletal (axial) muscles in the neck. Further dorsal again (Pax6-positive, Nkx2.2.-negative), consecutive domains of neuroepithelium generate V2, V1 and V0 interneurons. This takes us to the dorsal-ventral midline, where the organizing influence of the floor plate cedes to that of the roof plate (Ericson et al., 1997; Briscoe and Ericson, 1999).
In the rodent, PDGFRa+ oligodendrocyte progenitors appear to arise in the same Pax6-low, Nkx2.2-negative region of the neuroepithelium that generates somatic motor neurons (Sun et al., 1998; Lu et al., 2000), implying some kind of lineage relationship between motor neurons and oligodendrocytes. The idea of a common motor neuron-oligodendrocyte lineage is consistent with clonal analysis in the embryonic chick spinal cord in vivo (Leber et al., 1990; Leber and Sanes, 1995), and in the rat spinal cord in vitro (Kalyani et al., 1997). The precise form of the lineage tree that connects these cell types is unknown, but any lineage model must take into account the fact that oligodendrocyte progenitors are generated after motor neurons. In the rat, for example, post-mitotic motor neurons are born between E11 and E13 (Nornes and Das, 1974) whereas PDGFRa+ progenitors do not appear until E14 (Pringle and Richardson, 1993). One possibility is that motor neuron precursors and PDGFRa-negative oligodendrocyte "pre-progenitors" are segregated early on in the VZ through division of neuroglial precursors, and the oligodendrocyte pre-progenitors then remain dormant in the VZ until after motor neuron production is complete. Alternatively, motor neuron and oligodendrocyte progenitors might be formed sequentially from a common neuroglioblast in the VZ (Richardson et al., 2000) (see Fig. 4).
Figure 4 Two possible cell lineages connecting
motor neurons (MN), O-2A progenitors (O) and astrocytes (A) in
the ventral spinal cord. Left: separate neuroblasts (N)
and glioblasts (G) exist side-by-side in the neuroepithelium from
early times; the glioblast lies dormant during motor neuron
production and is activated later. The separate N and G cells
would be formed by asymmetric division of an earlier NG
neuroepithelial cell. Right: a single population of
neurogliobasts (NG) switches fates with time or successive
divisions to generate motor neurons, O-2A progenitors and
astrocytes sequentially. Other schemes combining elements of both
models are also possible. back
to index
________________________________________________________________________________________________________
Whatever the form of the relationship, one idea worth exploring is that production of PDGFRa+ oligodendrocyte progenitors might be triggered by feedback signals from newly-generated motor neurons. This could explain why oligodendrocyte progenitors appear after motor neurons. However, mature motor neurons do not seem to be required for oligodendrogenesis, because oligodendrocytes are generated as normal in explant cultures of spinal cords from Isl-1 knockout mice, which fail to develop any motor neurons (Sun et al., 1998). Nevertheless, it remains possible that motor neuron progenitors, which do not express Isl-1 and presumably are not affected in the Isl-1 knockouts, might provide a feedback signal. Motor neuron progenitors lie closer to the VZ than differentiated motor neurons and might therefore be better placed to signal back to the neuroepithelium.
Role of Sonic hedgehog in oligodendrogenesis
Explant cultures of dorsal or intermediate spinal cord neuroepithelium, which on their own do not give rise to any ventral cell types, can be induced by pure recombinant Shh to give rise to a variety of ventral cells including floor plate cells, motor neurons and oligodendrocytes (Roelink et al., 1994; Pringle et al., 1996; Poncet et al., 1996; Ericson et al., 1997; Orentas et al., 1999). The concentrations of Shh required to induce motor neurons and oligodendrocytes are similar - and lower than needed to induce floor plate cells - consistent with the putative gradient of [Shh] in the ventral neural tube and also with the coincident origin of somatic motor neurons and oligodendrocytes in vivo (Pringle et al., 1996; Orentas et al., 1999). Shh is necessary for oligodendrogenesis in vivo, because oligodendrocytes do not develop in chick spinal cords that are exposed to anti-Shh neutralizing antibody in ovo (Orentas et al., 1999). Moreover, PDGFRa+ oligodendrocyte progenitors do not appear in the spinal cords of perinatal Shh null mice (personal communication from Christer Betsholtz, University of Göteborg, Sweden).
Two periods of Shh exposure are required for motor neuron induction - an early exposure to ventralize the spinal cord and another closer to the time of motor neuron production (Ericson et al., 1996). Oligodendrocyte specification requires an extended period of exposure throughout the time of motor neuron production until shortly before the appearance of the first O4-positive progenitor cells in chick ventral spinal cord explants (Orentas et al., 1999; N.P Pringle unpublished). It is not clear what this implies for the mode of action of Shh.
How does Shh work? back to index
Constitutive activation of Shh signal transduction pathways through mutation or deletion of its inhibitory receptor Ptc results in tumour formation in mice and humans, implying that Shh signalling positively regulates the cell cycle (Stone et al., 1996; Goodrich et al., 1997). In vitro, Shh can act as a mitogen for neural precursor cells from cerebellum, retina and spinal cord (Levine et al., 1997; Jensen and Wallace, 1997; Dahmane and Ruiz i Altaba, 1999; Rowitch et al., 1999). However, the concentration of Shh required to induce DNA synthesis in vitro is generally higher than is needed to specify cell fates in vitro. It could be that the mitogenic effect of Shh is spurious, resulting from cross-talk between intracellular signal transduction pathways at excessively high, non-physiological levels of Shh signalling. On the other hand, neutralizing anti-Shh antibodies have been reported to reduce bromodeoxyuridine (BrdU) incorporation in vivo, implying a physiological role for Shh in stimulating DNA synthesis (Dahmane and Ruiz i Altaba, 1999; Wallace and Raff, 1999).
A key intracellular transducer of Hedgehog signalling in Drosophila is the transcription factor Cubitus interruptus (Ci). The vertebrate equivalent is the Gli family of proteins, three of which have been described to date (Ruiz i Altaba, 1999). The Gli proteins mediate different aspects of Shh signalling and have both complementary and antagonistic activities (Matise et al., 1998; Ruiz i Altaba, 1998). All three are expressed in the ventral neural tube. Ventral neurons appear to develop normally in mice lacking Gli1, Gli2 or both, even though Gli2-deficient mice lack a floor plate (Matise et al., 1998). Mice with a Gli3 mutation (Gli3xt) have defects in the dorsal neural tube but no obvious ventral defects (Hui and Joyner, 1993). The Gli family is named for the fact that Gli1 was first identified as the product of a gene amplified in human glioma (Kinzler et al., 1987), suggesting a link with glial cell biology. However, the possibility that glial development might be specifically affected in Gli-deficient animals has not yet been examined. Intriguingly, Gli1 is expressed in the ventral neuroepithelium of the mouse spinal cord close to the time and place where PDGFRa + oligodendrocyte progenitors first arise (Ruiz i Altaba, 1998).
Negative influences on oligodendrocyte lineage specification
Oligodendrocyte development is subject to negative as well as positive control. For example, bone morphogenetic proteins (BMPs) inhibit oligodendrocyte development and promote astrocyte development in cultures of embryonic precursor cells from cerebral cortex, while the BMP inhibitor Noggin has the opposite effect (Zhu et al., 1999; Mabie et al., 1999; Mehler et al., 2000). BMPs and Dorsalin, another member of the extended transforming growth factor beta (TGFb ) family, are expressed in the dorsal neural tube and/or adjacent epidermis and are known to influence differentiation of dorsal cells (Basler et al., 1993; Liem et al., 1995; Lee and Jessell, 1999). These or related factors might normally be required to inhibit oligodendrogenesis in the dorsal neural tube.
Oligodendrogenesis in the brain
Analogies between forebrain and spinal cord
Oligodendrogenesis at more anterior levels of the neuraxis has not yet been studied as extensively as in the spinal cord and a consensus has yet to be reached. However, what we know suggests that there might be strong analogies between oligodendrogenesis in the embryonic spinal cord and forebrain.
1. Dorsal versus ventral forebrain cultures. Cell cultures established from E14 or E15 rat cerebral cortex (i.e. dorsal forebrain) have a much lower capacity for generating oligodendrocytes than do equivalent cultures of ventral forebrain (Birling and Price, 1998; Tekki-Kessaris et al., 2000). Moreover, fragments of E15 cortex transplanted into the eye do not produce myelinating cells (Kalman and Tuba, 1998). In contrast, E17 or E18 cortical cells generate oligodendrocytes readily in culture and also give rise to ectopic myelin when transplanted into the eye. In principle, these experiments resemble the earlier spinal cord experiments of Warf et al. (1991) (see above) and can be interpreted in a similar way that oligodendrocyte progenitors originate in the ventral forebrain and later migrate dorsally into the cerebral cortex.
2. Lineage markers in the forebrain. The myelin gene PLP/DM-20 is expressed in the neuroepithelium of the ventral diencephalon from as early as E9 in the mouse (Timsit et al., 1992). In addition, neuroepithelial cells in the pre-optic area of the ventral diencephalon (beneath the medial ganglionic eminence) express PDGFRa weakly from around E12 in the mouse (E13 rat) (Tekki-Kessaris et al., 2000) (Fig. 1). Subsequently, small, strongly PDGFRa-labelled cells appear in the neuroepithelium, increasing in number and spreading laterally and dorsally into all parts of the forebrain, including the cerebral cortex, before birth (Tekki-Kessaris et al., 2000). These PDGFRa+ cells have been purified from late embryonic (E19) rat forebrain and cerebral cortex by immunoselection with an anti-PDGFRa Ig (Tekki-Kessaris et al., 2000). Like their counterparts in spinal cord and optic nerve, they can all differentiate into oligodendrocytes in culture. At E12 in the mouse, olig-1 and sox10 are co-expressed in the ventral forebrain neuroepithelium with PDGFRa. Olig-2 is more widespread, though still restricted to the ventral diencephalon (Tekki-Kessaris et al., 2000). Thus established lineage markers indicate a ventral origin for oligodendrocytes in the forebrain.
3. Chick-quail chimeras. Direct evidence that dorsal forebrain precursors do not normally generate oligodendrocytes comes from chick-quail chimeras. When a fragment of quail dorsal prosencephalon is grafted into the equivalent position of a chick embryo at the same stage of development, the quail graft can integrate seamlessly into the host and contributes to neurogenesis in the chimeric cortex. However, no donor-derived (quail) oligodendrocytes develop in such chimeras (personal communication from Salvador Martinez, University of Murcia, Spain).
There is no floor plate per se in the forebrain, but nevertheless Shh is expressed in the neuroepithelium at the ventral midline of the diencephalon, adjacent to the region of olig-1, sox10, PDGFRa co-expression, suggesting that Shh is involved in oligodendrocyte lineage specification in the forebrain as in the spinal cord. In support of this, production of oligodendrocytes and their progenitors in explant cultures of rat or chick prosencephalon is inhibited by anti-Shh neutralizing antibodies (Tekki-Kessaris et al., 2000). It is not useful to look in Shh null mice as these animals lack a recognizable forebrain. However, in nkx2.1 null mice Shh expression in the ventral diencephalon is specifically ablated, leaving more caudal sites of expression intact (Sussel et al., 1999). These mice have a morphologically normal forebrain. They die at birth but nevertheless allow us to investigate early oligodendrogenesis in the forebrain in the absence of Shh. We found that (PDGFRa+, Sox10+) cells do not appear on cue (E12) in the ventral forebrains of nkx2.1 null mice, although they do appear later, possibly by migration from more caudal regions (Tekki-Kessaris et al., 2000). Overall, the evidence supports the view that oligodendrocytes in the embryonic forebrain develop from migratory PDGFRa+ progenitors that are generated in the ventral neuroepithelium by a Shh-dependent process, as in the spinal cord.
Naturally, oligodendrocytes in the forebrain cannot be lineally related to motor neurons as they appear to be in the spinal cord, since there are no motor neurons in the forebrain. However, it remains possible that oligodendrocytes are related to some analogous class of neurons in the forebrain, perhaps evolutionarily related to motor neurons. There are neurons in the ventral diencephalon that are involved in motor control, for example.
One or more oligodendrocyte lineages in the forebrain?
Neuroepithelial cells in the ventral diencephalon express the myelin gene PLP/DM-20 from early times (e.g. before E9 in the mouse) and it has been proposed that even at these early times these cells might be dedicated oligodendrocyte precursors (Timsit et al., 1992; Ikenaka et al., 1993; Timsit et al., 1995; Spassky et al., 1998; Perez et al., 1999; Spassky et al., 2000). The relationship between these PLP/DM-20+ neuroepithelial cells and the PDGFRa+ neuroepithelial cells that appear later in the diencephalon (E12 mouse, see above) is still unclear. It might turn out that they are one and the same cell population, or one might be contained within the other.
The picture is complicated because a different population of PLP/DM-20 cells appears later during embryonic development of the mouse (Timsit et al., 1995; Dickinson et al., 1996; Fanarraga et al., 1996; Peyron et al., 1997). In contrast to the weakly labelled, tight-packed neuroepithelial cells discussed above, these are strongly labelled small cells outside the VZ. Relatively small numbers of these latter cells are present in the brainstem of the mouse from E12.5 and in the spinal cord from E14. Similar cells are also present in the ventral diencephalon of the chick (Perez et al., 1999). These PLP/DM-20+ cells do not co-express PDGFRa and on that basis, together with their early appearance, it has been suggested that they might be progenitors of a novel oligodendrocyte lineage (Spassky et al., 1998; Perez et al., 1999; Spassky et al., 2000). However, they do not readily incorporate bromodeoxyuridine (BrdU) as would be expected of proliferating progenitors; nor do they increase in number noticeably in the mouse spinal cord between E14 and E17 (Hardy and Friedrich, 1996; Fruttiger et al., 1999). They co-express mRNA encoding MBP and CNP (Peyron et al., 1997) and in this and their complex, process-bearing morphology they resemble differentiating oligodendrocytes. One possibility is that the PLP/DM-20+ cells are post-mitotic, non-myelinating (or pre-myelinating) oligodendrocytes formed by precocious differentiation of PDGFRa+ progenitors. PDGFRa is known to be rapidly down-regulated during oligodendrocyte differentiation (Hart et al., 1989; Hall et al., 1996; Butt et al., 1997a), which would explain the lack of overlap between PDGFRa and PLP/DM-20 expression.
Where do astrocytes come from? Glial-restricted precursors and pre-progenitors
We have emphasized the developmental connections between oligodendrocytes and neurons. Where do astrocytes fit in? Glial-restricted precursors (GRPs) that can generate both astrocytes and oligodendrocytes have been isolated from E13.5 rat spinal cord by immunoselection with the A2B5 monoclonal antibody (Rao et al., 1998). In clonal density cultures, each of these A2B5+ cells could give rise to oligodendrocytes and one of two types of astrocytes (A2B5+ or A2B5), depending on culture conditions. At E13.5, A2B5 immunoreactivity was detected in a large part of the central spinal cord neuroepithelium overlapping, but not restricted to, the oligodendrogenic domain defined by the olig genes and PDGFRa (Rao et al., 1998). This is consistent with the finding of Hall et al. (1996) that many A2B5+ cells in E14 rat spinal cord cultures do not co-express PDGFRa. Whether all the A2B5+ GRPs studied by Rao et al. (1998) generated oligodendrocytes via a PDGFRa+ intermediate is not known.
Pre-progenitor cells that can give rise to oligodendrocyte (O-2A-like) progenitors have also been described in mixed cell cultures established from neonatal rat cerebral hemispheres. These cells are responsive to PDGF and express the polysialylated form of the neural cell adhesion molecule (PSA-NCAM) (Grinspan et al., 1990; Ben-Hur et al., 1998; Vitry et al., 1999). They can be purified from mixed neural cultures by immunoselection with an antibody against PSA-NCAM and stimulated to proliferate as free-floating clusters ("spheres") with thyroid hormone (TH) together with fibroblast growth factor (FGF) or PDGF. When induced to differentiate on an adherent substratum, the PSA-NCAM pre-progenitors generate mainly oligodendrocytes and astrocytes so in this respect they resemble the GRPs discussed above (Ben-Hur et al., 1998). Apart from their different tissue of origin (brain versus spinal cord), the PSA-NCAM+ pre-progenitors differ from GRPs in that they express PDGFRa immunoreactivity and survive and proliferate in response to PDGF (Grinspan and Franceschini, 1995; Ben-Hur et al., 1998), whereas GRPs do not (Rao et al., 1998).
Retroviral lineage studies frequently indicate that astrocytes and oligodendrocytes are generated from separate progenitors in vivo (Parnavelas et al., 1991; Grove et al., 1993; Williams and Price, 1995). However, pluripotent precursors that can generate oligodendrocytes, astrocytes and even neurons have also been described (Leber et al., 1990; Galileo et al., 1990; Parnavelas et al., 1991; Williams et al., 1991; Levison and Goldman, 1997). Common oligodendrocyte-astrocyte precursors (some of which also generate neurons) have been identified following stereotactic injection of retroviral vectors into the subventricular zones (SVZ) that underlie the lateral tips of the lateral ventricles of the postnatal forebrain (Levison and Goldman, 1993; Zerlin et al., 1995; Levison and Goldman, 1997) (see Chapter X by Goldman). How these precursors in the postnatal SVZ relate to the A2B5+ GRPs or PSA-NCAM+ pre-progenitors discussed above remains to be determined.
II Control of cell proliferation and survival in the oligodendrocyte lineage
Since the first discovery of oligodendrocyte (O-2A) progenitors in rat optic nerve cell cultures (Raff et al., 1983), attention has focussed on the factors that control their proliferation. This is because it was recognized early on that proliferation and differentiation are usually mutually exclusive, so that the key to understanding oligodendrocyte differentiation lies in unravelling the control of progenitor cell proliferation and cell cycle exit (Raff et al., 1985). Later, it became clear that proliferation is only part of the story, and that extracellular signals are required continuously in order to prevent apoptotic death of oligodendrocyte lineage cells (Barres et al., 1992; reviewed by Raff et al., 1993). In the following section I review what we know about the control of proliferation and survival, concentrating on the roles of extracellular signalling molecules.
Platelet-derived growth factor (PDGF)
PDGF is mitogenic for O-2A progenitors in vitro (Raff et al., 1988; Richardson et al., 1988; Noble et al., 1988) and in vivo (Calver et al., 1998; Fruttiger et al., 1999). Presently we recognize three isoforms of PDGF (PDGF-A, -B, -C), and two receptors (PDGFRa, PDGFRb ). All three PDGF isoforms bind to PDGFRa, whereas only PDGF-B binds PDGFRb at high affinity (Heldin and Westermark, 1989; Li et al., 2000). Active PDGF is a covalent homo- or hetero-dimer. PDGF-A-containing dimers acting through PDGFRa on progenitor cells are responsible for most of the PDGF-mediated mitogenic response in the spinal cord, optic nerve and cerebellum, since knockout mice lacking the PDGF A-chain have profoundly reduced numbers of progenitors in these tissues (Fruttiger et al., 1999). PDGF-B seems much less important, because PDGF-B null mice have normal numbers of progenitors (Fruttiger et al., 1999). PDGF-C might play a role because there is a relatively modest (~50%) reduction in progenitor cell numbers in the brainstem of the PDGF-A null mouse, which might indicate residual activity of PDGF-CC, or an unrelated factor such as neuregulin/glial growth factor (NRG/GGF, see below).
PDGF-A is made by many neurons in the CNS and also by astrocytes (Yeh et al., 1991; Mudhar et al., 1993; Ellison et al., 1996). Transgenic mice expressing a PDGF-A transgene in neurons under the control of the neuron-specific enolase promoter (NSE-PDGF-A mice) have many more PDGFRa+ oligodendrocyte progenitors than normal in grey matter areas (Calver et al., 1998), but normal numbers in the optic nerve, an isolated white matter tract (Fruttiger et al., 2000). In contrast, transgenic mice that express PDGF-A under the GFAP promoter (GFAP-PDGF-A mice) have greatly increased progenitor cell numbers everywhere including the optic nerve, which is visibly hypertrophic as a result (Fruttiger et al., 2000). This suggests that PDGF-AA can be delivered into the optic nerve (and other white matter tracts) by astrocytes that reside within the nerve, but not from the axons of neurons that project through the nerve. This might apply to other diffusible factors that are synthesized by neurons. For neurons to deliver factors directly into axon tracts it might be that the factors need to be surface-bound, and transferred into the nerve by lateral diffusion or transport in the plane of the axonal membrane.
The fact that oligodendrocyte progenitor numbers are increased in PDGF-A transgenic mice demonstrates that PDGF is present in limiting concentrations in the normal developing CNS. This conclusion is also supported by the observation that exposure to saturating concentrations of PDGF in vitro causes freshly immuno-purified O4+ pro-oligodendrocytes to revert transiently to O4 progenitors (Gard and Pfeiffer, 1993). The fact that PDGF is normally in short supply is likely to keep progenitor cell division in check, contributing to cell cycle exit and oligodendrocyte differentiation in vivo (see below).
Although white matter astrocytes synthesize PDGF mRNA and protein, they might not secrete PDGF constitutively. O-2A progenitors do not proliferate when optic nerve cell cultures are maintained in defined medium including high concentrations of insulin but without added PDGF, even though optic nerve astrocytes in the cultures themselves make PDGF. However, adding pure PDGF to the cultures induces O-2A progenitor proliferation (e.g. Richardson et al., 1988; Robinson and Miller, 1996). This suggests that optic nerve astrocytes do not release PDGF constitutively in culture. Preventing action potential propagation through the optic nerve in vivo by injecting tetrodotoxin (TTX, a sodium channel blocker) into the eye inhibits O-2A progenitor cell proliferation in the optic nerve; this inhibition can be overcome by simultaneously delivering exogenous PDGF-AA to the nerve (Barres and Raff, 1993). Together, these data raise the possibility that release of PDGF and/or other factors from optic nerve astrocytes might be regulated by electrical activity in retinal ganglion cell axons (Barres and Raff, 1993).
Note that PDGF by itself is not a particularly potent mitogen for purified O-2A progenitors (Barres and Raff, 1994; Robinson and Miller, 1996). Additional polypeptide growth factors can potentiate the activity of PDGF; frequently this effect is masked because other cells (e.g. astrocytes) that are present in the cultures provide co-factors. Therefore, while PDGF is undoubtedly crucial for cell proliferation in vivo, it normally acts in combination with other factors (see below).
When O-2A progenitors start to differentiate into oligodendrocytes, they rapidly lose PDGF receptors. Immunoreactive PDGFRa cannot be detected on newly-formed oligodendrocytes, or even on the majority of O4+ pro-oligodendrocytes in culture (Hall et al., 1996). This explains why PDGF is not mitogenic for pro-oligodendrocytes or oligodendrocytes in culture (Gard and Pfeiffer, 1990; Gard and Pfeiffer, 1993). Nevertheless, PDGF can still act as a survival factor for newly formed GC+ oligodendrocytes (Barres et al., 1992; Barres et al., 1993a). Much lower levels of PDGF signalling are needed to stimulate survival versus proliferation of O-2A progenitors in vitro (Barres et al., 1993a) so very low, even undetectable amounts of PDGFRa on newly-formed oligodendrocytes might be adequate for a survival response. Oligodendrocytes that are more than a few days old cannot any longer be kept alive in vitro by PDGF (Barres et al., 1992; Barres et al., 1993a), presumably because they eventually lose their PDGF receptors altogether. They evidently do not dismantle their mitotic machinery irreversibly because some factors like neuregulin/glial growth factor (NRG/GGF) can induce them to re-enter the division cycle (Canoll et al., 1996) (see below).
Myelinating oligodendrocytes presumably depend on factors other than PDGF for their survival in vivo. This is strikingly illustrated by our NSE-PDGF-A transgenic mice. Although these animals contain many more progenitor cells and produce many more immature oligodendrocytes than normal, the additional oligodendrocytes are all removed by programmed cell death (PCD) soon after they are formed so that the final number of mature, myelin-forming oligodendrocytes is the same as in wild type animals (Calver et al., 1998). Similar data have been obtained by injection of PDGF-AA into the cerebrospinal fluid of new-born rats (Butt et al., 1997b); in those experiments, oligodendrocyte differentiation and myelination in the anterior medullary vellum (AMV) was found to be retarded, presumably due to prolonged progenitor cell proliferation, but the ultimate number of myelinating oligodendrocytes was unaltered. PDGF-AA is clearly not a long-term survival factor for differentiated oligodendrocytes.
It is likely that survival factors for oligodendrocytes are provided by the axons that they ensheath, so that oligodendrocyte number is matched to the surface area of axons to be myelinated (Barres and Raff, 1994). Since there is no evidence that either axon number or diameter is increased by augmenting the supply of PDGF in the experiments described above, axon-associated survival activity should remain at normal levels and support a normal number of oligodendrocytes, as observed. One strong candidate for an axon-associated survival signal is NRG/GGF (see below).
Fibroblast growth factor
The fibroblast growth factor (FGF) family has more than twenty known members to date. They all appear to exert their effects through four tyrosine kinase receptors, FGFR1-FGFR4. Rat oligodendrocyte lineage cells express FGFR1, FGFR2 and FGFR3, though their relative abundance changes as the cells mature from early O4 progenitor to GC+ oligodendrocyte; FGFR3 is expressed most strongly in O4+ pro-oligodendrocytes, while FGFR2 is expressed only by GC+ oligodendrocytes (Bansal et al., 1996). FGFR1 is expressed at all stages of the lineage but most strongly in mature oligodendrocytes (Bansal et al., 1996). Oligodendrocyte lineage cells respond to FGF in a developmental stage-specific way (see below) and this might partly result from their stage-specific complement of receptors.
FGF2 is a mitogen for O4+ rat pro-oligodendrocytes, and prevents their maturation to GC+ oligodendrocytes in vitro (Gard and Pfeiffer, 1993). When cultured in a combination of FGF2 and PDGF, early progenitors continue to divide indefinitely as O4 cells and further differentiation is blocked (Bögler et al., 1990). FGF2 also causes differentiated oligodendrocytes to re-express the O4 antigen, down-regulate myelin genes, change shape and synthesize DNA without entering mitosis (Bansal and Pfeiffer, 1997). This is not simply reversion to pro-oligodendrocyte, as with NRG/GGF (see below), but conversion to a novel phenotype. Intraventricular injection of FGF2 into perinatal rats caused an increase in the number of O4+ pro-oligodendrocytes concomitant with a reduction in the number of myelinating oligodendrocytes, reductions in myelin protein and mRNA expression and morphological changes to myelin sheaths, consistent with the in vitro observations (Goddard et al., 1999).
It is often stated that FGF up-regulates PDGFRa in early progenitor cells, implying that this augments their responsiveness to PDGF and accounts for the prolonged proliferation and inhibition of differentiation observed when progenitors are cultured in the presence of FGF plus PDGF (Bögler et al., 1990). The crucial experimental observation is that the level of PDGFRa mRNA is higher in cultures of O-2A progenitors maintained in FGF2 plus PDGF than it is in parallel cultures maintained in PDGF alone (McKinnon et al., 1990). However, this does not distinguish cause and effect; it could be that the primary effect of FGF is to inhibit differentiation, thus maintaining (rather than up-regulating) PDGFRa expression in progenitor cells it is known that PDGFRa is rapidly down-regulated once O-2A progenitors stop dividing and differentiate (Hart et al., 1989; Butt et al., 1997a). Hence, one can not necessarily conclude that FGF directly controls PDGFRa gene expression, or that the synergy displayed by FGF and PDGF results from an enhanced PDGF-driven response.
We do not have any idea as yet which of the large number of potential FGF ligands might act on oligodendrocyte lineage cells in vivo. There is likely to be redundancy among FGF ligands in any case so identifying those with an in vivo role will be difficult. Some obvious candidates e.g. FGF1 (acidic FGF) and FGF2 (basic FGF) have been deleted in mice with no obvious deleterious effects, although close examination of the FGF2 null revealed defects including reduced numbers of neurons in most layers of the neocortex (Ortega et al., 1998; Miller et al., 2000). The FGF1/FGF2 double knockout has no additional defects (Miller et al., 2000). FGF3-FGF10 have also been targeted in mice; most of these are either viable with no dysmyelinating phenotype (FGF3, FGF5, FGF6, FGF7), or else they die very early in utero (FGF4, FGF8, FGF9) (Mansour et al., 1993; Hebert et al., 1994; Feldman et al., 1995; Guo et al., 1996; Fiore et al., 1997; Meyers et al., 1998; D. Ornitz, personal communication). The FGF10 null mouse dies at birth from respiratory failure (Min et al., 1998; Sekine et al., 1999); oligodendrogenesis has not yet been examined. The receptor knockouts might be expected to be informative. However, both FGFR1 and FGFR2 null mutations are embryonic lethal (Deng et al., 1994; Yamaguchi et al., 1994; Deng et al., 1997) while the FGFR3 null has abnormal bone growth and is deaf but has no dysmyelinating phenotype (Colvin et al., 1996; Wang et al., 1999). Clearly, we still have some way to go before the precise roles of FGF signalling in vivo are established.
Neurotrophins
Neurotrophin-3 (NT-3) stimulates BrdU incorporation in O-2A progenitors and pro-oligodendrocytes from rat optic nerve, and promotes survival of GC+ oligodendrocytes in vitro (Barres et al., 1993a; Barres et al., 1994a; Cohen et al., 1996; Kumar et al., 1998). Although nerve growth factor (NGF) has no mitogenic effect on its own, it can potentiate the mitogenic effect of FGF (but not PDGF) and also promotes survival of differentiated oligodendrocytes (Cohen et al., 1996). In keeping with these observations, progenitor cells and oligodendrocytes from rat brain express TrkA and TrkC, the high affinity receptors for NGF and NT-3, respectively (NT-3 can also bind at lower affinity to certain TrkA isoforms) (Barres et al., 1994a; Cohen et al., 1996). The low-affinity neurotrophin receptor p75 and a truncated TrkB receptor that lacks the tyrosine kinase domain are also expressed at low levels in progenitors and are up-regulated during oligodendrocyte differentiation (Cohen et al., 1996). NT-3 and related neurotrophins are expressed by many neurons in the CNS (Ernfors et al., 1990), so NT-3 could mediate neuron-oligodendrocyte interactions in vivo. Note that Robinson and Miller (1997) found that although optic nerve progenitors were TrkC immunoreactive and responsive to NT-3, spinal cord progenitors were not, indicating that there is regional variation in the response to NT-3 (also see Ibarrola et al., 1996).
NT-3 influences oligodendrocyte development in vivo. When exogenous NT-3 was delivered to the developing rat optic nerve in vivo, there was a significant increase in the number of oligodendrocyte progenitors and oligodendrocytes in the nerve (Barres et al., 1994a). Conversely, when an anti-NT-3 neutralizing antibody was delivered to the nerve over a period of several days in vivo, the numbers of oligodendrocyte lineage cells were reduced (Barres et al., 1994a). Consistent these findings, there is a ~30% reduction in the number of PDGFRA+ progenitors in the spinal cords of NT-3 knockout mice and a ~15% reduction in TrkC knockouts (Kahn et al., 1999). The disparity between receptor and ligand knockouts possibly indicates that NT-3 exerts part of its effect through TrkA or another receptor in vivo. There were also modest reductions in the number of MBP+ and GC+ oligodendrocytes in the neonatal spinal cord; since both knockouts die shortly after birth it is difficult to look later than this. The cross-sectional area of the spinal cord is reduced by ~30-40% in the NT-3 null mice and by somewhat less in the TrkC null (Kahn et al., 1999), so some of the reduction in oligodendrocyte lineage cells could be caused indirectly by the loss of neurons, astrocytes or other cells in the cord. Overall, the evidence indicates that NT-3 plays a significant, though perhaps not a critical role in oligodendrocyte progenitor cell proliferation in vivo. The role of neurotrophins in oligodendrocyte survival remains to be fully explored in vivo, perhaps using cell type-specific or inducible knockout mice.
Insulin and insulin-like growth factors
Oligodendrocytes and their progenitors express both insulin receptors and insulin-like growth factor I (IGF-1) receptors (McMorris and Dubois-Dalcq, 1988; Baron-Van Evercooren et al., 1991). It is not clear whether insulin itself is available in the developing CNS but IGF-1 is expressed by astrocytes (Ballotti et al., 1987; Ayer-le Lievre et al., 1991) and possibly other cells. The effect of adding IGF-1 to cultures containing oligodendrocyte lineage cells is to increase the number of differentiated, GC+ oligodendrocytes (McMorris et al., 1986; McMorris and Dubois-Dalcq, 1988). Part of the effect could be due to increased progenitor cell proliferation, since IGF-1 has been shown to be a necessary mitogenic co-factor for FGF and NT-3 (Barres et al., 1993a; Barres and Raff, 1994). However, IGF-1 also enhances oligodendrocyte survival and this is probably the major effect (Barres et al., 1993a). For example, IGF-1 promotes survival of immuno-purified oligodendrocytes in culture and delivering exogenous IGF-1 to the developing rat optic nerve prevents most of the naturally-occurring death among newly-formed oligodendrocytes (Barres et al., 1993b). Moreover, intraventricular injection of IGF-1 leads to an increase in the number of myelinating oligodendrocytes in the postnatal AMV (Goddard et al., 1999). Transgenic mice that over-express IGF-1 constitutively under the control of the metallothionine promoter have brains that are 50% heavier than normal, increased numbers of oligodendrocytes, more myelin per oligodendrocyte and about twice the total weight of myelin (Carson et al., 1993). Conversely, transgenic mice that over-express the inhibitory IGF binding protein IGFBP-1 have about half the normal weight of myelin (Ye et al., 1995). In principle, therefore, IGF-1 could be an important regulator of oligodendrogenesis and myelination in vivo.
IGF-1 knockout mice have smaller brains than normal mice, but brain weight is not reduced in proportion to the reduction of total body weight (~30% versus ~60%, respectively) (Beck et al., 1995). There are no neurological signs of dysmyelination and, although IGF-1 null brains have less myelin than wild type brains, the reduction is in proportion to the reduced brain weight and the reduced number of axons (Cheng et al., 1998). An exception is the optic bulb where there is a disproportionate lack of myelin (Cheng et al., 1998). Therefore, though IGF-1 has an important influence on oligodendrocyte survival and myelin synthesis in vivo this might be indirect, through IGF-mediated effects on neuronal growth and survival. The overriding influence on oligodendrocyte number seems to be an axon-associated factor(s), which in most CNS regions seems not to be IGF-1. As discussed below, a likely candidate is NRG/GGF.
Neuregulin/glial growth factor
It has been known for a long time that axon membrane (axolemma) preparations are mitogenic for Schwann cells in culture, and this was presumed to reflect the presence of a growth factor(s) bound to the axonal surface (Wood and Bunge, 1975). Early attempts to identify mitogens for Schwann cells led to the identification of an activity in pituitary extracts that was called glial growth factor (GGF) (Raff et al., 1978; Lemke and Brockes, 1984). When the molecule responsible was cloned it was recognized as a member of the neuregulin (NRG) growth factor family (Marchionni et al., 1993). The NRG family includes both membrane-associated and diffusible factors that are the products of three genes in mammals, NRG1, -2 and -3, each of which encodes several alternative-splice products. GGF is one such product of the NRG1 gene; others include the neu differentiation factors (NDFs), acetylcholine receptor-inducing activity (ARIA), sensory and motor neuron-derived factor (SMDF) and the heregulins (for reviews see Burden and Yarden, 1997; Adlkofer and Lai, 2000). Here, I refer to all NRG1 isoforms collectively as NRG/GGF. A major portion of the axon-associated Schwann cell mitogenic activity has been ascribed to NRG/GGF, and it is now clear that it is important for oligodendrocyte lineage development too.
The NRGs all exert their effects through the ErbB family of tyrosine kinase receptors, ErbB1-ErbB4, (also called HER1-HER4). ErbB3 and ErbB4 are direct binding partners for NRG/GGF while ErbB1 (the epidermal growth factor receptor) and ErbB2 are co-receptors that are recruited by ligand-bound ErbB3 or ErbB4. Canoll et al. (1996) and Raabe et al. (1997) found that primary O4+ pro-oligodendrocytes and GC+ oligodendrocytes express ErbB2, -B3 and -B4, ErbB3 being the most abundant, whereas Vartanian et al. (1997) found only ErbB2 and -B4 on these cells. This discrepancy probably reflects differences in the culture conditions or developmental stage of the cells being studied. RNA analyses indicate that erbB4 is most abundant in the embryonic CNS, while erbB3 comes up during early postnatal development. Since the ErbB receptors dimerize in various combinations and each dimer pair could mediate a different subset of biological responses, it is possible that NRG/GGF might affect oligodendrocyte lineage cells in different ways at each stage of their progression from early precursor to myelinating cell.
Vartanian et al. (1999) reported that development of O4+ oligodendrocytes in explant cultures of embryonic mouse spinal cord requires NRG/GGF. They found that oligodendrocytes did not appear in explants of ventral spinal cord from NRG/GGF null mice, but the cultures could be "rescued" by adding pure NRG/GGF to the medium. Conversely, oligodendrocyte development in wild type explants could be blocked by NRG/GGF inhibitors. However, since O4 is a relatively late lineage marker in mice (Fanarraga et al., 1995), it is not clear from these experiments whether NRG/GGF is required early or late during oligodendrogenesis.
Some investigators have found NRG/GGF to be a potent mitogen for rat oligodendrocyte progenitors, particularly O4+ pro-oligodendrocytes (Canoll et al., 1996; Canoll et al., 1999), whereas other studies failed to find a mitogenic effect (Vartanian et al., 1994; Raabe et al., 1997). Shi et al. (1998) found that NRG/GGF by itself is not strongly mitogenic for purified perinatal or adult O-2A progenitors but requires intracellular cAMP levels to be raised (e.g. by forskolin) as well as the input of other factors such as PDGF. It seems likely that differences in the observed mitogenic effects of NRG/GGF might reflect differences in the culture conditions or the source, purity or state of differentiation of the cells which could influence their repertoire of ErbB receptors, for example. Most workers in the field are agreed that NRG/GGF is a potent survival factor for differentiated oligodendrocytes, however (Canoll et al., 1996; Raabe et al., 1997; Canoll et al., 1999).
Apart from its mitogenic and survival-promoting effects, NRG/GGF has been found to arrest lineage progression at the pro-oligodendrocyte stage in vitro, preventing them from differentiating further to GC+ oligodendrocytes (Canoll et al., 1996). Conversely, NRG/GGF was found to induce phenotypic reversion of GC+ oligodendrocytes to O4+ pro-oligodendrocytes, causing them to re-enter the cell division cycle and proliferate in vitro (Canoll et al., 1999). This could be important in the context of demyelination/remyelination during multiple sclerosis and other demyelinating diseases.
NRG/GGF is expressed by some neuroepithelial cells in the embryonic spinal cord and brain, by cells in the forebrain SVZ and by a variety of differentiated neurons (Orr-Urtreger et al., 1993; Corfas et al., 1995; Meyer et al., 1997; Vartanian et al., 1999). The expression pattern is therefore consistent with NRG/GGF playing both early and late roles in oligodendrocyte lineage progression. Mice that are homozygous null for NRG1, erbB2, erbB3 and erbB4 all demonstrate a requirement for NRG/ErbB signalling in formation of the peripheral nervous system and the heart (Gassmann et al., 1995; Lee et al., 1995; Meyer and Birchmeier, 1995; Kramer et al., 1996; Riethmacher et al., 1997). They die mid-gestation, precluding analysis of oligodendrogenesis in vivo. However, it has recently proved possible to prolong the life of at least one of these knockout mice by expressing the missing gene product(s) under the control of a heart-specific gene promoter (Woldeyesus et al., 1999). This should allow analysis of later functions of the NRG/ErbB signalling system including its role in the CNS.
Chemokine GRO-a
Recently, the growth-regulated oncogene-a (GRO-a ), a human CXC chemokine, has entered the fray as a mitogenic co-factor for oligodendrocyte progenitors. This factor is homologous to KC, a PDGF-induced immediate early gene product in NIH-3T3 cells (i.e. it is transcriptionally activated by PDGF in 3T3 cells in the absence of new protein synthesis; Cochran et al., 1983). GRO-a peptides are involved in growth regulation of a variety of cell lineages including fibroblasts and hemopoeitic precursors. CXC chemokines including GRO-a are up-regulated in the CNS during experimentally-induced demyelination (Glabinski et al., 1998) and in the dysmyelinating mouse mutant jimpy (Wu et al., 2000).
As mentioned before, PDGF is not by itself strongly mitogenic for purified O-2A progenitor cells in culture. The mitogenic response of progenitors to PDGF is greatly potentiated by factors present in mixed cell cultures of spinal cord or in medium conditioned by spinal cord cultures (Robinson and Miller, 1996); it now seems that a large part of this potentiating activity is due to GRO-a (Robinson et al., 1998). The polypeptide by itself is not mitogenic but it enhances BrdU incorporation in O4 early progenitors several-fold over that observed with PDGF alone. The enhancement is only observed with saturating concentrations (10 ng/ml) of PDGF and is optimum only over a narrow range of GRO-a concentrations (Robinson et al., 1998). There is no mitogenic effect on O4+ pro-oligodendrocytes. A potential source of GRO-a in vivo is astrocytes, since cultured astrocytes secret GRO-a and immunoreactive GRO-a can be detected in white matter glia, probably astrocytes, in situ (Robinson et al., 1998). It was suggested that transient synergy between PDGF and GRO-a might stimulate a local burst of progenitor cell proliferation, such as the late burst that is thought to occur in white matter when progenitors have stopped migrating and before they finally differentiate into oligodendrocytes (Robinson et al., 1998).
Other factors involved in oligodendrocyte proliferation or survival
A variety of other diffusible factors have been implicated in the regulation of proliferation and/or survival in the oligodendrocyte lineage. These include the cytokines ciliary neurotrophic factor (CNTF), interleukin-6 (IL-6) or leukemia inhibitory factor (LIF), all of which enhance oligodendrocyte survival in culture (Barres et al., 1993a; Mayer et al., 1994; Vos et al., 1996). The survival-promoting activities of cytokines, neurotrophins and insulin/IGF-1/IGF-2 are additive, so that oligodendrocytes can be kept alive for weeks in a combination of factors including at least one member of each of these groups of molecules (Barres et al., 1993a). Delivery of exogenous CNTF has been shown to promote oligodendrocyte development in the optic nerve in vivo (Barres et al., 1993a; Barres et al., 1996), apparently by stimulating progenitor cell proliferation, since CNTF has been shown to enhance PDGF-driven progenitor cell proliferation in vitro (Barres et al., 1996). Conversely, progenitor cell proliferation is reduced and oligodendrocyte production retarded in transgenic mice lacking CNTF, although myelination is ultimately normal (Barres et al., 1996). Therefore, CNTF plays a non-essential role in oligodendrocyte development. LIF also plays a subtle role because a reduction in MBP immunoreactivity in the brains of female (not male) LIF knockout mice has been reported (Bugga et al., 1998).
Apart from the diffusible factors discussed above, several other classes of molecules are also important for oligodendrocyte lineage development. Interactions between the cell and extracellular matrix are known to provide a mitogenic input, sometimes in synergy with growth factors. For example, oligodendrocyte lineage cells express an evolving set of a and b integrin subunits as they mature and these appear to influence not only cell-substrate adhesion and migration, but also proliferation, differentiation and survival (Frost et al., 1999; Blaschuk et al., 2000). Non-peptide agents such as thyroid hormones, retinoic acid and progesterone have also been implicated in oligodendrocyte development. Of particular note are the thyroid hormones (TH), which have been implicated in the proliferation of PSA-NCAM+ neuroepithelial glial precursors (Ben-Hur et al., 1998) and which are also required later for progenitor cells to differentiate into oligodendrocytes (see below). Finally, there is a growing awareness of the role of neurotransmitters and ion channels in oligodendrocyte proliferation control; this is the subject of Chapter X (Gallo).
III Controls on oligodendrocyte differentiation
A cell division limiter in oligodendrocyte progenitors; the role of thyroid hormones
When a single oligodendrocyte progenitor is cultured on its own in defined medium in the absence of other cells or added mitogens, it stops dividing and differentiates within a day or two into an oligodendrocyte. This suggested that oligodendrocyte differentiation is a default pathway that does not require signals from other cells. However, it turns out that the thyroid hormone triiodothyronine (T3), a constituent of the defined medium used for the early experiments (Bottenstein and Sato, 1979) is important for timely oligodendrocyte differentiation in vitro.
When perinatal optic nerve O-2A progenitors are cultured in mitogens such as PDGF, they proliferate for a time but not indefinitely; eventually they stop dividing and, as long as TH is present in the medium, they differentiate. If there are survival factors present they survive and mature, otherwise they die. What makes the progenitors stop dividing? The cells seem to have a built-in division limiter, or "clock". The evidence for this is as follows. If a single perinatal optic nerve progenitor is placed in a microwell in the presence of mitogens (e.g. astrocyte-conditioned medium) it divides a few times then all its progeny differentiate more-or-less synchronously into oligodendrocytes (Temple and Raff, 1985). A given progenitor can go through any number of divisions between zero and approximately eight. However, if the two daughters of the first division are separated and transferred to different microwells, both siblings usually go though the same number of further divisions before differentiating (Temple and Raff, 1986). It seems that each progenitor remembers its mitotic history, even in isolation, and retires from the cell division cycle after a pre-set time or number of divisions. The fact that not all progenitors go through the same number of divisions in this experiment is presumed to be because they have already used up a variable amount of their time or divisions in vivo, before the beginning of the experiment. This implies that new migratory progenitor cells must be produced continuously within the VZ (the part that supplies the optic nerve, at least) over an extended period of time. It is not known if this is correct but it might be testable.
A role for TH in these timing events was suggested by the observation that when T3 was omitted from the culture medium, the progenitor cells seemed to divide indefinitely in response to PDGF or NT-3 (Barres et al., 1994b). In fact they do not go on for ever but eventually stop dividing even in the absence of TH; nonetheless, the normal limit on their proliferation is greatly extended (Gao et al., 1998). It appears that TH is not required for timing per se, but rather for triggering cell cycle exit at the appropriate time; if progenitors are cultured in the absence of TH beyond the time that they would normally differentiate and then TH is added belatedly, the cells differentiate more rapidly and synchronously than they would have done in the continuous presence of the hormone (Barres et al., 1994b). It is as if they remember that they have exceeded the division limit (i.e. the intrinsic timer still works) but they are unable to exit the cell cycle unless TH is present. This "effector" function of TH can be mimicked by glucocorticoids or retinoic acid (RA) (Barres et al., 1994b).
There is an established literature on the role of TH in brain development and myelination (for review see Rodriguez-Peña, 2000). For example, the start of myelination is delayed in hypothyroid rats (Balazs et al., 1969; Walters and Morell, 1981; Ibarrola and Rodriguez-Peña, 1997) and accelerated in hyperthyroid rats (Marta et al., 1998) or rats that receive postnatal injections of T3 (Barres et al., 1994b). Moreover, TH normally only becomes available in the rat when the thyroid gland becomes active after birth, which is around the time myelin first starts to appear. Oligodendrocytes and their progenitors possess receptors for T3 (Baas et al., 1994; Fierro-Renoy, 1995; Gao et al., 1998; Carre et al., 1998) so it is likely that T3 acts directly on oligodendrocyte lineage cells to control their differentiation in vivo.
Cell-intrinsic and cell-extrinsic controls on progenitor cell proliferation and population growth
How does the cell-intrinsic division limiter (timer) work? There is evidence that inhibitors of cyclin-dependent kinases (cdks) - in particular, p27Kip1 - are key. As progenitor cells age in culture the amount of immunoreactive p27Kip1 in their nuclei increases and remains high after they differentiate into oligodendrocytes (Durand et al., 1997). This suggests a model in which p27Kip1 accumulates with time until eventually a critical threshold is reached, triggering cell cycle exit and oligodendrocyte differentiation. Consistent with this, p27Kip1 null mice are about 20% larger than normal, suggesting that progenitor cells of all sorts go through more divisions than normal in the absence of p27Kip1. Moreover, oligodendrocyte progenitors from optic nerves of p27Kip1 null mice undergo more divisions than normal in vitro (Durand et al., 1998) and oligodendrocyte differentiation is perturbed (Casaccia-Bonnefil et al., 1997). Other candidate components of the cell-intrinsic timing mechanism are TH receptors (b isoforms) (Barres et al., 1994b; Gao et al., 1998), the transcription factor SCIP/Oct-6 (Collarini et al., 1992) and the helix-loop-helix protein ID4 (Kondo and Raff, 2000).
A strict cell-intrinsic timing mechanism might not be adequate to explain the time course of oligodendrogenesis in all parts of the CNS. Cell clones containing both oligodendrocyte progenitors and differentiated oligodendrocytes commonly develop in low density spinal cord cultures, indicating that the progeny of single precursor cells do not necessarily differentiate synchronously as first described for optic nerve progenitors (Zhang and Miller, 1995; Ibarrola et al., 1996). In the spinal cord, oligodendrocytes continue to be generated for several weeks after birth, implying - if a cell-intrinsic timer were solely responsible for timing differentiation - that new progenitor cells must be generated continuously in the VZ over a similar extended period. This is possible but seems unlikely since oligodendrocyte lineage markers do not persist that long in the spinal cord VZ. - for example, olig-1 expression is detected in the ventral VZ for only a few days before birth (Lu et al., 2000; N. Tekki-Kessaris and WDR, unpublished observations).
Another type of observation indicates that progenitor cell proliferation is not always determined purely from within the cell. By inter-breeding two independent lines of NSE-PDGF-A transgenic mice we have been able to generate offspring with up to four transgene loci and a large (up to ~20-fold) increase in the expression of PDGF-A transcripts (van Heyningen et al., 2000). If the number of progenitor cell divisions were limited by a cell-autonomous mechanism, we might expect progenitor cell numbers to saturate at some point, regardless of transgene dose and amount of PDGF available. However, we found that the number of progenitor cells continued to increase in proportion to the level of PDGF-A mRNA (which presumably correlates with the rate of PDGF protein synthesis) until there were around twenty times the normal number of progenitors before birth and no sign of an approaching plateau (Calver et al., 1998; van Heyningen et al., 2000). At this point the new-born animals were no longer viable. At the other extreme, PDGF-A (-/-) embryos had very few progenitors indeed (<10% wild type) and PDGF-A (+/-) littermates had half of the wild type number. This demonstrates vividly that progenitor cell numbers in the cord are not determined by cell-intrinsic mechanisms - at least not to an extent that is useful to the animal - but rather are set by rates of supply of extracellular mitogens such as PDGF.
Thus, there is evidence that the progenitor cell proliferation is influenced both by cell-intrinsic and cell-extrinsic controls on cell division. This is to be expected, given that cell cycle control proteins (e.g. cdks, cdk inhibitors) are controlled by factors outside the cell and, conversely, cdks and cdk inhibitors modulate the cells response to extracellular factors. While there clearly are intracellular mechanisms that limit a cell's capacity to proliferate in vitro, they do not necessarily override environmental controls that operate in vivo. From experiments quoted above it seems likely that the steady-state number of progenitor cells and the time of onset of oligodendrocyte maturation in the embryonic and early postnatal spinal cord is influenced as much (or more) by limiting amounts of extracellular factors (e.g. PDGF and TH) as by cell-autonomous mechanisms. However, the intracellular timer is clearly present and might kick in to terminate cell proliferation later in development - for example in the optic nerve, which is an inherently late-developing part of the CNS. There is also evidence that long-term changes in the properties and behaviour of progenitor cells (e.g. progenitor cells in the adult versus those in the perinatal CNS) are driven by cell-intrinsic mechanisms (Gao and Raff, 1997; Tang et al., 2000)
There is also evidence that the Notch signalling pathway plays a role in timing oligodendrocyte differentiation. Oligodendrocytes express Notch1 and their differentiation is strongly inhibited in vitro by Notch ligands Jagged or Delta (Wang et al., 1998). Notch1 is expressed on optic nerve oligodendrocytes and its ligand Jagged-1 on retinal ganglion cell axons; neuronal expression of Jagged-1 declines during the period of active myelination in vivo, consistent with a role in timing oligodendrocyte differentiation (Wang et al., 1998). An attractive idea is that down-regulation of Jagged might correlate with when axons reach their targets and establish synaptic contacts, thus preventing myelination of growing axons. A picture is emerging in which cell contact-mediated signals (e.g. Jagged, NRG/GGF) combine with short-range paracrine signals (e.g. PDGF, FGF) and long-range systemic signals (e.g. TH, retinoic acid) to control oligodendrocyte differentiation and maturation in a way that serves the requirements of both the local environment and the organism as a whole.
Conclusion and Outstanding questions
There is a plethora of signalling molecules that influence development of the oligodendrocyte lineage from specification to myelination (Fig. 5). However, while there is an intimidating amount of information available, there are still fundamental questions to which we need answers.
There are still uncertainties about the sites of origin of oligodendrocyte precursors in the germinal zones, particularly of the brain. We need to know where and when the lineage is specified if we are to stand a chance of discovering how they are specified. We also need to make a major effort to understand how precursor cells defined by different assays or isolated from different regions of the CNS relate to one another tripotential GRPs in the embryonic spinal cord, PSA-NCAM+ precursors in the embryonic brain, pluripotent precursors in the postnatal and adult CNS and, of course, how these all relate to the familiar O-2A progenitor cell.
It is noteworthy that several of the factors we have discussed seem to influence the oligodendrocyte lineage in stage-specific ways. A clear example is TH, which influences proliferation of PSA-NCAM+ pre-progenitor cells as well as controlling later events concerned with cell cycle exit and oligodendrocyte maturation. This suggests that cells' responses to a given factor changes with their state of maturity whether this be through changing repertoires of receptors or different downstream signal transduction molecules is unknown. Clearly there is a need to learn more about intracellular signalling, a subject which has not been broached in this Chapter. The changing responses of cells to their environment is a general developmental issue that might be illuminated by studying the oligodendrocyte lineage.
We do not understand enough about the complex dynamics of how a population of progenitor cells can generate differentiated, post-mitotic progeny while continuing to divide and maintain the progenitor pool. For example, there is still a lot of uncertainty about the scale of cell death. What is the clearance rate of dead cells and does this vary according to area grey versus white matter, for example? Knowing clearance rates accurately would allow us to calculate death rates with more confidence. Do progenitor cells die, or do they always start to differentiate before dying as immature oligodendrocytes? What is relationship between proliferation and migration? New methods of imaging cells in vivo (in zebrafish, for example) or in explant cultures might allow us to follow cell division, migration,. differentiation and death in real time. How important is cell-autonomous behaviour compared to the influence of the extracellular environment? To answer this we need to know more about the local environment that oligodendrocyte lineage cells experience in vivo - the local concentrations of growth factors and so on - and also we need more detailed information about how the cells actually behave in vivo. For this we need to get back to basics and measure, for example, stage- and region-specific rates of cell division in vivo.
Crucially, we need to test more of our ideas in an in vivo setting - in transgenic animals, for example. This is difficult, time-consuming and expensive but the promise is that by gaining a thorough understanding of oligodendrocyte development in vivo we will come to discover more about development (and Biology) in general and, in doing so, learn how to manipulate the clinically important oligodendrocyte lineage to the future benefit of individuals with dysmyelinating disease.
Figure 5 Summary diagram of oligodendrocyte
lineage progression. Shown below the diagram are lineage markers
that have been used to identify various stages of the lineage,
mainly in rat. Differences in the specificity of markers A2B5 and
O4 are indicated for mouse and chick. Above the diagram are
listed some of the signalling molecules that are thought to
influence the lineage as it develops. For details see the text. back to index
_________________________________________________________________________________________
Acknowledgements
The author thanks members of his laboratory and many others, too numerous to name here, for ideas, discussions and helpful comments. Work in the author's laboratory is funded by the Medical Research Council and the Wellcome Trust.
Adlkofer, K. and Lai, C. (2000). Role of neuregulins in glial cell development. Glia 29, 104-11.
Miller, R.H. (1996). Oligodendrocyte origins. Trends in Neurosciences 19, 92-6.