Oligodendrocytes are the myelinating glial cells of the central nervous system (Bunge, 1968). These cells differentiate from bipolar, migratory precursor cells that arise from specific zones of the neuroepithelium and then disperse throughout the developing gray matter. In the spinal cord, oligodendrocyte precursors (OLPs) emerge from the ventral ventricular zone in a restricted domain near the floor plate (Noll and Miller, 1993; Pringle and Richardson, 1993; Timsit et al., 1995). This domain is thought to coincide with the early established motor neuron progenitor domain (Richardson et al., 1997) and expresses basic helix–loop–helix transcription factors Olig1 and Olig2 (Lu et al., 2000; Zhou et al., 2000). Recent gain- and loss-of-function studies indicated that Olig2 is a critical determinant in the generation of both motor neurons and oligodendrocytes (Lu et al., 2002; Takebayashi et al., 2002; Zhou and Anderson, 2002). Studies in chick have implicated Shh as a critical determinant in specifying spinal cord OLP development (Poncet et al., 1996; Pringle et al., 1996; Orentas et al., 1999; Soula et al., 2001). Although Shh−/− mutants fail to develop oligodendrocytes (Alberta et al., 2001), this finding is complicated by the lack of ventral motor neuron progenitor domains from which OLPs are thought to emerge (Lu et al., 2000; Zhou et al., 2000).
Intracellular responses to Shh are primarily mediated by the Gli family of zinc-finger transcription factors (Litingtung and Chiang, 2000a; Ruiz i Altaba et al., 2003). Genetic studies have indicated that Gli2 primarily functions as an activator of Hh signaling (Ding et al., 1998; Matise et al., 1998), while Gli1 plays a minor role in this process (Park et al., 2000). In contrast, Gli3 is a bipartite transcription factor capable of functioning both as a repressor and an activator in a context-dependent manner (Dai et al., 1999; Ruiz-i-Altaba, 1999; Sasaki et al., 1999). The repressor function is highlighted by the fact that loss of Gli3 function can largely rescue neural patterning defects observed in Shh−/− spinal cord (Litingtung and Chiang, 2000b; Persson et al., 2002). On the other hand, the observation that loss of Gli3 function can exacerbate Gli2−/− patterning defects in the ventral neural tube is consistent with the activator function of Gli3 (Motoyama et al., 2003).
The role of Gli2 in oligodendrocyte development has been reported recently. In Gli2−/− spinal cord, the appearance of OLPs is initially delayed, but the total number of OLPs remains unaffected at later stages (Qi et al., 2003), suggesting that either Shh signaling is dispensable for the generation of OLPs or Gli2 shares a redundant role with other Gli genes. Because Olig2+ motor neuron progenitors are generated in Shh−/−;Gli3−/− embryos (Persson et al., 2002), we decided to examine whether Shh signaling is essential to switch from motor neuron specification to OLPs and their subsequent differentiation into oligodendrocytes. Here, we provide genetic evidence that generation of OLPs can proceed independent of Shh signaling, whereas OLP differentiation into mature oligodendrocytes requires Shh signaling.
Olig2 Expression Is Maintained During Oligodendrocyte Development in Shh−/−;Gli3−/− Mutants
In the developing spinal cord, motor neurons and oligodendrocytes are sequentially generated from Olig2+ motor neuron progenitor domains (Lu et al., 2002; Zhou and Anderson, 2002). At the onset of oligodendrocyte development, Olig2 expression becomes up-regulated and it is continuously expressed in the OLPs (Qi et al., 2003). While motor neurons (Litingtung and Chiang, 2000b) and Olig2 (Persson et al., 2002) expressions are restored in Shh−/−;Gli3−/− mutants during neurogenesis, it is not known whether Shh signaling is required for the maintenance of Olig2 expression during oligodendrocyte development. We, therefore, examined Olig2 expression in Shh−/−;Gli3−/− and Gli3−/− mutants from embryonic day (E) 12.5 to E14.5 as OLPs are being generated. As noted previously, Olig2 expression at E12.5 is restricted to a discrete domain near the floor plate of the ventral neural tube (Fig. 1 A). In Shh−/−;Gli3−/− mutants, Olig2 expression is similarly localized but with a moderate reduction in the expression level (Fig. 1 B). At E13.5, many Olig2-expressing cells begin to populate the spinal cord outside of the ventricular zone in wild-type embryos (Fig. 1 E). In contrast, Olig2-expressing cells are still largely restricted to the ventricular zone in Shh−/−;Gli3−/− mutants (Fig. 1 G). However, by E14.5, a large number of Olig2+ cells dispersed throughout the spinal cord can be detected in both wild-type and mutant (Fig. 1 I,IK). The delay in Olig2 appearance is not due to a general developmental delay, because we observed many differentiated neurons in E11.5 Shh−/−;Gli3−/− mutants as indicated by HuCD, a general neuronal differentiation marker (see Supplementary Figure S1, which can be viewed at http://www.interscience.wiley.com/jpages/1058-8388/suppmat). Furthermore, the development of skeletal elements in the limb proceeds normally in Shh−/−;Gli3−/− mutants (Litingtung et al., 2002). The expression of Olig2+ cells is comparable between wild-type and Gli3 mutants (Fig. 1 D–L).
The delayed and reduced Olig2 expression suggests that OLP development is compromised during early stages of gliogenesis in Shh−/−;Gli3−/− mutants. To confirm this idea, we examined the expression of two OLP markers, Sox10 and Pdgfrα, in the spinal cord. Similar to Olig2, expressions of Sox10 and Pdgfrα are restricted to the ventricular zone at E12.5 (Fig. 2 AA,BA). In contrast to wild-type and Gli3−/− (Fig. 2 AD,BD), the expressions of Sox10 and Pdgfrα do not appear until E13.5 in Shh−/−;Gli3−/− mutant spinal cord (Fig. 2 AC,AG,BC,BG). However, by E14.5, the number of OLPs is comparable between wild-type and mutants (Fig. 2 AI,AK,BI,BK). Collectively, these results indicate that Shh signaling plays an important role in regulating the normal level of Olig2 expression in the neural epithelium and in the timing of OLP specification. However, in the absence of Gli3, Shh is not required for the generation and expansion of OLP population in the spinal cord.
Persistent Neurogenesis and Neurogenin Expression in Shh−/−;Gli3−/− Mutants
Misexpression studies in mice suggest that Shh can regulate neural precursor cell proliferation in the spinal cord from E12.5 onward (Rowitch et al., 1999). Therefore, it is possible that delay in OLP appearance in Shh−/−;Gli3−/− spinal cord may be due to reduced progenitor cell proliferation during early stages of oligodendrogenesis. To test this possibility, we examined the rate of cell proliferation in the spinal cord after one hour of 5-bromodeoxyuridine (BrdU) administration at E13.5. As indicated in Figure 3, we observed more BrdU+ cells throughout the neural tube (Fig. 3 I), indicating that cell proliferation is increased rather than decreased in Shh−/−;Gli3−/− mutants compared with wild-type controls. Similar results were also obtained with phosphohistone antibody (Fig. 3 E,F), which labels cells undergoing mitosis.
Because motor neurons and oligodendrocytes are thought to emerge sequentially from the ventral progenitor domain (Richardson et al., 1997), we reason that the delay in OLP generation may be attributed to the delay in neuron–glia fate switching. Therefore, we compared the capacity for motor neuron differentiation between wild-type and Shh−/−;Gli3−/− mutants at E12.5, at a time when motor neuron generation diminishes significantly. It has been established that newly generated motor neurons emerge from the ventricular zone and migrate laterally (Tsuchida et al., 1994). We determined the number of recently generated motor neurons at E12.5 after BrdU pulsing for 24 hr. At E12.5, when OLPs begin to emerge, there are only a few BrdU+ Isl1+ cells (Fig. 3 G,J) near the ventral progenitor domain, indicating that motor neuron generation is already or nearly completed. In contrast, the number of Isl1+ BrdU+ cells (Fig. 3 H,J) is much higher in Shh−/−;Gli3−/− spinal cord (Fig. 3 J, arrows), consistent with the reduced and delayed OLP production. This finding further supports the notion that motor neurons and OLPs are generated sequentially from the motor neuron progenitor domain.
It has been proposed that extinction of Ngn2 expression in the motor neuron progenitor domain is required to promote oligodendrocyte specification mediated through Olig1 and Olig2 genes (Zhou and Anderson, 2002). At E11.5, just before the appearance of OLPs, Ngn2 expression becomes down-regulated in the presumptive Olig2+ progenitor domain in the normal spinal cord (Fig. 4 A). At E12.5, Ngn2 expression is almost excluded from the Olig2+-expressing domain (Fig. 4 C), and by E14.5 when oligodendrogenesis is fully activated, Ngn2 expression is significantly diminished in the spinal cord (Fig. 4 G). In contrast, strong Ngn2 expression persists in the ventricular zone of the presumptive Olig2+ progenitor domain (Fig. 4 B,D,F) and only becomes down-regulated after E13.5 in Shh−/−;Gli3−/− spinal cord (Fig. 4 H). These results, together with the enhanced ability of Shh−/−;Gli3−/− mutants to generate motor neurons at E12.5, and the known inhibitory role of Ngn2 in the generation of OLPs in the chick spinal cord (Zhou et al., 2001), strongly suggest that delayed OLP specification in Shh−/−;Gli3−/− spinal cord can be attributed to the enhanced neurogenic activity and persistent expression of Ngn2 in the ventral progenitor domain of the Shh−/−;Gli3−/− neural tube.
Differentiation of OLPs Is Defective in Shh−/−;Gli3−/− Mutants
We next determined the capacity of OLP cells to differentiate into mature oligodendrocytes in Shh−/−;Gli3−/− spinal cord. To detect mature oligodendrocytes, we examined the expression of myelin basic protein (MBP) and proteolipid protein (PLP) in the caudal region of the spinal cord at E16.5. In wild-type and Gli3−/− mutants, a group of MBP+ and PLP+ mature oligodendrocytes can be clearly observed (Fig. 5 A,D,E,H); however, in Shh−/−;Gli3−/− mutants, mature oligodendrocytes were not generated (Fig. 5 C,G). These data indicate that, despite near-normal production of OLPs in Shh−/−;Gli3−/− mutants, differentiation of OLPs into oligodendrocytes is impaired in double mutants. Because Shh−/−;Gli3−/− embryos die around E17.5, we cannot rule out the possibility that lack of myelin-specific gene expression is due to delayed differentiation of OLPs.
Previous studies have implicated Nkx2.2 as a key transcription factor in promoting OLP differentiation (Qi et al., 2001; Zhou et al., 2001). In Nkx2.2−/− mutants, Pdgfr+, or Sox10+, OLPs are generated but maturation of OLPs is defective during embryogenesis (Qi et al., 2001). We, therefore, examined the pattern of Nkx2.2 expression in Shh−/−;Gli3−/− mutants. At E12.5, the Nkx2.2+ domain is situated in the ventral neural tube just below the Olig2+ motor neuron domain in wild-type and Gli3−/− mutants (Fig. 5 I,L). As development proceeds, Nkx2.2+ cells begin to migrate into lateral regions (Fig. 5 M,Q,U) and acquire Olig2 expression (Zhou et al., 2001; Fu et al., 2002). However, in Shh−/−;Gli3−/− mutants, Nkx2.2 expression is not restored (Fig. 5 V–X), even at E16.5 when a large number of OLPs is present (see Fig. 2 AK,BK). Because Nkx2.2 has been shown to repress Olig2 expression, the absence of Nkx2.2 is also consistent with the ventral expansion of Olig2 expression domain in the double mutant (compare Fig. 4 M,N). Therefore, the inability of Shh−/−;Gli3−/− mutants to generate mature oligodendrocytes is consistent with the lack of Nkx2.2 expression rather than a delay in OLP differentiation during embryogenesis.
It is thought that OLPs emerge from the ventral neural tube under the influence of inductive signaling by Shh derived from the ventral midline. Experimental abrogation of Shh signaling, either by means of Shh blocking antibody or chemical inhibitor, results in inhibition of OLP specification (Orentas et al., 1999; Alberta et al., 2001; Soula et al., 2001; Chandran et al., 2003). Conversely, exogenous application of Shh protein leads to the generation of ectopic OLPs (Pringle et al., 1996). In light of these observations, our genetic analysis of Shh and Gli3 mutant embryos reveals an interesting finding that OLPs can develop in the absence of Shh and Gli3 and further suggests that a redundant signal may be involved in the switch from motor neuron specification to OLPs. This notion is also consistent with recent findings that OLPs can be generated from primary cultures derived from Shh−/− or Hh inhibitor-treated neural tube explants in the presence of the exogenous fibroblast growth factor-2 (FGF2) growth factor (Chandran et al., 2003; Kessaris et al., 2004). Of interest, Fgf2 is expressed in the developing neural tube, including the floor plate and motor neurons (Zuniga Mejia Borja et al., 1996). Although FGF2 is not required for the development of oligodendrocytes from ventral neural tube explants, nevertheless, it is possible that FGF2 or members of the FGF family may play a compensatory role in the generation of OLPs in Shh signaling mutants.
A parallel compensatory signal, independent of Shh, has also been proposed in the generation of motor neurons in Shh−/−Gli3−/− spinal cord (Litingtung and Chiang, 2000b; Persson et al., 2002), and retinoic acid (RA) has been implicated in Shh-independent motor neuron induction (Novitch et al., 2003). What FGF2 and RA signals have in common is their ability to induce Olig2 expression in the absence of Hh pathway activation. Although FGF2 is sufficient to induce Olig2+ cells destined to form oligodendrocytes in culture (Chandran et al., 2003; Kessaris et al., 2004), a combination of FGF2 and RA appears to be required to induce Olig2+ progenitors destined to form motor neurons (Novitch et al., 2003). Further analysis is required to determine whether FGF2 is an obligatory factor in the differentiation of both motor neurons and oligodendrocytes.
Our studies also indicate that Shh signaling plays an important role in the timely generation of OLPs. This requirement is strikingly similar to that observed in Gli2−/− embryos, in which OLPs also fail to emerge at the appropriate developmental stage (Qi et al., 2003), and further implicates Gli2 as the primary mediator of Shh signaling in the development of OLPs. Whereas Olig2 plays a pivotal role in the generation of oligodendrocytes, a moderate reduction of Olig2 expression at E12.5 cannot fully account for the delay in oligodendrogenesis observed in Shh signaling mutants. On the other hand, enhanced motor neuron production and persistent Ngn2 expression in Shh−/−;Gli3−/− mutants suggest that extended neurogenesis in the ventral progenitor domain likely contributes to the delay in OLP generation. There is increasing evidence indicating that Ngn1/2 regulates neuron–glia fate switching in the central nervous system (Vetter, 2001; Bertrand et al., 2002). Ngn1 has been shown to promote neurogenesis and inhibit astrocyte specification (Sun et al., 2001). Furthermore, extinction of Ngn2 expression appears to be a prerequisite for the generation of OLPs in the ventral spinal cord (Zhou and Anderson, 2002). Intriguingly, Small eye (Sey) mutant spinal cord that is deficient in Pax6 protein also showed delay in OLP development (Sun et al., 1998). Although Ngn2 expression was not directly examined in Sey mutants at the onset of oligodendrogenesis, it was reported that motor neuron development is delayed by nearly 1 day and largely occurred after E10.5, suggesting presence of extended Ngn2 expression in the ventral progenitor domain as in Shh−/−;Gli3−/− mutants.
The generation of Olig2 progenitor domain in Shh−/−;Gli3−/− spinal cord indicates that Gli3 functions as a negative regulator of OLP development in the absence of Shh. However, the role of Gli3 in OLP specification and differentiation in the spinal cord is likely indirect. First, Gli3 expression is not observed in OLPs and becomes restricted to the dorsal spinal cord at E9.5, before the initiation of oligodendrogenesis (Ruiz i Altaba, 1998; Sasaki et al., 1999). Second, the development of OLPs in Gli3−/− embryos is comparable to the control (Figs. 1, 2). Therefore, the observed OLPs in Shh−/−;Gli3−/− spinal cord are likely attributed to the antagonistic role of Gli3 repressor on the Olig2+ ventral progenitor domain during the period when motor neurons are specified.
In addition to the repressor function of Gli3 in the generation of OLPs, analysis of Shh−/−;Gli3−/− spinal cord suggests that Gli3 contributes to Shh signaling in the generation of Nkx2.2+ progenitor domain and subsequent differentiation of OLPs. Although Gli3−/− spinal cord has normal Nkx2.2 expression domain, loss of both Gli2 and Gli3 results in complete abrogation of Nkx2.2 expression (Motoyama et al., 2003). The fact that Gli2−/− spinal cord expresses Nkx2.2 (Ding et al., 1998; Matise et al., 1998), albeit at reduced level, is also consistent with the redundant functions of Gli2 and Gli3 in the generation of Nkx2.2+ progenitors.
The generation and identification of Shh and Gli3 mutant mice were performed as described (Litingtung and Chiang, 2000b).
Analysis of Cell Proliferation
The BrdU in vivo labeling and detection were performed as previously described (Litingtung et al., 1998).
Labelings using antibodies against Isl-1 (Developmental Studies Hybridoma Bank, Iowa City, IA) and phosphohistone were performed on 15-μm tissue sections from OCT-embedded embryos fixed in 4% paraformaldehyde for 2 hr at 4°C as previously described (Litingtung and Chiang, 2000b). For Isl-1 and BrdU double-labeling experiments, E12.5 embryos were collected after 24 hr of initial pulse with BrdU. Shh−/−;Gli3−/− and wild-type embryos were processed and embedded in paraffin blocks. Serial 5-μm sections at the lumbar level were collected onto glass slides. Experimental slides were subject to an antigen retrieval process (Dako, Carpinteria, CA) before immunolabeling with rabbit anti-BrdU and mouse anti-Isl1 antibodies. Fluorescent images were obtained using an LSM510 META laser scanning confocal microscope (Carl Zeiss, Thornwood, NY).
In situ hybridizations were performed as described (Henrique et al., 1995). The following cDNAs were used as templates for synthesizing digoxigenin-labeled riboprobes: PDGFRα (gift of Dr. Christer Betsholtz), Sox10 (gift of Dr. Michelle Southard-Smith), Olig2 (gift of Dr. Hirohide Takebayashi), Nkx2.2 (gift of Dr. John Rubenstein), Ngn2 (gift of Dr. Francois Guillemot), PLP (IMAGE ID#5867358), and MBP (IMAGE ID# 1448814).
We thank John Rubenstein, Hirohide Takebayashi, Christer Betsholtz, Michelle Southard-Smith, and Francois Guillemot for reagents. C.C. was supported by the NIH, and S.O. received a postdoctoral fellowship program from the Korea Science and Engineering Foundation (KOSEF)