Dual origin of spinal oligodendrocyte progenitors and evidence for the cooperative role of Olig2 and Nkx2.2 in the control of oligodendrocyte differentiation

Oligodendrocytes are the myelinating macroglial cells distributed in all regions of the central nervous system (CNS). Despite their wide distribution throughout the entire CNS, recent studies have indicated that oligodendrocytes are derived from a restricted domain of neuroepithelium in the ventral CNS (Warf et al., 1991; Noll and Miller, 1993) under the influence of sonic hedgehog (Shh) signaling (Trousse et al., 1995; Poncet et al., 1996; Pringle et al., 1996; Orentas et al., 1999). In the spinal cord region, expression of early oligodendrocyte genes, such as Pdgfra, Sox10, Olig1, Olig2, Nkx2.2, Plp and O4 antigen, is initially confined to the ventral neuroepithelium adjacent to the Shh-expressing floor plate (Pringle and Richardson, 1993; Yu et al., 1994; Spassky et al., 1998; Ono et al., 1995; Xu et al., 2000). Soon after oligodendrocyte progenitor cells (OLPs) are generated from the ventral ventricular zone, they migrate into the surrounding gray and white matter regions where they undergo rapid proliferation prior to their terminal differentiation (Barres and Raff, 1994; Ono et al., 2001; Xu et al., 2000).

Despite the consensus view on the ventral origin of oligodendroyctes, the precise site of oligodendrocyte generation in the spinal cord remains under intense investigation. In the rodent spinal cord, expression of the oligodendrocyte marker gene Pdgfra is initially mapped to the lower region of the Pax6 gradient but dorsal to the Nkx2.2 domain (Sun et al., 1998). This domain corresponds to the motoneuron precursor domain (pMN domain) which lies dorsal to the Nkx2.2+ p3 domain but ventral to the Irx3+Nkx6.1+ p2 domain (Briscoe et al., 1999; Briscoe et al., 2000). Thus, it is believed that motoneurons and oligodendrocytes are generated from the same neuroepithelial domain but during different time windows (Richardson et al., 1997; Richardson et al., 2000; Spassky et al., 2000). In support of this hypothesis, expression of two novel oligodendrocyte-specific genes, Olig1 and Olig2, also appears to be mapped to the pMN domain at early stages (Lu et al., 2000; Zhou et al., 2000; Takebayashi et al., 2000).

In recent studies of the developing chicken spinal cord, other investigators have found results that are not entirely supportive of this hypothesis, and a different site of origin of oligodendrocytes has been suggested. In the chicken embryos, expression of early markers of the oligodendrocyte lineage, such as Pdgfra and O4 antigen, was initially detected in the Nkx2.2+ neuroepithelium (Xu et al., 2000; Soula et al., 2001). The OLPs that migrate away from the ventricular zone retain the Nkx2.2 expression and gradually acquire expression of late oligodendrocyte markers such as GalC and proteolipid protein (PLP) (Xu et al., 2000). Based on these observations, it was proposed that oligodendrocyte progenitors could originate from the Nkx2.2+ p3 domain (Xu et al., 2000; Soula et al., 2001), and that oligodendrocytes and motoneurons may not share the same lineage in embryonic chicken spinal cord (Soula et al., 2001).

These contradictory observations in mouse and chicken spinal cords have raised several important possibilities on the origin and lineage of oligodendrocyte progenitors in the spinal cord. One possibility is that oligodendrocyte progenitors may arise from different neuroepithelial domains in rodents and birds. It is conceivable that OLPs originate from the pMN domain in mammals, but from the more ventral p3 domain in avians. The second possibility is the dual origin of oligodendrocytes, i.e. that two distinct populations of oligodendrocyte progenitors arise from distinct sites of neuroepithelium in the same spinal cord tissue. It is possible that the Olig2+ progenitor cells might originate from the pMN domain, whereas the Nkx2.2+ progenitors could be generated from the ventral p3 domain. Finally, it is also possible that the Olig2+ progenitor cells and the Nkx2.2+ progenitors may represent the same population of progenitors that arise from a merged region of the Nkx2.2 and Olig2 domains at later stages of spinal cord development.

To investigate these possibilities, including the relationship of Olig2+ and Nkx2.2+ OLPs, we first compared the expression of Olig2, Nkx2.2 and other oligodendrocyte markers in embryonic chicken and mouse spinal cords. Our expression studies revealed that at early stages of spinal cord development, the Olig2+, Pdgfra+ and Sox10+ OLPs originate from the pMN domain of the ventral neuroepithelium in both mouse and chicken. Interestingly, this population of OLPs gains Nkx2.2 expression before their migration in chicken, but after migration in mouse. In addition, the Nkx2.2+ p3 domain can also produce OLPs which are initially Nkx2.2+/Olig2–, but appear to gain Olig2 expression as they migrate and differentiate. At later stages of embryogenesis, nearly all OLP cells in the spinal cord parenchyma co-express the Nkx2.2 and Olig2 transcription factors. The co-expression of Nkx2.2 and Olig2 in OLPs precedes, and is necessary for, OLP differentiation and myelin gene expression. Inhibition of expression of these two transcription factors in dissociated culture by antisense oligonucleotides has an additive inhibitory effect on OLP differentiation.

Fertilized chick eggs (White Horn, SPAFAS) were incubated at 38°C in a humidified incubator and embryos were staged according to the criteria set by Hamburger and Hamilton (Hamburger and Hamilton, 1951). Anti-Nkx-2.2 hybridoma supernatants were obtained from the Developmental Studies Hybridoma Bank (University of Iowa, Iowa City). The Alexa-488 or Alexa-594 conjugated secondary antibodies were obtained from Molecular Probes.

Embryos from various stages of chicken development were fixed in 4% paraformaldehyde at 4°C overnight. Tissue preparation and in situ hybridization with digoxigenin-labeled riboprobes were performed according to Schaeren-Wiemers and Gerfin-Moser (Schaeren-Wiemers and Gerfin-Moser, 1993) with minor modifications.

Spinal cord tissues from the thoracic or brachial regions were isolated from day 3-12 chicken embryos, fixed in 4% paraformaldehyde and sectioned on a cryostat. For immunofluorescence, slides were incubated with anti-Olig2 polyclonal antibody (1:3000 dilution), anti-Nkx2.2 (1:10) or anti-GalC (5 μg/ml from Boeringer Mannheim) overnight at 4°C. Sections were then washed five times with phosphate-buffered saline (PBS), incubated with Alexa-488- or Alexa-594-conjugated secondary antibodies (50 μg/ml). Fluorescent images were collected by Nikon epifluorescence microscope. The combined immunohistochemistry and in situ hybridization was previously described in Schaeren-Wiemers and Gerfin-Moser (Schaeren-Wiemers and Gerfin-Moser, 1993)

Spinal cord tissues were isolated from day 5 chicken embryos in 1×PBS, minced and then physically dissociated by repeated pipetting with fine-tip plastic pipettes. The dissociated cells were subsequently grown on poly-L-Lysine coated cover-glass in DMEM + 5% fetal bovine serum +N2 supplement (Gibco) for 4-5 hours at 37°C followed by oligonucleotide treatment, which PAGE-purified sense (for Nkx2.2) or antisense (for Nkx2.2 or Olig2) phosphorothiate oligodeoxynucleotides (p-ON, from Intergrated DNA Science) was added to culture medium at the final concentration of 1 μM each. The oligonucleotide sequences are as follows: Nkx2.2 sense (5′-GCTGTTCAGACGCTGCCT-3′), Nkx2.2 antisense (5′-AGGCAGCGTCTGAACAGC-3′) and Olig2 antisense (5′-TCATCTGCTTCTTGTCCT-3′).

Five days after treatment, cells were fixed for 10 minutes in 4% paraformaldehyde, washed twice with PBS and blocked with 5% goat serum. The coverslips were then incubated overnight with anti-GalC hybridoma supernatant overnight at 4°C. After several washes with PBS, Alexa-594-conjugated goat secondary antibodies and DAPI were applied for 1 hour at room temperature. The coverslips were then washed three times with PBS and mounted for immunofluorescent detection. The GalC+ cells and total cell numbers were scored under Nikon epifluorescent microsope. For each score, three coverslips have been used and at least 10 fields and about 2000-3000 cells have been counted for each coverslip. Only the relative number of GalC+ cells from various treatments was plotted, with the Nkx2.2 sense control as 100.

The effects of antisense treatments on PLP expression were assessed by in situ hybridization with Plp riboprobe as described by Schaeren-Wiemers and Gerfin-Moser (Schaeren-Wiemers and Gerfin-Moser, 1993). PLP+ cells were counted under light microscope and the number of positive cells from each 10 field (n=3) was plotted.

E13.5 mouse spinal cords were bisected into the dorsal and ventral explants which were dissociated and cultured separated in NEP basal with 35 ng/ml basic fibroblast growth factor (FGF) for 2 days or 5 days. Cells were then fixed in 2% paraformaldehyde and processed for double immunostaining with anti-Nkx2.2 monoclonal antibody (1:10) and anti-Olig2 (1:3000) (Takebayashi et al., 2000) polyclonal antibody as previously described (Rao et al., 1998; Qi et al., 2001).

Previous studies in rodents have demonstrated that OLPs are produced from Olig2+ pMN domain which is dorsal to the Nkx2.2+ p3 domain in the spinal cord (Sun et al., 1998; Lu et al., 2000; Zhou et al., 2000; Takebayashi, 2000). However, oligodendrocytes appear to be generated from the more ventral Nkx2.2+ p3 domain in chicken (Xu et al., 2000; Soula et al., 2001). To investigate this potential species difference in the origin of oligodendrogenesis, we examined in detail the expression of two early oligodendrocyte marker genes, Sox10 and Pdgfra, in relation to that of Olig2 and Nkx2.2 in chicken.

During neurogenesis, Olig2 is precisely expressed in the pMN domain dorsal to the Nkx2.2+ p3 domain during neurogenesis (Fig. 1A). In E3 and E4 chick spinal cord, the MNR+ motoneurons are exclusively produced from the entire Olig2+ domain of neuroepithelium (Fig. 1B,C). Comparison of the expression of Olig2 with Sox10 or Pdgfra on immediately adjacent sections from E7-9 chick spinal cord revealed that Sox10+ and Pdgfra+ OLPs are situated within or immediately adjacent to the Olig2 domain of neuroepithelium (Fig. 1D-G). Thus, the Pdgfra+ and Sox10+ oligodendrocyte progenitors originate from the Olig2+ motoneuron precursor domain in chicken.

Intriguingly, when E7 chicken spinal cord was simultaneously stained with Nkx2.2 and Sox10 or Pdgfra, we found that expression of these two OLP markers falls into the Nkx2.2+ neuroepithelium (Fig. 1H,I), instead of being dorsal to the Nkx2.2+ cells. One possible explanation is that Nkx2.2 expression is upregulated in the Olig2 domain before oligodendrogenesis. To examine this possibility, we performed double labeling experiments with Olig2 and Nkx2.2 in chicken spinal cord during the crucial stages of oligodendrogenesis in chick. Spinal cord sections prepared from E3 to E12 chicken embryos were subjected to Olig2 in situ hybridization followed by anti-Nkx2.2 immunohistochemistry. At E3-4, Olig2 and Nkx2.2 expression is restricted in adjacent non-overlapping domains of neuroepithelial cells, with each domain spanning a width of four to five cell bodies (Fig. 1A, Fig. 2A). At E5, Nkx-2.2 expression appears to expand dorsally into the Olig2 domain, whereas both the width and the position (relative to the floor plate) of the Olig2 domain remain relatively unchanged (Fig. 2B). By E6, the entire Olig2 domain gained a weak expression of Nkx2.2. A few Nkx2.2+ but Olig2- progenitor cells start to migrate away from the Nkx2.2+/Olig2- domain (p3 domain), first ventrally and then dorsally (Fig. 2C). From E7 to E10, the p3 domain gradually moves down along the floor plate as the central canal decreases and the floor plate elongates (Fig. 2D-G). As a result, the original pMN domain is concomitantly dragged downwards along the central canal (compare Fig. 2D-G with the p3 domain). At E10, cells in both the pMN domain and p3 domain are rapidly decreased. In the ventricle, only a small number of Olig2+/Nkx2.2+ neuroepithelial cells remain associated with the reduced central canal (Fig. 2G). By E12, expression of Olig2 and Nkx2.2 is nearly completely absent in the ventricular cells of the spinal cord (Fig. 2H).

Since the Olig2 domain appears to gain expression of Nkx2.2 before the onset of migration of Sox10+, Pdgfra+ and Olig2+ progenitors, it is expected that these migratory OLP cells in the surrounding regions would co-express Nkx2.2. Starting around E6-E7, a few Olig2+ OLP start to migrate ventrolaterally (Fig. 2C,D). As development proceeds, an increasing number of Olig2+ migratory OLP cells was detected in the surrounding gray and white matter. Closer examination revealed that at these stages, all migratory Olig2+ progenitors in the surrounding regions co-express Nkx2.2. However, many of the Nkx2.2+ progenitors in the gray matter, especially those adjacent to the ventricle and the floor plate, do not express Olig2 (Fig. 2C,D,F,G,J). These Nkx2.2+/Olig2– progenitors are generated slightly earlier than Olig2+ OLPs and migrate first ventrally and then dorsolaterally (Fig. 2C-D). This population of OLPs is likely to be produced from the p3 domain.

Interestingly, in the gray and white matter region further away from the ventral ventricular zone, the proportion of Nkx2.2+/ Olig2– is decreased with time, whereas the percentage of Nkx2.2+/Olig2+ cells is increased. By E12, nearly all Nkx2.2+ progenitor cells are positive for Olig2 (Fig. 2H). Moreover, the staining intensity of Olig2 also appears to become stronger at E8 and later stages (Fig. 2I-L). The increasing intensity of Olig2 expression, together with the observation that an increasing percentage of Nkx2.2+ progenitors express Olig2, strongly suggests that the Nkx2.2+ OLPs may gain expression of Olig2 during the process of migration and maturation.

In search of evidence for the capability of migratory Nkx2.2+ OLPs to gain Olig2 expression, we performed a similar double labeling experiment on other regions of the CNS, including the hindbrain, midbrain and forebrain, at the critical stages of oligodendrogenesis. Transverse sections of the brain tissues prepared from E6-8 chicken embryos were double-labeled with Nkx2.2 and Olig2. We found that in the hindbrain region, many Nkx2.2+ OLPs acquired Olig2 expression during their migration process. Intriguingly, in this region, only Nkx2.2 is initially expressed in the ventral neuroepithelium at the onset of oligodendrogenesis (Fig. 3A-D). Olig2 is not expressed in the ventricular zone dorsal to the Nkx2.2 domain as is seen in the spinal cord (Fig. 3A-D). However, Olig2 expression is clearly detected in groups of Nkx2.2+ cells in the subventricular zone and in individual migratory cells at E6-E7 (arrows in Fig. 3B,D). By E8, Nkx2.2+ cells are dispersed into surrounding regions, where many Nkx2.2 cells co-express Olig2. However, a few Nkx2.2+ migratory cells adjacent to the ventral midline remain Olig2 negative (arrowheads in Fig. 3F). These results provide further evidence that Nkx-2.2+ oligodendrocyte progenitors could acquire Olig2 expression during the process of migration and proliferation.

To investigate whether the origin of oligodendrogenesis is conserved between mouse and chicken, we re-examined the expression relationship of early OLP markers with Olig2, Nkx2.2 and other neural identity genes. Consistent with previous suggestions (Takeyabashi et al., 2000), we confirmed that during neurogenesis stages, Olig2 is precisely expressed in the pMN domain, flanked ventrally by Nkx2.2 expression and dorsally by Irx3 expression (Fig. 4A,B) (Briscoe et al., 2000). However, unlike in chicken, Nkx2.2 is not upregulated in the Olig2 domain at the onset stage of oligodendrogenesis. At E12.5, Olig2 and Nkx2.2 are still expressed in the adjacent domains of ventral neuroepithelium with no or little overlapping (Fig. 4C). At this stage, Pdgfra + OLPs are exclusively born from the Olig2+ domain, as expression of Pdgfra is confined within or immediately adjacent to, the Olig2+ neuroepithelium (Fig. 4D-E), dorsal to the Nkx-2.2+ neuroepithelial cells (Fig. 4F).

Although Nkx2.2 is not upregulated in the Olig2+ neuroepithelium at the early stage of oligodendrogenesis, it is possible that these two domains are merged at later stages. To examine this possibility, we performed detailed expression studies with Nkx2.2 and Olig2 on spinal cord sections prepared from E12.5 to E14.5 mouse embryos by in situ hybridization. As described above, at E12.5, Olig2 and Nkx2.2 are expressed in adjacent non-overlapping domains. At this stage, a few Nkx2.2+ cells start to migrate out of the p3 domain into the adjacent gray matter (Fig. 5A,B). At E13.5, the Olig2+ OLPs start to migrate away from the pMN domain into the dorsal and ventral spinal cord. The number of Nkx2.2+ OLPs in the ventral gray matter surrounding the p3 domain is also increased (Fig. 5C,D). At the same time, the Nkx2.2 expression in the neuroepithelial cells starts to expand dorsally into the Olig2 domain, similar to the observations in embryonic chicken spinal cord. At E14.5, Olig2 domain and Nkx2.2 domain are almost completely merged (Fig. 5E,F). Thus, the dorsal expression of Nkx2.2 into the pMN domain occurs about two days after OLPs start to migrate out.

Contrary to the embryonic chicken spinal cord, Nkx2.2 is not co-expressed in the Olig2+ migratory OLPs in mouse during the early stages of oligodendrogenesis. By E14.5 when the pMN domain expresses Nkx2.2, Olig2+ progenitor cells have already spread into the entire spinal cord (Fig. 5C-F). Although Nkx2.2 is initially not expressed in these Olig2+ OLPs, it is conceivable that this population of OLPs could acquire Nkx2.2 expression at later stages of oligodendrocyte development, as suggested by the co-expression of Olig2 and Nkx2.2 in chicken OLPs that are derived from the pMN domain. To examine this possibility, we performed Nkx2.2 in situ hybridization on postnatal day one (P1) mouse spinal cord sections followed by anti-Olig2 immunohistochemistry. As shown in Fig. 5G,H, most, if not all, of the Olig2+ cells are also positive for Nkx2.2 (arrows), although expression of Nkx2.2 in some Olig2+ cells is fairly weak (arrowheads), probably because they are in the early phase of gaining Nkx2.2 expression. Similarly, all Nkx2.2+ cells are also immunoreactive for anti-Olig2.

To test further whether Olig2+ OLPs can gain Nkx2.2 expression after migration, spinal cords from E13.5 mouse embryos were bisected and the dorsal halves were used for dissociated cell culture. At this stage, Nkx2.2+ OLPs are restricted to the ventral half (Fig. 5C,D), but the Olig2+ OLPs have already migrated into the dorsal cord. Two hours after dissociation, no Olig2+/Nkx2.2+ double positive OLP cells were observed from the dorsal culture (data not shown). However, after 2 days in vitro (2 DIV) culture, about 45% of Olig2+ cells (n=60) became immunoreactive to Nkx2.2 although Nkx2.2 expression in many of these cells was relatively weak (Fig. 6A-C). By 5 DIV, the percentage of Nkx2.2+ cells in Olig2+ OLPs increased to 85% (n=130), and the intensity of Nkx2.2 immunostaining also appeared to be increased (Fig. 6D-F). These experiments provide direct evidence that in rodents, the Olig2+ OLPs can acquire Nkx2.2 expression after they migrate away from the ventricular zone.

To summarize our expression analyses, we found that Pdgfra+, Sox10+ and Olig2+ OLPs are produced exclusively from the Olig2+ pMN domain in both mouse and chicken. The Nkx2.2+ p3 domain can also produce OLPs which are initially Nkx2.2+/Olig2-, but might gain Olig2 expression as they migrate. The pMN domain gains Nkx2.2 expression during oligodendrogenesis but at different time window in chicken and mouse. In chicken, Nkx2.2 is upregulated in the pMN domain before the migration of Olig2+ and Pdgfra+ OLPs; however, in the mouse the upregulation occurs 2 days after the onset of OLP migration. Thus, the Olig2+ OLPs gain Nkx2.2 expression before their migration in chicken, but after migration in mouse.

Based on the dynamic expression of Olig2 and Nkx2.2 in neuroepithelium and OLP cells, we hypothesize that there are two separate populations of OLPs at early stages of mouse spinal cord development. The Olig2+/Pdgfra+/Nkx2.2- OLPs are produced from the pMN domain, whereas the Nkx2.2+/Olig2–/Pdgfra– OLPs are generated from the more ventral p3 domain. If our hypothesis on the dual origins of OLPs is true, the number of these two OLP populations would be differentially affected by the mutation of PDGFA, an oligodendrocyte mitogen required for proliferation of Pdgfra+ OLPs (Fruttiger et al., 1999).

To examine this possibility, adjacent spinal cord sections from E13.5 embryos to P7 pups were examined for the expression of Pdgfra, Olig2 and Nkx2.2. At E13.5, numerous Pdgfra+ and Olig2+ progenitor cells have already spread into the ventral and dorsal spinal cord in the wild-type embryos (see Fig. 9A,B). However, only few Pdgfra+ and Olig2+ cells are detected within or adjacent to the ventricular zone in the mutants (Fig. 7D,E). The drastic reduction in the number of Pdgfra+ and Olig2+ OLPs in the mutants is also detected throughout the later stages of animal development (Fig. 7, Fig. 8). The parallel delay and reduction of the Pdgfra+ cells and Olig2+ cells is consistent with our hypothesis that they represent the same population of OLPs, and that Pdgfra is likely to be transiently expressed in Olig2+ cells.

By contrast, the number and distribution of Nkx2.2+ cells are not affected in the mutants at early stages. At E13.5 and E16.5, similar patterns of Nkx2.2 expression in the ventral gray matter are observed in the wild-type and mutant embryos (Fig. 7C,F,I,M). At P0, Nkx2.2 expression in the ventral gray matter is not significantly affected in the mutants (Fig. 8C,D). However, the number of Nkx2.2+ cells in the white matter is dramatically reduced (Fig. 8D). The parallel reduction of Pdgfra+/Olig2+ OLPs and Nkx2.2+ OLPs in the white matter is consistent with our hypothesis that Olig2+/Pdgfra+ cells can acquire Nkx2.2 expression after migration (Fig. 8C,G). At P7, expression of Nkx2.2 is greatly reduced in the entire spinal cord, especially in the white matter (Fig. 8K,O), similar to that of Pdgfra and Olig2.

We next investigated how the differential reduction of two populations of OLPs in the PDGFA mutants affects oligodendrocyte differentiation and distribution using myelin basic protein (MBP) as a marker. MBP expression can be observed in the ventral gray matter of both normal and mutant embryos as early as E16.5 (Fig. 7J,N). The PDGFA mutation does not reduce MBP expression at this stage. In the wild-type P0 animal, MBP is expressed in both gray matter and to a larger extent in the white matter (Fig. 8D). In the mutants, MBP expression in the gray matter is more or less the same as in the wild type, whereas in the white matter, it is drastically reduced (Fig. 8H), similar to the preferential reduction of Nkx2.2 expression in the white matter (Fig. 8H). By P7, expression of MBP in the mutants is decreased in both the gray and white matter (Fig. 8L,P).

Comparison of expression of MBP, Nkx2.2 and Olig2 in both wild-type and mutant animals revealed that the expression pattern of MBP appears to closely follow the overlapping region of Nkx2.2 and Olig2, especially from E16.5 to P0. At P7, the relatively large number of MBP+ oligodendrocytes in the mutants could be due to the slow but steady differentiation and accumulation of oligodendrocytes in the white matter after birth.

The co-expression of Nkx2.2 and Olig2 in OLPs in both chicken and mouse suggests their important and perhaps collaborative role in the control of oligodendrocyte differentiation. To examine this possibility, we tested the effects of inhibition of their expression on oligodendrocyte differentiation in culture by antisense approach. Dissociated cells prepared from E5 chicken spinal cord were plated on cover slips at the same density and cultured for 5 days in the presence of synthetic sense or antisense phosphorothiate oligonucleotides (at the final concentration of 1 μM for each) derived from the chicken Nkx2.2 or Olig2 sequences. Cells were then examined for GalC expression (Ranscht et al., 1982; Bansal et al., 1989) by immunofluorescence or PLP expression (Dubois-Dalcq et al., 1986; Knapp et al., 1987) by in situ RNA hybridization as indicators of oligodendrocyte differentiation.

After 5 days in vitro culture, a significant decrease in the number of GalC+ and PLP+ oligodendrocytes was observed in dissociated spinal cord following Olig2 or Nkx2.2 antisense treatment compared with the Nkx2.2 sense control group (Fig. 9). When the Nkx2.2 and Olig2 antisense oligonucleotides were applied together, we detected a further decrease of GalC+ and PLP+ oligodendrocytes, indicating an additive inhibitory effect on GalC and MBP expression. Under the same conditions, Nkx2.2 expression in dissociated culture is similarly decreased in the cells treated with Nkx2.2 antisense oligonucleotides, but not with the Olig2 antisense (data not shown), indicating the efficiency and specificity of antisense treatment. These observations, together with the fact that Olig2 and Nkx2.2 are co-expressed in OLPs before oligodendrocyte differentiation, strongly suggest that these two transcription factors may cooperate to control oligodendrocyte differentiation.

In this study, we investigated the relationship of Olig2+ and Nkx2.2+ oligodendrocyte progenitors by comparing the expression pattern of Olig2 and Nkx2.2 in the embryonic chicken and mouse spinal cords during the critical stages of oligodendrogenesis. At the early stages of oligodendrogenesis, Olig2+, Pdgfra+ and Sox10+ OLPs migrate out from the Olig2+ pMN domain in both chicken and mouse (Fig. 1, Fig. 4). Pdgfra+ cells, Olig2+ cells and Sox10+ cells are likely to represent the same population of OLPs, as suggested by the parallel delay and reduction of production of these cell types in PDGFA mutants (Fig. 7, Fig. 8; data not shown). Soon after they are generated, this population of OLPs rapidly spreads into all regions of the spinal cord and later differentiates into mature oligodendrocytes (Hall et al., 1996; Kuhlbrodt et al., 1998; Lu et al., 2000; Zhou et al., 2000).

At the same time or slightly earlier stages, many Nkx2.2+ but Olig2– cells also migrate out into the surrounding gray matter area. In the chicken embryos, these Nkx2.2+ cells quickly migrate dorsally and laterally into both gray and white matter (Fig. 2D-F). Our previous studies have indicated that all Nkx2.2+ cells are OLPs, but not astrocytes or neurons (Xu et al., 2000). Many Nkx2.2+ cells from the p3 domain start to express O4 before they migrate away from the ventricular zone (Soula et al., 2001). A recent study further confirmed that some Nkx2.2+/Olig2– OLPs are indeed immunoreactive to O4 antigen (Zhou et al., 2001). In the mouse, the Nkx2.2+ cells that are generated from the p3 domain migrate relatively slowly and remain in the ventral gray matter until at least E16.5. The evidence to indicate that these Nkx2.2+/Olig2– cells to become oligodendrocyte is not as strong in mouse as it is in chicken. One piece of evidence is that all Nkx2.2+ cells in rodents express oligodendrocyte markers, but not neuronal or astrocytic markers, both in vivo and in vitro (Qi et al., 2001) (Fig. 5G,H). Furthermore, in the dissociated cell culture from the Nkx2.2-rich ventral halves of the E13.5 mouse spinal cord, every Nkx2.2+ cell became immunoreactive to anti-Olig2 after 2 days in vitro (data not shown).

Based on the present and previous studies, we propose that oligodendrocytes can be generated from both the pMN domain and p3 domain of ventral neuroepithelium during the early stages of oligodendrogenesis in both mouse and chicken. The hypothetical model is proposed in Fig. 10. In this model, the pMN domain gives rise to the well-characterized Olig2+/Pdgfra+/Sox10+ OLPs, whereas the p3 domain gives rise to Nkx2.2+/Olig2–/Pdgfra– OLPs. The generation of the Olig2+, Pdgfra+ and Sox10+ OLPs from the pMN domain in both mouse and chicken indicates that the lineage relationship of somatic motoneurons and oligodendrocytes is evolutionally conserved (Richardson et al., 1997; Richardson et al., 2000). This is in contrast to the recent interpretation of distinct origin sites for oligodendrocytes and somatic motoneurons in the chicken (Soula et al., 2001).

Although the Olig2+ OLPs arise from the pMN domain dorsal to the Nkx2.2+ p3 domain, this population of OLPs can acquire Nkx2.2 expression either before migration in chicken or after migration in rodents. Before oligodendrogenesis, Olig2 and Nkx2.2 are expressed in adjacent non-overlapping domains in both chicken and mouse (Fig. 1, Fig. 4) (Lu et al., 2000; Zhou et al., 2000). Interestingly, at stages when neurogenesis is switched to gliogenesis, the expression boundary of Olig2 and Nkx2.2 breaks down and the Nkx2.2 expression expands dorsally into the Olig2 domain. In chick, Nkx2.2 expansion occurs between E5 and E7, right before generation of Olig2+/Pdgfra+ OLPs from the pMN domain (Fig. 2). Thus, all migratory Olig2+ and Pdgfra+ OLPs in chicken are also Nkx2.2+ at the beginning of their migration. However, merging of these two domains occurs relatively late in mouse, about 2 days after OLPs have already migrated out into the surrounding regions. Therefore, many of the Olig2+ OLPs at the early stages of oligodendrogenesis (before E16.5) do not express Nkx2.2. These observations could certainly explain the apparent species difference in the pattern of Nkx2.2 expression during early stages of spinal cord development (Xu et al., 2000; Qi et al., 2001). In the chicken, Nkx2.2 is expressed in OLPs derived from both the pMN domain and the p3 domain from the beginning. However, Nkx2.2 expression is only initially restricted to the ventral region where the OLPs derived from the p3 domain.

Although the Olig2+/Pdgfra+ OLPs originating from the pMN domain is initially Nkx2.2-negative in rodents, this population of OLPs appears to acquire Nkx2.2 expression after their emigration into the spinal cord parenchyma based on the following observations. First, around the time of birth, nearly all the Olig2+ OLPs in white matter co-express Nkx2.2, although the expression level in some cells is relatively low (Fig. 5G,H). Second, the OLPs derived from E13.5 dorsal spinal cord gradually gain Nkx2.2 expression with time in dissociated culture (Fig. 6). Similarly, the percentage of Nkx2.2+ cells in immunopurified A2B5+ cells also increases with time in vitro (Qi et al., 2001) (data not shown). Third, in the PDGFA mutant embryos, reduction of Olig2+/Pdgfra+ OLPs in white matter is accompanied by a decrease in Nkx2.2 expression at P0, although Nkx2.2 expression in ventral gray matter is not affected (Fig. 8).

The OLPs that originate from the p3 domain initially express Nkx2.2, but not Olig2 and Pdgfra (Fig. 2, Fig. 7), which could explain why the number of Nkx2.2+ OLPs in the ventral gray matter is not reduced in the Pdgfa mutants at P0 and earlier stages. There is some good evidence to suggest that this population of OLPs might gain Olig2 expression during or after migration. In chicken spinal cord, the p3 domain actively produces Nkx2.2+/Olig2– OLPs between E6 and E9. As these OLPs are dispersed into the surrounding gray and white matter region, the percentage of Nkx2.2+/Olig2– OLPs decreases from the ventral ventricular zone to the white matter at E8-10 (Fig. 2). Similarly, the percentage of Olig2+ cells in the white matter increases with time between E8 and E10. Further evidence for the capability of Nkx2.2+ migratory OLPs to gain Olig2 expression comes from the subventricular expression of Olig2 directly beneath the Nkx2.2+ neuroepithelial cells in the developing hindbrain (Fig. 3). However, we are also aware of the alternative explanation that the increase of Olig2+ cells in the embryonic spinal cord results from preferential proliferation of the Olig2+/Nkx2.2+ population that arises from the pMN domain.

In the mouse, the p3-derived Nkx2.2+ OLP cells might similarly gain Olig2 expression after they migrate out into the gray matter. There is only some weak evidence to support this hypothesis. First, the Olig2 expression in the Nkx2.2+ region is higher than the rest of the spinal cord at E16.5 in both the wild-type and Pdgfa mutant embryos (Fig. 7). At birth, all the Nkx2.2+ cells in this region, and in the white matter as well, are positive for Olig2 expression (Fig. 5H, data not shown). Second, in the dissociated cell culture with the ventral halves of the E13.5 mouse spinal cord, every Nkx2.2+ cell is immunoreactive for anti-Olig2 after two days in vitro (data not shown). However, the evidence for the acquisition of Olig2 expression by this population of OLPs is still indirect, and further immunological and genetic labeling studies are required to further confirm this hypothesis.

Side-by-side comparison of expression of Olig2, Nkx2.2 and Mbp revealed that expression of MBP closely follows that of Nkx2.2 during oligodendrocyte differentiation. In E16.5 mouse spinal cord, MBP expression is detected only in the ventral gray matter where Nkx2.2 is expressed. At P0, expression of both Nkx2.2 and MBP is observed in the white matter. Reduction of Nkx2.2 expression in the white matter in the Pdgfa mutants is accompanied by the reduction of MBP expression in this region. As nearly every Nkx2.2+ cell is co-labeled with Olig2 at these stages (Fig. 5G,H; data not shown), it is reasonable to speculate that myelin gene expression and oligodendrocyte differentiation may be initiated by the interaction of Nkx2.2 and Olig2. There are at least two lines of evidence to support this concept. First, terminal differentiation of oligodendrocytes appears to require simultaneous expression of these two transcription factors, as suggested by the additive inhibitory effects of the Nkx2.2 and Olig2 antisense treatments on PLP gene expression (Fig. 9). Inhibition of either Nkx2.2 or Olig2 expression is accompanied by a smaller but significant reduction of GalC+ and PLP+ cells in culture. The inhibition of oligodendrocyte differentiation by Nkx2.2 antisense treatment is in agreement with our previous findings that mutation of the Nkx2.2 gene can cause a dramatic reduction of MBP and PLP expression (Qi et al., 2001). Second, expression of either Olig2 or Nkx2.2 alone is not sufficient for oligodendrocyte differentiation in vivo (Zhou et al., 2001) (J. C., H. F. and M. Q., unpublished). However, co-transfection of Nkx2.2 and Olig2 can result in ectopic and precocious oligodendrocyte differentiation in embryonic chicken spinal cord (Zhou et al., 2001). Based on these observations, it appears that co-expression of the Olig2 and Nkx2.2 is both necessary and sufficient for oligodendrocyte differentiation and myelin gene expression.

If co-expression of Olig2 and Nkx2.2 is directly responsible for myelin gene expression, we would predict a species difference in the expression of MBP as Nkx2.2 expression is upregulated much earlier in the pMN-derived OLPs in the chicken than in the mouse. Our preliminary results confirm this species difference. In the chick, MBP expression starts relatively early (E8-9) and is initially detected in white matter and to a lesser extent in the pMN domain (H. F. and M. Q., unpublished) in which Nkx2.2 expression is upregulated (Fig. 2). In mouse, MBP+ cells are initially observed in the ventral gray matter, the region where Olig2 and Nkx2.2 are co-expressed (Fig. 7). Thus, MBP expression closely follows the co-expression of Olig2 and Nkx2.2 in the chicken, as in the mouse. It is conceivable that the pace and pattern of myelin gene expression could be regulated during evolution by controlling the timing and location of the co-expression of these two transcription factors.

It is worthwhile mentioning that co-expression of Olig2 and Nkx2.2 is not required for the initial specification of OLPs. In the embryonic mouse spinal cord, OLP production from the pMN and p3 domain occurs prior to merging of these two domains. In the hindbrain of chicken embryo, Olig2 is not expressed in the ventricular zone when Nkx2.2+ OLPs are born (Fig. 5). Moreover, in the Nkx2.2 mutants, production of Olig2+/Pdgfra+ OLPs is normal or even slightly increased, although their terminal differentiation is greatly reduced and delayed (Qi et al., 2001). The residual expression of Mbp and Plp in the Nkx2.2 mutants might imply that the function of Nkx2.2 could be weakly compensated by a related unidentified transcription factor.

We thank Drs David Anderson, Chuck Stiles and C.C. Hui for providing cDNA probes. We are particularly grateful to Dr David Anderson for providing chicken Olig2 before publication. We also thank Dr David Stapp and the anonymous reviewers for their insightful comments and suggestions. The anti-Nkx2.2 hybridoma cells were obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Iowa City. This study is supported by NSF (IBN-9808126), NIH (NS 37712) and National Multiple Sclerosis Society.