Oligodendrocyte progenitor cell numbers and migration are regulated by the zebrafish orthologs of the NF1 tumor suppressor gene

Abstract

Neurofibromatosis type 1 is the most commonly inherited human cancer predisposition syndrome. Neurofibromin (NF1) gene mutations lead to increased risk of neurofibromas, schwannomas, low grade, pilocytic optic pathway gliomas, as well as malignant peripheral nerve sheath tumors and glioblastomas. Despite the evidence for NF1 tumor suppressor function in glial cell tumors, the mechanisms underlying transformation remain poorly understood. In this report, we used morpholinos to knockdown the two nf1 orthologs in zebrafish and show that oligodendrocyte progenitor cell (OPC) numbers are increased in the developing spinal cord, whereas neurons are unaffected. The increased OPC numbers in nf1 morphants resulted from increased proliferation, as detected by increased BrdU labeling, whereas TUNEL staining for apoptotic cells was unaffected. This phenotype could be rescued by the forced expression of the GTPase-activating protein (GAP)-related domain of human NF1. In addition, the in vivo analysis of OPC migration following nf1 loss using time-lapse microscopy demonstrated that olig2-EGFP⁺ OPCs exhibit enhanced cell migration within the developing spinal cord. OPCs pause intermittently as they migrate, and in nf1 knockdown animals, they covered greater distances due to a decrease in average pause duration, rather than an increase in velocity while in motion. Interestingly, nf1 knockdown also leads to an increase in ERK signaling, principally in the neurons of the spinal cord. Together, these results show that negative regulation of the Ras pathway through the GAP activity of NF1 limits OPC proliferation and motility during development, providing insight into the oncogenic mechanisms through which NF1 loss contributes to human glial tumors.

INTRODUCTION

Neurofibromatosis type 1 is a dominantly inherited autosomal disease that affects 1 in 3500 individuals worldwide. Inactivating mutations of the neurofibromin (NF1) gene responsible for the disease can arise in a diverse set of tissues, producing symptoms that range from mild to severe, the most prevalent of which include café-au-lait pigment spots in the skin and the cutaneous development of neurofibromas from Schwann cells of the peripheral nervous system (PNS). Plexiform neurofibromas—non-circumscribed, thick, irregular and invasive benign tumors composed of Schwann cells, fibroblasts, mast cells and vascular components—often occur in this disease and are a major cause of disfigurement because they invade essential structures. NF1 loss is the most common syndrome predisposing to tumorigenesis in humans, increasing the risk for the development of low grade, pilocytic optic pathway gliomas of the central nervous system (CNS) (1), as well as malignant peripheral nerve sheath tumors (2) and high-grade astrocytomas (3). Recently, a large-scale genomic analysis identified somatic mutations and loss of heterozygosity of NF1 in 25% of sporadic glioblastoma cases (4).

The NF1 gene encodes a very large 2818 amino acid cytoplasmic protein (5). To date, the principal domain identified and analyzed in this protein is a 360 amino acid GTPase-activating protein (GAP)-related domain (GRD) whose activity is known to hydrolyze GTP to GDP bound to Ras (5,6). The loss of GRD function in NF1-deficient malignant peripheral nerve sheath tumors leads to increased Ras activity, which in turn activates downstream signaling cascades including the Ras/ERK and PI3K/Akt pathways, providing putative underlying mechanisms for the tumor suppressor function of NF1 in these cancers. Studies using murine models suggested that Nf1 loss of function in specific cell populations (e.g. Schwann cells), possibly in combination with a heterozygous microenvironment (e.g. mast cells), was critical for the formation of plexiform neurofibromas (7–9). However, conflicting data supporting the micro-environmental contribution of NF1 may reflect both tissue-specific and developmental timing effects (10,11).

Astrocytes and oligodendrocytes comprise the main glial cell types in the CNS, the latter being responsible for producing the myelin that forms sheaths around axons. In both mammals and zebrafish, oligodendrocytes arise from the motoneuron progenitor (pMN) domain of the ventral spinal cord, where motoneurons form first and oligodendrocytes develop later (12). The transcription factor gene olig2 is critical for oligodendrocyte specification, and its expression is first observed in the pMN domain and then maintained throughout the subsequent differentiation of oligodendrocyte progenitor cells (OPCs) (13). Although the role of oligodendrocytes in tumors associated with NF1 loss remains unclear, optic pathway gliomas in neurofibromatosis type 1 patients expressed PEN5, a marker for oligodendrocyte precursors (14). Consistent with this finding, knockout mice and in vitro cell culture experiments demonstrated that deletion of the Nf1 gene in neuroglial cells leads to their overproliferation and abnormal differentiation (15,16). Aberrant oligodendroglial phenotypes may also contribute to other neurofibromatosis type 1 CNS abnormalities, such as macrocephaly and learning disabilities, presumably due to abnormal increases in glial cell production (17,18).

We recently identified two zebrafish orthologs of NF1 and analyzed their cardiac and vascular functions during embryologic development (19). Here we report the in vivo roles of nf1 in zebrafish oligodendrocyte development. Both genes are expressed ubiquitously early in development, but become restricted to the CNS after 48 hpf (hours postfertilization). Knockdown of the nf1 genes resulted in an increase in OPC proliferation in the spinal cord, and time-lapse imaging demonstrated an enhancement of OPC migration during development. The increased numbers of OPCs in the spinal cord caused by nf1 deficiency was rescued by forced expression of the GRD domain of human NF1, suggesting an essential role for Ras pathway activation and the GAP activity of NF1 in regulating OPC numbers during development.

RESULTS

The number of OPCs is increased upon nf1 knockdown

NF1 patients often have symptoms that may be related to glial abnormalities such as macrocephaly, and they are prone to develop glia-derived cancers including optic pathway glioma and astrocytoma (1,3,18). In order to test the role of nf1a and nf1b in zebrafish oligodendrocyte development, we used morpholinos (MOs) to knockdown each gene in the Tg(olig2:EGFP) transgenic zebrafish line (20) that had been bred into wild-type and p53 e7/e7 backgrounds (21). The Tg(olig2:EGFP) transgenic line expresses EGFP in OPCs throughout oligodendrocyte development, allowing their analysis in vivo (20). The Tg(olig2:EGFP) transgenic line in the p53 mutant background was used to circumvent off-target effects that are known to be due to MO toxicity (Supplementary Material, Fig. S2) (22). To knockdown nf1 function, we designed and injected nf1-specific splice-blocking MOs [two for nf1a (nf1a-e1 and nf1a-e7) and one for nf1b (nf1b-e4)] into the Tg(olig2:EGFP) transgenic zebrafish lines (19). The two MOs used to block nf1a function, nf1a-e1 and nf1a-e7, yielded similar results, both in efficacy and in the resultant phenotypes (Supplementary Material, Figs S3 and S4); thus, for consistency, we only present our results with nf1a-e7 and nf1b-e4 to block nf1a and nf1b expression, respectively. The efficiency of the MOs was confirmed by RT–PCR at 3 dpf (days postfertilization), when the observed aberrant bands upon MO knockdown demonstrated inappropriate splicing [MOs for nf1a (nf1a-e7) and nf1b (nf1b-e4) are shown in Supplementary Material, Fig. S3]. The efficacy of the nf1b MO was further evaluated by quantitative real-time RT–PCR, which indicated an ∼90% knockdown of the wild-type nf1b transcript (Supplementary Material, Fig. S3D). Because nf1a-e7 injection led to several aberrant bands in addition to multiple cryptic donor splice sites within exon 7 of nf1a, the quantitative knockdown of nf1a transcripts could not be similarly evaluated by quantitative real-time RT–PCR. However, when these aberrant bands were subcloned and sequenced, they were found to contain either deletions of the targeted exons or insertions of additional introns that resulted in frameshifts in coding sequence leading to the early termination of transcripts (see Supplementary Material, Fig. S3B for details).

In Tg(olig2:EGFP) transgenic zebrafish, a subset of GFP⁺ OPCs that arise in the ventral spinal cord migrate dorsally and differentiate into oligodendrocytes (brackets in Fig. 1A–C), whereas another GFP⁺ OPC subpopulation remains in the ventral spinal cord, intermingled with olig2-EGFP⁺ motoneurons and interneurons (Fig. 1A–C) (20,23). These olig2-EGFP⁺ motoneurons and interneurons can often be distinguished from the OPCs because of their lower level of EGFP expression, relatively round shape devoid of processes and their failure to migrate dorsally. When nf1a and nf1b were knocked down by MOs, we consistently found an increased number of dorsally positioned OPCs at 3 dpf (Fig. 1A–C; see also Supplementary Material, Fig. S4). The most dramatic OPC increase relative to the control was observed when both the nf1a and nf1b genes were knocked down simultaneously in the p53 e7/e7 background (Fig. 1C), although significant increases in OPC numbers were also observed in embryos with the knockdown of an individual nf1 gene (Fig. 1M; Supplementary Material, Fig. S4). These results indicate that both nf1 genes contribute to the regulation of normal OPC numbers and that nf1 knockdown results in an increased number of dorsally positioned OPCs.

Figure 1.

Knockdown of nf1a and nf1b leads to increased numbers of OPCs. (A–C) Projected confocal images showing a lateral view of the spinal cord in live transgenic animals at 3 dpf. (A, D–F) Uninjected Tg(olig2:EGFP): p53+/+ transgenic zebrafish. (B, G–I) Control MO-injected Tg(olig2:EGFP): p53 e7/e7 embryo. (C, J–L) Tg(olig2:EGFP): p53 e7/e7 injected with nf1a and nf1b MOs [nf1a + 1b Kd (knockdown)]. (D–L) Transverse sections through the spinal cord of embryos at 3 dpf, with GFP (green) and sox10 (red) labeling to specifically visualize co-expressing OPCs (arrows). (M) Statistical analysis of mean ± SEM numbers of GFP⁺/sox10⁺ cells per section in nf1a or nf1b alone or in combination at 3 dpf. Asterisks indicate statistical significance (*P < 0.05; **P < 0.005; ***P < 0.0005). Scale bars in (A–C) = 50 μm; (F, I and L) = 20 μm.

We then extended our analysis to include ventrally positioned OPCs by co-immunostaining with an anti-sox10 antibody that labels OPCs, but not olig2-EGFP⁺ motoneurons and interneurons that are also located in the ventral spinal cord at this stage. Although the increase in olig2-EGFP⁺;sox10⁺ OPC numbers upon nf1 knockdown was not apparent in transverse sections of the spinal cord at 2 dpf, when the dorsal migration of OPCs commences (Supplementary Material, Fig. S5), the increased numbers of OPCs did become evident by 3 dpf (Fig. 1J–L, compared with D–F and G–I). The mean number of OPCs was determined for nf1a + nf1b morphants and control embryos by counting these cells in immunostained transverse sections. The nf1 morphants were found to have 36% more OPCs than control p53 e7/e7 embryos (9.0 versus 6.6, with n = 23 and 26, respectively, P < 0.0001; Fig. 1M). These findings indicate that nf1 negatively regulates the number of OPCs during development.

The observed increase in OPCs in nf1 morphants could be due to an aberrant dorsal migration of OPCs that normally remain in the ventral spinal cord or to an overall increase in OPCs throughout the spinal cord. To distinguish between these possibilities, we assessed the differences in OPC numbers in dorsal and ventral spinal cord regions. A 36% increase in dorsally migrating OPCs was found in nf1 versus control morphants in the p53 background (4.0 versus 2.9; n = 11; P < 0.05), whereas a 26% increase in ventrally localized OPCs was observed (4.6 versus 3.7; n = 11; P < 0.05). These results indicate that increases in OPC numbers occur in both locations rather than reflecting an aberrant displacement of ventral cells to more dorsal regions. Although a slight increase in OPCs in the p53 mutant background was observed, presumably due to the survival of cells that would otherwise succumb to MO toxicity, it is important to note that significant increases in OPCs in nf1 morphants were also observed in the wild-type background (Fig. 1M; Supplementary Material, Fig. S4).

In order to further document the OPC increase upon nf1a + 1b knockdown, we also employed a second approach to block apoptosis and rescue the early cell death in the p53 wild-type embryos that occurs with nf1a + 1b MO injection. In this approach, we co-injected zebrafish bcl-X_L RNA, which is known to block most cell death during early embryogenesis (24,25). Most of the embryos (>90%) co-injected with bcl-X_L RNA together with nf1a + 1b MOs did not exhibit cell death in the brain at 1 dpf (Supplementary Material, Fig. S12D) and developed relatively normally to 3 dpf. The numbers of OPCs in these Tg(olig2:EGFP) embryos were significantly increased compared with those in embryos co-injected with bcl-X_L RNA and the control MO (63.3 versus 52.9, P < 0.05; Supplementary Material, Fig. S12E–G), or embryos injected with control MO alone (Fig. 4D and E).

Figure 4.

Forced expression of NF1 GRD is sufficient to rescue the proliferative OPC phenotype. (A–D) Projected confocal images showing the lateral view of the spinal cord of Tg(olig2:EGFP): p53 e7/e7 embryo at 3dpf: (A and B) nf1a + 1b Kd; (C and D) control MO; (A andC) GRD RNA injected; (B and D) mCherryRed (control) RNA injected. Insets show the bright field images of the embryos used for confocal imaging and demonstrate normal overall morphologies in all cases. Scale bar = 50 μm. (E) Graph showing the number of OPCs, represented by individual points, in the dorsal spinal cord (SC) of 10 embryos for each condition at 3 dpf (*P < 0.05; ***P < 0.001). The data are presented as means ± SEM.

To determine whether nf1 loss affects the differentiation of OPCs into oligodendrocytes, we counted myelin basic protein-positive oligodendrocytes in the dorsal spinal cord at 3 dpf in whole-mount RNA in situ hybridized embryos. No significant differences between control and nf1a + 1b knockdown embryos were found at 3 dpf (Supplementary Material, Fig. S6). This finding suggests that the extra OPCs fail to differentiate during the 3 day time period shown in Supplementary Material, Figure S6; however, it is possible that the effects of MOs are diluted over time and genetic analysis at time points beyond 3 days will require the isolation of stable nf1 mutants.

nf1 loss specifically affects OPCs from the pMN of the ventral spinal cord

Since OPCs and motoneurons are derived from common precursors in the olig2-positive pMN domain of the ventral spinal cord during development (26), we examined whether nf1 deficiency might affect the development of motoneurons and contribute to the increased numbers of OPCs. Islet protein (Isl) is expressed in primary and secondary motoneurons, interneurons and sensory neurons of the spinal cord, and Zn5 is a marker specific for developing secondary motoneurons (27,28). Quantification of cell numbers using anti-Isl and Zn5 antibodies did not exhibit significant differences in cells expressing either of these markers in nf1a + 1b knockdown embryos compared with control at 80 hpf (Fig. 2F versus B; N versus J; Q and R, n = 15). In addition, nf1a + 1b loss did not affect the overall size or shape of developing secondary motoneurons (Fig. 2N versus J). These results suggest that nf1 knockdown acted specifically on the cells of the oligodendrocyte lineage, whereas the neuronal lineages from the pMN domain, such as motoneurons, remain unaffected.

Figure 2.

Spinal cord neurons are not affected by nf1 loss. (A–D, I–L) Tg(olig2:EGFP): p53 e7/e7 embryos injected with the control MO (control). (E–H, M–P) An Tg(olig2:EGFP): p53 e7/e7 embryo injected with nf1a and nf1b MOs (nf1a + 1b Kd). (A–H) Projected confocal images of transverse sections labeled with the anti-isl antibody show subsets of primary/secondary motor neurons (bracket 1), interneurons (bracket 2) and putative Rohon Beard sensory neurons (asterisk). Green, olig2-GFP; red, anti-isl; blue, DAPI. Arrows in (A), (C), (E) and (G) designate dorsal OPCs. (I–P) Fluorescence images of transverse sections labeled with the zn5 antibody showing secondary motor neurons. Statistical analyses of the numbers of anti-isl-positive (Q) or zn5-positive (R) cells show no significant differences. Green, olig2-GFP; red, zn5; blue, DAPI. Arrows in (I), (K), (M) and (O) indicate dorsal OPCs; asterisks in (J), (K), (N) and O indicate the dorsal lateral fasciculus; brackets in (J), (K), (N) and (O) indicate secondary motoneurons. Scale bar = 20 μm.

Loss of nf1 causes an increase in OPC proliferation

The increase in OPC numbers caused by nf1 deficiency could occur through several different mechanisms. We first investigated whether cell death played a role in this phenotype by comparing the numbers of TUNEL-positive OPCs in the spinal cord of control versus nf1a + 1b morphants; the numbers of TUNEL-positive OPCs were identical and, in fact, were extremely rare in embryos injected with either specific or control MOs (total 2–4 cells in >40 sections), suggesting that the increase OPC numbers due to nf1 loss does not occur because of decreased levels of apoptosis.

Next we investigated whether the increase in OPC numbers in nf1 morphants was due to the increased proliferation of OPCs. We pulse-labeled control- and MO-injected embryos with BrdU at 54 and 80 hpf, fixed them immediately after BrdU treatment and examined BrdU incorporation in GFP⁺/sox10⁺ OPCs. At 80 hpf, BrdU⁺ cells were much more abundant in sections of the spinal cord (>40% of total sections), and BrdU⁺-positive OPCs could be identified in both control and nf1a + 1b morphants. Importantly, when GFP⁺/sox10⁺/BrdU⁺ cells were counted; the number of BrdU-labeled as well as total OPCs in nf1a + 1b morphants was significantly higher compared with those in controls (1.08 versus 1.83 per section, n = 12, P < 0.005; Fig. 3). These results indicate that nf1-deficient OPCs exhibit a higher proliferation rate, thus accounting for the observed increase in OPC numbers. In contrast, at 54 hpf, when OPC numbers were not yet affected by nf1 loss (Supplementary Material, Fig. S5), GFP⁺/sox10⁺/BrdU⁺ cells in the spinal cord of both control and nf1a + 1b morphants were very rare and not significantly different (1/60 sections of control and 4/43 sections of nf1a + 1b morphant, Fisher's exact test, P > 0.05; Supplementary Material, Fig. S7).

Figure 3.

nf1 loss leads to increased BrdU incorporation in OPCs. Projected confocal images of transverse spinal cord sections of Tg(olig2:EGFP): p53 e7/e7 embryos at 80 hpf. (A and E) green, olig2-GFP; (B and F) magenta, anti-sox10; (C and G) red, anti-BrdU; (D and H) overlay merged with DAPI (blue). (A–D) Embryos injected with the control MO (control). (E–H) Embryos injected with nf1a + nf1b MOs (nf1a + 1b Kd). (A–H) Arrows indicate dividing OPCs (GFP⁺/sox10⁺/BrdU⁺), and arrowheads denote non-dividing OPCs (GFP⁺/sox10⁺/BrdU⁻). (E–H) Asterisks denote a GFP⁻/sox10⁻/BrdU⁺ cell in the spinal cord, representing a dividing neuron. (I) Statistical analysis of mean ± SEM numbers of GFP⁺/sox10⁺/BrdU⁺ (‘BrdU’) or total OPCs (‘tOPCs’) per section in control or nf1a + 1b Kd at 3 dpf. OPCs incorporating BrdU are increased in the nf1a + 1b morphants compared with controls, indicating a higher proliferative activity. Asterisks indicate statistical significance (**P < 0.005; ***P < 0.0005). Scale bar = 20 μm.

Another potential contributor to increased OPC numbers may be through an increase in OPC progenitor cells. olig2-EGFP⁺ radial glial cells are slowly dividing cells that are thought to give rise to newly forming OPCs in the juvenile and adult spinal cord (29). Thus, hyperproliferation or transdifferentiation of radial glial cells into OPCs could contribute to the increase in OPCs observed in our nf1a + 1b morphants. However, the numbers of olig2-EGFP⁺ radial glia, which are co-labeled with GFAP, did not differ significantly in nf1-deficient versus control embryos (1.17 versus 1.11, n = 18, P > 0.7, Supplementary Material, Fig. S8A–F), and BrdU-incorporating radial glial cells in the spinal cord at 54 and 80 hpf were unchanged and extremely rare without or with nf1 loss (3 BrdU⁺ radial glial cells/109 sections in control and 1 BrdU⁺ radial glial cell/138 sections in nf1a + 1b morphants at 80 hpf, Fisher's exact test, P > 0.3; Supplementary Material, Fig. S8G–N). Taken together, these data suggest that following OPC specification, nf1 loss causes an increase in their proliferation between 54 and 80 hpf, which is responsible for the observed increase in OPC numbers in nf1a + 1b morphants.

The increase in OPCs due to nf1 knockdown depends upon the GRD of NF1

The best known function of human NF1 is to act as a Ras-GAP to downregulate Ras signaling. To test whether the overproliferation of OPCs in nf1 morphants was dependent upon GRD function, we co-injected mRNA encoding the GRD domain of human NF1 (5,6,30) into nf1a + 1b-morphant embryos [Tg(olig2:EGFP): p53 e7/e7 background] and examined whether its expression could rescue this phenotype. To monitor GRD expression, we fused the coding sequence of the mCherryRed fluorescent protein in-frame at the N-terminus of the sequence encoding the GRD domain. The nf1a + 1b-morphant embryos [Tg(olig2:EGFP): p53 e7/e7 background] co-injected with mCherryRed RNA served as a positive control and exhibited the expected increase in OPC numbers relative to control MO- and control RNA-injected Tg(olig2:EGFP): p53 e7/e7 embryos (Fig. 4B–D). However, nf1a + 1b-morphant embryos that were co-injected with the GRD RNA exhibited a relative reduction in OPC numbers in the dorsal spinal cord from 67.7/embryo to 48.5/embryo (n = 10, P < 0.001; Fig. 4A, B and E), which did not differ significantly from the number of OPCs present in control MO- and control RNA-injected embryos (48.5 versus 47.5, P > 0.7, and 48.5 versus 57.5, P > 0.1, respectively; Fig. 4A and C–E). These results demonstrate that the expression of NF1-GRD alone can rescue the increased OPC phenotype due to nf1 deficiency, indicating that NF1 regulates OPC proliferation through its GAP activity.

nf1 loss promotes OPC migration

To examine the behavior of OPCs following nf1 knockdown, we performed in vivo live time-lapse imaging of these cells. We monitored the Tg(olig2:EGFP) transgenic animals for 12 h, beginning at 60 hpf when OPCs start to actively migrate away from the ventral spinal cord. Compared with the findings in uninjected control animals, more OPCs in the nf1a + 1b-knockdown embryos migrated into the dorsal spinal cord (Fig. 5A and B; Supplementary Material, Movies S1 and S2), consistent with our observations from fixed sections (Fig. 1). Interestingly, the time-lapse study also revealed that the nf1a + 1b-morphant OPCs traveled longer distances relative to control animals (Fig. 5E; Supplementary Material, Movies S1 and S2). In the movies, nf1a + 1b-knockdown OPCs migrated farther in the dorsal and rostro-caudal directions than did control OPCs, as indicated by the examination of representative migratory traces of individual cells (Fig. 5C and D). Analysis of the movement of all individual OPCs that could be observed over the 12 h imaging period showed a 35% increase in total distance traveled by the OPCs in nf1 knockdown animals (32.05 µm in control versus 43.27 µm in nf1a + 1b-morphant, P < 0.05; n = 9 and n = 23, respectively; Fig. 5E). In all cases, the OPCs display intermittent movements consisting of repeated cycles of active migration, separated by pauses before continuing, often in different directions from the original path. After nf1 knockdown, the OPCs paused for shorter periods of time relative to controls (413.9 min in control versus 324.3 min in nf1a + 1b-morphant; P < 0.005; Fig. 5F), whereas the frequency of the pauses did not differ significantly (2.28 h⁻¹ in control versus 2.58 h⁻¹ in nf1a + 1b-morphant; P > 0.1). Furthermore, the migration velocity of OPCs was determined by dividing the distance traveled by the total traveling time, which excluded the periods when they were stationary. The migration velocity was unaffected by nf1 knockdown (0.5203 µm/5 min in control versus 0.5343 µm/5 min in nf1a + 1b-morphant; P > 0.7; Fig. 5G). Together, these findings show that OPCs in nf1a + 1b-morphants exhibit a novel phenotype in vivo, spending more time actively migrating with shorter pauses than controls.

Figure 5.

nf1 loss affects the migration of OPCs. (A and C) An uninjected Tg(olig2:EGFP): p53 e7/e7 control embryo; (B and D) A Tg(olig2:EGFP): p53 e7/e7 embryo injected with nf1a + 1b MOs. (A and B) Montages of 12 h time-lapse images from 60 to 72 hpf, showing the movements of OPCs every hour. Numbers in each panel denote the hour(s) after the start of imaging. For the complete movies, refer to Supplementary Material, Movies S1 and S2. (C and D) Cell migration paths are shown for five OPCs that traveled the farthest during the observation period. In both conditions, migration patterns are highly dynamic. Arrowheads indicate the endpoint of the cell's migration. The dashed lines represent the dorsal-most GFP⁺ domains of the ventral spinal cord. Top, dorsal (d); left, rostral (r). (E) Graph showing the total distance OPCs travel with or without nf1 loss. (F) Graph showing the total time individual OPCs spent pausing. (G) Graph showing the velocity of OPCs calculated from when they were actively migrating. All individual OPCs (represented by single points) that could be observed in the field of view throughout the entire period, spanning the 12 h imaging period, were traced and used in this analysis (n = 9 for uninjected control embryos, n = 23 for nf1a + 1b-morphant embryos). The data in (E–G) are reported as mean ± SEM; asterisks indicate non-parametric statistical significance (*P < 0.05; ns, not significant). Scale bars in (A) and (B) = 25 μm; (C) and (D) X-axis = 10 μm and Y-axis = 5 μm.

nf1 loss leads to the hyperactive ERK signaling in the spinal cord

To further investigate the effect of nf1 deficiency on Ras signaling during OPC development, we examined two principal pathways downstream of activated Ras by evaluating the status of phosphorylated ERK and phosphorylated S6 to assess the Raf/ERK and PI3K/Akt/mTOR pathways, respectively (31). Labeling transverse sections of the spinal cord at 54 hpf of Tg(olig2:EGFP): p53 e7/e7 embryos with a phospho-ERK antibody revealed few phospho-ERK⁺ cells in either control or nf1-knockdown conditions, suggesting low ERK signaling in the spinal cord at this stage that did not respond to nf1 loss (Supplementary Material, Fig. S9). However, at 80 hpf, nf1 knockdown resulted in a marked increase in phospho-ERK-positive cells throughout the spinal cord of Tg(olig2:EGFP): p53 e7/e7 embryos relative to controls, indicating an aberrant activation of the Ras pathway due to nf1 deficiency (Fig. 6). Interestingly, phospho-ERK was not increased in OPCs (arrows in Fig. 6A–H), but rather was expressed in neighboring neurons that were identified by the co-expression of HuC/D, a pan-neuronal marker (arrows in Supplementary Material, Fig. S10). There was a subpopulation of OPCs that were weakly pERK positive in both control and nf1a + 1b morphants (asterisks in Fig. 6A–H and Supplementary Material, Fig. S9F–J); however, no significant change in pERK levels was detected upon nf1 loss. We also assayed for the phosphorylation of the S6 ribosomal protein in the spinal cord at 3 dpf following nf1 knockdown in Tg(olig2:EGFP): p53 e7/e7 embryos and did not observe any significant differences in the numbers of phospho-S6-positive OPCs or other cell types (Supplementary Material, Fig. S11). Thus, nf1 deficiency causes an aberrant upregulation of the ERK signaling pathway in neighboring spinal cord neurons, coincident with the overproliferation and abnormal migration of OPCs.

Figure 6.

nf1 loss activates the ERK pathway. Projected confocal images of transverse spinal cord sections at 80 hpf of Tg(olig2:EGFP): p53 e7/e7 embryos. (A and E) green, olig2-EGFP; (B and F) magenta, anti-sox10; (C and G) red, anti-phospho-ERK; (D and H) overlay merged with DAPI (blue). (A–D) Control MO-injected embryo. (E–H) nf1a + 1b Kd. Increased numbers of phospho-ERK-positive cells were detected in the spinal cord of nf1a + 1b morphants (arrows in G), compared with control (arrow in C). Arrowheads denote GFP⁺/sox10⁺ OPCs, which are predominantly phospho-ERK1/2 negative. A minor OPC population is phospho-ERK positive (asterisks). Scale bars = 20 μm.

DISCUSSION

Inactivation of the NF1 tumor suppressor gene has been found to contribute to a wide variety of pathologies that can affect glial development and predispose to tumorigenesis in both the PNS and CNS. In this study, we gain insight into the role of NF1 in gliogenesis by investigating how the loss of nf1 affects oligodendrocyte development in zebrafish. We determined the expression patterns of the two zebrafish nf1 genes during embryogenesis and showed that the loss of the nf1 orthologs by MO knockdown resulted in increased numbers of OPCs. The increased number of OPCs was found to be due to overproliferation of the OPCs themselves, which was dependent upon the activation of the Ras pathway through the expression of the NF1-GRD. nf1 deficiency also increased the motility of OPCs, leading to an enhancement of migration. The hyperproliferation and enhanced migration of OPCs after nf1 knockdown occurred along with the aberrant activation of pERK signaling in neighboring neurons in the developing spinal cord.

Duplication of the whole genome in teleosts, which followed their divergence from tetrapods, accounts for the presence of two zebrafish nf1 genes, in contrast to the single gene found in mammals (32). Both nf1 genes are not only functional but also act redundantly during OPC development in zebrafish because: (i) they show exceptionally high amino acid sequence homology, especially in the GRD; (ii) both genes conserve syntenic relationships with human NF1; (iii) both are expressed during embryogenesis with partially overlapping expression patterns; and (iv) both genes contribute independently to an increase in OPC numbers as a result of MO knockdown. Our analysis indicates that zebrafish nf1a and nf1b act in an additive manner during OPC development, in that simultaneous knockdown of these genes increased OPC numbers more than knockdown of either gene alone.

Studies in vivo and in vitro indicate that astrocytes and oligodendrocytes, the two main glial cell types in the CNS, overproliferate in the context of either complete Nf1 knockout or heterozygous Nf1 mutant cells (11,15,16,33). Schwann cells, the myelin-producing glial population in the PNS, also overproliferate upon Nf1 loss (34,35). We augment these findings by showing in vivo that the number of zebrafish OPCs in the spinal cord is increased upon nf1 MO knockdown. We found that the numbers of zebrafish olig2⁺ OPCs co-labeled by sox10 in the developing spinal cord are increased upon nf1 MO knockdown. Furthermore, our TUNEL and BrdU pulse-labeling assays clearly show that excessive OPCs in nf1-deficient embryos results from the aberrant overproliferation of OPCs themselves (Fig. 4), rather than decreased OPC death or increased production from olig2-EGFP⁺ radial glial cells (29). Consistent with our finding, cultured cells from the E12.5 spinal cord of the nf1 knockout mouse showed an increased number of BrdU-incorporating cells (15), and the numbers of both olig2-expressing cells and BrdU-incorporating cells were increased in the brains of BLBP-cre/Nf1 flox mice, when Nf1 was selectively deleted in neuroglial progenitors (16). Moreover, our GRD rescue experiment establishes that the OPC phenotype is due to the loss of GAP activity normally provided by Nf1. Our results are important because Nf1 has also been observed to have tissue-specific roles in development that are independent of the GRD. For example, GRD rescue in the context of Nf1 deficiency can restore cardiovascular development, but does not inhibit the overgrowth of neural crest-derived tissues (30).

During development, neurons and glia of the vertebrate CNS originate from common neuroglial precursor cells in the pMN domain of the ventral spinal cord (26). Since MO injections were performed at the one-cell stage, it is possible that the increased OPC phenotype might reflect an earlier effect of nf1 deficiency on neuroglial precursor cell proliferation, which might be evident from abnormal cell numbers in both neuronal and glial lineages. However, analysis of the neuronal derivatives (including motoneurons) in the nf1 morphants revealed numbers that were essentially unchanged, indicating that the effect of nf1 deficiency was specific to OPCs. These findings are interesting in light of the marked increase in levels of phosphorylated ERK observed after nf1 knockdown in the neurons surrounding the OPCs (Fig. 6; Supplementary Material, Fig. S7). Although we do not believe that earlier changes in phospho-ERK signaling affect the OPCs themselves (Supplementary Material, Fig. S8), it is worthwhile to note that there remain additional signaling pathways regulated by Ras that may contribute to the intrinsic control of OPC numbers in nf1 morphants (31).

In vitro assays indicate that Nf1 loss in astrocytes and Schwann cells can lead to enhanced motility. Astrocytes missing one or both alleles of Nf1 exhibit an increase in cell motility, together with cytoskeletal abnormalities (36), that is dependent upon the mTOR pathway and requires nucleophosmin (37). Cultured Schwann cells deficient in Nf1 have also demonstrated increased invasiveness (34) and motility via the enhanced activity of TC21/R-Ras2 (38). However, the motility of these cells has not been analyzed in vivo due to technical limitations, and the effect of NF1 on oligodendrocyte motility has not been addressed. Here, we report the first in vivo analysis of OPC migration following nf1 knockdown using time-lapse microscopy. Our studies show conclusively that the nf1 deficiency leads to the enhanced migration of OPCs, which move intermittently as they migrate and cover greater distances owing to the decreased duration of pauses rather than to increased velocity of movement during active migration (Fig. 4). The increased migratory behavior of OPCs in nf1 morphants might result from the cell-autonomous effects of nf1 deficiency, such as cytoskeletal defects possibly mediated by the overexpression of GAP43 and T-cadherin (36), similar to the role of nf1 in motility in other glial cell types. The increased numbers of OPCs might also indirectly affect their migratory properties in nf1 morphants (39). At this time, we do not know whether the OPCs are responding to a non-cell-autonomous effect of nf1 loss, possibly mediated by increased phosopho-ERK activity in adjacent neurons, or whether the effects are mediated by cell-autonomous effects of nf1 unrelated to phosopho-ERK activation. The increased OPC numbers and motility due to NF1 loss may contribute to the phenotypes observed in human patients, such as neurofibromatosis type 1-associated macrocephaly (18) and the increased invasiveness of NF1-deficient glioma cells (40).

In summary, we have shown that OPC numbers and migration are enhanced upon nf1 knockdown and that forced expression of the GRD restores normal OPC numbers in nf1-deficient embryos. The experimental accessibility of the zebrafish embryo provides an excellent platform for further analysis of the mechanisms that underlie these changes. Furthermore, the eventual establishment of stable nf1-deficient zebrafish lines and the ability to conduct modifier genetic and drug screens in the zebrafish embryo should contribute to the dissection of proteins and pathways that can be targeted to improve the treatment of human pathologies resulting from NF1 loss.

MATERIALS AND METHODS

Zebrafish maintenance

All zebrafish were grown and maintained in accord with the DFCI IACUC-approved protocol. Tg(olig2:EGFP) transgenic animals were obtained from Dr Bruce Appel (20) and were crossed to our p53 e7/e7 mutant fish (21) to generate the Tg(olig2:EGFP): p53 e7/e7 line.

Cloning and plasmids

The predicted full-length cDNA for nf1a and nf1b was obtained from the Ensembl database (http://www.ensembl.org). EST clones CN024089 (GI: 45796859) for nf1a and CR930343 (GI: 56559576) for nf1b were identified and cloned into pBluescript to generate probes for whole-mount in situ hybridization assays.

MO injections

Approximately 1 nl of MO(s) in water with phenol red (0.5%), corresponding to 150 μm nf1a MO and 200 μm nf1b MO, was injected individually or together into embryos at the one-cell stage. The standard control MO was also used (Gene Tools). MO sequences: (1) nf1a-e1: 5′-GTC CAA GTA GTG TTT TCC TTA CCT G-3′; (2) nf1a-e7: 5′-TAG TAA ACA AGT GTC ACT CAC CGG C-3′; (3) nf1b-e4: 5′-CTC AGT ATT TAT CTG CAC CTG GTG G-3′; (4) MO control; 5′-CCT CTT ACC TCA GTT ACA ATT TAT A-3′.

Whole-mount in situ hybridization and immunofluorescence assays

Paraformaldehyde-fixed embryos were cryosectioned and immunostained using primary antibodies (rabbit anti-sox10, anti-GFP, anti-isl, zn5, anti-BrdU and anti-phospho-ERK), followed by incubation with Alexa 488- or 568-conjugated secondary antibodies. WISH assays were performed as by Thisse and Thisse (41).

Reverse transcription and PCR

Cells from the spinal cord of Tg(olig2:EGFP) at 3 dpf were FACS-sorted based on GFP expression. RNA from these cells were reverse-transcribed (RT) to prepare cDNA, which was used to perform PCR to test the expression of nf1 and nf1b. Similarly, RT–PCR of MO-injected embryos detected aberrant splicing events, confirming the whole or partial deletion of the targeted exons, or insertions leading to aberrantly truncated proteins.

Imaging and statistical analyses

A Zeiss LSM 510 META confocal microscope and a Zeiss compound microscope Axio Imager.Z1 were used to capture confocal images and fluorescence/brightfield DIC images, respectively. For time-lapse movies, a Zeiss 200M inverted microscope equipped with a spinning-disk confocal system (Yokogawa) was used. The statistical significance of changes to OPC numbers and migration upon nf1 knockdown was determined by t-test analysis (see Supplementary Material for further methodological details).

SUPPLEMENTARY MATERIAL

Supplementary Material is available at HMG online.

FUNDING

This work was supported by the Department of Defense (NF050175 to J.A.E. and A.T.L.). J.S.L. was supported by a Young Investigator Award (Children's Tumor Foundation) and A.P. by a fellowship from the Sarnoff Cardiovascular Research Foundation.