The E3 Ligase Mind Bomb-1 (Mib1) Modulates Delta-Notch Signaling to Control Neurogenesis and Gliogenesis in the Developing Spinal Cord

Background: Mib1 is a ubiquitin ligase that modifies Delta, a ligand for the Notch signaling pathway.

Results: Absence of Mib1 results in a reduced number of neural progenitors, spinal interneurons, and astrocytes.

Conclusion: Mib1 controls neurogenesis and gliogenesis in the spinal cord.

Significance: Novel insights about the role of Mind bomb1 in the regulation of early spinal cord development via Delta signaling are presented.

Abstract

The Notch signaling pathway is essential for neuronal and glial specification during CNS development. Mind bomb-1 (Mib1) is an E3 ubiquitin ligase that ubiquitinates and promotes the endocytosis of Notch ligands. Although Mib1 is essential for transmitting the Notch signal, it is still unclear whether it is a primary regulator of Notch ligand activity in the developing spinal cord. In Mib1 conditional knock-out mice, we observed depletion of spinal progenitors, premature differentiation of neurons, and unbalanced specification of V2 interneurons, all of which mimic the conventional Notch phenotype. In agreement with this, the reduction of progenitors in the absence of Mib1 led to a loss of both astrocytes and oligodendrocytes. Late removal of Mib1 using a drug-inducible system suppressed glial differentiation, suggesting that Mib1 continues to play a role in the formation of late progenitors mainly designated for gliogenesis. Finally, misexpression of Mib1 or Mib1 deletion mutants revealed that the ring domain of Mib1 is required for the specification of V2 interneurons in the chick neural tube. Together, these findings suggest that Mib1 is a major component of the signal-sending cells required to provide Notch ligand activity for specifying neurons and glia in the spinal cord.

Introduction

Notch signaling is one of the fundamental pathways that specify the fates of diverse cell types during embryonic development based on cell-to-cell interaction, mediated by Delta ligands and Notch receptors. Notch signaling converts initially homogeneous progenitors into cells with different neurogenic potential by lateral inhibition; cells expressing Delta adopt neuronal fates and simultaneously prevent Notch-expressing adjacent cells from becoming neurons. The role of Notch in controlling the neurogenic potential was initially suggested by classic genetic studies using Drosophila; disruption of the Notch pathway resulted in the production of extra neurons and a reduction in the number of progenitors (1). In zebrafish and mice, down-regulation of Notch or Delta led to premature and excessive neuronal differentiation, whereas ectopic expression of Delta inhibited neurogenesis (2–4). Thus, the role of the Notch signal in creating a balance between maintenance of neural progenitors and neuronal differentiation is evolutionarily conserved.

In addition to the general role of Notch in regulating neuronal fates, the Notch signal has been proposed to serve a more specific and refined role in the vertebrate nervous system. For instance, multiple neuronal subtypes are derived from distinct progenitor populations along the dorsal-ventral axis of the spinal cord (5). Previous studies have shown that the fate of interneuronal subtypes is compromised when Notch activity is altered (6–8). The binary choice between V2a and V2b interneurons, both derived from p2 ventral progenitors, is sensitive to the Notch signal. When Notch activity is absent, as demonstrated in Notch1 conditional mutant mice or Presenilin null mice, more V2a interneurons and less V2b interneurons are generated. Together, cell lineage and genetic studies suggest that Notch-expressing cells turn into inhibitory V2b interneurons marked by Gata3, whereas Delta-expressing cells become excitatory V2a interneurons labeled by Chx10 (7, 8).

In the developing CNS, gliogenesis begins after neurogenesis and produces astrocytes and oligodendrocytes in the late period of embryonic development (9). Because Notch maintains progenitor populations, it could potentially contribute to glia formation by reserving a progenitor pool for gliogenesis. Consistent with this idea, misexpression of NICD,² an active form of Notch, induces radial glial cells, a population recently identified as embryonic neural stem cells that later transform into astrocytes (10, 11). Likewise, the Delta-Notch interaction between intermediate progenitor cells and radial glia maintains the stem cell potential of radial glial cells (12). Furthermore, the formation of other specialized glial cells, including Müller glia in the retina and Bergman glia in the cerebellum, is also promoted by active Notch (13, 14). In contrast, it is still unclear how Notch signaling affects oligodendrocyte development. Notch is known to inhibit oligodendrocyte differentiation, whereas oligodendrocyte precursor cells appear to be induced by the Notch signal (2, 15–17).

Notch signaling is initiated by interaction between the Notch receptor and DSL family ligands (Delta, Serrate, Lag-2). This triggers a series of protease-driven cleavages of the Notch receptor. As a result, intracellular fragments of Notch (NICD) are produced and translocated to the nucleus where they directly activate gene transcription together with other transcriptional activators (18). Studies over the past several decades have identified numerous core elements of the Notch pathway, mostly implicated in signal transduction following receptor activation. Nevertheless, our understanding of the regulation of Delta within signal-sending cells, the key first event triggering Notch signaling, is still limited. In the signal-sending cells, DSL family ligands are present on the cell membrane and become endocytosed to be incorporated into vesicles. Endocytosis of the Delta ligands appears to be required for Notch signal transduction because blockade of endocytosis in Drosophila perturbs Notch signal transduction (19). Genetic screens in Drosophila and zebrafish have identified two genes, neuralized (Neur) and mind bomb-1 (Mib1), both of which are E3 ubiquitin ligases that ubiquitinate and promote endocytosis of DSL family ligands (20, 21). Elimination of Mib1 in zebrafish and mice resulted in premature differentiation of neurons, a phenotype common to mutants in which Notch signaling is defective (12, 21). Although it is not yet known whether Mib1 is involved in the maturation of Delta ligand by recycling or trafficking, or provides a pulling-force for cleavage, it is clear that it is essential for Notch signaling (22, 23).

In this study, we investigated the role of Mib1 in neurogenesis and gliogenesis during spinal cord development. The absence of Mib1 resulted in premature differentiation of neurons and mis-specification of spinal V2 interneurons. In addition, radial glia and later differentiation of astrocytes and oligodendrocytes were inhibited in Mib1 mutant mice. Down-regulation of Mib1 late in development using a drug-inducible system resulted in various extents of reduction of mature glia. The developmental defects found in Mib1 mutant mice were identical to those of other Notch-related mutants, supporting the view that Mib1 is a critical regulator of Notch signaling in the developing spinal cord.

EXPERIMENTAL PROCEDURES

Animals

Mib1 floxed (f) mice and Nes-Cre ERT2 transgenic mice have been described previously (24, 25). Nestin-Cre transgenic mice were obtained from The Jackson Laboratory. Mib1 conditional knock-out (cKO) mice were generated by mating Mib1^f/f homozygous mice with Mib1^f/+;Nestin-Cre mice. Mib1 drug-inducible knock-out (iKO) mice were produced by mating Mib1^f/f homozygous mice with Mib1^f/+;Nes-Cre ERT2 mice. To delete the flox allele, tamoxifen (5 mg/ml) was administered intraperitoneally daily for 3 days as indicated.

Expression Constructs

Mouse Mib1 was generated by PCR from mouse cDNA and cloned into the pCAGGS1 vector. The Mib1dn and Mib1ΔRD3 constructs lack all the ring finger domains (amino acids 781–1006) and the third domain (amino acids 959–1006), respectively (21). The NICD construct was purchased from Addgene. The plasmid Dll1 was kindly provided by Dr. S. Sockathanan. Mouse Dll1, Dll4, and Delta mutants were cloned into pCAGGS1 with the FLAG tag at the C terminus. Dll4^ΔICD is a deletion mutant that contains amino acids 1–555 of Dll4. Dll4^ΔICD-Ub-HA was designed to have Ub-HA conjugated with Dll4^ΔICD in-frame (21).

Chick in Ovo Electroporation

Chick eggs (Hwasun Hyangsang Farm) were incubated in a humidified chamber, and DNA constructs were injected into the lumens of Hamburger Hamilton (HH) stage 10–12 spinal cords. Electroporation was performed using a square wave electroporator (BTX). Co-electroporation resulted in >80% of cells expressing all constructs. Incubated chicks were harvested and analyzed at HH stage 20–25.

In Utero Electroporation

Timed pregnant mice were ethically anesthetized with inhalation anesthetics, isoflurane combined with oxygen/nitrous oxide. CAG promoter-based plasmid DNAs (up to 1 μg/μl) were injected into the lateral ventricles of embryos. Pulse conditions were 5 pulses separated by 950 ms at 40 V for E13.5 embryos. Embryonic brains were harvested at E16.5 and fixed with 4% paraformaldehyde in 0.1 m phosphate buffer. After PBS washing and cryoprotection, brains were sectioned coronally at 12 μm for immunohistochemistry.

Immunohistochemistry and in Situ Hybridization

Embryos were fixed in 4% paraformaldehyde for immunohistochemistry. Antibodies used in this study were as follows: guinea pig anti-Olig2 (26); rabbit anti-Hb9 (27); guinea pig anti-Chx10 (28); monoclonal anti-Gata3 (Santa Cruz Biotechnology); rabbit anti-En1 (Dr. Jessell); rabbit anti-Sox1 (Cell Signaling Technology); rabbit anti-NFIA (Active Motif); rat anti-Nestin (Chemicon); rabbit anti-p-Vimentin (MBL); rabbit anti-GFAP (Dako); and rabbit anti-GFP (Invitrogen) antibodies. Fluorophore-conjugated species-specific secondary antibodies were used as recommended (The Jackson Laboratory and Invitrogen). For endogenous Dll1 staining, autoclaved antigen retrieval and signal amplification using TSA kit (Molecular Probes) were performed. The slides were processed at 121 °C in 0.01 m tri-sodium citrate buffer, pH 6.0, for 10 min and immunostained with anti-Dll1 antibody (H-265, Santa Cruz Biotechnology) at 1:100 (29).

For in situ hybridization, transverse sections were hybridized with digoxigenin-labeled probes specific for each gene amplified from mouse or chick embryonic cDNA using an Advantage cDNA PCR kit (Clontech).

Ubiquitination Assay

HEK293T cells were transiently transfected with pCAGGS1-Dll1-flag, pcDNA3Ub-HA, pEGFP-Mib1, and pCAGSS1-Mib1dn as indicated (21). pcDNA3-Ub-HA was kindly provided by Dr. Y. J. Yoo. After 30 h, cells were lysed in IP lysis buffer (50 mm Tris-HCl, pH 7.4, 150 mm NaCl, 1 mm EDTA, 1% Triton X-100) with 10 mm iodoacetamide (Sigma), 20 μm MG132 (Biomol), and protease inhibitor mixture (Calbiochem). Lysates were immunoprecipitated with anti-FLAG M2 affinity gel (Sigma). Immunoprecipitates were washed and analyzed by Western blot analysis using anti-FLAG M2 peroxidase (Sigma), rat anti-HA peroxidase (Roche Applied Science), or anti-α-tubulin (AbD Serotec) antibodies.

Subcellular Localization Analysis

Dll4, Dll4-flag, Mib1wt-GFP (30), or Mib1dn-Myc was transfected in COS7 cells. Primary antibodies used for immunocytochemistry were anti-FLAG (Sigma), anti-Myc, and anti-GFP (Invitrogen) antibodies.

Primary Cortical Culture

Primary astrocytes were cultured from the 1- to 3-day-old Sprague-Dawley rat cortices. Cortices were dissected, dissociated using papain (Worthington Biochemical Corp.), and grown on poly-d-lysine (Sigma)-coated culture dishes for 2 weeks in neurobasal medium (Invitrogen), 10% fetal bovine serum (Hyclone). To deplete microglia and meningeal cells, cortical cultures were incubated with serum-free neurobasal medium for 2 days before use. Cells were trypsinized and re-plated into poly-d-lysine-coated coverslips. Neocortices from E14.5 Mib1^f/f mouse embryos were dissected and dissociated and were grown on astrocyte-plated coverslips. Cultures were maintained in neurobasal medium with B27 (Invitrogen). GFP or Cre recombinase-expressing retrovirus was added to the culture after plating. Coverslips were fixed and immunostained at 3 days in vitro.

Quantification and Statistical Analysis

Cells were counted in immunostained transverse sections containing the brachial spinal cords of various embryonic stages as indicated. At least three embryos were harvested from each of three independent experiments and used for quantification. Cell counts are shown as mean ± S.E. Unpaired Student's t tests were used for statistical analysis.

RESULTS

Delta-Notch Signaling Components Are Expressed in the Progenitor Domains of the Spinal Cord

To define the location of Notch signaling in the developing spinal cord, we examined the expression of Notch-Delta signaling components in E11.5 mouse spinal cords by in situ hybridization. Three Notch receptors (Notch1, Notch2, and Notch3) and five DSL ligands (Jagged1, Jagged2, Jagged3, Dll1, and Dll4) are present in the neural tube (31). Expression of Mib1, a major component required for Delta trafficking and thereby transmitting the Notch signal, and a Notch effector gene, Hes5, were also examined. At E11.5, Mib1 transcript was found throughout the progenitor domains along the entire dorsoventral axis (Fig. 1A). Progenitors tended to express higher levels of Mib1 as they proceed laterally within the progenitor domains where its substrates Delta and Jagged co-exist. Notch1 mRNA was relatively uniformly expressed in the ventricular zone along the dorsovental axis of the spinal cord (Fig. 1C). Notch ligands Jag1, Dll1, and Dll4 were expressed in distinct ventral progenitor domains in a mutually exclusive fashion as follow: Jagged1 in the p1 domain and Dll1 and Dll4 in the p0 and p2 domains (Fig. 1, E, G, and I). Hes1 was restricted to the dorsal ventricular zone, and Hes5 was highly expressed in all ventral progenitor domains, implying that Delta-Notch signaling is active in the majority of ventral progenitors (Fig. 1, K and M).

FIGURE 1.

Patterning of the neural tube in E11.5 Mib1 mutant mice. A–R, expression of Mib1, Notch1, Jag1, Dll1, Dll4, Hes1, Hes5, Ngn1, and Mash1 was examined in E11.5 Mib1 mutant mice by in situ hybridization. Progenitor domains were marked by dashed lines. S–Y, protein expression of Ngn2, Olig2, and Chx10 was examined in adjacent sections. Scale bar in Y, 200 μm for A–Y.

Defective Notch Signaling in Mib1 Mutant Mice

Mib1 mutant mice were severely growth-retarded around E9.5 and die around E11.5, which makes it impossible to examine the role of Mib1 in neural tube patterning (30). To overcome this problem, we crossed conditional Mib1 mutant mice with Nestin-Cre transgenic mice to selectively eliminate Mib1 within neural progenitors (12). Although the Mib1 conditional knock-out mice developed hemorrhages in the forebrain that affected brain development before E15.5, the spinal cord remained intact, which allowed us to examine the overall patterning of the spinal cord. We confirmed that Mib1 transcript was significantly down-regulated in Mib1 mutant mice as expected (Fig. 1B). We next examined the expression of Notch signaling components, including Notch1, Jag1, Dll1, Dll4, Hes1, and Hes5 in E11.5 Mib1 mutant spinal cords by in situ hybridization. Expression of Notch1, Dll1, and Hes5 was significantly reduced in the ventral spinal cord, indicating that Notch activity was low. Furthermore, the ventral midline was slightly distorted due to extinction of the progenitor domain (Fig. 1, D, P, and R). Jag1, Dll4, and Hes1 expression was unaltered in the mutant (Fig. 1, F, J, and L). We also compared the expression of the proneural genes Ngn1, Mash1, and Ngn2 together with the pMN domain marker Olig2 and the V2a interneuron marker Chx10 (Fig. 1, O–Y). Expression of Ngn1, Mash1, Ngn2, and Olig2 was greatly reduced in the ventral progenitor domain, although expression of Chx10 was prematurely increased in the progenitor domain. Together, the low Notch activity and depletion of ventral progenitors in the Mib1 mutant indicate that Mib1 is essential for the Notch signaling that maintains progenitor cell populations.

Mib1 Is Essential for Internalization and Ubiquitination of Delta

Mib1 is an E3 ligase that mono-ubiquitinates the Delta ligand (21). Because the RING domain is critical for its enzymatic activity, we generated deletion mutant constructs that either lack all three RING domains (Mib1dn) or the third one (Mib1ΔRD3) (21). We co-transfected Dll1 and ubiquitin into 293T cells, immunoprecipitated Dll1, and measured its ubiquitin level. Without Mib1, the level of ubiquitin in the immunoprecipitate was almost undetectable (Fig. 2A, 2nd lane). When Mib1 was introduced with Dll1 and ubiquitin, a significant level of ubiquitin in Dll1 was detected (Fig. 2A, 3rd lane). However, co-transfection of Mib1dn or Mib1ΔRD3 greatly inhibited ubiquitination of Dll1 (Fig. 2A, 4th and 5th lanes). The weak ubiquitination level in the presence of Mib1ΔRD3 is likely due to the residual activity coming from the other two intact RING domains.

FIGURE 2.

Mib1 is critical for ubiquitination and internalization of Delta1. A, ubiquitination of Dll1 by Mib1 in vitro. Delta-FLAG, ubiquitin-HA (Ub-HA), Mib1, Mib1ΔRD3, or Mib1dn was transfected in 293T cells as indicated. Cell lysates were immunoprecipitated (IP) with anti-FLAG antibody and immunoblotted (IB) with anti-HA or anti-FLAG antibody as shown. Asterisk marks the predicted size of Delta protein. Anti-tubulin antibody was used as a loading control. B and C, distribution of Dll4 and Mib1 or Mib1dn in COS7 cells. Dll4 and Mib1 co-localize in intracellular vesicles (B, B′, and B″). Dll4 is present in the plasma membrane when Mib1dn is introduced (C, C′, and C″). D and E′, localization of Notch1 was assessed in E12.5 Mib1 mutant mice by immunohistochemistry. Notch1 expression is broad and occasionally internalized into puncta in the control (D′, arrowheads), although it is reduced and mostly localized in the membrane in Mib1 mutant mice (E′, arrowhead). Scale bar in C″ is 15 μm for B and C″; in E, 50 μm for D and E; and in E′, 50 μm for D′ and E′.

We next tested whether ubiquitination of Delta by Mib1 induces internalization of Delta, a process necessary for Delta signaling transduction. To analyze subcellular localization of Delta, we co-transfected Dll4 and Mib1 into COS7 cells. When Dll4 and Mib1 were introduced, both were internalized and co-existed in intracellular vesicles (Fig. 2, B, B′, and B″). However, co-transfection of Dll4 with Mib1dn resulted in diffuse distribution of both proteins in the plasma membrane and cytoplasm. Together, ubiquitination of Dll1 by Mib1 internalizes Dll1 into the cells (Fig. 2, C, C′, and C″).

If Dll1 is mainly ubiquitinated by Mib1 and becomes internalized, we expect to see mis-localization of Dll1 in Mib1 mutant mice. To test our hypothesis, we examined expression of endogenous Delta1 in E11.5 Mib1 mutant spinal cord. In littermate control, Delta expression was restricted to the ventricular zone similar to the distribution of Mib1 transcript (Figs. 1 and 2, D and D′). In a high magnification view, Delta1 is localized in the plasma membrane in some cells, although it is present in intracellular vesicles in others, showing a salt and pepper pattern (32). In E11.5 Mib1 mutant, expression of Delta1 was reduced in the ventricular zone and mostly localized in the plasma membrane (Fig. 2, E and E′). Thus, it is likely that internalization of Delta is defective when Mib1 is absent.

Mis-specification of Neuronal Subtypes in the Ventral Spinal Cord without Mib1

Recent studies suggest that Notch signaling specifies selective neuronal subtypes, including V2a and V2b interneurons in the p2 domain within the developing spinal cord (7, 8). In Notch1 conditional knockouts and presenilin 1 null mice where Notch signaling is compromised, p2 progenitors become V2a interneurons instead of V2b interneurons. To determine whether the absence of Mib1, like other Notch-related genetic defects, influences the choice between V2a and V2b interneurons, we examined the neuronal subtypes in the ventral spinal cord of Mib1 mutants. In addition to the conditional mutant mice (cKO), we examined a drug-inducible Mib1 mutant mice (iKO) containing the Nes-Cre allele that allows drug-dependent elimination of Mib1 (25). When tamoxifen was administered from E10 to E12, Mib1 expression was almost undetectable at E12.5 (data not shown). Expression of Chx10, Gata3, and Sox1, which label V2a, V2b, and V2c interneurons, respectively, was examined by immunohistochemistry in E12.5 spinal cord (Fig. 3). We found that the ventral progenitors were virtually absent in Mib1 mutant mice, as indicated by the loss of the progenitor marker Sox1 and the appearance of the postmitotic markers Chx10 and Gata3 in the ventricular zone where only progenitors are supposed to be located (Fig. 3, B, C, E, F, H, and I). This is consistent with the phenotype found in Notch1 mutants, which display premature differentiation of neurons with reduced numbers of neural precursors (8). In the absence of Mib1, the number of V2a neurons increased (153.1 ± 5.437, p < 0.001) as compared with littermate controls (94.3 ± 4.89). The number of V2b interneurons declined (15 ± 1.592, p < 0.001) as compared with littermate controls (49.6 ± 2.633) (Fig. 3, A–F and M). To locate V2c interneurons, a population recently shown to originate from the p2 domain, we examined the expression of Sox1 (33). To focus on postmitotic V2c interneurons and eliminate Sox1⁺ progenitors, we only counted cells that expressed both Sox1 and NeuN, a postmitotic neuronal marker (data not shown). Sox1⁺ V2c interneurons were slightly increased in Mib1-inducible mutant mice (8 ± 1.125) as compared with the littermate controls (4.47 ± 0.583) (Fig. 3, G–I and M).

FIGURE 3.

Interneuron specification in the ventral neural tube is perturbed when Mib1 is down-regulated. A–L, expression of Chx10, Gata3, Sox1, and Hb9 in E12.5 Mib1 cKO, iKO, and littermate controls. For Mib1 iKO, tamoxifen was injected daily from E10 to E12. Premature Chx10⁺ V2a interneurons appeared in the ventricular zone close to the midline in Mib1 cKO and iKO mice (B and C, arrowheads). Numbers of Gata3⁺ V2b interneurons are reduced, although those of Sox1⁺ V2c interneurons (G–I, arrowheads) and Hb9⁺ motor neurons are mostly unchanged. M, numbers of cells expressing individual markers. Means ± standard errors are shown. Asterisks indicate statistically significant compared with controls (*, p < 0.05; ***, p < 0.001 in the unpaired Student's t tests). Scale bar in L, 200 μm for A–L.

Expression of progenitor markers that label most progenitor cells such as Sox1, Sox2, and NFIA revealed that numbers of progenitors were reduced in a broad region of the ventral spinal cord, including the p2 domain. To examine in more detail the specification of other ventral cell types in the absence of Mib1, we assessed the expression of additional markers such as Hb9, Evx1, and En1, which label motor neurons and V0 and V1 interneurons, respectively. We observed expansion of En1⁺ V1 interneurons and no change in motor neurons and V0 neurons in Mib1 mutant mice, although premature expression of Hb9 in the progenitor domain was found (Fig. 3, J–L and M, and data not shown). These results suggest that Mib1 mainly controls the fate specification of V1 and V2 interneurons but is dispensable for generation of motor neurons and V0 interneurons.

Specification of Glial Progenitors Is Compromised in Mib1 Mutant Spinal Cords

In addition to maintaining neural progenitors, the Notch signal is implicated in gliogenesis because the late progenitor pools preserved by the Notch signal retain the potential to generate glial cells after neurogenesis. Consistent with this, previous studies showed that activation of the Notch receptor triggered the generation of radial glia and astrocytes in the cortex (10). To examine whether the late progenitors that give rise to glial cells are under the influence of Mib1, we examined the development of glial progenitors in Mib1 mutants, using markers for progenitors or immature astrocytes such as NFIA, Nestin, p-Vimentin, Sox2, and Fgfr3 (34–36). NFIA is known to be present in progenitors as well as postmitotic motor neurons (34). Expression of NFIA in the progenitor domain was significantly reduced in E12.5 Mib1 mutant spinal cords, although its expression within the motor column was unchanged (Fig. 4, A–C and P). Examination of Nestin expression showed that radial processes were reduced in number and disorganized when Mib1 was absent. We also observed a similar reduction in p-Vimentin, Sox2, and Fgfr3 transcripts in the ventral spinal cord, consistent with the depletion of ventral progenitors in Fig. 3 (Fig. 4, D–P). Together our results suggest that Mib1-mediated Notch signaling is necessary for specification of glial cells.

FIGURE 4.

Assessment of glial progenitors in Mib1 mice. A–O, expression of NFIA, Nestin, p-Vimentin, Sox2, and Fgfr3 in E12.5 Mib1 mutants and controls. Drug was injected daily from E10 to E12 for Mib1 iKO. P, numbers of cells expressing NFIA and p-Vimentin. Asterisks indicate significant differences compared with controls (**, p < 0.01; ***, p < 0.001 in the unpaired Student's t test). Means ± S.E. are shown. Scale bar in O, 300 μm for A–O.

Mib1 Continues to Play a Role in the Later Period of Astrocytic Development

There are two types of astrocytes defined by their final destinations, morphology and marker expression, GFAP⁺ white matter astrocytes and GFAP⁻ gray matter astrocytes. In both types, immature astrocytes arise in the progenitor domain and migrate laterally. To decide whether the initial mis-specification of astrocytic progenitors in the absence of Mib1 affects the terminal differentiation of astrocytes, we monitored expression of GFAP in E17.5 Mib1 mutant spinal cords. In littermate controls, GFAP⁺ astrocytes with fibrous morphology were located within the white matter along the entire edge of the spinal cord. In contrast, GFAP expression in the Mib1 conditional mutant was significantly reduced in the white matter (Fig. 5, B and G). A very low level of disorganized GFAP immunoreactivity was found in the dorsal gray matter where GFAP expression is normally absent, indicating that astrocyte differentiation is either compromised or delayed (Fig. 5, B and G). The expression of AldhL1, a marker that labels both gray matter and white matter astrocytes, revealed that numbers of both types of astrocyte were significantly reduced in the absence of Mib1 (Fig. 5L) (37). Thus, it seemed likely that the initial loss of glial progenitors in the absence of Mib1 has an ongoing suppressive effect on astrocyte differentiation.

FIGURE 5.

Mib1 continues to be required for astrocyte generation as determined by removal of Mib1 at different developmental times. A–E, expression of GFAP in E17.5 spinal cords of Mib1 mutants and controls. Drug was administered at different time points as indicated. F–J, higher magnification views of A–E. K–M, expression of AldhL1 in Mib1 mutant and control animals. AldhL1 transcripts are reduced in the Mib1 mutant mice to a similar extent as GFAP. N, numbers of cells co-expressing NFIA and GFAP to quantify mature astrocytes. Means ± S.E. are shown. Asterisks indicate statistically significant difference compared with control (***, p < 0.001 in the unpaired Student's t test). Scale bars in E, 400 μm for A–E; in J, 100 μm for F–J; in M, 100 μm for K–M.

Recent studies have suggested that the Notch signal remains active in mature astrocytes and neurons (37–39). Consistent with this, Mib1 expression persisted among late progenitors when gliogenesis occurred (data not shown). Thus, it is possible that Mib1 plays a role in transmitting the Notch signal during late gliogenesis or the differentiation process. To locate the potential critical period for Notch signaling during glial differentiation, we eliminated Mib1 at various stages of embryonic development using tamoxifen-inducible Cre transgenic mice (25). We injected tamoxifen into the pregnant mice for 3 days at either E10 to E12, E12 to E14, or E13 to E15, and we harvested the embryos at E17.5. GFAP and AldhL1 expression was significantly reduced when the drug was applied from E10 to E12 (E10–12 Mib1 iKO) (Fig. 5, C, H, M, and N). We only observed a partial reduction of GFAP-expressing astrocytes in E12–14 iKO mice (33.4%) and almost no reduction in E13–15 iKO mice (0.8%) (Fig. 5, D, E, I, and J). Thus, it appears that, in addition to the early role of Mib1 during neurogenesis, this protein continues to be required for proper glial differentiation at least until around E14.

Generation of Oligodendrocytes Is Compromised in the Absence of Mib1

To decide whether Mib1 also plays a role in oligodendrocyte development, we examined the production and differentiation of oligodendrocytes by assessing the expression of Olig2, a marker of oligodendrocytes. Oligodendrocytes are derived from the pMN domain after motor neurons were generated. Oligodendrocyte precursors migrate and populate the entire region of the spinal cord. At E13.0, Olig2⁺ cells in the pMN domain were reduced in both Mib1 cKO and iKO mutants injected between E10 and E12 (Fig. 6, A–C and N). We also examined the number of oligodendrocytes in the Mib1 iKOs at E17.5. When the drug was delivered between E10 and E12 or E12 and E14, the number of oligodendrocytes was greatly reduced (67.5 and 44.4% compared with the control) (Fig. 6, C, F, G, K, L, and O). In contrast, it was unchanged when the drug was applied between E13 and E15 (Fig. 6, H, M, and O). This suggests that at least the early period of oligodendrocyte development (until ∼E14) was dependent on the Mib1-Notch signal.

FIGURE 6.

Oligodendrocyte generation is affected in Mib1 mice. A–C, expression of Olig2 in the spinal cord of Mib1 iKO injected between E10 and E12 and controls at E13.0. Oligodendrocyte progenitors in the pMN domain are depleted in the mutants as indicated by Olig2 expression. D–H, expression of Olig2 in E17.5 spinal cords of Mib1 and controls. Drug was administered at different time points as indicated. I–M, higher magnification views of D–H. N and O, quantification of Olig2-expressing cells. Means ± S.E. are shown. Asterisks indicate statistically significant differences compared with control (***, p < 0.001 in the unpaired Student's t test). Scale bar in C, 50 μm for A–C; in H, 200 μm for D--H; in M, 100 μm for I–M.

Functional Interaction between Mib1 and Other Notch Signaling Components during Neural Development

It is known that Mib1 induces the endocytic trafficking of DSL family ligands that trigger activation of the Notch signal pathway (3). To characterize the molecular events behind this, we generated Mib1 deletion constructs and tested their ability to induce Notch signaling in chicks by in ovo electroporation (21, 40). Misexpression of wild-type Mib1 did not have any effect on V2 interneuron specification, suggesting that Mib1 is not sufficient to initiate Notch signaling without Delta (Fig. 7, A and E). Mib1dn, a deletion mutant lacking the C-terminal RING domain, induced Chx10⁺ V2a interneurons at the expense of Gata3⁺ V2b interneurons when misexpressed in chick neural tubes (Chx10, 114.6%; Gata3, 57.4%) (Fig. 7, B, F, and Q). Mib1ΔRD3, which does not contain the third terminal RING domain, caused mis-specification of V2 interneurons to a greater extent, indicating that the last RING domain is critical for Mib1 activity (Chx10, 114.6%; Gata3, 70.7%) (Fig. 7, C, G, and Q). Furthermore, ectopic V2a interneurons were observed in the ventricular zone dorsal to the p2 domain in both Mib1dn and Mib1ΔRD3-expressing neural tubes. This indicates that premature differentiation of V2a interneurons occurs as in Mib1 mice (Fig. 7, B, C, and G). In contrast, activation of Notch signaling by introducing the NICD resulted in an increase in V2b interneurons and a reduction in V2a interneurons (Chx10, 88.6%; Gata3, 143.3%) (Fig. 7, D, H, and Q). Next, we electroporated Dll1 that reduced both V2a and V2b interneurons without changing the ratio between them (Chx10, 73.4%; Gata3, 81.5%) (Fig. 7, I, M, and Q). The reduction of cell numbers in V2 interneurons in the presence of Dll1 could be due to early depletion of p2 progenitors. Misexpression of Dll4 also showed no significant difference in a ratio between V2a and V2b interneurons (data not shown).

FIGURE 7.

Assessment of the requirement for Mib1 activity for specifying V2 interneurons. A–P, misexpression of Mib1, Mib1dn, Mib1ΔRD3, NICD, Dll1, Δicd, Δicd-Ub, or Δicd-Ub + Mid1dn in the chick neural tube by in ovo electroporation. Generation of Chx10⁺ V2a interneurons or Gata3⁺ V2b interneurons was measured. Note that ectopic V2 interneurons are found in the progenitor zone dorsal to the p2 domain (C and H, arrowheads). Q, % change compared with control. Means ± S.E. are shown. Asterisks indicate significant difference compared with control (*, p < 0.05; **, p < 0.01; ***, p < 0.001 in the unpaired Student's t test). Scale bar in P, 50 μm for A–P.

The fact that introduction of exogenous Dll ligands alone is relatively ineffective in altering V2 interneuron identity prompted us to reason that it may require additional modification such as ubiquitination. To test this, we generated Dll4Δicd that lacks an intracellular domain where Mib1 interacts (21). When we electroporated Dll4Δicd, both V2a and V2b interneurons were reduced, but the ratio between V2a and V2b was unchanged (Chx10, 73.0%; Gata3, 66.1%) (Fig. 7, J, N, and Q). This suggests that Delta activity to assign V2 interneuron specification may require Mib1. If the dominant-negative effect by Mib1dn is mainly due to Delta, one would expect that conjugating a mono-ubiquitin with Delta would rescue the dominant-negative effect of Mib1dn. To test this, we fused Dll4Δicd with ubiquitin to generate Dll4Δicd-Ub (21). Introduction of Dll4Δicd-Ub alone induced more V2b interneurons and lesser V2a interneurons, indicating that mono-ubiquitination of Dll4 is sufficient to transmit Notch signaling without Mib1 (Chx10, 81.0%; Gata3, 126.5%) (Fig. 7, K, O, and Q).

We next electroporated Δicd-Ub with Mib1dn. As expected, Δicd-Ub rescued the phenotype of Mib1dn to produce more V2b interneurons at the expense of V2a interneurons (Chx10, 96.8%; Gata3, 101.4%) (Fig. 7, L, P, and Q). Together, these results suggest that an ability of Mib1 to ubiquitinate Delta ligand via the RING domain is essential for V2 interneuron specification.

Cell-autonomous Role of Mib1 in Neuronal Specification

When Notch-Delta signaling occurs, the Delta-expressing cell, the signal-sending cell, becomes a neuron, whereas the Notch-expressing cell, the signal-receiving cell, remains a progenitor. Although we had demonstrated premature neuronal differentiation when Mib1 was absent, as a result of failure to transmit the Notch signal, it was unclear whether the cells with reduced Mib1 activity maintained their neurogenic ability when Delta activity could be compromised. To test this, we isolated neuronal progenitors from E14.5 Mib1^f/f animals and infected them with either a GFP or Cre retrovirus. To provide Notch signals from the neighborhood, we grew progenitors at low density on top of primary astrocytes, cells in which Notch signaling is known to be active (37, 41). If Mib1 was required for delivering but not receiving the Notch signal, the Mib1-deficient cells would remain progenitors. If Mib1 was also necessary for receiving the Notch signal, they would become neurons. At 3 days in vitro, most cells infected by GFP virus became neurons, as determined by MAP2 immunostaining (88.5% ± 3.073%) (Fig. 8, A–E). In contrast, the Mib1-deficient cells infected by Cre virus generated fewer neurons (59.9% ± 2.16%). There was no significant difference in Nestin⁺ progenitors between the two groups, indicating that Delta-expressing cells lacking Mib1 can receive the Notch signal and hence tend to remain as progenitors.

FIGURE 8.

Cell autonomous role of Mib1 in neuronal specification. A–D, representative images of cortical neuronal cultures from E14.5 Mib^f/f animals infected by GFP or Cre-GFP retroviruses and grown on top of astrocytes for 3 days in vitro. Coverslips were immunolabeled for GFP and Nestin or MAP2 as indicated. E, quantification of Nestin and MAP2-expressing GFP⁺ cells. F–G′, misexpression of Mib1dn or vector in E13.5 embryos by in utero electroporation. Pups were harvested at E16.5 and immunostained for Ngn2 (red) and GFP (green). H, quantification of Ngn2⁺ GFP⁺ cells per mm². Means ± S.E. are shown. Asterisks indicate significant differences compared with control (***, p < 0.001 in the unpaired Student's t test). Scale bars in D, 100 μm for A–D; in G, 200 μm for F and G; in G′, 100 μm for F′ and G′. I, schematic model depicting the roles of Mib1 in spinal neuron development. Mib1 maintains neural progenitors and regulates V2a-V2b interneuron specification as well as gliogenesis.

To test the cell-autonomous role of Mib1 in vivo, we electroporated Gfp or Mib1dn plasmids into E13.5 mouse embryonic brains and harvested them at E16.5. The number of neural progenitors located in the ventricular zone was counted by co-labeling them with the neural progenitor marker, Ngn2 (Fig. 8, F–H). We found that the density of Ngn2⁺GFP⁺ cells was reduced when Mib1dn was expressed. Although detailed mechanisms need to be further characterized, our result suggests that Mib1 is involved in the neuronal specification of Delta-expressing cells in a cell-autonomous manner.

DISCUSSION

Mib1 Is a Major Mediator of Notch-Delta Signaling

Numerous signaling components of the Notch-Delta signaling pathway have been identified (42). Nevertheless, the constituents of the signal-sending cells that function along with Delta and modulate its activity are less well known. One of the major steps in the activation of Delta ligands is their ubiquitination, which promotes internalization and signal transmission to the neighboring cells (21, 23). At least two ubiquitin ligases, Mind bomb (Mib) and Neuralized (Neur), are known to ubiquitinate DSL family members Delta and Jagged in the CNS (20, 21). Jagged ligands are expressed in the p1 domain, whereas Delta ligands are expressed in the p2 domain. The domain-specific and complementary expression of Delta and Jagged ligands suggests that individual ligands may deliver differential Notch signals. Nevertheless, we found that Mib1 is uniformly expressed in the ventral spinal cord, suggesting that it is a common ligase in the spinal cord that modifies several DSL ligands (data not shown). In contrast, Neur appears to play only a minor role in the developing neural tube because the elimination of both Neur1 and Neur2 does not cause any obvious developmental defects (24).

Mib1 is also known to ubiquitinate other proteins that are mainly known in different cellular context (40, 43). Although we cannot completely eliminate the possibility that these proteins may participate, several lines of evidence suggest that defective modification of Delta by Mib1 could be a major factor for the phenotype in Mib1 mutant mice. First, we observed premature neuronal differentiation and mis-specification of selective interneurons when Mib1 was absent. Development of radial glia, astrocytes, and oligodendrocytes was perturbed, all of which exactly phenocopy the mutants with defective Notch signaling (Fig. 8I) (7, 12). Second, Yoon et al. (12) demonstrated that the Mib1 mutant phenotype is rescued by activating Notch signaling using mouse genetics, indicating that the Notch receptor is downstream of Mib1. Third, expression of DAPK, another substrate of Mib1, is unchanged in the Mib1 mutant nervous system, which suggests that DAPK may not be relevant (44). Finally, we and others showed that introduction of the mono-ubiquitinated form of Delta is sufficient to rescue the defects caused by the dominant-negative Mib1 (21). Taken together, our findings suggest that Mind bomb-1 is the major mediator that transmits Delta-Notch signaling during CNS development.

Mib1 Conveys Delta-Notch Signaling for Interneuron Identity

In addition to its pan-neuronal role in neurogenesis, Delta-Notch signaling assigns cell identity to the specific interneuronal subtypes in the spinal cord, such as in the binary choice between V2a and V2b interneurons in the ventral spinal cord (6–8, 45). In mutants where Delta-Notch signaling is compromised, there is an imbalance in the ratio of the interneurons that share lineages (7, 8). Recently, another V2 interneuron called V2c, originating from the same p2 progenitors, has been identified. Fate mapping and analysis of marker expression revealed that V2c interneurons are in the same lineage as V2b but later segregate and display distinct sets of transcription factors (33, 46). In this study, we examined the production of all three V2 interneurons, as well as other ventral cell types, in Mib1 mutant mice. As expected, abnormal expansion of V2a interneurons and the loss of V2b interneurons were observed, supporting the view that Mind bomb-1 is the major mediator of Notch signaling in the spinal cord. Interestingly, the generation of V2c interneurons was only slightly altered in the Mib1 mutants although they derive from the same progenitors as V2b interneurons (33). V2c neurons are far fewer in number than V2a or V2b interneurons. This argues against a simple binary cell fate decision that should generate equal numbers of V2b and V2c interneurons. Thus, the V2c identity may not be directly controlled by Notch activity, unlike the other V2 interneurons. One possibility is that two different pools of p2 progenitors exist that generate either V2a/V2b neurons or V2b/V2c neurons. An alternative possibility is that transiently amplifying p2 progenitors exist that preferentially generate greater numbers of V2a and V2b neurons; Notch would then dictate the choice between V2a and V2b neurons whereas other factors such as Foxn4 or Sox1 would occasionally induce the V2c identity, independent of Notch signaling (45).

Several lines of evidence suggest that the role of Notch signaling in cell fate specification may not be restricted to V2 interneurons (6, 8). An imbalance of neuronal subtypes was reported involving other cell types such as dorsal interneurons and ventral cell types when Notch signaling was defective (6, 8, 47, 48). For instance, the generation of motor neurons in the pMN domain, a domain next to the p2 domain, is inhibited by Jagged2 activity (48). Likewise, we observed the premature appearance of motor neuron markers in chick neural tubes when a dominant-negative Mib1 was introduced (data not shown). In addition, premature extinction of the ventricular zone in the p1 domain as well as the p2 domain was found in Mib1 mutant mice. Finally, the glial defects found in Mib1 mutant mice suggest that Notch signaling influences gliogenesis in multiple ventral progenitor domains. In the absence of Mib1, the numbers of astrocytes, which arise in multiple ventral domains, and of oligodendrocytes, which originate from the pMN domain, were both reduced. Together, these observations suggest that Mib1 plays a major role in cell specification in the ventral spinal cord.

Spatiotemporal Segregation of Neurogenesis and Gliogenesis Is Controlled by Mib1-mediated Notch Signaling

During neural development, neural progenitors generate neurons and glia at different times and locations (9). Given that Notch inhibits neuronal differentiation and maintains neural progenitors, it seems likely that its early action in neurogenesis influences the behavior of reserved progenitors that later give rise to glial cells. In this way, the Notch signal appears to be the major pathway that links neurogenesis and gliogenesis by controlling the timing of progenitor competence.

Astrocytes and oligodendrocytes are two major glial cell types that arise in different regions of the developing spinal cord. In the cortex, astrocytes are generated in the dorsal telencephalon, whereas oligodendrocytes are produced in the basal forebrain area (9, 49). Likewise, in the spinal cord, astrocytes arise in the ventral progenitor domains after the appearance of neurons. Oligodendrocytes mainly originate in the pMN domain, and some from the dorsal spinal cord in later development. Thus, the region-specific generation of glial cells may be regulated by Notch signaling, in the same way as it affects spinal interneuronal identity. In Mib1 mutant ventral spinal cords, we observed a significant loss of radial glial cells, the precursors of astrocytes. Consistent with this, the number of astrocytes in both the gray and white matter of the spinal cords was reduced in Mib1 mutant mice. It is noteworthy that elimination of Mib1 in drug-inducible Mib1 mutant animals reduced numbers of astrocytes and oligodendrocytes at least up to ∼E14, indicating that Notch continues to act in late glial progenitors.

In addition to its primary role in astrocytes, Notch is thought to serve a dual role in oligodendrocyte generation; it promotes oligodendrocyte precursor cell development and suppresses later differentiation of oligodendrocytes (2, 15, 16, 48). This binary action of Notch was confirmed in vitro where astrocytes were induced and oligodendrocytes were reduced in a stepwise manner in clonal neurosphere stem cell cultures (50). Consistent with this, in Mib1 mutants, we first observed a reduction of Olig2 expression in the pMN domain after motor neurons had arisen. Likewise, the number of mature oligodendrocytes in the entire spinal cord was reduced in Mib1 mutant mice. This was observed only when Mib1 was eliminated in the early period between E10 and 12, a similar time window to the one in which astrocyte specification is perturbed. Together, these findings indicate that Mind bomb-1 mediates the Notch signaling that controls the generation of astrocytes and oligodendrocytes during the early period of glial development.

Cell Autonomous Role of Mib1 in Neuron-Glial Specification

There are two mechanisms by which Delta ligands control Notch signaling activity as follows: they trans-activate Notch in surrounding cells and cis-inhibit Notch in their own cells (51–54). Previous studies have shown that endocytic internalization of Mib1 is essential for transmitting Notch signals to neighboring cells (21). However, it is unclear whether the inhibition of Notch within the same cells is also modulated by endocytic events or ubiquitin modification by Mib1. Previous studies indicated that the cis-inhibition process might not require Mib1 (23, 55). First, Dll3, which does not undergo ubiquitination, acts primarily on cis-inhibition but not transactivation (56). Next, co-expression of Mib1 and Dll1 does not inhibit the cis-inhibitory activity of Delta ligands in vitro (21). Furthermore, transplantation experiments in zebrafish have demonstrated that Mib1-deficient cells surrounded by wild-type cells do not readily become neurons because they can still accept Notch signals without Mib1 (21). Finally, a Serrate mutant that cannot interact with Mib1 still cis-inhibits Notch activity in Drosophila (57). Using low titer retroviral infections and lineage tracing of progenitors, we demonstrated that Mib1-deficient cells can receive Notch signal and remain as progenitors. Interestingly, the portion of cells destined to become neurons were reduced without Mib1, implying that the neurogenic potential in Mib1-deficient cells could be compromised. Recently, the roles of Notch are rapidly expanding to mature neurons and the adult brain (58, 59). It is tempting to say that Mib1 may participate in later events in neuronal maturation in a similar context, which need to be further characterized in future studies.