Abstract
In developing neural tube, the basic helix-loop-helix (bHLH) transcription factor Olig2 interacts with the homeodomain transcription factor Nkx2.2 at two distinct stages. During neuronogenesis, a cross-repressive interaction appears to establish the precise boundary between the p3 and pMN domains. At later times, a cooperative interaction is noted because Nkx2.2 promotes maturation of oligodendrocyte progenitor cells specified by expression of Olig2. We show here that the Olig2 protein can form a physical complex with Nkx2.2 protein in mammalian cells and yeast two-hybrid trap assay. This interaction is specific because Olig2 does not bind to a biologically irrelevant homeodomain transcription factor (Nkx6.1), and Nkx2.2 does not interact with a biologically irrelevant bHLH protein (NeuroD). Deletion mapping analysis suggests that formation of an Olig2-Nkx2.2 physical complex is insufficient for the induction of oligodendrocyte progenitors in developing spine; however, the protein-protein interaction observed might be important for the cross-repressive interaction between Olig2 and Nkx2.2 that helps to establish the pMN-p3 boundary in the developing spinal cord.
Introduction
During vertebrate CNS development, coincident spatial-temporal expression of transcription factors is thought to comprise a combinatorial code that specifies formation of dedicated neural progenitor cells and emerging waves of specialized neuronal and glial progeny (Jessell, 2000; Bertrand et al., 2002). Two structurally distinct types of transcription factors are principal components of this code. Homeodomain (HD) transcription factors are defined by a 60 amino acid DNA-binding motif that forms a highly conserved helix-turn-helix structure. Within the NK-2 subclass of HD transcription factors, Nkx2.2 defines the p3 progenitor domain and is specifically required for production of V3 interneurons (INs) that express Sim1 (Briscoe et al., 1999) and maturation of oligodendrocytes (Qi et al., 2001).
Transcription factors with basic helix-loop-helix (bHLH) domains have fundamental roles during neuronal and glial production (Gowan et al., 2001; Bertrand et al., 2002). The basic domain is required for interaction with specific DNA elements, whereas the helix-loop-helix motif mediates homodimerization or heterodimerization, which is required for biological activity. Cell type-specific bHLH proteins such as those involved in spinal cord patterning typically do not form homodimers. Rather they interact with E proteins (E12, E47, E2-2, and HEB), a family of bHLH factors that are broadly expressed (Murre et al., 1989; Zhuang et al., 1998). Among the cell type-specific bHLH transcription factors expressed in the developing spinal cord, function of Olig2 is required for motor neuron and oligodendrocyte development in the pMN domain (Lu et al., 2002; Park et al., 2002; Takebayashi et al., 2002a; Zhou and Anderson, 2002).
Gain-of-function and loss-of-function genetic analyses suggest that Olig2 interacts with Nkx2.2 at two distinct stages in developing spine. During neuronogenesis, a cross-repressive interaction between Olig2 and Nkx2.2 contributes to formation of the boundary between the p3 and pMN domains in the ventral neural tube (Mizuguchi et al., 2001; Novitch et al., 2001; Sun et al., 2001; Zhou et al., 2001). At later times, a cooperative interaction is noted. Delayed differentiation of oligodendrocyte precursors in Nkx2.2 null mice indicates that Nkx2.2 promotes maturation of oligodendrocyte progenitor cells specified by expression of Olig2 (Qi et al., 2001). In chick neural tube, this cooperative interaction can be visualized by ectopic formation of oligodendrocytes after co-electroporation of Olig2 and Nkx2.2 (Sun et al., 2001; Zhou et al., 2001).
The biochemical mechanisms underlying cooperative and cross-repressive interactions between Olig2 and Nkx2.2 are poorly defined. In principle, these structurally distinct transcription factors should function independently of each other within the cell nucleus by regulating the expression of distinct gene sets. In this study, however, we show that Nkx2.2 and Olig2 physically interact with each other to form a specific intranuclear complex. The domains required for this biochemical interaction are not sufficient for oligodendrocyte formation during CNS development but they are sufficient for a cross-repressive interaction.
Materials and Methods
Plasmid constructs. The full-length mouse Olig2 cDNA was subcloned into a pcDNA3.1 expression vector (Invitrogen, San Diego, CA) with the cytomegalovirus (CMV) promoter and an in-frame Myc- or V5-tag at the C terminus. A series of deletions of Olig2 were amplified by PCR using the full-length mouse Olig2 cDNA as template and then subcloned into a pRK5 vector (gift from Dr. Anthony Lanahan, Johns Hopkins Medical Institute, Baltimore, MD) with the CMV promoter and an in-frame Myc-tag at the N terminus.
The full-length mouse Nkx2.2 cDNA was subcloned into a pFLAG-CMV-2 expression vector (Sigma, St. Louis, MO) with the CMV promoter and an in-frame Flag-tag at the N terminus, or pcDNA3.1 expression vector (Invitrogen) with an in-frame Myc-tag at the C terminus. A series of deletions of Nkx2.2 were amplified by PCR using the full-length mouse Nkx2.2 cDNA as template and then subcloned into a pRK5 vector with the CMV promoter and an in-frame hemagglutinin (HA)-tag at the N terminus. Similarly, the full-length mouse NeuroD cDNA was subcloned into pRK5 expression vector with the N terminus Myc-tag. The full-length mouse Nkx6.1 (a gift from Dr. M. German, University of California San Francisco) and E12 cDNA were subcloned into a pRK5 vector with the N terminus HA-tag.
Yeast two-hybrid assay. Bait vector harboring full or partial Olig2 cDNA and target vector harboring full or partial Nkx2.2 cDNA were cotransformed into yeast (pJ69-4A or Y190 strains) using a standard protocol (Clontech, Cambridge, UK). Full-length Olig2 and its bHLH domain were subcloned into a pAS2-1 vector (Clontech) or the pPC97 vector, respectively, with the in-frame Gal4 DNA binding domain at the N terminus. Full-length Nkx2.2 and a series of Nkx2.2 deletions were subcloned into the pACT2 vector (Clontech) or the pPC86 vector, respectively, with the in-frame Gal4 DNA binding domain at the N terminus. The PC97 and PC86 vectors were gifts from Dr. Anthony Lanahan (John Hopkins Medical Institute). Colonies displaying two-hybrid interactions were visualized on Leu-Trp-His-plates, as well as on the more stringent β-galactosidase filter assay.
Cell culture and transient transfection. Cos7 cells were cultured in DMEM with 10% fetal bovine serum and penicillin (100 U/ml)-streptomycin (50 μg/ml) at 37°C with 5% CO2 to a density of 2-3 × 106 cells per 100 mm plate at the time of transfection. Transfections were performed for 1-4 hr by using the Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's protocol. Equal amounts of plasmid DNA were used for cotransfection. For immunostaining experiments, Cos7 cells were plated on coverslips in the 24-well plates.
Western blot analysis and co-immunoprecipitation. Cos7 cells were washed in PBS 24-40 hr after transfection, and protein lysates were harvested in buffer H (50 mm Tris, pH 7.5, with 5 mm EDTA, 0.5 mm Na3VO3, 5 mm NaF, adding fresh 0.2% NP-40, 1 mm DTT, 1 mm PMSF, 1 μg/μl aprotinin, 1 μg/μl leupeptin before use) and rotated at 4°C for 20 min; 0.11 volume of 170 mm NaCl was then added to extract the nuclear protein. The lysates were further rotated at 4°C for 20 min, centrifuged at 13,000 rpm for 10 min at 4°C, and finally frozen at -80°C until they were used.
For each co-immunoprecipitation (Co-IP), 1-2 μg of monoclonal anti-Myc (Santa Cruz Biotechnology, Santa Cruz, CA) or anti-Flag (Sigma) antibodies were mixed (1:1) with 100 μl of protein G Sepharose beads (Amersham Biosciences, Arlington Heights, IL), preswelled in buffer H; 600 μl of buffer H was then added. The antibody-protein-G bead mixture was rotated at 4°C for 1 hr and then centrifuged at 13,000 rpm. Between 100 and 400 μl of cell lysate was added in the mixture, and Buffer H was added to the final volume of 600-700 μl. The mixture was rotated at 4°C for 1-2 hr. Precipitates were then washed twice with 1 ml of ice-cold buffer H, resuspended in 120 μl 1.5× SDS sampling buffer, and boiled for 5 min. After centrifugation, the supernatant was size fractionated on a 12 or 15% SDS-PAGE gel and transferred to a polyvinylidene difluoride membrane (Millipore, Bedford, MA).
The Western blot analysis was performed using monoclonal or polyclonal anti-Myc antibodies (1:500; Santa Cruz Biotechnology), monoclonal anti-HA antibodies (1:5000; Babco, Richmond, CA), monoclonal anti-Flag antibodies (1:800; Sigma), and monoclonal anti-Nkx2.2 antibodies [1:500; Developmental Studies Hybridoma Bank (DSHB), Iowa City, IA]. The secondary alkaline phosphatase-conjugated anti-mouse or rabbit antibodies (1:5000; Promega, Madison, WI) were used and reacted with AttoPhos Fluorescent Substrate (1:25; Promega). Proteins were visualized with STORM 860 Scanner (Molecular Dynamics, Sunnyvale, CA).
Immunohistochemistry and in situ hybridization. Transfected cells were washed briefly in cold PBS, fixed in 4% PFA in PBS, pH 7.0, at 4°C for 10 min, and then washed with PBS. Cells were labeled with the first antibodies in PBS with 0.1% Triton X-100 and 10% normal goat serum at room temperature for 1 hr. After they were washed with PBS, cells were labeled with the second antibodies at room temperature for 1 hr followed by 4′,6′-diamidino-2-phenylindole (DAPI) staining to visualize the nuclei. For immunohistochemistry on sections, chick embryos were fixed in 4% PFA in PBS and cryosectioned. The following primary antibodies were used at the indicated dilutions: murine Nkx2.2 (1:500; DSHB), rat Nkx2.2 (1:5000; gift from T. Jessell, Columbia University, New York), Olig2 (1:4000), Myc (1:200; Invitrogen), and HA (1:5000; Babco). Secondary antibodies used were anti-mouse IgG FITC (1:100; Jackson ImmunoResearch, West Grove, PA), and anti-rabbit IgG Cy3 (1:200, Jackson ImmunoResearch). Photomicrographic images were collected on a Nikon E600 microscope and RT Monochrome SPOT digital camera.
In situ hybridization was performed using a modified protocol (Ma et al., 1997), and detailed instructions are available on request. Digoxygenin (DIG)-labeled antisense mRNA probes for cSox10 and cSim1 were prepared by in vitro transcription.
Electroporation in ovo. The chick neural tubes at Hamburger and Hamilton (HH) stage 10-12 were electroporated unilaterally (five 50 msec pulses at 25 V) with cDNAs (3 μg/μl) using an ECM830 electro-squareporator (BTX Inc.). For co-electroporation, equal amounts of plasmid DNA (3 μg/μl each) were mixed before electroporation. Embryos were analyzed after 48 hr incubation (HH stages 21-23). At least 10 electroporated embryos were analyzed for each plasmid DNA construct. Results documented in the figures are representative of at least five embryos. Expression constructs driven by the CMV promoter were used in all experiments. The advantage of these CMV-driven constructs is that fate changes are confined to the primary targets of the electroporation. By contrast, replication-competent avian retrovirus expression vectors allow secondary infection of neural progenitors (Morgan and Fekete, 1996). Note that gene expression driven by CMV vectors is attenuated in vivo 2-3 d after the electroporation.
Results
Specific physical interaction between Nkx2.2 and Olig2
To identify co-regulator proteins that might interact with Olig2 and Nkx2.2, we performed a series of antibody pull-down experiments. Cos7 cells were transfected or cotransfected with expression constructs encoding Flag-tagged Nkx2.2 and Myc-tagged Olig2. Somewhat to our surprise, these pull-down experiments showed that Olig2 interacts with Nkx2.2 when the two transcription factors are coexpressed (Fig. 1A).
Specific physical interaction between Nkx2.2 and Olig2. A, Cos7 cells were cotransfected with expression constructs encoding Flag-tagged Nkx2.2 and Myc-Tagged Olig2. Western blots from anti-Flag immunoprecipitates (IP) and also total cell lysates were immunoblotted (IB) with anti-Myc and anti-Nkx2.2 antibodies. Olig2 was immunoprecipitated by full-length Nkx2.2 (arrow). Note that Olig2 protein is distributed within four separate bands in total cell lysates. Two of these bands may reflect alternate translation initiation sites in the Olig2 cDNA sequence (Lu et al., 2000), and the other two may reflect degradation-post-translational modification products. B, Cos7 cells were cotransfected with expression constructs encoding Flag-tagged Nkx2.2 and Myc-Tagged NeuroD. Western blots from anti-Flag immunoprecipitates and also from total cell lysates were immunoblotted with anti-Myc and anti-Nkx2.2 antibodies. NeuroD was not immunoprecipitated by the full-length Nkx2.2 (arrow indicates the expected NeuroD band, and arrowhead indicates nonspecific IgG band.). C, Cos7 cells were cotransfected with expression constructs encoding Myc-tagged Olig2 and HA-Tagged Nkx6.1. Western blots from anti-Myc immunoprecipitates and also from total cell lysates were immunoblotted with anti-HA and anti-Myc antibodies. Olig2 fails to coimmunoprecipitate with Nkx6.1 (arrow indicates the expected Nkx6.1 band, and arrowhead indicates nonspecific IgG band).
The bHLH transcription factor NeuroD is expressed in postmitotic neurons in the CNS (Lee et al., 1995; Roztocil et al., 1997) and has no known cooperative or cross-repressive interactions with Nkx2.2. As shown in Figure 1B, Nkx2.2 does not interact with this biologically irrelevant bHLH protein. The HD transcription factor Nkx6.1 is broadly expressed in the ventral region of developing spine and thus overlaps the expression domain of Olig2 at early times in development as does Nkx2.2 (Sander et al., 2000; Novitch et al., 2001); however, Nkx6.1 has no known biological interactions with Olig2. As shown in Figure 1C, Olig2 does not coprecipitate with Nkx6.1. Together these results suggest that Olig2 interacts with Nkx2.2, either directly or indirectly, to form a protein complex and this interaction is not merely adventitious.
Physical interactions are mediated by the Nkx2.2 homeodomain and the Olig2 bHLH domain
To identify regions of Nkx2.2 and Olig2 that are required for physical interaction, we constructed a series of deletion mutants of Nkx2.2 with HA-tag and Olig2 with Myc-tag. To identify regions of Nkx2.2 that are necessary for physical interaction with Olig2, we performed antibody pull-down experiments using full-length Olig2-Myc and the panel of Nkx2.2 deletion mutants. Full-length Nkx2.2 contains a homeodomain and an NK-2-specific domain (NK2-SD) (Fig. 2). As shown, Nkx2.2 deletions without homeodomains do not interact with Olig2. Conversely, Nkx2.2 deletions that contain an HD are coimmunoprecipitated with Olig2 (Fig. 2). Thus, for physical interaction with Olig2, the homeodomain of Nkx2.2 is both necessary and sufficient.
The homeodomain of Nkx2.2 is both necessary and sufficient for physical interaction with Olig2. A, Cos7 cells were cotransfected with expression constructs encoding Myc-tagged full-length Olig2 and HA-tagged deletions of Nkx2.2. Western blots of the anti-Myc coimmunoprecipitates were immunoblotted with anti-HA and anti-Myc antibodies. The scheme of full-length Nkx2.2 is shown on the top left, and schemes of Nkx2.2 deletions are shown above each pair of lanes. Arrows indicate the predicted positions of coimmunoprecipitated Nkx2.2 deletion fragments. Nonspecific bands of IgG heavy chain and light chain are indicated by black arrowheads and white arrowheads, respectively. As shown, deletion mutations of Nkx2.2 containing HD domains can be coimmunoprecipitated by the full-length Olig2, whereas deletion mutants that lack HD cannot be coimmunoprecipitated. B, Western blots of total cell lysates were immunoblotted with anti-HA and anti-Myc antibodies.
In reciprocal experiments, we defined the regions of Olig2 that are required to form a complex with full-length Nkx2.2. As shown in Figure 3, salient structural features of Olig2 include the bHLH domain, a cluster of contiguous serine-threonine (S/T) residues in the N-terminal region, and an alanine-rich region (Ala) at the C terminus. In repeated attempts, we were unable to express constructs encoding the Olig2 bHLH domain alone at levels sufficient to detect by immune precipitation (see Fig. 8); however, combinatorial cotransfections summarized in Figure 3 show that the C- and N-terminal domains of Olig2 are individually dispensable to formation of a stable complex with Nkx2.2. Conversely the bHLH domain is essential and possibly sufficient.
The bHLH domain of Olig2 is both necessary and sufficient for physical interaction with Nkx2.2. A, Full-length and deletion mutations of Olig2 (all Myc tagged) were cotransfected with the Flag-tagged full-length Nkx2.2 into Cos7 cells. Anti-Flag immunoprecipitates were immunoblotted with anti-Flag and anti-Myc antibodies as in Figure 2. The scheme of full-length Olig2 is shown on the top left, and schemes of Olig2 deletions are shown on the top. The predicted positions of coimmunoprecipitated Olig2 deletion fragments are indicated by arrow. Nonspecific bands of IgG heavy chain and light chain are indicated by black arrowheads and white arrowheads, respectively. As shown, Olig2 mutations containing bHLH domains are coimmunoprecipitated by the full-length Nkx2.2. B, Western blots of total cell lysates were immunoblotted with anti-Myc and anti-Nkx2.2 antibodies. The doublet of Nkx2.2 bands shown here may reflect partial degradation of the Flag-tagged Nkx2.2 during handling or a distortion of thick gel during transferring on the membrane.
Interaction domains of Olig2 and Nkx2.2 are sufficient for cross-repressive function. Expression constructs encoding full-length Nkx2.2 (with Flag tag) and full-length Olig2 and its mutations with Myc tag were co-electroporated into the neural tube of stage 10-12 chick embryos. After 48 hr, embryos (stage 21-23) were harvested for analysis. Adjacent sections were immunolabeled with polyclonal (red) or monoclonal (green) antibodies targeted to the Nkx2.2 or Olig constructs as indicated and hybridized in situ with a probe targeted to mRNA for chick cSim1. Ectopic expression was directed to the left half of the neural tube, and the contralateral (non-electroporated) side served as a control. Schemes of deletions of Olig2 and Nkx2.2 with amino acid count are shown on the left. The level of induction of Sim1 as visualized by in situ hybridization is summarized by + (induction) or - (no induction) on the right. The embryos shown and results summarized are representative of at least five separate embryos in each case. A, The ectopic expression of full-length Nkx2.2 induces Sim1 expression. B, Cross-repressive interaction: co-electroporation of Nkx2.2 and Olig2 inhibits induction of Sim1. C, Olig2 lacking the bHLH domain is not sufficient for cross-repressive interaction with Nkx2.2. As shown, co-electroporation of the ΔOlig2 (1-110 aa) lacking the bHLH domain with full-length Nkx2.2 does not inhibit Sim1 induction by the Nkx2.2. D, Olig2 fragment containing the amino terminal and bHLH domains is sufficient for cross-repressive interaction. Note however that loss of the C-terminal domain attenuates the repression somewhat. E, Olig2 fragment containing the C terminal and bHLH domains is sufficient for cross-repressive interaction.
Olig2 and Nkx2.2 interact in yeast two-hybrid assays
Antibody pull-down experiments of the sort shown in Figures 1, 2, 3 are vulnerable to artifact. Depending on conditions of salt and pH used to wash the Sepharose beads, protein-protein interactions can be made to appear or disappear. Accordingly we looked for evidence of a physical interaction between Olig proteins and Nkx2.2 in living cells. As shown in Table 1, Olig proteins and Nkx2.2 interact in yeast two-hybrid assays. As was noted in antibody pull-down experiments, the only protein domains required for this interaction are the bHLH domain of Olig2 and the homeodomain of Nkx2.2. In fact, the two-hybrid interaction is more robust when the sequences flanking the bHLH and homeodomain are deleted (Table 1).
Olig2 and Nkx2.2 interact via bHLH and HD domains in yeast two-hybrid assays
Olig2 and Nkx2.2 interact within mammalian cells
Another indication of an Olig2-Nkx2.2 interaction in living cells is suggested by an interesting relocalization of Olig2 protein in Cos7 cell transfection experiments. Within the ventral neural tube of developing mice and chicks, endogenous Olig2 protein appears to be sequestered within the nucleus as expected of a bHLH transcription factor (Novitch et al., 2001; Sun et al., 2001; Zhou et al., 2001; Takebayashi et al., 2002b). As shown in Figure 4A, however, Olig2 protein is sequestered within the cytosol and perinuclear region of Cos7 cells when it is expressed by itself. Coexpression with a class E bHLH partner protein (E12) allows translocation of Olig2 to the nucleus (Fig. 4B); however, as shown, coexpression with Nkx2.2 will also relocalize Olig2 to the nucleus (Fig. 4C). By contrast, coexpression with Nkx6.1 does not relocalize Olig2 to the nucleus (Fig. 4D).
Olig2 and Nkx2.2 interact within mammalian cells. A, Full-length Olig2, Nkx2.2, and HA-tagged Nkx6.1 were transfected into Cos7 cells individually. Transfected cells were labeled with anti-Olig2 antibodies (red), anti-Nkx2.2 antibodies (green), and anti-HA antibodies (green). Olig2 is localized in the cytoplasm and nucleus, whereas Nkx2.2 and Nkx6.1 are within the nuclei. B, Full-length Olig2 and HA-tagged E12 were cotransfected into Cos7 cells. Transfected cells were labeled with anti-Olig2 antibodies (red) and anti-HA antibodies to visualize E12 (green). Olig2 is recruited into nuclei by cotransfection with E12 (yellow in overlay). C, Full-length Olig2 and Nkx2.2 were cotransfected into Cos7 cells. Transfected cells were labeled with anti-Olig2 antibodies (red) and anti-Nkx2.2 antibodies (green). Olig2 is localized into nuclei by cotransfection with Nkx2.2 (yellow in overlay). D, Full-length Olig2 and Nkx6.1 were cotransfected into Cos7 cells. Transfected cells were labeled with anti-Olig2 antibodies (red) and anti-HA antibodies to visualize Nkx6.1 (green). Nkx6.1 fails to bring Olig2 into nuclei in cotransfected cells.
To map the domains of Olig2 and Nkx2.2 that mediate this response in mammalian cells, we first examined the location of deletion mutants in single transfection assays. As shown in Figure 5A, the default location of Nkx2.2 is within the nucleus of Cos7 cells. The HD domain is both necessary and sufficient for this nuclear localization. Conversely, the default location of Olig2 is within the cytosol-perinuclear region of Cos7 cells, and the entire protein sequence is required to maintain this position. Deletion of any material to the N- or C-terminal side of the bHLH domain allows Olig2 to gain access to the nucleus (Fig. 5B) (see Discussion).
Mapping Nkx2.2 and Olig2 domains that specify subcellular localization of the wild-type proteins. A, Schemes of a series of Nkx2.2 deletion mutations. Full-length Nkx2.2 with Flag tag and deletion mutations with HA tag are illustrated schematically according to amino acid content. The HD (135-190 aa) and the NK2-SD (199-215 aa) are labeled separately. Subcellular localizations of each deletion fragment in transfected Cos7 cells were established by using anti-HA antibodies and DAPI to visualize nuclei. As shown, the HD domain of Nkx2.2 is essential for nuclear localization. B, Schemes of a series of Olig2 deletion mutations. The full-length Olig2 and each deletion mutation with Myc tag are illustrated according to amino acid content. The bHLH domain, S/T, and Ala-rich domain are labeled separately. Subcellular localization of each deletion fragment in transfected Cos7 cells was visualized by using anti-Myc antibodies and DAPI to visualize nuclei. Full-length Olig2 is localized in the cytoplasm and the nucleus, whereas all deletions of Olig2 with bHLH domains are localized in nuclei.
To define the regions of Nkx2.2 essential for recruiting full-length Olig2 into the nucleus, we cotransfected full-length Olig2 and Nkx2.2 deletion mutants into Cos7 cells. As shown in Figure 6, Nkx2.2 deletion mutants that lack HDs do not affect Olig2 localization; however, Nkx2.2 deletion mutants that contain only HDs likewise fail to recruit Olig2 into the nucleus (Fig. 6). The ΔNkx2.2 (112-273 aa) mutant, which contains both HD and NK2-specific domains, mimics full-length Nkx2.2 in recruiting Olig2 into the nucleus (Fig. 6). Thus, in this mammalian cell-based assay, the Nkx2.2 homeodomain is essential but not sufficient for interaction with Olig2. The NK2-specific domain contributes to the interaction.
The HD and NK2-specific domains of Nkx2.2 are required for relocalization of full-length Olig2 into the nucleus. A series of Nkx2.2 deletion mutations with HA tag and full-length Olig2 with Myc tag were cotransfected into Cos7 cells. Transfected cells were labeled with anti-Olig2 antibodies (red) and anti-HA antibodies (green). Schemes of Nkx2.2 deletion mutations are shown on the left. As shown, only Nkx2.2 mutations with the HD and the NK2-specific domains can bring Olig2 into the nuclei of Cos7 cells.
Interaction domains of Olig2 and Nkx2.2 are not sufficient for cooperative biological function
Data shown to this point indicate that Olig2 and Nkx2.2 form a specific physical complex in vitro and in vivo. To define the role of this interaction in their cooperative function to form oligodendrocytes, we electroporated (singly or together) a panel of Olig2 and Nkx2.2 expression vectors into developing chick neural tubes. In confirmation of previous studies (Sun et al., 2001; Zhou et al., 2001), neither Olig2 nor Nkx2.2 alone promotes oligodendrocyte development in chick neural tube (Fig. 7A,B), whereas the combination stimulates ectopic formation of oligodendrocyte progenitor cells marked by Sox10 (Fig. 7C). Deletion of any material flanking the HD domain ablates the ability of Nkx2.2 to cooperate with Olig2 in this biological response. Conversely, deletion of any material outside the bHLH domain ablates the ability of Olig2 to cooperate with Nkx2.2 (Fig. 7D-F) (data not shown). Thus the domains of Olig2 and Nkx2.2 that are sufficient to promote a physical interaction of the two proteins are not sufficient to promote this developmental response. The physical interaction of these two proteins may be irrelevant to their cooperative function in promoting formation of oligodendrocytes (see Discussion). Alternatively, other subregions of both proteins are required to initiate formation of oligodendrocytes in developing neural tube.
Interaction domains of Olig2 and Nkx2.2 are not sufficient for cooperative biological function. Expression constructs encoding full-length Olig2 (with Myc tag), Nkx2.2 (with HA tag) and their deletion mutations were electroporated into the neural tube of stage 10-12 chick embryos. At 48 hr after electroporation, the embryos (stage 21-23) were harvested for analysis. Adjacent sections were immunolabeled with polyclonal or monoclonal antibodies against the Olig2 or Nkx2.2 expression products in red or green as indicated. Ectopic expression was directed to the left half of the neural tube, and the contralateral (non-electroporated) side served as a control. The embryos shown and the data summarized represent results from at least five analyzed embryos in all cases. Schemes of the Olig2 and Nkx2.2 expression constructs are shown to the left. The level of induction of cSox10 as visualized by in situ hybridization is summarized by the + (induction) and the - (no induction) on the right. A, The ectopic expression of full-length Olig2 does not induce Sox10 expression. B, The ectopic expression of full-length Nkx2.2 does not induce Sox10 expression. C, Cooperative interaction: co-electroporation of Nkx2.2 and Olig2 induces Sox10 expression. D, The homeodomain of Nkx2.2 is not sufficient for cooperative interaction with Olig2. As shown, co-electroporation of the ΔNkx2.2 (112-194 aa) with full-length Olig2 does not induce Sox10. E, Olig2 lacking the C terminus is not sufficient for cooperative interaction with Nkx2.2. As shown, co-electroporation of the ΔOlig2 (1-170 aa) with full-length Nkx2.2 does not induce Sox10. F, Interaction of the homeodomain of Nkx2.2 and the bHLH domain of Olig2 is not sufficient for the induction of Sox10. Co-electroporation of ΔOlig2 (1-170 aa) containing the bHLH domain and ΔNkx2.2 (112-273 aa) lacking the N terminus does not induce Sox10 expression.
Interaction domains of Olig2 and Nkx2.2 are sufficient for cross-repressive function
Previous studies show that Nkx2.2 is necessary and sufficient to direct the development of Sim1-expressing V3 INs (Briscoe et al., 1999, 2000) and that the trophic effect of Nkx2.2 on V3 INs is antagonized by Olig2 (Sun et al., 2001). To assess whether this cross-repressive interaction requires a direct physical interaction between Nkx2.2 and Olig2 proteins, we conducted co-electroporation experiments as for Figure 7 above. Positive and negative controls are presented in Figure 8, A and B. As shown, Nkx2.2 alone promotes ectopic expression of the Sim1 marker for V3 INs (Fig. 8A), and this response is blocked by co-electroporation with full-length Olig2 (Fig. 8B). As noted also in our experiments with Cos7 cells (Fig. 3), we were unable to detect expression of a deletion fragment containing only the Olig2 bHLH domain in chick neural tube; however, combinatorial cotransfections show that an N-terminal deletion of Olig2 has no effect on the cross-repressive interaction (Fig. 8E), whereas a C-terminal deletion has only a minor effect (Fig. 8D). Thus, the domains of Olig2 required for a physical interaction with Nkx2.2 are at least qualitatively sufficient for cross-inhibitory effects, in contrast to cooperative responses.
Discussion
The transcription factors Olig2 and Nkx2.2 belong to structurally distinct protein families (Chien et al., 1996; Harvey, 1996). As shown here, however, these two proteins recognize each other in antibody pull-down and yeast two-hybrid protocols. In each of these two conventional assays for detecting protein-protein interactions, the bHLH domain of Olig2 and the HD domain of Nkx2.2 appear necessary and sufficient for the interaction.
Additional evidence for an interaction between Olig2 and Nkx2.2 is obtained by monitoring the subcellular localization of Olig2 in Cos7 cells. The open reading frame of Olig2 does not encode a canonical nuclear localization signal (Lu et al., 2000). Deletion analysis (Fig. 5) shows that the Olig2 bHLH domain itself can function as a nuclear translocation signal. It would appear, however, that the bHLH domain is somehow masked in native Olig2 because the full-length protein is excluded from the nucleus in single transfection experiments (Fig. 5). Coexpression with a class E bHLH partner protein (E12) allows translocation of Olig2 to the nucleus (Fig. 4B); however, as shown, coexpression with Nkx2.2 will also relocalize Olig2 to the nucleus (Fig. 4C). In this relocalization assay, the NK2-specific domain of Nkx2.2 contributes to the interaction with Olig2. Therefore, as suggested schematically in Figure 9, a partner protein is required to unmask the nuclear localization signal (NLS) of Olig2 (Fig. 9).
An occluded NLS motif in Olig2. Our current model of Olig2 proteins is shown. The bHLH domain, S/T domain, and Ala-rich domain are labeled separately. The model suggests that Olig2 has a functional NLS that may possibly be the bHLH motif itself. Olig2 monomers fold in such a way that the bHLH-NLS motif is occluded. The motif is unmasked by (1) interaction with bHLH partner proteins such as E12 (Fig. 4), (2) interaction with co-regulator proteins such as Nkx2.2 (Fig. 4), or deletion of flanking amino acids (Fig. 5).
Taken individually, the antibody pull-down experiments, the yeast two-hybrid studies, and the nuclear relocalization experiments must be interpreted with caution. Collectively, however, the results are internally consistent with each other and support the conclusion that Olig2 and Nkx2.2 interact to form a specific physical complex. Our data do not address whether the interaction is direct or indirect. In addition, we cannot quantitate the relative strength of this interaction for different deletion constructs. On a qualitative level, some of the interactions, for example that between full-length Olig2 and ΔNkx2.2 (112-273 aa), appear weaker than others. Within cells that express both transcription factors, it is conceivable that only a minor fraction of each protein participates in formation of the complex. Biological functions of the complex, if any, are unknown; however, we can gain some insight into functional relevance of the complex by combinatorial analysis of our deletion mutants in the developing chick neural tube. Our structure-function analysis is summarized in Table 2.
Summary of distinct regions of Olig2 and Nkx2.2 essential for nuclear localization, physical interaction, and cross-repression
As indicated in Table 2, the domains required for the physical interaction are insufficient for, and perhaps irrelevant to, the cooperative interaction between Olig2 and Nkx2.2 that results in formation of Sox10-positive oligodendrocyte progenitor cells. This is perhaps not surprising. The bHLH and HD transcription factors have very distinct DNA recognition motifs. Conventional wisdom would suggest that their cooperative functions are exerted independently of each other via regulation of factor-specific gene sets; however, although the interaction between Olig2 and Nkx2.2 is not sufficient for oligodendrocyte development, our data do not preclude the possibility that the interaction is necessary for oligodendrocyte formation. To assess this possibility, we attempted to determine whether Nkx2.2 deletion mutations act in a dominant-negative manner to preclude development of endogenous oligodendrocytes in chick neural tube. None of our experiments (data not shown) suggested such a dominant-negative action; however, these negative results are ambiguous because endogenous oligodendrocytes appear at later times [embryonic day 5 (E5)-E6] in the ventral neural tube of the chick at a point where expression of ectopic genes from the CMV promoter used in our experimental system is much attenuated (see Materials and Methods).
We do not know how tightly Olig2 and Nkx2.2 bind to each other in vivo. Within cells that express both transcription factors at comparable levels, it is possible that only a minor fraction of each transcription factor is involved in the physical complex that we describe here. Conversely, however, domains required for the protein-protein interaction overlap to a first approximation with domains required for cross-inhibitory effects in the chick neural tube assay (Table 2). This overlap might have functional overtones for establishment of a precise boundary between adjacent p3 and pMN progenitor domains during development.
Recent work has established that Olig2 expression defines the pMN and is necessary for motor neuron development in mice (Lu et al., 2002; Takebayashi et al., 2002a; Zhou and Anderson, 2002). In the neural tube of Nkx2.2-/- embryos, the expression domain of the Olig2 extends ventrally and MN formation is augmented (Briscoe et al., 1999; Qi et al., 2001). Conversely, misexpression of Nkx2.2 in the chick neural tube suppresses Olig2 expression in a cell-autonomous manner and inhibits motor neuron development (Novitch et al., 2001; Zhou et al., 2001); thus the cross-repressive action of Nkx2.2 may be mediated at the transcriptional level. In contrast, ectopic expression of Olig genes does not repress expression of Nkx2.2 (Novitch et al., 2001). The data summarized in Figure 8 are consistent with the view that the cross-repressive action of Olig2 on Nkx2.2 is exerted at the post-transcriptional level via a physical interaction between the two proteins. In this scenario, the p3-pMN boundary formed within a narrow zone wherein Olig2 and Nkx2.2 are coexpressed. Within a few cells ventral to the developing boarder, transcription of Olig2 is suppressed by Nkx2.2 (present in relative excess). In a few cells dorsal to the border, formation of V3 interneurons is inhibited at the post-transcriptional level by entrapment of Nkx2.2 in a physical complex with Olig2 (also present in relative excess).
If a physical interaction with Olig2 antagonizes the ability of Nkx2.2 to induce V3 interneurons, it might be expected that the loss of Olig2 function (knock-out) would lead to an increase in V3 interneuron cell number; however, a dorsal expansion of V3 interneuron domain is not found in either Olig2 or Olig1/2 knock-out neural tube. Instead, a ventralization of the p2 domain, evidenced by ectopic high levels of Pax6 and Irx3 in the pMN, is observed (Lu et al., 2002; Zhou and Anderson, 2002). A cross-repressive interaction of Pax6 and Nkx2.2 has been shown previously by others (Ericson et al., 1997; Briscoe et al., 1999, 2000). It is likely therefore that ventralization of the Pax6 expression accounts for repression of Nkx2.2 expression and maintenance of the pMN-p3 border in Olig2-/- mice.
A post-transcriptional switching mechanism in neural cell fate choice has been noted in other contexts. For example, Lhx3 binds to the nuclear LIM interactor (NLI) to direct V2 interneuron formation. In motor neurons, however, Isl1 is available to compete for binding to NLI, displacing Lhx3 to a high-affinity binding site on the C-terminal region of Isl1 and thereby transforming Lhx3 from an interneuron-promoting factor to a motor neuron-promoting factor (Thaler et al., 2002; Lee and Pfaff, 2003). This switching mechanism enables specific LIM complexes to form in each cell type and ensures that neuronal fates are tightly segregated.
Interactions of HD proteins and bHLH transcription factors have been demonstrated previously in pancreas and pituitary development and in muscle cell differentiation. The ubiquitously expressed bHLH protein E47 can form a complex with the homeodomain proteins Lmx1.1 or PDX-1 and activate Insulin transcription (Johnson et al., 1997; Ohneda et al., 2000). During pituitary development, the bHLH protein NeuroD1 is expressed in the corticotroph cells. Within these cells, NeuroD1 is dimerized with the ubiquitously expressed bHLH protein Pan1. This heterodimer forms a complex with the HD protein Pitx1 and regulates target gene POMC expression (Poulin et al., 2000). In developing muscle, myogenic bHLH proteins form heterodimers with the pan-bHLH protein E2a and go on to interact with the HD protein Pbx1-Meis1/Prep (Knoepfler et al., 1999). Therefore regulation of target gene expression via interactions of HD proteins and bHLH transcription factors may be a general mechanism in multiple developmental processes.
Footnotes
T.S. thanks the National Multiple Sclerosis Society for an advanced postdoctoral fellowship. D.H.R. is a Claudia Adams Barr Investigator and Harry Weaver Neuroscience Scholar of the National Multiple Sclerosis Society. These studies were funded by National Institute of Neurological Diseases and Stroke Grant NS4051 (D.H.R.), National Institute of Child Health and Development Grant HD24296 (C.D.S.), and the Dana-Farber/Mahoney Center for Neuro-Oncology.
Correspondence should be addressed to either of the following: David H. Rowitch, Department of Pediatric Oncology, Dana-Farber Cancer Institute, Harvard Medical School, 44 Binney Street, Boston, MA 02115, E-mail: david_rowitch{at}dfci.harvard.edu; or Charles D. Stiles, Department of Cancer Biology, Dana-Farber Cancer Institute, Harvard Medical School, 44 Binney Street, Boston, MA 02115, E-mail: charles_stiles{at}dfci.harvard.edu.
Copyright © 2003 Society for Neuroscience 0270-6474/03/239547-10$15.00/0
↵* T. S. and H. D. contributed equally to this work.
