PAR-1 phosphorylates Mind bomb to promote vertebrate neurogenesis

PAR-1 and aPKC influence neurogenesis in Xenopus embryos

To study a function for apical-basal polarity proteins for neuronal fate determination in the vertebrate brain and spinal cord, we examined effects of the polarity kinase PAR-1 and its regulatory kinase aPKC (Goldstein and Macara, 2007; Hurov et al., 2004; Suzuki et al., 2004) on primary neurogenesis in Xenopus embryos (Fig. 1, Fig. S1). Overexpressed PAR-1A/MARK3 (later referred to as PAR-1) increased the number of N-tubulin-positive primary neurons in 42% of injected embryos (n=55) (Figs. 1A, K). The same effect was observed with PAR-1B/MARK2 (data not shown). Conversely, embryos injected with two previously described PAR-1 morpholino oligonucleotides (MO), which specifically deplete PAR-1B/MARK2 protein (Kusakabe and Nishida, 2004; Ossipova et al., 2005)(Fig. S1A), but not a control MO (COMO), revealed a significant reduction in primary neuron number in the majority of injected embryos (90.5%, n=74; Fig. 1B, K, and data not shown). Of note, PAR-1BY MO did not have any effects on Wnt/β-catenin signaling (Ossipova et al., 2005), indicating that the function of PAR-1 in neurogenesis is not mediated by the Wnt pathway. By contrast, aPKC-CAAX, an activated membrane-targeted form of aPKC (Ossipova et al., 2007), suppressed neurogenesis (63%, n=119), whereas aPKC-N, a dominant interfering form of aPKC, increased the number of primary neurons (38%, n=106; Fig. 1C, D). In support of these observations, in situ hybridization for a different neuronal gene, Hox11L2 (Patterson and Krieg, 1999), revealed enlarged clusters of primary sensory neurons in PAR-1 RNA-expressing embryos, while a kinase-dead form of PAR-1 had an inhibitory effect (Fig. S1B, C, F), consistent with its dominant negative activity (Sun et al., 2001). Other neuronal markers, including Dll1 (Delta-like 1 or Delta1)(Chitnis et al., 1995) and Xaml-1 (Tracey et al., 1998), were also sensitive to the modulation of PAR-1 and aPKC functions (Fig. S1D-F and data not shown). These results were confirmed in embryo sections (Fig. S1G–L). Our observations demonstrate that PAR-1 and aPKC modulate neurogenesis in the opposite manner, consistent with the negative regulation of PAR-1 by aPKC during the establishment of apical-basal polarity.

To evaluate the number of neural progenitors and the size of the neural plate in the manipulated embryos, we studied the expression of pan-neural markers, such as Sox2 and Sox3 (Mizuseki et al., 1998; Pevny and Placzek, 2005), using in situ hybridization and immunohistochemistry. No significant effects on the Sox2 expression domain (Fig. 1E–H) and the number of Sox3-positive progenitors (Fig. 1I, J, L and data not shown) were detected in stage 15/16 embryos, injected with the same phenotypically active doses of PAR-1, aPKC-CAAX, PKC-N RNAs (n>30 for each group) and PAR-1BY MO or COMO (n > 80 for each group). Moreover, western analysis did not reveal any alteration in Sox3 protein levels (Fig. 1M). By contrast, manipulating BMP signaling by injection of BMP7 (Fig. 1M) or a truncated BMP receptor (tBR, Fig. 1L) readily affected Sox3 protein levels and progenitor number. Furthermore, we did not observe significant changes in mitotic cell number, when measured by phospho-histone 3 staining in embryo sections (Fig. S2A–F). Nevertheless, the same PAR-1 RNA coexpressed with the neuralizing factor Chordin activated the N-tubulin gene in ectodermal explants (Fig. S1M), further supporting our data obtained by in situ hybridization. To assess the location of ectopic neurons more precisely, we performed double in situ hybridization for N-tubulin positive cells and the pan-neural marker Sox2 or the epidermal keratin XK70. These experiments show that overexpressed PAR-1 expands the existing N-tubulin stripes within the neurogenic territory, which extends into the XK70-positive area, adjacent to the Sox2-positive domain (Fig. S1N, O). No ectopic neuronal differentiation was detected in response to PAR-1 in ventral ectoderm (data not shown), indicating that PAR-1 activity is only manifested within the normal neurogenic territory, and, therefore, differs from the neuralizing influence of Chordin and other BMP antagonists. Although we cannot rule out a possibility that PAR-1 and aPKC influence progenitor proliferation at early stages, our results suggest that PAR-1 influences neuronal differentiation rather than progenitor pool expansion.

PAR-1 and aPKC control mammalian neural progenitor differentiation

The observed changes in cell fate may result from intrinsic activities of these polarity proteins or be an indirect consequence of disrupted cytoskeletal organization and lack of tissue integrity (Ghosh et al., 2008; Kim and Walsh, 2007). To address this question, we used mouse cerebellar multipotent progenitor C17.2 cells, which do not exhibit epithelial morphology and are capable of neuronal differentiation (Snyder et al., 1992). These cells were infected with lentiviruses expressing GFP-tagged aPKC and PAR-1 proteins (Fig. 2A, B, S3A, B) and allowed to differentiate at low serum conditions. After five days of differentiation, frequency of neuron-specific tubulin TuJ1 staining in the control GFP-virus-infected (or uninfected) cells increased to 31.9 %, while the population of cells expressing the progenitor marker Nestin was reduced to 9.2 % (Fig. 2A, B and data not shown). Many cells expressing aPKC-CAAX retained Nestin expression (31.5 %) and did not differentiate into neurons (Fig. S3C, D). By contrast, overexpression of aPKC-N, wild-type PAR-1 or PAR-1T560A with a mutation in the aPKC phosphorylation site (Ossipova et al., 2007) caused extensive morphological differentiation, revealed by the formation of extended bipolar processes, and high frequency of TuJ1 staining (47.3 %, 52 % and 63.3 % of cells, respectively, Fig. 2A, B and data not shown). Together, these results demonstrate that PAR-1 enhances, whereas aPKC suppresses neuronal differentiation of C17.2 cells, and extend the data obtained in Xenopus embryos to the mammalian system. Of note, in contrast to what we observed in Xenopus embryos, PAR-1 and PKC-N decreased, whereas aPKC-CAAX increased C17.2 cell proliferation, when measured by phospho-histone 3 staining (Fig. 2C, S3E) and assessing cell growth in vitro (Fig. S3F). These findings indicate that the observed increase in TuJ1-positive neurons stimulated by PAR-1 is due to enhanced cell differentiation in the absence of cell proliferation, rather than the expansion of the progenitor pool. Our experiments also suggest that the observed effects of PAR-1 and aPKC on neurogenesis do not require epithelial tissue organization and likely represent intrinsic signaling properties of PAR-1 and aPKC.

Regulation of Notch signaling in signal sending cells

As the Notch pathway is a major regulator of neurogenesis, we hypothesized that the observed effects of PAR-1 and aPKC may be mediated by Notch signaling and measured Notch transcriptional activity in C17.2 cells cotransfected with various aPKC and PAR-1 constructs, using a luciferase reporter containing multimerized CSL-binding sites. We observed that the reporter was activated by aPKC-CAAX and inhibited by wild-type PAR-1 and PAR-1T560A, whereas PAR-1KD did not have a significant effect (Fig. 2D). Thus, our results link the antagonistic effects of PAR-1 and aPKC on neurogenesis to Notch pathway regulation, consistent with other studies implicating polarity proteins in Notch signaling (Bayraktar et al., 2006; Smith et al., 2007).

Since PAR-1 modulated Dll1, but not Notch ICD, signaling in Xenopus ectoderm (Ossipova et al., 2007), we hypothesized that PAR-1 regulates Dll1, rather than Notch, function. We therefore asked whether PAR-1 is active in signal-sending cells expressing Dll1, which are cocultured with the population of signal-receiving cells expressing a Notch reporter (Itoh et al., 2003; Lindsell et al., 1995)(Fig. 2E). Both PAR-1 and PAR-1T560A repressed Dll1-dependent reporter activity, while aPKC-CAAX and PAR-1KD had a modest enhancing effect, consistent with the view that PAR-1 inhibits Dll1 activity in signal-sending cells.

PAR-1 phosphorylates Mib and regulates its expression levels in vitro and in vivo

We next wanted to test the hypothesis that PAR-1 regulates Dll1 by phosphorylating E3-ubiquitin ligases, such as Neuralized or Mib. Supporting this possibility, a mass-spectrometry-based approach identified a mammalian PAR-1 homolog among proteins associated with Mib (Choe et al., 2007). The physical interaction of PAR-1 and Mib was confirmed by protein co-precipitation from cell lysates (Fig. 3A). Similar results were obtained using PAR-1 and Mib constructs with different epitope tags (data not shown). An immune complex kinase assay revealed that PAR-1, but not PAR-1KD, phosphorylated Mib and MibC1001S, an E3 ubiquitin ligase mutant (Itoh et al., 2003)(Fig. 3B), identifying Mib as an in vitro substrate for PAR-1.

We next examined the effect of PAR-1 on Mib in vivo. PAR-1 or PAR-1T560A downregulated Mib levels in lysates of embryos and cultured cells, whereas PAR-1KD had no effect (Fig. 3C, D, Fig. S4A, B). PAR-1 did not have a significant effect on endogenous α-tubulin or exogenous coexpressed GFP (Fig. 3C–E). Levels of the E3 ligase-deficient Mib mutant remained unaffected (Fig. S4B), indicating that the E3 ligase activity of Mib is necessary for the PAR-1 effect. Moreover, PAR-1 did not decreased Mib protein levels in cultured cells and Xenopus embryos treated with the proteasome inhibitor MG132, suggesting that PAR-1 promotes Mib degradation via the proteasome pathway (Fig. 3E and data not shown).

To assess whether endogenous PAR-1 functions to downregulate Mib, we examined Mib levels in Xenopus embryos depleted of PAR-1. Lysates of embryos injected with a PAR-1BY MO revealed a specific increase in Mib protein levels (Figs. 3F and S4C). A similar result was obtained with a second PAR-1BY MO with a different nucleotide sequence, but not with a control MO (data not shown). Together, these observations support a model, in which PAR-1 phosphorylates Mib to stimulate its proteasome-dependent degradation.

PAR-1 inhibits Dll1 ubiquitination and signaling activity, but does not influence a Mib-independent form of Dll1

Since Mib is a known major regulator of Dll1 ubiquitination and signaling activity, our results predict that PAR-1 should influence Dll1 ubiquitination. Consistent with this prediction, ubiquitination of Dll1 in both Xenopus embryos and mammalian cells was inhibited by PAR-1, but not PAR-1KD (Fig. 4A, B). These results support the model, in which PAR-1 inhibits Dll1 signaling by antagonizing Mib function.

As Dll1 signaling activity depends on the ubiquitination of its intracellular domain by Mib, we fused Dll1 to the ubiquitin moiety to generate a Mib-independent construct (Dll1ΔC-Ub)(Itoh et al., 2003). We next examined PAR-1 effects on the ability of Dll1 and Dll1ΔC-Ub to promote Notch signaling and inhibit neuronal differentiation. PAR-1 rescued neurogenesis downregulated by Dll1, but not Notch ICD (Fig. 5A, B, S5), supporting the idea that PAR-1 functions at the level of Dll1, upstream of Notch receptor activation. By contrast, embryos expressing Dll1ΔC-Ub showed reduced number of neurons even in the presence of PAR-1 RNA (Fig. 5A, B). Our observations were further supported in the Dll1-dependent co-culture assay by measuring Notch reporter activity in mammalian cells. Whereas both Dll1 and Dll1ΔC-Ub activated the reporter to comparable levels, PAR-1 suppressed Dll1, but not Dll1ΔC-Ub responses (Fig. 5C). These experiments demonstrate that PAR-1 inhibits Dll1 activity both in Xenopus ectoderm and mammalian cells in a Mib-dependent manner.

Specific phosphorylation sites in Mib are required for PAR-1-dependent changes in Dll1 ubiquitination

To assess the functional significance of Mib phosphorylation by PAR-1, we wanted to identify sites in Mib that are phosphorylated by PAR-1. We generated a series of mutated Mib proteins with substitutions in the putative serine/threonine phosphorylation motifs (Fig. 6A), and assessed their levels in the absence or presence of PAR-1 in Xenopus embryo lysates. Whereas the majority of Mib mutant proteins remained sensitive to PAR-1-mediated degradation, the levels of the M2/M8 Mib construct, carrying mutations in the M2 and M8 sites, were not altered by exogenous PAR-1 (Fig. 6B), indicating that these sites are critical for Mib regulation. In vitro kinase assays using wild-type and mutated Mib constructs confirmed that these sites are subject to phosphorylation by PAR-1 (Fig. 6C, D). The phosphorylation analysis of S804/S806 residues was carried out in the Mib construct with deleted amino-terminal phosphorylation sites. Similarly to other PAR-1 sites identified in Dsh and PAR-3 (Sun et al., 2001, Benton and St. Johnston, 2003), the S804 has an immediately adjacent upstream basic residue (Fig. 6A). Importantly, these phosphorylation sites were conserved among vertebrate Mib1, but not Mib2 homologs (Fig. 6A), suggesting functional significance.

We next analyzed whether PAR-1 can inhibit the activity of the M2/M8 mutant in the Dll1 ubiquitination assay. PAR-1 inhibited Dll1 ubiquitination by the wild-type Mib, but not by the M2/M8 Mib (Fig. 6E). This finding suggests that the inhibitory effect of PAR-1 on Delta ubiquitination involves Mib phosphorylation.

Embryo culture and microinjections, in situ hybridization and lineage tracing

Fertilization and embryo culture have been performed as described (Itoh et al., 2005). DNA constructs, RNA synthesis, morpholino oligonucleotides, RT-PCR and mutagenesis are described in detail in Supplementary on-line materials. Embryos were microinjected in 1/3 x MMR, 2 % Ficoll-400 (Pharmacia) in the animal pole with 10 nl of a solution containing 0.1 – 2 ng of RNA per blastomere at the two to four-cell stage, and cultured in 0.1 x MMR until desired stages. In loss-of-function experiments, PAR-1BY MO and PAR-1BX MO (Ossipova et al., 2005) or control MO were injected at 5 –10 ng per blastomere. Of note, PAR-1BY MOs did not have an effect on Wnt/β-catenin signaling.

In situ hybridization and XGal staining were carried out using standard techniques (Harland, 1991) with the following anti-sense probes: N-tubulin (Oschwald et al., 1991), Sox2 (Mizuseki et al., 1998), XDll1 (Chitnis et al., 1995), Hox11L2 (Patterson and Krieg, 1999), Xaml (Tracey et al., 1998), epidermal type I keratin XK70 (Winkles et al., 1985). For double in situ hybridization, probes were synthesized using digoxygenin- or fluorescein RNA Labeling Mix (Roche) and detected with corresponding antibodies (Roche). Magenta phosphate, Fast Red and NBT/BCIP (Roche) were used for chromogenic reactions. For 10–12 μm sections, embryos were embedded in coldwater fish gelatin/sucrose mixture as described (Fagotto and Gumbiner, 1994) and cryosectioned using the Leica cryostat CM3050. Images were digitally acquired on a Zeiss Axiophot microscope. Quantification of results is presented as a percent of embryos with a conspicuous phenotypic change. RNA injected embryos usually revealed a range of phenotype severity.

Immunoprecipitation, western analysis and kinase assay

Immunoprecipitation (IP) and Western blotting were carried out with cell and embryo lysates essentially as described (Gloy et al., 2002; Itoh et al., 1998). Ubiquitination assay in cultured cells was as described (Itoh et al., 2003) with the following modifications. HEK-293T cells were transiently transfected with 0.4 μg total of expression plasmid DNA per well in 6-well plates using Effectene (Qiagen). Cells were collected 48 hr after transfection, lysed in 300 μl of lysis buffer (Gloy et al., 2002) containing protease inhibitor cocktail (Roche Diagnostics) and analyzed by immunoblotting or IP/immunoblotting. For IP, the 200 μl lysates were cleared by centrifugation at 10,000 g at 4°C for 5 min to remove cellular debris and incubated with 20 μl of anti-myc antiboby (9E10) overnight at 4°C. The immunoprecipitates were incubated with protein A-agarose beads (12.5 μl of 50 % suspension) for 1 hr at 4°C, washed 3 times with ice-cold lysis buffer and resuspended in a sample buffer. One-fourth of the material was analyzed by 8% SDS PAGE. The equivalent of 0.5 embryo per lane was loaded for analysis of embryo lysates, and the equivalent of 10–15 embryos per lane was loaded for analysis of immunoprecipitated proteins. Peroxidase activity was detected by enhanced chemiluminescence as described (Itoh et al., 1998). Protein detection was performed using Fujifilm Image Reader LAS-3000 System and software (FujiFilm). Quantification of band intensities by densitometry was carried out using NIH Image J software, using a representative of at least three independent experiments.

The following antibodies were used: anti-HA and anti-myc monoclonal antibodies are hybridoma supernatants of 9E10 and 12CA5 cells (Roche), anti-myc monoclonal antibodies 9E10 (Roche), anti-β-catenin 8E7 (Millipore, USA), anti-tubulin (BioGenex), anti-GFP monoclonal antibodies (Santa Cruz Biotechnology), anti-GFP JL8 monoclonal antibodies (Clontech), anti-Flag M2 (Sigma), anti-tubulin B512 (Sigma), anti-PAR-1 (Ossipova et al., 2005). Immune complex kinase assays were as described in (Ossipova et al., 2005) using in vitro synthesized myc-tagged proteins in TNT system (Promega) and Molecular Dynamics PhosphorImager. The following DNA templates were used: pCS2-Myc-PAR-1, pCS2-Myc-PAR-1 kinase dead, pCS3-Myc-Mib, pCS3-Myc-MibC1001S, pCS3-Myc-M2, pCS3-Myc-M8 (691–1031).

Cell culture, MG132 treatment, and Notch reporter assay

HEK-293T and NIH3T3 cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM, Gibco) supplemented with 10% fetal bovine serum (FBS, Gemini-Bioscience) and penicillin/streptomycin (Sigma). C17.2 progenitor cells were cultured in DMEM containing 10 % FBS, 5 % horse serum (Snyder et al., 1992; Zhang et al., 2006). For differentiation assays, C17.2 cells were seeded on coverslips in 12-well plates (1×10⁵ per well) and shifted the next day to DMEM containing 2 % FBS. For MG132 treatment, the inhibitor was added at 10 μm to HEK293T cell culture medium for 6 hrs before harvesting the cells. Staining with anti-Nestin (R401, DSHB) or anti-TuJ1 (Covance) mouse monoclonal antibodies was carried out in the beginning (day 0) and at the end (day 5) of differentiation.

For Notch reporter assays, 5×10⁴ C17.2 cells were plated in 12-well plates for transfection the next day using linear polyethylenimine (M.W. 25,000, Polysciences). For transfection, each well received 0.25 μg of pGL3-11CSL-Luc (a gift of J. Kitajewski), 0.02 μg of pTK-Renilla-Luc (Promega) and 0.5 μg of pCS2-Flag-aPKC-CAAX, pCDNA3.1-aPKC-N, pCS2-Myc-PAR-1, pCS2-Myc-PAR-1T560A or pCS2-Myc-PAR-1KD were used. Specificity of pGL-11CSL-Luc responses to Notch ICD was established in separate experiments (Ezan et al., data not shown). Total DNA amount was adjusted to 1 μg by supplementing pCDNA3 DNA. For the co-culture assay, 293T cells were transfected in 6-well plates at 80 % confluence with 2 μg pCS2-Dll1-myc or empty pCDNA3 together with aPKC-CAAX and PAR-1 constructs and NIH3T3 cells were transfected in 24-well plates at 50 % confluence with 0.3 μg of pGL3-11CSL-Luc, 0.005 μg of TK-Renilla-Luc, 0.2 μg of pBOS-Notch1 and 0.5 μg of pCDNA3. One day after transfection, HEK293T cells were added on to NIH3T3 transfected cells, co-cultured for an additional 24 h and harvested in 100 μl of lysis buffer (Promega) for luciferase activity measurement. Firefly and Renilla luciferase activities were determined using the Dual Luciferase assay system (Promega) and the Veritas microplate luminometer (Turner Biosystems). Normalized results are expressed as the means ± s. d. of three independent experiments performed with triplicate samples. Statistical analysis was performed using two-tailed paired Student’s t-test.

Immunocytochemistry of Xenopus embryos and cultured cells

Indirect immunofluorescence analysis on Xenopus embryo cryosections was performed as described previously (Ossipova et al., 2007; Fagotto and Gumbiner, 1994). Embryos were cryosectioned at 10μm using the Leica cryostat CM3050. The following antibodies were used: anti-XSox3c (Zhang et al., 2003, 1:500), anti-GFP (B2, Santa-Cruz 1:200), anti-phospho-histone H3 (Cell Signaling, 1:300). Images were digitally acquired on a Zeiss Axiophot microscope at 200x magnification. Aphidicolin treatment (Sigma) was performed as described (Hardcastle and Papalopulu, 2000). Representative sections of 5–10 embryos per experimental group are shown.

C17.2 cells were fixed with 4% paraformaldehyde for 10 min, permeabilized with PBS, 0.2% Triton X-100, blocked with PBS-5% goat serum, and probed with anti-β-tubulin III (TuJ1, Covance; 1:500) or anti-Nestin (R401, DSHB, 1:50) and Cy3-conjugated goat anti-mouse antibody (Jackson ImmunoResearch; 1:100). Coverslips were mounted in Vectashield/DAPI (Vector Laboratories). The number of labeled cells (at least 50 cells per field) was determined under a fluorescence microscope (AxioImager, Zeiss) with Axiovision imaging software (Zeiss). The ratio of TuJ1- or Nestin-positive cells to GFP-expressing cells for each field is presented. Results are expressed as the means ± s. d. of five random fields per sample with triplicate samples per group from three to five independent experiments in parallel cultures.

Lentiviral vector construction, lentivirus production and transduction

Self-inactivating lentiviral vector pVVPW, pVVPW-IRES-eGFP, the envelope plasmid pMD.G and the packaging plasmid pCMV R8.2 were kindly provided by Luca Gusella (Battini et al., 2006). GFP-tagged versions of aPKC-CAAX, aPKC-N, PAR-1 and PAR-1T560A were generated from cDNAs encoding rat PKC zeta, accession number NM_022507 (Parkinson et al., 2004), and Xenopus PAR-1A, accession number AF509737, and PAR-1T560A (Ossipova et al., 2005; Ossipova et al., 2007) by Pfu-mediated PCR. GFP-tagged aPKC-CAAX and aPKC-N were subcloned in frame into pXT7-GFP vector (Itoh et al., 2000) as Eco RI inserts. PAR-1 and PAR-1T560A were subcloned as Bgl II/Sal I inserts. To generate GFP–tagged versions in pVVPW, aPKC-CAAX and aPKC-N were amplified using pXT7-eGFP constructs as templates and inserted into the Nhe I and Eco RI sites of VVPW. GFP–tagged PAR-1 and PAR-1T560A in pVVPW were subcloned as Nhe I/Not I inserts. Plasmids were verified by sequencing. Primer oligonucleotide sequences and further construction details are available upon request.

Lentiviruses were produced by transient transfection as described (Battini et al., 2006), with several modifications. Briefly, 3×10⁶ HEK293T cells were plated into 10 cm² plate and adherent cells were transfected 24 h later using 150 μl of linear polyethylenimine (1 mg/ml, Polysciences) with a total of 16 μg of plasmid DNA: 8 μg of the lentiviral vector, 6 μg of pCMV R8.2, and 2 μg of pMD.G. Cell culture medium was replaced 24 hours after transfection with medium supplemented with 4 mM sodium butyrate. Culture supernatant were collected at 48–96 h after transfection, filtered and concentrated by ultracentrifugation (50,000 g, for 2.5 h at 4°C). The viral pellet was resuspended in complete culture medium and aliquots were stored at −80°C. Viral titers estimated from GFP expression in HEK293T cells transduced with serial dilutions of concentrated lentiviruses, ranged from 10⁶ to 10⁸ transducing units per milliliter. For transduction, cells were seeded in 12-well plates (1×10⁵ per well) and shifted the following day to virus-containing medium in the presence of 8 μg/ml polybrene (Fluka). Viruses were removed after 24 h of incubation, and cells were expanded for differentiation, western analysis and immunocytochemistry. More than 90% of cells expressed GFP at least 7 days after transduction as assessed by flow cytometry, western analysis and immunofluorescence.