A role for the ELAV RNA-binding proteins in neural stem cells: stabilization of Msi1 mRNA

In mammals, adult neurogenesis was demonstrated to be active in two main brain regions, namely the subventricular zone (SVZ) and the dentate gyrus of the hippocampus, where neural stem cells (NSCs) were isolated (Alvarez-Buylla and Garcia-Verdugo, 2002; Galli et al., 2003). NSCs are characterized by plasticity and multipotency so that they may proliferate and/or undergo differentiation into functional progeny according to the environmental conditions (Bottai et al., 2003). These features make them ideal candidate targets for the cure of neurodegenerative diseases, although the molecular mechanisms that determine a NSC to proliferate or to differentiate towards a neuronal or astroglial fate are only partially elucidated (Qian et al., 2000; Roegiers and Jan, 2004). Post-transcriptional regulation, exerted by RNA-binding proteins (RBPs) as an alternative mechanism to the activity of transcription factors in modulating gene expression, has recently been shown to have a role in the development of the central nervous system (CNS) and, in particular, in the control of the NSC division mode (Okano et al., 2002). Several neural-specific RBPs have been described so far to affect splicing, transport, translation and stability of target mRNAs, and can cause severe neurological disorders when altered (Perrone-Bizzozero and Bolognani, 2002).

The Musashi-1 (Msi1) RBP was first reported to be required for the proper development of the neural sensory organ in Drosophila (Nakamura et al., 1994), whereas in mammals it is commonly considered a specific marker for stem/progenitor cells of neural origin (Kaneko et al., 2000; Maslov et al., 2004). Msi1 acts as a translational suppressor by binding to the 3′-untranslated region (3′UTR) of specific mRNA targets. In this way the proliferation state of NSCs is maintained by inhibiting the translation of the membrane protein Numb, involved in the Notch/Delta signaling cascade (Imai et al., 2001), and of the cyclin-dependent kinase inhibitor p21^WAF-1 (Battelli et al., 2006).

The neuronal-specific ELAV (nELAV) RBPs, which are associated with the pathological condition of paraneoplastic encephalomyelitis (Szabo et al., 1991), are necessary and sufficient to induce neuronal differentiation in mammalian cells (Akamatsu et al., 1999; Kasashima et al., 1999), as is their elav ortholog in Drosophila (Robinow and White, 1991). The three mammalian nELAV family members, HuB, HuC and HuD, are commonly referred to as early and specific markers of post-mitotic neurons during CNS development (Barami et al., 1995; Okano and Darnell, 1997; Wakamatsu and Weston, 1997), whereas the fourth member, HuR, is ubiquitously expressed. The ELAV RNA-binding activity promotes the stabilization and/or translation of an array of transcripts containing the AU-rich consensus element (ARE) in their 3′UTR (Antic and Keene, 1997). ARE sequences represent well-documented cis-acting regulatory motifs usually identified in mRNAs of genes with a high turnover rate (Shaw and Kamen, 1986), and are recognized by several ARE-binding proteins which exert a different and often opposite role on mRNA fate (Bevilacqua et al., 2003). nELAV target genes are endowed with a wide variety of biological functions (Gao et al., 1994), from cell growth regulation (Levine et al., 1993) to brain maturation and maintenance (Antic et al., 1999; Aranda-Abreu et al., 1999; Chung et al., 1997).

The expression of Msi1 and nELAV RBPs has been initially described to be spatiotemporally sequential during the development of murine CNS (Sakakibara et al., 1996), although recent data suggest that nELAV proteins might have an important role as early as the neuronal-specific commitment of NSCs (Akamatsu et al., 2005).

In the present work we show that nELAV proteins are expressed and colocalize with Msi1 in the adult mouse SVZ and in cultured NSCs/progenitors, where they show a specific and ARE-dependent binding activity for the Msi1 transcript. In particular, we find that nELAV RBPs exert a stabilizing activity on Msi1 mRNA decreasing its turnover rate in vitro and promoting its translation in vivo. Such findings suggest a mechanistic correlation between nELAV and Msi1 RBPs in controlling the proliferation/differentiation activities of neural stem/progenitor cells.

Examination of the mouse Msi1 gene revealed the presence of a putative ARE consensus sequence in its 3′UTR, catalogued as belonging to group III in the ARE-mRNA database (ARED) (Bakheet et al., 2001). Since AREs are cis-acting regulatory sites, they are usually well conserved throughout phylogenesis (Asson-Batres et al., 1994), and indeed a comparative analysis of Msi1 3′UTRs in ortholog genes revealed that the human sequence showed a 75% identity with its murine counterpart (Fig. 1A). The EST-expanded rat Msi1 3′UTR displayed the same degree of conservation (74%), and alignment of the three sequences gave almost full identity of a region of 24 bp, including the putative functional ARE. On the other hand, Danio rerio, Xenopus laevis, Caenorhabditis elegans and Drosophila melanogaster Msi1 orthologs neither contained the ARE signature nor showed any significant homology along the entire 3′UTR region. Also Msi2 gene, the ubiquitously expressed member of the Musashi family (Sakakibara et al., 2001), didn't show any ARE consensus sequence in its 3′UTR. Therefore, we hypothesized that the ARE sequence present in the Msi1 gene could have a role in influencing its mRNA stability and in the post-transcriptional regulation of its expression only in mammals. Prediction of the mouse Msi1 3′UTR secondary structure by the Sfold RNA-folding algorithm (Ding et al., 2004) clearly showed that the ARE sequence is exposed in a single-stranded loop, which is likely to be accessible to ARE-binding proteins (Fig. 1B).

In order to investigate the putative RNA-binding activities associated with the Msi1 transcript, we performed RNA/protein UV crosslinking experiments in the presence of radiolabeled Msi1 3′UTR riboprobe and protein extracts from both mouse brain and cultured NSCs. After UV irradiation, the resulting mRNA-protein (mRNP) complexes were separated by SDS-PAGE and the electrophoretic pattern was compared with that obtained for the growth-associated protein 43 (Gap43), a well-known mRNA target of the nELAV RBPs. A partially overlapping profile was evident with common bands of approximately 95, 65 and 42 kDa (Fig. 1C), the latter being previously shown to correspond to the RNA-binding activity of HuD (Chung et al., 1997). We observed two additional mRNP complexes specific to the Msi1 3′UTR, with apparent molecular weights of 37 and 32 kDa, the lower band being more abundant in NSCs and the upper one in whole brain lysate.

These preliminary findings prompted us to investigate the possible role of the nELAV RBPs as trans-acting factors in the binding and regulation of the Msi1 transcript. To this purpose after UV crosslinking assays, the resulting mRNP complexes were immunoprecipitated with the pan anti-nELAV antibody (16A11), which is known to recognize the three neuronal RBPs HuB, HuC and HuD, but not HuR (Marusich et al., 1994) (Fig. 1D). A 42 kDa complex was immunoprecipitated in the presence of the Msi1 3′UTR riboprobe, confirming the existence of an nELAV RNA-binding activity for the Msi1 transcript both in mature brain and in NSCs.

As several RBPs are reported to bind to their own mRNA (Chu et al., 1991; Schaeffer et al., 2001), including nELAV proteins (Abe et al., 1996; Samson, 1998), we also immunoprecipitated the products of the UV crosslinking reaction using the anti-Msi1 antibody, but the Msi1 transcript was not recovered (Fig. 1E).

Our preliminary data suggest the existence of a functional correlation between the Msi1 gene, associated with the maintenance of NSC proliferation, and the nELAV RBPs, usually considered early markers of newly generated neurons. NSCs isolated from developing or mature brain can be grown in vitro as floating cell aggregates denominated neurospheres, which are characterized by an indefinite proliferation potentiality and by the multipotency to differentiate into the principal neural phenotypes (Morshead and van der Kooy, 2004). We therefore looked for nELAV protein expression in neurospheres obtained from adult mouse brains. Immunolabeling images with the anti-Msi1 antibody showed colocalization with the nELAV antigens throughout the sphere (Fig. 2A-C), suggesting a stem identity of nELAV-positive cells. To further characterize the expression of these RBPs, from tridimensional neurospheres we obtained a monolayer of stem/progenitor cells after dissociation and exposure to adhesive substrate for 1 hour. This limited period of time allows the preservation of the typical NSC heterogeneous morphologies and it is not sufficient to commit cells towards a specific fate (Bez et al., 2003; Suslov et al., 2002). As in neurosphere cultures, double fluorescent signals for nELAV and Msi1 were still present although their cellular distribution appeared slightly different, Msi1 signals being prevalent in the nuclei and nELAV RBP signals more diffuse in the cytoplasm (Fig. 2D-F). Coexpression of the intermediate filament nestin, widely used as a neural stem/progenitor cell marker (Wiese et al., 2004), with both nELAV and Msi1 demonstrated the still undifferentiated and proliferating state of our neurosphere-derived cell culture (data not shown). The specificity of the anti-Msi1 antibody used was previously verified in committed progenitor cells with neuronal morphology, which resulted positive for the differentiation marker β-tubulinIII and negative for Msi1, and in immature astrocytes, positive for both GFAP and Msi1 (Sakakibara et al., 1996) (data not shown).

Furthermore, the uncommitted mitotic state of the neurosphere-derived cells was attested by the positive labeling for the proliferation-associated antigen Ki67, a protein produced in all active phases of the cell cycle, but absent in G₀ (Kee et al., 2002). nELAV RBPs and Ki67 were co-expressed in actively proliferating cells with a distinct cellular distribution, nuclear for Ki67 and mainly cytoplasmic for nELAV proteins (Fig. 2G-I). nELAV protein staining was also present in non-proliferating Ki67-negative cells.

Since our immunostaining data clearly indicate that nELAV proteins are not limited to early post-mitotic and mature neurons, we wanted to investigate nELAV expression in vivo in the SVZ region, one of the germinal areas where neurogenesis occurs in the adult. Brain sections from adult rats were immunolabeled with anti-nELAV and anti-Ki67 antibodies (Fig. 3A,B): nELAV proteins were clearly expressed in the SVZ and overlapped the Ki67 fluorescent signals with a complementary intracellular distribution pattern (Fig. 3C,D). As neuronal-specific markers, nELAV proteins were also strongly expressed in the cerebral areas surrounding the lateral ventricle, where Ki67 labeling was absent. Immunostaining with the anti-Msi1 antibody confirmed the presence of stem/progenitor cells in the SVZ. The concomitant expression of Msi1 and nELAV proteins in this neurogenic region (Fig. 3E-G), mainly confined to the sub-ependymal cell layer (Fig. 3H), suggests the importance of both these RBPS in regulating NSC proliferation.

We have found that nELAV proteins are endowed with an RNA-binding activity for the Msi1 transcript and are expressed in neural stem/progenitor cells in vitro and in vivo. A ribonomic approach (Tenenbaum et al., 2002) was then used to isolate nELAV-containing mRNP complexes from adult mouse neurospheres. Endogenous mRNPs were immunoprecipitated from neurosphere lysates with anti-nELAV, with an irrelevant isotype-matched antibody (anti-His₅) and with no antibody, respectively, in three independent experiments. Precipitated mRNAs from each sample were subjected to quantitative real-time RT-PCR using specific sets of primers for Msi1, Rpl10a, Gap43 and Mapt (microtubule-associated protein tau) genes. The housekeeping Rpl10a, encoding a ribosomal large subunit protein, was used as negative control because it does not contain an ARE consensus sequence in its 3′UTR. Gap43 and Mapt are two well-known mRNA targets of nELAV RBPs both in developing and mature neurons. In the nELAV-containing mRNP complexes, Msi1 mRNA was enriched 62-fold compared with the control sample (anti-His₅) (Fig. 4), whereas Rpl10a was absent in all the experimental conditions tested, although its content in the input lysate was 100-fold more abundant than Msi1 (see Table S1 in supplementary material). Neither Gap43 nor Mapt transcripts were significantly immunoprecipitated from neurospheres by the anti-nELAV antibody, even if these two genes were expressed in the input lysate and were recovered from mRNPs of mouse mature brain (data not shown).

We also immunoprecipitated the endogenous mRNP complexes containing the Msi1 protein in order to test the presence of the Msi1 transcript itself among its targets. No significant results were found in line with the negative data obtained in UV crosslinking assays and confirming the observation that the Msi1 3′UTR sequence does not show any consensus binding site (G/AU_nAGU) for Msi1 (Imai et al., 2001).

We then tested whether nELAV protein binding to Msi1 mRNA was dependent on the presence of the putative ARE consensus sequence. Two deletion transcripts, ARE⁺ and ARE^-, were obtained from the mouse Msi1 3′UTR sequence analysis. UV-crosslinking assays on neurosphere lysates revealed that the ARE⁺ probe showed the same RBP-binding pattern as the full-length Msi1 3′UTR, whereas all RNA-protein interactions were abolished when the ARE^- fragment was used (Fig. 5B, left panel). This finding suggests that the presence of RBPs or associated proteins in the Msi1-containing mRNP complexes is strictly dependent on the cis-acting regulatory sequence of the ARE region. We have demonstrated that nELAV proteins exhibit an RNA-binding activity for the Msi1 transcript both in in vitro and in vivo immunoprecipitation assays. In order to further characterize this binding affinity, we expressed a recombinant His-tagged HuD protein and performed UV-crosslinking experiments. RNA-protein complexes formed with the Msi1 full-length 3′UTR and ARE⁺, but not with the ARE^- riboprobe, confirming a strictly ARE-dependent binding activity of the HuD protein (Fig. 5B, right panel). To demonstrate the specificity of such in vitro binding, competition assays were performed using a 100× molar excess of ARE⁺ and ARE^- riboprobes together with Msi1 full-length 3′UTR (Fig. 5C, left and right panels). Although ARE^- had no ability to bind to HuD, cold ARE⁺ and Msi1 3′UTR efficiently competed with the radiolabeled probe. The HuD-specific binding to Msi1 ARE sequence was also confirmed by the absence of interaction with an irrelevant sequence (As). The binding was dose dependent, as shown by UV-crosslinking assays performed with increasing amounts of recombinant HuD protein (Fig. 5D).

ARE sequence elements are usually present in mRNAs with a high turnover rate (Chen and Shyu, 1995), on which they determine increased or decreased stability, probably depending on the bound RBPs (Lal et al., 2004). We wanted to investigate whether the presence of the HuD-bound ARE in the Msi1 gene was associated to a cis-acting regulatory role. To this end, the stability profile of the Msi1 mRNA was studied by means of a functional approach which allowed us to reproduce and follow the mRNA deadenylation/degradation kinetic profile in vitro (Ford et al., 1999). For this assay short capped and polyadenylated (polyA₆₀) transcripts were synthesized to represent the putative functional region of the Msi1 3′UTR. We produced a 114 bp ³²P-labeled fragment containing the ARE consensus site (μmsi1ARE⁺) and an irrelevant 148 bp transcript as a control (μmsi1AS, the antisense sequence of the ARE^- fragment). The deadenylation kinetic and the associated degradation of the body of these two synthetic mRNAs, composed only of putative regulatory sequences, were followed by simply visualizing their decrease in length on a denaturing gel. We observed that the deadenylation of the μmsi1ARE⁺ transcript was progressive and became complete within 30 minutes compared with the fully deadenylated probe (A₀), used as a reference marker (Fig. 6A). By contrast, high-molecular-weight intermediates of the μmsi1AS probe (i.e. smear length and consistence) were still present after a longer incubation time (45 minutes), indicating a slower degradation kinetic rate.

When the same deadenylation/degradation assay was repeated in the presence of the recombinant HuD protein, the decay rate of the μmsi1ARE⁺ transcript markedly decreased compared with the control, and resembled that of the μmsi1AS mRNA in the degradation profile (Fig. 6B). Indeed, after a 45-minute incubation, the shortening of the polyA₆₀ tail was still incomplete compared with the same sample incubated in the absence of HuD, with the persistence of high-molecular-weight species corresponding to the almost intact transcript. When the μmsi1AS mRNA was used, the addition of the recombinant HuD protein did not alter the deadenylation pattern (data not shown). These results demonstrate that the conserved AU-rich region in the Msi1 3′UTR acts as a cis-acting regulatory sequence on which the nELAV HuD protein specifically exerts its stabilizing activity by decreasing its mRNA turnover rate.

We have recently shown that the nELAV proteins undergo translocation from the cytosolic to the cytoskeletal compartment after treatment with phorbol esters (PMA) in human neuroblastoma SH-SY5Y cells (Pascale et al., 2005). Such translocation is accompanied by nELAV RBP threonine phosphorylation as a consequence of PKCα isozyme activation. A PKCα-dependent stabilization of the Gap43 mRNA and an increase in its protein levels were also specifically observed in the cytoskeleton. We investigated the effect of nELAV protein activation on Msi1 content in the same cell model after treatment with PMA or solvent alone (DMSO). Msi1 protein levels were examined on cellular fractions by western blotting (Fig. 7A). PKCα translocation to the cytoskeletal compartment was clearly evident (+63%, P<0.05), together with nELAV protein increase (+98%, P<0.001), reproducing the nELAV RBP activation process already described. We also found that the Msi1 protein content significantly increased in the cytoskeleton (+107%, P<0.001) after nELAV RBP activation by phorbol ester treatment.

In this paper we reveal the existence of a post-transcriptional regulatory mechanism involving Msi1 and HuD, two neural-specific RNA-binding proteins implicated in the control of the proliferation/differentiation switch during neuronal development. Msi1 overexpression has been associated with the proliferation state and degree of malignancy of brain tumors such as gliomas (Kanemura et al., 2001; Toda et al., 2001). As a consequence, we expect that Msi1 expression has to be properly regulated both at transcriptional and translational levels during the proliferation phases of neural development. We focused on the presence of a putative ARE consensus motif in Msi1 3′UTR gene. The high degree of conservation of the entire 3′UTR sequence in only the mammalian Msi1 orthologs strongly suggests a putative role in a phylogenetically recent post-transcriptional regulatory mechanism. In our UV-crosslinking experiments, we found several RBPs showing a binding activity for the mouse Msi1 3′UTR sequence in a strictly ARE-dependent manner, in a pattern which partially overlaps the one previously described for the Gap43 mRNA, a well-characterized target of the ELAV protein family (Chung et al., 1997; Quattrone et al., 2001). By this analogy, the common 65 kDa and 95 kDa bands could be hypothesized to correspond to the RNA-binding activities of PTB (pyrimidine-tract binding) and FBP (far-upstream-element binding) proteins, respectively (Irwin et al., 1997; Kohn et al., 1996), whereas here we demonstrate that the common 42 kDa complex does indeed represent the HuD protein. These different RBPs could play opposite roles on mRNA fate, because they were shown to bind in a competitive manner to the same cis-acting regulatory element in the Gap43 3′UTR sequence (Irwin et al., 1997). Moreover, it is described that the balance resulting from distinct RNA-binding activities on different regulatory elements of the same transcript may contribute to finely and gradually regulate the passage from proliferation to differentiation programs (Briata et al., 2003; Yano et al., 2005). More generally, a coordinated expression of neural genes could be obtained by the association of their mRNAs in common mRNP complexes (Keene, 2001), composed of RBPs which are temporally specified.

From our work we have an indirect evidence that nELAV activity could be developmentally regulated, because by selectively immunoprecipitating nELAV-containing mRNP complexes from adult neurospheres we were unable to show any binding to two well-known mRNA targets, Gap43 and Mapt, which nonetheless are expressed in proliferating neurospheres (Esdar et al., 1999). Therefore, nELAV binding to ARE-bearing mRNAs could depend on various factors like nELAV activation and phosphorylation state (Pascale et al., 2005), and on the interaction or competition with other RBPs in a cell-state and cell-compartment specific manner.

Immunoprecipitation of neurosphere mRNPs and in vitro binding experiments with the recombinant HuD protein clearly demonstrated that nELAV proteins are able to recognize the Msi1 transcript in an ARE-specific and ARE-dependent way. We looked for a functional meaning of this association by the approach described by Ford and Wilusz (Ford and Wilusz, 1999), a highly reproducible system that mimics the whole mRNA catabolism in vitro. Results obtained by this deadenylation and degradation assay showed that the basal decay pattern of a synthetic mRNA bearing the Msi1 ARE sequence was faster than in its absence, and that addition of the recombinant HuD protein stabilized this transcript, as similarly described for HuB and TNF-α (Ford et al., 1999). Therefore, the HuD activity on the Msi1 transcript suggests a positive effect on its translation, as already described for other nELAV target genes (Antic et al., 1999; Mazan-Mamczarz et al., 2003). Our functional data on neuroblastoma cells showed that phorbol-ester-induced stimulation of diacylglycerol-dependent PKC isozymes determines an increase of PKCα and nELAV protein content in the cytoskeleton. This effect, previously described to be associated with PKCα and nELAV protein colocalization and with an increase in the threonine phosphorylation state of nELAV RBPs (Pascale et al., 2005), probably reflects an induction of nELAV activity at the polysomal level (A.Q., unpublished results). We showed that these phenomena were also associated with an increase in the Msi1 protein content, which was evident especially in the cytoskeletal compartment, as already shown for the nELAV protein target GAP-43.

Immunolabeling studies in cultured NSCs and in the SVZ of adult rat brains clearly showed that the nELAV RBPs are expressed and colocalize with the Msi1 protein in neural stem/progenitor cells, thus extending their biological role in NSC fate determination. This observation is in agreement with the recent description of the HuD-null mouse phenotype, where a perturbation of the proliferation and differentiation processes in neurospheres was observed (Akamatsu et al., 2005). An overall distinct and complementary expression pattern for Msi1 and nELAV RBPs has been previously reported during rodent brain development (Kaneko et al., 2000; Sakakibara et al., 1996; Sakakibara and Okano, 1997), which correlates with the observation of opposite effects of thyroid hormone on the transcription of the Msi1 and HuD genes (Cuadrado et al., 2002; Cuadrado et al., 2003). However, analyzing Msi1 and nELAV protein expression in neurogenic regions, they were described to colocalize in vivo at the border of the mouse embryonic subventricular and intermediate zones (Sakakibara et al., 1996), and in differentiation-committed areas including the anterior horn of the spinal cord. Furthermore, their coexpression was also observed in vitro in MAP2-positive cells from embryonic neuroepithelial cultures (Kaneko et al., 2000).

Neurogenesis is a multi-step cell process that gradually leads a self-renewing undifferentiated NSC to acquire a completely differentiated phenotype. During embryogenesis this is achieved through symmetric division of the initial NSC, which increases the body mass as a first step. Progenitors, slightly differentiated and committed cells, originate from NSCs by sequential asymmetric division and, although for a limited number of cell cycles, they maintain the proliferative capacity. Therefore the cell programs of proliferation and differentiation are both active in progenitors, and need to be finely co-regulated. A molecular pathway known to be essential for the maintenance of actively proliferating NSCs is the Notch/Delta cascade, which controls cell division and neural differentiation (Hitoshi et al., 2002). The Msi1 protein contributes to the Notch-mediated proliferation of NSCs by binding and preventing the translation of Numb (Imai et al., 2001), which is known to negatively affect Notch activation (Cayouette and Raff, 2002; Zhong et al., 1996). On the other hand, the nELAV proteins have been widely demonstrated to be necessary and sufficient to induce neuronal differentiation in vitro. The exact timing and cascade of events that cause an NSC to completely exit from the cell cycle is hard to define, but the process is accompanied by the inactivation of cyclin-dependent kinases by specific inhibitors, such as p21^WAF-1. Interestingly, its mRNA was shown to be regulated positively by HuD (Joseph et al., 1998) which competes with hnRNP K for the same 3′UTR motif (Yano et al., 2005), and negatively by Msi1 whose binding to an adjacent site blocks p21^WAF-1 translation (Battelli et al., 2006). According to our results, we propose that the HuD-mediated stabilization of Msi1 mRNA may serve to prolong Msi1 activity in the proliferating neural progenitor cell which is going to exit the cell cycle (Fig. 8). This would allow the stem/progenitor cell to continue to divide even after Msi1 transcriptional inactivation.

The post-transcriptional regulative mechanism of nELAV proteins on Msi1 mRNA seems to be evolutionarily restricted to mammals from embryonic to adult neurogenesis. Regenerative events in the SVZ have already been demonstrated to occur in vivo in the adult CNS after an ischemic insult or a damage (Douen et al., 2004; Yagita et al., 2002), whereas in the hippocampus, neurogenesis has been strictly linked to the processes of learning, neuronal plasticity and memory formation (Kempermann et al., 2004; Schinder and Gage, 2004). Upregulation and cytoskeletal translocation of nELAV proteins and the subsequent positive effects on Gap43 mRNA levels have been recently described in tight association with synaptic plasticity and learning processes in rat dentate gyrus (Bolognani et al., 2004; Pascale et al., 2004; Quattrone et al., 2001). We speculate that our findings of a positive nELAV modulation of Msi1 transcript and protein level could be required for the transition from proliferation to differentiation of the stem/progenitor cells residing in the hippocampal subgranular zone, as well as in the SVZ. In adult neurogenic areas, nELAV activity could be assumed to be transiently induced by signals present only in restricted spatial and/or temporal conditions. We have recently demonstrated that nELAV proteins are activated by a PKCα-dependent pathway in human neuroblastoma cells (Pascale et al., 2005), and it is known that PKC-dependent signal transduction is at the basis of many aspects of CNS development (Metzger and Kapfhammer, 2003). The combination of these observations with the present results could open the possibility of a pharmacological modulation of NSC dynamics.

In conclusion, our results suggest that in mammals nELAV proteins may have a key biological role in positively regulating Msi1 gene expression along a molecular cascade leading a proliferating stem cell towards a multi-step neural differentiation process. Since the intrinsic features of NSCs make them attractive candidates for cell-based repair therapy of the CNS (Cova et al., 2004; Lakshmipathy and Verfaillie, 2005), the nELAV/Msi1 pathway could also represent a new pharmacological target for enhancement of neurogenesis in the treatment of neurodegenerative disorders.

Neural stem/progenitor cell cultures were obtained from CD1 mouse brains as previously published (Gritti et al., 1999). After isolation of adult SVZ, tissue was papain-digested and mechanically dissociated. Single cells were plated with a density of 3.5×10³ cells/cm² in NS-A basal serum-free medium (Euroclone), supplemented with 20 ng/ml epidermal growth factor (EGF) and 10 ng/ml basic fibroblast growth factor (bFGF, PeproTech) and optimized for stem-cell growth. In 7-10 days neurospheres were obtained, and every 5-7 days they were dissociated to single cells for 3-5 passages in order to amplify the neural stem compartment. For immunocytochemical assays, they were plated, as neurospheres or dissociated stem/precursor cells, on an adhesive substrate (Matrigel™, Becton Dickinson) for 1 hour.

Human neuroblastoma SH-SY5Y cells were grown in MEM (Eagle's minimal essential medium) with 10% fetal calf serum, penicillin/streptomycin, non-essential amino acids and 1 mM sodium pyruvate (Invitrogen). Cells were exposed to 100 nM phorbol 12-myristate-13-acetate (PMA, Sigma) or to the solvent alone (DMSO) for 15 minutes and then the incubation was stopped with ice-cold PBS.

Total mouse brain and NSC cultures were homogenized in lysis buffer (150 mM NaCl, 20 mM Tris-HCl, 1% Triton X-100, protease inhibitor cocktail; Roche) and 400 U/ml RNase inhibitor (Promega). Samples were centrifuged at 12,000 g for 20 minutes at 4°C and supernatant was collected. Proteins from different cell fractions were obtained as previously published with slight modifications (Pascale et al., 1996). Briefly, SH-SY5Y cells were homogenized in buffer A (20 mM Tris-HCl pH 7.4, 2 mM EDTA, 0.5 mM EGTA, 50 mM mercaptoethanol, 0.32 mM sucrose and protease inhibitor cocktail) and centrifuged at 100,000 g for 1 hour. The supernatant containing the cytosolic fraction was collected. The pellet was resuspended in the same buffer containing 1% Triton X-100, sonicated, and incubated for 45 minutes at 4°C, then centrifuged again at 100,000 g for 1 hour. The supernatant containing the membrane fraction and the pellet with the cytoskeletal component were collected. Proteins were resolved by SDS-PAGE, transferred to nitrocellulose membranes and immunoblotted with Msi1 (1:200, Chemicon), PKCα (1:1000, Transduction Laboratories), α-tubulin (1:5000, Santa Cruz Biotechnology) antibodies, and Hu-positive serum (1:1000).

Total RNA was extracted from whole mouse brain with TriZol® reagent (Invitrogen) and used to amplify Msi1 and Gap43 3′UTR sequences by RT-PCR with the following primers: Msi1_fw TGAGGACCAGACTGAGCCAGCAAG and Msi1_rev GGGGCCTCAGTCTGCAGCAG; GAP-43_fw ATGCCTGAACTTTAAGAAATGGCT and GAP-43_rev ATGAGGAAACAAAATGGTTTTTG. The products were cloned into pGem®-T Easy Vector (Promega). ARE⁺ and ARE^- fragments were obtained by PCR amplification of the above Msi1-pGem® construct with the following primers: Msi1_fw and ARE⁺_rev GTAGGGCAACTGGCTAATC; ARE^-_fw GATTAGCCAGTTGCCCTAC and Msi1_rev. The μmsi1ARE⁺ insert was subcloned from the Msi1-pGem® construct (primers: CTCATGTCTGGCTCCCCTACT and GTAGGGCAACTGGCTAATC) into the pCR®II-TOPO® (Invitrogen) vector and selected in order to have the T7 promoter at the 5′ end and the HindIII restriction site at the 3′ end of the cloning site for the mRNA decay assay. The cDNA fragment encoding the open-reading frame for HuD (GenBank accession number D31953) was amplified by RT-PCR using the following modified primers: HuD_fw TGATCTCATGAAGCCTCAGGTGTCAAATGGAC and HuD_rev CTGCATCCCGGGGGATTTGTGGGCTTTGTTGGTT. The product was digested with RcaI/SmaI and directionally cloned into the expression vector pIVEX2.3d (Roche) with a His tag at the C-terminus. All the clones and their orientation were validated by sequencing.

Radiolabeled riboprobes were obtained by transcribing 0.5 μg linearized construct DNA with 20 U T7 RNA polymerase (Roche), 20 μCi [α-³²P]UTP, 0.5 mM NTPs, 20 U RNase inhibitor (Promega) for 30 minutes at 37°C. The reaction was stopped at 65°C and template DNA was removed by DNaseI digestion (20 U, Roche). The resulting ³²P-labeled riboprobe was purified on ProbeQuant G-50 microcolumns (Amersham Biosciences).

Recombinant HuD protein was obtained in a cell-free system in which 0.5 μg HuD-pIVEX2.3d plasmid was incubated at 30°C for 6 hours in the reaction solution, containing E. coli lysate and amino acids according to the manufacturer's instructions (Roche). The fusion protein was purified on Ni-NTA spin kit columns and visualized by western blotting using an Anti-His₅ antibody (Qiagen).

300,000 cpm of ³²P-labeled RNA transcripts were incubated with 40 μg of protein extract or with 300 ng of recombinant HuD protein in 15 μl ligation buffer (1.3 mM MgCl₂, 19 mM HEPES-KOH pH 7.4, 1.5 mM ATP, 19 mM creatine phosphate) for 10 minutes at 30°C. After addition of 5 μg tRNA, samples were irradiated with UV (Stratalinker®, Stratagene) for 5 minutes on ice and RNaseA treated (25 U) for 30 minutes. Samples were run on a 12% SDS-PAGE, and analyzed by autoradiography. For competition experiments, a 100× molar excess of cold riboprobe was added to the sample before UV irradiation. Immunoprecipitation was conducted on UV crosslinked samples by the addition of 4 μg of the selected antibody for 2 hours at 4°C. Samples were then incubated with 30 μl Protein A/G Sepharose beads (Amersham Biosciences) for 2 hours at 4°C. Immunocomplexes were then collected by centrifugation at 14,000 g for 30 seconds, washed several times in lysis buffer, run on a 12% SDS-PAGE and analyzed by autoradiography.

Cells were fixed with 4% paraformaldehyde in PBS (pH 7.4) for 20 minutes, blocked with 10% normal goat serum (NGS) and permeabilized with 0.3% Triton X-100 (Gritti et al., 1999). 20-μm-thick coronal brain sections from adult rats intracardially perfused with 4% paraformaldehyde were mounted on glasses pre-coated with poly-L-lysine. Sections were boiled for 15 minutes in 50 mM Tris-HCl (pH 8.0), permeabilized with 0.5% Triton X-100 and blocked with 10% NGS. Samples were exposed to the selected antibodies overnight at 4°C (antibodies and dilutions used are specified in Table S2 in supplementary material). Slides were mounted with Fluorsave™ (Calbiochem) and acquired with a camera connected to a DMIRE2/HCS microscope (Leica Microsystems). For negative controls, the primary antibody was replaced with NGS.

About 5×10⁶ NSCs were harvested for each condition, washed several times with cold PBS and resuspended in 1:1 (v/v) RNP buffer (100 mM KCl, 5 mM MgCl₂, 10 mM HEPES pH 7.4, 0.5% NP-40) (Tenenbaum et al., 2002). 50 μl protein A/G Sepharose beads, pre-coated with 8 μg of the selected antibody, were added to 300 μg NSC lysate in 1 ml NT2 buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM MgCl₂, 0.05% NP-40), containing 400 U RNase inhibitor, 1 M DTT and 20 mM EDTA. 10% of the reaction mix was collected as the initial input and RNA was extracted by TriZol® reagent (Invitrogen). After a 2-hour incubation, the immunoprecipitated mRNPs were washed several times with cold NT2 buffer, incubated with 30 μg of proteinase K for 30 minutes and phenol-chloroform extracted. After DNaseI digestion, the isolated mRNAs were retro-transcribed using SuperScriptII RT (Invitrogen), oligo dT and random primers.

Real-time PCR was performed for 45 cycles with SYBRGreen PCR Master mix (Applied Biosystems) and processed on the ABI Prism 7900HT sequence detection system. Oligonucleotide pairs for each gene were designed with Primer Express 2.0 software (Applied Biosystems) on exon boundaries. Reactions were conducted in triplicate for each sample and a dissociation curve was produced at the end. For the mRNP assays, the threshold cycle (C_t) values of immunoprecipitated samples were normalized to the C_t value of the corresponding input (ΔC_t) to account for differences in initial mRNA quantities. ΔΔC_t was calculated by subtracting the ΔC_t value of the sample with no antibody to the ΔC_t of the sample with anti-nELAV or anti-His₅ antibodies and fold differences were then expressed as 2^-ΔΔCt (Chakrabarti et al., 2002). Statistical analysis was conducted by one-way ANOVA with Bonferroni t-test correction. For primer sequences and real-time PCR data see Tables S1 and S3 in supplementary material.

The mRNA decay experiments were conducted as described (Ford and Wilusz, 1999) with some modifications. Whole-brain cytoplasmic extracts were used to reproduce a neural-like environment in vitro. The addition of the polyA₆₀ tail to the μmsi1ARE⁺ or μmsi1AS constructs was carried out by the PCR-ligation procedure. Briefly, the double-stranded linker AGCTT(A)₆₀TATTTACCTCGAGCACTC was ligated to the HindIII-digested plasmids, which were then amplified with the T7 promoter primer and the complementary linker primer (GAGTGCTCGAGGTAAATAT). Products were digested with SspI and in-vitro transcribed in the presence of 7mGpppG (Roche) and [α-³²P]UTP. Riboprobes were run on a denaturing 5% polyacrylamide gel (19:1) containing 8 M urea. After a short autoradiographic exposure, bands were excized from the gel, eluted overnight in HSCB buffer (400 mM NaCl, 25 mM Tris-HCl pH 7.6, 0.1% SDS), phenol-chloroform extracted and ethanol precipitated. 200,000 cpm riboprobes were incubated with 200 μg of mouse brain extract and 0.5 μg polyA (Amersham Biosciences) in 15 μl deadenylation buffer (2.6% polyvinyl alcohol, 1 mM ATP, 13 mM creatine phosphate) at 30°C. At the indicated time intervals reactions were stopped with 400 μl HSCB buffer. RNA was extracted in phenol-chloroform, separated on an urea-denatured 5% polyacrylamide gel and visualized by autoradiography.

We thank C. Fenoglio, A. Grifa and M. G. Savino for their kind help. This work was supported by Italian Ministry of Health (Mal. Neurodegenerative, ex art.56 Anno 2003), Fondazione I.Monzino and Fondazione FiorGen.