Characterization of a neural-specific splicing form of the human neuregulin 3 gene involved in oligodendrocyte survival Free

Neuregulin 1 (NRG1) gene products form a family of neurally expressed proteins that perform a wide range of functions in the developing nervous system, such as controlling the number of acetylcholine receptors at the neuromuscular junction, promoting Schwann cell proliferation and survival, inducing differentiation of oligodendrocytes and regulating the tangential migration of cortical interneurons (Falls, 2003; Flames et al., 2004). NRG1 has been identified recently as a potential susceptibility gene for schizophrenia (Corfas et al., 2004). NRG1 gene products are also expressed in non-neural tissues, where they are involved in heart and breast development (Falls, 2003; Zhao et al., 1999). The NRG1 gene subfamily is broad, as evidenced by the 15 isoforms described so far. This molecular diversity is further increased by the presence of three additional genes encoding for neuregulins in mammals, namely NRG2, NRG3 and NRG4 (Falls, 2003). Whereas NRG4 has not been found to be expressed in neural tissues, NRG2 and NRG3 are abundantly expressed in the mammalian central nervous system (CNS) (Falls, 2003; Longart et al., 2004). NRG3 mRNA is expressed in the mouse brain from embryonic day (E) 13 to adult (Zhang et al., 1997; Longart et al., 2004). NRG3 is also expressed in non-neural tissues, where it is involved in mammary gland development and carcinogenesis (Dunn et al., 2004; Howard et al., 2005).

A previously described isoform encoded by the NRG3 gene contains a characteristic, extracellular epidermal growth factor (EGF)-like domain which, when purified from genetically engineered bacteria, can bind and activate erbB4, but not erbB2 or erbB3 receptors (Zhang et al., 1997), suggesting a non-overlapping role with NRG1. However, no physiological or cellular expression studies have been performed on NRG3 gene products yet.

ErbB4 homozygous knock-out mice show aberrant migration of a subpopulation of hindbrain-derived cranial neural crest cells, which - along with heart trabeculae defects - cause death during mid-embryogenesis (Tidcombe et al., 2003). Furthermore, the erbB4 pathway has been proposed to participate in several aspects of neural cell biology, such as cortical interneuron migration (Flames et al., 2004), synaptic plasticity (Huang et al., 2000), neuronal cell death (Goldshmit et al., 2001), long-term potentiation (Eilam et el., 1998; Huang et al., 2000; Ma et al., 2003) and oligodendrocyte biology (Calaora et al., 2001; Schmucker et al., 2003; Sussman et al., 2005). All these findings correlate with the nervous-system-specific expression pattern of NRG3 and suggest that this gene could be involved in brain development and/or function.

NRG1 is a highly spliced gene with at least 15 splicing forms described so far (Falls, 2003). Alternative splicing in genomes is a major mechanism for generating proteome diversity (Nurtdinov et al., 2003). Recent analysis has estimated that 60% of human and 41% of mouse genes are alternatively spliced. Like the NRG1 subfamily, it seems likely that the NRG3 gene suffers post-transcriptional processing, giving rise to a diverse family of splicing forms. Here, we have addressed this issue and used a computer-based search of expressed sequence tag (EST) databases to identify new isoforms of the NRG3 gene. We report the recognition of a novel NRG3 splicing variant - human foetal brain neuregulin 3 (hFBNRG3) - isolated from a human foetal brain cDNA library. Using reverse transcriptase (RT)-PCR we show that this transcript is specifically expressed in the human foetal CNS, but not in other tissues. The hFBNRG3 product is a 56 kDa protein that differs at the N-terminus from human neuregulin 3 (hNRG3), a previously described isoform of NRG3 (Zhang et al., 1997). The protein is glycosylated and its lifetime is strongly regulated by the ubiquitin/proteasome pathway. EGF-like activity from hFBNRG3 can be found in conditioned media of transiently transfected mammalian cells, suggesting that the protein is shed into the extracellular space. Notably, when the recombinant protein is added to oligodendrocyte cultures, it promotes oligodendrocyte survival by activating the phosphoinositide 3 (PI 3)-kinase pathway. Taken together, our results are consistent with a model in which a highly regulated NRG3 gene isoform is expressed in the developing CNS, where it is involved in oligodendrocyte survival.

We used a computer-based approach to search EST databases for sequences identical to the EGF-like motif of NRG3. The inquiry identified a full-length cDNA corresponding to a putative new isoform of the NRG3 gene. The clone was named hFBNRG3 (human foetal brain NRG3) because it was found in a human foetal brain library (clone CS0DF004YG18, GenBank accession numbers AL534571 and AL565572) at the full-length cDNA collection of RESGEN (Invitrogen Corporation). After nucleotide sequencing of the insert and BLAST-searching the human genome, we found that this transcript is an alternative splice product of the NRG3 gene. Unlike hNRG3, a previously described splicing form, exon 1 is spliced out of the hFBNRG3 transcript and two additional exons (2 and 3) are included (Fig. 1A). The absence of exon 1 in the hFBNRG3 transcript suggests that a different promoter is used to produce this mRNA. The isolated clone contained an open reading frame (ORF) encoding a protein with a predicted molecular mass of 56 kDa. The encoded polypeptide is identical to hNRG3 from the EGF-like domain to the C-terminus, but differs greatly in both size [54 versus 275 amino acids (aa)] and sequence (8.4% aa identity) at the N-terminal domain (Fig. 1B). Hydropathy analysis indicates the presence of two hydrophobic segments that are located in the N-terminal region (aa V10-F29 and V140-F162). Segment V140-F162 is highly hydrophobic and long enough to span the width of the plasma membrane. Thus, it probably is the unique membrane-anchor domain of the protein, akin to some isoforms of the NRG1 gene (Cabedo et al., 2002). The predicted extracellular segment of the protein contains several putative O- and N-glycosylation sites. However, in contrast to most of the NRG1 splicing forms, hFBNRG3 does not contain IgG-like or kringle domains, a property that is shared with the previously reported hNRG3 isoform (Zhang et al., 1997). The large C-terminal domain (Y163-K499) is probably cytoplasmic, and accommodates a coiled-coil region (L433-G460) and a dileucine motif (L433-L434), suggesting that it is involved in DNA binding, protein-protein interactions or protein sorting and trafficking.

To study the expression pattern of the hFBNRG3 mRNA, RT-PCR from several human embryonic tissues was performed. Using specific primers to amplify the complete ORF of hFBNRG3 mRNA we amplified the cDNA from human foetal brain, kidney, testis, spleen, muscle and liver supplied in the Multiple-Choice First-strand cDNA (OriGene Technologies, see Materials and Methods). As shown in Fig. 2, a band of the expected molecular size was amplified from the human foetal brain cDNA, but not from other tissues. The housekeeping β-actin mRNA was used as control (Fig. 2, bottom). To verify the identity of the amplified band, the PCR product was cloned using TA technology in the pCRII dual-promoter vector and sequenced using T7 and SP6 primers; their sequences corresponded to the 5′- and 3′-end of the mRNA encoding the hFBNRG3 protein, respectively. These results, therefore, imply that the hFBNRG3 transcript is expressed in human foetal brain.

We next investigated the expression of hFBNRG3 protein in cultured mammalian cells. The ORF of hFBNRG3 was cloned in a mammalian expression vector with a myc-epitope at the C-terminus of the protein. COS-7 cells were transfected with hFBNRG3-myc and the expression of the protein was studied by immunocytochemistry. As shown in Fig. 3A, expression of hFBNRG3-myc protein could be detected in transfected cells and was increased by proteasome inhibition with MG-132 (see below). By contrast, an inversely oriented (inv) hFBNRG3 construct [(inv)-hFBNRG3] did not express any protein (data not shown). The polypeptide segment encompassing V140-F162 of hFBNRG3 is highly hydrophobic and long enough to span the width of the plasma membrane, thus it probably is a membrane-anchor domain for the protein. To check this possibility, we used different approaches. First, using the anti-myc antibody, we performed immunocytochemical studies on hFBNRG3-transfected COS-7 cells. Confocal images suggested a membrane localization of the protein (Fig. 3A). This notion was supported by confocal microscopy of cells transfected with construct (CFP)-hFBNRG3 [cyan fluorescent protein (CFP) coupled to hFBNRG3]. CFP-hFBNRG3 displayed a cortical distribution of the fluorescence (Fig. 3C) similar to CFP-SMDF (Fig. 3D), a well documented plasma membrane protein (Cabedo et al., 2002). By contrast, the cytoplasmic protein CFP showed a homogenous distribution pattern (Fig. 3E). Neuregulins are expressed in neural tissues. To gain insight of the subcellular location of hFBNRG3 in a neuron-like cell, immunocytochemistry was performed in transfected phaeochomacytoma 12 (PC12) cells. hFBNRG3 was detected in the plasma membrane of PC12 cells (Fig. 3B) suggesting that, in neuron-like cells too, the protein is located at the plasma membrane.

These findings were substantiated further by biochemical methods. Enriched plasma membrane fractions from hFBNRG3-expressing COS-7 cells were western blotted with anti-myc antibody. As shown in Fig. 3F, most of the immunoreactivity was found in the pellets, not in the soluble fraction, as is the case for the membrane protein SMDF. These results demonstrate a plasma membrane location of the hFBNRG3 polypeptide.

Finally, to know whether the protein is exposed on the cell surface, we performed biotinylation experiments on hFBNRG3-expressing cells. When biotin is added to cells in culture it binds only to proteins exposed to the cell surface. As depicted in Fig. 3G, anti-myc immunoreactivity was clearly found in the streptavidin pellets of the biotinylated cells, supporting the notion of a cell-surface-exposed protein. Taken together, our results demonstrate that hFBNRG3 is a membrane protein that is expressed and accessible on the cell surface.

During our experiments we noticed that the proteasome inhibitor MG-132 increased the number of hFBNRG3 immunoreactive cells. Therefore, we asked whether the cellular stability of this polypeptide is regulated by proteasomal degradation. Transiently transfected COS-7 cells were incubated with the proteasome inhibitors MG-132 or Z-Leu-Leu-Phe-CHO or the highly specific inhibitor Lactacystin for 18 hours. Thereafter, cells were lysed, and cell extracts subjected to SDS-PAGE, transferred to nitrocellulose membranes and blotted with polyclonal anti-myc antibody. Under our conditions, the hFBNRG3-myc polypeptide could not be immunodetected in western blots unless proteasomes were inhibited. Incubation of the cells with the proteasome inhibitors promoted the expression of two immunoreactive bands of approximately 50 and 70 kDa (Fig. 4A). Notice that, the fast-migrating band displays a slightly lower molecular mass than that predicted from the sequence, suggesting postranslational processing of the protein. Likewise, the 70 kDa immunoreactive band suggests that the polypeptide is post-translationally modified. As expected, hFBNRG3-type bands were not observed in cell extracts from cells transfected with the inversely oriented cDNA (Fig. 4A). Similar results were obtained with the neuron-like cell line PC12 (Fig. 4A), substantiating the notion that the protein stability is cell type independent. To gain further insights into protein expression, we performed a time-course study of protein stability after proteasome inhibition in COS-7 cells. As displayed in Fig. 4B, hFBNRG3 can be detected after 4 hours of proteasome inhibition. Similar results were obtained with Lactacystin (data not shown). Taken together, our results suggest that hFBNRG3 is degraded, at least in part, by the proteasome, although the participation of other cellular proteases or lysosomes can not be completely ruled out.

The proteasome is known to degrade ubiquitylated proteins. Ubiquitin is an 8 kDa peptide that is attached to the amide group of lysines in polypeptide chains that are destined for degradation. Multi-ubiquitylated chains of at least four ubiquitin molecules are necessary for efficient recognition and degradation of ubiquitylated proteins by the proteasome (Pickart, 2004). Hence, we next investigated the role of this post-translational modification in hFBNRG3 stability. For these studies, we used the cell line ts20, that harbours a mutant E1 (ubiquitin ligase), active only at the permissive temperature of 30°C, and inactive at the nonpermissive temperature of 42°C (Kulka et al., 1988). To investigate the role of ubiquitylation in hFBNRG3 biogenesis, we transfected hFBNRG3-myc and (inv)-hFBNRG3-myc into ts20 cells, and monitored the expression of the protein at permissive and nonpermissive temperatures. As illustrated in Fig. 4C, anti-myc immunoreactivity was not found in cell extracts when the cells were incubated at the permissive temperature for 18 hours. By contrast, the nonpermissive temperature induces the appearance of the 50 and 70 kDa immunoreactive bands. Notice that, the ∼65 kDa band is unspecific, given that it was also found in the (inv)-hFBNRG3-myc transfected cells incubated at 42°C. These results suggest that hFBNRG3 can be degraded by the proteasome.

Unexpectedly, the hFBNRG3 protein appears on immunoblots as two different forms of approximately 50 and 70 kDa. The fast-migrating band is consistent with proteolytic post-translational processing of the protein, presumably by truncation of a signal peptide. To reveal the molecular identity of the polypeptide with the approximate size of 50 kDa, an HA epitope was fused to the N-terminal domain of the hFBNRG3-myc construct and used as a reporter. Unlike an HA-tagged control protein, expression of the HA-hFBNRG3-myc protein was not detected on western immunoblots from cell extracts (Fig. 6A). The lack of HA-hFBNRG3-myc expression was not due to a defect in protein biogenesis because immunoblotting with an anti-myc antibody showed that the HA-tagged neuregulin was expressed at a similar level to hFBNRG3-myc (Fig. 6B). Notice that the expression of both proteins was only detected when the proteasome was inhibited. These findings indicate that the N-terminus of the protein is proteolytically processed, and suggest that the hydrophobic segment V10-F29 is a signal-sequence that is cleaved and degraded during the insertion of the protein into the plasma membrane.

To determine the molecular determinants of protein instability, we deleted the first 52 aa in the N-terminus of hFBNRG3 [hFBNRG3-myc(Δ3-55)] (Fig. 5). In contrast to hFBNRG3-myc, the mutated neuregulin appeared in western immunoblots as two distinct bands of approximately 50 and 65 kDa even in the absence of proteasome inhibitors (Fig. 6C), indicating that the protein lacking the N-terminus is highly stable in cells. Similar results were obtained with deletion mutants of the CFP-hFBNRG3 protein. As illustrated in Fig. 6D, deletion mutants that do not have the N-terminal domain were readily detected. These results suggest that the N-terminus of the protein conveys instability to the protein. Indeed, fusion of the N-terminal domain of hFBNRG3 to the type I membrane protein synaptotagmin I [(▿1-92)^hFBNRG3-synaptotagmin I-myc], significantly decreased the stability of synaptotagmin (Fig. 6C). Taken together, our data indicate that hFBNRG3 bears a cleavable signal-sequence and a downregulation signal at its N-terminal domain.

We next investigated the molecular identity of the 70 kDa polypeptide detected on western blots. Because the protein contains several glycosylation sites, we first focused on the question whether the protein is glycosylated in cells and, if so, what type of sugar is attached. Post-translational modification with sugar derivatives has been shown to affect the aggregation state and subcellular localization of other neuregulins (Cabedo et al., 2004). To address this issue, we took advantage of the sugar-specificity of two agarose-conjugated lectins. The Triticum vulgaris lectin wheat germ agglutinin (WGA) binds to β-GlcNac-modified proteins, whereas soybean agglutinin (SBA) from Glycine max interacts with both, α- and β-Gal-Nac-glycosylated proteins. Whole-cell extracts from hFBNRG3-myc-transfected COS-7 cells were incubated with WGA- or SBA-agarose and the presence of the protein in the precipitates was evaluated by western blots using the anti-myc antibody. As depicted in Fig. 7A, a 70 kDa polypeptide was precipitated by the SBA resin but not by WGA. Notice that galactose eluted the slow-migrating band (lane 6), ruling out the possibility of nonspecific binding to the resin. Thus, this result demonstrates that the 70 kDa polypeptide is an α- or β-Gal-Nac-glycosylated form of hFBNRG3.

The EGF-like domain of neuregulins is released into the extracellular medium to activate erbB receptors expressed in target cells. To test this, conditioned medium from cells transiently transfected with hFBNRG3 was concentrated and assayed for erbB4 autophosphorylation in erbB4-transfected COS-7 cells. To validate our assay, a dose-response curve for the activation of the erbB4 receptor was performed with increasing amounts of recombinant glutathione-S-transferase (GST)-NRG3 purified from bacteria (Fig. 7B). Quantification of the blot showed that detection of as low as 3 nM of the recombinant protein was possible. As illustrated in Fig. 7C, conditioned medium from MG-132-treated, hFBNRG3-expressing cells (lane 6) induced a noticeable increase in erbB4 tyrosine phosphorylation. The increase in erbB4 activation was not due to higher load of protein, as demonstrated by reprobing the same blot with an anti-PKCα antibody (bottom). Conditioned medium from hFBNRG3-transfected cells that had not been treated with MG-132 did not induce statistically significant erbB4 phosphorylation (lane 4). Therefore, our results suggest that, hFBNRG3 can be released into the extracellular medium where it can activate erbB4 receptors.

Neuregulin 3 is expressed during mammalian CNS development (Longart et al., 2004). It has been shown that erbB3 activation by neuregulin 1 has an anti-apoptotic effect on Schwann cells (Dong et al., 1995). Recombinant neuregulin 1 has also been shown to act as a survival factor for oligodendrocyte precursors (Flores et al., 2000). Oligodendrocytes express erbB2, erbB3 and erbB4 receptors (Schmucker et al., 2003; Park et al., 2001). Since erbB3 is dispensable for oligodendrocyte development both in vitro and in vivo (Schmucker et al., 2003), and erbB2 does not bind neuregulin ligands (Burden and Yarden, 1997), we reasoned that NRG3 gene products - through erbB4 signalling - might also be involved in oligodendrocyte survival. To address this issue, we studied the effect of recombinant NRG3 on the survival of serum-deprived oligodendrocyte-precursor-enriched primary cultures (Lubetzki et al., 1991). After 7 days in vitro, cultures were serum-starved and GST or GST-NRG3 was added at 150 nM. Forty-eight hours later, cells were immunostained with the anti-NG2 antibody as an oligodendrocyte-precursor-marker and counter-stained with Hoechst dye (Fig. 8A). Apoptotic nuclei of NG2-positive cells were counted for both treatments. Images of five microscopic fields per slide from five different slides per treatment (∼750-1500 NG2-positive cells) were evaluated in each experiment. The results of three independent experiments are shown in Fig. 8B. Serum-starved cultures showed 8.8±0.5% of apoptotic oligodendrocyte precursors. Recombinant neuregulin 3 decreased the number of apoptotic oligodendrocytes to 4.5±0.9%, close to the level of apoptosis in nonstarved cultures (3.5±0.7%).

Akt is a protein kinase that promotes cell survival by inhibiting apoptosis through its ability to inactivate caspase 9 and the transcription factor Bad (Franke et al., 1997). It has been shown previously that neuregulin signalling can activate the PI 3-kinase pathway (Flores et al., 2000) leading to Akt activation by phosphorylation of Ser473. To test whether NRG3 activates the PI 3-kinase pathway in glial cells we studied the effect of GST-NRG3 on Akt phosphorylation in serum-deprived C6 glioma cells. As shown in Fig. 8C, GST-NRG3 promoted the phosphorylation of Akt in a dose-dependent way. The time course of Akt phosphorylation in NRG3-treated serum-deprived oligodendrocyte-precursor cultures was investigated to determine whether the anti-apoptotic effect of NRG3 in oligodendrocyte precursors can also be mediated by Akt activation. Notably, GST-NRG3 induced the phosphorylation of Akt after 5 minutes of incubation (Fig. 8D). These results suggest that NRG3 acts as a survival factor for oligodendrocyte precursors by activating the PI 3-kinase pathway.

To test this hypothesis further, we checked whether inhibition of PI 3-kinase blocks the cell-survival-inducing activity of NRG3. Oligodendrocyte-precursor cultures were serum-starved and treated with 150 nM of GST or GST-NRG3, with or without 20 μM of the PI 3-kinase inhibitor LY294002. The result of three independent experiments is depicted in Fig. 8E. In this series of experiments GST-treated serum-starved cultures contained 13.7±1.5% of apoptotic oligodendrocyte precursors; LY 294002 did not induce a significant change in the number of apoptotic cells (13.7±3.3%). However, LY294002 reduced the number of NG2-positive cells in our cultures suggesting that, as in other cell systems (Jeong et al., 2005), PI 3-kinase inhibition can induce cell-cycle arrest in oligodendrocyte precursors. Recombinant NRG3 decreased the number of apoptotic oligodendrocytes to 6±0.6% and this effect was completely reversed by co-incubation with LY 294002 (17.7±2%). Taken together, our results show that the anti-apoptotic effect of NRG3 is mediated by activation of the PI 3-kinase pathway.

NRG1 gene products regulate numerous aspects of neuronal biology, such as interneuron migration, and fate, differentiation and proliferation of glial cells, as well as synaptic plasticity and the expression of diverse neurotransmitter receptors (Falls, 2003; Rieff et al., 1999). NRG2 and NRG3 mRNAs have also been found to be expressed in the mammalian CNS (Longart et al., 2004). However, so far, their biological role is undetermined. Expression of the NRG3 gene is restricted to the CNS (Zhang et al., 1997), suggesting that this gene has a key role in neural biology. Here, we used a computer-based approach to identify new NRG3 transcripts. We report a new transcript of NRG3 (denoted hFBNRG3), whose expression is restricted to the human embryonic brain.

Alignment with the human genome suggests that this transcript is produced by alternative splicing and different promoter usage. Alternative splicing (observed in approximately 50% of mammalian genes) is a crucial step in the regulation of many vertebrate gene products and is thought to have a pivotal role in the generation of proteomic diversity (Nurtdinov et al., 2003; Pan et al., 2004; Lee et al., 2003). Three main types of alternative-splicing cassettes have been described. Conserved alternative splicing is the less common process in mammalian genomes (approximately 15% of the spliced genes). Species-specific alternative splicing of conserved exons, by contrast, is much more frequent and has been estimated to affect approximately 35% of the spliced genes in human and mouse genomes. Finally, the most frequent event is alternative splicing that is genome-specific. This type of splicing involves genes, which contain exons that are only present in the genome of a single species. More than 50% of the spliced genes in humans and mice are spliced in this genome-specific alternative-splicing process (Ahn et al., 2005; Pan et al., 2004; Modrek and Lee, 2003). Interestingly, multiple new transcription initiation sites have been described for the human NRG1 gene (Steinthorsdottir et al., 2004). Some of the NRG1 isoforms produced this way are rare brain transcripts that contain internal exons that are not conserved between the human and the mouse genomes. Much less is known about the splicing of the NRG3 gene. Recently, evidence was presented that multiple NRG3 isoforms are expressed in the mammary gland of mice where they might be involved in mammary-bud induction (Howard et al., 2005). The hFBNRG3 transcript appears to be specific for humans because ESTs with homology to the 5′ region of this mRNA were not found in EST databases from other organisms, including rats and mice. This is consistent with the observation that this sequence has been found in the human genome but not in that of mouse or rat. Interestingly, we found the sequence of exon 2, but not that of exon 3, in the draft of the chimpanzee genome that has recently been published (Chimpanzee Sequencing and Analysis Consortium, 2005). Collectively, our data support the existence of a genome-specific splice isoform of NRG3 that includes newly created exons in the primate genome.

Confocal microscopy and biochemical studies support a plasma membrane localization of the protein. Like other neuregulins, hFBNRG3 is probably a type I transmembrane protein with an extracellular N-terminus and a cytosolic C-terminus. The N-terminal domain contains an EGF-like region and a hydrophobic sequence (V10-Y29). Fusion of an HA-epitope to the N-terminus of the protein reveals that this segment is cleaved, rendering a shorter mature polypeptide. In addition, the extracellular domain contains sites for N- and O-glycosylation. Our data indicate that the protein binds to the N-acetyl-galactosamine-specific SBA lectin, implying that the protein is glycosylated. Furthermore, activity of the protein can be detected in the extracellular milieu, as demonstrated by the erbB4-activating activity of conditioned medium from hFBNRG3-expressing cells. Taken together, these experiments suggest that, like other neuregulins, hFBNRG3 is exported to the cell surface and released into the extracellular medium.

Heterologous expression of the hFBNRG3 clone in mammalian cells can be enhanced by incubation with Lactacystin, a bacterial metabolite with a high specificity for proteasome inhibition (Fenteany et al., 1995). Proteosomal degradation of the hFBNRG3 protein seems to be mediated by ubiquitylation, as suggested by the increase in protein stability and expression following inhibition of ubiquitylation in ts20 cells. However, we could not directly detect ubiquitylation of the hFBNRG3 protein (data not shown), suggesting that its degradation is driven by an interaction with other proteins (Diamonti et al., 2002; Qiu and Goldberg, 2002). In support of this notion, removal of the N-terminal domain spanning residues C3-F55, completely abrogates protein degradation, suggesting the existence of a protein-protein interacting domain in this region.

Accumulating evidence shows that the ubiquitin/proteasome system is involved in numerous aspects of neural cell biology and pathology. In general, the degradation of native proteins has a high energy cost, not only because the elimination of resources, but also, because suppression of natively folded proteins is thermodynamically unfavourable. However, targeted degradation of signalling molecules might provide a mechanism to enable rapid responses to a changing environment. For instance, it has been proposed that regulation of DCC/Fra (the receptor for axonal attractant netrin-1) depends on its proteasome-mediated degradation (Bashaw and Goodman, 1999; Johnston and Madura, 2004). Furthermore, synaptic plasticity is controlled by ubiquitin/proteasome degradation (DiAntonio and Hicke, 2004). Studies on long-term potentiation (LTP) have illustrated that PKA, a kinase pivotal in learning, is activated in the absence of cAMP because the regulatory subunits of the enzyme complex are degraded by the proteasome (Johnston and Madura, 2004). These data provide compelling evidence for a central role of the ubiquitin/proteasome pathway in nervous system biology. Thus, regulation of hFBNRG3 stability by proteasomal degradation might be a mechanism used to control the activity of NRG3.

Myelination of the mammalian CNS is a highly regulated process in which ensheathing oligodendrocytes are matched to the number and length of neuronal axons. Oligodendrocytes are postmitotic cells, which develop from oligodendrocyte progenitors that migrate into the developing white matter from their germinal zones in the neuroepithelium. Their development is regulated by cell-intrinsic- as well as cell-extrinsic-factors (Barres and Raff, 1999; Rowitch et al., 2002). Previous work has shown that platelet-derived growth factor (PDGF) secreted by neurons is required for proliferation of oligodendrocyte progenitors during development (Calver et al., 1998; van Heyningen et al., 2001). Overexpression of PDGF-A in transgenic mice under the control of the neuron-specific enolase promoter (NSE-PDGF-A mice) caused hyper-proliferation resulting in a large increase in the number of oligodendrocyte progenitors. The extra progenitors generated excess oligodendrocytes but these all died soon after their production, so that a normal number of myelinating oligodendrocytes survived long-term (Calver et al., 1998; van Heyningen et al., 2001). This is presumably because the final number of oligodendrocytes is controlled by an unidentified survival factor generated by neurons. There is some evidence that the survival factor is a member of the neuregulin family (Richardson, 2001; Fernandez et al., 2000; Canoll et al., 1996; Vartanian et al., 1999; Kim et al., 2003). Here, we show that the neuregulin 3 gene product is able to activate pAkt, a protein kinase involved in cell survival by inactivating the transcription factor Bad and caspase 9 (Franke et al., 1997). We also report that recombinant NRG3 promotes survival of serum-starved cultured oligodendrocyte precursors and that the inhibition of the PI 3-kinase pathway with LY294002 blocks the effect of NRG3 on oligodendrocyte-precursor survival. Taken together, our results demonstrate that NRG3 gene products are survival factors for oligodendrocyte precursors in culture. Ongoing experiments in our laboratory address the question whether NRG3 is a crucial survival factor in vivo.

The clone CS0DF004YG18 (GenBank accession numbers AL534571 and AL565572) was from the full-length cDNA collection of RESGEN (Invitrogen Corporation). Primers, Dulbecco's modified Eagle's medium (DMEM), foetal bovine serum (FBS), antibiotics, Ni-NTA agarose, TA cloning kit and pcDNA3.1/myc-His-A were obtained from Invitrogen. Percoll and nitrocellulose membranes were from Amersham Biosciences. The pcDNA3-erbB4 vector was provided by the Yossef Yarden (Weizmann Institute of Science, Rehovot, Israel). Pfu turbo DNA polymerase was from Stratagene. Rabbit polyclonal anti-myc antibody, mouse monoclonal anti-phosphotyrosine antibody (clone PT-66), horseradish peroxidase (HRP)-conjugated anti-rabbit IgG and anti-mouse IgG secondary antibodies, sulfo-NHS biotin, streptavidin agarose beads, reduced glutathione (GSH)-agarose beads and bisbenzimide H33258 (Hoechst dye) were obtained from Sigma. Mouse monoclonal anti-ubiquitin antibody was from Zymed Laboratories. Mouse monoclonal anti-HA antibody (clone 12CA5) and lumi-light plus substrate were from Roche. pEYFP-C1 and the anti-green fluorescent protein (GFP) antibody (Living Colors A.v. peptide antibody) were from Clontech. Rabbit polyclonal anti-NG2 antibody was from Chemicon. LY294002 was from Biosource. Rabbit polyclonal p44/42 MAP kinase antibody and pAkt(Ser 473) antibody were from Cell Signaling Technology. Streptavidin Alexa Fluor 546 and other fluorescent conjugated secondary antibodies were from Molecular Probes. Recombinant PDGF AA (PDGFAA) was from Biovision.

hFBNRG3-encoding cDNA was amplified from CS0DF004YG18 clone by PCR using Pfu Turbo DNA polymerase with the primers hFBNRG3-NotI-sense (5′-CCAGAATGGCGGCCGCGCCAGTAAC) and hFBNRG3-NotI-antisense (5′-CTACATCTGCGGCCGCTTGGTCAATG) and cloned in frame, after mutation of the stop codon, with the myc epitope and poly-His tag. To introduce an HA-epitope at the N-terminal end of the hFBNRG3-myc construct, inverse PCR with the PAC I bearing primers HA-hFBNRG3 sense (5′-GCTAGCTTAATTAAGTATCCTTATGATGTGCCTGATTATGCGTGTGGTATACCTCCAACCCTTG) and HA-FBNRG3 antisense (5′-CGCGACCTTAATTAACTCCATAATTCTATCAACTGC) were used. PCR products were digested with DpnI and PacI, ligated and transformed. To clone the pECFP-C1 chimeras, the ORF of hFBNRG3 was amplified and cloned in frame using the primers hFBNRG3-HindIII-sense (5′-CTTTTGAAGCTTTTGATAGAATAATG) and hFBNRG3-SalI-antisense (5′-CTACATGTCGACTCACTTGGTCAATG). Other constructs were obtained using standard molecular biology techniques. All constructs were verified by automatic sequencing. To check the tissue expression pattern of hFBNRG3 in human embryos, seminested PCR was performed on human Multiple Choice First-strand cDNA (Origene). External primers, sense (5′-CATATTTTCCCAGGAGGTGT) and antisense (5′-GTGTTGACATGACATAACCCGTG), were used for the first round PCR. The same antisense primer and an internal sense primer (5′-TCTACCCTGAAGACCCCAGA) were used for the second round of PCR. Amplified products were checked in agarose horizontal gels stained with ethidium bromide, cloned into pCRII 2.1 dual-promoter vector using the Invitrogen TA cloning Kit and sequenced.

COS-7 and C6 glioma cells (from ECACC) were cultured in DMEM containing 10% FBS. Rat pheochromocyctoma (PC12) cells were maintained in DMEM supplemented with 10% heat-inactivated horse serum and 5% heat-inactivated bovine serum. Ts20 cells (kindly provided by Alice Dautry-Varsat, Institut Pasteur, Paris) were cultured in DMEM 10% FBS at 30°C unless indicated. Cells were plated on 2-cm²-wells at 250,000 cells/well. After 20-24 hours, cells were transfected with 1 μg of plasmid DNA using lipofectamine 2000 following the manufacturer's recommendations. To prepare conditioned mediums, cells were serum-starved 24 hours post-transfection for 18 hours, and the medium was collected and concentrated using Amicon Ultra-4 filters (Millipore).

Cultures were grown on poly-L-lysine-coated 12 mm glass coverslips as described previously (Lubetzki et al., 1991). Briefly, forebrains were removed from newborn mice, dissociated mechanically in 1× Hanks medium and then digested with trypsin (0.0025%) for 20 minutes at 37°C. After centrifugation and washing, the pellet was passed sequentially through two nylon meshes with the pore size of 150 μm and 63 μm. Cells were layered on a Percoll density-gradient and centrifuged at 20,000 g for 45 minutes at 4°C. Glial progenitor cells were carefully pipetted out and washed three times with 1× Hanks. They were the resuspended in DMEM supplemented with 10% FBS, and 1% penicillin/streptomycin. Cells were plated at the density of 5× 10⁵ cells per well for 45 minutes to facilitate attachment. Then, 1 ml of Bottenstein and Sato medium (Bottenstein and Sato, 1979) supplemented with 0.5% FBS, 1% antibiotics and recombinant PDGF-AA at 10 ng/ml was added to each well.

All the inhibitors used were obtained from Sigma. To study the effect of proteasome inhibition on protein stability, hFBNRG3-transfected cells were incubated with 25 μM Lactacystin, 2 μM MG-132 (Z-Leu-Leu-Leu-al) or 20 μM Z-Leu-Leu-Phe-CHO for 18 hours (unless indicated) before lysis. Nontreated cells were incubated with vehicle (DMSO) for the same period of time. After that, cells were submitted to anti-myc immunoblotting or immunocytochemistry as indicated.

After transfection, cells were mixed with μ-mercaptoethanol containing SDS sample buffer, heated and separated by SDS-PAGE. Proteins were electrotransferred onto nitrocellulose membranes, blocked with 10% fat-free milk in TBS and incubated with the primary antibody in blocking buffer for 1 hour. Membranes were washed with TBS-Tween (0.3%), incubated with a secondary HRP-conjugated antibody and developed with the ECL plus system. For lectin binding assays, transfected COS-7 cells were lysed with 10 mM Tris-HCl pH 7.5, 100 mM NaCl, 0.4% sodium deoxycholate, 0.3% SDS and 1% Nonidet P-40, and cell extracts were incubated with 80 μl of SBA- or WGA-agarose lectins and extensively washed with lysis buffer. Immunocomplexes and lectin-bound proteins were analysed by immunoblotting. To rule out nonspecific binding, lectin-bound protein was eluted with 0.3 M galactose.

Crude plasma membranes were essentially prepared as described by Cabedo et al., (Cabedo et al., 2002). Briefly, transfected cells were rinsed with PBS and lysed in hypotonic buffer A (2 mM MgCl₂; 1 mM EDTA; 1% phenylmethylsulfonyl fluoride; 20 mM Hepes pH 7,4). Cell lysates were centrifuged at 9000 g 4°C for 10 minutes to prepare low-speed pellets containing the plasma membranes and supernatants, which had the soluble proteins and the remaining membranes. Low-speed pellets were washed first in buffer A supplemented with 1M NaCl and centrifuged at 4°C, second in buffer A for desalting. Final pellet and supernatants were mixed with μ-mercaptoethanol-containing SDS sample buffer, and heated and separated by SDS-PAGE.

Transfected cells were incubed with 0,1 mg/ml of sulfo-NHS-biotin during 30 minutes at 4°C, and lysed in 10 mM Tris pH 8.8, 150 mM NaCl, 1% sodium deoxycholate, 2% SDS and 2% nonidet P-40. Biotin-conjugated cell-surface proteins were purified with streptavidin agarose, mixed with μ-mercaptoethanol-containing sample buffer and separated by SDS-PAGE.

Cells were seeded on poly-L-lysine-coated cover-slips at 50 to 200×10³ cells/well and transfected as indicated previously. At the indicated times, cells were washed with PBS and fixed with 4% paraformaldehyde, blocked with 20% FBS in DMEM (DMEM/FBS) 45 minutes and incubated with primary antibodies in DMEM/FBS 0.1% Triton X-100 at room temperature for 180 minutes. Afterwards, coverslips were washed with PBS and incubated with secondary antibodies in the same medium for 45 minutes. Washes were repeated and cells were incubated with Hoechst dye and mounted on slides. Images of the slides were obtained with a Leica confocal or fluorescence microscope.

After 7 days of culture, oligodendrocyte progenitors were serum- and growth-factor-starved and incubated with 150 nM of GST- or GST-NRG3-purified proteins for 48 hours. Then, cells were immunostained with an anti-NG2 antibody, a specific oligodendrocyte-precursor marker, and counter-stained with the nuclear marker Hoechst dye. Numbers of NG2-positive cells with pycnotic nuclei were counted in control and treated cultures. For Akt phosphorylation assays, oligodendrocyte precursor or C6 glioma cells were serum-deprived and cell extracts submitted to anti-pAkt specific antibody immunoblot.

cDNA encoding the EGF-like domain of NRG3 was cloned in pGEX-4T-1 and transformed in DH5α bacterial cells, which were grown at 37°C until reaching 0.6-0.8 OD. Thereafter, cells were induced with isopropyl-beta-D-thiogalactopyranoside (IPTG) at 0.2 mM for 4 hours and pelleted. The pellet was resuspended in PBS and disrupted by sonication. Triton X-100 was added to reach 1% and the cell extract centrifuged at 3000 g for 15 minutes. Recombinant GST-NRG3-EGF-like polypeptide was purified from the supernatant with reduced glutathione (GSH)-agarose beads. After extensive washing, fusion-protein was eluted from the beads with 10 mM GSH in Tris-HCl pH 8. GST was purified using the same procedure. The protein concentration was determined using the method of Bradford (Bradford, 1976).

Neuregulin-induced tyrosine phosphorylation of erbB receptors was carried out by a modified protocol of Ho et al. (Ho et al., 1995). To ensure comparable levels of the receptor per well, pcDNA3-erbB4 transiently transfected COS-7 cells were trypsinised, reseeded and grown to μ80% confluence in 24-well plates. Under these conditions the amount of erbB4 per total protein was equal. Thereafter, cells were serum-starved for 18 hours and incubated with concentrated serum-free conditioned medium or recombinant NRG3-EGF-like domain for 5 minutes at 37°C as indicated. Medium was removed and cells were harvested with 100 μl of μ-mercaptoethanol-containing SDS sample buffer. Whole-cell extracts were heat-denatured and separated by SDS-PAGE. Nitrocellulose membranes were blocked with 10% fat-free dried milk in TBS and incubated with the monoclonal anti-phosphotyrosine antibody in blocking buffer for 2-4 hours. Membranes were washed with TBS-Tween (0.3%), incubated with the anti-mouse IgG-HRP-conjugated antibody and developed with the ECLplus system. To verify identical protein load in each lane the membranes were reblotted with anti-PKCα.

We thank W. D. Richardson, at University College London, for insightful comments on the manuscript. We are grateful to C. Lubetzki for advice in oligodendrocyte culture techniques. We are also grateful to Yossef Yarden, Alice Dautry-Varsat, and Cruz Morenilla-Palao for providing the pcDNA-erbB4 clone, ts20 cells and synaptotagmin I plasmid, respectively. We like to thank Consuelo Martínez-Moratalla for technical assistance. C.C. holds a predoctoral FPU fellowship from the Spanish Ministry of Education and Science. This work was supported by grants from `Instituto de Salud Carlos III' (FIS 01/3081) and `Fundación Principe de Asturias' (Beca Severo Ochoa 2002) to H.C., and from MEC (SAF 2003-0509) to A.F.M.