Essential role for the Prader–Willi syndrome protein necdin in axonal outgrowth

Abstract

Necdin and Magel2 are related proteins inactivated in Prader–Willi syndrome (PWS), a sporadic chromosomal deletion disorder. We demonstrate that necdin and Magel2 bind to and prevent proteasomal degradation of Fez1, a fasciculation and elongation protein implicated in axonal outgrowth and kinesin-mediated transport, and also bind to the Bardet–Biedl syndrome (BBS) protein BBS4 in co-transfected cells. The interactions among necdin, Magel2, Fez1 and BBS4 occur at or near centrosomes. Centrosomal or pericentriolar dysfunction has previously been implicated in BBS and may also be important in the features of PWS that overlap with BBS, such as learning disabilities, hypogonadism and obesity. Morphological abnormalities in axonal outgrowth and fasciculation manifest in several regions of the nervous system in necdin null mouse embryos, including axons of sympathetic, retinal ganglion cell, serotonergic and catecholaminergic neurons. These data demonstrate that necdin mediates intracellular processes essential for neurite outgrowth and that loss of necdin impinges on axonal outgrowth. We further suggest that loss of necdin contributes to the neurological phenotype of PWS, and raise the possibility that co-deletion of necdin and the related protein Magel2 may explain the lack of single gene mutations in PWS.

INTRODUCTION

NDN (encoding necdin) and MAGEL2 are two of four protein-coding genes inactivated in individuals with Prader–Willi syndrome (PWS), a sporadic chromosome deletion disorder marked by profound neonatal hypotonia, global developmental delay, hypoventilation, childhood onset hyperphagia, obesity and hypogonadism ( 1 ). Hypothalamic dysfunction underlies many aspects of this disorder and, at least one of the inactivated PWS genes, is predicted to be critical for the normal development and function of the hypothalamus ( 2 ). Decreased fetal movements, decreased myelination, hypoplasia of cortical commissures and enlarged lateral ventricles are found in fetuses and young children with PWS ( 3 , 4 ). The partial resolution of both neonatal hypotonia and failure to thrive suggested to us that a delay in the maturation of neuronal circuits might underlie the early course of PWS. In adults, additional abnormalities include chronically elevated levels of the orexigenic hormone ghrelin, elevated levels of serotonin metabolites and, to a lesser extent, dopamine metabolites in the cerebrospinal fluid ( 2 , 5 ) and peripheral insensitivity to pain. Defects in parasympathetic innervation of the gut and/or failure of serotonergic neuronal transmission could contribute to the voracious appetite of individuals with PWS.

Simultaneous deletion of all PWS-equivalent genes in mice is usually lethal in the first postnatal week because of hypoventilation and failure to thrive ( 6 , 7 ). Necdin and MAGEL2 are part of a multiprotein family related by a MAGE homology domain. This family also includes MAGED1 (NRAGE), which interacts with the p75 neurotrophin receptor and facilitates nerve growth factor (NGF)-mediated apoptosis through a Jun kinase-dependent pathway ( 8 , 9 ). Expression of murine Magel2 is highest in neurons of the developing hypothalamus, particularly the suprachiasmatic nucleus and supraoptic tract. Magel2 null mice have not been described. Murine Ndn is expressed in many but not all postdifferentiation stage neurons ( 10 , 11 ), and also in developing muscle, skin and cartilage ( 11 ). A subset of necdin null mice exhibit a defect in respiratory rhythm generation in the medulla, causing hypoventilation with high neonatal mortality ( 12 – 15 ). A role for necdin in neuronal terminal differentiation is supported by experiments showing that necdin-transfected PC12 cells display increased differentiation and accelerated neurite extension in response to nerve growth factor ( 16 ), that ectopic necdin expression induces neurite outgrowth in neuroblastoma cells ( 17 ) and that repression of necdin in embryonic dorsal root ganglion cells suppresses their differentiation ( 18 ).

We now present a novel role for necdin and Magel2 in neuronal function. We identified an interaction between both necdin and Magel2 and fasciculation and elongation (Fez) proteins implicated in centrosome-mediated cytoskeletal rearrangement after neuronal differentiation and in axonal outgrowth. We identified a second interaction of necdin and Magel2 with BBS4, another protein implicated in centrosome function. BBS4 is one of several genes mutated in Bardet–Biedl syndrome (BBS), a complex disorder in which affected individuals display learning disabilities, retinopathy and obesity, together with hypogonadism, cardiac, limb and kidney malformations ( 19 ). We show that necdin null embryos have defects in cortical commissural formation and axonal extension, bundling and pathfinding. These results suggest that necdin is required to facilitate the intracellular processes that underlie neurite and axonal outgrowth in embryonic neurons, leading us to propose that loss of necdin impairs these processes in necdin null mice and in PWS. Furthermore, if the functions of necdin and Magel2 are partially redundant in key neurons, their combined loss in PWS may abrogate this shared function. We postulate that PWS is one of an emerging class of neurodevelopmental disorders that includes BBS, schizophrenia and lissencephaly, which are in part caused by defects in centrosome function in cytoskeletal rearrangement during neurite extension.

RESULTS

Necdin and Magel2 interact with fasciculation and elongation proteins Fez1 and Fez2

We performed a screen for cytoplasmic proteins interacting with necdin using the yeast two-hybrid Ras rescue system ( 16 ). We identified a necdin-interacting protein corresponding to amino acids 46–282 of the 353 amino acid Fez2 protein. Fez2 (UniGene Cluster Hs.258563), which is widely expressed, and Fez1 (UniGene Cluster Hs.79226), which is primarily expressed in the brain, are mammalian homologues of Caenorhabditis elegans and Drosophila melanogaster unc-76 ( 20 – 26 ). UNC-76 protein is found in the cell body and axonal cytoplasm of C. elegans neurons during development, and is essential for axon bundling ( C. elegans ) and transport mediated by the microtubule motor protein kinesin ( D. melanogaster ). The unc-76 mutation can be partially rescued by ectopic expression of human Fez1 , suggesting retention of function between species ( 21 ).

We confirmed the interaction of Fez1 and Fez2 (together called Fez1/2) with necdin by co-immunoprecipitation of transiently transfected epitope-tagged full-length proteins. Lysates of HEK293 cells containing Xpress-tagged necdin (XNdn) and HA-tagged Fez1 or lysates containing Xpress-tagged necdin (XNdn) and HA-tagged Fez2 (HAFez1/2) were immunoprecipitated with an anti-Xpress monoclonal antibody, and then HAFez1 or HAFez2 were detected on anti-HA immunoblots (Fig.  1 A). We then co-transfected XNdn with an irrelevant protein, HA-tagged neurofilament 3 (HANef3). In anti-Xpress immunoprecipitates, HANef3 was not co-immunoprecipitated, indicating that XNdn is not interacting non-specifically with the HA-tagged proteins (Fig.  1 A). In anti-HA immunoprecipitates of co-transfected HEK293 cells, XNdn was detected on an anti-Xpress immunoblot, verifying the interaction of necdin with both Fez1 and Fez2 (Fig.  1 B). Similarly, we detected Fez1–Magel2 interaction by co-transfection of HEK293 cells with Xpress-tagged Magel2 (XMagel2) and HAFez1 followed by immunoprecipitation with anti-Xpress (Fig.  1 C).

We noted that co-transfection with HAFez1 and either XNdn or XMagel2 increased the amount of HAFez1 detected in the lysates, keeping the total amount of plasmid DNA in the transfection constant with pXpress insertless vector (Fig.  1 D). We performed similar transfections varying the amounts of the pXNdn but maintaining constant amounts of pHAFez1. Increasing the amount of XNdn while maintaining a stable amount of HAFez1 increased the amount of HAFez1 (Fig.  1 E). After immunoprecipitation with an anti-Xpress antibody, the amount of HAFez1 co-immunoprecipitated is proportional to the amount of HAFez1 transfected, suggesting that Fez1 interacts stoichiometrically with necdin. We found similar results for XNdn co-transfected with HAFez2, and with XMagel2 co-transfected with either HAFez1 or HAFez2 (e.g. Fig.  1 F). Co-transfection of an Xpress-tagged irrelevant protein (XpressFoxC1) had no effect on the levels of HAFez1 detected (data not shown).

Necdin and Magel2 may increase Fez1 levels by increasing Fez1 transcription or stability. We first determined by RT–PCR that equivalent amounts of Fez1 RNA are present in cells transfected with pHAFez1 with or without pXNdn (data not shown). Co-transfected cells were then treated with each of three different proteasome inhibitors. In the presence of proteasome inhibitors, HAFez1 is ∼3-fold more abundant than with DMSO alone (Fig.  1 G), and is slightly more abundant than when co-expressed with XNdn (Fig.  1 D). Similarly, co-transfection of pXNdn with pHAFez2 increases the amount of HAFez2, and proteasome inhibitors stabilize HAFez2 in the absence of necdin co-transfection (Fig.  1 G). This suggests that necdin stabilizes Fez1/2 by preventing their degradation by the proteasomal pathway.

Co-localization of necdin and Fez1 near centrosomes

Endogenous Fez1 is present in SK-N-SH neuroblastoma cells in a punctate fashion in the cytosol, along organized filamentous structures ( 23 ). In cultured rat hippocampal neurons, Fez1 also colocalizes with F-actin in growth cones ( 23 ). Endogenous necdin is highly concentrated in the cytoplasm of differentiated neurons, and moves to the nucleus under specific conditions ( 27 ). To visualize the intracellular location of the interaction between necdin and Fez1/Fez2, HEK293 cells were transiently transfected with combinations of HAFez1/HAFez2 and XNdn, then stained for the epitope tags after 18 h. Cells containing XNdn stained in both the nucleus and the cytoplasm, with a distinctive juxtanuclear body noted in many cells. Cells containing HAFez1 had punctate staining in the cytoplasm, in a juxtanuclear body and no staining in the nucleus. Staining of co-expressed HAFez1 and XNdn revealed an overlap in the juxtanuclear body (Fig.  2 A). To ascertain whether the HAFez1- and XNdn-positive juxtanuclear body is centrosome-associated, we used an antibody directed against endogenous γ-tubulin, a component of centrosomes and other microtubule-associated structures. In HEK293 cells co-transfected with pHAFez1 and pXNdn, HAFez1 immunoreactivity was observed surrounding and partially overlapping the γ-tubulin immunoreactive signals at the centrosomes, apparent as two bright juxtanuclear dots (Fig.  2 B).

Interaction of necdin with BBS4

We next investigated whether necdin or Magel2 interact with other proteins located in centrosomes. One of the proteins implicated in the genetic disorder BBS, BBS4, localizes to basal bodies of ciliated cells and to centrosomes. BBS4 is proposed to act in intracellular microtubule-associated transport but not in the formation of the cilia themselves ( 19 , 28 ). Given the overlapping phenotypes of PWS and BBS and the common centrosomal localization of BBS4 and the necdin/Magel2-interacting protein Fez1, we explored the interactions among necdin, Magel2, Fez1 and BBS4. We co-transfected HEK293 cells with pMycBBS4 and/or pHAFez1. As expected, staining for both epitopes revealed an overlap in centrosomes, with HAFez1 also present in the cytoplasm (Fig.  2 C). On co-transfection of pHABBS4 and pXpressNdn, both proteins were present in centrosomes, although transfected necdin is also widely distributed in the cell (Fig.  2 D). Transfected Magel2 is present in a punctate pattern in the cytoplasm, and is also detected in an overlapping juxtanuclear location with transfected HABBS4 (Fig.  2 E).

We then examined whether Fez1 could interact directly with the centrosome component γ-tubulin. Lysates of HEK293 cells transfected with HAFez1 and/or XNdn were immunoprecipitated with an anti-HA antibody. A protein of the correct size for γ-tubulin was detected with an anti-γ-tubulin antibody in the immunoprecipitates, even when no HA containing vector was transfected (Fig.  2 F). This represents the non-specific precipitate of this highly expressed protein that carries through the immunoprecipitation process. However, in the presence of HAFez1 and XNdn, increased levels of co-immunoprecipitated γ-tubulin were consistently detected (Fig.  2 F), confirming the localization of the Fez/necdin complex to γ-tubulin-containing centrosomal structures. Lysates of HEK293 cells containing either HANdn or HAFez1 together with mycBBS4 were immunoprecipitated with an anti-myc antibody. Using an anti-HA antibody, HAFez1 and HANdn were detected in their respective immunoprecipitates, indicating interactions between mycBBS4 and HANdn, and mycBBS4 and HAFez1 (Fig.  2 G). Likewise, transfected XMagel2 interacts with HABBS4 in co-immunoprecipitation experiments (data not shown).

Axonal extension and bundling of primary sympathetic neurons are impaired in cell culture

Because Fez-related proteins are essential for axonal elongation in C. elegans ( 21 ) and kinesin-dependent axonal transport in D. melanogaster ( 25 ) and because of the role of the centrosome in organizing microtubules during axonal outgrowth in postmitotic neurons ( 29 ), we investigated whether the absence of necdin caused a defect in neuronal fasciculation and elongation in the developing murine embryonic nervous system. We assayed superior cervical ganglion (SCG) neurons, which extend axons in a reproducible manner when plated in compartmented chambers ( 30 ); these neurons normally express necdin ( 11 ). In this assay, dissociated neurons are plated in the central compartment of a tissue culture dish, and their axons extend along collagen tracks and cross under grease barriers into side compartments that contain different culture medium. To observe differences in neuronal bundling, we isolated SCGs from control or necdin null embryonic day 17.5 (E17.5) mouse embryos, then dissociated and plated the neurons in compartmented culture dishes supplemented with rat serum in the center compartment and NGF in the side compartments. The necdin null ganglia were typically more difficult to dissect, and the neurons grew poorly in culture, with thinner, less bundled and more branched axons in surviving neurons (Fig.  3 A). At higher magnification, we observed necdin null axons with varicosities and atypically localized thickenings, changes in direction without branching and backtracking towards the barrier. These anomalies are rarely seen in control axons (Fig.  3 B). The necdin null growth cones are swollen rather than flattened, with ruffles as is typical of control growth cones (Fig.  3 C). Overall, the outgrowth, bundling and morphology of axons are significantly compromised in necdin null SCG neurons.

Axon projections and tracts are reduced or misrouted in necdin null mice

The compartmented culture data established the importance of necdin for axon outgrowth in vitro. We then compared the histology of axon tracts in necdin null embryonic brain sections with those of control littermates. Thionin labels neuronal Nissl substance and cell nuclei, and is excluded from axonal tracts. On sagittal sections at E18.5, the axonal bundle leading into anterior commissure is typically reduced or missing in the necdin null embryos (Fig.  4 A′). In contrast, the corpus callosum axonal bundle and the hippocampal commissure axonal bundle are variably affected in the mutant embryos. The lateral and fourth ventricles were consistently moderately enlarged in necdin null embryos, from E13.5 to birth. Abnormalities in the corpus callosum and fornix were also apparent by anti-neurofilament immunohistochemistry of transverse brain sections at E18.5 (Fig.  4 B). To visualize the optic chiasm, we traced the trajectory of retinal ganglion cell (RGC) axons in E16.5 mice by placing DiA powder in the optic cup of the right eye, then 10 days later examining coronal sections through the telencephalon. The optic chiasm is larger and less compact as visualized by thionin staining in the necdin null embryos (Figs  4 C and 5 E), and the DiA labeling shows that most of the axons reach the contralateral side, as in the control (Fig.  4 D). We then examined the fasciculus retroflexus (Fr), a long fasciculated tract that carries axons within its core from the medial habenula to the interpeduncular nucleus, and carries axons within its sheath from the lateral habenula to specific midbrain targets. In thionin-stained sagittal (Fig.  4 E) and coronal (Fig.  4 F) E18.5 control sections, the Fr is visualized as a single, non-staining bundle of axons. In contrast, the Fr is less tightly bundled in the necdin null (Fig.  4 E′) and appears as a cluster of non-staining ectopic circular axonal bundles (arrowheads in Fig.  4 F′) that surround what is normally a tight axonal bundle. This confirmed a fasciculation defect in both cultured and in vivo neurons, and led us to perform a more detailed immunohistological investigation of the necdin null brain.

On thionin-stained sections of necdin null embryos from E14.5 onwards, an ectopic axonal bundle was detected in the anterior hypothalamic region dorsolateral to the optic chiasm and rostral to the zona incerta and lateral hypothalamus (Lh) (Fig.  5 A′ and B′). At E16.5, this bundle has not increased in its extent, the internal capsule is poorly defined and ectopic axonal whorls become evident in the striatum of the necdin null embryos (Fig.  5 C′). We supposed that the ectopic bundle could be misrouted thalamocortical axons (Tca), which normally course ventrally from the thalamus, then rostrally through the internal capsule before extending dorsolaterally into the cortex ( 31 ), or alternatively could be misrouted RGC axons ( 32 ). At E16.5, the necdin null thalamic axons labeled with the L1 cell adhesion molecule do not form an internal capsule, and a reduced number of projections extend to the intermediate zone of the cortex (Fig.  5 D′, arrow). Both the ectopic bundle in the rostral lateral hypothalamus and the smaller ectopic whorls in the thalamus stain strongly with L1, which labels both thalamic efferent projections and RGC axons. RGC axons are normally fasciculated within the optic nerve, course through the optic chiasm at the base of the ventral hypothalamus and partially defasciculate as they grow over the diencephalon to reach the superior colliculus ( 32 ) (Fig.  5 E). In the necdin null E16.5 embryos, DiA-labeled RGC axons reach the optic chiasm and extend partway along the lateral border of the thalamus, but we failed to detect any axonal labeling, with DiA extending into the superior colliculus on coronal sections (Fig.  5 E′) or on transverse sections. Some of the RGC axons deviate dorsomedially into the dorsal aspect of the ectopic patch (Fig.  5 E′). In both the control and the necdin null brain, a minority of RGC axons does not cross at the optic chiasm but instead course along the ipsilateral aspect of the hypothalamus. In summary, the defects in the thalamus and hypothalamus of necdin null embryos include failures in both extension and routing of thalamocortical and retinal ganglion cell axons towards their respective targets.

Developmental timing of the axonal defect

We analyzed mid-gestation embryos to define the timing of the axonal defects, particularly the ectopic bundle in the developing hypothalamus. The hypothalamus is of interest because of its major role in endocrine function and appetite regulation, which are profoundly affected in PWS. Necdin and Fez1 are co-expressed in the developing hypothalamus by E10.5 (Fig.  6 A–D). We then investigated whether the expression of early hypothalamic markers was perturbed in the necdin null mice. The obesity-related transcription factor Sim1 and Magel2 are expressed in regions overlapping with Ndn expression in the early hypothalamic neuroepithelium ( 11 , 33 ), and Sim1 is also expressed in the zona limitans. The expression of Sim1 and Magel2 in the necdin null embryo is comparable to control in serial sections through the hypothalamus at E12.5 (Fig.  6 E–G). At E13.5, a misrouting of thalamocortical L1-positive axons is evident in the ventral part of the dorsal thalamus (Fig.  6 H′), the region later to become part of the ectopic bundle shown in Figure 5 C. Thus no evidence for defects in hypothalamic neuroepithelial differentiation was found at E12.5, but misrouting of hypothalamic axons was seen in necdin null embryos by E13.5.

Serotonergic and catecholaminergic projections are reduced in the necdin null embryo

The rostral and caudal projections of the serotonergic raphe nuclei participate in many functional systems, and are implicated in behavioral disorders including PWS ( 34 ). At E12.5, we detected a comparable number and placement of rostral serotonergic cells using RNA in situ hybridization with a probe for the serotonin transporter ( SERT ) in control and necdin null embryos (Fig.  7 A). Immunohistochemistry with an anti-serotonin antibody reveals that while these cell bodies were indeed producing serotonin, the ascending fibers containing serotonin are not detectable in the mesencephalon of the necdin null embryo (Fig.  7 B′). Furthermore, serotonin immunohistochemistry on sagittal sections at E15.5 revealed a continuing paucity of ascending serotonergic-positive fibers (Fig.  7 C′). Immunohistochemistry with anti-tyrosine hydroxylase (TH) antibody at E15.5 normally detects the noradrenergic neurons of the brainstem and their rostral and caudal projections, and the mesencephalic dopaminergic neurons and their striatal, cortical and limbic projections. In a section adjacent to a thionin-stained section showing the ectopic thalamocortical bundle (Fig.  8 A), we noted a severe disruption of the TH-positive projections in the nigrostriatal pathway and caudoputamen (Nsp) (Fig.  8 B). In a section ∼600 µm rostral to that in Figure 8 B, few TH-positive axons were detected in the necdin null section (Fig.  8 C), but a section ∼600 µm caudal to the ectopic bundle had fairly normal TH staining of the cell bodies in the dorsal medial hypothalamus (Dmh) and the Nsp (Fig.  8 D). Thus both the serotonergic and catecholaminergic projection systems are disrupted in the necdin null embryo.

DISCUSSION

We propose that deficiency of necdin causes a delay or dysfunction in axonal extension, which accounts for the consistent finding of profound hypotonia, reduced myelination, enlarged ventricles and commissural defects described in fetuses and children with PWS. A defect in parasympathetic innervation of the stomach has been suggested in PWS ( 2 ), and parallels with the axonal defects we see in necdin null sympathetic neurons. Defects in serotonin pathways and altered responses to psychoactive drugs targeting dopaminergic pathways are consistently seen in PWS, mirroring the dysfunction in the late embryonic necdin null embryo. In this model, hypoventilation and central sleep apneas prevalent in individuals with PWS are also ascribed to necdin deficiency ( 15 , 35 ).

We propose that necdin and the related, co-deleted protein Magel2 act in Fez and BBS4 centrosome-related activities that lead to cytoskeletal rearrangements during neurite outgrowth. Centrosomes are critically dependent on the function of BBS proteins (including BBS4) ( 19 ), providing a tantalizing mechanistic link between these two rare disorders that each result in impaired mental development and obesity ( 19 ). BBS4 is thought to transport the scaffold protein PCM1 to centrosomal satellites through interactions with the dynein microtubule-based molecular motor, allowing for formation of the centrosomal microtubule organizing center ( 19 ). In D. melanogaster , the interaction of the Fez orthologue UNC-76 with the molecular motor kinesin is essential for axonal transport ( 25 ), and loss of UNC-76 causes disruption of fast axonal transport and ‘axon clogs’ similar in appearance to the varicosities we describe in necdin null cultured sympathetic neurons. Necdin enhances neurite outgrowth in NGF-stimulated PC12 cells, whereas Fez1, when phosphorylated by PKC zeta, causes neurite outgrowth in PC12 cells ( 22 ). We identified a necdin- and Magel2-mediated protection of Fez proteins from proteasomal degradation, and co-localization of necdin and Fez to a juxtanuclear compartment overlapping centrosomes. Moreover, Ndn , Magel2 and Fez1 mRNAs are co-expressed in the embryonic ventral and caudal hypothalamus. Unfortunately, Magel2 null mice are not available to test whether specific hypothalamic neurons are dependent on Magel2 function for appropriate axonal extension. Together, our data support a model whereby up-regulation of necdin/Magel2 in postmitotic neurons stabilizes Fez proteins to facilitate centrosome-mediated cytoskeletal rearrangements required for axonal outgrowth and kinesin‐mediated transport.

The necdin null pathology also emulates a subset of neuronal patterning and specification mutant phenotypes that cause ectopic whorls and bundles ( 32 , 36 ). Although we have clearly shown an axonal extension defect in necdin null mice, it remains possible that some of the axonal misrouting and whorling defects are related to dysfunctional specification or patterning of the ventral telencephalon. However, the up-regulation of necdin in most postmitotic, differentiated neurons, the localization in proximal axons and cell bodies but not in distal axons in SCG cultures and the role of necdin in cytoskeletal rearrangements and the dysfunction of necdin null neurons in culture are more strongly suggestive of an intrinsic growth defect in cytoskeletal dynamics than of a direct role in patterning and/or specification of subsets of neurons during development.

Necdin and Fez1 interact in a juxta–centrosomal compartment also containing the schizophrenia candidate protein DISC1 in a complex with the cytoplasmic dynein-regulating proteins LIS1 and NUDEL ( 37 , 38 ). Although we have not fully characterized the domains of Fez1/2 interacting with necdin, in the initial yeast two-hybrid screen necdin interacted with amino acids 46–282 of the 353 amino acid Fez2 protein, while the more C-terminal domain of this protein family is implicated in binding to the KHC tail of kinesin ( 25 ) and to DISC1 ( 23 ). LIS1 mutations cause neuronal migration defects and lissencephaly in humans and mice, through a disruption of dynein and microtubular dynamics mediated in part through centrosomes ( 37 ). This complex set of interactions points to overlapping dysfunctions in centrosome–microtubule dynamics in PWS, BBS, lissencephaly and possibly other psychiatric disorders. The possibility that concurrent deletion of necdin and Magel2 in PWS may abrogate the functional redundancy of these two related proteins in specific neurons in the hypothalamus provides a plausible explanation for the lack of single gene mutations in PWS. The neurons involved in the hypothalamic circuits implicated in PWS may indeed have delayed or insufficient axonal connections. Therefore, strategies applied early in postnatal life that functionally restore some of these connections could provide a novel line of therapy for infants diagnosed with PWS.

MATERIALS AND METHODS

Mouse breeding and genotyping

The Animal Welfare Committee at the University of Alberta approved procedures for animal care. Ndn^tm2Stw necdin null mice were bred through the maternal line with C57BL/6J male mice as previously described ( 15 ). Male offspring carrying a maternally inherited Ndn^tm2Stw are phenotypically normal and were bred to C57BL/6J females to produce experimental embryos. Because of genomic imprinting that silences the maternally inherited allele, inactivation of the paternally inherited allele is sufficient to cause a complete loss of necdin expression. In these litters, half the embryos are control and half carry a paternally inherited necdin deficiency and are functionally necdin null. Genotyping was performed by LacZ PCR and histochemistry as described ( 15 ).

Interacting proteins

Full-length Fez1, Fez2 and necdin cDNAs were cloned into a pCI vector with an N-terminal in-frame HA tag (Promega Corp.) to form constructs pHAFez1, pHAFez2 and pHANdn. Full-length necdin (325 amino acids) and full length Magel2 (521 amino acids) were cloned into the pcDNA4 HisMax mammalian expression vector (Invitrogen) with an N-terminal Xpress epitope tag to form pXNdn and pXMagel2. BBS4 constructs (pmycBBS4, pHABBS4) were provided by Dr N. Katsanis ( 19 ). Appropriate insertless vectors were co-transfected to maintain constant amounts of each plasmid subtype in each co-transfection. Transfection of purified plasmid was performed by calcium phosphate precipitation. In one set of experiments, the proteasome inhibitors MG132 (20 µ M ), proteasome inhibitor I (25 µ M ) or lactacystin (10 µ M ) (Calbiochem), each diluted in DMSO, were added to the media 4 h before sample recovery. DMSO vehicle alone was added to the control samples with no inhibitor. Protein lysates were collected in ECB buffer (150 m M NaCl, 0.5% IGEPAL, 50 m M , Tris–Cl pH 8.0) 36–48 h posttransfection. Lysate supernatants were immunoprecipitated using a monoclonal anti-Xpress antibody (Invitrogen), a polyclonal anti-HA antibody (Santa Cruz) or a monoclonal anti-myc antibody (Sigma), whereas detection on immunoblots of SDS–PAGE gels used these antibodies and an anti-Xpress-HRP conjugated antibody (Invitrogen), monoclonal anti-HA antibody (HA-7 from Sigma) or anti-gamma-tubulin antibody (Sigma). Detection was by ECL (Amersham). For immunocytochemical co-localization, HEK293 cells were co-transfected on microscope coverslips with various plasmids, keeping the total plasmid molar concentrations identical across the experiment by co-transfection with insertless vector. After 18 h, the media was removed and the cells were fixed in 2% paraformaldehyde (PFA) in 1×PBS. Transfected proteins were labeled with antibodies against their respective epitope tags, then detected with Alexa 594 goat anti-mouse IgG and Alexa 488 goat anti-rabbit IgG (Molecular Probes).

Primary cultures of sympathetic neurons from necdin null mice

Superior cervical ganglions were dissected from E17.5 mouse embryos. Dissociated cells were plated into the center chamber of collagen coated compartmented culture dishes. L15 medium with penicillin/streptomycin, L -glutamine, glucose and additives ( 30 ) such as bicarbonate and methylcellulose was used as basal medium, to which were added 2.5% rat serum and 1 mg/ml vitamin C for the center compartment and 50 ng/ml NGF for the side compartments. During the first week after plating, 10 µ M cytosine arabinoside and 20 ng/ml NGF were included in the center compartment, after which their use was discontinued.

Thionin staining and immunohistochemistry

Cryosections of PFA-fixed embryos (20–30 µm thick) were stained with thionin, and adjacent sections were prepared for immunohistochemistry. Results were confirmed in control/necdin null littermate pairs from multiple litters. Antibodies were: monoclonal rat anti-L1 (Chemicon International), goat anti-5-HT (Immunostar, used at 1:1000), monoclonal anti-neurofilament (2H3, Developmental Studies Hybridoma Bank) and rabbit anti-TH (Chemicon). Secondary antibodies for immunofluorescence were Alexa 594 goat anti-rabbit IgG, Alexa488 goat anti-mouse IgG and Texas Red goat anti-rat IgG (Molecular Probes) and Cy3 donkey anti-goat IgG (Jackson Immunoresearch).

Retinal ganglion cell axonal tracing

A powdered form of the dialkylaminostyryl neuronal tracer DiA (Molecular Probes) was placed in the optic cup of the right eye of E16.5 embryos. The embryos were placed in 4% PFA for 10 days at 37°C, cryoprotected overnight in 22% sucrose, embedded in OCT and sectioned at 60 µm in the coronal plane.

RNA in situ hybridization to mouse embryos

RNA in situ hybridization with DIG-labeled probes on 20 µm cryosections was performed as previously described ( 11 ). The serotonin transporter (SERT) cDNA probe corresponds to 88–886 bp of GenBank accession no. AF013604 and was provided by Dr Paul Gray (Harvard University). The Magel2 and Ndn probes were previously described ( 11 ), and the Sim1 probe was kindly provided by Dr J. Michaud (Hôpital Ste. Justine, Montreal). Control experiments with no probe or sense probes gave either no signal or a uniformly low background staining, as expected (data not shown).

ACKNOWLEDGEMENTS

We thank Drs Joe Casey, Alan Underhill, Robert Campenot, Andrew Waskiewicz, Jean Vance, John Greer and Silvia Pagliardini at the University of Alberta, Dr Phil Barker (Montreal Neurological Institute) and Dr Carol Schuurmans and Lin Ma (University of Calgary) for helpful discussions. We thank Dr Nicholas Katsanis at Johns Hopkins University for BBS4 cDNAs, Dr Jacques Michaud at Hôpital Sainte-Justine for the Sim1 probe, Dr Paul Gray for the SERT probe and Dr Fred Berry for the pCI-HA vector and helpful advice. This work was supported by research grants from the March of Dimes Birth Defects Foundation (research grant #6-FYOO-196) and the Canadian Institutes of Health Research (CIHR) (MOP 57678) to R.W. S.L. held a studentship from the Alberta Heritage Foundation for Medical Research (AHFMR) and A.A.T. was partially funded by a CIHR Training Program Grant in Maternal, Fetal and Newborn Health. R.W. is a senior scholar of the AHFMR.

†

Present address: Beth Israel Deaconess Medical Center, 345 Research North, 99 Brookline Avenue, Boston, MA 02215, USA.

Figure 1. Necdin and Magel2 interact with and stabilize Fez1/Fez2. ( A ) Lysates of HEK293 cells co-transfected with combinations of pXNdn and pHAFez1, pHAFez2 or pHANef3 (as a negative control), immunoblotted (IB) with α-HA (input), or immunoprecipitated (IP) with α-Xpress then immunoblotted with α-HA. ( B ) Lysates of co-transfected HEK293 cells, immunoblotted with α-Xpress (input) or immunoprecipitated with α-HA then immunoblotted with α-Xpress. ( C ) Lysates of HEK293 cells co-transfected with pXMagel2 and pHAFez1, immunoblotted (IB) with α-HA (input) or immunoprecipitated with α-Xpress then immunoblotted with α-HA. ( D ) Lysates of HEK293 cells co-transfected with combinations of pXNdn, pXMagel2 and pHAFez1 and immunoblotted with α-HA. ( E ) Lysates of HEK293 cells co-transfected with pHAFez1 (0.5 pmol) and 0–0.5 pmol of pXNdn, immunoblotted with α-HA. The ratio of pXNdn to pHAFez1 is indicated across the top. Insertless pXpress was co-transfected to keep the total pXpress-type plasmid at 0.5 pmol. ( F ) As in (E), but with XMagel2 instead of XNdn. ( G ) Lysates of HEK293 cells, transfected with pHAFez1 or pHAFez2 and treated with proteasome inhibitors, and immunoblotted with α-HA. UT, untreated; DMSO, vehicle alone; MG132; PI1, proteasome inhibitor 1; LC, lactacystin.

Figure 2. Necdin, Fez1 and BBS4 interactions. ( A ) Immunofluorescent staining of HEK293 cells co-transfected with XNdn (green) and HAFez1 (red). Co-localization (yellow) is indicated by the arrow. Nuclei are stained blue with DAPI. ( B ) Immunofluorescent staining of HEK293 cells co-transfected with XNdn and HAFez1 and then immunostained for HAFez1 (red) and endogenous γ-tubulin (green). Co-localization (yellow) is indicated by the arrow. ( C ) Immunofluorescent staining of HEK293 cells co-transfected with mycBBS4 (green) and HAFez1 (red). Co-localization (yellow) is indicated by the arrow. ( D ) Immunofluorescent staining of HEK293 cells co-transfected with HABBS4 (green) and XNdn (red). BBS4 positive centrosomes are indicated by the arrow. ( E ) Immunofluorescent staining of HEK293 cells co-transfected with HABBS4 (green) and XMagel2 (red). Co-localization (yellow) is indicated by the arrow. Nuclei are stained blue with DAPI. ( F ) Lysates of HEK293 cells co-transfected with X Ndn and/or HAFez1, immunoblotted with α-γ-tubulin (input), or immunoprecipitated with α-HA then immunoblotted with α-γ-tubulin. ( G ) Lysates of HEK293 cells co-transfected with combinations of pHANdn, pHAFez1 and pmycBBS4, immunoblotted with α-HA (input) or immunoprecipitated with α-myc then immunoblotted with α-HA. HAFez1 (*), HANdn (>).

Figure 3. Axons from necdin null sympathetic neurons elongate more slowly and are less bundled when grown in compartmented cultures. ( A ) Overlapping images of a typical set of axons from E17.5 sympathetic neurons grown in compartmented culture, 6 days after axotomy. The necdin null sympathetic neurons ( A ′) have axons that are less bundled and have more branches than the control. The barrier separating the neuronal soma and proximal axons from the distal axons is at the top. ( B ) Magnified images of cultured neurons reveal anomalies in the necdin null axons, including varicosities (circled in B ′) and changes in direction without branching (arrow in B ′) ( C ) On further magnification, growth cones in the control are flattened (arrows in C) whereas necdin null growth cones are swollen and/or stumpy (arrows in C ′).

Figure 4. Morphological defects in the axonal tracts of necdin null mice. ( A ) Thionin-stained sagittal sections at E18.5. Rostral is to the left. ( A ′) Loss of the axonal bundle leading into the anterior commissure (Ac, >) with relatively normal corpus callossum axonal bundle (Cc). Hc, hippocampal commissural bundle. ( B ) Transverse sections labeled with an antibody to neurofilament (red), at E18.5. Rostral is towards the top. Cc. corpus callosum; F, fornix. ( C ) Thionin-stained sagittal sections at E18.5. The optic tract (Ot, >) is less compact in the necdin null ( C ′). ( D ) Coronal section at E16.5. DiA placed in the optic cup of the right eye (left on the image) labels projections through the optic chiasm (Oc) and is visualized as red fluorescence. A white line marks the midline of the third ventricle. Abnormal RGC axonal bundling is noted at the optic chiasm in D ′. Hoechst counterstain is blue. ( E ) The fasciculus retroflexus (Fr) (>) is less well defined in the necdin null embryo in a thionin-stained sagittal section ( E ′) at E18.5. ( F ) In a cross-section through the Fr, a tight non-staining axonal bundle is seen in the control (F) but multiple smaller axon bundles (>) surround a central bundle in the necdin null ( F ′). Scale bars: 200 µm in A and C and 100 µm in B, E and F.

Figure 5. Ectopic axonal bundles in necdin null mice. ( A ) Coronal section at E14.5 at the level of the anterior hypothalamus (Ah) stained with thionin. The ectopic axonal bundle is boxed in red. 3v, third ventricle; Ot, optic tract. ( B ) Higher magnification of the red boxed region from A and A′. ( C – E ), Serial coronal sections at E16.5. (C) The ectopic bundle and an ectopic axonal whorl in the striatum (>) and abnormalities of the internal capsule (Ic) are visualized as non-thionin staining regions in the necdin null. (D) L1-positive axons (red) are misrouted around the region of the mutant internal capsule (>) and thalamocortical tracts (Tca) are reduced in the necdin null cortex (arrow). The ectopic bundle and ectopic whorls also label with L1. (E) RGC axons labeled with DiA in the right optic cup (left on this image) do traverse partway around the ventral thalamus, and a small number axons deviate into the ectopic bundle, outlined by a white dotted line. Necdin null RGC axons are not visible in the superior colliculus (Sc). Ctx, cortex; Hy, hypothalamus. Hoechst counterstain is blue in (D) and (E). Scale bars: 500 µm in A and D, 50 µm in B and E, 1 mm in C.

Figure 6. Embryonic origins of axonal defects and overlap in expression in hypothalamic neuroepithelium ( A , B ) RNA in situ hybridization with Ndn (A, C ) and Fez1 (B, D ) on sagittal (rostral to the left) (A, B) or transverse (C, D) sections of E10.5 embryos. The arrows point to co-expression in the mesencephalon. Is, isthmus; Me, mesencephalon; Po, pons; 4v, fourth ventricle. ( E – H ) Serial coronal sections at E12.5, stained with thionin (E), Sim1 by RNA in situ hybridization (F) and Magel2 by RNA in situ hybridization (G) marking specific zones in the hypothalamic neuroepithelium (Hne) and zona limitans (Zl). Oc, optic chiasm. (H) L1 staining is aberrant in the ventral region of the E13.5 necdin null dorsal thalamus (D, arrowhead in H ′). Hy, hypothalamus. Hoechst counterstain is blue. Scale bar, 100 µm.

Figure 7. Serotonergic rostral projections. ( A ) Serotonin transporter (SERT) RNA in situ hybridization at E12.5 detects serotonergic soma in control and necdin null hindbrains. Arrow points to SERT expression (blue) in the rostral serotonergic group (Rg). ( B ) Serotonin (5-HT) immunohistochemistry (red) of a transverse section at E12.5. Arrow marks the ascending serotonergic fibers in the mesencephalon, which are reduced in the necdin null ( B ′). ( C ) 5-HT labeling of sagittal sections at E15.5 reveals a reduction in ascending projections (white box) from the rostral group. Me, mesencephalon; Po, pons. Rostral is to the left. Scale bars: 100 µm in A and B and 500 µm in C.

Figure 8. Tyrosine hydroxylase (TH) immunohistochemistry for catecholaminergic cell bodies and projections. ( A ) E15.5 thionin stained section at the level of the ectopic axonal bundle. Arrowheads mark ectopic axonal whorls. Ic, internal capsule; Lv, lateral ventricle; 3v, third ventricle. ( B ) adjacent sections stained with TH reveal disorganization of the nigrostriatal pathway (Nsp), arrowhead in (B′). ( C ) a TH-stained section ∼600 µm rostral to that in (B) reveals little TH staining in the Nsp in the necdin null (nn). ( D ) A TH-stained section ∼600 µm caudal to that in (B), showing relatively normal TH staining in the dorsomedial hypothalamus (Dmh) and Nsp. Hoechst counterstain is blue. Scale bar, 500 µm.