Podocalyxin Is a Novel Polysialylated Neural Adhesion Protein with Multiple Roles in Neural Development and Synapse Formation

Neural development and plasticity are regulated by neural adhesion proteins, including the polysialylated form of NCAM (PSA-NCAM). Podocalyxin (PC) is a renal PSA-containing protein that has been reported to function as an anti-adhesin in kidney podocytes. Here we show that PC is widely expressed in neurons during neural development. Neural PC interacts with the ERM protein family, and with NHERF1/2 and RhoA/G. Experiments in vitro and phenotypic analyses of podxl-deficient mice indicate that PC is involved in neurite growth, branching and axonal fasciculation, and that PC loss-of-function reduces the number of synapses in the CNS and in the neuromuscular system. We also show that whereas some of the brain PC functions require PSA, others depend on PC per se. Our results show that PC, the second highly sialylated neural adhesion protein, plays multiple roles in neural development.


Introduction
Neural migration and axonal guidance are governed by several families of extracellular cues, which elicit either attractive or repulsive responses on leading edges and axonal growth cones. Prominent members of these families include Netrins and various classes of Semaphorins [1,2,3]. In addition, neural development involves cell-to-cell contact and adhesion to the extracellular matrix, which also contribute to the assembly of brain regions and the formation of axonal connections [4,5]. Adhesion molecules, such as NCAM, L1, or TAG1, have pivotal roles in axonal growth and fasciculation, neural cell migration and synaptogenesis [6,7,8,9,10]. Moreover, some of these proteins cooperate in signaling events triggered by extracellular factors [6,11,12]. In previous studies, we have shown that the highly sialylated renal anti-adhesin Podocalyxin (PC) is expressed in the developing brain [13,14]. PC is the main glycoprotein expressed on the apical surface of glomerular podocytes. PC is a 140-160 kDa type I transmembrane protein composed of a highly sialylated ectodomain and a short cytoplasmic tail [15,16]. PC has a strong negative charge and it has been proposed as an anti-adhesin responsible for maintaining the filtration slits open [17,18]. podxldeficient mice die soon after birth because of defects in kidney development and mutant podocytes do not form foot processes, which leads to glomerular reduced permeability and anuria [19]. PC is also expressed in vascular endothelia, mesothelial cells, hematopoietic stem cells and in several types of tumors [20,21,22,23,24]. In most circumstances, PC blocks adhesion. In the endothelial venules, however, PC acts as an adhesive ligand for L-selectin-expressing leukocytes [23]. The cytosolic tail may also contribute to the unique organization of podocytes. Two cytosolic adaptor proteins, Na + /H + -Exchanger Regulatory Factor 2 (NHERF2) and Ezrin, interact with PC in kidney [25]. Given the crucial role of PSA in multiple steps during neural development [26,27,28], here we examined the role of PC in brain development. We show that PC is involved in axonal fasciculation and neuritogenesis, and in synaptogenesis.

Results
Brain PC is a poly-sialylated protein widely expressed during brain development In agreement with a previous study [14], PC mRNA was widely expressed in the developing brain from E12 to adult stages ( Fig. 1a and Fig. S1). To analyze the fine distribution of PC, we used immunohistochemistry (Fig. 1b-f and Fig. S1). Two antibodies recognizing the PC extracellular domain (chicken anti-mouse PC and mouse anti-human PC, gift of D. Kershaw; [29]), gave a similar staining pattern to that of the mRNA expression, labeling preferentially cell bodies (Fig. 1b-f and Fig. S1). In contrast, the rat monoclonal antibody, which also recognizes an extracellular epitope [30], preferentially stained the neuropile and fibers (Fig.  S1). Using this antibody, we detected PC in many axonal fascicles throughout the brain during perinatal stages (Fig. S1a). PC mRNA and protein were detected in the proliferative ventricular zones, and especially in postmitotic neurons, including pyramidal cells in the cerebral cortex and hippocampus, periglomerular and granule cells in the olfactory bulb, and Purkinje and granule cells in the cerebellum (Fig. 1a-f and Fig. S1). Postnatal proliferative zones, such as the EGL in the cerebellum and the SVZ in the forebrain were also intensely labeled (Fig. 1c,e and Fig. S1). PC was highly expressed in laminated regions, such as the olfactory bulb, cerebral cortex, hippocampus and cerebellum, and expression was also detected in many nuclei throughout the brain ( Fig. 1a-f, and Fig.  S1). Immunostaining of sections from podxl (2/2) brains did not reveal immunolabeling (Fig. S1).
To localize PC with greater precision, hippocampal neuronal cultures were analyzed by immunofluorescence ( Fig. 1g-l). At 2-4 DIV PC immunoreactivity was detected in neuronal cell bodies, dendrites and in axons, including growth cones ( Fig. 1g-l). Hippocampal explants stained with the PC antibodies confirmed that this protein was present in axons and in growth cones ( Fig. 1j-l). At later stages (1-2 weeks in vitro) PC protein was in axonal presynaptic bouton-like structures (Fig. S1 d,h).
Expression of PC protein in brain was corroborated by Western Blot with expression peaking at prenatal and early postnatal stages (Fig. 2a,b). No band was detected in extracts obtained from podxl (2/2) brains (Fig. 2a). In some cases a second, weak PC band was apparent in Western Blots (see Fig. 2c). This band was at a similar height as Neuraminidase-treated PC (see below). In kidney, PC is a highly poly-sialylated protein [16,31]. To assess whether PC is sialylated in brain, extracts of dissociated neuronal cultures and brain lysates were incubated with a2-3, 6, 8, 9-Neuraminidase (Fig. 2c). Western Blot analysis revealed that the 140-kDa band is sensitive to Neuraminidase activity since the mobility of PC was decreased (arrow in Fig. 2c; Fig. S2), as it is in kidney lysates. This lower mobility may be caused by the lack of negative charges [17]. In contrast, incubation with EndoN did not remove PSA from PC but specifically degraded PSA-NCAM (Fig. S2). This differential degradation supports that PSA in PC is bound by alpha2-6 and alpha2-3 links (O-glycosylation), whereas PSA-NCAM is formed by alpha2-8 links (N-glycosylation) [32].
Brain PC forms a protein complex with Ezrin, NHERF1/2 and RhoA/G In kidney and MDCK cells, PC interacts with Ezrin and the sodium-hydrogen exchanger regulatory factor 1 and 2 (NHERF1/ 2) [25,33,34]. To study this interaction in neural tissue, coimmunoprecipitation analyses were performed in E16 brains. When PC antibodies were used to immunoprecipitate lysates, a band of about 80 kDa was detected by Western Blot using anti-Ezrin antibodies (Fig. 2d). PC protein was identified in complementary immunoblots from brain lysates immunoprecipitated with anti-Ezrin antibodies (Fig. 2d). Experiments using an antibody recognizing the Ezrin/Radixin/Moesin family of proteins gave similar results (Fig. 2e). Co-immunoprecipitation experiments also yielded an interaction of brain PC and NHERF1 (Fig. 2f). Given that NHERF2 has also been found to be highly expressed in neuronal tissue, we also demonstrated interaction of PC with this additional member of the NHERF family (Fig. 2g). Finally, in MDCK cells, PC is linked to small G protein RhoA [34]. Coimmunoprecipation analyses in brain lysates revealed that neural PC also interacts with RhoA and with the related small GTPase RhoG (Fig. 2h,i). Controls, including GFP-imunoprecipitation, did not reveal immunolabeling with PC, Ezrin, Ezrin/Radixin/ Moesin, RhoA/G or NHERF1/2 antibodies (Fig. 2d-i). Further immunoprecipitation controls using WB with an irrelevant antibody (the transcription factor EGR1) did not show signals ( Fig. 2d-i).

PC is not required for neuronal migration
To analyze the role of PC in brain development, we used podxldeficient mice [19]. Given that these mice die after birth, embryos were analyzed at E18. Nissl-staining of brain sections revealed that the overall organization and cytoarchitecture of podxl (2/2) brains was similar to that of wt embryos. Thus, brain regions, nuclei and layers were clearly recognizable throughout the brains of podxl (2/2) embryos, including the layered organization of the cerebral cortex, olfactory bulb and cerebellum, and the nuclear distribution in the thalamus and brain stem ( Fig. 3a-j). These data suggest that PC is not required for cell proliferation or neuronal migration.
To substantiate these observations, pregnant females were injected with single BrdU-pulses at E12 or E15 and the pattern of radial migration was studied in the neocortex. E12-labeled cohorts were fated to the subplate and layer VI in wt and podxl (2/2) embryos ( Fig. 3k-l); similarly, E15-labeled neurons were properly fated in the upper cortical layers in both genotypes ( Fig. 3m-n). Correct positioning of migrating neurons in the neocortex was also observed with anti-Tbr1 antibodies (Fig. 3o,p), a marker expressed by deep neurons of the cerebral cortex [35]. Taken together, these findings indicate that radial migration is not altered in podxldeficient brains.
To examine the involvement of PC in tangential migration, we analyzed neurophilic chain migration using lower rhombic lip explants (Fig. 3q,r). The neurons that migrate from this proliferative region form the circumferential migratory stream, which will give rise to the precerebellar nuclei [36,37,38]. In wt rhombic lip explants co-cultured with aggregates of Netrin-1expressing cells, typical chains of migrating neurons were formed that were chemoattracted by Netrin-1-expressing cells (Fig. 3q,r). A similar chemoattractive response was observed in podxl (2/2) rhombic lip explants, which showed normal exit of neurons and the formation of migratory chains (Fig. 3q,r). We conclude that neither radial nor tangential neuronal migration is impaired by the absence of PC.

Axonal extension, fasciculation and branching is impaired in PC-deficient neurons
To examine the contribution of PC to axonal growth, E18 brain sections were stained with antibodies against the adhesion proteins L1 and TAG-1 (Fig. S3). Immunostained sections showed that all major axonal tracts were correctly formed in podxl (2/2) brains ( Fig.  S3, S4). For example, in the forebrain the major commissures (corpus callosum, hippocampal and anterior commissure) and axonal pathways (anterior olfactory tract, reciprocal thalamocor-  tical pathway and cortical white matter) were well developed in these mutant embryos (Fig. S3). Similarly, the topography and radial distribution of fibers was not substantially altered, as seen in the hippocampus and neocortex. Also in the midbrain and hindbrain, all the major axonal tracts in podxl (2/2) brains were distributed normally (not shown). To further analyze axonal targeting in podxl (2/2) embryos, DiI injections were performed in several regions, including the neocortex, olfactory bulb, dorsal thalamus, entorhinal cortex and cerebellum. As illustrated in Fig.  S4, cortical injections in the somatosensory cortex labeled corticothalamic fibers which extended through the internal capsule and terminated appropriately in the dorsal thalamus in wt and podxl (2/2) brains. Similarly, we found no evidence of aberrant axonal targeting after DiI injections in the remaining brain areas (e.g., the LOT, Fig. S4e,f). These observations indicate that axonal guidance and targeting is not impaired by the lack of PC.
A closer examination of axonal tracts, however, revealed that the shape and size of axonal fascicles differed. Thus, in the mutant hippocampus axonal bundles in the white matter were smaller, less compact and occupied a wider zone in the adjacent stratum oriens than those in wt littermates (Fig. S3). Similarly, other axonal tracts, such as the habenulo-peduncular tract in the dorsal thalamus ( Fig.  S3) or the fornix, displayed reduced fasciculation. These data indicate that PC is involved in neuron-to-neuron adhesion and in the fasciculation of developing axonal tracts. Next, we examined the role of PC in axonal growth in vitro. Hippocampal wt explants gave rise to numerous axons that fasciculated and grew along straight courses (Fig. 4a). The pattern of axonal outgrowth in podxl (2/2) explants differed dramatically: axons grew aberrantly following sinusoidal trajectories with little axonal fasciculation and with profuse branching (Fig. 4b). Overall, the pattern of axonal growth appeared as a dense meshwork of crisscross fibers. An identical phenotype was observed when wt explants were cultured on coverslips coated with soluble PC-ectodomain, which blocks the PC function (Fig. 4c).
To substantiate these findings, dissociated hippocampal neurons from wt and podxl (2/2) mice were cultured and neurite length and the number of branching points per neuron were calculated (Fig. 4d, e, g-i). podxl (2/2) neurons exhibited increased neuritogenesis, extended neuritis and a two-fold increase in the number of branching points per neuron compared to wt neurons ( Fig. 4g-i). Similar results were observed when wt hippocampal neurons were cultured on a PC ectodomain substrate, used as a PC blocking reagent, in comparison with control cultures incubated with Fc protein ( Fig. 4f and j-l). These findings indicate that PC is involved in axonal elongation, fasciculation and axonal branching.

The lack of PC impairs synaptogenesis
We next examined whether PC participates in synapse formation. First, PC localization was analyzed in 7-10 DIV hippocampal cultures. PC was enriched in axonal-like varicosities, where PC co-localized with synaptic proteins such as Synaptophysin and Synapsin I and II (Fig. 5a-c and 6a-c). To confirm that PC was expressed in presynaptic terminals, we prepared adult synaptosomes and analyzed the presence of PC protein by Western Blot. PC was enriched in adult synaptosomal (6-fold) compared to total brain homogenates (H, total homogenates; and SS, synaptosomal fraction; in Fig. 5d). Further fractionation of synaptosomal preparations through a sucrose gradient showed that PC was enriched in the vesicular fractions (F10-F13, displaying Synaptophysin, VAMP2 and SNAP25 immunoreactivities) compared to the cytosolic (F3-F5, tubulin) and membrane (F6-F9, MBP immunoreactivity) fractions (Fig. 5e). Finally, immunogold techniques were used to localize PC at the fine structural level. Hippocampal dendrites and postsynaptic dendritic spines were devoid of immunolabeling. In presynaptic axon terminals gold particles were preferentially associated with synaptic vesicles (Fig. 5f,g). These data indicate that PC is enriched in developing and mature presynaptic axon terminals in the adult brain.
To determine the possible involvement of PC in synaptogenesis, we first studied the formation of synaptic-like appositions in vitro ( Fig. 6d-i). Hippocampal cultures (7DIV) were immunostained with MAP2 and with the presynaptic marker Synapsin I and II, and presynaptic appositions over MAP2-positive dendrites were counted. podxl (2/2) neurons showed about a 25% decrease in the density of presynaptic-like structures apposed to MAP2-immunolabeled dendrites (Fig. 6j). A similar decrease was observed when wt neurons were incubated with the blocking PC ectodomain protein, but not with control Fc protein (Fig. 6k). We next examined synaptic development in E18 brains in vivo by electron microscopy (Fig. 6l,m). Presynaptic terminals containing a few synaptic vesicles were identified in the embryonic hippocampus in wt and podxl (2/2) embryos. The density of synaptic contacts was decreased by 23% in the stratum lacunosum moleculare and by 25% in the stratum radiatum of the hippocampus of the latter compared to wt littermates (Fig. 6n). podxl-deficient mice do not survive after birth, therefore to ascertain whether decreased synaptogenesis is transient or persistent, we prepared organotypic hippocampal slices from E18 embryos that were cultured for 7 DIV. Electron microscopy examination did not show changes in the morphology of synaptic contacts between both genotypes. Again, the density of synaptic contacts was decreased (by 28%) in the hippocampus of the former (Fig. 6o-q).
Because PC mRNA and protein were also expressed in motorneurons (Fig. 7a,b), we analyzed the role of PC in the formation of neuromuscular synapses. Sections from the soleus muscle were stained with a-bungarotoxin (to stain cholinergic receptors and thus synaptic junctions) and anti-Neurofilament 200/Synaptophysin antibodies (staining axons) to label post-and pre-synaptic terminals respectively. The density of neuromuscular synaptic specializations in 1610 4 um 2 surface samples (open squares in Fig. 7d,g) was calculated. Neuromuscular synapses were decreased by 32% in podxl (2/2) E18 embryos compared to wt littermates ( Fig. 7c-h). These findings implicate PC protein in synaptogenesis in the central and peripheral nervous system.

PSA-dependent and -independent functions of PC in neural development.
To determine whether the functions of brain PC are dependent on PSA or on PC per se, we carried out in vitro experiments by incubating cells with PSA-PC ectodomain or with this ectodomain treated with Neuraminidase (non-sialylated PC). We also used EndoN (cleaving specifically PSA-NCAM) as a control (Fig. 8). We addressed whether axonal phenotypes depended on PSA-bound PC. In hippocampal neurons, the increase in neurite length induced by PSA-PC was blocked when PSA was removed from PC ( Fig. 8a-d). In contrast, the increased neuronal branching seen in PSA-PC treated cultures remained unchanged upon removal of PSA ( Fig. 8a-d). Treatment with EndoN did not result in modifications. These findings suggest that while PSA-PC is required for the neurite growth phenotype, non-sialylated PC is responsible for the branching phenotype.
Using a similar approach, we also found that both sialylated and non-sialylated forms of PC induced an identical decrease in hippocampal synaptogenesis (Fig. 8e-k), thereby indicating that PC per se (and not PSA) is responsible for this reduction. Taken together, our results strengthen the notion that the neurite extension function of PC is mediated by PSA carried by the protein, while non-sialylated PC is responsible for neurite branching and synapse formation phenotypes.

Discussion
Cell adhesion molecules play an essential role during several developmental stages in neural tissue [7,39]. PSA-NCAM has a pivotal role in these functions [7,40]. Here we show that PC is widely expressed throughout CNS development and in the adult brain, and that it is highly sialylated in neuronal tissue. The presence of a second, weak PC band, particularly enriched in synaptosomal fractions, suggests that both sialylated and nonsialylated forms of PC are expressed in brain. To our knowledge, PC is the second highly-sialylated adhesion glycoprotein expressed in neurons, in addition to NCAM. Here we show that the lack of PC function results in increased axonal elongation, branching and synaptogenesis. Furthermore, our study demonstrates the PC effects on branching and synaptogenesis do not require PSA, but appear to be dependent on PC protein alone. The highly negatively charged PSA-PC is believed to act as an anti-adhesive protein by a charge repulsion mechanism in kidney podocytes [25,41]. Such an anti-adhesive role is consistent with the increased neurite length phenotype, which is blocked after PSA removal. Other PC phenotypes, however, particularly in processes that are independent of PSA (axonal branching and synaptogenesis), are unlikely to respond to an anti-adhesive function, but correlate better with an adhesive role of PC. However, the putative PC ligands that mediate brain adhesion remain to be identified.
The highly polysialylated form of the NCAM is required for the proper tangential migration of interneuron precursors along the Rostral Migratory Stream (RMS). Mice deficient for all the splice variants [42] or for only the 180-kDa isoform [10] show a reduction in olfactory bulb size and an accumulation of migrating neurons along the RMS. At early postnatal stages, this effect is phenocopied by genetic or enzymatic removal of PSA, thereby indicating the relevance of PSA modification in this process [27,43]. In the present study we found no defects either in the olfactory bulb or in the RMS of podxl (2/2) embryos. This observation indicates that, at least in prenatal stages, PC is not essential for chain migration in the RMS. In contrast, NCAM is not required for radial migration in laminated brain regions, such as the cerebral cortex, the cerebellum or the olfactory bulb in vivo [10,42]. Interestingly, ST8SiaII and ST8SiaIV double-mutant mice show reduced radial migration in the neocortex [26], thereby implicating an additional PSA-containing protein in radial migration. As our phenotypic studies in podxl (2/2) embryos did not show radial migration alterations, NCAM and PC may have redundant functions in neuronal radial migration. One of the most dramatic alterations produced by the lack of PC function in neurons is enhanced growth and branching of neurites and axons. This finding suggests that PC is a negative regulator of neurite branching, at least in embryonic stages. Although neurite branching is crucial to development, its mechanisms are not fully understood. A number of adhesion and extracellular molecules, including Netrins, Semaphorins and BDNF regulate neurite branching [44,45,46]. In contrast, RhoA is a negative regulator of dendrite outgrowth in various organisms and cell types [47,48]. In kidney podocytes PC forms a complex with Ezrin and the NHERF2, which promotes RhoA activation and stabilizes the actin cytoskeleton [25,34]. In the present study, we found that brain PC interacts with the Ezrin/NHERF1-2/RhoA-G complex, thereby suggesting that activation of this protein complex is responsible for the reduction in neurite growth and branching observed in PC loss-of-function models.
The cell adhesion molecule NCAM has been shown to act as a co-receptor for the neurotrophic factor GDNF [12]. A recent study has demonstrated that in the hematopoietic system Podocalyxin co-associates with the SDF-1 receptor CXCR4 thus modulating the biochemical signaling response to this cytokine [49]. It is therefore possible that part of the functions of Podocalyxin in neural development observed in the present study are related to the above molecular mechanism, i.e., modulation of neurotrophic factor and cytokine signaling by this novel adhesion protein.
Deficiency of NCAM, and other adhesion molecules such as L1, often results in axonal defasciculation [9,50,51]. ncamdeficient mice, for instance, show defasciculation of mossy fibers in the hippocampus [50] and pathfinding errors and reduced fasciculation of the corticospinal tract [52]. Interestingly, defects in mossy fiber lamination and major axonal tracts, including malformations in the anterior commissure, corticospinal tract and the corpus callosum, are more dramatic in ST8SiaII and ST8SiaIV double-mutant mice lacking PSA [26,53]. These observations suggest that, in addition to NCAM, PSA activity is essential for the formation and fasciculation of axonal tracts. Our findings show that PC controls axonal fasciculation in vitro. Since the phenotype of podxl (2/2) embryos is relatively subtle in vivo (though dramatic in vitro), our results suggest compensatory mechanisms in podxl (2/2) embryos, possibly caused by PSA-NCAM expression, thereby suggesting functional redundancy of PC and PSA-NCAM in axonal guidance and fasciculation. PSA-NCAM is essential for neural plasticity [54,55,56]. In vitro experiments show that PSA-NCAM is involved in axonal target selection and stabilization of the synapse [40,57]. However, ncam (2/2) mice do not show differences in the number of CNS and PNS synapses [57,58]. Fasciclin II, the NCAM homologue in Drosophila [59], and the Aplysia cell adhesion molecule (apCAM) [60] have an essential role in synapse formation, which suggests functional genetic redundancy in mammals. A recent study, however, provides evidence that the removal of PSA-NCAM from the cerebral cortex at postnatal stages leads to a selective precocious maturation of GABAergic synapses [61]. This observation implies that PSA-NCAM may play synaptogenic functions in specific neural populations and developmental ages. Here we report that the lack of PC results in fewer synaptic contacts both in vivo and in vitro in the CNS and in the neuromuscular system, thereby suggesting that PC is required for the correct formation or stabilization of synapses. In neural development, signals that trigger increased axonal branching and elongation, such as BDNF, often lead to an increased number of synapses [62]. This evidence suggests that the increase in synapses is secondary to longer axonal lengths. However, in our study, the increment of axonal branches in PC loss-of-function models (expected to be associated with increased synaptogenesis) produced an opposite effect, i.e. reduced number of synapses, thereby supporting a direct role of PC in synapse formation or maintenance. Given that the lack of PSA does not affect overall synapse numbers [40], the role of PC in this process may be independent of PSA. Our experiments showing that either PSA-PC or non-sialylated PC triggers identical decreases in synaptogenesis support this view (Fig. 7h).
In summary, here we demonstrate that, in addition to PSA-NCAM, the PSA-containing PC protein plays pivotal roles in several processes during early brain development, including neurite outgrowth and branching, fasciculation and synaptogenesis. These developmental processes may be mediated by an NHERF/Ezrin/RhoA pathway, linking PC and the actin cytoskeleton. Our results indicate that PC has a dual antiadhesive/adhesive role in successive steps of neural development, which are likely to be mediated by highly sialylated and nonsialylated forms of the protein. Thus, our data provide evidence for a role of the Sialomucin/CD34 protein family in the development of the nervous tissue.

Materials and Methods
All procedures were performed in accordance with the guidelines approved by the Spanish Ministry of Science and Technology and following the European Community Council Directive 86/609 EEC.

Animals
PC-deficient embryos (podxl (2/2) ) were obtained and genotyped as described previously [19]. OF1 embryos and postnatal mice (Iffra Credo, Lyon, France) were used in this study. The mating day was considered as embryonic day 0 (E0) and the day of birth as postnatal day 0 (P0). For OF1, the following developmental stages were studied: E12, E14, E16, E18, P0, P5, P10, P15, P21, and adult (three to five animals for each stage). Animals were anesthetized with 4% halothane. Since podxl 2/2 mice die within 24 h of birth, embryos were used. Wt embryos of the same littermates were used as controls in all the experiments.

PC ectodomain-Fc preparation
DNA encoding the Fc of rabbit IgG was cloned into the 59-Pst and 39-XhoI sites of the pSecTag2A vector (Invitrogen, Carlsbad, CA). The cDNA corresponding to the extracellular domain from mouse PC were generated by PCR and cloned 59of the Fc to produce PC ectodomain-Fc fusion protein. The PC ectodomain fusion construct was expressed in EBNA-293 cells for 2-3 days following transfection with Lipofectamine Plus (Invitrogen, Carlsbad, CA). The supernatant was collected, concentrated, filtered and maintained at 4uC for dissociated and explant culture treatment.

Immunohistochemistry and in situ hybridization
Embryos and postnatal mice were perfused transcardially with 4% paraformaldehyde. Thereafter, the brains were cryoprotected and frozen on dry ice. Coronal sections (thickness: E14-E18: 50 mm, P0-adult: 30 mm) were obtained and processed for immunohistochemistry. Sections were washed in PBS and PBS-Triton X-100, blocked for 2 h and incubated with primary anti-PC antibodies diluted in blocking solution, overnight at 4uC. The sections were then incubated with secondary biotinylated antibodies (Vector Laboratories, Inc., Burlingame, CA) in blocking solution for 2 h and with a streptavidin-horseradish peroxidase complex (Amersham Pharmacia Biotech). Sections were developed with 0.03% diaminobenzidine and 0.002% hydrogen peroxide, mounted on slides, dehydrated, and coverslipped. Sections were also stained by immunofluorescence. Incubation with nonimmunized IgGs and omission of primary antibodies prevented immunostaining. In situ hybridization was performed on freefloating sections as described previously [14].
The soleus muscles of E18 podxl (2/2) and control littermates were dissected. The muscles were cryoprotected in 30% sucrose-PBS and frozen on dry ice. Longitudinal sections (thickness: 50 mm) were obtained with a Leica CM 1325 cryostat and cryoprotected in a solution containing 30% glycerol, 30% ethylene glycol, 40% 0.1 M (PBS), and stored at 230uC until use. For immunohistochemistry, muscle sections were processed as described above. FITC-conjugated bungarotoxin (10 28 M, Molecular Probes, Eugene, USA) was added with secondary antibodies overnight at 4uC. Sections were recorded in a Leica TCS 4D laser scanning confocal microscope (Leica Lasertechnik). For quantitative analysis of AChR clusters, the numbers of AChR clusters in a matching 1610 4 mm 2 area from control (n = 3) and mutant (n = 3) soleus were counted.
Post-embedding immunocytochemistry OF1 adult mice (n = 2) were perfused with 4% PFA-0.1% glutaraldehyde in 0.12 M phosphate buffer. Brains were removed and small samples of hippocampus were dissected, cryoprotected gradually in sucrose and cryofixed by immersion in isopentane. Freeze-substitution was performed at 290uC over 3 days in an ''Automatic Freeze Substitution System'' (AFS, Leica), using methanol containing 0.5% uranyl acetate as substitution medium. Brains were infiltrated in Lowicryl HM20 at 250uC and then polymerized with UV lamps. Ultrathin sections were collected and processed for post-embedding PC immunostaining using a chicken anti-PC antibody (1:25), and biotinylated rabbit anti-chicken (1:100; R&D System) and 15 nm colloidal gold-coated secondary antibodies (BBI; 1:25). In control experiments, the primary PC antibody was omitted. No immunogold labeling occurred in these conditions.

Electron microscopy
E18 mutants (n = 3) and control (n = 3) littermates were perfused with 1% glutaraldehyde-1% paraformaldehyde in 0.12 M phosphate buffer. Brains were fixed in the same solution overnight. Tissue slices were post-fixed with 2% osmium tetroxide, stained with 2% uranyl acetate and embedded in Araldite. Ultrathin sections were collected on formvar-coated slot grids and stained with lead citrate. Electron micrographs covering 64 mm 2 (final magnification 20,0006) were randomly taken from the stratum radiatum and the stratum lacunosum-moleculare of the hippocampus, and the number of synaptic contacts was counted (n = 50-77 micrographs per layer and group).
In addition, hippocampal slice cultures were prepared from E18 podxl (2/2) mice and control littermates (n = 2 embryos per group) as described [63]. Mice were anesthetized by hypothermia, their brains were removed, and the hippocampal formations were dissected out. Horizontal sections (350 mm thick) were obtained using McIlwain tissue chopper. Selected slices were cultured using the interphase membrane method [64]. After 7 DIV, hippocampal cultures were fixed in 1% glutaraldehyde-1% paraformaldehyde in 0.12 M phosphate buffer and processed for electron microscopy (see above). Ultrathin sections of the stratum radiatum from two hippocampi per animal type were obtained and electron micrographs covering 64 m 2 (n = 42 per group) were randomly taken and the density of synaptic (both symmetric and asymmetric) contacts was analyzed.

DiI tracing
For tracing developing connections, the neocortex, dorsal thalamus, olfactory bulb, entorhinal cortex and cerebellum of E18 embryos were injected with a small crystal of the lipophilic tracer DiI (1,19 dioctadecyl-3,3,39,39 tetramethylindocarbocynanine; Molecular Probes, Eugene, USA). Three podxl (2/2) and three wt counterparts were used for these studies. After some weeks in fixative, vibratome coronal sections were stained with bisbenzimide and viewed under epifluorescence.
Samples were loaded and run in polyacrylamide gels at 100 V. After running, transfer to nitrocellulose membranes was per-formed in 120 mM glycine, 125 mM Tris, 0.1% SDS, and 20% methanol, 10% mercaptoethanol. Transfer was performed at 35 V ON. Filters were then blocked in 5% powder milk in TBS and incubated with primary antibodies (anti-PC or -PSA-NCAM). Secondary antibodies were used diluted 1:2500 in TBS containing 5% powder milk. Labeling was visualized with ECL plus (Amersham Pharmacia Biotech).
For neurite length and branching analysis, 2-3 day-old wt, podxl (2/2) and wt primary cultures incubated with conditioned medium containing PC ectodomain (30-50 ml conditioned medium in 450 ml of total medium) or with control medium (rabbit Fc) were fixed and labeled with anti-TUJ-1. Quantification of axonal length from hippocampal cells was performed using IMAT software (developed by the Technical Services of the University of Barcelona). For quantification of neurite length and branch number, fixed cells were viewed at 206 magnification. There were about 80 neurons per group. Neurite length was measured from the cell body to the distal end of the process. Total neurite length is the sum of all primary neurites and branches produced by a single neuron. Branching points are the points at which a neurite extends from another neurite. Branching number is the sum of each branching point from a single neuron. Primary neurites are those that extend directly from the soma. To count synaptic contacts, 7-day-old wt, podxl (2/2) and wt primary cultures incubated with conditioned medium containing PC ectodomain (30-50 ml conditioned medium in 450 ml of total medium) or control medium (rabbit Fc) were fixed and labeled with the MAP-2 and Synapsin I and II antibodies. The density of synaptic appositions was counted. A total of 56-75 neurons per group, from at least 3 separate experiments, were counted.

Explant Cultures
Wt and podxl 2/2 explants from the E16 CA3 hippocampal region, the E13 lower rhombic lip (lRL) and E16 optic nerve were used in this study. Hippocampal explants were cultured for 3 days on a laminin substrate and were stained with the TUJ1 antibody, which was visualized with immunofluorescence. For functionblocking analysis, wt hippocampal explants were cultured on a substrate with laminin and the PC ectodomain (1:2). After 2-3 days of culture, the explants were labeled as described above.
lRL explants of wt and podxl (2/2) embryos were co-cultured with aggregates of Netrin 1-expressing 293T cells or aggregates of control 293T cells for 2 days in a Matrigel matrix (Becton Dickinson). Explants were stained with TUJ-1 mAb and bisbenzimide (Sigma-Aldrich, St Louis, MO) and visualized by immunofluorescence.

Neuraminidase and EndoN treatments in cultures
The PC ectodomain-Fc fusion construct (or its control rabbit Fc) was expressed and concentrated as above. Aliquots of PCectodomain were incubated with Neuraminidase or EndoN for 1-3 h (see above). Western Blots with anti-PC or PSA-NCAM antibodies were routinely run to assess the PC sialylation patterns. Dissociated hippocampal cultures were prepared as above and cultured for either 3 DIV or 7 DIV in the following conditions: untreated with PC, control substrate with rabbit Fc, PCectodomain substrate, PC-Ectodomain substrate incubated with EndoN or PC-Ectodomain substrate incubated with Neuraminidase. Cultures were fixed and stained with the TUJ-1 antibody or with the MAP2/Synaptohysin antibodies as above.

Statistical analysis
Data were analyzed with the program Statgraphics Plus 5.1 using the ANOVA and the Student's t test or the Mann Whitney (W) test. Minimal statistical significance was fixed at p,0.05. In Figures, * indicates P,0.05, ** indicates P,0.01 and *** indicates P,0.001. Data are expressed as mean 6 s.e.m.