Schwannomin/merlin is the product of a tumor suppressor gene mutated in neurofibromatosis type 2 (NF2). Although the consequences of NF2 mutations on Schwann cell proliferation are well established, the physiological role of schwannomin in differentiated cells is not known. To unravel this role, we studied peripheral nerves in mice overexpressing in Schwann cells schwannomin with a deletion occurring in NF2 patients (P0–SCH–Δ39–121) or a C-terminal deletion. The myelin sheath and nodes of Ranvier were essentially preserved in both lines. In contrast, the ultrastructural and molecular organization of contacts between Schwann cells and axons in paranodal and juxtaparanodal regions were altered, with irregular juxtaposition of normal and abnormal areas of contact. Similar but more severe alterations were observed in mice with conditional deletion of the Nf2 gene in Schwann cells. The number of Schmidt–Lanterman incisures, which are cytoplasmic channels interrupting the compact myelin and characterized by distinct autotypic contacts, was increased in the three mutant lines. P0–SCH–Δ39–121 and conditionally deleted mice displayed exuberant wrapping of nonmyelinated fibers and short internodes, an abnormality possibly related to altered control of Schwann cell proliferation. In support of this hypothesis, Schwann cell number was increased along fibers before myelination in P0–SCH–Δ39–121 mice but not in those with C-terminal deletion. Schwann cell numbers were also more numerous in mice with conditional deletion. Thus, schwannomin plays an important role in the control of Schwann cell number and is necessary for the correct organization and regulation of axoglial heterotypic and glio-glial autotypic contacts.
- neurofibromatosis type 2
- nodes of Ranvier
- Schmidt–Lanterman incisures
Neurofibromatosis type 2 (NF2) is a dominant autosomic disease characterized by the occurrence of multiple schwannomas, as well as ependymomas and meningiomas. The Nf2 tumor suppressor gene codes for a protein closely related to ezrin, radixin, and moesin (ERM) termed schwannomin (Rouleau et al., 1993) or merlin (Trofatter et al., 1993). Cell type-specific mutations of schwannomin or its absence lead to tumor formation in various mouse tissues (McClatchey et al., 1998; Giovannini et al., 1999, 2000; Kalamarides et al., 2002). Schwannomin is thought to play a critical role at the plasma membrane, controlling cell–cell interactions and signaling pathways triggered by cell contacts (McClatchey and Giovannini, 2005; Okada et al., 2007). It stabilizes adherens junctions (Lallemand et al., 2003) and interacts with the cytoskeleton, cytoskeleton-associated proteins, many transmembrane and adaptor-scaffold proteins, and a variety of signaling proteins (Okada et al., 2007). However, besides its importance as a tumor suppressor, the physiological role of schwannomin is still poorly understood, and its function in differentiated Schwann cells is not known.
During development of peripheral nerves, immature Schwann cells give rise to myelinating nonmyelinating cells (Sherman and Brophy, 2005). Nonmyelinating Schwann cells ensheath small-diameter axons (<1 μm), whereas myelination takes place in Schwann cells wrapped around large-diameter axons. In myelinated fibers, the voltage-gated Na+ channels are concentrated at nodes of Ranvier between adjacent Schwann cells, allowing the rapid saltatory conduction of action potentials. Myelination requires a complex series of interactions between myelinating Schwann cells and axons, which results in highly differentiated domains along the axon (Arroyo and Scherer, 2000). These domains are centered by nodes of Ranvier, which are flanked on either sides by paranodal junctions and juxtaparanodal regions. Paranodal junctions separate Na+ channels at the node and shaker-type K+ channels at the juxtaparanode and function as barriers to restrict the lateral diffusion of axonal membrane proteins and extracellular molecules between the node and the internodal space (Poliak and Peles, 2003). Although a number of molecules involved in axoglial contacts have been identified during the past few years (Girault and Peles, 2002; Poliak and Peles, 2003; Salzer, 2003), their organization remains poorly understood. Because of its proposed role in cell contacts, schwannomin is an interesting candidate for playing a role in the organization of axoglial contacts. Schwannomin-like immunoreactivity has been reported in paranodal regions and Schmidt–Lanterman incisures (SLIs) (Scherer and Gutmann, 1996), which are cytoplasmic channels interrupting the compact myelin. In neurons, schwannomin can associate with the cytoplasmic tail of the paranodal protein paranodin/Caspr (Denisenko-Nehrbass et al., 2003), whereas in Schwann cells, it is localized to the plasma membrane through a paxillin-mediated interaction with β1-integrin (Obremski et al., 1998; Fernandez-Valle et al., 2002).
To address the function of schwannomin in vivo, we used mouse lines with targeted overexpression of mutated schwannomin in Schwann cells or bearing a conditional deletion of the Nf2 gene in Schwann cells. Our results indicate an important role of schwannomin in the organization of Schwann cell contacts and provide new clues about its function.
Materials and Methods
Transgenic mice overexpressing mutated forms of human schwannomin under the control of the P0 promoter were obtained as described previously (Giovannini et al., 1999). In P0–SCH–Δ39–121 mice (line 27), exons 2 and 3, which code for amino acid residues 39–121 within the FERM domain of isoform 1, have been deleted (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). This deletion corresponds to a known mutation in Nf2 patients. In P0–SCH–ΔCter mice (line 3), the C-terminal residues beyond amino acid 314 were deleted (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). Conditional deletion of schwannomin in Schwann cells was obtained as described previously (Giovannini et al., 2000) by crossing two lines of transgenic mice: one in which exon 2 of Nf2 was flanked by LoxP sequences (supplemental Fig. 1, available at www.jneurosci.org as supplemental material), and the other expressing the Cre recombinase under the control of the P0 promoter (P0–CreB). The animals used for experiments (P0CreB; Nf2flox2/flox2) had two floxed Nf2 alleles and were hemizygous for P0–CreB. These animals are referred as P0Cre;Nf2flox2/flox2 mice.
Electron microscopy and morphometry.
Mice were anesthetized with pentobarbital and perfused with 0.9% NaCl, followed by 4% paraformaldehyde and 3% glutaraldehyde in 0.1 m phosphate buffer (PB). The sciatic and phrenic nerves were removed and placed in fresh fixative overnight at 4°C, rinsed in PB, postfixed in 2% OsO4 in PB, dehydrated in an ascending series of ethanol, and embedded in epoxy resin. Semithin sections (0.5 μm) were stained with toluidine blue and viewed with a Leica DMRAX light microscope. Morphometric analyses were performed on 0.5 μm semithin transversal sections from phrenic nerves of three different mice of each phenotype. G-ratios were calculated as ratios of internal to external perimeters of the myelin sheath, measured automatically after segmentation, in >100 fibers per section (Michailov et al., 2004). Ultrastructural studies were performed on transversal sections of the phrenic nerves and on longitudinal sections of the sciatic nerves. Ultrathin sections (40 nm) were cut, stained with Reynold's lead citrate and uranyl acetate, and viewed with a Philips CM-120 TEM electron microscope. Ten Ranvier nodes, from at least two different 12-month-old mice of each phenotype, were observed, and three paranodal regions were selected from 3- and 6-month-old mice. The size of 10 paranodal loops from 10 heminodes was evaluated by measuring the length of membrane of each loop in contact with the axonal membrane.
Rabbit antibodies against paranodin (L51), Caspr2, and syndecan 4 have been described previously (Menegoz et al., 1997; Denisenko-Nehrbass et al., 2003; Goutebroze et al., 2003). The anti-neurofascin antibody reacting with NF155 and NF186 was generated by immunizing rabbits with the common intracellular region of the protein (residues 1065–1175; GenBank accession number AY061639) fused to glutathione S-transferase. The other antibodies were from the following sources: NF2 rabbit polyclonal antibodies (immunoblotting, A-19 sc-331; Santa Cruz Biotechnology); voltage-gated Na+ channel α subunit mouse monoclonal antibody (PAN Nav, clone K58/35); polyclonal β-catenin antibody (Sigma-Aldrich); Kv1.1 α subunit monoclonal antibody (clone K20/78; Millipore); neuronal class III β-tubulin (TUJ1) and nonphosphorylated neurofilament H (SMI-32) monoclonal antibodies (Covance Research Products); FITC-conjugated sheep anti-rabbit antibodies (Eurobio); cyanine 3 (Cy3)-conjugated goat anti-mouse antibodies (Invitrogen); and IRDye800CW-conjugated donkey anti-mouse and anti-rabbit antibodies (Rockland Immunochemicals).
Immunofluorescence and quantitative studies.
Immunostaining of cryostat sections (10 μm) or teased fibers of sciatic nerves was performed as described previously (Goutebroze et al., 2003). Images were acquired using a Leica epifluorescence microscope equipped with a CCD camera (Micromax; Roper Scientific), or a Leica SP2 confocal laser scanning microscope. For quantification of internodal length, nodes were identified by voltage-gated Na+ channel α subunit and paranodin immunostaining on teased fibers. The diameter of individual fibers was measured after immunolabeling with the Cy3-conjugated goat anti-mouse antibodies that underlined fibers surface. To determine the distance between Schmidt–Lanterman incisures, teased fibers were labeled with an antibody against β-catenin. Distances between nodes or incisures and axon diameters were measured using MetaMorph software (Molecular Devices). Three-dimensional reconstructions from confocal optical sections (0.162 μm apart) and surface rendering techniques to visualize immunoreactivity as solid objects were done using Imaris 4.0 software (Bitplane).
Lysates preparation and immunoblotting.
Sciatic nerves were dissected out and homogenized in a Dounce vessel containing 200 μl of a lysis buffer containing 150 mm NaCl, 50 mm Tris, pH 8, 0.5% deoxycholate, 0.1% SDS, 1% NP-40, and Complete proteases inhibitors (Roche Diagnostics). Homogenates were centrifuged 15 min at 4°C at 20,000 × g, and protein concentration in the supernatants were determined by the bicincholinic acid method (Sigma-Aldrich). Equal amounts of protein (20 μg) were fractionated by SDS-PAGE and transferred to nitrocellulose. Membranes were incubated with primary antibodies followed by appropriate IRDye-conjugated secondary antibodies, and developed and quantified using Odyssey (LI-COR Biosciences).
In mice overexpressing mutated schwannomin in Schwann cells, myelin sheaths are preserved but nonmyelinated fibers are altered
To investigate the role of schwannomin in peripheral nerves, we studied two transgenic mouse lines that overexpress in Schwann cells mutated forms of human schwannomin deleted of either amino acid residues 39–121 within the FERM domain (P0–SCH–Δ39–121 mice) or the C-terminal residues beyond amino acid 314 (P0–SCH–ΔCter mice) (Giovannini et al., 1999) (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). In semithin transverse sections of phrenic nerves at low magnification at various ages (3, 6, and 12 months), the endoneurial space appeared to be increased in P0–SCH–Δ39–121 mice in which fibers often had an irregular shape (supplemental Fig. 2, available at www.jneurosci.org as supplemental material). However, no dramatic alteration of the number of fibers was found (Table 1). Myelination appeared globally normal, although myelin thickness was marginally increased in P0–SCH–ΔCter mice, as indicated by a slight reduction of the G-ratio (Table 1).
Electron microscopy study of transverse sections of phrenic nerve revealed several abnormalities (Fig. 1). In register with increased endoneurial space, collagen pockets between fibers were observed in P0–SCH–Δ39–121 mice but not in P0–SCH–ΔCter or wild-type mice (data not shown). In wild-type mice, groups of unmyelinated fibers were regularly surrounded by a single Schwann cell process (Remak fibers) (Fig. 1A,B), whereas in P0–SCH–Δ39–121 mice, Schwann cell processes often made several turns around small-caliber axons (Fig. 1C,D). These abnormalities were not found in P0–SCH–ΔCter mice, although some features of aberrant myelination were occasionally present, such as partially undefasciculated bundles of unmyelinated axons surrounded by the same Schwann cell process (Fig. 1E,F). The observation of abnormal wrapping of nonmyelinated fibers indicated that regulation of autotypic cell contacts might be altered in the presence of mutated schwannomin. However, such abnormalities did not affect significantly the formation of myelin sheath.
Overexpression of mutated schwannomin in Schwann cells alters the organization of paranodal and juxtaparanodal proteins
We examined the role of schwannomin in interactions between Schwann cells and axons, by studying the distribution of proteins enriched at nodes of Ranvier and paranodal junctions, and playing crucial roles in their integrity. In longitudinal sections of sciatic nerves from P0–SCH–Δ39–121 and P0–SCH–ΔCter mice, nodal axonal proteins, Na+ channels (Fig. 2 Aa–Af, arrows), and ankyrin G (Kordeli et al., 1990 and data not shown) appeared normally clustered, as in wild-type mice. The distribution of proteins enriched in Schwann cells microvilli that surround the axon at nodes of Ranvier (Melendez-Vasquez et al., 2001; Goutebroze et al., 2003), including syndecan 4 (Fig. 2 Aa–Ac), syndecan 3, and ERM proteins (data not shown), was also indistinguishable in mutant and wild-type mice. In contrast, the localization of the paranodal protein paranodin/Caspr was altered in both P0–SCH–Δ39–121 and P0–SCH–ΔCter mice, in sciatic nerve sections (data not shown), and in teased fibers (Fig. 2 Ba–Bc). The labeling was irregular with areas of intense immunoreactivity alternating with regions of low or absent staining. The limits of paranodin distribution, which normally ends rather abruptly at the borders of the paranodal junctions (Fig. 2Ba, arrowheads), with the exception of the labeling along the mesaxon (Fig. 2Ba, arrows), were less regular than in wild-type mice and appeared to protrude into the juxtaparanodal regions (Fig. 2Bb,Bc). Quantification of the number of paranodes with altered paranodin immunoreactivity revealed that the proportion of abnormal labeling tended to be higher in large fibers (58% in P0–SCH–ΔCter and 76% in P0–SCH–Δ39–121) than in small fibers (36 and 61%, respectively). We also examined the distribution of neurofascin, a cell adhesion molecule localized in the axon at nodes (NF186 isoform) and in glial loops at paranodes (NF155 isoform) (Tait et al., 2000). Alterations in the localization of paranodal neurofascin in teased fibers of sciatic nerves from P0–SCH–Δ39–121 and P0–SCH–ΔCter mice were similar to those observed for paranodin (Fig. 2 Bd–Bf). Three-dimensional reconstructions of paranodin immunolabeling clearly showed its irregular and fragmented organization in mutant mice (Fig. 2C).
We then examined the distribution of Caspr2, a protein normally enriched in juxtaparanodes (Poliak et al., 2001), which are differentiated regions of internodes flanking the paranodes. Caspr2 distribution was altered, and its intensity appeared diminished in both P0–SCH–Δ39–121 and P0–SCH–ΔCter mice compared with wild-type mice (Fig. 2 Ad–Af,Bg–Bi). Areas of strong labeling alternated with almost negative patches, and the labeled regions were usually smaller and less intensely stained than in wild-type mice. At many nodes, Caspr2 labeling extended into the paranodal regions and came in contact of nodal markers (Fig. 2Ae,Af). Similar abnormalities were observed for Kv1.1 shaker-type K+ channels, which are normally enriched in the juxtaparanodal axonal membrane (Wang et al., 1993) (Fig. 2 Bj–Bl). The distribution of paranodal and juxtaparanodal proteins displayed the same abnormalities in 3- and 6-month-old mutant mice as in 1-year old animals (data not shown), indicating that these perturbations were not age dependent.
We evaluated the levels of expression of the endogenous and mutated forms of schwannomin in sciatic nerves extracts from wild-type and transgenic mice. Endogenous schwannomin expression was not diminished in mutant mice, and the levels of expression of mutated schwannomin were higher than those of the endogenous protein in both P0–SCH–Δ39–121 and in P0–SCH–ΔCter mice (Fig. 2D). To determine whether the altered immunoreactivity of paranodal and juxtaparanodal proteins resulted from a reduction in their levels of expression, we then examined the levels of paranodin/Caspr and Caspr2 in sciatic nerve extracts, using β-tubulin as a control for loading (Fig. 2E). The levels of paranodin and Caspr2 were virtually unchanged in P0–SCH–Δ39–121 and were slightly increased in P0–SCH–ΔCter mice (Fig. 2E). Thus, the abnormalities in the distribution of the paranodal and juxtaparanodal proteins could be related directly to the overexpression of the mutated forms of schwannomin.
Overexpression of mutated schwannomin in Schwann cells alters the ultrastructure of paranodal regions
We then used electron microscopy to determine whether the abnormalities observed by immunofluorescence corresponded to alterations in the ultrastructure of axoglial contacts. Ten nodes of Ranvier were examined in longitudinal sections of sciatic nerves from two 12-month-old mice of each genotype. In wild-type mice, the paranodal myelin loops were tightly and regularly positioned at paranodes, with a symmetrical organization across the heminode (Fig. 3A), and transverse bands, the ultrastructural hallmark of paranodal junctions, were clearly visible (Fig. 3B, arrowheads). In contrast, abnormal paranodal loops and interactions between axons and Schwann cells were frequently observed in mutant mice. Paranodal myelin loops were often asymmetrical in P0–SCH–Δ39–121 and P0–SCH–ΔCter mice with variable width, appearing either enlarged or shrunk from place to place (Fig. 3 C–F). Enlarged loops, generally located close to the juxtaparanode (Fig. 3C, arrow), were observed in 33 and 46% of the heminodes in P0–SCH–Δ39–121 and P0–SCH–ΔCter mice, respectively. Loop width was more variable in the two lines of mutant mice than in wild-type mice (Fig. 4). Electron-dense transverse bands were visible at the level of the normal loops (Fig. 3C, inset, arrowheads) but were absent from those that were massively enlarged (Fig. 3C, arrow). Atrophic loops, filled with electron-dense material and devoid of transverse bands, were observed in 77% of the heminodes examined in P0–SCH–Δ39–121 mice but not in wild-type or P0–SCH–ΔCter mice (Fig. 3D). Formations similar to axon Schwann cell networks (Gatzinsky et al., 1991), in which axonal digitations protruded within large glial formations, were observed in mutant (Fig. 3E,F, arrows) but not in wild-type mice. Other features of mutant paranodes included protrusions of Schwann cell cytoplasm into the axon (P0–SCH–Δ39–121, 22% of the heminodes; P0–SCH–ΔCter, 14% of the heminodes) or axonal protrusions, bulging out of the axon limits into the cells (P0–SCH–Δ39–121, 22% of the heminodes; P0–SCH–ΔCter, 36% of the heminodes). Overall, only 22% of the heminodes were devoid of the features described above in P0–SCH–Δ39–121 and 28% in P0–SCH–ΔCter mice, whereas they were not observed in heminodes from matched wild-type mice. Similar features, in approximately the same proportions, were found in sciatic nerves from 3- and 6-month old mice, indicating that they were not secondary to a late degeneration of fibers.
Conditional deletion of schwannomin in Schwann cells alters axoglial contacts
The results obtained in transgenic mice demonstrated that the targeted overexpression of mutated schwannomin in Schwann cells resulted in a marked disorganization of axoglial contacts. Importantly, many alterations found in P0–SCH–Δ39–121 mice were also observed in P0–SCH–ΔCter transgenic mice, albeit to a slightly lesser degree. Such alterations could theoretically result from a gain of function of mutated proteins or from a partial or complete loss of function with a dominant-negative effect. The similar phenotype of the two different mutations argued in favor of the second hypothesis. To distinguish more precisely between these two hypotheses, we used mutant mice in which NF2 exon 2 was conditionally deleted in Schwann cells (P0Cre;Nf2flox2/flox2) (supplemental Fig. 1, available at www.jneurosci.org as supplemental material). The line we used (P0–CreB) had no major postnatal lethality, in contrast with the initially reported P0–CreA line (Giovannini et al., 2000), allowing the study of the peripheral nervous system in adult mice. In these mice, a large proportion of Schwann cells is devoid of schwannomin, because deletion of exon 2 destabilizes the molecule (Giovannini et al., 2000). However, the total amount of schwannomin in sciatic nerve homogenates was not decreased (Fig. 2D) because the deletion occurred in Schwann cells, whereas the protein is also expressed in other cell types, including axons (Denisenko-Nehrbass et al., 2003).
In P0CreB;Nf2flox2/flox2 mice, analysis of semithin sections of sciatic nerve (Fig. 5A) showed a slight decrease (−12%) in the number of myelinated fibers (Table 1) compared with wild-type mice. The myelin thickness was slightly increased as indicated by a minor decrease in the G-ratio (Table 1). Electron microscopy of phrenic nerve sections showed some indirect evidence of axonal loss in P0CreB;Nf2flox2/flox2 mice, including collagen pockets in the endoneurial space, fibers, and collapsed Schwann cells (Fig. 5B, asterisk, arrow, and arrowhead, respectively). Abnormal Schwann cell wrapping was evidenced by small axons surrounded by multiple wraps of nonmyelinating Schwann cells (Fig. 5C, arrow) and by bundles of undefasciculated small axons wrapped by a myelinating Schwann cell (data not shown). Thus, these mice displayed signs of moderate axonal loss and aberrant myelination.
Electron microscopic study of longitudinal sections of sciatic nerves, at the level of nodes of Ranvier, revealed abnormalities of paranodes. The size of adjacent loops was irregular (Fig. 4). Atrophic paranodal loops (22%) and axonal protrusions (11%), sometimes with electron-dense material, were also present (Fig. 5 D–F). Transverse bands were not visible at the level of these altered paranodal loops (Fig. 5E,F). Enlarged paranodal loops and axon Schwann cell networks containing mitochondria and other vesicular material were frequently observed (44% of the heminodes), generally located close to the juxtaparanode (Fig. 5G) (data not shown).
We examined molecular markers of nodal regions in sections and teased fibers of sciatic nerves from P0CreB;Nf2flox2/flox2 mice. There was no consistent difference in the distribution of proteins enriched in the nodal axon, including sodium channels (Fig. 6 Aa–Ad) and proteins of glial microvilli syndecan 3 (data not shown) and 4 (Fig. 6Ac,Ad). In contrast, the distribution of the paranodal and juxtaparanodal markers was severely altered (Fig. 6B). Paranodin/caspr immunoreactivity was distributed irregularly, with a combination of areas of normal and decreased labeling, often not delineating paranodal regions and invading the nodal or juxtaparanodal spaces (Fig. 6Ba,Bb). We noticed that paranodin immunoreactivity was more frequently altered in large fibers than in small ones. The three-dimensional reconstructions further indicated the overall alterations of paranodin distribution (Fig. 6C). The labeling of paranodal neurofascin was less regular in P0Cre;Nf2flox2/flox2 than in wild-type mice, and its overall intensity was dramatically reduced (Fig. 6B), as confirmed by quantification (wild type, 61 ± 5; mutant, 28 ± 3; mean ± SEM, arbitrary units; n = 5; t test, p < 0.00035). Caspr2 immunoreactivity was also decreased, irregular, and abnormally located, often in contact with the nodal proteins (Fig. 6B). Kv1.1 immunoreactivity was similarly altered (data not shown). However, immunoblotting experiments on sciatic nerve extracts showed that the levels of paranodin as well as those of Caspr2 were virtually unchanged in P0Cre;Nf2flox2/flox2 mice (Fig. 6D), indicating that the alterations of immunoreactivity of the paranodal and juxtaparanodal proteins did not correspond to changes in their levels of expression. Thus, mice bearing a targeted deletion of schwannomin in Schwann cells displayed severe alterations of the organization of axoglial contacts.
Mutation of schwannomin in Schwann cells alters the number of Schmidt–Lanterman incisures, the internodal length, and the number of Schwann cells
The compact myelin in Schwann cells is regularly interrupted by SLIs, which are helicoidal cytoplasmic channels thought to facilitate the communication between the cytoplasmic compartments on adaxonal and abaxonal sides of the myelin sheath. These incisures are characterized by the presence of Schwann cell autotypic contacts, including adherens junctions, tight junctions, and gap junctions, containing cadherins, claudins, and connexins, respectively (Arroyo and Scherer, 2000). On transverse sections, SLIs can appear as an empty space between double concentric myelin rings (Robertson, 1958). While studying transverse sections of phrenic nerves of P0Cre;Nf2flox2/flox2 mice, we noticed that myelin sheaths appeared more frequently as double concentric myelin rings than in wild-type mice (Fig. 5A), suggesting an increased number of SLIs. We counted the number of fibers with double concentric rings on transverse phrenic nerve sections (Table 1). Their number was significantly higher in P0–SCH–Δ39–121 and P0Cre;Nf2flox2/flox2 mice compared with wild-type or P0–SCH–ΔCter mutant mice (Table 1). This increased number of double concentric rings was observed in 3-, 6-, and 12-month-old mutant mice (data not shown). To quantify precisely the number of SLIs per Schwann cells in mutant mice, we measured internodal lengths and distances between adjacent SLIs (visualized by β-catenin immunostaining) on teased sciatic fibers (Fig. 7A). A characteristic feature of two of the mutant mouse lines was an increased proportion of very short internodes (<350 μm), which was significantly higher in P0–SCH–Δ39–121 transgenic mice (22%) and P0Cre;Nf2flox2/flox2 mice (63%) than in P0–SCH–ΔCter (0%) or wild-type (4%; χ2 = 35.7; p < 0.0001) mice. The average internodal length was not significantly different between wild-type and P0–SCH–ΔCter and P0–SCH–Δ39–121 transgenic mice but was decreased in P0Cre;Nf2flox2/flox2 mice (Fig. 7B). Because a correlation between the caliber of a myelinated axon and the internodal length is well established (Friede, 1983), we normalized internodal distances to axon caliber. Although the average fiber diameter was smaller in P0Cre;Nf2flox2/flox2 than in wild-type mice, the internodal distance normalized to axonal diameter was significantly decreased (Fig. 7C,D). The distance between SLIs was also decreased in mutant mice, the diminution being more pronounced in P0–SCH–Δ39–121 and P0Cre;Nf2flox2/flox2 than in P0–SCH–ΔCter mice (Fig. 7E), confirming the increased number of SLIs in these mice.
To determine whether an increased proportion of very short internodes and the decreased distance between SLIs observed in P0–SCH–Δ39–121 mice could be associated with an increased proliferation of Schwann cells at the early onset of myelination, we counted Schwann cell nuclei along teased sciatic nerves derived from postnatal day 1 wild-type, P0–SCH–Δ39–121, and P0–SCH–ΔCter mice (Fig. 7F,G). At the early onset of the myelination, the number of Schwann cells along the nerve fibers was higher in P0–SCH–Δ39–121 mice than in wild-type or P0–SCH–ΔCter mice (Fig. 7F,G). Importantly, the number of Schwann cells was not altered in P0–SCH–ΔCter mice. We also looked for alterations in the number of Schwann cells in P0Cre;Nf2flox2/flox2 using 3-month-old mice (Fig. 7H,I). At this age, the cell density was much lower than in neonates. In wild-type mice, the Schwann cell nuclei were isolated and never in close vicinity of the node. In contrast, in P0Cre;Nf2flox2/flox2 mice, we often observed pairs of nuclei, close to the nodal region (Fig. 7H, right panel, arrows). Cell counts showed that the number of nuclei per length of fiber was significantly increased (Fig. 7I). These observations clearly supported the hypothesis of an increased number of Schwann cells during myelination in P0–SCH–Δ39–12 and P0Cre;Nf2flox2/flox2 but not in P0–SCH–ΔCter mice.
We also measured, by immunoblotting, the amounts of β-catenin and NF155, the 155 kDa glial isoform of neurofascin, two proteins enriched in SLIs (Fannon et al., 1995; Chang et al., 2000). The levels of these two proteins were increased in sciatic nerve extracts from the three lines of mutant mice compared with wild-type mice (Fig. 7J). Altogether, these results show that the number of SLIs is altered by mutations in schwannomin and that overexpression of P0–SCH–Δ39–121 or conditional knock-out of schwannomin decreases the internodal length, i.e., increases the number of myelinating Schwann cell per length unit of axon.
Our observations in mice overexpressing specifically mutated schwannomin in Schwann cells, under the control of P0 promoter, demonstrate an important role of this protein in the organization of axoglial interactions and Schwann cell autotypic contacts at the level of SLIs. It was shown previously that mice overexpressing SCH–Δ39–121, but not those overexpressing SCH–ΔCter, display Schwann cell hyperplasia and develop tumors (Giovannini et al., 1999). These findings were confirmed in mice with a selective deletion of schwannomin in Schwann cells (Giovannini et al., 2000). These previous observations, as well as those from other mutant mice (McClatchey et al., 1998) and a number of studies in cells in culture (McClatchey and Giovannini, 2005), demonstrated the tumor suppressor activity of schwannomin, providing a basis for the phenotype of patients with NF2. The present study shows that schwannomin is also important for the organization of myelinated fibers.
The phenotype of schwannomin mutant mice included some abnormalities of nonmyelinated fibers, which may be attributable to the expression of P0 promoter in nonmyelinating Schwann cells (Zhang et al., 1995). A partial deficit in defasciculation of axon bundles by Schwann cells was observed in schwannomin mutants. Interestingly, it has been shown recently that focal adhesion kinase is involved in defasciculation (Grove et al., 2007), suggesting a possible link with schwannomin that is associated with focal adhesion kinase (FAK) and β1-integrins (Taylor et al., 2003) and can alter FAK activity (Poulikakos et al., 2006). Several other features suggested that Schwann cell membrane interactions were altered in schwannomin mutant mice. Nonmyelinating Schwann cell extension wrapped several fold around small axons in P0–SCH–Δ39–121 and P0Cre;Nf2flox2/flox2 mice, instead of stopping after a single turn. Exuberant myelination of large-caliber fibers was reported in a different line of P0Cre;Nf2flox2/flox2 mice, expressing higher levels of Cre than the line used in this study (Giovannini et al., 1999), and we observed a relative increase in myelin thickness in P0–SCH–ΔCter and P0Cre;Nf2flox2/flox2 mice. These findings suggest that the inhibitory signals that normally stop the extension of Schwann cell lamellipodia are defective when schwannomin is mutated or deleted. This could be a consequence of the ability of schwannomin to inhibit Pak1-mediated recruitment of Rac to the plasma membrane and matrix adhesions (Shaw et al., 2001; Morrison et al., 2007; Okada et al., 2007). Schwannomin may also have a regulatory role through its interactions with ErbB2 (Fernandez-Valle et al., 2002), a receptor for the axonal factor neuregulin-1, critical for the control of myelin thickness (Garratt et al., 2000; Michailov et al., 2004). Additional signs of altered axon–Schwann cell interactions included the frequent protrusions originating from one cell type into the other, including the formation of axon–Schwann cell networks. These complex interactions between axons and Schwann cells have been described in paranodal regions of other species (Gatzinsky et al., 1991). It is striking that we did not observe them in sciatic nerves of wild-type mice but only in mutant animals. Thus, it is tempting to suggest that their frequent occurrence also resulted from a deficient regulation of Schwann cell membrane expansion.
In contrast with the exuberant axoglial interactions discussed above, we also observed atrophic paranodal loops containing electron-dense material. Thus, a distinctive feature of schwannomin mutant mice was the coexistence of apparently normal loops and loops with an increased size and/or atrophic loops. In these latter two types, transverse bands that are characteristic for intact axoglial contacts were usually not present. The combination of regions of normal and abnormal axoglial junctions provided the basis for the irregular distribution of juxtaparanodal and paranodal proteins, and their extension toward nodes and internodes. This phenotype was different from that described previously in other mutant mice in which the transverse bands were completely absent (Dupree et al., 1999; Bhat et al., 2001; Boyle et al., 2001; Mathis et al., 2001). In these mice, paranodal markers were diffusely distributed along the axon, and juxtaparanodal markers spread through paranodes to the vicinity of nodal markers. These observations provide evidence for an important role of schwannomin in the organization and stability of Schwann cells membrane domains. This role may be related to the partitioning of schwannomin in lipid rafts (Stickney et al., 2004), which are important for the segregation of membrane proteins and organization of axoglial contacts (Schafer et al., 2004).
An increased number of incisures was found in the three Nf2 mutant genotypes, although it was much more striking in P0Cre;Nf2flox2/flox2 and, to a lesser degree, P0–SCH–Δ39–121 mice than in P0–SCH–ΔCter mice. A similar increased number of incisures was observed in mutant mice with altered Schwann cell–axon interactions, including myelin basic protein-deficient shiverer mice (Gould et al., 1995), sulfatide-deficient mice (Hoshi et al., 2007), and desert hedgehog-null mice (Sharghi-Namini et al., 2006). In contrast to desert hedgehog-null mice, both shiverer (Sharghi-Namini et al., 2006) and schwannomin mutant mice displayed an increase in β-catenin and NF155 levels, indicating that these various mutations altered SLI numbers through different mechanisms. Little is known about the regulation of the formation of SLIs in which cadherin-mediated autotypic interactions are a key component (Tricaud et al., 2005; Perrin-Tricaud et al., 2007). Schwannomin immunoreactivity has been reported in SLIs (Scherer and Gutmann, 1996), suggesting that alteration of SLI numbers in schwannomin mutant mice may result from the role of schwannomin in stabilizing cadherin-based adherens junctions (Lallemand et al., 2003). This may involve the associated protein erbin (Rangwala et al., 2005) and Rac1 (Tricaud et al., 2005). In support of this hypothesis, β-catenin appears to stabilize SLIs through an inhibition of Rac1 (Tricaud et al., 2005) and schwannomin is capable of suppressing the recruitment of Rac1 to the plasma membrane (Okada et al., 2005).
Another conspicuous feature observed in schwannomin mutant mice was the shortening of internodal distances. The shortening of internodal distance was dramatic in P0Cre;Nf2flox2/flox2 mice, and an increased number of very short internodes was also found in P0–SCH–Δ39–121 mice. We found that reduced axon caliber could not account for these short internodes. Conversely, cell proliferation and apoptosis are known to control Schwann cell number in the developing peripheral nervous system (Jessen and Mirsky, 2005). Our observation of an increased number of Schwann cells aligned along axons before the onset of myelination in P0–SCH–Δ39–121 but not P0–SCH–ΔCter mutant mice provides an explanation for the abnormally short internodes. Moreover, an increased number of Schwann cells was found in P0Cre;Nf2flox2/flox2 mice. These alterations are likely to result from the impaired control of cell cycle in the absence of schwannomin (McClatchey et al., 1998; Giovannini et al., 1999, 2000). A similar mechanism has been proposed to explain the existence of short internodes in young mice deficient for PMP22, which is also implicated in cell proliferation and apoptosis (Neuberg et al., 1999). In support of this hypothesis, short internodes were not observed in P0–SCH–ΔCter mice that displayed no tendency to increased Schwann cell proliferation (Giovannini et al., 1999).
The combination of the three types of mutant mice used in the present study allows a better understanding of the function of schwannomin. Most of the findings in P0–SCH–Δ39–121 and P0–SCH–ΔCter mice that overexpressed a mutated protein were qualitatively similar to those in P0Cre;Nf2flox2/flox2 that had a dramatic decrease in its expression levels (McClatchey et al., 1998; Giovannini et al., 1999, 2000). This indicated that the phenotype was attributable to a loss of function of the normal Nf2 gene product, because of its absence or a dominant-negative effect of overexpressed mutated protein. Moreover, it is striking that the tumorigenic (SCH–Δ39–121) and nontumorigenic (SCH–ΔCter) mutations had similar phenotypes, clearly dissociating the dual role of schwannomin in cell growth control and in cell–cell interactions.
In addition, our results provide novel clues about the pathophysiology of the human familial NF2 disease, raising the possibility that peripheral nerves might be altered besides the occurrence of tumors. Our results suggest that some schwannomin mutations, even at the hemizygous stage, before the occurrence of tumors, may have a dominant-negative effect. It is also possible that a gene dosage effect occurs, especially taking into account the long lifespan of humans compared with mice. Neuropathies have been reported in NF2 patients that may not be all accounted for by microcompressions (Sperfeld et al., 2002). Our results suggest that primary alterations in Schwann cell biology may also play a role in these neuropathies and provide an incentive for further exploring this issue, which may have diagnostic and therapeutic implications.
This work was supported in part by grants of the National Multiple Sclerosis Society, Association pour la Recherche sur la Sclerose en Plaque, Association Française contre les Myopathies, Action Concertée Incitative Développement et Physiologie, Association pour la Recherche sur le Cancer, Foundation Schlumberger pour l'enseignement et la Recherche, and Agence Nationale de la Recherche-Neuro-05-NEUR-A05158DS. We are indebted to P. Ezan, A. Boisquillon, C. Fayet, P. Bozin, and E. Valjent for their help with some experiments.
- Correspondence should be addressed to Jean-Antoine Girault, Inserm, Unité Mixte de Recherche en Santé 839, Institut du Fer à Moulin, 17 rue du Fer à Moulin, Paris, 75005, France.