Coordinated internodal and paranodal adhesion controls accurate myelination by oligodendrocytes

Myelin is a multilamellar membrane sheath produced by two types of glial cells: oligodendrocytes in the central nervous system (CNS) and Schwann cells in the peripheral nerve system (PNS). It allows fast and efficient propagation of electrical impulses along axons, thereby modulating various neuronal circuits (McKenzie et al., 2014; Etxeberria et al., 2016). During development, several cell adhesion molecules (CAMs) mediate a complex set of dynamic intercellular interactions between myelinating glia and their underlying axons (Pan and Chan, 2017). These axoglial systems are thought to be involved in various stages in the formation of the myelin sheath, including axonal selection, membrane ensheathment and wrapping, longitudinal growth of the myelin unit, and segregation of the axonal membrane into distinct domains, which is required for saltatory conduction (Barros et al., 2009; Snaidero et al., 2014; Rasband and Peles, 2015; Elazar et al., 2019). However, oligodendrocytes and their underlying axons express a relatively large number of CAMs (Zhang et al., 2014; Sharma et al., 2015), making it challenging to decipher the role specific molecules have in myelination. Furthermore, it is presently unclear whether the function of axoglial CAMs in CNS myelination is at all necessary, as oligodendrocytes are able to ensheath nanofibers, which lack specific axonal adhesive molecules (Lee et al., 2012; Bechler et al., 2015).

Distinct axoglial adhesion complexes are located along the internodes, which reside beneath the compact myelin, and at the paranodal junctions that are formed between the terminal loops of the myelin sheath and the axon. At the paranodal junction, intercellular contact is mediated by glial neurofascin 155 (NF155) and a neuronal complex of Caspr and contactin (Einheber et al., 1997; Menegoz et al., 1997; Rios et al., 2000; Tait et al., 2000; Charles et al., 2002). During myelination, Caspr and NF155 are first clustered at initial contact sites between oligodendrocyte processes and the axon, but then redistribute from the internodes to the forming paranodal junction in a spiral pattern that corresponds to each turn of the myelin sheath (Eisenbach et al., 2009; Pedraza et al., 2009). While the absence of these molecules results in the detachment of the paranodal loops from the axon, it did not affect myelin membrane ensheathment (Bhat et al., 2001; Zonta et al., 2008), suggesting that axoglial contact at the paranodes is not required for myelination. Along the internode, the interaction between myelinating glial cells and the axon is mediated by binding of glial cell adhesion molecule 4 (Cadm4), also known as synaptic CAM (SynCAM) or Nectin-like (Necl) adhesion molecule, to axonal Cadm2 and Cadm3 (Maurel et al., 2007; Spiegel et al., 2007). However, oligodendrocytes form normal myelin in mice lacking either the Cadm4 (Zhu et al., 2013; Elazar et al., 2019) or Cadm3 (Park et al., 2008; unpublished data) genes, suggesting that Cadm proteins are dispensable for CNS myelination. These observations may also indicate the presence of compensatory CAMs that mediate redundant axon–glia interaction along the internodes. A potential candidate is myelin-associated glycoprotein (MAG), an immunoglobulin superfamily member that is located with Cadm4 at the adaxonal (inner) membrane facing the axon in the CNS and the PNS (Bartsch et al., 1989; Trapp et al., 1989) and is present on oligodendrocyte processes during the initial stages of ensheathment (Bartsch et al., 1989). In line with a possible functional redundancy between Cadm4 and MAG, genetic deletion of the latter in mice had largely no effect on either PNS or CNS myelination (Li et al., 1994; Montag et al., 1994). Notably, while Cadm4-Cadm3 and NF155-Caspr/contactin complexes represent internodal and paranodal adhesion systems, respectively, MAG may contribute to axoglial adhesion at both domains (Marcus et al., 2002; McGonigal et al., 2019).

We have recently shown that increased expression of the extracellular domain of Cadm4 in oligodendrocytes impaired myelin targeting, membrane wrapping, and lateral extension of the myelin sheath (Elazar et al., 2019), indicating that axon–glia adhesion should be tightly regulated during myelination. Given the presence of multiple axoglial CAMs, we reasoned that these molecules should have overlapping functions. Here we tested this idea by genetically ablating Cadm4 together with MAG or with Caspr. We found that accurate ensheathment by oligodendrocytes depends on the coordinated action of different axoglial adhesion systems that limit membrane overgrowth during CNS myelination.

To test for a possible functional redundancy between Cadm4 and MAG, we have generated mice lacking both genes (Cadm4^−/−/Mag^−/−) by crossing Cadm4^−/− (Golan et al., 2013) and Mag^−/− (Li et al., 1994) heterozygous mice. The complete absence of Cadm4 and MAG proteins in the progeny was confirmed by Western blot analysis of brain lysates (Fig. 1 A). Double Cadm4^−/−/Mag^−/− mice were viable and had normal life spans but, in contrast to the single Cadm4^−/− or Mag^−/− mutants, exhibited a complex neurological motor deficit. On a horizontal beam, which requires precise foot placing and assesses motor balance and coordination (Luong et al., 2011), Cadm4^−/−/Mag^−/− mice displayed a significantly higher rate of leg slippage from the beam and often dragged their pelvis instead of walking (measured as percentage of pelvis dragging), compared with WT or the single Cadm4^−/− or Mag^−/− mutants (Fig. 1, B–D). Catwalk analysis used to assess motor function and coordination revealed that Cadm4^−/−/Mag^−/− mice had abnormal interpaw coordination between several paw combinations, most notably between the right front and left hind paws (Fig. 1 E). Similarly, double Cadm4^−/−/Mag^−/− mice fell more rapidly from a rotarod compared with WT or the single mutants (Fig. 1 F). In contrast to the prominent motor coordination phenotype, when placed in an open field, Cadm4^−/−/Mag^−/− mice showed general activity levels and gross locomotor and exploration behavior similar to their littermate controls (not depicted).

To examine the formation of myelin in these mice, we performed immunofluorescence labeling of postnatal day 30 (P30) mice brain sections using an antibody to myelin basic protein (MBP). As depicted in Fig. 1 F, this analysis did not reveal any significant difference in either gross morphology or the presence of myelinated tracts between Cadm4^−/−/MAG^−/− and WT or the single gene deletions (Fig. 1 G). Similarly, EM analysis of the corpus callosum of 3-mo-old mice revealed a slight and insignificant decrease in the number of myelinated axons in Cadm4^−/−/Mag^−/− mice (527 ± 61) compared with WT (637 ± 54), Cadm4^−/− (766 ± 47), or MAG^−/− (638 ± 39) mice (Fig. 1, H and I). Unlike Cadm4^−/− or MAG^−/− mice, mice lacking both genes also exhibit several myelin abnormalities, including myelin outfolds (Fig. 1 J). In contrast, EM analysis of sciatic nerves isolated from 1-mo-old mice demonstrated that peripheral myelin in Cadm4^−/−/Mag^−/− mice was indistinguishable from the other genotypes (Fig. 1, K and L).

To further examine the ultrastructure of the myelin, we performed EM of both cross and longitudinal sections of the corpus callosum of 3-mo-old mice. We found that the most conspicuous abnormality present in Cadm4^−/−/Mag^−/− mice was the appearance of axons surrounded by multiple myelin sheaths (Fig. 2, A–D). These structures contained double, triple, and often quadruple myelin sheaths, which were frequently interrupted by cytoplasmic pockets that are typically present only at the periaxonal space and the paranodal loops. The frequency of these multimyelinated axon profiles was significantly higher in Cadm4^−/−/Mag^−/− mice (7.8 ± 2.8% per field of view) compared with WT (0.1 ± 0.2%), Cadm4^−/− (0.4 ± 0.4%), or Mag^−/− (1.9 ± 1%) mice (Fig. 2 E). In addition, non-multimyelinated axons in Cadm4^−/−/Mag^−/− mice displayed thicker myelin (i.e., lower g-ratio: 0.66 ± 0.07 compared with 0.726 ± 0.07 in WT mice), with 22% of myelinated profiles having a g-ratio <0.65 compared with 5, 8, and 9% in WT, Cadm4^−/−, and Mag^−/− mice, respectively (Fig. 2 F). We next analyzed longitudinal sections of optic nerves isolated from Cadm4^−/−/Mag^−/− mice. We observed paranodal loops that were present on top (Fig. 2 G) or underneath (Fig. 2 H) compact myelin. To further examine the cellular basis for the formation of double myelin, we followed individual axons by performing 3D reconstruction using a focused ion beam scanning EM (FIB-SEM) on optic nerves isolated from adult Cadm4^−/−/Mag^−/− mice. As depicted in Fig. 2 I, we identified axons that were myelinated by two separate sheaths containing paranodal loops that arose from one sheath and attached on top of another (Fig. 2 I, a–d). These results suggest that multimyelinated axons are formed by the longitudinal overgrowth of the myelin lamellae beneath the forming paranodal junction. Developmental EM analysis of corpus callosum sections obtained from Mag^−/− and Cadm4^−/−/Mag^−/− mice revealed that double myelin is already formed during the first three postnatal weeks (Fig. 2, J–M), indicating that they represent an abnormal growth of the myelin sheath during myelination. The increased number of multiple myelin profiles in the absence of both Cadm4 and MAG reveals that these two internodal CAMs play compensatory roles in keeping the inner membrane of oligodendrocytes in register with the axons, as previously suggested for MAG (Li et al., 1994; Montag et al., 1994).

Longitudinal overgrowth of the myelin membrane in Cadm4^−/−/Mag^−/− mice (Fig. 2, G–I), and hence the formation of multiple myelin profiles, may depend on loosening of the attachments of the paranodal loops to the axon, as detected in several paranodal mutant mice (Dupree et al., 1998; Bhat et al., 2001). To begin to clarify how the deletion of two internodal CAMs results in such a phenotype, we labeled mice brain sections and myelinated spinal cord cultures using different combinations of MAG, Cadm4, and Caspr antibodies. We found that in contrast to the PNS (Maurel et al., 2007; Spiegel et al., 2007), MAG, but not Cadm4, is enriched at the paranodes (Fig. S1), suggesting that it may be involved in paranodal adhesion. This idea is in line with previous studies showing that, in addition to mediating axoglial adhesion along the internodes, MAG regulates the formation and integrity of the paranodal junction (Marcus et al., 2002). We reasoned that if the contribution of MAG to the Cadm4^−/−/Mag^−/− phenotype is due to its role in paranodal adhesion, one should expect to obtain a similar phenotype in mice lacking Cadm4 and other paranodal CAMs. To test this possibility, we reduced both internodal and paranodal adhesion by removing Cadm4 and the paranodal protein Caspr (Fig. 3 A). Cadm4^−/−/Caspr^−/− mice were generated by crossing heterozygous Cadm4^−/− (Golan et al., 2013) and Caspr^−/− (Gollan et al., 2003) mice, and the absence of both proteins was assessed by immunolabeling of P45 optic nerves (Fig. 3 B). Cadm4^−/−/Caspr^−/− mice displayed a more severe neurological phenotype than Caspr^−/− or Cadm4^−/−, resulting in premature death starting at P15. In mice surviving >2 wk, the neurological abnormalities deteriorated over time, resulting in complete hindlimb paralysis at P40.

EM analyses of sciatic nerve cross sections obtained from Caspr^−/− and Cadm4^−/−/Caspr^−/− mice at P40 showed the presence of normal myelin profiles in both genotypes (Fig. 3 C). In contrast, analysis of optic nerves isolated from WT and single and double gene deletion demonstrated that Cadm4^−/−/Caspr^−/− exhibited a significant reduction in the density of myelinated axons (414 ± 37) compared with Cadm4^−/− (606 ± 86) and Caspr^−/− (550 ± 44) mutants or WT mice (650 ± 80; Fig. 3, D and E). Consistent with our hypothesis that reduced adhesion at the internode and the paranode will result in myelin overgrowth, Cadm4^−/−/Caspr^−/− mutant mice exhibited multimyelin abnormalities that phenocopied that of mice lacking Cadm4 and MAG (Fig. 4, A–E). Cadm4^−/−/Caspr^−/− mutant mice formed a significantly higher percentage of aberrant myelin, with double-myelinated axons in both the optic nerve (13.8 ± 3.1) and the corpus callosum (9.7 ± 2.2) compared with WT (0 ± 0 in both), Cadm4^−/− (0 ± 0 and 0.32 ± 0.2 in optic nerve and corpus callosum, respectively), and Caspr^−/− (5.2 ± 2.5 and 3.3 ± 0.5) mice (Fig. 4, F and G). Longitudinal sections of WT (Fig. 4 H) and Cadm4^−/−/Caspr^−/− nerves revealed what appeared to be overgrowth of paranodal loops over existing compact myelin (Fig. 4 I), as well as the presence of compact myelin covering paranodal loops that contacted the axon (Fig. 4, J and K). 3D reconstruction of serial sections obtained from Cadm4^−/−/Caspr^−/− optic nerves using FIB-SEM revealed two sheaths growing on top of each other, resulting in two paranodal junctions forming either on top or beneath compact myelin (Fig. 4, L and M, a–f). These multimyelin profiles recapitulate those detected in Cadm4^−/−/Mag^−/− mice, further supporting the interpretation that the combined reduction in paranodal and internodal adhesion results in overgrowth of the myelin sheath. Interestingly, even in the absence of paranodal and internodal axoglial CAMs, the paranodal loops still formed, suggesting that they are being held together by an additional adhesion system.

Our findings reveal that the coordinated action of several axoglial adhesion systems controls the growth of the myelin membrane in the CNS. In the absence of this adhesion, myelin targeting and axonal ensheathment still occurs, but the lateral elongation of the myelin sheath is unreliable, demonstrating that axoglial CAMs limit the lateral overgrowth of oligodendrocyte myelin. According to the prevailing model of CNS myelination (Snaidero et al., 2014), the inner membrane of the myelin sheath moves around the axon underneath previously formed myelin lamellae, which grow laterally. During this process, internodal CAMs mediate intercellular contact between the inner myelin lip and the axon (Bartsch et al., 1989; Elazar et al., 2019), while paranodal CAMs mediate the interaction between the axon and the paranodal loops that are forming at the edge of the laterally growing myelin sheath (Eisenbach et al., 2009; Pedraza et al., 2009). We propose that the synergistic action of these internodal and paranodal CAMs regulates the growth of the myelin sheath. In Cadm4^−/− mice, reduced internodal adhesion does not result in myelin overgrowth due to the presence of paranodal adhesion that attaches the myelin sheath to the axon. In paranodal mutants such as Caspr^−/− (Bhat et al., 2001) or mice lacking ceramide galactosyltransferase (Dupree et al., 1998), paranodal adhesion is impaired, resulting in only limited generation of double myelin profiles due to the presence of internodal adhesion by Cadm4. In contrast, by reducing both internodal and paranodal adhesion in Cadm4^−/−/Caspr^−/− mice, the growth of the inner myelin membrane is not limited to the axons, and it extends beneath the detached paranodal loops and could continue to grow either over or under an adjacent forming myelin segment (Fig. 5). Notably, our data do not exclude the possibility that some of the double myelin profiles observed are not the result of longitudinal overgrowth of the inner lip, but rather are formed by direct oligodendrocyte ensheathment of preexisting myelin. In this case, Cadm4, but likely not MAG (Chong et al., 2012), may inhibit wrapping of other myelin segments. Surprisingly, this does not affect myelin wrapping by Schwann cells, suggesting that in peripheral nerves the basal lamina and additional CAMs present at the Schwann cell microvilli that contact the nodes (Eshed et al., 2005; Bang et al., 2018) are sufficient to keep the growing myelin membrane in register with the axon.

Cadm4^−/−, Mag^−/−, and Caspr^−/− mice were previously described (Li et al., 1994; Gollan et al., 2003; Golan et al., 2013). Cadm4^−/− and Mag^−/− mice were bred on a C57BL/6 background (Jackson Laboratory), while Caspr^−/− mice were bred on an ICR background (Jackson Laboratory). Homozygous Cadm4^−/−/Mag^−/− and Cadm4^−/−/Caspr^−/− mice were obtained by a conventional breeding scheme. Genotypes were determined by performing PCR on genomic DNA extracted from mice tails or sciatic nerves using previously described primers. Genotyping was done by standard PCR using the following primers: Cadm4 WT, 5′-CGC​AGT​TCT​GAT​CCA​GCA​TC-3′ and 5′-ATC​AGA​GGA​CAG​CTC​ATG​AG-3′; Cadm4^−/−, 5′-CGC​AGT​TCT​GAT​CCA​GCA​TC-3′ and 5′-GGC​AGG​AGG​GTT​GCC​ATG​A-3′; Mag^−/− 5′-TGC​CGC​TGT​TTT​GGA​TAA-3′, 5′-CGC​CTC​GGA​AAT​AGT​ATT​TG-3′, and 5′-CTT​GGG​TGG​AGA​GGC​TAT​TC-3′; Caspr WT, 5′-GAG​AGG​GAA​GGG​TGG​ATA​AGG​AC-3′ and 5′-ATT​GCG​GAG​CGC​TGG​GGA​GAG​G-3′; Caspr^−/− 5′-ATT​TCC​CAA​CGG​CAG​GTT-3′ and 5′-TCG​CCT​TCT​TGA​CGA​GTT​C-3′. Both male and female animals were used in the study, with no detectable difference in myelin morphology. Animals were maintained in the animal facility at the Weizmann Institute and housed in a temperature-controlled animal room with a 12-h light/dark cycle. Water and food were available ad libitum. All experiments were performed in compliance with the relevant laws and institutional guidelines and were approved by the Weizmann Institute’s Animal Care and Use Committee.

For motor tests, a total of 42 male and female mice 4–6 mo old were evaluated (10 male WT, 7 male and 3 female Cadm4^−/−, 3 male and 7 female Mag^−/−, and 6 male and 6 female Cadm4^−/−/Mag^−/−). In each behavioral experiment, animals were evaluated one at a time. The behavioral studies were performed in the behavioral Laboratory in the animal facility at the Weizmann Institute. Rotarod was done using a standard rotarod apparatus (Panlab) as previously described (Novak et al., 2011). Coordination motor ability was assessed by training mice to walk on a horizontal beam (35-mm width, 50-cm length) raised 30 cm above the working surface to get back to their home cage. Trials consisted of mice forced to cross a narrow beam (6 mm wide) to get back to their home cages uninterrupted five times. Test sessions were video recorded using an overhead camera and manually scored for time spent on the beam and slips per step percentage, together with groin dragging. Gait was analyzed using a Catwalk gait automatic analyzer (CatWalk; Noldus Information Technology). The position, timing, and dimensions of each footfall were recorded by a video camera positioned underneath a glass plate.

Mice were anesthetized and perfused with 2% PFA/PBS. Brains and optic nerves were isolated and postfixed on ice for 30 min. For immunofluorescence, tissues were incubated in 30% sucrose/PBS at 4°C overnight, embedded in optimal cutting temperature medium, and sectioned (12 µm). Sections were washed with PBS and blocked with PBS containing 5% normal goat serum, 0.5% Triton X-100, and 0.05% sodium azide for 1 h at RT. Samples were incubated overnight at 4°C with primary antibodies diluted in PBS containing 5% normal goat serum, 0.5% Triton X-100, and 0.05% sodium azide; washed three times in PBS; incubated for 1 h with secondary antibodies; washed in PBS; and mounted with Fluoromount-G. Immunohistochemistry of paraffin sections (6 µm) was performed after deparaffinization, rehydration (graded ethanol solutions), and antigen retrieval (10 mmol/liter citrate, pH 6.0, at 95°C for 10–15 min). Immunofluorescence labeling was done essentially as described above. For immunocytochemistry, cells were fixed using 4% PFA for 10 min at RT and subsequently washed with PBS and immunolabeled as described above. Images were acquired at room temperature using an LSM700 confocal microscope (Carl Zeiss) by using a 40×/1.2-NA lens imaged with Carl Zeiss immersol W 2010, a 63×/1.4-NA lens imaged with Carl Zeiss immersol 518F, or a Panoramic digital slide scanner (3DHISTECH) using a 20×/0.8-NA air lens. For image analysis, images were taken in equivalent spatial distribution from all slides. Image analysis was performed using ImageJ and ZEN 2011 (Carl Zeiss) software. The following primary antibodies were used for immunostaining: rat anti-MBP (1:300, R&D Systems), mouse anti-Cadm4 (1:500, NeuroMab), rabbit anti Cadm4 (1:500), and rabbit anti-Caspr (1:500). Secondary antibodies coupled to Dylight 405, 488, Cy3, Dylight 647, Cy5, and Cy3 were obtained from Jackson ImmunoResearch.

Mice were anesthetized and perfused with a fixative containing 4% PFA, 2.5% glutaraldehyde, and 0.1 M cacodylate buffer. Brains and spinal cords were isolated and incubated in the fixative overnight at room temperature and processed as previously described (Yang et al., 2016). Samples were examined using a Tecnai Spirit or T12 transmission EM equipped with a FEI Eagle camera or a Gatan ES500W Erlangshen camera, respectively. EM micrographs were analyzed using computer-assisted ImageJ analysis software, for axon diameter and total outer axon diameter containing myelin. G-ratio was calculated by dividing the measured inner axonal diameter by the measured total outer axonal diameter. For FIB-SEM, samples were fixed (Yang et al., 2016) and embedded in Epoxy resin (Agar Scientific). FIB-SEM was performed on a Crossbeam 550 FIB-SEM (Carl Zeiss). Specimens were coated (6 nm) with a conductive layer of iridium and prepared with sputter coating (Safematic). Ion beam current at 20 pA to 30 nA was used. Scanning EM images were acquired at 2 kV, 490 pA, with Inlens detector and noise reduction of line averaging (n = 80). Slicing was performed with an ion beam of 700 pA, by Serial Surface Imaging algorithm SmartFIB. Image series of consecutive sections were collected using an isotropic voxel size of 7 × 7 × 7 nm. Image processing was performed by Fiji (National Institutes of Health). FIB-SEM image stacks were aligned and segmented in Microscopy Image Browser (University of Helsinki) and 3D reconstructed in Amira-Avizo SW (Thermo Fisher Scientific).

Spinal cords isolated from genotyped embryonic day 13.5 mice embryos were transferred to HBSS and dissociated using 0.25% trypsin and 0.1% collagenase. Cells were collected, resuspended in spinal cord plating medium (DMEM containing 25% horse serum, 25% HBSS, and 4 mM glutamine), and plated (300,000/13 mm coverslip) on 0.1% poly-l-lysine–coated coverslips. 3 h after seeding, the medium was changed to differentiation medium (DMEM containing 0.5% hormone mix [1 mg/ml apo-transferrin, 20 mM putrescine, 4 mM progesterone, and 6 mM selenium], 10 ng/ml biotin, and 50 nM hydrocortisone). Insulin (10 mg/ml) was added to the differentiation medium for the first 12–14 d. Medium was changed every other day for 28 d.

Mouse brains were lysed in RIPA buffer (10 mM Tris-Cl, pH 8.0, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, 140 mM NaCl, and 1 mM PMSF) supplemented with protease inhibitors. Equal amounts of protein were loaded onto a 10% SDS-PAGE gel and transferred to polyvinylidene difluoride membrane. Membranes were blotted with antibodies to Cadm4, MAG, and Gapdh, and then incubated with HRP-conjugated secondary antibodies. Blots were developed by ECL and imaged on a Bio-Rad ChemiDOC.

Data are presented as the mean ± SEM or average ± SD. Statistical analyses were performed using an unpaired Student’s t test with two tails or one-way ANOVA with Tukey’s or Dunnett’s multiple comparisons test correction. For in vivo experiments, sample size was not predetermined but was based on similar studies in the field.

Fig. S1 shows that MAG exhibits higher paranodal localization than Cadm4.

We thank Jeff Dupree and Mika Simons for discussions.

This work was supported by the National Institutes of Health (R01NS097428), the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation, the National American Brain Foundation, Lilly Fulop Fund for Multiple Sclerosis Research, Estate of David Georges Eskinazi, and research grants from Dahlia and Philip Lawee, Gary Clayman, and Ellie Adiel. E. Pelez is the Incumbent of the Hanna Hertz Professorial Chair for Multiple Sclerosis and Neuroscience.

The authors declare no competing financial interests.

Author contributions: N. Elazar and E. Peles: conceptualization; N. Elazar, A. Vainshtein, Y. Eshed-Eisenbach, K. Rechav, and Y.M. Tsoory: investigation. N. Elazar and E. Peles: manuscript preparation. E. Peles secured funding and supervised the project.