The formation of paranodal spirals at the ends of CNS myelin sheaths requires the planar polarity protein Vangl2

During axonal ensheathment, noncompact myelin channels formed at lateral edges of the myelinating process become arranged into tight paranodal spirals that resemble loops when cut in cross section. These adhere to the axon, concentrating voltage‐dependent sodium channels at nodes of Ranvier and patterning the surrounding axon into distinct molecular domains. The signals responsible for forming and maintaining the complex structure of paranodal myelin are poorly understood. Here, we test the hypothesis that the planar cell polarity determinant Vangl2 organizes paranodal myelin. We show that Vangl2 is concentrated at paranodes and that, following conditional knockout of Vangl2 in oligodendrocytes, the paranodal spiral loosens, accompanied by disruption to the microtubule cytoskeleton and mislocalization of autotypic adhesion molecules between loops within the spiral. Adhesion of the spiral to the axon is unaffected. This results in disruptions to axonal patterning at nodes of Ranvier, paranodal axon diameter and conduction velocity. When taken together with our previous work showing that loss of the apico‐basal polarity protein Scribble has the opposite phenotype—loss of axonal adhesion but no effect on loop–loop autotypic adhesion—our results identify a novel mechanism by which polarity proteins control the shape of nodes of Ranvier and regulate conduction in the CNS.

the myelin sheath also play a critical role in regulating conduction along, and providing metabolic support to, the axon. Noncompact myelin consists of specialized channels that are formed as the myelin sheath is compacted and oligodendroglial cytoplasm is extruded to the edges of myelinating processes (Figure 1a). While the cytoplasmfilled channels formed once wrapping and compaction is complete are continuous within the sheath, they are subdivided into functionally and molecularly distinct regions (reviewed by Simons, Snaidero, & Aggarwal, 2012). The channel at the leading edge of the myelinating process forms the adaxonal inner tongue that runs along the surface of the axon under the sheath. Here adaxonal myelin proteins including the transporter MCT1 and CNPase are required for long-term metabolic support of the axon (Lappe-Siefke et al., 2003;Lee et al., 2012). By contrast, the paranodal channels, so named because of their location abutting the node of Ranvier, are formed at the lateral edges of the myelinating process and form a tight spiral. When cut in cross section along the axon, the turns of the spiral give the appearance of a series of loops-the so-called paranodal loops (Figure 1b).
F I G U R E 1 Organization of the CNS myelin sheath. (a) Myelin sheaths produced by oligodendrocytes consist of compacted layers of plasma membrane wrapped around the axon. The lateral edge of each sheath, which abuts the node of Ranvier (light pink) is characterized by a cytoplasm-filled channel that spirals around the paranodal axon (yellow), separating the node from the juxtaparanode (orange). The channel continues with a trajectory roughly parallel to the axon into the internodal axon (red). On the right is a partially unfurled myelin sheath, showing the division between compact myelin (dark blue) and the cytoplasmic channel (light blue). (b and c) A longitudinal cross section of the paranode reveals that the spiral appears as a series of cytoplasm-filled loops that are adhere to both the axon and each other. The adhesions present in each paranodal loops, and the polarity of the axis along which these adhesion complexes are patterned, are shown in c. (d and e) Paranodal axoglial adhesions are disrupted in the CNS following conditional elimination of Scribble expression in oligodendroglia [Color figure can be viewed at wileyonlinelibrary.com] This paranodal spiral forms a close association with the underlying axon through an adhesion complex comprising the 155 kDa isoform of neurofascin (NFC155) on the oligodendrocyte binding to contactin-1 and Caspr on the axon (Charles et al., 2002;Rios et al., 2000). The resulting tight axoglial adhesion at the paranodal domain acts as a physical and electrical barrier separating the voltage-gated sodium channels required for saltatory conduction from the delayed rectifier potassium channels at the juxtaparanode (reviewed by Zollinger, Baalman, & Rasband, 2015). Another distinct set of adhesion proteins hold adjacent turns of the spiral together (so maintaining a tight spiral), including the tight junction protein Claudin-11 and gap junction protein Connexin-32 (Cx32), which are both essential for normal conduction in the CNS (Figure 1b, Devaux & Gow, 2008;Sargiannidou et al., 2009). These adhesion molecules are confined to lateral membranes of the channel, where they form autotypic junctions between neighboring paranodal loops (Gow et al., 1999;Kamasawa et al., 2005). While therefore essential for the complex three-dimensional organization and stability of the myelin sheath, the mechanisms responsible for the precise localization of the distinct sets of adhesion proteins within the paranodal channel and its resultant polarization (Figure 1c, arrow) are unknown.
The precise arrangement of axoglial and autotypic adhesions within the paranodal channel represents a remarkable example of subcellular polarization. In other cell types, the localization of proteins to specific membrane subdomains is controlled by evolutionarily conserved regulators of cell polarity. In epithelial cells, where the function of polarity proteins is best understood, this occurs along two perpendicular axes: Apico-basal polarity (ABP) signaling controls the distribution of proteins between the inner and outer faces of the epithelium, while the planar cell polarity (PCP) pathway operates in an orthogonal plane to the ABP pathway, conferring an "anterior" and "posterior" identity to lateral membranes within epithelial cells, and a directionality to the epithelium itself (reviewed by Campanale, Sun, & Montell, 2017). We have previously demonstrated that Scribble, a key regulator of ABP in epithelia (Bilder & Perrimon, 2000), is required for paranodal axoglial adhesion, as revealed by the detachment of loops from the axon following the conditional elimination of Scribble expression in oligodendroglia (Figure 1d,e;Jarjour et al., 2015). However, very little is known about the role played by the other major polarity signaling pathway, PCP, in CNS myelination. Among the highly evolutionarily conserved core PCP proteins, the four-pass transmembrane protein Van Gogh-like 2 (Vangl2), a vertebrate orthologue of Drosophila Van Gogh/Strabismus, is of particular interest owing to its close physical and functional interaction with Scribble (Courbard, Djiane, Wu, & Mlodzik, 2009;Montcouquiol et al., 2003;Yates et al., 2013). In neurons, Vangl2 has been implicated in the regulation of neurite outgrowth (Dos-Santos Carvalho et al., 2020), axon guidance (Leung et al., 2016;Sun, Purdy, & Walsh, 2016), dendritic spine morphology (Okerlund, Stanley, & Cheyette, 2016) and glutamatergic synapse formation (Thakar et al., 2017). It has previously been reported that the Vangl2 binding protein Prickle1 is expressed by oligodendroglia (Liu et al., 2013) and promotes their differentiation (Zilkha-Falb, Gurevich, Hanael, & Achiron, 2017), but the role played by Vangl2 itself in paranode formation and other oligodendroglial functions is unknown. Here, we have addressed this question by removing Vangl2 from oligodendrocytes. We reveal a striking phenotype in which the paranodal spiral loosens and unwinds and demonstrate that, as in epithelia, Vangl2 regulates the precise localization of junctional adhesion proteins. Within the paranodal spiral these adhesion proteins normally stabilize the paranodal channel in a tight spiral around the axon, a spiral that patterns the underlying axon.
Thus, we reveal a novel role for PCP signaling molecules in both shaping and patterning the paranodal axon.

| Contact for reagent and resource sharing
Requests for further information, resources, and/or reagents should be addressed to Andrew Jarjour (andrew.jarjour@merck.com).
Reagents used:

| Experimental model and subject details
Mice in which the Vangl2 gene was targeted for condition elimination by insertion of LoxP sites flanking exon 4 (Ramsbottom et al., 2014) were crossed with mice expressing Cre recombinase under the control of the 2 0 -3 0 cyclic nucleotide phosphodiesterase (CNP) gene (Lappe-Siefke et al., 2003) to generate CNP +/Cre :Vangl2 fl/fl mice, in which Vangl2 expression is conditionally eliminated in myelinating glia. Both male and female mice (1-8 weeks of age) were used. A total of 110 mice were used for all studies described in this study. All animal work conformed to UK legislation (Scientific Procedures) Act 1986 and to the University of Edinburgh regulations.

| Mouse oligodendroglial cultures
Eight postnatal Day 7 mouse pups were killed by injection with a 1:1 mix of Dormitor and Vetelar and their cortices removed. Cortices were dissociated using a gentleMACS dissociator using C tubes with the Neural Tissue Dissociation Kit (Miltenyi Biotec). Cell suspensions were subjected to immunopanning, first by negative selection using "bacterial" (untreated) culture plates coated with anti-Griffonia Simplicifolia Lectin, then by positive selection using antibodies against platelet-derived growth factor receptor α (PDGFRα). Positively selected cells were cultured on PDL-coated T-75 flasks in OPC-Sato medium (high-glucose DMEM with, 1% apo-Transferrin, 1% bovine serum albumin, 9.934 mM Putrescine dihydrochloride, 19.88 μM progesterone, 0.2313 μM sodium selenite, B27 supplement, 0.3678 mM N-acetyl cysteine, 5 μg/ml insulin, Trace Element B mix, 4.093 nM biotin and 5 nM forskolin) with 10 ng/ml PDGF-AA and 5 ng/ml neurotrophin 3 (NT3) at 37 C in 7.5% CO 2 . 50% medium changes were carried out every day for 7 days to allow OPCs to proliferate. When OPCs reached confluence, cells were seeded on PDL-coated 22 mm glass coverslips, and differentiated into oligodendrocytes in OPC-Sato medium with 5 ng/ml NT3 and 10 ng/ml ciliary neurotrophic factor (CNTF) for at 37 C in 7.5% CO 2 for 2 days. For immunocytochemical analyses, cultures were fixed with 4% paraformaldehyde and washed twice in PBS, and blocked with blocking solution (4% heat-inactivated donkey serum, 2% BSA, and 0.1% Triton X-100 in PBS). Cultures were incubated overnight in primary antibodies diluted in blocking solution. Following repeated washes with PBS, cultures were incubated with secondary antibodies, washed with PBS, and then mounted. Confocal z-stacks were acquired using Leica SP8 inverted confocal microscope (×63 objective, zoom 2). Images were processed using ImageJ and Photoshop (Adobe) software.

| Teased spinal cord fibers
Thirty-four postnatal Day 14 or 60 mice were fixed by cardiac perfusion with 4% formaldehyde in PBS. The cervical spinal cord was dissected out, and postfixed for 30 min in cold 4% formaldehyde in PBS before being stored in PBS at 4 C for at most 48 hr. The meninges were removed, and the ventral white matter isolated into cold PBS, and cut into pieces roughly 2 mm long. Fibers were then teased onto SuperFrost Plus slides (for standard confocal imaging) or TESPAcoated coverslips (for gSTED super-resolution imaging) using acupuncture needles. Slides were stored at −20 C in airtight containers until immunolabeling. Immunohistochemical labeling was carried out as described above for cultures. For standard confocal imaging, zstacks were acquired using Leica SP8 inverted confocal microscope (×63 objective, 1.40 N.A.). Maximum intensity projection images of Zstacks were used for analysis. Images were processed using ImageJ and Photoshop (Adobe) software. For immunolabeling of STED images, the secondary antibodies used were Atto647N-conjugated goat anti-rabbit IgG for Claudin-11 immunolabeling and Alexa 568-conjugated goat anti-mouse IgG (H + L) for Caspr labeling. For STED images, z-stacks were acquired using a Leica TCS SP8 STED ×3 microscope (×100 objective, 1.40 N.A.), using the 775 nm laser for depletion. Stacks were deconvolved using the Huygens STED deconvolution package (Scientific Volume Imaging B.L., Netherlands) integrated in Leica LAS AF software prior to analysis. Images were acquired using a Gatan OneView camera (Gatan). and recorded using a Cygnus ER-1 differential amplifier (Cygnus Technologies, Inc.) and pClamp 10 software (Molecular Devices). Data were filtered at 10 kHz and sampled at 50 kHz.

| Quantification and statistical analyses
All confocal and TEM images were analyzed using FiJi software (Schindelin et al., 2012). For analysis of separation between paranodal loops, neighboring loops were considered separated if they were clearly separated by more than one membrane width for at least one-third of the length of the border between the loops. For analysis of detachment of paranodal loops from the axon, a loop was considered detached if a separation of at least 100 nm was measured between the loop and axon.
G-ratio was determined by dividing axon diameter by the total diameter of the axon and overlying myelin sheath. Axon diameter was calculated from measured axon perimeter based on an assumption of circularity.
For measurements taken from optic nerve, the diameter of the myelinated axon was determined similarly. For measurements of g-ratio in spinal cord, where myelin sheaths often appeared uncompacted, myelin sheath thickness was measured directly at the point at which the myelin sheath was most compact, and myelinated axon diameter was determined by adding this value to the calculated axon diameter. Internodal axonal diameter was measured at the location in the field where the internodal axon was thickest. Paranodal axonal diameter was measured at the innermost border of whichever second loop from the inside is located closer to the node (where inner means toward the internode). For gSTED analysis of immunofluorescence intensity of Claudin-11 or Caspr labeling, image stacks were cropped to eliminate half the paranodal spiral so that "top" and "bottom" edges of each turn of the spiral were not double-counted and a maximum intensity projection was performed. On the projected image, a 2 pixel-wide line was drawn longitudinally across the paranodal spiral starting at the nodal end and an intensity plot was generated. Maxima and minima were identified using the BAR Scripts 1.5.1 "Find Peaks" macro (Ferreira et al., 2017). 3 | RESULTS

| Vangl2 is expressed by oligodendroglia and is localized to CNS paranodes
To ask whether Vangl2 plays a role in organizing the myelin sheath, we first examined its distribution within oligodendroglia and in CNS myelin. We confirmed that Vangl2 is expressed by MBP-positive oligodendroglia. Vangl2 is localized to cell bodies and processes in wildtype oligodendrocytes but is excluded from myelin-like membrane sheets (

| Oligodendroglial Vangl2 organizes CNS paranodes and nodes of Ranvier
Our previous findings demonstrated that the apico-basal polarity protein Scribble is required for normal CNS myelin initiation, thickness and paranodal axoglial adhesion (Jarjour et al., 2015). By contrast, Vangl2 cKO animals had no reduction in the proportion of myelinated  To ask whether this loosening of the paranodal spiral seen at P60 reflects a failure of the formation of the paranodal structure or a loss of its maintenance once formed, we analyzed P14 spinal cord-a time when axons are still growing in length. Here we found using EM that slight separations between loops could be seen even at this early time point ( Figure S3), although statistical significance was not reached until the axons lengthened at P60. We conclude, therefore, that Vangl2 is required for normal adhesion between paranodal loops in the CNS, with the consequent phenotype of a loosened spiral fully revealed once the axon has lengthened. Vangl2 is not, however, required for the adhesion between loops and axon.
To further quantify the spacing between turns of the paranodal spiral, paranodal Caspr immunolabeling was imaged using superresolution gated STED (gSTED) confocal microscopy. The distance between turns of the Caspr spiral, measured by determining the distance between Caspr intensity peaks, was increased at Vangl2 cKO paranodes ( Figure 3m) relative to wild-type paranodes (Figure 3k, quantified in L). Increased spacing between loops became more pronounced with distance from the node of Ranvier, as visualized by the increase in slope of the best-fit line when plotting distance between turns against distance from the node of Ranvier (Figure 3n).

| Vangl2 organizes the CNS myelin sheath and patterns the underlying axon
Two of the key functions of CNS paranodal axoglial adhesions are to promote the concentration of an axolemmal protein complex containing voltage-gated sodium (Na V ) channels at nodes of Ranvier (Zonta et al., 2008), and to form a barrier that separates the nodal complex from a juxtaparanodal complex containing voltage-gated potassium (Kv) channels (Bhat et al., 2001). To determine whether paranodal function was disrupted in the absence of Vangl2 expression by oligodendroglia, we examined axonal organization in teased spinal cord preparations from P60 Vangl2 cKO mice. We observed that, while Caspr-immunolabeled paranodes flank Na V -immunopositive nodes of Ranvier in wild-type CNS (Figure 4a,c,e), paranodal Caspr extends into the nodes of Ranvier of the Vangl2 cKO mice, narrowing the nodal "gap" between neighboring paranodes (Figure 4b [arrow], f,g), and also in toward the internode (Figure 4b, arrowhead). Reduced distance across the node of Ranvier could also be detected by TEM in P60 spinal cord ( Figure S4). Surprisingly, despite the narrowing of the node of Ranvier in the Vangl2 cKO CNS, the nodal Na V -immunolabeled domain was widened, with Na V labeling extending laterally into the paranodes ( Figure 4d [arrowheads], f,h) with no significant difference in the mean intensity of nodal Na V immunofluorescence observed in Vangl2 cKO teased spinal cord at P60 (Figure 4i). This widening of the Na V -labeled F I G U R E 4 Oligodendroglial Vangl2 expression is required for domain organization of myelinated CNS axons at nodes of Ranvier. In wild-type CNS, Caspr-immunopositive paranodes (green) flank Na V -immunopositive nodes of Ranvier (red), with a clear separation existing between neighboring paranodes (a, c, and e). In Vangl2 cKO CNS (b, d, and f) the Caspr domain extends both into the internode (b, arrowhead) and into the node of Ranvier (b, arrow). This shortens the node (quantified in g, distance between paranodal Caspr immunolabeling across node of Ranvier (μm): WT: 1.34 ± 0.08, cKO: 0.90 ± 0.05, p = .0079). The nodal Na V -immunolabeled domain is widened and spreads laterally into the paranodes (d, arrowheads, quantified in h; width of nodal Na V immunolabeling (μm): WT: 1.9 ± 0.1, cKO: 3.2 ± 0.2, p = .0079). No significant change in nodal sodium channel immunolabeling intensity was observed in Vangl2 cKO CNS (i; WT: 100.8 ± 10.6, cKO: 94.4 ± 6.4, p = .8413). In wild-type CNS, a clear paranodal separation is observed between nodal Na V (red) and juxtaparanodal K V (green) immunolabeling (j, l, and n). In Vangl2 cKO (k, m, and o), the K V -immunolabeled area is widened (p; width of K V 1.1 immunolabeling (μm): WT: 17.8 ± 1.0, cKO: 28.9 ± 3.2, p = .0079) and invades the paranodal domain (k and o, arrowheads), with the distance between Na V and K V 1.1 immunolabeling reduced (q; WT: 2.5 ± 0.2 μm, cKO: 1.5 ± 0.2 μm, p = .0159). (q-aa): Septin-8 is mislocalized to paranodes in Vangl2 cKO P60 teased spinal cord fibers. Septin-8 (green), which is restricted to the internodal inner tongue in wild-type fibers (r, t, and v), extends into Caspr-immunopositive (red) paranodal regions in Vangl2 cKO fibers (s, u, and w, arrowheads), with the percentage length of the Caspr-immunopositive domain that overlapped with Septin-8 labeling being significantly increased (x; WT: 19.6 ± 3.7, cKO: 35.3 ± 3.1, p = .0317). Myelin was immunolabeled for MAG (blue) and axons for NFH (white). Super-resolution imaging using gated STED reveals that Septin-8 (green, y, aa) and Caspr (red, z, aa) do not colocalize at paranodes. All data are represented as mean ± SEM. n = 5 mice per genotype. Statistical analyses were performed using the Mann-Whitney test. a-h: At least 20 nodes of Ranvier were analyzed per animal. i-p: At least 40 paranodes were analyzed per animal. Scale bars: (a-f and j-o): 5 μm, (y-aa): 300 nm [Color figure can be viewed at wileyonlinelibrary.com] domain in Vangl2 cKOs was observed in all rather than a subset of nodes, and was not dependent on axon diameter (data not shown).
We also observed a similar change when we examined the distribution of axonal spectrins at Vangl2 cKO nodes of Ranvier. The mutual exclusion of the βIV spectrin-containing nodal cytoskeleton ( Figure S5A,E) and the βII spectrin-containing complex underlying the paranodal cytoskeleton ( Figure S5C,E) is a key mediator of the paranode's Na V clustering and barrier functions (Amor et al., 2017). We observed that in Vangl2 cKO CNS, the width of nodal βIV spectrin domains were widened ( Figure S5B,F,H), while the distance across the nodal gap in βII spectrin labeling was decreased (Figure S5D

| Vangl2 is required for local regulation of axon diameter at CNS paranodes
In addition to patterning the underlying axon into distinct molecular domains, the myelin sheath physically shapes the axon. Contact between oligodendrocytes and axons promotes radial axonal growth during development (Sanchez, Hassinger, Paskevich, Shine, & Nixon, 1996), with the internodal axon diameter being greater than that of the paranode and node in the adult CNS (Hildebrand & Skoglund, 1971). When the ratio of paranodal axon diameter (dP) to internodal axon diameter (dI) was measured in Vangl2 cKO spinal cord at P60, an increased dP/dI ratio relative to that observed in wild-type spinal cord ( Figure 5a,b,e) was observed. Interestingly, this difference between Vangl2 cKO and WT axons was not seen in P14 spinal cord, a stage at which axons are still increasing in diameter (Figure 5c,d,f). We conclude that oligodendoglial Vangl2 is essential for the local regulation of axon diameter at CNS paranodes.

| Increased conductivity in Vangl2 cKO CNS
The observed disruptions to axonal domain organization and the regulation of axon diameter at nodal regions in the Vangl2 cKO raised the question of how these changes, particularly to the length and diameter of the node of Ranvier, affect conduction. When compound action potentials (CAPs) were measured in the optic nerve at P60, we observed three peaks ( Figure 6a, arrowhead, black arrow, gray arrow), as has previously been reported (Foster, Connors, & Waxman, 1982). Across all three peaks, conduction velocities were increased in Vangl2 cKO nerves compared to wildtype controls (Figure 6b). Further analysis revealed that the minimum stimulus intensity required to elicit each component of the CAP was decreased in Vangl2 cKO nerves (representative traces shown in Figure 6c, quantification in Figure 6d). These findings suggest that myelination by Vangl2 cKO oligodendrocytes results in increased conductivity in the CNS. F I G U R E 7 Vangl2 controls the distribution of the autotypic tight junction protein Claudin-11 within CNS paranodal myelin. (a-i) Standard and super-resolution gated stimulation emission depletion (gSTED) confocal microscopy revealed altered distribution of Claudin-11 in P60 Vangl2 cKO spinal cord paranodes. The spiral of paranodal oligodendroglial Claudin-11 immunolabeling is loosened and elongated laterally along the axon in teased spinal cord preparations from Vangl2 cKO mice (b) relative to those from Vangl2 WT (a). In Vangl2 cKO paranodes (d and f), the intensity of Claudin-11 distribution (red in c and d, white in e and f) appeared diffuse relative to that observed in wild-type paranodes (c and e). Caspr-immunolabeled axoglial adhesions are shown in green (c and d). When Claudin-11 immunolabeling intensity is plotted across individual paranodes (along the yellow lines in e and f), analysis of intensity maxima (red dots in g and i) and minima (blue dots in g and i) revealed that the difference between maxima and minima was decreased at Vangl2 cKO paranodes (h). (j-l) Microtubule density is reduced in the Vangl2 cKO paranodal spiral. In P60 Vangl2 cKO spinal cord (l), the density of microtubules (white arrowheads) within paranodal loops was decreased relative to that observed in wild-type animals (j, quantification in k; WT: 129.5 ± 10.2 microtubules/μm 2 , cKO: 51.6 ± 5.4 microtubules/μm 2 , p = .0079). All data are represented as mean ± SEM. n = 5 mice per genotype. 3.6 | Vangl2 regulates the distribution of autotypic adhesion proteins in CNS paranodes The observed disruption to axonal shape and patterning at nodes of Ranvier and, consequently, on conduction as a result of a "loosened" paranodal spiral following conditional Vangl2 deletion in oligodendroglia raises the question of the mechanism by which Vangl2 "tightens" the spiral. Autotypic tight junctions containing Claudin-11 (Gow et al., 1999) and gap junctions comprised of Connexin-32 (Cx32, Kamasawa et al., 2005) have been reported to bridge loops in CNS paranodal myelin, suggesting that these complexes may be responsible for mediating adhesion between turns of the paranodal spiral. Vangl2 has previously been reported to regulate tight junction organization and claudin-1 localization in uterine luminal epithelium (Yuan et al., 2016), raising the possibility that it may function similarly to localize Claudin-11-containing tight junctions in paranodal myelin.
To test this and so provide a mechanism for the loosening of the paranodal spiral in the Vangl2 cKO mice, we imaged Claudin-11 immunolabeling within the loops using super-resolution gSTED confocal microscopy. Standard confocal imaging revealed that the spiral of It is clear, however, that much remains to be learned about the mechanisms that organize the paranodal spiral. The phenotype observed in Vangl2 cKO CNS cannot be explained solely by disruption of Claudin-11. In the Claudin-11 null CNS, disruption to paranodes was not observed and impulse propagation is slowed and not hastened in these animals (Devaux & Gow, 2008;Maheras et al., 2018); its principal function is believed to be as a diffusion barrier for water and other small molecules (Denninger et al., 2015). One candidate for these additional effects in the Vangl2cKO mice is mislocalization of Cx32, which we were unable to examine by super resolution microscopy due to a lack of suitable antibodies. The primary role of Cx32 gap junctions in myelin is believed to be as a conduit permitting the transport of water, ions and other small molecules between turns of the paranodal spiral, and between abaxonal myelin and astrocytes, allowing for the buffering and recycling of K + ions released by neurons (Menichella et al., 2006). However, it has been reported that homotypic Cx32 interaction can drive cell-cell adhesion (Cotrina, Lin, & Nedergaard, 2008), and that enlarged and irregularly shaped paranodal loops are present in the CNS of Cx32 null animals, with neurons in Cx32 mutant displaying hyperexcitability in response to weak stimuli (Sutor, Schmolke, Teubner, Schirmer, & Willecke, 2000). Given that Cx32 gap junctions are found only in oligodendrocytes, and that oligodendrocytes do not appear to form gap junctions with neurons (Rash, Yasumura, Dudek, & Nagy, 2001) these observations could indeed be consistent with this protein playing a role in organizing the paranodal spiral. While mice lacking both it and another oligodendroglial connexin, Cx47, experience a severe CNS phenotype including oligodendrocyte death and CNS demyelination, most likely due to a loss of astrocyte-oligodendrocyte gap junctions (Menichella, Goodenough, Sirkowski, Scherer, & Paul, 2003) Cx32 KO CNS does not show any gross myelin pathology (Scherer et al., 1998). A detailed analysis of myelin organization and axonal patterning at Cx32 KO nodes and paranodes could address the question of whether Cx32 plays an asyet unidentified role in stabilizing the paranodal spiral in the CNS.
Other adhesion proteins may also play a role in stabilizing the paranodal spiral. In the PNS, E-cadherin is concentrated at adherens junctions between Schwann cell paranodal loops (Fannon et al., 1995).
Vangl2 can bind directly to, and control the plasma membrane localization of, both E-cadherin (Nagaoka, Inutsuka, Begum, Bin Hafiz, & Kishi, 2014) and N-cadherin . While E-cadherin is not expressed in oligodendroglia, N-cadherin is expressed in myelinating oligodendrocytes, where it plays a role in myelin initiation (Schnadelbach, Ozen, Blaschuk, Meyer, & Fawcett, 2001) highlighting this protein as an attractive candidate to mediate adhesion within paranodal spiral in the CNS. Another possible candidate for such a role in the CNS is the transmembrane protein opalin, which has been localized to sites between paranodal loops in CNS myelin (Yoshikawa et al., 2008). However conditional elimination of opalin in oligodendroglia does not affect paranodal organization or conduction, suggesting that it is not essential for the compaction of the paranodal spiral (Yoshikawa et al., 2016). Given the complexity of this structure, it seems likely that Vangl2 regulates the integrity of the paranodal spi- as well as disrupted paranodal axoglial adhesion and mislocalization of juxtaparanodal K V channels to paranodes in older animals (Saifetiarova & Bhat, 2019), suggesting that the other causes of disruption to axoglial interaction at the paranode may contribute to the observed trailing-loop phenotype.
A surprising insight gained through our study of the Vangl2 cKO phenotype relates to the ability of paranodal myelin to locally regulate axon diameter. The initial observation in CNS that axonal diameter is greatest at internodes and is decreased at the node of Ranvier and paranodes was made approximately 50 years ago (Hildebrand, 1971), and provides a mechanism to decrease the amount of energy that must be spent to repolarize the membrane following passing of an action potential. Both our own findings and those of Arancibia-Carcamo and colleagues suggest that sodium channel density appears to remain constant as the dimensions of the node of Ranvier change (Arancibia-Carcamo et al., 2017), suggesting that regulation of the amount of nodal membrane present is the primary factor in controlling the number of sodium ions that enter the nodal axon during depolarization. As restoration of membrane resting potential by the Na+/K+ ATPase is responsible for as much as 70% of the energy expenditure of neurons with long-projecting axons in the brain (Howarth, Gleeson, & Attwell, 2012), tightly controlling the dimensions of the node of Ranvier is an important mechanism that ensures that brain energy usage is efficient. However, the mechanism underlying nodal diameter reduction remains unknown. It has been observed in PNS that in cross sections of internodal regions of myelinated axons, neurofilaments are relatively spread out, while they are more densely packed together at paranodes (Price, Lasek, & Katz, 1990) suggesting an external compressive force (Price, Lasek, & Katz, 1993). In the PNS, it was observed that a dense actin cytoskeleton encircles the paranodal axon, leading to the hypothesis that actin contractility within the Schwann cell generates force that squeezes the paranodal axon (Zimmermann, 1996). Vangl2 has a previously reported role in generating contractile force through regulation of the actin cytoskeleton during gastrulation (Ossipova, Chuykin, Chu, & Sokol, 2015), but as filamentous actin is largely absent from CNS myelin (Zuchero et al., 2015) any such compression would have to be driven by an actinindependent process. One possible interpretation of our findings is that increased paranodal axon diameter is consistent with an inability of the paranodal spiral to "lock" by adhesion between loops at Vangl2 cKO CNS paranodes, leading to relaxation and lengthening of the spiral and an inability to squeeze the enlarging axon at paranodes as radial growth proceeds between P14 and P60. The presence of a compressive action of the paranodal spiral on the axon has not been proven at CNS or PNS nodes, however, and further investigation will be required to determine the mechanism by which Vangl2 regulates paranodal axon diameter.
We observed in the Vangl2 cKO mice that the length of the Na Vimmunopositive domain was increased, extending into the paranode even though the length of the node as measured by electron microscopy was not altered. This lengthening presumably occurs through spiral diffusion of the channels along the widened areas of axon membrane created between turns of the loosened paranodal loops-the so-called "pathway 3" described by Mierzwa in 2010 (Mierzwa, Shroff, & Rosenbluth, 2010). This in turn may explain why nerves lac- A key assumption of this model, however, is that the density of sodium channels at the nodal plasma membrane is held constant. Indeed, we found that no significant difference in the mean intensity of nodal Na V immunofluorescence could be observed in Vangl2 cKO teased spinal cord at P60, suggesting that the functionally longer nodes of Ranvier in the Vangl2 cKO contain more voltage-gated sodium channels than do those in wild-type CNS.
Implicit in our assumption is that the entire sodium channel domain (and not just the region of bare axon between adjacent myelin segments) contributes to conduction in Vangl2 cKO, with the sodium channels mislocalized into paranodal regions in these animals accessible and actively contributing to depolarization.
Our results showing increased conduction velocity with loss of loop-loop adhesion differs from that reported for animals lacking another paranodal autotypic adhesion molecule, JAM-C (Scheiermann et al., 2007). Here decreased conduction is seen in the PNS of JAM-C deficient mice. However, a deficiency of JAM-C causes a loss of both loop-loop and loop-axon adhesion, as shown in figs. 2B and S4D of that paper. We suggest that there will be a much greater perturbation of the paranodal sealing required for saltatory conduction resulting from loss of adhesion between loops and the underlying axon than from a loop-loop defect, and that this loss of sealing then dominates the electrophysiological phenotype.
A predicted consequence of an increased number of sodium channels at Vangl2 cKO nodes of Ranvier would be a corresponding increase in the amount of energy required to repolarize the membrane following passing of an action potential. Interestingly and consistent with this, we observed that an increased proportion of nodes of Ranvier in Vangl2 cKO P60 spinal cord contained mitochondria compared to wild-type nodes (data not shown), which could represent a response to this increased metabolic demand of the Na + /K + ATPase as has been observed in the peripheral nervous system (PNS,C. L. Zhang, Ho, Kintner, Sun, & Chiu, 2010). Increased energy demand by axons could also explain why myelin sheaths in P60 Vangl2 cKO spinal cord contain significantly more internodal Schmidt-Lanterman incisure-like myelinic channels (data not shown), as these structures are thought to play a key role in delivering metabolic support from the oligodendrocyte to the axon (Philips & Rothstein, 2017).
Finally, our finding that Vangl2 is a key regulator of paranodal architecture also has important implications for myelin pathology and plasticity. Paranodal loosening occurs with CNS aging (Hinman et al., 2006) and elongation of paranodes has been observed in regions of the brain that have undergone remyelination in multiple sclerosis patients (Howell et al., 2006). The similarity of these phenotypes with those we see here suggests that Vangl2 may be compromised in aging or disease. Such a conclusion is potentially important, as a possible consequence of "looser" paranodal assembly could be increased susceptibility of the myelin sheath to age-related demyelination, or to further damage following remyelination, at least in part due to the presence of circulating antibodies in the sera of individuals suffering from multiple sclerosis recognizing paranodal epitopes that are normally inaccessible (Mathey et al., 2007). This could for example provide an explanation for why MS lesions often reoccur in regions where lesions have already occurred (Prineas et al., 1993).

ACKNOWLEDGMENTS
We thank David Lyons and Anna Williams for their helpful comments about the manuscript, and Bertrand Vernay for his assistance with image analysis. All authors reviewed and commented on the article before submission. Funding for this project was provided by the Wellcome Trust and the UK Multiple Sclerosis Society.

CONFLICT OF INTEREST
The authors declare no competing interests.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.