Whirlin, a cytoskeletal scaffolding protein, stabilizes the paranodal region and axonal cytoskeleton in myelinated axons

Background Myelinated axons are organized into distinct subcellular and molecular regions. Without proper organization, electrical nerve conduction is delayed, resulting in detrimental physiological outcomes. One such region is the paranode where axo-glial septate junctions act as a molecular fence to separate the sodium (Na+) channel-enriched node from the potassium (K+) channel-enriched juxtaparanode. A significant lack of knowledge remains as to cytoskeletal proteins which stabilize paranodal domains and underlying cytoskeleton. Whirlin (Whrn) is a PDZ domain-containing cytoskeletal scaffold whose absence in humans results in Usher Syndromes or variable deafness-blindness syndromes. Mutant Whirlin (Whrn) mouse model studies have linked such behavioral deficits to improper localization of critical transmembrane protein complexes in the ear and eye. Until now, no reports exist about the function of Whrn in myelinated axons. Results RT-PCR and immunoblot analyses revealed expression of Whrn mRNA and Whrn full-length protein, respectively, in several stages of central and peripheral nervous system development. Comparing wild-type mice to Whrn knockout (Whrn−/−) mice, we observed no significant differences in the expression of standard axonal domain markers by immunoblot analysis but observed and quantified a novel paranodal compaction phenotype in 4 to 8 week-old Whrn−/− nerves. The paranodal compaction phenotype and associated cytoskeletal disruption was observed in Whrn−/− mutant sciatic nerves and spinal cord fibers from early (2 week-old) to late (1 year-old) stages of development. Light and electron microscopic analyses of Whrn knockout mice reveal bead-like swellings in cerebellar Purkinje axons containing mitochondria and vesicles by both. These data suggest that Whrn plays a role in proper cytoskeletal organization in myelinated axons. Conclusions Domain organization in myelinated axons remains a complex developmental process. Here we demonstrate that loss of Whrn disrupts proper axonal domain organization. Whrn likely contributes to the stabilization of paranodal myelin loops and axonal cytoskeleton through yet unconfirmed cytoskeletal proteins. Paranodal abnormalities are consistently observed throughout development (2 wk-1 yr) and similar between central and peripheral nervous systems. In conclusion, our observations suggest that Whrn is not required for the organization of axonal domains, but once organized, Whrn acts as a cytoskeletal linker to ensure proper paranodal compaction and stabilization of the axonal cytoskeleton in myelinated axons.


Background
Nervous system function depends on proper molecular organization between neurons and glial cells. In myelinated neurons, the segregation and enrichment of proteins in the defined domains, the Node of Ranvier, paranode, and juxtaparanode, is critical for saltatory action potential propagation [1][2][3][4]. Without proper organization, electrical nerve conduction is delayed and can result in significant motor and sensory deficits. The paranodal domain is a region of direct interaction between the glial myelin membrane loops and neuronal plasma membrane. Axoglial septate junctions (AGSJ) link these glial membrane loops to the axonal membrane and establish the paranodal ionic barrier separating sodium (Na + ) channel-enriched node from the potassium (K + ) channel-enriched juxtaparanode [5]. AGSJ's are composed of three transmembrane proteins: Contactin (Cont) [6], Contactin-associated protein (Caspr) [7], and Neurofascin155 (glial-derived 155 kDa isoform) [8]. Genetic ablation of any of the three molecules disrupts the paranodal barrier function and results in degraded action potential propagation [6][7][8][9][10][11]. While much research has focused on these three proteins, other molecules contribute to domain formation and stabilization. For example, Caspr2 and TAG-1 are required for the organization of the juxtaparanodal domain and localization of potassium (K + ) channels [12]. For long-term stability, Caspr and Caspr2 are thought to rely on cytoskeletal scaffold proteins to link their associated complexes with the axonal cytoskeleton. Loss of Caspr results in significant cytoskeletal disorder [9], and since its extracellular partner Cont lacks an intracellular c-terminus, the c-terminus of Caspr is likely a critical region of interaction between AGSJ's and the axonal cytoskeleton. Previous studies using 4.1B mutant mice revealed that the 4.1B cytoskeletal protein contributes to paranodal and juxtaparanodal stability [13,14]. A recent report also highlights 4.1G as an organizer of internodes in the peripheral nervous system [15]. The neuron is abundant with cytoskeletal scaffolds however, and such findings do not exclude the possibility that other cytoskeletal elements may contribute to paranodal maintenance and long-term stability.
Whrn is a cytoskeletal scaffold protein that plays an important role in vision and hearing [16,17]. In humans, WHIRLIN (WHRN) (DFNB31) mutations have been linked to Usher syndrome Type II (USH2), an autosomal recessive vision-hearing impairment disorder [18,19]. In mice, the Whrn coding sequence consists of 12 exons with two dominant splice variants, a full length and short (exon [5][6][7][8][9][10][11][12] isoform. Both variants contain a Proline-rich domain and PDZ-domain(s) which have been shown to link submembranous cytoskeletal elements to transmembrane complexes, as well as to self-oligomerize [20]. In the eye, Whrn interacts with the transmembrane proteins Usherin (USH2) and Very Large G protein-coupled Receptor-1 (VLGR1/GPR98) within the periciliary membrane complex of photoreceptors. Likewise in the inner ear, Whrn interacts with Usherin and VLGR1/GPR98, which form the stereociliary ankle-links. Additionally, Whrn interacts indirectly with 4.1B [21] and 4.1R through Mpp1/p55 [22], and directly with Myosin XVa [23]. Finally, Whrn is implicated with Esp8 in stereocilialength regulation [24]. While the function of Whrn in the ear and the eye has received significant attention, very little is known about its function in the central and/or peripheral nervous system. There are reports of Whrn protein expression in the cerebrum, cerebellum, and brainstem in wild-type mice and the protein is absent in Whrn knockout (Whrn −/− ) and whirler (Whrn wi/wi ) mutant mice [25]. In Drosophila, the closest homolog to Whrn is dyschronic (dysc) [26]. In dysc mutants there is arrhythmic locomotor behavior but the eclosion circadian rhythms and clock protein cycling is unaffected [26]. Here we report that Whrn is involved in proper compaction of the paranodal region in myelinated axons and for proper stabilization of the axonal cytoskeleton.

Results and discussion
Whrn is expressed in central and peripheral nervous system tissues throughout development Whrn is a cytoskeletal scaffolding protein which functions to link membrane protein complexes to the cytoskeleton within hair cell stereocilia of the ear and photoreceptors in the eye. To begin assessing its function in myelinated neurons, we obtained Whrn exon 1 knockout mice [17]. As reported previously, the murine Whrn locus consists of 12 exons with two dominant splice variants, a full length (~4 kb) isoform and a short (~2.5 kb) isoform ( Figure 1A). Both variants contain PDZ-domains ( Figure 1A, yellow box) and a Proline-rich domain ( Figure 1A, blue box). After initial back crossing to C57BL6 mouse strain, we identified and confirmed the Whrn genotype using PCR methods. To begin characterizing mRNA expression of Whrn in the central (CNS) and peripheral nervous systems (PNS), we examined the relative expression of Whrn by reverse transcriptase polymerase chain reaction (RT-PCR) in dorsal root ganglia (DRG), sciatic nerves (SN), and spinal cord (SC) tissues ( Figure 1B) in postnatal 21 dayold mice. With this subset of tissues we could delineate the origin of Whrn expression as SC tissue is a combination of glial and neuronal nuclei, DRG is predominantly neuronal, and SN is principally glial cytoplasm and nuclei. Since Whrn has two major isoforms, we designed specific primers to distinguish the full length isoform (Exons 1-4) alone and those common to the full length and short isoforms (Exons 9-10). No expression of Whrn isoforms was observed in Whrn −/− mice ( Figure 1B). Robust expression of Whrn (Exons 1-4) mRNA was observed in DRG and SC tissue while weak expression was observed in SN tissue. Interestingly, weak expression of Whrn (Exons 9-10) mRNA was observed in DRG and SC tissue but no significant expression was observed in SN tissue. Actin (Exons 2-4) mRNA expression was used as a control for total RNA present. Relative expression was quantified as a ratio of Whrn to Actin mRNA between three total RT-PCR analyses ( Figure 1C). After finding Whrn mRNA expression in the wild-type CNS and PNS neuronal tissues, we next sought to determine its protein expression in P30 Whrn +/+ versus Whrn −/− mice by immunoblot analysis. In order to pursue these experiments, we generated several antibodies and affinitypurified one to the non-domain encoding, c-terminal region (aa699-804) of Whrn ( Figure 1A, red bar). After affinity-purification, wild-type (+) and Whrn knockout (−) lysates derived from DRG, SN, SC, and whole eye were immunoblotted ( Figure 1D, upper blots). Whole eye lysates were used as a positive control for protein expression based on previous reports [17]. A 110 kDa Whrn band representing the full length protein was present in DRG, SC, and whole eye lysates. No 110 kDa Whrn band was observed in Whrn −/− lysates or in wild-type SN lysate. Each tissue type showed similar total protein levels between genotypes based on total Tubulin on immunoblots ( Figure 1D, lower blots). Having confirmed Whrn −/− tissues were deficient in Whrn mRNA and protein, we next sought to determine if loss of Whrn resulted in altered protein expression of known nodal, paranodal, and juxtaparanodal proteins. We prepared spinal Figure 1 Whirlin (Whrn) is a PDZ-containing protein expressed throughout the central and peripheral nervous system. A. Schematic showing the relative organization of the twelve exons which make up the Whrn full-length sequence including untranslated exon regions (white boxes) and coding sequence (alternating grey boxes). Whrn' s second, short isoform begins with an alternative transcriptional start site (asterisks). Both variants contain PDZ-domains (yellow boxes) and a Proline-rich domain (blue box). A red rectangle highlights the region used for antibody creation (RbWhrn349). B. RT-PCR analysis shows absence of any Whrn transcripts in homozygous Whrn exon 1 knockout mice from dorsal root ganglia (DRG, peripheral neuronal nuclei), sciatic nerves (SN, peripheral glial nuclei), and spinal cords (SC, combination neuronal/glial nuclei). mRNA transcripts were reverse transcribed and amplified using primers located on Whrn exon 1 and 4 (top panel), Whrn exon 9 and 10 (middle panel), or actin (bottom panel). C. Relative quantification of Whrn mRNA from Figure 1B expressed as a ratio of Whrn to actin band intensity. D. Immunoblots of 110 kDa Whrn protein band in wild-type and Whrn mutant DRG, SN, SC, and whole eye using affinity-purified RbWhrn349 antibody. αTubulin served as a loading control. E. Immunoblots of wild-type and Whrn knockout mutant 4, 6, 8, and 16-week-old spinal cord lysates. The expression profile using various myelinated axonal domain markers includes Caspr, Neurofascin (186 and 155), 4.1B, Caspr2, as well as CASK. αTubulin served as a loading control. F. Immunofluorescence of teased sciatic nerve fibers from wild-type (upper panel) and Whrn knockout mice (lower panel) mice. Neurofascin (NFCt, red) and paranodal Caspr (green) reveal paranodal compaction is disrupted in Whrn knockout fibers. Note NFCt (red) detects both paranodal NF155 and nodal NF186 isoforms. cord lysates from 4, 6, 8, and 16 week-old wild-type (+/+) and Whrn knockout (−/−) mice and immunoblotted for Whrn (110 kDa) ( Figure 1E). Whrn protein expression levels were similar from 4 to 16 weeks in wild-type mice while no expression was observed in the Whrn −/− at any time point. Next, we immunoblotted for the paranodal protein Caspr (190 kDa) and found similar expression levels between 4-16 week-old in both Whrn +/+ and Whrn −/− mice. Using a pan-Neurofascin-c-terminal (NFct) antibody, we observed similar levels of nodal Neuro-fascin186 and paranodal Neurofascin155 protein expression in wild-type and Whrn knockout mice across 4-16 weeks of age. Examination of neuronal cytoskeletal protein 4.1B and juxtaparanodal protein Caspr2 shows steady protein expression levels in wild-type and Whrn knockout mice from 4-16 weeks of age. Finally, previous reports showed in vivo interaction between rat Whrn/ CIP98 and calmodulin-dependent serine kinase (CASK) [27], a synaptic organization protein. After immunoblotting for CASK, no difference in its protein level was observed across Whrn genotype or ages from 4-16 weeks. Total protein levels in spinal cord lysates were consistent across the ages and genotypes based on Tubulin levels ( Figure 1E). In summary, these developmental expression profiles suggest that loss of Whrn expression does not affect the overall expression or stability of other axonal domain markers.
Whrn knockout mice reveal a quantifiable paranodal compaction phenotype in peripheral myelinated axons To determine the effects of Whrn loss on axonal domain organization, we examined well-characterized nodal, paranodal, and juxtaparanodal markers by immunofluorescence in wild-type and Whrn −/− mice. Extensive and repeated immunostaining with our Whrn antibody revealed no consistent localization or differences in Whrn localization between wild-type to Whrn −/− fibers (data not shown). Compared to wild-type, the most striking observation in teased Whrn −/− sciatic nerve fibers was the spring-like separation ( Figure 1F) of the paranodal axo-glial septate junction (AGSJ) loops beginning along the paranodaljuxtaparanodal border as observed by Caspr and NF155 immunostaining. Irregular paranodal compaction is rarely observed in wild-type fibers, so we sought to quantify the overall observation of this phenotype in Whrn +/+ and Whrn −/− fibers. We utilized a blinded counting strategy to count spring-like phenotypes from wild-type (N = 4676) or Whrn −/− (N = 2798) Caspr-stained paranodes. A statistically significant difference (student t-test p = 0.02) was observed in the percentage of irregular paranodal compaction at 0.3% (SEM=0.07%) in wild-type fibers compared to 1.8% (SEM=0.45%) in Whrn knockout mice. Having confirmed the significance of this springlike paranodal phenotype in 7 week-old Whrn knockout fibers, we expanded our analysis to 4-8 week-old time points.
Sciatic nerves from Whrn +/+ or Whrn −/− littermate mice at 4, 6, and 8 weeks, were immunostained with antibodies against nodal, paranodal, and juxtaparanodal markers ( Figure 2). In the 4 week-old wild-type sciatic nerve (Figure 2Aa To determine if such peripheral nerve phenotypes could be the result of differences in inner mesaxons, we immunostaining 7 week-old wild-type and Whrn knockout fibers with MAG but observed no striking phenotypic differences in localization (data not shown). In summary, the paranodal and juxtaparanodal regions displayed phenotypes that suggest that normal compaction of the paranodal loops fails to occur in Whrn knockout mice at 4, 6, and 8 weeks of age.
During initial quantification we observed that larger diameter myelinated fibers had proportionally more paranode compaction defects compared to thinner caliber myelinated fibers, so we imaged and assembled twenty Caspr-stained confocal images for each genotype and time point (4, 6, and 8 week-old) to assess any subtle, sub-micron paranodal changes by light microscopy due to Whrn loss. Next, we measured various dimensions of the paranode ( Figure 2I). Note images collected were from~10 micron caliber myelinated neurons. Using Zeiss software, micron measurements were recorded, tabulated, and reported as averages with standard deviations for the nodal gap ( Figure 2J), the paranodal diameter ( Figure 2K), paranodal width ( Figure 2L), as well as the percentage of paranodes which display paranodal compaction abnormalities ( Figure 2M). The Figure 2 Loss of Whirlin in the peripheral nervous system results in disrupted paranodal compaction. A-H. 4, 6, 8-week-old teased sciatic nerve fibers either wild-type (Aa-Ad, Ca-Cd, Fa-Fd) or Whrn knockout (Ba-Bd, Da-Dd, Ea-Ed, Ga-Gd, Ha-Hd) immunostained against K v 1.2 (Aa-Da, Fa, Ga, red), NFCt (Ea, Ha, red), Caspr (Ab-Hb, green), NF186 (Ac-Dc, Fc, Gc, blue), AnkG (Ec, Hc, blue), and merged images (Ad-Hd). In all Whrn mutant panels, Caspr (Bb, Bd; Db, Dd; Eb, Ed; Gb, Gd; Hb, Hd, green) and paranodal NF155 (NFCt) (Ea, Ed; Ha, Hd, red) fail to compact properly at the paranodes. Nodal NF186 or AnkG are not affected (Ac,d-Hc,d, blue). Scale bars (Ad-Hd) = 5 μm. I. Sample image shows parameters of various domain measurements in (nodal gap in white, paranodal diameter in blue, paranodal width in red, and counting of springlike phenotype in purple) using~10 micron caliber, Caspr-immunostained wild-type and Whrn −/− fibers. J-L. No statistically significant differences were observed comparing 4, 6, and 8-week-old wild-type and mutant fibers with concern to nodal gap (J), paranodal diameter (K), or paranodal width (L) (N=20 for each genotype/age combination). Note the greater percentage of paranodes with compaction issues in mutant fibers (M, light purple bars) likely contributes to the increased deviation in paranodal widths (L, light red bars).
Whrn knockout mouse sciatic nerve and spinal cord myelinated fibers display paranodal compaction abnormalities throughout development To expand on the characterization of Whrn loss with respect to myelinated domain organization, we examined a larger developmental window from postnatal ages 2 weeks to 1 year. Wild-type and Whrn knockout sciatic nerves revealed the following percentages of Caspr-stained phenotype-positive paranodes ( Figure 2): 1.5% and 2.7% at 10 weeks, 0.9% and 1.8% at 20 weeks, 0.8% and 1.5% at 30 weeks, 0.5% and 1.4% at 40 weeks, and 0.7% and 1.5% at 1 year, respectively. Similar to the 4-8 week studies (Figure 2), we immunostained fibers with nodal, paranodal, and juxtaparanodal markers, as well as the axonal cytoskeletal protein markers 4.1B and heavy chain Neurofilament (Nfl-H) given Whrn's known cytoskeletal scaffolding role. We observed no differences in nodal formation using NF186 (Figure 3Ac (Figure 3Ig) which was not observed in the wild-type fibers (Figure 3Hg).
We also sought to examine the effects of Whrn loss on axonal domain organization in the central nervous system (CNS). Utilizing white matter tracts in the spinal cord, we were able to identify by immunostaining subtle but consistent differences in paranodal compaction. As in the PNS (Figure 3), we stained longitudinal spinal cord sections with nodal, paranodal, and juxtaparanodal markers (Figure 4). Wild-type and Whrn knockout spinal cord fibers revealed the following percentages of Casprstained phenotype-positive paranodes ( Figure 3): 0.6% and 1.7% at 10 weeks, 1.2% and 3.4% at 20 weeks, 1.3% and 2.5% at 30 weeks, and 1.0% and 2.4% at 40 weeks. No differences in nodal organization were observed (

Whirlin knockout mice have cerebellar Purkinje cells with bead-like, axonal swellings
To determine the effects of Whrn loss on cerebellar Purkinje cell morphology, we immunostained cerebellar slices from 6 week-old wild-type, Whrn knockout, and double Whrn and 4.1B [14] null animals ( Figure 5). Given that Caspr [9] and CGT [28,29], two genes critical for formation of a proper paranode, display Purkinje axonal swellings and cytoskeletal disorganization, we were curious as to the effects of the combined loss of Whrn and 4.1B. While no obvious differences in localization was observed using our Whrn antibody (data not shown) in any of these genotypes, we immunostained against 4.1B (Figure 5Aa (Figure 5Ab,c,e,f ). The secondary loss of 4.1B protein resulted in more swellings observed with Calb (Figure 5Cb,c-5Db,c vs. 5Bb,c, white arrowheads) and MBP (Figure 5Ce,f-5De,f vs. 5Be,f, white arrowheads) when compared with Whrn null animals alone. In summary, the cytoskeletal elements Whrn and 4.1B likely have an assistive role in preventing cytoskeletal accumulation and disorganization within cerebellar Purkinje cell axons in a similar phenotypic manner to Caspr null mice [9].

Ultrastructural abnormalities in Whrn knockout sciatic nerve, spinal cord fibers, and cerebellar Purkinje axons
To further understand Whrn's role in myelinated axons, transmission electron microscopy was performed in order to examine the ultrastructural architecture in myelinated axons in the PNS ( Figure 6) and CNS (Figure 7) of 7 week-old and 3 month-old wild-type and Whrn −/− mice. Low-magnification, cross-section electron micrographs of wild-type ( Figure 6A) and Whrn knockout ( Figure 6B, C) myelinated sciatic nerve fibers showed the typical organization of tightly bound, electrondense myelin wraps around the internodal region of the axonal membrane. Accumulation of mitochondria and lipid vesicles ( Figure 6B, C vs. 6A and 6E, F vs. 6D, flat arrowheads) in the internodal regions was clearly observed in Whrn knockout animals compared to wild-type sciatic nerve fibers. Given the potential role of the mesaxon in the observed light microscope phenotype [15], we found no striking differences in the ultrastructural organization or arrangements of the inner mesaxon along the internodal region at 7 weeks or 3 months of age (data not shown). Consistent with our immunostaining data, no obvious differences in nodal organization were observed in either genotype ( Figure 6G vs 6H,I). Higher magnification along the wild-type paranodal region ( Figure 6J, concave arrowheads) revealed the hallmark electrondense AGSJs formed between the glial paranodal loops and axolemma and accompanying parallel arrays of cytoskeletal elements in the axon. In contrast, the Whrn knockout paranodal region of 7 week-old ( Figure 6K) and 3 month-old ( Figure 6L) displayed poorly defined but present AGSJs ( Figure 6K,L, concave arrowheads), less organized neurofilaments and microtubules, and consistent accumulation of mitochondria and lipid vesicles ( Figure 6H, 6I, 6L, flat arrowheads) in the paranodal Figure 6 Ultrastructural examination of Whirlin knockout sciatic nerves reveals organelle accumulation and cytoskeletal disruption. Low magnification electron micrographs through the internodal regions of sciatic nerves in wild-type (7 week-old) and Whrn knockout mice (7 week-old, 3 month-old) in cross section (A vs. B, C) and longitudinal orientations (D vs. E, F). Overall cellular organization between Whrn knockout and wild-type sciatic nerve fibers is conserved with tightly compacted myelin around each axon. Low magnification, longitudinal electron micrographs through the nodal and paranodal regions of sciatic nerves are presented for wild-type (7 week-old, G) and Whrn knockout mice (7 week-old, H; 3 month-old, I). At a higher magnification, the wild-type (J) paranodal loops have clearly defined characteristic transverse, electron-dense septa (concave arrowheads) and parallel arrays of cytoskeletal elements. In contrast, Whrn mutant paranodal septa (K, L, concave arrowheads) are less definitive and fuzzy with associated accumulation of organelles (flat arrowheads), particularly mitochondria and transport vesicles. Scale bars: A-I, 2 μm, J-L, 400 nm. Figure 7 Ultrastructural examination of Whirlin knockout central nervous system tissues reveals organelle accumulation in myelinated axons. Electron micrographs of internodal cross sections (A-C) and nodal-paranodal longitudinal (D-F) regions in spinal cords from wild-type (7 week-old, A, D) and Whrn knockout mice (7 week-old, B, E; 3 month-old, C, F). Overall cellular organization is conserved between wild-type and Whrn knockout and sciatic nerve fibers with tightly compacted myelin around each axon. Enrichment of mitochondria (flat arrowheads) and occasional myelin ruffling is observed in Whrn knockout mice compared to wild-type. At higher magnification, the wild-type (G) paranodal loops have characteristic electron-dense septa (concave arrowheads) and parallel arrays of cytoskeletal elements. In contrast, Whrn knockout CNS fibers accumulate organelles (flat arrowheads), particularly mitochondria and transport vesicles, and have paranodal septa (H, I, concave arrowheads) that are less defined. Also, electron micrographs from mice show cerebellar Purkinje myelinated axons (J-L) running through the granular layer. This region reveals axonal swellings filled with densely-packed organelles (flat arrowheads), particularly mitochondria and vesicles, in 7 week-old (K) and 3 month-old (L) Whrn knockout animals compared to 7 week-old wild-type (J). Scale bars: A-F, 2 μm, G-I, 400 nm, J-L: 1 μm. region. The ultrastructural phenotypes displayed at both 7 weeks and 3 months of age suggest that Whrn is important for the long-term maintenance and the overall structure of the myelinated axons.
In the central nervous system, low-magnification, electron micrographs of wild-type ( Figure 7A) and Whrn knockout ( Figure 7B, C) myelinated fibers showed slight differences in spinal cord cross-sections. Like the peripheral nerves, the overall organization between glial cell and neuron remained similar between wild-type and Whrn −/− fibers. However, mitochondria were slightly more abundant in Whrn knockout fibers. No obvious ultrastructural differences were observed in the node of either genotype ( Figure 7D-F), but greater accumulation of mitochondria (flat arrowheads) and lipid vesicles was observed in the paranodal and juxtaparanodal regions. Higher magnification along the paranodal region ( Figure 7G-I, concave arrowheads) revealed the expected electron-dense AGSJs formed between myelin loops and the axolemma. Like in the PNS, the Whrn knockout paranodal region of 7 week-old ( Figure 7H) and 3 month-old ( Figure 7I) displayed poorly defined but present AGSJs (concave arrowheads) and accumulation of mitochondria and lipid vesicles ( Figure 7H, flat arrowhead) in the paranodal region. Examining the cerebellum, we observed Purkinje axon swellings in Whrn knockout fibers ( Figure 7K-L). Low-magnification electron micrographs of 7 week-old wild-type ( Figure 7J) and Whrn knockout at 7 weeks ( Figure 7K) and 3 months ( Figure 7L) of age showed striking differences in Purkinje axon myelinated fibers in the cerebellum granular layer. This region shows axonal swellings filled with densely-packed organelles, particularly mitochondria and vesicles ( Figure 7K-L, flat arrowheads). Taken together, the ultrastructural analyses of Whrn knockout mice demonstrate that Whrn is critical for the stability of paranodal organization, proper axonal cytoskeletal arrangements, and prevention of sub-cellular organelle accumulation in myelinated axons.

Conclusions
Cellular and molecular interactions between neurons and glia establish and stratify the numerous tasks of the nervous system. In particular, linkage of cellular membranes with the underlying cytoskeleton via cytoskeletal linker proteins helps maintain the specialized cellular arrangements necessary to glial and neuronal function. The potential of Whrn to link plasma membrane proteins with multiple cytoskeletal protein partners has been well established [16,17,[19][20][21][22][23][24], yet a precise role for Whrn in the central or peripheral nervous system or even more specifically in myelinated neurons has not been examined. In myelinated axons, axonal membrane proteins like Caspr and Caspr2 help stabilize domain organization through linkage of 4.1B to the underlying cytoskeleton. This loss of organization at both the light and electron microscopy level is readily apparent in mutant mice lacking Caspr [7], Caspr2 [12], and 4.1B [14]. Here we report that Whrn knockout animals reveal defective cytoskeletal organization and accumulation of organelles in the myelinated fibers. Our phenotypic analyses of Whrn knockout mice demonstrate that loss of Whrn disrupts proper paranodal compaction and long-term stability of the myelinated axons throughout development.

Whrn alternative splicing and phenotypes
To understand the function of Whrn in the nervous system, one must be considerate of Whrn mRNA splice variants and rule out confounding mouse genetic strain differences. Currently, two mutant mouse strains exist for Whirlin: the whirler (Whrn wi/wi ) mouse has a spontaneous genomic deletion of exons 6-9 while the Whrn (Whrn −/− ) knockout mouse [17] has a targeted exon 1 deletion. We examined the localization of several myelinated axon markers (Caspr, 4.1B, K v 1.2, and NF186) in Whrn wild-type mice from each background strain and observed no disruptions in the localization of these proteins or the morphology of the paranodes (data not shown). Our initial studies comparing Whrn −/− to Whrn wi/wi mouse strains suggested myelinated domain organization, early paranodal disorganization, and nerve conduction (~30 m/s) was indistinguishable between each mutant mouse line. Given the novelty of the phenotype, we additionally substantiated no mouse strain effects after back-crossing to isogenic C57BL6 mice for two generations. Given the consistency of the mutant phenotype regardless of background mouse genetics, we were assured the phenotype was attributed to loss of Whrn function and not alternative Whrn splice variants or mouse genetic background variation.

Scaffolding by Whrn and previously established protein networks may underlie paranodal compaction and stabilization
Whrn has several established roles derived from its complex and numerous interactions with other protein partners within the ear and eye. This complexity comes from identifying and comparing the human and mouse Whrn mutants and splice variants. The significant bulk of Whrn research has been performed in the ear and eye since human WHRN mutations contribute to a subset of Usher syndromes. In the eye, full-length Whrn colocalizes with the transmembrane proteins Usherin and VLGR1 at the periciliary membrane complex in photoreceptors [17]. Studies demonstrate the two Nterminal PDZ domains of the full-length Whrn isoform are responsible for this interaction. In the ear, the shorter Whrn isoform has a more significant role since mutation and/or loss of Whrn's C-terminus correlates to poorer hearing deficits compared to Whrn's N-terminus. The Whrn short isoform interacts through its Prolinerich domain and last PDZ domain with Myosin XVa and Mpp1/p55. Additional reports demonstrate Whrn protein expression in the mouse cerebrum, cerebellum, and brainstem [25], rat cerebellum [27], and in the Drosophila central nervous system [26]. Taken together, Whrn has a well-documented history of shared proteinprotein interaction across multiple model systems and cells.
We propose that Whrn as a cytoskeletal scaffold crosslinks a subcellular axonal meshwork to stabilize the paranode, juxtaparanode, or both, and that in the absence of Whrn, subcellular compaction of the paranode and organization of underlying microtubules, neurofilaments, cause cytoskeletal disorganization leading to an accumulation of mitochondria, and lipid vesicles along myelinated axons. Here we propose the potential sites of interaction for Whrn within these regions given its established protein-protein interactions in vivo. Within the paranode, the intracellular c-terminus of Caspr contains a SH3 domain, a potential site of interaction with Whrn's Proline-rich domain. The paranode and juxtaparanode are also enriched for 4.1B, a known protein partner in the ear stereocilia for Mpp1/p55 which interacts with Whrn [22]. Caspr2 also contains a PDZbinding motif which could potentially interact with one of Whrn's PDZ domains [12]. Finally, the c-terminus of Whrn has the potential for self-oligomerization [20], allowing for even more complex networks of proteinprotein interaction within the myelinated axon.
Domain organization in myelinated axons is a complicated developmental process, culminating from intrinsic and extrinsic cellular factors. Here we demonstrate that Whrn expression is important for proper axonal domain organization and expand the role of Whrn outside the ear and eye. The phenotypes observed in the myelinated axons highlight that the paranodal-juxtaparanodal interface represents a substantial region for insight into paranodal stabilization and potential interaction with the axonal cytoskeleton. In light of recent evidence of 4.1G's role at the internode and mesaxon, this paranodal-juxtaparanodal interface may represent an important subdomain in the study of myelinated fibers. Whrn's protein domains have the potential to stabilize the paranodal myelin loops and associated cytoskeleton through direct or indirect interactions with Caspr, 4.1B, or other unidentified cytoskeletal proteins. These observations are correlated using several techniques including biochemistry, light and electron microscopy. Our observed paranodal phenotypes are consistent throughout development (2 wk-1 yr) and similar between central and peripheral nervous systems. One final important consideration about cytoskeletal linker proteins, both Whrn and 4.1B null mice have no statistical difference in conduction velocities in sciatic nerves compared to wild-type mice, despite having clear paranodal instability in Caspr-stained myelinated fibers. Such data suggest cytoskeletal linker proteins may be functionally redundant with respect to myelinated domain organization and may require secondary or tertiary genetic ablations to achieve any measurable electrophysiological effects. To this point, we observed the increase in Purkinje axonal swellings in the double Whrn; 4.1B null mouse cerebellum compared to the single Whrn knockout or wild-type mouse cerebellum. In conclusion, our observations indicate Whrn acts as a cytoskeletal scaffolding protein that is essential for proper paranodal compaction and stabilization of the axonal cytoskeleton for long-term health of myelinated axons.

Animals
Whrn exon 1 homozygous mutants used were obtained from Dr. Jun Yang's lab (University of Utah, Salt Lake City, Utah, 84132). The mice were backcrossed to C57BL6 mice (JAX Laboratories #000664, Bar Harbor, Maine) for two generations and maintained as heterozygous Whrn +/− breeding stocks. All animal experiments were performed according to Institutional Animal Care and Use Committee approved guidelines for ethical treatment of laboratory animals at the University of North Carolina at Chapel Hill and the University of Texas Health Science Center at San Antonio.

Generation of Whrn antibody
We generated rabbit, guinea pig, and rat polyclonal anti-Whrn antibodies similar to previous literature [23]. A full length Whrn mouse cDNA construct in pcDNA3.1 was obtained from Dr. Jun Yang's lab. Regions encoding amino-acid residues 220-326 and 699-804 of mouse Whirlin (Genbank: NP_001008791.1) were subcloned individually into pGEX4T1 and expressed in Escherichia coli (BL21; Stratagene, La Jolla, CA). Fusion proteins were isolated by incubating with Glutathione Sepharose 4 Fast Flow (GE Healthcare, Sweden). Each fusion protein was used to immunize a rabbit, guinea pig, or rat (Cocalico Biologicals Inc, Reamstown, PA). cDNAs encoding amino acids 220-326 or 699-804 of mouse Whrn were also introduced into pMAL-c2x (New England Biolabs, Beverely, MA), transformed into E. coli (DE3 BL21; Stratagene) and induced to express the corresponding maltose binding (MBP) fusion protein. The expressed MBP-fusion proteins were purified using amylose resin (New England Biolabs, Beverely, MA) and then linked to a NHS-activated Sepharose 4 Fast Flow (GE Healthcare, Sweden). Only antisera from the immunized rabbit #349 (RbWhrn349) was affinity purified using the corresponding MBP-Whrn fusion protein.

Immunostaining
Briefly, sciatic nerves were removed from anesthetized littermate wild-type and Whrn mutants of either sex and fixed in 4% paraformaldehyde in PBS for 15-30 min. The nerves were washed with PBS three times (10 min each) and stored at 4°C until teased. The nerves were teased into individual fibers in PBS, mounted on glass slides, and dried overnight at room temperature. Fibers were either immediately used for immunostaining or stored at −80°C until needed. Teased nerve slides were submerged in acetone (methanol instead for anti-MBP staining) at −20°C for 20 min then washed with PBS, followed by immunostaining [4]. For spinal cord sections and cerebellar sections, wild-type and mutants were deeply anesthetized and intra-cardially perfused with PBS followed by ice-cold 4% paraformaldehyde in PBS. The spinal cord or cerebellum was dissected out and post-fixed in 4% paraformaldehyde overnight at 4°C. The tissues were rinsed with PBS and sectioned to 30 um using a Vibratome (Leica). The sections were then immediately immunostained as previously described [9,10]. Primary antibodies for immunostaining were used at the following concentrations overnight at 4°C: RbCaspr @1:500, GPNF186 @1:400, MsIgG2b-K v 1.2 @1:200, GPNFCt @1:400, GP-beta-IVspectrin @1:1000, GP4.1B @1:10000, and MsIgG-Nfl-H @1:1000, MsIgG1-Calb @1:1000, and MsIgG1-MBP @1:200.

Immunoblotting
Sciatic nerves and dorsal root ganglia from littermate wild-type and mutants of either sex were excised and processed using a glass homogenizer in ice-cold lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM EDTA, 1% Triton X-100, 1% SDS, and a protease mixture tablet). The lysate was incubated for 30 min on ice and then centrifuged at 16,000×g for 20 min at 4°C. The sciatic nerve or dorsal root ganglia supernatant was saved for further processing. Spinal cords from littermate wild-type and mutants of either sex were excised and either directly processed or frozen at −80°C. Spinal cords were homogenized using a glass mortar and pestle on ice with lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM EDTA, 1% Triton X-100, and a protease inhibitor mixture tablet) and incubated for 30 min on ice with occasional trituration. The homogenate was centrifuged at 1000xg for 10 min at 4°C. The supernatant was collected and subjected to an additional centrifugation at 100,000 × g for 30 min at 4°C. The resulting second supernatant was collected and saved for further processing. Protein concentrations of final lysates were determined using the Lowry assay (BC assay; Bio-Rad). Lysates were resolved by SDS-PAGE and transferred onto nitrocellulose membranes, followed by immunoblotting procedures described previously [4]. Primary antibodies for immunoblotting were used at the following concentrations for 1 hr at room temperature: Affinitypurified RbWhrn349 @ 1:1000 (overnight at 4°C), GP Caspr @ 1:2000, GPNFCt @ 1:2000, Rb4.1B @ 1:50000, Ms IgG1-CASK @ 1:50000, RbCaspr2 @ 1:50000, RbTubulin @ 1:2000 (overnight at 4°C).

Image analysis and software
Confocal images were captured with a Zeiss LSM510 microscope. Scanning parameters were optimized for wild-type tissues and maintained for scanning the mutant tissues. Immunofluorescence images for sciatic nerves and spinal cords are composite projections from Z stacks of three to six sections (0.6um scan step) or stacks of ten to twenty sections (0.6um scan step) for cerebellar slices. Software used for assembling figures included Zeiss LSM Image Browser (v4.2), ImageJ (v1.47d), GIMP (v2.82), and OpenOffice (v3.4.1).

Quantification of phenotype and statistics
For the initial quantification, we utilized a blinded counting strategy to best estimate the spring-like phenotype. Teased sciatic nerves were prepared from wild-type or Whrn −/− mice. One individual prepared all teased slides and randomly assigned a number to each slide. Once completed, the individual compiled a table of genotypes matched to assigned numbers. Blinded to that table, a second individual immunostained the numbered slides and counted wild-type and Whrn −/− Caspr-stained paranodes. Immunostained paranodes were counted under a fluorescent microscope at 40× magnification. Any paranode with 3 or more spring-like, loops were considered phenotypepositive. Data was compiled as the percentage of phenotype-positive paranodes out of total paranodes. Phenotype percentages per slide were matched to genotype and then a final average phenotype percent for wild-type or Whrn −/− fibers was calculated as well the standard error of the mean (SEM). A standard t-test was used to calculate the statistical significance (p-value) between the percent for 7 week-old wild-type or Whrn −/− fibers. Similar phenotype quantitation was applied to 10, 20, 30, and 40 wk old sciatic nerve and spinal cords. All measured data was tabulated in Microsoft Excel. A pvalue of 0.05 was considered to indicate a significant difference between groups.
For secondary quantitation of paranodal parameters in 4-8 week-old animals, Caspr-stained paranodes were imaged using a confocal LSM510 microscope and accompanying Zeiss Image software. A composite projection of Z-stacks was performed for each image. Large-caliber sciatic nerve images (~10um total diameter glial-edge to glial edge) were analyzed given the initial observation of more phenotype-positive paranodes in large vs. small caliber axons. Twenty wild-type and twenty Whrn −/− paranode images were selected for each time point (60 total images). Using Zeiss software tools, each paranode was measured in microns for nodal gap (distance between nodal-paranodal boundaries in white), paranodal diameter (distance across axon caliber in blue), paranodal width (distance from nodal-paranodal boundary to paranodaljuxtaparanodal boundary in red), and phenotype percentage (a paranode was considered phenotype-positive if the purple line from the paranodal-juxtaparanodal boundary crossed a Caspr AGSJ line three times (i.e. 1.5 circular myelin wrapping loops) and phenotype-negative if less than three times). All measured data was tabulated in Microsoft Excel. The average and standard deviation of each age and genotype was graphed for nodal gap, paranodal diameter, and paranodal width. Phenotype percentage was reported as a percent of phenotypepositive paranodes out of twenty total counted for that age and genotype.