doi:10.1093/brain/awp341 Brain 2010: 133; 389–405 | 389 BRAIN A JOURNAL OF NEUROLOGY

Novel forms of neurofascin 155 in the central nervous system: alterations in paranodal disruption models and multiple sclerosis

Anthony D. Pomicter,1 Seema M. Shroff,1, Babette Fuss,1 Carmen Sato-Bigbee,2 Peter J. Brophy,3 Matthew N. Rasband,4 Manzoor A. Bhat5 and Jeffrey L. Dupree1

1 Department of Anatomy and Neurobiology, Virginia Commonwealth University, School of Medicine, Richmond, VA, USA 2 Department of Biochemistry and Molecular Biology, Virginia Commonwealth University, School of Medicine, Richmond, Virginia, USA 3 Centre for Neuroscience Research, University of Edinburgh, Edinburgh, Scotland, UK 4 Department of Neuroscience, Baylor College of Medicine, Houston, Texas, USA 5 Department of Cell and Molecular Physiology, University of North Carolina School of Medicine, Chapel Hill, North Carolina, USA

Present address: New York University Langone Medical Center, 455 First Ave., PHL Building 863, New York, NY 10016, USA

Correspondence to: Jeffrey L. Dupree, PhD, Department of Anatomy and Neurobiology, 1101 E. Marshall Street, Box 980709, Richmond, VA 23298, USA E-mail: [email protected]

Stability of the myelin-axon unit is achieved, at least in part, by specialized paranodal junctions comprised of the neuronal heterocomplex of contactin and contactin-associated protein and the myelin protein neurofascin 155. In multiple sclerosis, normal distribution of these proteins is altered, resulting in the loss of the insulating myelin and consequently causing axonal dysfunction. Previously, this laboratory reported that mice lacking the myelin-enriched lipid sulphatide are characterized by a progressive deterioration of the paranodal structure. Here, it is shown that this deterioration is preceded by significant loss of neurofascin 155 clustering at the myelin paranode. Interestingly, prolonged electrophoretic separation revealed the existence of two neurofascin 155 bands, neurofascin 155 high and neurofascin 155 low, which are readily observed following N-linked deglycosylation. Neurofascin 155 high is observed at 7 days of age and reaches peak expression at one month of age, while neurofascin 155 low is first observed at 14 days of age and constantly increases until 5 months of age. Studies using conditional neurofascin knockout mice indicated that neurofascin 155 high and neurofascin 155 low are products of the neurofascin gene and are exclusively expressed by oligodendrocytes within the central nervous system. Neurofascin 155 high is a myelin para- nodal protein while the distribution of neurofascin 155 low remains to be determined. While neurofascin 155 high levels are significantly reduced in the sulphatide null mice at 15 days, 30 days and 4 months of age, neurofascin 155 low levels remain unaltered. Although maintained at normal levels, neurofascin 155 low is incapable of preserving paranodal structure, thus indicating that neurofascin 155 high is required for paranodal stability. Additionally, comparisons between neurofascin 155 high and neurofascin 155 low in human samples revealed a significant alteration, specifically in multiple sclerosis plaques.

Keywords: sulphatide; cerebroside sulphotransferase; myelin; paranode; Caspr

Received June 16, 2009. Revised December 14, 2009. Accepted December 16, 2009 ß The Author (2010). Published by Oxford University Press on behalf of the Guarantors of Brain. All rights reserved. For Permissions, please email: [email protected] 390 | Brain 2010: 133; 389–405 A. D. Pomicter et al.

Abbreviations: CST = cerebroside sulphotransferase; Caspr = contactin-associated protein; CNP = 20,30-cyclic nucleotide 30-phosphodiesterase; CNP-Cre;Nfascflox/flox = neurofascin conditional knockout mice; neurofascin 155H = neurofascin 155 high; neurofascin155L = neurofascin 155 low; neurofascin’ = deglycosylated neurofascin; Nfasc = neurofascin gene; PNGase F = Peptide N-glycosidase F

of myelin sheath compaction, altered paranodal loop orientation, Introduction increased nodal length (Marcus et al., 2006), and a reduction of Axonal saltatory conduction is facilitated by the myelin sheath, a nodal protein domains (Ishibashi et al., 2002). In the present lipid rich multi-lamellar wrap generated by oligodendrocytes in the study, quantitative analysis advances the previous observation of CNS. Throughout much of its length the mature sheath lacks cyto- neurofascin 155 cluster loss (Marcus et al., 2006), while western plasm resulting in compaction of the myelin membrane. At the blot analysis indicates that the loss of neurofascin 155 clusters edges of the sheath, cytoplasm is retained in sacs known as para- does not result strictly from decreased neurofascin 155 levels. nodal loops (Peters et al., 1976). These loops ‘bend’ toward the Equally important, the current study demonstrates that neurofascin axonal membrane and establish septate-like junctions that main- 155 is comprised of two forms, termed neurofascin 155 high tain clusters of voltage-gated sodium and potassium channels at (155H) and neurofascin 155 low (155L). It is shown here that the node of Ranvier and the juxtaparanode, respectively (Chiu, neurofascin 155H and neurofascin 155L are oligodendrocyte- 1980; Waxman and Richie, 1985; Chiu and Schwarz, 1987; specific products of the neurofascin (Nfasc) gene that differ with Poliak and Peles, 2003; Rasband, 2004). These specialized junc- regard to N-linked glycosylation and temporal expression. tions are also critical in tethering the myelin sheath to the axon Neurofascin 155H is a paranodal protein whose levels are reduced through trans interactions between the myelin protein neurofascin in CST knockout mice and is nearly absent by 4 months of age. In 155 and the axonal heterocomplex of contactin and contrast, levels of neurofascin 155L are not altered in the CST contactin-associated protein (Caspr) (Rios, 2000; Tait et al., knockout mice as compared to littermate wild-type animals. 2000; Bhat et al., 2001; Boyle et al., 2001; Charles et al., With age in both genotypes, neurofascin 155L becomes more 2002). Although the specifics of the interactions between neuro- abundant and less glycosylated. However, its function remains fascin 155 and contactin/Caspr remain unresolved (Gollan et al., unknown. Evaluation of human brain samples showed a multiple 2003; Bonnon et al., 2007), it is clear that mice lacking either sclerosis plaque-specific alteration between neurofascin 155H and contactin, Caspr or neurofascin 155 fail to form the paranodal neurofascin 155L following deglycosylation. These results indicate septate-like junctions, do not properly establish paranodal and that, in both mouse and human samples, the expression pattern of juxtaparanodal protein domains (Bhat et al., 2001; Boyle et al., neurofascin 155H and neurofascin 155L is altered coincident with 2001; Sherman et al., 2005; Zonta et al., 2008, Pillai et al., 2009), paranodal decay. and exhibit compromised paranodal loop orientation (Bhat et al., 2001). Whereas compromised paranode formation and stability are consistent with the absence of either contactin, Caspr or neu- rofascin 155, it is more difficult to reconcile the loss of paranode Materials and methods structure in mice that lack netrin-1 or deleted in colorectal cancer (DCC) (Jarjour et al., 2008), myelin and lymphocyte (MAL) pro- Animals tein (Schaeren-Wiemers et al., 2004), galactocerebroside and sul- Wild-type and CST knockout (Honke et al., 2002), also known as phatide (Dupree et al., 1998) or sulphatide alone (Honke et al., sulphatide null, mice were genotyped as described (Shroff et al., 2002). Paranodal deterioration, reminiscent of these mutant mice, 2009). Other animals used in this study include Caspr knockout also occurs in multiple sclerosis, prior to demyelination, as para- (Bhat et al., 2001), neurofascin conditional knockout, referred to as flox/flox nodal loops face away from the axon (Suzuki et al., 1969), para- CNP-Cre;Nfasc , (Pillai et al., 2009), and littermate wild-type mice. Sprague-Dawley rats were purchased from Charles River nodal clustering of Caspr and neurofascin 155 is reduced Laboratories (Wilmington, MA, USA). Animal use was conducted in (Wolswijk and Balesar, 2003; Howell et al., 2006), and juxtapar- accordance with the guidelines from the National Institutes of Health anodal proteins invade the paranode (Howell et al., 2006). and approved by the Virginia Commonwealth University’s Animal Care Interestingly, in addition to reporting a possible reduction of sul- and Use Committee. phatide in normal-appearing white matter in patients with multiple sclerosis, Marbois et al. (2000) showed a significant alteration in the sulphatide species present in multiple sclerosis. These findings Antibodies emphasize the need to elucidate further the role of sulphatide in The neurofascin antibodies used in this study have been previously the regulation of paranode stability. characterized and are known as FIGQY (Ogawa et al., 2006) and The sulphatide null mice, which contain a disruption in the NF-C1 (Tait et al., 2000). For immunocytochemical labelling of gene that encodes the enzyme cerebroside sulphotransferase frozen tissue sections, FIGQY and NF-C1 were used at 1:200 and (CST) (Honke et al., 1996, 1997), initially form abundant, com- 1:1000 dilutions, respectively. For western blot analysis, both antibo- pacted myelin sheaths and structurally normal nodes and para- dies were diluted 1:2000. These antibodies are directed against distinct nodes; however, with age the CST knockout mice reveal a loss epitopes of the cytoplasmic tail common to neuronal neurofascin Discovery of Nfasc155H and Nfasc155L Brain 2010: 133; 389–405 | 391

186, oligodendrocyte neurofascin 155, and the uncharacterized pro- aprotinin] followed by centrifugation for 1 min at 16 100g in a table tein neurofascin 140. Other antibodies used in this study include: top microcentrifuge. Supernatant concentrations were determined anti-Caspr (1:1000; Bhat et al., 2001); anti-pan voltage gated using the Micro BCA Protein Assay Kit (Pierce, Rockford, IL, USA). sodium channel (1:200; Sigma-Aldrich, St Louis, MO, USA); anti- Twenty–thirty micrograms of spinal cord or 40–50 mg of brain proteins extracellular signal-regulated kinase 2 (1:10 000; Santa Cruz in Laemmli sample buffer (Bio-Rad, Hercules, CA, USA) containing Biotechnology, Santa Cruz, CA,USA); anti-myelin basic protein 5% b-mercaptoethanol were subjected to SDS–polyacrylamide gel (1:100; Millipore, Billerica, MA); anti-20,30-cyclic nucleotide electrophoresis for 30 min at 70 V followed by either 40 or 150 min 30-phosphodiesterase (CNP; 1:1000, Covance, Emeryville, CA); horse- (see below) at 180 V using 10% Ready Gels (Bio-Rad). Precision Plus radish peroxidase conjugated secondary antibodies (1:10 000; Santa Protein Kaleidoscope Standards (Bio-Rad) allowed monitoring of the Cruz); and fluorescently conjugated secondary antibodies (1:200; gel run and transfer. For the extended electrophoresis protocol, the Vector Laboratories, Burlingame, CA, USA). electrophoresis was allowed to proceed until the 75 kDa standard band reached the bottom of the gel (150 min) and transferred to nitrocel- lulose for 2 h at 100 V. The membranes were then blocked with 5% Immunocytochemistry non-fat dry milk (Bio-Rad) in phosphate buffered saline containing 0.05% Tween 20 (Sigma-Aldrich) prior to incubation with primary Cervical spinal cord samples from three wild-type and four littermate and secondary antibodies, and the same solution was utilized for anti- CST knockout mice at 15 and 30 days of age and a rat at 7 days of body incubations. Immunoreactive bands were visualized with horse- age were single, double or triple labelled as previously described radish peroxidase chemiluminescent substrate (Pierce or Millipore, (Dupree et al., 1999; Marcus et al., 2006). Billerica, MA, USA) and either an Alpha Innotech FluorChem SP imager (San Leandro, CA, USA) or radiograph film. Quantitation of Nfasc 155 localization Immunolabelled sections were analysed using a Leica TCS2 AOBS laser Glycosidase reaction scanning confocal microscope. Maximum projection images were com- Peptide N-glycosidase F (PNGase F) and related reagents were piled using eight optical sections spanning 3.6 mm and collected using obtained from New England Biolabs (Ipswich, MA, USA). Following a pin hole of one Airy unit with a resolution of 1024 1024 pixels. the manufacturer’s suggested protocol, a typical 25 ml reaction All images were collected using a PL APO 63 oil immersion objective included 50 mg of protein, 2 ml of PNGase F (1000 U, as defined lens (numerical aperture of 1.4–0.7) with a digital zoom of two result- by the manufacturer), and 1 ml of protease inhibitor cocktail (Sigma- m m ing in a microscopic field of 119 m 119 m. To ensure that similar Aldrich, P8340). Following 3 h at 37C, samples were mixed with one regions of the cervical cord were sampled, the cerebellum was used to volume of Laemmli Sample Buffer containing 5% b-mercaptoethanol discern ventral and dorsal columns. All images were collected from the and prepared for SDS–polyacrylamide gel electrophoresis as above. cervical ventral columns. The posterior portion of the cerebellum was also used as the most anterior extent of the spinal cord and thus served as the starting point for image collection. For quantitation, six Densitometry, molecular mass and maximum projection images per section were collected; neurofascin glycosylation analyses 155 paranodal clusters were counted, and the average number of clusters per image was determined for each mouse. Two sets of criteria Except where stated otherwise, protein band densities were measured were used to identify paired neurofascin 155-containing paranodes. with the AlphaEase software version 4.1.0 by Alpha Innotech First, solid rectangular clusters ranging between 5 and 15 mm with Corporation. All western blot densities were normalized to extracellular well demarcated ends were identified as a single paranodal pair (see signal-regulated kinase 2, data were analysed by a two-tailed t-test, Supplementary Fig. 2A’ and B’). Second, paired rectangular clusters and significance was defined as P50.05. The same software was (corresponding to two adjacent paranodes) separated by a small paired with Magic Mark XP Western Protein Standard (Invitrogen, region void of labelling (corresponding to the node of Ranvier) were Carlsbad, CA, USA) to determine molecular masses and shifts due to also identified as a single paranodal pair (Supplementary Fig. 2C’). deglycosylation utilizing the 220, 120, 100 and 80 kDa bands of the Unpaired clusters were not counted (Supplementary Fig. 2D’). All PNGase F treated and untreated samples. counts were statistically compared using two-tailed t-tests and signif- For the quantitative analysis shown in Fig. 8, ImageJ 1.14o was icance was defined as P50.05. utilized to determine the density of a rectangular selection encompass- ing neurofascin 155H’ and neurofascin 155L’ (neurofascin’ refers to deglycosylated neurofascin). The resulting graphical plot was observed Western blot analysis to contain either one or two peaks and the area under the curve was Animals were deeply anaesthetized with 2, 2, 2-tribromoethanol determined with the wand tool. In the case of one peak, the straight (Sigma-Aldrich); brain and spinal cord were dissected, rapidly frozen line selection function was utilized to divide the plot into the main separately in liquid nitrogen, and stored at 80C. Homogenization region corresponding to the neurofascin 155H’ band and the lesser utilized a Glas-Col (Terre Haute, IN, USA) motor-driven homogenizer, region corresponding to the shadow-like neurofascin 155L’ band. The a Potter Elvehjem mortar, and a Teflon pestle (Wheaton Industries, area under each curve was then determined as above. In the case of Inc., Millville, NJ, USA). Tissues were homogenized (5 min per brain two peaks, the wand tool was utilized to divide the plot into two and 3 min per spinal cord) on ice in modified radioimmunoprecipitation regions, corresponding to neurofascin 155H’ and neurofascin 155L’. assay buffer [137 mM NaCl, 10 mM Na2HPO4, 1.8 mM KH2PO4, For examples of this procedure and representative plots of 2.7 mM KCl, pH 7.4 with 0.1% sodium dodecyl sulphate (SDS), non-multiple sclerosis and multiple sclerosis plaque samples, see 0.5% sodium deoxycholate, 0.1% nonidet P-40] with protease inhib- Supplementary Fig. 3. The density values obtained from this analysis itor cocktail [Sigma-Aldrich P8340, comprised of 4-(2-aminoethyl)- were first utilized to create a ratio of neurofascin 155H’/neurofascin benzenesulphonyl fluoride, pepstatinA, E-64, bestatin, leupeptin and 155L’ for each sample. These ratios were then grouped as 392 | Brain 2010: 133; 389–405 A. D. Pomicter et al. non-multiple sclerosis and normal-appearing white matter versus mul- macroscopic and light microscopic levels. Macroscopic analysis tiple sclerosis plaque, and compared using a two-tailed Student’s t-test revealed well circumscribed, grey, glistening depressed plaques only with unequal variance. Multiple sclerosis sample 119 was removed in multiple sclerosis samples. Microscopic analysis was used to assess from quantitative analysis due to a Cook’s distance value 41, calcu- myelin pallor and cellularity. Using this combination of evaluations, all lated with the Predictive Analytics SoftWare Statistics 17.0 software, disease tissue samples used in this study were collected from con- indicating that a data point is an outlier and therefore a candidate for firmed multiple sclerosis cases. removal from the data set (Cook, 1977); it is noteworthy that this Extent of cellularity was used to classify multiple sclerosis plaque patient with confirmed multiple sclerosis died of complications due samples further as ‘active’ versus ‘inactive’. Active multiple sclerosis to leukaemia and suffered multiple brain haemorrhages. samples exhibited hypercellularity demonstrated by lymphocytes, The NetNGlyc Server 1.0 (http://www.cbs.dtu.dk/services/ macrophages and/or reactive astrocytes. In contrast, inactive plaques NetNGlyc/) was used to predict locations of N-linked glycosylation were identified by the absence of hypercellularity. All multiple sclerosis (PubMed accession AAL27854, rat neurofascin 155 kDa isoform) as samples were collected within the lesion site whereas normal appear- well as the likelihood that a given site is glycosylated in vivo. ing white matter samples were collected from brains with confirmed ‘Highly likely’ is defined as above 0.5 threshold, 9/9 neural networks multiple sclerosis but from regions that appeared normal based on agreeing on the outcome, and a ++ result. ‘Somewhat likely’ is defined macroscopic and microscopic evaluation. For specific age, gender, as above 0.5 threshold, 6 or more neural networks agreeing on the post-mortem harvest times and pathologic summary of each donor outcome, and a + result. Both likelihoods were considered to be poten- used in this study, see Table 1. tial sites of glycosylation. Only NXS/T sequons free of proline and For analysis within this study, non-multiple sclerosis (n = 4), normal located in the extracellular region were considered. Molecular mass appearing white matter from multiple sclerosis donors (n = 5) and mul- was calculated with the ExPASy ‘Computer pI/MW’ tool available tiple sclerosis plaque samples from active (n = 4) and inactive (n =5) at www.expasy.ch/tools/pi_tool.html. lesions were evaluated. An 2 1 cm sample was collected from the received sub-cortical white matter samples and processed for western Human samples blot and deglycosylation analyses as described above with 20–25 mg of protein per lane. Human brain samples from non-multiple sclerosis and multiple sclerosis donors (both plaque and normal-appearing white matter) were pro- vided by the Rocky Mountain Multiple Sclerosis Centre (Englewood, CO, USA). All samples were collected at time of autopsy, which was Results between one and 9 h post-mortem. At the time of death, donors ranged between 30 and 87 years of age. All samples were collected Paranodal clusters of neurofascin from sub-cortical white matter regions and immediately snap frozen in liquid nitrogen. Based on medical history and autopsy evaluation, none 155 are significantly reduced in CST of the multiple sclerosis individuals included in this study presented knockout mice any neurodegenerative disorders other than multiple sclerosis. Brains were classified as ‘multiple sclerosis’ or ‘non-multiple sclerosis’ based The NF-C1 pan-neurofascin antibody was utilized to quantify on medical history and post-mortem pathological analyses at both the paranodal clusters of neurofascin 155 in the ventral columns of

Table 1 Donor information for the Rocky Mountain Multiple Sclerosis Centre brain samples used in this study

Disease summary No. Age Gender HTTC Pathological evaluation of the brain Non-MS 2 63 – 9 Myelin stain revealed no evidence of MS Non-MS 5 55 F 2 No plaques observed in coronal sections Non-MS 27 65 M 6.5 History of myocardial infarction Non-MS 32 70 F 8 No plaques observed in multiple brain regions NAWM-MS 243a 58 F 5 Numerous shadow plaques in PVWM NAWM-MS 216a 71 F 4.5 Numerous plaques with near total myelin loss NAWM-MS 235 41 F 3 Neurons preserved, no anoxic injury; cavitation NAWM-MS 237 50 M 4 PVWM extensively demyelinated, no anoxic injury NAWM-MS 137a 30 M 5 Grossly visible plaques in multiple regions Active MS 137a 30 M 5 Grossly visible plaques in multiple regions Active MS 73 44 M 4 Myelin pallor and shadow plaques throughout Active MS 136 43 F 3.5 PVWM extensively demyelinated, hypercellularity Active MS 168 37 F 3.5 PVWM extensively demyelinated Inactive MS 119 38 F 1 12 year history of multiple sclerosis Inactive MS 164 – – 3 Axons present in demyelinated regions Inactive MS 224 87 F 4 Myelin pallor in frontal and parietal lobes Inactive MS 243a 58 F 5 Numerous shadow plaques in PVWM Inactive MS 216a 71 F 4.5 Numerous plaques with near total myelin loss Averages and totals 52.1 9 F, 4 M 4.5 18 samples 2 NA

MS = multiple sclerosis; HTTC = hours to tissue collection; PVWM = periventricular white matter; NAWM = normal appearing white matter; = data not available (NA). aTwo samples (normal appearing white matter and plaque) from the same donor. Discovery of Nfasc155H and Nfasc155L Brain 2010: 133; 389–405 | 393

section). In 15-day-old CST knockout mice, the number of para- nodal clusters of neurofascin 155 was reduced by 50% and by 30 days of age the CST knockout mice contained fewer than 10% of the number of clusters observed in littermate wild-type mice (Fig. 1B, D and E; Table 2). Similar observations were made with the FIGQY pan-neurofascin antibody (data not shown). Additionally, the 30-day-old CST knockout mice showed a preva- lence of small, unpaired clusters (Fig. 1D, Supplementary Fig. 2D’). Double labelling with a sodium channel antibody (data not shown) revealed that the unpaired clusters were at the node of Ranvier, consistent with the distribution of neurofascin 186 (Tait et al., 2000).

Levels of neurofascin 155 are significantly reduced in spinal cords of CST knockout mice The reduced paranodal clusters of neurofascin 155 observed in the CST knockout mice could result from either decreased protein expression or improper protein localization. To discern between these two possibilities, the levels of neurofascin 155 were deter- mined by western blot analysis. Figure 2A shows that, compared to littermate wild-type mice, neurofascin 155 was reduced by 40% in the spinal cords of 15- and 30-day-old CST knockout mice (Fig. 2D; Table 2). However, as shown in Fig. 2B and C, the decrease in neurofascin 155 is not accompanied by a widespread reduction of myelin proteins as neither levels nor Figure 1 Paranodal clustering of neurofascin 155 is significantly developmental expression patterns of the myelin-specific proteins reduced in the absence of sulphatide. Cervical spinal cord CNP or myelin basic protein was altered in the CST sections immunolabelled with the NF-C1 pan-neurofascin anti- knockout mice. body revealed abundant paired clusters of neurofascin (Nfasc) 155 in the myelin paranode of wild-type (WT) animals at 15 (A) and 30 (C) days of age. Note that the labelling pattern in the 15-day-old animals was frequently rectangular in shape repre- Extended electrophoresis reveals that senting immunolabelling of both neurofascin 155 in the myelin paranode and neurofascin 186 in the (axonal) node. In contrast, neurofascin 155 is comprised of two by 30 days of age, nodal labelling was less apparent while bands—neurofascin 155 high and paranodal labelling remained prevalent. In the CST knockout (KO) spinal cord sections, the number of paranodal clusters of neurofascin 155 low neurofascin 155 was reduced by 50 and 90% at 15 (B) and 30 Consistent with previous studies (Schafer et al., 2004; Sherman (D) days of age, respectively, compared to littermate wild-type et al., 2005), standard electrophoretic separation (40 min; see mice. Also, note that by 30 days of age, the prevalence of small, ‘Materials and methods’ section) of spinal cord homogenates single clusters (arrows) was increased in the CST knockout. (E) Quantitative analyses revealed that the reduction of para- from wild-type mice up to 30 days of age resulted in the detection nodal clustering of neurofascin 155 was significantly reduced at of a single band representing neurofascin 155 (Fig. 2A). However, both 15 and 30 days of age. Scale bar = 10 mm; see Table 2. extended electrophoresis of the same samples revealed two bands, P50.05. now named neurofascin 155 high (155H) and neurofascin 155 low (155L), at 30 days of age in the CST knockout mice (Fig. 3A–C) as well as in the wild-type brain (Fig. 3C) between wild-type and CST knockout mice at 15 and 30 days of age neurofascin 186 and neurofascin 140. Neurofascin 155L was not (Fig. 1; see Supplementary Fig. 1 for colour). At both ages, detected by extended electrophoresis at 15 days of age in either wild-type mice exhibited numerous paranodal clusters of neuro- fascin 155 (Fig. 1A, C and E; Table 2). While paranodal clusters of wild-type or CST knockout animals. Neurofascin 155H and neu- neurofascin 155 were also observed in the CST knockout mice, rofascin 155L were detected by the pan-neurofascin antibodies comparison with the wild-type mice indicated significant reduc- FIGQY (Fig. 3A and C, and all other figures unless otherwise tions at each age examined (for an explanation of how the clusters stated) and NF-C1 (Fig. 3B), and were present in both spinal were defined and quantified, see the ‘Materials and methods’ cord (Fig. 3A and B) and brain (Fig. 3C) homogenates. 394 | Brain 2010: 133; 389–405 A. D. Pomicter et al.

Table 2 Neurofascin paranodal clusters and protein in the CST knockout spinal cord

Age Wild-type (n) SD Knockout (n)SD P Reduction (%) Neurofascin 155 positive clusters/field 15 days 75.4 (3) 14.1 36.8 (4) 8 0.0001 51.2 30 days 118.6 (3) 51.1 9.5 (4) 9.3 0.001 92.2 Neurofascin 155 (protein), % of wild-type 15 days 100 (3) 3.90% 64.2 (4) 13.80% 0.001 35.8 30 days 100 (4) 17.00% 60.9 (2) 15.80% 0.027 39.1

Figure 3 Extended electrophoresis reveals two forms of neurofascin 155. At 15 days of age, the FIGQY (A) and NF-C1 (B) pan-neurofascin antibodies labelled one band representing neurofascin (Nfasc) 155 in both the wild-type (WT) and the CST knockout (KO) spinal cord samples. By 30 days of age, wild-type samples displayed the protein now called neurofascin 155H with a shadow-like band immediately below, while the CST knockout samples showed both neurofascin 155H and neurofascin 155L. (C) Mouse brain samples incubated with FIGQY show neuro- fascin 155H and neurofascin 155L in both wild-type and CST knockout at 30 days of age and, when compared to 15-day-old samples, suggest a developmental pattern for neurofascin 155L in wild-type as well as CST knockout mice. ERK2 = extracellular signal-regulated kinase 2.

Neurofascin 155H and neurofascin Figure 2 CST knockout spinal cords contain reduced amounts of neurofascin 155 at 15 and 30 days of age. (A) Western blot 155L are distinctly regulated through analysis showed that CST knockout (KO) mice contain 40% development less (Table 2) neurofascin (Nfasc) 155 at 15 and 30 days of age than wild-type (WT) littermates. However, no change in the To examine whether the observations of Fig. 3 were specific to levels of either the CNP monomers (B) or myelin basic proteins mouse and to determine the expression profiles of neurofascin (MBP) (C) was observed in the absence of sulphatide, demon- 155H and neurofascin 155L in wild-type animals, spinal cord strating that the reduction in neurofascin 155 is not due to a homogenates from rats ranging between 7 days and 5 months global decrease in myelin proteins. (D) Graphical representation of age were separated by the extended electrophoresis protocol of the reduction in neurofascin 155. ERK2 = extracellular signal- (Fig. 4A). Neurofascin 155H was visible at all ages, increasing in regulated kinase 2. P50.05. intensity from barely detectable at 7 days to maximal levels at 27 days of age, followed by a reduction at 3 and 5 months. Discovery of Nfasc155H and Nfasc155L Brain 2010: 133; 389–405 | 395

Figure 4 Neurofascin 155H and neurofascin 155L are developmentally regulated glycoproteins in rat spinal cord. (A) Rat spinal cord homogenates revealed low levels of neurofascin 155H at 7 days of age. Neurofascin 155H steadily increased with age, reaching a peak at 27 days of age followed by a decrease which is maintained through 5 months of age. Neurofascin 155L was first seen as a clearly delineated band at 3 months of age and remained prevalent at 5 months. Interestingly, neurofascin 155L appeared to form gradually from the shadow-like band below neurofascin 155H. (B) Following PNGase F treatment neurofascin 155L’ was observed in 3 and 5 month samples (as expected from Fig. 4A) as well as at 14, 21 and 27 days of age, with increasing abundance with increasing age. Neurofascin 155H’ was barely visible at 7 days of age, became increasingly intense through 27 days of age, then decreased in abundance at 3 and 5 months of age consistent with neurofascin 155H in Fig. 4A. (C) At 7 days of age, when neurofascin 155L’ was not present, the pan-neurofascin antibody FIGQY labelled neurofascin 186 at the node of Ranvier and neurofascin 155H at the paranode in the rat spinal cord. Voltage gated sodium channels are indicated in dark blue (a and e), Caspr is indicated in red (b and f), and FIGQY labelling is shown in green (c and g). The panels on the left (a–d) show a fully formed node:paranode region and in the merged image (d) light blue represents voltage gated sodium channels and neurofascin 186 at the node and the flame colour represents Caspr and neurofascin 155H at the paranode. The panels on the right (e–h) show an immature node:paranode region with a pair of voltage gated sodium channel/ neurofascin 186-labelled clusters flanking an unlabelled region that is presumed to develop into the node. Notice that the flame colour is observed in the merged image (h) indicating that Caspr and neurofascin 155H are present in the developing paranode. d = days;

m = months; P = PNGase F treated; ERK2 = extracellular signal-regulated kinase 2; Nfasc = neurofascin; Nav = voltage gated sodium. Scale bar = 5 mm. 396 | Brain 2010: 133; 389–405 A. D. Pomicter et al.

An increasingly dark shadow-like band, perhaps representing mul- used to identify the nodal and paranodal domains, respectively, in tiple glycosylation states, was present immediately below neuro- established [Fig. 4C, (a–d)] and developing [Fig. 4C, (e–h)] fascin 155H beginning at 14 days of age. By 3 months of age the node:paranode regions. The unlabelled region between the shadow-like band was no longer visible but a distinct neurofascin labelled clusters in Fig. 4C (e–h) is consistent with the developing 155L band was observed and remained prominent until at least node:paranode region described by Shrager and colleagues 5 months of age. (Dugandzija-Novakovic et al., 1995). In both series of images, the FIGQY pan-neurofascin antibody strongly labels the nodal Neurofascin 155H and neurofascin 155L and paranodal regions. Paranodal reactivity against the FIGQY pan-neurofascin antibody at an age when neurofascin 155L is contain N-linked carbohydrates not detected by western blot analysis indicates that neurofascin Previous work (Volkmer et al., 1992; Davis et al., 1993; Maier 155H is the myelin paranodal protein previously referred to as et al., 2005) has shown that neurofascin proteins are glycosylated neurofascin 155 (Tait et al., 2000; Poliak and Peles, 2003). and analysis with the NetNGlyc 1.0 Server indicated that neuro- fascin 155 contains multiple (predicted) sites for N-linked glycosy- Neurofascin 155H and neurofascin lation. To test the possibility that neurofascin 155H and neurofascin 155L are the same polypeptide that differ according 155L are products of the Nfasc gene to the extent of N-linked glycosylation, samples were treated with Bhat and colleagues recently reported the generation of a mouse PNGase F, which specifically removes N-linked carbohydrates line with a floxed Nfasc transgene that is removed via CNP without compromising peptide bond integrity (Maley et al., promoter-regulated Cre expression (CNP-Cre;Nfascflox) (Pillai 1989). If the molecular mass difference between neurofascin et al., 2009). Since CNP promoter activity in the CNS has been 155H and neurofascin 155L resulted exclusively from N-linked shown to be exclusive to oligodendrocytes (Trapp et al., 1988), glycosylation, then a single, faster migrating band would be these mice allow for specific deletion of the Nfasc gene from detected following PNGase F treatment. Interestingly, following oligodendrocytes. CNP-Cre;Nfascflox/flox mice die by 17 days of N-linked deglycosylation, neurofascin 155H and neurofascin age; accordingly, spinal cords were harvested from 16-day-old 155L remained as distinct bands (Fig. 4B; see Fig. 6A for wild-type (+/+), CNP-Cre/Nfasc+/flox (heterozygote) and side-by-side untreated and treated lanes). The bands resulting CNP-Cre;Nfascflox/flox (homozygote) mice and are shown in from PNGase F treatment are referred to as neurofascin 155H’ Fig. 5A and B. Both wild-type and CNP-Cre;Nfasc+/flox samples and neurofascin 155L’ to acknowledge that neurofascin 155H exhibited the normal pattern of neurofascin 155H plus the and neurofascin 155L exist in vivo while neurofascin 155H’ and shadow-like band, together termed neurofascin 155H&L, prior to neurofascin 155L’ are generated by in vitro deglycosylation. PNGase F treatment (Fig. 5A) and both neurofascin 155H’ and Unexpectedly, PNGase F treatment revealed neurofascin 155L’ from samples of 14, 21 and 27-day-old animals, as well as at the expected ages of 3 and 5 months (Fig. 4B). This transition from shadow-like band to distinct band, possibly due to variable, age-related states of glycosylation (see ‘Discussion’ section), sug- gested that neurofascin 155L was the source of the increasingly dark shadow-like band observed in Fig. 4A. Fig. 4B shows that neurofascin 155H’ increases steadily between 7 and 27 days of age before sharply decreasing between 27 days and 3 months of age. This temporal pattern of expression of neurofascin 155H’ is consistent with neurofascin 155H playing a role in myelination, as has been previously shown for neurofascin 155 (Collinson et al., 1998). In contrast, neurofascin 155L’, which was not detected at 7 days of age, first appeared in the rat spinal cord between 7 and 14 days of age and steadily increased in abundance through 5 months of age. The slowly increasing abundance of neurofascin 155L’ suggested that, unlike neurofascin 155H, neurofascin 155L does not play a role in the initial stages of myelination.

Neurofascin 155H is a component of the myelin paranode Figure 5 Neurofascin 155H and neurofascin 155L are products of the Nfasc gene. Spinal cords from 16-day-old Since neurofascin 155L was not detected in 7-day-old rat spinal CNP-Cre;Nfascflox/flox mice contained no ‘neurofascin 155H&L’ cord (Fig. 4A and B), this age provided the opportunity to deter- and no neurofascin 155H’ or neurofascin 155L’ (arrows) when mine the localization of neurofascin 155H. As shown in Fig. 4C, observed without (A) or with (B) PNGase F treatment. the localizations of voltage gated sodium channels and Caspr are ERK2 = extracellular signal-regulated kinase 2. Discovery of Nfasc155H and Nfasc155L Brain 2010: 133; 389–405 | 397

Figure 6 Neurofascin 155H’ is reduced in CST knockout spinal cords. (A) Following PNGase F treatment, spinal cord homogenates showed an 50% decrease in neurofascin 155H’ at 15 and 30 days of age. Surprisingly, at 4 months of age no definitive neurofascin 155H’ (or neurofascin 155H) band was observed in the CST knockout samples (arrow). Quantitative analyses revealed that neurofascin 155H’ (B) was significantly reduced in the CST knockout mice at all ages examined while the levels of neurofascin 155L’ (C) were not altered; see Table 3. = untreated; P = PNGase F treated; KO = knockout; WT = wild-type; Nfasc = neurofascin; ERK2 = extracellular signal-regulated kinase 2. P50.05. neurofascin 155L’ after PNGase F treatment (Fig. 5B). In contrast, altered in the CST knockout mice. Accordingly, PNGase F was neither neurofascin 155H/neurofascin 155H’ nor neurofascin 155L/ employed to quantitatively compare levels of neurofascin 155H’ neurofascin 155L’ were observed in the CNP-Cre;Nfascflox/flox sample and neurofascin 155L’ in the wild-type and CST knockout mice. (arrows). These results demonstrate that both neurofascin 155H and As shown in Fig. 6A, PNGase F treated spinal cord samples from neurofascin 155L are products of the Nfasc gene and suggest that CST knockout mice contain 50% less neurofascin 155H’ both neurofascin 155H and neurofascin 155L are exclusively than wild-type littermates at 15 and 30 days of age (Fig. 6B; expressed by oligodendrocytes since expression is eliminated by Table 3). In contrast, neurofascin 155L’ was not reduced at CNP promoter-regulated Cre expression. either 15 or 30 days of age (Fig. 6C; Table 3). At 4 months of age, neurofascin 155H’ was also significantly reduced and no Neurofascin 155H, but not neurofascin definitive band was observed for neurofascin 155H’ (or neurofas- 155L, is reduced in the absence of cin 155H) in the CST knockout spinal cords, while the amount of neurofascin 155L’ remained unchanged (Fig. 6C; Table 3). sulphatide Running the untreated and PNGase F treated samples side-by-side The combined results of Figs 3 and 4 suggest that the levels of facilitates the comparison of the shifts in migration induced by neurofascin 155H’ and neurofascin 155L’ could be differentially deglycosylation. Note that neurofascin 155L was not observed in 398 | Brain 2010: 133; 389–405 A. D. Pomicter et al.

Table 3 Neurofascin 155H’ and neurofascin 155L’ in the CST knockout spinal cord

Age Wild-type (n) SD (%) Knockout (n)SD(%)P Reduction (%) Neurofascin 155H’, % of wild-type 15 days 100 (3) 14.40 50.1 (4) 32.30 0.009 49.9 30 days 100 (3) 11.30 55.0 (4) 23.50 0.005 45 4 months 100 (2) 8.40 34.4 (2) 28.10 0.019 65.6 Neurofascin 155L’, % of wild-type 15 days 100 (3) 22.90 77.5 (4) 26.90 0.249 0 30 days 100 (3) 14.40 91.2 (4) 20.80 0.515 0 4 months 100 (2) 11.60 87.1 (2) 9.70 0.342 0

the untreated lanes at 15 days of age from either genotype or in the 30 day wild-type sample; however, following PNGase F treat- ment, neurofascin 155L’ was clearly distinguished in these lanes, consistent with the idea that the band labelled ‘neurofascin 155H&L’ contains both proteins, and that neurofascin 155L is variably glycosylated with age. For this reason, only neurofascin 155H’ and neurofascin 155L’ were quantified. Age-dependent extinction of the neurofascin 155H band is not a normal event, as neurofascin 155H (and neurofascin 155L) have been observed in spinal cord homogenates from wild-type mice as old as 22 months of age (data not shown).

Reduced levels of neurofascin 155H Figure 7 Spinal cords from Caspr knockout mice, similar to CST knockout animals, contain reduced amounts of neurofascin are not limited to the CST 155H and neurofascin 155H’. Comparable to 15-day-old CST knockout mice knockout mice, spinal cords samples from 19-day-old Caspr knockout mice treated with PNGase F showed dramatically To determine whether there is a relationship between reductions reduced neurofascin 155H and no band representing neurofas- in neurofascin 155H/neurofascin 155H’ and paranodal stability, cin 155H’ was observed following PNGase F treatment (arrow). Caspr knockout mice (Bhat et al., 2001) were analysed via Neurofascin 155L’ appears modestly reduced. The Magic Mark PNGase F treatment and the extended electrophoresis protocol. Western Protein standard (M) revealed the shift in molecular These mice, reminiscent of the CST knockout mice, exhibit abnor- mass following treatment with PNGase F. CP = Caspr; mal paranodes and lack septate-like paranodal junctions. The life = untreated; P = PNGase F treated; KO = knockout; expectancy of these mice is 421 days (Bhat et al., 2001), so WT = wild-type; Nfasc = neurofascin; ERK2 = extracellular signal-regulated kinase 2. analysis was restricted to younger ages. Similar to 15-day-old CST knockout mice, 19-day-old Caspr knockout mice revealed noticeably reduced levels of neurofascin 155H and neurofascin banding pattern was observed in brain samples taken from regions 155H’, while levels of neurofascin 155L and neurofascin 155L’ of white matter that appear normal in multiple sclerosis brains were only modestly altered (Fig. 7). This reduction in neurofascin (normal-appearing white matter). Normal-appearing is defined as 155H and neurofascin 155H’, taken together with the findings revealing no visible signs of myelin deterioration as determined by from the CST knockout mice, suggests that the loss of neurofascin both macroscopic and light microscopic examination. All multiple 155H may be a general defect associated with paranodal sclerosis plaque samples also revealed neurofascin 155H. However, instability. samples from two (#164 and 224) of the eight donors analysed revealed a band with an apparent molecular mass consistent with The pattern of neurofascin 155H and neurofascin 155L as identified in rodent. To assess this possible neurofascin 155L is altered in alteration better, multiple sclerosis plaque samples were run multiple sclerosis plaques side-by-side with either non-multiple sclerosis or normal-appearing white matter-multiple sclerosis samples (representative examples Analysis of human brain samples by the extended electrophoresis are presented in Figs 8A2 and Figure 8A3, respectively). This addi- protocol (Figure 8A1) indicated that neurofascin 155H is abundant tional analysis further indicated that the banding pattern for neu- in non-multiple sclerosis brain samples while no clearly delineated rofascin 155H and neurofascin 155L was altered in certain neurofascin 155L was observed. (‘Clearly delineated’ refers to a samples, with clearly delineated neurofascin 155L being observed band that is distinct and separate from adjacent bands.) A similar only in multiple sclerosis plaque samples. Discovery of Nfasc155H and Nfasc155L Brain 2010: 133; 389–405 | 399

Figure 8 The pattern of neurofascin 155H’ and neurofascin 155L’ is altered in multiple sclerosis plaques. (A) Following the extended electrophoresis protocol, clearly delineated neurofascin 155L was not observed in non-multiple sclerosis (A1 and A2) or normal-appearing white matter-multiple sclerosis (A1 and A3) samples but was observed in multiple sclerosis plaque samples (black stars). (B) Following PNGase F treatment, no clearly delineated neurofascin 155L’ band was observed in either the non-multiple sclerosis or normal-appearing white matter-multiple sclerosis samples. In contrast, neurofascin 155L’ was clearly observed in the majority of multiple sclerosis brains analysed (white stars). (C) The ratio of neurofascin 155H’/neurofascin 155L’ was significantly reduced in multiple sclerosis plaque samples (n =8; P50.006) compared to non-multiple sclerosis and normal-appearing white matter-multiple sclerosis samples (n = 7). (D) Side-by-side analysis of PNGase F treated samples of normal-appearing white matter-multiple sclerosis and multiple sclerosis plaque samples from the same donors shows neurofascin 155L’ only in the plaque samples. It is noteworthy that the pattern of neurofascin 155H’ and neurofascin 155L’ is similar between inactive (#243) and active (#137) plaque samples. Additionally, neurofascin 186’ was frequently not observed in multiple sclerosis plaque samples. The commonly used loading control proteins b-actin, glyceraldehyde 3-phosphate dehydrogenase and extracellular signal-regulated kinase 2 are shown and demonstrate a lack of autolysis within the samples. = untreated; P = PNGase F treated; NM or NAWM = normal appearing white matter from multiple sclerosis donor; M = Magic Mark XP Protein Standard; MS = multiple sclerosis; Nfasc = neurofascin; ERK2 = extracellular signal-regulated kinase 2; GAPDH = glyceraldehyde 3-phosphate dehydrogenase. P50.05. 400 | Brain 2010: 133; 389–405 A. D. Pomicter et al.

Interestingly, samples from donors #164 and #224 revealed correlation between the absence of neurofascin 186 and the an apparent reduction of neurofascin 155H, which may have length of time between death and tissue collection (Table 1), in increased the visibility of neurofascin 155L. Thus it was postulated agreement with a previous study demonstrating preservation of that the apparent absence of neurofascin 155L in some multiple nodal and paranodal proteins up to 24 hours post-mortem sclerosis plaque (donors #168 and 73), normal-appearing white (Howell et al., 2006). matter-multiple sclerosis (donor #235) and non-multiple sclerosis (donor #5) samples resulted from an intense, bulky neurofascin 155H band which was obscuring a less intense neurofascin 155L band, as is the case with rodent (compare 30-day-old wild-type to Discussion 30-day-old knockout in Figs 3A, B). To test this hypothesis, non-multiple sclerosis, normal-appearing white matter-multiple In the absence of sulphatide, paranodes appear structurally normal sclerosis and multiple sclerosis plaque samples were treated with during development but deteriorate with age (Marcus et al., PNGase F. Non-multiple sclerosis and normal-appearing white 2006). In the current study, quantitative analysis revealed that matter-multiple sclerosis samples (donors #5 and 235; Fig. 8B) neurofascin 155 accumulated in the paranode in the absence of showed a prominent neurofascin 155H’ band with a faint sulphatide; however, the number of anti-neurofascin-positive shadow-like neurofascin 155L’ band that was continuous with paired clusters in the spinal cord of the CST knockout mice was the neurofascin 155H’ band. Additionally, the remaining reduced by 50 and 90% at 15 and 30 days of age, respectively. non-multiple sclerosis (donor #2) and normal-appearing white Western blot analyses indicated that total levels of neurofascin matter-multiple sclerosis (donor #237) samples showed a similar 155 in the CST knockout mice are reduced 40% at both 15 pattern. All multiple sclerosis plaque samples treated with PNGase and 30 days of age. Therefore, the significant reduction of neu- F yielded a neurofascin 155H’ band while five of the eight samples rofascin 155 clustering observed at 15 days of age may be a (Fig. 8B, donors #137, 216, 224; see #243 in Fig. 8D, and #164 consequence of reduced neurofascin 155 expression; however, not shown) revealed a clearly delineated neurofascin 155L’ band. decreased protein levels alone do not provide adequate explana- To compare quantitative changes in the neurofascin 155H’ and tion for the continued reduction of paired paranodal clusters of neurofascin 155L’ pattern, a densitometric ratio of neurofascin neurofascin 155 by 30 days of age. These differences suggest a 155H’ to neurofascin 155L’ was determined for each sample. relationship between sulphatide and maintenance and localization This ratio was utilized instead of direct comparisons to account of neurofascin 155 at the paranode. In contrast, sulphatide is not for the variability that is inherent in human tissue sampling. As required for initial paranode formation (Marcus et al., 2006). The shown in Fig. 8C, the ratio of neurofascin 155H’ to neurofascin mechanistic differences between paranode development and 155L’ was significantly reduced in the multiple sclerosis plaques maintenance are yet to be determined. 5 (n = 8; 4.2 3.5, P 0.006) compared to the non-multiple sclerosis Extended electrophoresis revealed the presence of a neurofascin and normal-appearing white matter-multiple sclerosis samples 155 doublet. These bands have been termed neurofascin 155H (n = 7; 14.2 6.6). and neurofascin 155L to reflect their relative molecular masses. In addition to the decrease in the neurofascin 155H’/neurofas- The 40% reduction in neurofascin 155 at both 15 and 30 cin 155L’ ratio in multiple sclerosis samples, alterations in the days of age can be solely attributed to the reduction of neurofas- levels of neurofascin 186 were also observed. Although neurofas- cin 155H, as levels of neurofascin 155L remain unchanged in the cin 186 was always detected in the non-multiple sclerosis samples, CST knockout mice. The observation of neurofascin 155H at the one (donor #216) of five normal-appearing white matter-multiple paranode along with the specific reduction of neurofascin 155H in sclerosis samples failed to reveal neurofascin 186 while three mice that exhibit paranodal instability is consistent with neurofas- (donors # 137, 216, and 224) of eight multiple sclerosis samples cin 155H, but not neurofascin 155L, being a paranodal protein yielded no neurofascin 186 (Figs 8A1,8A2 and 8A3). Importantly, required for preservation of paranodal integrity. Although the multiple sclerosis donor #164 presented both neurofascin 155L loss of neurofascin 155H correlates with paranodal instability in and neurofascin 186 (Fig. 8A1 and 8A3), demonstrating that the the sulphatide and Caspr null mice, the possibility persists that presence of neurofascin 155L is not dependent on the absence of the loss of neurofascin 155H is the result of paranodal breakdown, neurofascin 186; and suggesting that neurofascin 155L is a not a rather than its cause. breakdown product of neurofascin 186. Also note that neurofas- Together, these findings demonstrate that sulphatide is essential cin 186 is present in the normal-appearing white matter-multiple for paranodal integrity due to its role in maintaining established sclerosis samples from donors #243 and #137 (Fig. 8D) but is absent in the plaques samples from these same donors, suggest- paranodal protein domains. Ishibashi et al. (2002) reported a loss ing that the absence of neurofascin 186 is not the result of of both nodal sodium and juxtaparanodal potassium channel clus- post-mortem autolysis and is a multiple sclerosis plaque-specific ters in the CST knockout mice; however, this loss was not signif- alteration. Further supporting the absence of significant icant until 6 weeks of age. The reduction of neurofascin 155 post-mortem autolysis, no changes in the levels of b-actin, extra- paranodal clusters is significant as early as 15 days of age, while cellular signal-regulated kinase 2 or glyceraldehyde 3-phosphate the paranodal structure is normal (Marcus et al., 2006), suggesting dehydrogenase were observed between normal-appearing white that the reduction and loss of neurofascin 155 clusters from matter-multiple sclerosis and multiple sclerosis plaque samples the myelin paranode precedes, and may facilitate, neuronal from the same donors (Fig. 8D). Additionally, there was no deterioration. Discovery of Nfasc155H and Nfasc155L Brain 2010: 133; 389–405 | 401

What is the difference between This apparent emergence suggests that neurofascin 155L is a deriv- ative of neurofascin 155H; however, treatment with PNGase F neurofascin 155H and indicates that even at 14 days of age, when the two bands appar- neurofascin 155L? ently migrate together, neurofascin 155H’ and neurofascin 155L’ represent distinct protein populations. Figure 4B shows, particu- Possible sources for neurofascin 155H and neurofascin 155L larly following deglycosylation, that levels of neurofascin 155H’ include products of different genes, splice variants of the same decrease between 27 days and 3 months of age with a corre- gene, or the same amino acid sequence with differential post sponding increase in neurofascin 155L, suggesting that neurofas- translational modifications. Utilizing the CNP-Cre;Nfascflox mice cin 155L could be a post-translational product of neurofascin (Pillai et al., 2009) this study demonstrated that neurofascin 155H. However, in opposition to neurofascin 155L being derived 155H and neurofascin 155L are products of the same gene (the from neurofascin 155H, normal levels of neurofascin 155L’ are Nfasc gene); however, it remains to be determined whether neu- maintained in the absence of neurofascin 155H’ in the aged CST rofascin 155H and neurofascin 155L differ according to amino acid knockout mice, suggesting that levels of neurofascin 155L’ are not sequence and/or post-translational modifications. dependent on the presence of neurofascin 155H’. Alternatively, The production of multiple splice variants is consistent with the neurofascin 155L could be highly stable, such that the loss of current understanding of the Nfasc gene as it produces at least neurofascin 155H would have minimal effect on levels of neuro- three proteins (Davis et al., 1996; Basak et al., 2007) and as many fascin 155L. This supposition is not supported by the recent report as 50 transcripts (Hassel et al., 1997), lending validity to the idea that disruption of the Nfasc gene in the oligodendrocytes of that neurofascin 155H and neurofascin 155L are distinct isoforms 33-day-old mice results in the gradual loss of neurofascin 155 of the Nfasc gene. However, it is also plausible that neurofascin and complete loss within 60 days (Pillai et al., 2009). These find- 155H and neurofascin 155L consist of identical amino acid ings demonstrate that the stability of all oligodendrocyte-derived sequences and that their molecular mass difference results from Nfasc gene products is less than two months. Longer stability varying post-translational modifications. Maier et al. (2005) would be required for the maintenance of neurofascin 155L in reported the observation of a neurofascin 155 doublet and specu- the absence of neurofascin 155H, as is observed in the aged lated that variable glycosylation may account for this observation. CST knockout mice, since neurofascin 155H’ is lost as early as Here, it has been shown that N-linked deglycosylation of neuro- 3 months of age and neurofascin 155L is maintained through fascin 155H and neurofascin 155L does not result in the formation 15 months of age (unpublished observation). The possibility of a single anti-neurofascin band. Thus, the difference between cannot be excluded, however, that in the aged CST knockout neurofascin 155H and neurofascin 155L is not simply differential mice neurofascin 155H is rapidly converted into neurofascin N-linked glycosylation. Analyses with the AlphaEase software indi- 155L, precluding detection of neurofascin 155H or neurofascin cated that removal of N-linked glycans with PNGase F resulted in 155H’ by western blot analysis. shifts of 12 and 17 kDa, respectively, for neurofascin 155H to Hassel and colleagues (1997) identified a fifth fibronectin III-like neurofascin 155H’ and neurofascin 155L to neurofascin 155L’ in domain in the Nfasc gene of chicken and rat after the original animals 3 months of age and older. This difference suggests that report that only four fibronectin III-like domains were present neurofascin 155L either maintains N-linked glycans at more sites (Volkmer et al., 1992). This domain was also observed in cDNA than neurofascin 155H or that the carbohydrates on neurofascin clones and, unlike the other fibronectin III-like domains, is 155L are more extensive than those attached to neurofascin expressed as either a half (one exon) or a whole (two exons) 155H. domain. The whole fifth fibronectin III-like domain is expressed Other modifications that have been reported for neurofascin with the mucin-like domain, which is specific to neurofascin 186. 186, the neuronal isoform of the neurofascin gene, include phos- It is unclear whether or not this fifth fibronectin III-like domain is phorylation (Garver et al., 1997; Tuvia et al., 1997), O-linked expressed in known neurofascin proteins (Tait et al. 2000; Koticha glycosylation (Volkmer et al., 1992) and palmitoylation (Ren and et al., 2005). The two halves of the fifth fibronectin III-like domain Bennett, 1998). While Volkmer et al. (1992) have shown that are 5 kDa each and analysis with the NetNGlyc 1.0 Server indi- O-linked glycosylation does not occur in neurofascin 155, it cates two potential sites of N-linked glycosylation for the first half remains to be determined whether neurofascin 155H or neurofas- and one potential site for the second half. It is interesting to con- cin 155L are palmitoylated or phosphorylated. Following N-linked sider that either half of this domain may account for the observed deglycosylation, neither neurofascin 155H nor neurofascin 155L, differences in size between neurofascin 155H and neurofascin at any age analysed, migrated at 132 kDa, which is the predicted 155L. mass of neurofascin 155 based on amino acid sequence, suggest- ing that differences other than the published amino acid sequence for neurofascin 155 and N-linked glycosylation contribute to the Are neurofascin 155H and neurofascin molecular mass of neurofascin 155H and neurofascin 155L. 155L functionally distinct? However, it cannot be ruled out that the deglycosylated proteins may migrate anomalously in the gel, thus accounting for migration The functional significance of differential glycosylation of neuro- above 132 kDa. fascin 155H and neurofascin 155L is unknown. Since glycosylation In Fig. 4A, neurofascin 155L appears to emerge from the is a prime regulator of protein functions such as protease activity shadow-like band below neurofascin 155H with increasing age. (Van den Steen et al., 2001), intracellular trafficking (Yan et al., 402 | Brain 2010: 133; 389–405 A. D. Pomicter et al.

2008), protein binding (Milev et al., 1995; Zhou et al., 2008) and 155H’ shown in Fig. 4A has been labelled ‘neurofascin 155H&L’ stability (Buck et al., 2007; Goa and Nehta, 2007; Li et al., 2007), in later figures to denote the supposition that neurofascin 155H’ neurofascin 155H and neurofascin 155L may have distinct func- and neurofascin 155L’ are derived from this band plus the tions related to their different states of glycosylation, including shadow-like band (Figs 5A, 5B, 6A and 7). Presumably, neurofas- potential interactions with contactin and/or Caspr; however, this cin 155L migrates with/immediately below neurofascin 155H in remains to be determined. younger animals (compare Fig. 4A and B; 14, 21 and 27 days The only recognized function of neurofascin 155 is the binding of age, as well as the first four pairs of lanes in Fig. 6A). At all of neuronal contactin at the paranode (Tait et al., 2000; Gollan ages, the shift from neurofascin 155H to neurofascin 155H’ is et al., 2003; Bonnon et al., 2007). At an age when neurofascin 12 kDa while the shift from the shadow-like band to distinct 155H’, but not neurofascin 155L’, is detected (7 days of age; Fig. neurofascin 155L’ is 22 kDa in younger animals and 17 kDa 4B), anti-neurofascin reactive paranodes are prevalent (Fig. 4C) for neurofascin 155L in older animals, suggesting that neurofascin indicating that neurofascin 155H is a myelin paranodal protein 155L is variably glycosylated with age. Specifically, neurofascin expressed early in the process of myelination. In contrast, neuro- 155L may become less glycosylated with age, accounting for the fascin 155L’ is not detected until the end of the second week of gradual, age-related transition from the shadow-like band below life and continues to increase through at least 5 months of age, neurofascin 155H in younger animals to the emergence of the suggesting that neurofascin 155L is not involved in the onset of distinct neurofascin 155L band observed in older animals myelination. It cannot be concluded, however, that neurofascin (Fig. 4A). The same conclusion is suggested by Fig. 6A, where 155H is the form that binds contactin or that neurofascin 155L neurofascin 155L is not observed in the 15 or 30-day-old is not a paranodal protein involved in mediating myelin-axon inter- wild-type untreated lanes but neurofascin 155L’ is observed fol- actions. Interestingly, Maier et al. (2005), who detected an lowing PNGase F treatment. Age-related changes in glycosylation anti-neurofascin doublet in the vicinity of 155 kDa, reported that is documented in adhesion proteins of immune cells (Daniels et al., the lower band of the doublet is not detected in pure cultures of 2002), epithelial cells (Wang et al., 2008) and nervous system oligodendrocytes, suggesting that oligodendrocyte-axon interac- cells (Edelman, 1985). Based on these observations, it is reason- tion is a prerequisite for expression of the lower component of able to propose that neurofascin 155H’ is deglycosylated neuro- the doublet, which is likely to be neurofascin 155L. fascin 155H, that neurofascin 155L’ is deglycosylated neurofascin Neurofascin 155H and neurofascin 155H’ are virtually absent in 155L, and that results regarding neurofascin 155H’ and neurofas- 4-month-old CST knockout mice indicating that sulphatide is cin 155L’ can provide information regarding neurofascin 155H and required for long term maintenance of this protein. One possible neurofascin 155L. explanation for the loss of neurofascin 155H lies with the concept of membrane rafts (Harder and Simons, 1997), which are small, dynamic membrane domains enriched in cholesterol and sphingo- Levels of both oligodendrocyte and lipids (reviewed by Dupree and Pomicter, 2009). Sulphatide, a sphingolipid, is enriched in myelin fractions containing putative neuronal neurofascin are altered raft proteins (Arvanitis et al., 2005) such as fyn (Kra¨ mer et al., in multiple sclerosis 1999), flotillin-1 (Salzer and Prohaska, 2001) and caveolin In multiple sclerosis plaque samples, the ratio of neurofascin (Sargiacomo et al., 1993); and neurofascin 155 has been impli- 155H’/neurofascin 155L’ is significantly reduced compared to cated as a raft protein (Maier, 2007; Schafer et al., 2004). non-multiple sclerosis and normal-appearing white matter-multiple Therefore, the absence of sulphatide may result in abnormal raft sclerosis. This finding is intriguing in light of the significant reduc- formation facilitating neurofascin 155H instability at the myelin tion observed for neurofascin 155H’ in the CST knockout model in paranode. This is consistent with recent evidence implicating a the presence of unaltered levels of neurofascin 155L’. In both role for membrane rafts in regulating protease activity multiple sclerosis plaques and in CST knockout mice, the pattern (Abad-Rodriguez et al., 2004; Bae et al., 2008). It is interesting of neurofascin 155H’ and neurofascin 155L’ is disrupted, concur- that the Caspr knockout tissue reveals a decrease in neurofascin rent with paranodal deterioration that precedes axonal pathology. 155H and neurofascin 155H’. Since receptor–ligand binding, such It remains to be determined, however, in both human and mouse as occurs in the paranode between paranodal neurofascin and models, whether disruption of neurofascin 155H and/or neurofas- neuronal contactin:Caspr (Tait et al., 2000; Gollan et al., 2003; cin 155L causes or results from paranodal deterioration. Bonnon et al., 2007), may induce raft formation (Subczynski and In this study, the multiple sclerosis samples were categorized Kusumi, 2003), a disruption of the myelin-axon junction may also either as active or inactive based on the level of cellularity result in neurofascin 155H instability as a consequence of impaired within the harvested plaque. Neurofascin 155L’ was observed in raft structure in the Caspr knockout mice. one of three active (donor #137) and four of five inactive (donors #224, 164, 216, 243) plaques suggesting that the pathological Neurofascin 155L is variably event that results in a reduced ratio of neurofascin 155H’ to neu- rofascin 155L’ is initiated during active myelin deterioration. The glycosylated with age presence of clearly delineated neurofascin 155L’ in inactive pla- An increasingly intense, age-dependent shadow-like band ques is also consistent with the idea that neurofascin 155L’ is a attached to and just beneath the band labelled ‘neurofascin product of oligodendrocytes (although axons cannot be ruled out Discovery of Nfasc155H and Nfasc155L Brain 2010: 133; 389–405 | 403 as a source), as few other cell types are present in inactive plaques Neurobiology Microscopy Facility, supported, in part, with (Lassmann et al., 1998; Mycko et al., 2003). Neurofascin 155L’ as funding from National Institutes of Health-National Institute a product of oligodendrocytes is consistent with the rodent find- of Neurological Disease and Stroke Centre core grant ings since mice that exhibit a disruption in the neurofascin gene in (5P30NS047463). oligodendrocytes are incapable of synthesizing neurofascin 155L. It is intriguing that the neuronal form of neurofascin, neurofas- cin 186, is less frequently observed in multiple sclerosis plaque Supplementary material samples compared to non-multiple sclerosis and normal-appearing white matter-multiple sclerosis samples. The reduced ability to Supplementary material is available at Brain online. detect neurofascin 186 in multiple sclerosis is particularly interest- ing in light of the findings of Mathey et al. (2007) who reported that neurofascin antibodies, derived from serum of multiple scle- References rosis patients, selectively target the node of Ranvier, the specific Abad-Rodriguez J, Ledesma MD, Craessaerts K, Perga S, Medina M, location of neurofascin 186 (Davis et al., 1996; Poliak and Peles, Delacourte A, et al. Neuronal membrane cholesterol loss enhances 2003), resulting in the recruitment of complement, axonal degen- amyloid peptide generation. J Cell Biol 2004; 167: 953–60. eration, and exacerbated pathology in a mouse model. Therefore, Arvanitis DN, Min W, Gong Y, Heng YM, Boggs JM. Two types of the inability to detect neurofascin 186 in some normal-appearing detergent-insoluble, glycosphingolipid/cholesterol-rich membrane white matter-multiple sclerosis and multiple sclerosis plaque sam- domains from isolated myelin. J Neurochem 2005; 94: 1696–710. Bae JS, Yang L, Rezaie AR. Lipid raft localization regulates the cleavage ples is consistent with the early axonal degeneration that occurs in specificity of protease activated receptor 1 in endothelial cells. J multiple sclerosis. Thromb Haemost 2008; 6: 954–61. In conclusion, the present study demonstrates an essential role Basak S, Raju K, Babiarz J, Kane-Goldsmith N, Koticha D, Grumet M. for sulphatide in maintaining neurofascin 155 at the myelin para- Differential expression and functions of neuronal and glial neurofascin node and introduces the existence of two forms of neurofascin isoforms and splice variants during PNS development. Dev Biol 2007; 311: 408–22. 155—neurofascin 155H and neurofascin 155L—that exhibit differ- Bhat MA, Rios JC, Lu Y, Garcia-Fresco GP, Ching W, St Martin M, et al. ential developmental expression and variable stability in the Axon-glia interactions and the domain organization of myelinated absence of sulphatide. Moreover, neurofascin 155H is concen- axons requires neurexin IV/Caspr/Paranodin. Neuron 2001; 30: trated at the myelin paranode and is progressively lost from 369–83. both the paranode and the CNS concomitantly with structural Bonnon C, Bel C, Goutebroze L, Maigret B, Girault JA, Faivre-Sarrailh C. PGY repeats and N-glycans govern the trafficking of paranodin and its deterioration of the paranode. Similarly, the loss of paranodal selective association with contactin and neurofascin-155. Mol Biol Cell integrity observed in multiple sclerosis (Suzuki et al., 1969; 2007; 18: 229–41. Howell et al., 2006) coincides with a significant alteration in the Boyle ME, Berglund EO, Murai KK, Weber L, Peles E, Ranscht B. pattern of neurofascin 155H’ and neurofascin 155L’. Determining Contactin orchestrates assembly of the septate-like junctions at the the consequence of this altered pattern of expression in multiple paranode in myelinated peripheral nerve. Neuron 2001; 30: 385–97. Buck TM, Eledge J, Skach WR. Evidence for stabilization of aquaporin-2 sclerosis plaques will be an important step in understanding the folding mutants by N-linked glycosylation in endoplasmic reticulum. pathological mechanisms that underlie CNS degeneration. Am J Physiol Cell Physiol 2004; 287: C1292–9. Charles P, Tait S, Faivre-Sarrailh C, Barbin G, Gunn-Moore F, Denisenko- Nehrbass N, et al. Neurofascin is a glial receptor for the paranodin/ Caspr-contactin axonal complex at the axoglial junction. Curr Biol Acknowledgements 2002; 12: 217–20. Chiu SY. Asymmetry currents in the mammalian myelinated nerve. Human tissue samples were provided by the Rocky Mountain J Physiol 1980; 309: 499–519. Multiple Sclerosis Centre (Englewood, CO). The authors wish to Chiu SY, Schwarz W. Sodium and potassium currents in acutely demye- thank Haley J. Steinert, Multiple Sclerosis Research Coordinator linated internodes of rabbit sciatic nerves. J Physiol 1987; 391: 631–49. and Tissue Bank Research Associate at the Rocky Mountain Collinson JM, Marshall D, Gillespie CS, Brophy PJ. xTransient expression Multiple Sclerosis Centre for her assistance in acquiring donor of neurofascin by oligodendrocytes at the onset of myelinogenesis: information. Additionally, the authors thank Dr Nitya Ghatak, Implications for mechanisms of axon-glial interaction. Glia 1987; 3: Neuropathology Division Chair in the Department of Pathology 11–23. at Virginia Commonwealth University for his assistance in inter- Cook RD. Detection of influential observation in linear regression. Technometrics 1977; 19: 15–18. preting the pathology reports for each donor. Daniels MA, Hogquist KA, Jameson SC. Sweet ’n’ sour: The impact of differential glycosylation on T cell responses. Nat Immunol 2002; 3: 903–10. Funding Davis JQ, Lambert S, Bennett V. Molecular composition of the node of Ranvier: Identification of ankyrin-binding cell adhesion molecules This work was supported by grants from the National Multiple neurofascin (mucin+/third FNIII domain-) and NrCAM at nodal axon Sclerosis Society (PP 1440 to J.L.D.) and the A. D. Williams segments. J Cell Biol 1996; 135: 1355–67. Davis JQ, McLaughlin T, Bennett V. Ankyrin-binding proteins related to Foundation (to J.L.D.) and the National Institutes of Health nervous system cell adhesion molecules: candidates to provide trans- (NS066186 to J.L.D. and GM063074 to M.A.B.). All microscopy membrane and intercellular connections in adult brain. J Cell Biol 1993; was performed at the VCU Department of Anatomy and 121: 121–33. 404 | Brain 2010: 133; 389–405 A. D. Pomicter et al.

Dugandzija-Novakovic S, Koszowski AG, Levinson SR, Shrager P. Maier O, van der Heide T, van Dam AM, Baron W, de Vries H, Clustering of Na+ channels and node of Ranvier formation in remye- Hoekstra D. Alteration of the extracellular matrix interferes with raft linating axons. J Neurosci 1995; 15 (1 Pt 2): 492–503. association of neurofascin in oligodendrocytes. Potential significance Dupree JL, Pomicter AD. Myelin, DIGs, and membrane rafts in the cen- for multiple sclerosis? Mol Cell Neurosci 2005; 28: 390–401. tral nervous system. Prostaglandins Other Lipid Mediat [Epub ahead of Maley F, Trimble RB, Tarentino AL, Plummer TH Jr. Characterization of print, April 18, 2009; doi:10.1016/j.prostaglandins.2009.04.005]. glycoproteins and their associated oligosaccharides through the use of Dupree JL, Girault JA, Popko B. Axo-glial interactions regulate the loca- endoglycosidases. Anal Biochem 1989; 180: 195–204. lization of axonal paranodal proteins. J Cell Biol 1999; 147: 1145–52. Marbois BN, Faull KF, Fluharty AL, Raval-Fernandes S, Rome LH. Dupree JL, Coetzee T, Blight A, Suzuki K, Popko B. Myelin galactolipids Analysis of sulfatide from rat cerebellum and multiple sclerosis white are essential for proper node of Ranvier formation in the CNS. matter by negative ion electrospray mass spectrometry. Biochim J Neurosci 1998; 18: 1642–9. Biophys Acta 2000; 1484: 59–70. Edelman GM. Cell adhesion and the molecular processes of morphogen- Marcus J, Honigbaum S, Shroff S, Honke K, Rosenbluth J, Dupree JL. esis. Annu Rev Biochem 1985; 54: 135–69. Sulfatide is essential for the maintenance of CNS myelin and axon Gao Y, Mehta K. N-linked glycosylation of CD38 is required for its structure. Glia 2006; 53: 372–81. structure stabilization but not for membrane localization. Mol Cell Milev P, Meyer-Puttlitz B, Margolis RK, Margolis RU. Complex-type Biochem 2007; 295: 1–7. asparagine-linked oligosaccharides on phosphacan and protein-tyrosine Garver TD, Ren Q, Tuvia S, Bennett V. Tyrosine phosphorylation at a site phosphatase-zeta/beta mediate their binding to neural cell adhesion highly conserved in the L1 family of cell adhesion molecules abolishes molecules and tenascin. J Biol Chem 1995; 270: 24650–3. ankyrin binding and increases lateral mobility of neurofascin. J Cell Biol Mycko MP, Papoian R, Boschert U, Raine CS, Selmaj KW. cDNA micro- 1997; 137: 703–14. array analysis in multiple sclerosis lesions: detection of genes associated Gollan L, Salomon D, Salzer JL, Peles E. Caspr regulates the processing of with disease activity. Brain 2003; 126 (Pt 5): 1048–57. contactin and inhibits its binding to neurofascin. J Cell Biol 2003; 163: Ogawa Y, Schafer DP, Horresh I, Bar V, Hales K, Yang Y, et al. Spectrins 1213–8. and ankyrinB constitute a specialized paranodal cytoskeleton. J Harder T, Simons K. Caveolae, DIGs, and the dynamics of sphingolipid- Neurosci 2006; 26: 5230–9. cholesterol microdomains. Curr Opin Cell Biol 1997; 9: 534–42. Peters A, Palay S, Webster H, editors. The fine structure of the nervous Hassel B, Rathjen FG, Volkmer H. Organization of the neurofascin gene system. Philadelphia: W. B. Saunders Company; 1976. and analysis of developmentally regulated alternative splicing. J Biol Pillai AM, Thaxton C, Pribisko AL, Cheng JG, Dupree JL, Bhat MA. Chem 1997; 272: 28742–9. Spatiotemporal ablation of myelinating glia-specific neurofascin Honke K, Yamane M, Ishii A, Kobayashi T, Makita A. Purification and (Nfasc NF155) in mice reveals gradual loss of paranodal axoglial junc- characterization of 3’-phosphoadenosine-5’-phosphosulfate:GalCer tions and concomitant disorganization of axonal domains. J Neurosci sulfotransferase from human renal cancer cells. J Biochem 1996; Res 2009; 87: 1773–93. 119: 421–7. Poliak S, Peles E. The local differentiation of myelinated axons at nodes Honke K, Tsuda M, Hirahara Y, Ishii A, Makita A, Wada Y. Molecular of Ranvier. Nat Rev Neurosci 2003; 4: 968–80. cloning and expression of cDNA encoding human 3’-phosphoadeny- Rasband MN. It’s ‘‘juxta’’ potassium channel!. J Neurosci Res 2004; 76: lylsulfate: galactosyl -ceramide 3’-sulfotransferase. J Biol Chem 1997; 749–57. 272: 4864–8. Ren Q, Bennett V. Palmitoylation of neurofascin at a site in the mem- Honke K, Hirahara Y, Dupree J, Suzuki K, Popko B, Fukushima K, et al. brane-spanning domain highly conserved among the L1 family of cell Paranodal junction formation and spermatogenesis require sulfoglyco- adhesion molecules. J Neurochem 1998; 70: 1839–49. lipids. Proc Natl Acad Sci USA 2002; 99: 4227–32. Rios JC, Melendez-Vasquez CV, Einheber S, Lustig M, Grumet M, Howell OW, Palser A, Polito A, Melrose S, Zonta B, Scheiermann C, Hemperly J, et al. Contactin-associated protein (Caspr) and contactin et al. Disruption of neurofascin localization reveals early changes pre- form a complex that is targeted to the paranodal junctions during ceding demyelination and remyelination in multiple sclerosis. Brain myelination. J Neurosci 2000; 20: 8354–64. 2006; 129(Pt 12): 3173–85. Salzer U, Prohaska R. Stomatin, flotillin-1, and flotillin-2 are major Ishibashi T, Dupree JL, Ikenaka K, Hirahara Y, Honke K, Peles E, et al. A integral proteins of erythrocyte lipid rafts. Blood 2001; 97: 1141–3. myelin galactolipid, sulfatide, is essential for maintenance of ion chan- Sargiacomo M, Sudol M, Tang Z, Lisanti MP. Signal transducing nels on myelinated axon but not essential for initial cluster formation. molecules and glycosyl-phosphatidylinositol-linked proteins form a J Neurosci 2002; 22: 6507–14. caveolin-rich insoluble complex in MDCK cells. J Cell Biol 1993; 122: Jarjour AA, Bull SJ, Almasieh M, Rajasekharan S, Baker KA, Mui J, et al. 789–807. Maintenance of axo-oligodendroglial paranodal junctions requires DCC Schaeren-Wiemers N, Bonnet A, Erb M, Erne B, Bartsch U, Kern F, et al. and netrin-1. J Neurosci 2008; 28: 11003–14. The raft-associated protein MAL is required for maintenance of proper Koticha D, Babiarz J, Kane-Goldsmith N, Jacob J, Raju K, Grumet M. Cell axon-glia interactions in the central nervous system. J Cell Biol 2004; adhesion and neurite outgrowth are promoted by neurofascin NF155 166: 731–42. and inhibited by NF186. Mol Cell Neurosci 2005; 30: 137–48. Schafer DP, Bansal R, Hedstrom KL, Pfeiffer SE, Rasband MN. Does Kra¨ mer EM, Klein C, Koch T, Boytinck M, Trotter J. Compartmentation paranode formation and maintenance require partitioning of neurofas- of fyn kinase with glycosylphosphatidylinositol-anchored molecules in cin 155 into lipid rafts? J Neurosci 2004; 24: 3176–85. oligodendrocytes facilitates kinase activation during myelination. J Biol Sherman DL, Tait S, Melrose S, Johnson R, Zonta B, Court FA, et al. Chem 1999; 274: 29042–9. Neurofascins are required to establish axonal domains for saltatory Lassmann H, Raine CS, Antel J, Prineas JW. Immunopathology of multi- conduction. Neuron 2005; 48: 737–42. ple sclerosis: report on an international meeting held at the Institute of Shroff SM, Pomicter AD, Chow WN, Fox MA, Colello RJ, Henderson SC, Neurology of the University of Vienna. J Neuroimmunol 1998; 86: et al. Adult CST-null mice maintain an increased number of oligoden- 213–7. drocytes. J Neurosci Res 2009; 87: 3403–14. Li JG, Chen C, Liu-Chen LY. N-glycosylation of the human kappa opioid Subczynski WK, Kusumi A. Dynamics of raft molecules in the cell and receptor enhances its stability but slows its trafficking along the bio- artificial membranes: Approaches by pulse EPR spin labeling and single synthesis pathway. Biochemistry 2007; 46: 10960–70. molecule optical microscopy. Biochim Biophys Acta 2003; 1610: Maier O, Baron W, Hoekstra D. Reduced raft-association of NF155 in 231–43. active MS-lesions is accompanied by the disruption of the paranodal Suzuki K, Andrews JM, Waltz JM, Terry RD. Ultrastructural studies of junction. Glia 2007; 55: 885–95. multiple sclerosis. Lab Invest 1969; 20: 444–54. Discovery of Nfasc155H and Nfasc155L Brain 2010: 133; 389–405 | 405

Tait S, Gunn-Moore F, Collinson JM, Huang J, Lubetzki C, Pedraza L, Waxman SG, Ritchie JM. Organization of ion channels in the myelinated et al. An oligodendrocyte cell adhesion molecule at the site of assem- nerve fiber. Science 1985; 228: 1502–7. bly of the paranodal axo-glial junction. J Cell Biol 2000; 150: 657–66. Wolswijk G, Balesar R. Changes in the expression and localization of the Trapp BD, Bernier L, Andrews SB, Colman DR. Cellular and subcellular paranodal protein Caspr on axons in chronic multiple sclerosis. Brain distribution of 2’, 3’-cyclic nucleotide 3’-phosphodiesterase and its 2003; 126 (Pt 7): 1638–49. mRNA in the rat central nervous system. J Neurochem 1988; 51: 859–68. Yan Y, Scott DJ, Wilkinson TN, Ji J, Tregear GW, Bathgate RA. Tuvia S, Garver TD, Bennett V. The phosphorylation state of the FIGQY Identification of the N-linked glycosylation sites of the human relaxin tyrosine of neurofascin determines ankyrin-binding activity and pat- receptor and effect of glycosylation on receptor function. Biochemistry terns of cell segregation. Proc Natl Acad Sci USA 1997; 94: 12957–62. 2008; 47: 6953–68. Van den Steen PE, Opdenakker G, Wormald MR, Dwek RA, Rudd PM. Zhou F, Su J, Fu L, Yang Y, Zhang L, Wang L, et al. Unglycosylation at Matrix remodelling enzymes, the protease cascade and glycosylation. Asn-633 made extracellular domain of E-cadherin folded incorrectly Biochim Biophys Acta 2001; 1528: 61–73. and arrested in endoplasmic reticulum, then sequentially degraded Volkmer H, Hassel B, Wolff JM, Frank R, Rathjen FG. Structure of the by ERAD. Glycoconj J 2008; 25: 727–40. axonal surface recognition molecule neurofascin and its relationship to Zonta B, Tait S, Melrose S, Anderson H, Harroch S, Higginson J, et al. a neural subgroup of the immunoglobulin superfamily. J Cell Biol 1992; Glial and neuronal isoforms of neurofascin have distinct roles in the 118: 149–61. assembly of nodes of Ranvier in the central nervous system. J Cell Biol Wang P, Julian JA, Carson DD. The MUC1 HMFG1 glycoform is a pre- 2008; 181: 1169–77. cursor to the 214D4 glycoform in the human uterine epithelial cell line, HES. Biol Reprod 2008; 78: 290–8.