O -Glycosylation of the Tail Domain of Neurofilament Protein M in Human Neurons and in Spinal Cord Tissue of a Rat Model of Amyotrophic Lateral Sclerosis (ALS)*

Mammalian neurofilaments (NFs) are modified by post-translational modifications that are thought to regulate NF assembly and organization. Whereas phosphorylation has been intensely studied, the role of another common modification, the attachment of O- linked N acetylglucosamine (GlcNAc) to individual serine and threonine residues, is hardly understood. We generated a novel monoclonal antibody that specifically recognizes an O- glycosylated epitope in the tail domain of NF-M and allows determination of the glycosylation state at this residue. The antibody displays strong species preference for human NF-M, shows some reactivity with rat but not with mouse or bovine NF-M. By immu-nohistochemistry and Western blot analysis of biopsy-derived human temporal lobe tissue we show that immunoreactivity is highly enriched in axons parallel to hyperphosphorylated NFs. Treatment of cultured neurons available mouse antibodies belong to the subtype G1, double fluorescence analysis using isotype-specific secondary antibodies becomes possible. Characterization of the antibody by immunocytochemistry revealed that it was sensitive to glutaraldehyde fixation and that the signal was completely lost after fixation for 10 min with 0.3%(v/v) glutaraldehyde.

Neurofilaments (NFs) 1 are the most abundant structural components in large-diameter myelinated axons. Mammalian NFs consist of three major subunits of ϳ200, 160, and 70 kDa, termed NF-H, NF-M, and NF-L, respectively (1). Structurally, the NFs contain a N-terminal head and a helical rod domain that are required for filament assembly (2), and a C-terminal tail domain. NFs are the intrinsic determinant of radial growth of axons, which is most probably mediated by the tail domains of NF-M and NF-H. Whereas genetic deletion of NF-H or its tail domain has relative little effect on radial growth (3), the tail of NF-M has an important role in the formation of cross-links between NFs and modulation of axonal caliber (4). Misaccumulation of NFs is a frequent hallmark of several neurodegenerative diseases including amyotrophic lateral sclerosis (ALS), spinal muscular atrophy, and Alzheimer disease (5).
NFs are among the most highly phosphorylated neuronal proteins. Second messenger-dependent protein kinases such as protein kinase C and cyclic AMP-dependent protein kinase appear to be the principal kinases targeting the head domain. The tail domain is a preferred substrate for second messengerindependent kinases (6), e.g. casein kinase I and II (7), cdk5 (8), and mitogen-activated protein kinases (ERK1/2) (9). Phosphorylation of the head domain appears to regulate the assembly of NFs (10,11), whereas tail domain phosphorylation may affect their ability to control the axon caliber (12). Normally, tail domain phosphorylation appears to be restricted to the axon, but in ALS hyperphosphorylation of this domain occurs also in perikarya (13).
Attachment of O-linked N-acetylglucosamine (GlcNAc) to individual serine and threonine residues is a common post-translational modification of many nuclear and cytoskeletal proteins (14 -16). The enzyme catalyzing O-GlcNAc modification (O-GlcNAc transferase) is essential for cell viability in mammals (17). O-Glycosylation by GlcNAc appears to be as dynamic as protein phosphorylation and is often reciprocal to phosphorylation at the same or adjacent sites (18). Also NFs are modified by O-linked GlcNAc with several identified sites being located in the head region of NF-L and NF-M (19). It is not known how O-glycosylation of NFs is regulated, how it relates to the phosphorylation of NFs, and where O-glycosylated NFs are distributed.
Here we describe a novel monoclonal antibody (NL6) that specifically recognizes an O-glycosylated epitope in the projection domain of NF-M. The antibody displayed a strong species preference for human NF-M. NL6-positive NF-M was enriched in the axons of human neurons in situ. Pharmacological manipulations demonstrated that O-glycosylation as detected with the NL6 antibody is highly dynamic, reversible, and inversely regulated to the activity of mitogen-activated protein kinase kinase (MEK). In a rat model for ALS, NL6 immunoreactivity was strongly decreased compared with wild-type animals.
Preparation of the NL6 Antibody-Mice were immunized with a fraction enriched for components of the human neuronal cytoskeleton. The fraction was prepared from NT2-N neurons and SK-N-BE(2) neuroblastoma cells as described previously (22). The immunization was performed with the PBS-insoluble material that was solubilized with 8 M urea and diluted to a final concentration of 0.85 M urea in PBS. BALB/c mice were injected 4 times intraperitonally with 50 g of the protein fraction at intervals of 3 to 4 weeks. Four days after the final injection, spleens were removed and the spleenocytes were fused with the myeloma line X63Ag8.653 as described previously (23). The resulting hybridoma clones were screened for immunoreaction against cytoskeletal proteins. Positive clones were recloned twice to obtain stable antibody producing hybridoma lines and cultured in RPMI 1640, 10% fetal calf serum (FCS), 292 g/ml glutamine. Antibodies were concentrated by ammonium sulfate precipitation of the hybridoma supernatant, resuspended in PBS, dialyzed against PBS, and stored in the presence of 50% (v/v) glycerol at Ϫ20°C.
For isotype determination, the "isostrip mouse monoclonal antibody isotyping kit" from Roche Diagnostics was used according to the manufacturer's protocol. NL6 is a IgG2a antibody with a -chain. Because most of the commercially available mouse antibodies belong to the subtype G1, double fluorescence analysis using isotype-specific secondary antibodies becomes possible. Characterization of the antibody by immunocytochemistry revealed that it was sensitive to glutaraldehyde fixation and that the signal was completely lost after fixation for 10 min with 0.3%(v/v) glutaraldehyde.
Construction of Expression Vectors-Eukaryotic expression plasmids for hNF-M, mNF-M, and deletion constructs of hNF-M with the Cterminal fused GFP epitope were constructed using the CT-GFP fusion TOPO®TA expression kit (Invitrogen, Karlsruhe, Germany). To construct mutated hNF-M, respective codons were changed using the QuikChange® site-directed mutagenesis kit (Stratagene, Amsterdam, Netherlands). HNF-M(⌬385-472) was prepared by deleting the respective fragment with the restriction enzyme Eco0109I. Constructs with N-terminal-fused GFP were prepared in EGFP C3 Vector (Clontech, Palo Alto, CA). HNF-M(⌬891) was produced using BCUI385, hNF-M(⌬552), with BamHI. To prepare hNF-M(⌬603), an Eco32I restriction site was inserted by changing GT at positions 1813/1814 to AT followed by the respective digest. For hNF-M(⌬789), a Eco32I restriction site was inserted by changing AA at positions 2369/2370 to TG. HNF-M(T747A) was prepared by changing A at position 2248 against G. Mutated constructs were verified by dideoxy sequencing using T7 Sequenase (Amersham Biosciences).
Cell Culture and Transfection-NT2 cells were grown in serum-DMEM (Dulbecco's modified Eagle's medium supplemented with 10% FCS, 5% horse serum (HS), 292 g/ml glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin) and differentiated essentially as described previously (24). NT2-N neurons were obtained after 5 weeks differentiation with retinoic acid and 2 weeks treatment with mitotic inhibitors as described previously (20), plated at 4000 cells/cm 2 onto collagencoated coverslips and cultured for 3 days in serum/Dulbecco's modified Eagle's medium. After this time, almost all neurons had established polarity as judged from axon-specific tau staining. SK-N-BE(2) cells (25) were grown in serum/Opti-MEM (Opti-MEM supplemented with 5% iron enriched FCS, 292 g/ml glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin) and differentiated with 10 M retinoic acid for 2 weeks. 3T3 cells (26) were cultured in serum/Dulbecco's modified Eagle's medium supplemented with 10% FCS, 584 g/ml glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin, HeLa cells (27) in serum/minimal essential Earle's medium (MEM Earle's Medium supplemented with 10% FCS, 292 g/ml glutamine, 50 units/ml penicillin, and 50 g/ml streptomycin) and Neuro2A cells (28) in serum/MEM (minimal essential medium with Earle's salts supplemented with 10% FCS, 292 g/ml glutamine and nonessential amino acids for MEM). In some experiments cells were treated with PUGNAc or the MEK inhibitor PD-98059 for 1 h or the times indicated prior to analysis.
One day before transfections, cells were plated at a density of 3 ϫ 10 4 cells/coverslip on collagen-coated coverslips and transfected with Fu-GENE 6 (Roche Molecular Biochemicals) at a ratio of 1 g of DNA to 4 l of FuGENE per coverslip. Cells were processed for immunocytochemistry 48 h after transfection.
Preparation of Cellular and Tissue Lysates and Cell Fractionation-For preparation of lysates, cells were washed twice with PBS, scraped off into ice-cold PBS, and collected for 5 min at 300 ϫ g. Tissue was shock frozen in liquid nitrogen, cut with a cryostat in sections to allow a better penetration of the lysis buffer, and collected in a test tube. Pellet and tissue slices were resuspended in RIPA buffer (50 mM Tris/ HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS) containing protease and phosphatase inhibitors (1 mM PMSF, 10 g/ml each of leupeptin and pepstatin, 1 mM sodium orthovanadate, 20 mM NaF, 1 mM sodium pyrophosphate), incubated for 30 min at 4°C, centrifuged for 10 min at 13,000 ϫ g, and the supernatant (lysate) collected.
For detergent extraction, SK-N-BE(2) cells from one 15-cm TC dish were washed twice with PBS, scraped off into extraction buffer (10 mM PIPES/KOH, pH 6.8, 50 mM KCl, 3 mM MgCl 2 , 10 mM EGTA), and collected for 5 min at 300 ϫ g with all steps at 4°C. Cells were resuspended in 1 ml of extraction buffer containing 1%(v/v) Triton X-100 and protease inhibitors and incubated for 10 min. 250 l of the solution were transferred on a 0.85 M sucrose cushion and centrifuged for 30 min at 72,000 ϫ g. The pellet was resuspended in 100 l of extraction buffer.
For reassembly of NFs, spinal cords from several adult rats were prepared and frozen in liquid nitrogen. An equal amount (w/v) of disassembly buffer (50 mM MES, pH 6.8, 0.5 mM MgCl 2 , 1 mM EGTA, 1 mM PMSF) was added, the mixture was homogenized on ice using an electrical blade homogenizer and centrifuged for 30 min at 72,000 ϫ g. The supernatant was collected, cleared by centrifugation, and an equal volume of assembly buffer (50 mM MES, pH 6.8, 0.5 mM MgCl 2 , 1 mM EGTA, 8 M glycerol) was added. The mixture was incubated for 15 min at 4°C and centrifuged for 45 min at 75,000 ϫ g. The pellet was resuspended in 100 l of assembly buffer.
Immunohistochemistry, Immunocytochemistry, and Immunoelectron Microscopy-Human temporal lobe tissue was obtained neurosurgically from five patients suffering from focal epilepsy. Material was kept on ice, and gray and white matter were dissected apart. Tissue samples were transferred to liquid nitrogen and kept at Ϫ80°C. For immunohistochemistry, samples were immediately fixed in 4% formalin for 24 h after surgical removal and routinely paraffin-embedded. Deparaffinized 4-m thick sections were blocked and incubated with primary antibodies over-night at 4°C. Delay of formalin fixation for a few hours, or incubation for 1 h at room temperature resulted in substantial loss of staining intensity in 14 samples of surgically resected hippocampi (data not shown). Biotinlabeled secondary antibodies were visualized using peroxidase-conjugated streptavidin with diaminobenzidine as the substrate. All solutions for the visualization of the antigens were contained in a commercially available detection kit (DAKO ChemMate TM , DakoCytomation GmbH, Hamburg, Germany). Staining procedures were performed on an autostainer according to routine protocols provided by the manufacturer (DAKO TechMate 500 TM ). As negative controls, primary antibodies were omitted. The study was approved by the local ethics committee, and informed consent was received from all tissue donors.
Cultured cells were fixed for 20 min with 4% paraformaldehyde in PBS (10 mM phosphate buffer, pH 7.4, 2.7 mM KCl, and 137 mM NaCl) containing 4% (w/v) sucrose at room temperature, washed with PBS, incubated for 20 min with 0.1 M glycine in PBS, and permeabilized with 0.2% (v/v) Triton X-100 for 5 min. Staining was performed as described earlier (30) in PBS containing 1% (w/v) bovine serum albumin using appropriate antibody combinations. 5 g/ml 4,6-Diamidino-2-phenylindole was included in the mixture of the secondary antibodies to detect the nuclei. Cells were analyzed using a Leica DMIRB fluorescence microscope or a Leica TCS SP2 True confocal scanner (Leica, Solms, Germany).
For immunoelectron microscopy, reassembled rat NFs were transferred on pioloform-coated and glow-discharged copper grids (400 mesh), fixed for 20 min with 4% paraformaldehyde in PBS containing 4% (w/v) sucrose, incubated for 20 min with 0.1 M glycine in PBS, and labeled with primary antibodies. After five washes with PBS containing 1% (w/v) bovine serum albumin and 0.1% (v/v) Tween 20, grids were incubated with polyclonal rabbit anti-mouse antibody (ICN Biomedicals, Aurora, OH), washed five times with PBS containing 1% (w/v) bovine serum albumin and 0.1% (v/v) Tween 20, and incubated with protein A coupled to 15 nm gold. Grids were washed with PBS, postfixed for 30 min with 2% (v/v) glutaraldehyde in PBS, washed 3 ϫ 10 min with water, and contrasted with uranyl acetate. Electron microscopy was performed on a Zeiss 10CR electron microscope (Zeiss, Oberkochen, Germany).
Other Methods-Protein concentrations were determined using the BCA microplate assay (Pierce). SDS-polyacrylamide gel electrophoresis and transfer onto polyvinylidene difluoride were performed as described previously (31) using the benchmark protein molecular weight marker (Life Technologies). For dot-blot analysis, samples were diluted with 2 volumes of dot-blot buffer (30% methanol, 0.5% deoxycholate, 9 g/liter NaCl, 1.21 g/liter Tris, pH 7.4) and immobilized on Immobilon NC pure (Millipore Corp., Bedford, MA) using a dot-blot hybridization chamber (Loxo, Dossenheim, Germany). In some cases, blots were stained for total protein using Ponceau S or Coomassie Brilliant Blue according to standard procedures (32). Immunodetection used enhanced chemiluminescence (ECL) (Amersham Biosciences) and was performed according to the manufacturer's protocol. Image acquisition employed a chemo CAM-System (Intas Science Imaging Instruments, Göttingen, Germany) or Fuji LAS3000-System (Raytest, Straubenhart, Germany). Quantification of the blots was carried out with NIH Image 1.61/ppc (rsb.info.nih.gov/nih-image/index.html).
Phosphatase treatment was performed by incubating the membrane for 24 h at 37°C with 20 units/ml alkaline phosphatase (from calf intestinal mucosa) in 100 mM Tris/HCl, pH 8.0, 1 mM PMSF. Phosphatase-treated blots were subjected for immunodetection in parallel with control blots that have been washed with buffer only.
For chemical cleavage of ␤-O-glycosidic linkages, blots were washed for 3 ϫ 10 min with water, incubated for 16 h with 55 mM NaOH at 40°C, and washed again for 3 ϫ 10 min with water. Blots were subjected for immunodetection in parallel with control blots that have been incubated with water only. Enzymatic digestion of ␤-glycosidic linkages was performed with Triton-extracted fractions of BE2 cells. 20 g of protein were incubated with 0.5 unit of ␤-N-acetylglucosaminidase (from bovine kidney) in a total volume of 50 l in 10 mM sodium cacodylate, pH 5.0, 1 mM PMSF, and 10 g/ml each of leupeptin and pepstatin for 16 h at 37°C. The reaction was stopped by addition of 12.5 l of 5ϫ Laemmli buffer. Control samples were treated identical, but in the absence of ␤-N-acetylglucosaminidase. 50% of the samples were subjected for immunodetection and processed in parallel for NL6 and M20 staining.
Partial chymotryptic digest of intact NFs was performed with Tritonextracted fractions of BE2 cells. 10 g of protein were incubated with 50 ng of chymotrypsin in a total volume of 18 l in digestion buffer (100 mM Tris/HCl, pH 7.8, 10 mM CaCl 2 , 10 g/ml each of leupeptin and pepstatin). The reaction was stopped after different times by addition of PMSF to 4 mM. Samples were centrifuged for 20 min at 4°C and 12,800 ϫ g and equal amounts of pellet and supernatant were analyzed by immunoblot. Comparison between experimental groups was based on paired Student's t test (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001).

NL6, A Novel Monoclonal Antibody against a Human Neuronal Cytoskeleton Fraction, Recognizes an O-Glycosylated
Epitope of Human NF-M-In an immunological approach to obtain novel antibodies against components of the human neuronal cytoskeleton, a PBS-insoluble cytoskeletal fraction of human model neurons was prepared and solubilized with urea (see "Experimental Procedures"). The material was used to immunize mice and produce hybridoma lines. One of the antibodies (NL6) detected a protein with an apparent molecular mass of ϳ160 kDa. The antigen was highly enriched in the Triton X-100-insoluble fraction of homogenates from differentiated human neuroblastoma (SK-N-BE(2)) cells (Fig. 1A), indicating that NL6 recognizes an intermediate filament protein.
The NL6 antigen was up-regulated during differentiation of human neuroblastoma cell lines parallel to NF-M (SK-N-BE(2), SH-SY5Y cells; Fig. 1B, left panel). To determine whether the antigen of NL6 is also expressed in neuronal precursor cells and when it is produced during differentiation, the NT2/NT2-N system was employed. NT2 cells are a human teratocarcinoma cell line with characteristics of central nervous system neuronal progenitor cells that can be differentiated in vitro to yield terminally differentiated, polar, and postmitotic neurons (24). The NL6 antigen was present in lysates of NT2-N cells. In contrast, no signal was observed in NT2 cell lysates (Fig. 1B,  right panel), indicating that the NL6 antigen is expressed only in differentiated neurons. The expression pattern was similar albeit not identical to total NF-M as detected using a previously described antibody recognizing NF-M independent of posttranslational modifications (M20 (21,33)). Taken together the data strongly suggest that NL6 detects a subpopulation of NF-M.
It is known that in neurons NF-M is highly post-translationally modified (for a review, see Ref. 1). To test whether NL6 detects a phosphorylated epitope, cellular homogenates were separated by SDS-PAGE, blotted onto polyvinylidene difluoride, and treated with alkaline phosphatase. Phosphatase treatment resulted in a strong reduction of the immunoreactivity with an antibody that detects phosphorylated epitopes in NF-H and NF-M (SMI31) but caused no reduction in M20 or NL6 immunoreactivity (Fig. 1C, arrowheads) indicating that binding of NL6 does not depend on phosphorylation of the epitope. To test whether NL6 detects an O-glycosylated epitope, blots from cellular lysates were treated with 55 mM NaOH, which is known to remove ␤-glycosidically linked Glc-NAc from serine or threonine residues by a reaction called ␤-elimination (34). The treatment resulted in a complete loss of immunoreactivity with an antibody that recognizes O-linked GlcNAc independent of protein determinants (CTD110.6 (35); Fig. 1D, top left panel). Blots from fractions enriched for NFs were treated similarly and probed with NL6 and M20 antibody. Treatment with NaOH resulted in an almost complete loss of NL6 immunoreactivity, whereas M20 staining remained unchanged (Fig. 1D, top right panel). To check whether GlcNAc at the NL6 epitope can also be removed enzymatically, a fraction enriched for neurofilaments was incubated with ␤-N-acetylglucosaminidase from bovine kidney. The treatment resulted in a strong reduction in NL6 immunoreactivity compared with a control sample, whereas M20 staining remained unchanged (Fig. 1D, bottom left panel). Furthermore, inclusion of 1 M GlcNAc in the primary antibody reaction abrogated binding of the NL6 antibody (Fig. 1D, bottom right panel). Taken together the data indicate that NL6 recognizes an O-glycosylated epitope of NF-M.
To test whether NF-M remains O-glycosylated at the NL6 epitope during disassembly and reassembly, NFs were reconstituted from high-speed supernatants of rat spinal cord homogenates and analyzed by immunoelectron microscopy. NL6 immunoreactivity was evenly distributed on reconstituted NFs similar to M20 reactivity (Fig. 1E). The data indicate that O-glycosylated NF-M is assembly-competent and uniformly incorporates into NFs.
The NL6 Epitope Is Localized in the Tail Domain of NF-M-NF-M consists of a central ␣-helical rod domain that forms the filament backbone, a small globular head domain, and an extended carboxyl-terminal tail domain that protrudes from the backbone (1,36). To determine in which of these regions the NL6 epitope is localized, a partial chymotryptic digest of intact NFs was performed. It has previously been shown that, under these conditions, at first the tail domain is cleaved off from the remainder of the protein close to the filament backbone, which can then be separated by centrifugation (37). After short treatment with chymotrypsin (2 min), NL6 reactivity was exclusively present in the supernatant at molecular masses of about 100 and 110 kDa ( Fig. 2A), which closely corresponds to the apparent size of the tail domain as reported earlier (37). After longer incubation periods (Ն8 min) NL6 immunoreactivity disappeared because of further digestion of the protein (data not shown). The data indicates that the NL6 epitope is localized in the carboxyl-terminal tail domain of NF-M.
As a tool to determine the position of the NL6 epitope, a fusion protein between human NF-M and GFP was constructed and transfected into the human neuronal precursor cell line NT2. As shown above (Fig. 1B) and confirmed by immunocytochemistry (data not shown), these cells do not express detectable amounts of endogenous NF-M. Individual transfected cells showed a filamentous pattern of GFP fluorescence (Fig. 3B, top  left) indicative of the incorporation of the exogenous construct into the cellular intermediate filaments. NL6 staining colocalized with the GFP fluorescence (Fig. 3B, top right) indicating that exogenously expressed NF-M becomes O-glycosylated at the NL6 epitope. To map the region of NL6 immunoreactivity, a panel of deletion mutants and mutated constructs were prepared and tested in similar transfection assays. The only known glycosylation site in the tail domain of NF-M is localized at Thr 430 (19). However, NL6 immunoreactivity persisted after deletion of a part of NF-M containing Thr 430 (hNH-M(⌬385-472)) (Fig. 2B, middle), indicating that NL6 recognizes a so far unknown glycosylation site on NF-M. The deletion resulted in a loss of the filamentous pattern and the formation of immunoreactive aggregates because of the deletion of part of the core domain that is known to be required for filament assembly (2). A construct truncated at amino acid 789 was immunoreactive against NL6 (Fig. 2B, top), whereas immunoreactivity was lost with a construct truncated at amino acids 603 (Fig. 2B, bottom) or 552 (Fig. 2C). This indicates that the epitope is located between 604 and 789. This region encompasses the KSP repeat region (amino acids 613-691), which consists of 6 repeats con-  (2) cells (BE2), SH-SY5Y cells, and NT2-N cells similar to NF-M as detected using an antibody (M20) that recognizes NF-M independent of posttranslational modifications. C, effect of phosphatase treatment on immunoreactivity against the NL6 epitope. Phosphatase treatment strongly reduces immunoreactivity against the antibody SMI31 that reacts with a phosphorylated epitope in NF-H (arrow) and NF-M (arrowhead) but does not affect immunoreactivity against M20 or NL6 (arrowheads), indicating that the NL6 epitope is phosphorylation independently. D, effect of ␤-elimination, glucosaminidase treatment, and competition with GlcNAc on NL6 immunoreactivity. Treatment with NaOH strongly reduces reactivity of cellular lysates against an antibody that detects ␤-glycosidically linked N-acetylglucosamine and NL6 immunoreactivity (top). Glucosaminidase treatment resulted in a strong reduction of NL6 immunoreactivity (bottom left) and inclusion of GlcNAc in the antibody reaction abrogated NL6 binding (bottom right), indicating that NL6 recognizes an O-glycosylated epitope of NF-M. E, immunoelectron micrograph of reconstituted NFs. NL6 immunoreactivity is evenly distributed on NFs indicating assembly competence of O-glycosylated NF-M. Cell culture, fractionations, immunoblot analysis, treatment with phosphatase, ␤-elimination, and electron microscopy were performed as described under "Experimental Procedures." For A, 2% (v/v) of each fraction was separated by SDS-PAGE on 10% acrylamide and stained with Coomassie (left) or NL6 (right). For B, 20 (BE2/SH-SY5Y; immunoblot with NL6, M20), 10 (NT2/NT2-N; immunoblot with NL6 and M20), and 1 g (immunoblot with actin) of protein lysates were separated by SDS-PAGE on 7.5 (NL6/M20) and 10% (actin) acrylamide. For C, 10 or 20 g of protein lysates from BE2 cells were separated by SDS-PAGE on 7.5% acrylamide. For ␤-elimination in D, 1 (GlcNAc) and 10 g (NL6/M20) of Triton extracts from BE2 cells were separated by SDS-PAGE on 7.5% acrylamide, blotted, and incubated with or without 55 mM NaOH and immunostained. Following immunodetection, the blot was stained with Ponceau Red for total protein detection. After glucosaminidase treatment, 10 g of protein were sep-sites, to recognize the sequence context and surface accessibility (38). One of the predicted O-glycosylation sites was located outside the KSP repeat region at position Thr 747 . To test whether this site is part of the epitope we mutated Thr 747 to alanine. This construct was still immunoreactive with NL6 after transfection, suggesting that Thr 747 is not involved but that the glycosylated NL6 epitope is localized within the KSP repeat region. It was not possible to mutate the candidate sites within the KSP region directly by site-directed mutagenesis because of the repetitive nature of this region. Taken together, the data indicate that the NL6 epitope is localized in or close to the KSP region of human NF-M.
Immunoreactivity of the NL6 Epitope Shows a Strong Preference for Human Species-NFs are highly conserved across species. The region where most changes occur represents the KSP repeat region that displays a major deletion of about 60 amino acids in mouse and rat compared with the human sequence. Thus, also NL6 immunoreactivity may differ across species. In fact, NFs from mouse and bovine spinal cord were not reactive with NL6 in immunoblot analysis (Fig. 3A). In contrast, some signal, albeit less than with human material, was observed with rat NFs. In agreement with the immunoblot results, NL6 immunoreactive neurons were observed in organotypic cultures of rat but not of mouse brain (data not shown). To test whether glycosylation of the NL6 epitope depends on the sequence or is cell type-specific, GFP-tagged NF-M from mouse (mNF-M) was transfected into human neuronal (NT2) and non-neuronal (HeLa) cells. The construct was expressed as evidenced by the filamentous GFP fluorescence and reactivity with M20 (Fig. 3B, bottom panel). No staining with NL6 was observed in both cell lines, indicating the absence of the respec-tive immunoreactive epitope. As expected, staining was also absent in transfected neuronal and non-neuronal mouse lines (Neuro 2A and 3T3) (Fig. 3C). In contrast, expression of human NF-M (hNF-M) in any of these cell lines resulted in NL6 immunoreactivity, indicating that formation of the NL6 epitope depends on the NF-M sequence but not on the cell type where it is expressed.
The number and distribution of the repeats strongly differs among vertebrates. In human NF-M, 12 KSP repeats are located between residues 611 and 688, rat and mouse NF-M contain only 2 KSP repeats in the respective region (Fig. 3D). The repeats of the human sequence are very similar to each other and in five of six repeats the sequence PVPK, which reflects, except for the missing Gln, the glycosylation motif of O-GlcNAc transferase is located in front of the second KSP sequence. The rat KSP region is identical to human NF-M, whereas the mouse sequence exhibits a Val to Met exchange. The NL6 epitope may encompass the PVPKS sequence with Ser being O-glycosylated. Because this sequence is present five times in human NF-M but only once in rat NF-M, this may explain the stronger immunoreactivity of human versus rat material. The PVPKS motif is missing in the bovine NF-M sequence, which is consistent with the lack of immunoreactivity in this material.

O-Glycosylated NF-M as Recognized by NL6 Is Enriched in the Axons of Human Neurons in
Situ-It is known that phosphorylated subpopulations of NF-M are enriched in the axon in situ as detected by phosphorylation-sensitive antibodies (39). However, it is not known whether differentially glycosylated NFs also show a compartment-specific distribution. Because O-glycosylation may be very labile during post-mortem proces-

FIG. 2. Localization of the NL6 epitope on NF-M.
A, immunoblot after chymotryptic cleavage of intact NFs. After cleavage, NL6 detects protein fragments at 100 and 110 kDa in the supernatant (sup.), indicating the localization of the NL6 epitope in the tail domain as represented in the schematic drawing. B, double fluorescence staining for GFP and the NL6 epitope of NT2 cells transfected with GFP-tagged NF-M deletion constructs. NL6 reactivity is detected in NF-M truncated at amino acid 789 (hNF-M(⌬789)). No NL6 reactivity is detected in the fragment truncated after amino acid 603 (hNF-M(⌬603)), indicating that the NL6 epitope is located between amino acids 604 and 789. Note that the construct hNF-M(⌬385-472), where part of the core domain is deleted, is deficient to assemble into filaments but still forms immunoreactive aggregates. C, schematic representation of the constructs, position of known and predicted O-GlcNAc sites, and fluorescence staining results after transfection. The data suggest that the O-GlcNAc site that is detected by the NL6 epitope is located in or close to the KSP region. Chymotryptic cleavage, immunoblot analysis, construction of expression plasmids, and immunofluorescence staining were performed as described under "Experimental Procedures." For A, equal amounts of pellet and supernatant from samples that have been treated with or without chymotrypsin were separated by SDS-PAGE on 10% acrylamide. Scale bar, 10 m. sion of human material, we aimed to perform NL6 staining on as fresh material as possible. In tissue from human temporal lobe obtained at surgery of epilepsy patients, NL6 immunoreactivity was highly enriched in the subcortical white matter similar to staining with an antibody against phosphorylated NF-L that is known to be enriched in the axon (2F11; Fig. 4A). Both antibodies labeled radial and horizontal fibers as well as recurrent collaterals in the infragranular layers (layers IV-VI). In high power magnifications it was evident that NL6 immunoreactivity was exclusively found in axons, whereas 2F11 also labeled the perinuclear soma of some neurons (Fig. 4B, asterisks). We also performed immunoblot analysis with NL6 and M20 antibody of temporal lobe tissue lysates from white and gray matter obtained from 5 different patients (Fig. 4C). Quantitation revealed that the ratio of immunoreactivity of white versus gray matter was significantly higher (p Ͻ 0.001; n ϭ 5) with NL6 compared with M20 that recognizes total NF-M (Fig.  4D), which was well in accordance with the findings from immunohistochemistry. When autopsy-derived human brain material with considerable postmortem intervals (12-72 h) was used for immunoblot analysis, no significant enrichment of NL6 immunoreactivity in white matter was observed (data not shown), indicating that O-glycosylation of NF-M in neuronal axons is very sensitive to degradation. The data suggest that O-glycosylated NF-M as recognized by NL6 is a dynamic modification that is highly enriched in the axons of human neurons in situ in parallel to phosphorylated NF-L subpopulations.

O-Glycosylation of NF-M Is Dynamically Influenced by Inhibitors of GlcNAcase and Mitogen-activated Protein Kinase
Kinase-O-Glycosylation of cellular proteins appears to be highly dynamic similar to phosphorylation (40 -42). In some instances a direct reciprocity of phosphorylation and O-glycosylation on the same or neighboring serines or threonines was observed (43,44). To test the dynamicity of O-glycosylation at the NL6 epitope, the human neuroblastoma cell line SK-N-BE(2) was treated with the drug PUGNAc, a membrane-permeable derivative of N-acetylglucosamine that binds and competitively inhibits the cytoplasmic GlcNAcase (45). Immunoblot analysis of cellular lysates revealed that PUGNAc increased NL6 immunoreactivity relative to total NF-M in a concentrationdependent manner with a maximum of 140 -150% compared with the control (set at 100%) (Fig. 5A). PUGNAc caused a fast increase in NL6 immunoreactivity because already after 10 min a significant rise was observed (Fig. 5B). The changes in immunoreactivity were fully reversible indicating fast O-glycosylation and -deglycosylation at the NL6 epitope. Total Oglycosylation as detected using an antibody that recognizes O-linked GlcNAc independent of protein determinants showed a very similar increase after PUGNAc treatment (Fig. 5C), suggesting that the dynamic glycosylation/deglycosylation at the NL6 epitope closely reflects changes in the general level of O-glycosylation within the cell. It is known that the KSP repeat region of NF-M is phosphorylated by the proline-directed protein kinase ERK1/2 (9), which is the effector kinase of the mitogen-activated protein (MAP) kinase pathway. Thus, to test for the effect of phosphorylation in the KSP region, cells were treated with PD-98059, a specific inhibitor of activation of MEK (46). PD-98059 caused an about 35% decrease of activated (Thr 202 /Tyr 204 phosphorylated) ERK1/2 relative to total ERK1/2, indicating inhibition of MEK and an increase in relative NL6 immunoreactivity (Fig.  5D). This suggests that phosphorylation of NF-M in the tail domain dynamically affects O-glycosylation in the same region and is reciprocally related to O-glycosylation at the NL6 epitope. Treatment of the cells with the protein phosphatase 2A inhibitor cyclosporin A (5 nM, 2 h) resulted in an about 25% decrease in relative NL6 immunoreactivity, which is also consistent with a reciprocal O-glycosylation and phosphorylation (data not shown). Considering these findings and the fact that neurofilament shows an ongoing process of phosphorylation/ dephosphorylation in human brain tissue ex vivo, the rapid loss of NL6-reactive epitopes as described above might well be explained by a co-regulated deglycosylation (47). However, it cannot be excluded from these experiments that MEK inhibition and cyclosporin A treatment exert a systemic effect on NF-M glycosylation that is independent of the phosphorylation of the KSP repeats.
O-Glycosylation of NF-M at the NL6 Epitope Is Down-regulated in a Transgenic Rat Model of ALS-To determine a potential pathophysiological role of disturbed O-glycosylation in neurodegenerative diseases, an animal model for ALS was employed. Because NFs from mouse tissue were not reactive with NL6 (Fig. 3A), a previously established transgenic rat model where Cu 2ϩ /Zn 2ϩ -superoxide dismutase 1 harboring an ALS-linked familial genetic mutation (SOD1 G93A ) is overexpressed and that develops ALS-like pathology was used (48). NL6 immunoreactivity was strongly reduced in spinal cord extracts from mutant animals compared with wild-type rats (Fig. 6A). The reduction was evident at postnatal day 0 as well as after disease onset in end-stage animals that expressed full pathology. Also total NF-M reactivity decreased in these animals indicating neurofilament loss in axons. To determine the decrease in NL6 immunoreactivity compared with changes in total NF-M, blots were quantitated and relative ratios of NL6 and RMO44 reactivity to tubulin from transgenic animals were determined and compared with the respective ratios in wildtype animals. At all time points, the reduction in NL6 immunoreactivity was higher than in total neurofilament (Fig. 6B). In particular, in spinal cord material from end-stage animals, NL6 immunoreactivity was barely detectable. In contrast, reactivity for SMI31 (phosphorylated NF-M) was increased compared with total NF-M at every age analyzed. The data suggest that O-glycosylation at the NL6 epitope is selectively decreased during ALS, whereas phosphorylation is increased. DISCUSSION Using a cytoskeletal fraction prepared from human neurons as immunogenic material we generated a novel monoclonal antibody that specifically detects an O-glycosylated epitope in NF-M. To our knowledge, this is the first antibody that specifically detects this modification in NF-M and one of very few antibodies that recognizes O-linked GlcNAc in cellular proteins. An example for the latter is an antibody that detects O-GlcNAc at threonine 58 in c-Myc protein and that was used FIG. 4. Distribution of NL6 immunoreactivity in brain tissue. A, staining of human temporal lobe tissue with NL6 and 2F11. NL6 immunoreactivity is highly enriched in the subcortical white matter similar to 2F11 staining, which detects an axonally enriched phosphorylated NF-H isoform. B, temporal lobe tissue stained for NL6 and 2F11 epitopes at higher magnification. NL6 reactivity is restricted to radial and horizontal fibers, whereas staining of perinuclear soma of some neurons is observed with 2F11 (asterisks), as also shown at high power magnifications in the insets. C and D, immunoblot (C) of gray and white matter of temporal lobe tissue lysates from five different patients, and quantitation of the relative ratio (D) for NL6 and M20 immunoreactivity. The ratio of white versus gray matter is significantly increased for NL6 reactivity compared with M20 (***, p Ͻ 0.001, n ϭ 5; paired Student's t test), indicating a strong axonal enrichment of the NL6 epitope in situ. Tissue preparation, stainings, and immunoblot analysis were performed as described under "Experimental Procedures." For C, 10 or 30 g of protein lysates were separated by SDS-PAGE on 7.5% acrylamide. Nuclei in sections A and B were counterstained with hematoxylin. Scale bars, 20 m (insets of B), 1 mm (A), and 100 m (B).
to analyze the interplay between phosphorylation and glycosylation at this position (49). Several antibodies that more generally detect O-GlcNAc-modified proteins have been developed. However, most of the antibodies are limited in their use because they only detect part of all modified proteins and require additional epitopes (50 -52). An exception is an antibody that has been developed by Comer et al. (35) that detects O-GlcNAc independent from the protein backbone and is very useful to determine overall O-GlcNAc modifications (see Figs. 1D and 5C).
Many antibodies have been described that detect phosphorylated epitopes in NF proteins, mostly in NF-H, which is heavily phosphorylated with a stoichiometry of about 20 mol of phosphate/mol of protein (53). NF-M is much less phosphorylated with about 4 -9 mol of phosphate/mol of protein. The epitopes that are detected by phosphorylation-dependent antibodies are located predominantly in the tail region of NF-M and NF-H that contain multiple repeats of lysine-serine-proline (KSP). Especially the serines in KSP repeats appear to be heavily phosphorylated in axons (54 -56). The number and distribution of the repeats strongly differ among vertebrates. NF-H contains 42-51 of these repeats. In human NF-M, 12 KSP repeats are located between residues 614 and 689, rat and mouse NF-M contain only 2 KSP repeats in the respective region. We have mapped the position of the O-glycosylated epitope that is detected by the NL6 antibody in or close to the KSP repeat region between residues 604 and 789. This region contains 3 predicted sites for O-linked GlcNAc modifications with two of them being located in the KSP region and one (Thr 747 ) carboxyl-terminal flanking it. Mutating Thr 747 to alanine did not affect NL6 immunoreactivity, suggesting that the site that is recognized by the antibody is located within the KSP repeats. It should, however, be noted that because of the low number of identified sites that are modified by O-linked GlcNAc no strict consensus sequence can be discerned yet. Thus, it cannot be completely excluded that the NL6 antibody detects an O-glycosylated epitope outside the KSP repeats.
We have shown that O-glycosylation of NF-M can quickly change, is fully reversible, and is affected by changes in protein phosphorylation. It is also very sensitive to postmortem/ex vivo delay in the processing of neuronal tissue. Thus, O-glycosylation of NF-M in neurons appears to be highly dynamic similar to the rapid turnover of O-GlcNAc-modified proteins after lymphocyte activation (40) or the dynamic O-GlcNAcylation of the small heat shock protein ␣B-crystallin (42).
The level of glycosylation can be very different in various cell types (45). Thus, also during neuronal development, the cycling enzymes that are responsible for GlcNAcylation may be differentially regulated. This would result in a change of the level of NL6 reactivity relative to total NF-M. We did not observe obvious changes during the differentiation of NT2 cells, SK-N-BE(2) cells, or SH-SY5Y cells, indicating that the enzymes that are responsible for the GlcNAc modification are constitutively active in neuronal precursor cells and neurons. In agreement, strong glycosylation was observed also in non-neuronal cells both of human and mouse origin after transfection of human NF-M sequence. O-Glycosylation could be increased by pharmacological inhibition of the deglycosylating enzyme with PUGNAc by 40 -50%. Thus, the amount of glycosylation is in a medial range that would allow the cell to react to stimuli with a change in the glycosylation level. A similar increase after PUGNAc treatment was observed for the O-glycosylation of all proteins, indicating that NF-M glycosylation closely reflects the general activity of the cycling enzymes.
O-Glycosylation of NF proteins may affect their functional properties. We did not observe a differential stability of NL6positive NF-M versus total NF-M to urea-induced disassembly of NFs (data not shown) and O-glycosylated NF-M incorporated M PUGNAc on total O-glycosylation as determined from the relative ratio of immunoreactivity with an antibody that detects ␤-glycosidically linked GlcNAc and actin. D, effect of the MEK inhibitor PD-98059 on ERK1/2 phosphorylation and the relative ratio of NL6 to M20 reactivity. Incubations were for 1 h. Activated (Thr 202 /Tyr 205 phosphorylated) ERK1/2 is decreased relative to total ERK1/2, and reactivity against the NL6 epitope relative to total NF-M is increased after inhibition of MEK. Cell culture of SK-N-BE(2) cells and immunoblot analysis were performed as described under "Experimental Procedures." For quantitations by Western blot (A, B, and D), 20 g of protein lysates were separated by SDS-PAGE on 7.5% acrylamide. For dot-blot analysis (C), 15 g of protein lysates were immobilized on nitrocellulose. Mean Ϯ S.E. are shown. * indicates significant differences to carrier control experiments (p Ͻ 0.05; n ϭ 5; paired Student's t test).
into reassembled NFs similar to total NF-M. It is known that the head and rod of NF proteins, but not the tail, are essential for assembly and that phosphorylation of the head region of intermediate filaments affects assembly or disassembly (10,11,57,58). Because the NL6 epitope is located in the tail region, it is conceivable that O-glycosylation in this region does not affect assembly properties but other features that are thought to be mediated by the tail region. These could include modulation of the radial growth of axons that are critically affected by the tail domain of NF-M (4), interactions between filaments and other cellular components, and slow axonal transport rates that are also thought to be affected by phosphorylation in the tail region (1,6). Very recent data have implicated a role for NF-M for myelin-directed "outsidein" signaling cascades (59). It will be interesting to determine the effect of O-glycosylation on these features. Because Glc-NAc is neutral, it is possible that it modulates the surface properties of NFs different from a strongly negatively charged phosphate group that may repel the filaments and increase radial growth (54,60).
It is known that the tail regions of NF-M and NF-H are heavily phosphorylated in axons (54 -56). In some cases, a reciprocal relation between phosphorylation and O-glycosylation ("yin-yang" relationship) has been reported (15,61). However, in contrast to what might be expected from these results, also the GlcNAc modification in the tail region shows a strong enrichment in the white matter, which appears to be even more pronounced than the distribution of the phosphoepitope in that it is completely absent from neuronal cell bodies in situ. Thus, in white matter, phosphorylation and O-glycosylation of NF-M are increased in parallel. Dong et al. (62) suggested that Oglycosylation of the KSP region in NF-H occurs directly after synthesis before the subunits are incorporated into NFs and are transported into the axon. In the axon, GlcNAc would be FIG. 6. Expression of the NL6 epitope in spinal cords of SOD1 transgenic rats. A, immunoblots of spinal cord extracts from wild-type and mutant SOD1 G93A expressing rats at postnatal day 0 (top) and at the end-stage of the disease (bottom). NL6 immunoreactivity is strongly reduced in NF-M from mutant animals compared with wild-type rats. B, relative intensities of the immunoreactivity against total NF-M (RMO44), the NL6 epitope, and phosphorylated NF-M (SMI31) of spinal cord extracts from wild-type and mutant SOD1 G93A expressing rats of different ages. NF-M from mutant animals shows a stronger reduction of immunoreactivity against the NL6 epitope compared with total NF-M at every age analyzed, whereas phosphorylated NF-M is increased. Preparation of spinal cord extracts and immunoblot analysis were performed as described under "Experimental Procedures." Immunoblots of spinal cord extract from animals with different ages were quantified for NL6, SMI31, and RMO44 signals and normalized to tubulin (monoclonal antibody DM1A). Three animals of the same genotype were analyzed for every time point, and the data for wild-type animals were set as 1.0. Blots were performed at least twice. Mean Ϯ S.E. are shown. removed and substituted by phosphate groups. At least with respect to NF-M our data are not consistent with such an hypothesis. In contrast, the data argue for a synchronous phosphorylation and O-glycosylation of the tail region of NF-M within the axon. Despite this synchronicity of phosphorylation and O-glycosylation, inhibition of MEK with PD-98059 increased O-GlcNAcylation at the NL6 epitope, suggesting a possible reciprocal regulation between phosphorylation and Oglycosylation in the KSP region. Whether such a reciprocal regulation occurs also in vivo remains to be shown.
The development of a polar cytoarchitecture alone was not sufficient to cause a compartment-specific enrichment of Oglycosylation because NL6 immunoreactivity was present in all processes and in the cell body of dissociated human model neurons (data not shown). It is possible that O-glycosylation in the axon is ultimately linked to, and may be regulated by, myelination as has been suggested for phosphorylation (1). It will be very interesting to determine whether O-glycosylation of NF-M as detected using the NL6 antibody is affected by myelination during neuronal development or in neuroinflammatory diseases such as multiple sclerosis. Because phosphorylated neurofilaments and O-GlcNAc-linked glycopeptides are abnormally distributed in damaged motor neurons in infantile spinal muscular atrophy (63), and detergent-insoluble cytoskeletal proteins show increased O-GlcNAcylation in Alzheimer brains (64), NL6 might also become a valuable tool to further investigate the pathophysiological role of disturbed O-glycosylation in common neurodegenerative diseases.
We have shown that in an animal model for ALS, O-glycosylation of NF-M as detected using the NL6 antibody is decreased, whereas phosphorylation is increased relative to total NF-M. Hyperphosphorylation of NF proteins has been described in patients with ALS (65), and in mice and rats transgenic for SOD1 with familial ALS-linked mutations (48,66,67), which is in agreement with our observation. In ALS, also an increased nonenzymatic glycation (leading to an irreversible formation of advanced glycation end products) has been reported (reviewed in Ref. 68). However, this reaction is mechanistically completely different from the dynamic enzymatic O-GlcNAcylation of many nuclear and cytoskeletal proteins and both reactions appear to occur independently. In fact, the reciprocal changes in NF-M phosphorylation and O-GlcNAcylation are in agreement with a reciprocal regulation of phosphorylation and enzymatic O-glycosylation as reported earlier (18). The NL6 antibody might become a valuable tool to further investigate the pathophysiological role of disturbed O-glycosylation in common neurodegenerative diseases.