Cytoplasmic O-GlcNAc modification of the head domain and the KSP repeat motif of the neurofilament protein neurofilament-H.

Neurofilaments, the major intermediate filaments in large myelinated neurons, are essential for specifying proper axonal caliber. Mammalian neurofilaments are obligate heteropolymers assembled from three polypeptides, neurofilament (NF)-H, NF-M, and NF-L, each of which undergoes phosphorylation at multiple sites. NF-M and NF-L are known to be modified by O-linked N-acetylglucosamine (O-GlcNAc) (Dong, D. L.-Y., Xu, Z.-S., Chevrier, M. R., Cotter, R. J., Cleveland, D. W., and Hart, G. W. (1993) J. Biol. Chem. 268, 16679-16687). Here we further report that NF-H is extensively modified by O-GlcNAc at Thr53, Ser54, and Ser56 in the head domain and, somewhat surprisingly, at multiple sites within the Lys-Ser-Pro repeat motif in the tail domain, a region in assembled neurofilaments known to be nearly stoichiometrically phosphorylated on each of the ∼50 KSP repeats. Beyond the earlier identified sites on NF-M and NF-L, O-GlcNAc sites on Thr19 and Ser34 of NF-M and Ser34 and Ser48 of NF-L are also determined here, all of which are localized in head domain sequences critical for filament assembly. The proximity of O-GlcNAc and phosphorylation sites in both head and tail domains of each subunit indicates that these modifications may influence one another and play a role in filament assembly and network formation.

Neurofilaments (NFs), 1 major 10 nm neuronal intermediate filaments, are the most abundant cytoskeletal components in large myelinated axons (1) and have been proven to be intrinsic determinants of axonal caliber (2)(3)(4)(5)(6)(7), which in turn determines the axonal conduction velocity of action potentials (8). Abnormalities in NF organization may play a pivotal role in the etiology of motor neuron diseases, such as amyotrophic lateral sclerosis (7,9). Recent transgenic mouse models that overexpress wild type subunits (10,11) or express mutant forms (12) of NF subunits have closely mimicked the pathology of amyotrophic lateral sclerosis.
NFs are obligate heteropolymers assembled from three NF subunits, NF-H (115 kDa), NF-M (100 kDa), and NF-L (62 kDa) (13,14). Structurally, each subunit has a globular head and a variable length tail domain separated by a 310-amino acid helical rod domain (15). The three subunits are synthesized in neuronal cell bodies, transported into axons, and extensively post-translationally modified in their head and tail domains. The aberrant mobility of NF-H and NF-M on SDS-polyacrylamide gel electrophoresis (PAGE) is caused by the high degree of phosphorylation of tail domains (16 -19). For example, the rat NF-H tail contains 52 Lys-Ser-Pro (KSP) repeats (20), and the serine residue of each repeat can be phosphorylated yielding nearly stoichiometric modification post assembly in the internodal segments of axons (16 -19). It has been proposed that as a consequence of repulsive force mediated by the negative charged phosphates, phosphorylation of both NF-H and NF-M may contribute to the flexibility of the carboxyl tails and modulate NF density and interfilament spacing by altering cross-bridges between NFs or the interactions of NFs and other cytoskeletal components (1,17,(21)(22)(23)(24)(25), whereas hypophosphorylation is correlated with the close packing of filaments in cell bodies, in dendrites, at nodes of Ranvier (23)(24)(25), and in in vitro reassembled NFs (26).
The head domains of NF proteins are rich in serine and threonine residues and are also phosphorylated. Several sites have been identified on both NF-M and NF-L (27)(28)(29). The head domain itself is known to be crucial for the formation of 10 nm filaments (30 -32). Besides phosphorylation, we have previously reported that NF-M and NF-L are also modified posttranslationally with O-linked N-acetylglucosamine (O-GlcNAc) on serine and threonine residues (33). This type of glycosylation is the simplest protein modification with sugars (34). Initially discovered from a glycosylation study of murine lymphocytes (35), this modification has subsequently been found within the nucleoplasmic and cytoplasmic compartments of virtually all eukaryotic cells (36). O-GlcNAc transferase, an enzyme responsible for the addition of GlcNAc to the serine and/or threonine residues of peptides/proteins, and a cytoplasmic N-acetyl-␤-D-glucosaminidase that selectively cleaves O-GlcNAc from glycopeptides and proteins have both been purified (37,38). The O-GlcNAc modification (termed as O-GlcNAcylation) is thus both highly abundant and dynamic in a manner quite similar to phosphorylation (39 -43), suggesting a regulatory role for O-GlcNAcylation in a variety of biological processes. In order to explore the possible role(s) of O-GlcNAcylation in NF functions, we extend further our original study and show that not only NF-M and NF-L but also NF-H are modified by O-GlcNAc. Most strikingly, we document that multiple serines in the ϳ50 KSP repeats of the tail domain of NF-H are O-GlcNAcylated, suggesting that it regulates the properties of the tail domain by directly competing and modulating phosphorylation.

EXPERIMENTAL PROCEDURES
Materials-Spinal cords were taken from 6-month-old or older male rats. UDP-[6-3 H]galactose (38 Ci/mmol) was obtained from Amersham Corp. Bovine milk galactosyltransferase (GT'ase), chymotrypsin, and aprotinin were purchased from Sigma, and GT'ase (37 unit/ml) was pregalactosylated as described (36). Sequencing grade trypsin was from Boehringer Mannheim. Mixed monosaccharide standards were from Dionex. All other chemicals were of the highest quality commercially available.
Purification of NF Triplet Proteins-Total NFs comprising mainly NF triplet proteins were prepared from about 15 g of frozen rat spinal cord as described (33,44,45). The Triton X-100 insoluble pellet from repeated sucrose cushions was resuspended in 20 mM sodium phosphate, pH 7.0, 8.0 M urea, and 0.1% ␤-mercaptoethanol, filtered through a 0.20-m filter, and loaded onto a 45-ml DE52 cellulose column. NF-H was separated from the remaining of cytoskeletal proteins including NF-M and NF-L in a 250-ml sodium phosphate gradient from 20 to 500 mM of above buffer. NF-M and NF-L were further purified to homogeneity as described (33).
GT'ase Labeling of NF Proteins with UDP-[ 3 H]Galactose-Total NFs (25 g) or 5 g of each purified NF subunit (NF-H, NF-M, or NF-L) was labeled with UDP-[ 3 H]galactose and GT'ase on ice as described (36) with some modifications according to Dong et al. (33). The labeled mix was precipitated with 8 volumes of cold acetone. The pellet was resuspended in gel sample buffer and analyzed by SDS-PAGE.
Stoichiometry estimation of O-GlcNAc on NF-H was determined as described previously (33). Briefly, purified NF-H was acid-hydrolyzed in 6.0 M HCl for 5 h under nitrogen gas at 110°C (for glucosamine analysis) or for 24 h at 110°C (for amino acid composition analysis). The exact amount of protein was quantified by amino acid composition analysis. The amount of glucosamine was determined by the ratio of the peak area of glucosamine between the standard run and acid hydrolyzed NF-H run by high pH anion exchange chromatography-pulsed amperometric detection (HPAEC-PAD) on a CarboPAc PA1 column using 1 nmol of 2-deoxyglucose as an internal standard.
Preparation of Chymotryptic Fragments of NF-H Head Domain-Purified NF-H was first digested with chymotrypsin (weight ratio, 200:1) in 0.1% ammonium bicarbonate, pH 8.25, 10% CH 3 CN, for 2 h at room temperature. The sample was centrifuged in a Microcon-30 (Amicon), and the flow through containing small chymotryptic fragments was lyophilized.
Trypsin Digestion of NF-H-Purified NF-H (0.48 mg) was first digested with trypsin (weight ratio, 20:1) in 100 mM Tris, pH 8.5, and 10% CH 3 CN at 37°C for 18 h, and the digestion was continued for another 10 h with an addition of the same amount of trypsin. The reaction was stopped by adding 2 l of aprotinin (2 mg/ml). The trypsinized peptides were desalted through two Sep-Pak C18 cartridges (Waters), eluted with 60% CH 3 CN, and dried.
GT'ase Labeling of NF-H Peptides and Isolation of 3 H-Labeled Glycopeptides-Peptides from chymotrypsin digestion of NF-H were labeled with UDP-[ 3 H]galactose and 1.5-fold of nonradioactive UDP-galactose (27.5 M final) in GT'ase buffer in the presence of aprotinin at 37°C for 2 h. The reaction mixture was acidified with 1% trifluoroacetic acid and loaded onto a Sep-Pak cartridge. The labeled peptides were eluted with 60% CH 3 CN, dried, and then separated by reverse phase high performance liquid chromatography (RP-HPLC) using a C18 column (0.46 ϫ 25 cm, Rainin) in the following gradient of 0 -60% CH 3 CN with 0.1% trifluoroacetic acid: 0 -60 min, 0 -24%; 60 -70 min, 24 -60%; 70 -85 min, 60% (all RP-HPLC gradients described hereafter were linear in the specified time window, the flow rate was 1 ml/min, fractions were collected every minute, and the absorbance of the elutant was monitored at 214 nm).
Peptides from trypsin digestion of NF-H were labeled with UDP-[ 3 H]galactose followed by 20-fold excess of unlabeled UDP-galactose.

Isolation of Tryptic [ 3 H]Galactose-labeled Peptides of both NF-M and NF-L-Tryptic
H]galactose-labeled peptides of both NF-M and NF-L were prepared according to Dong et al. (33). For both subunits, there were four major radioactive peaks (named as Peaks 1, 2, 3, and 4 for NF-M, and Peaks A, B, C, and D for NF-L, in the order of elution time) on the first dimension RP-HPLC, and each radioactive peptide was separated again on the second dimension RP-HPLC (33). The first two radioactive peptide peaks of NF-M on the first dimension RP-HPLC, Peaks 1 and 2, were purified by the second dimension RP-HPLC, and both eluted as a single radioactive peak. The fractions of [ 3 H]galactoselabeled glycopeptides of Peak 2 were combined, dried, and resuspended in 30% CH 3 CN. The radioactive peptide peak of Peak 1 containing multiple peptides from two runs of RP-HPLC was further purified on a third dimension RP-HPLC in 0.1% trifluoroacetic acid, 0 -60% CH 3 CN gradient: 0 -60 min, 0 -24%; 60 -70 min, 24 -60%; 70 -85 min, 60%. The tritium-labeled glycopeptide was eluted at 24 and 25 min, and both fractions were dried.

Gas-Phase Sequencing and Manual Edman Degradation Sequencing of [ 3 H]Galactose-labeled
Glycopeptides-Following the first, second, or third dimension RP-HPLC, purified glycopeptides of NF proteins were sequenced by automated Edman degradation in a model 470A gasphase sequencer (Applied Biosystems, Inc.). MED sequencing of glycopeptides was performed by covalent coupling of peptides to Sequelon-AA membranes (MilliGen/Biosearch of Millipore), followed by repeated coupling of phenyl isothiocyanate and trifluoroacetic acid extraction as described (47), except that the membrane (or filter) from each cycle was extracted and washed with 0.5 ml of trifluoroacetic acid and then washed with methanol twice, and all extraction and washes were combined, dried, and neutralized with 100 l of 1.0 M Tris, pH 8.8, before scintillation counting.
SDS-PAGE and Autofluorography-NF proteins were analyzed by 8.0% SDS-PAGE (48). Gels were either stained with Coomassie Brilliant Blue or treated with En 3 HANCE (DuPont NEN), dried, and fluorographed.
Protein Assay-Protein concentration was determined by Amido Schwarz dye binding procedure (49) using bovine serum albumin as a standard.

NF Triplet Proteins Are [ 3 H]Galactose Labeled by GT'ase-A
common and sensitive method to detect O-GlcNAcylation of proteins is to covalently label the protein-bound sugar with UDP-[ 3 H]galactose and GT'ase (36,46). Total NFs, comprised primarily of the three NF subunits, NF-H (200 kDa), NF-M (145 kDa), and NF-L (62 kDa), were isolated from rat spinal cord (Fig. 1A, lane 4). Each protein subunit was purified to near homogeneity as described above (Fig. 1A, lanes 1, 2, and 3). Total NFs and each purified subunit were separately labeled with UDP-[ 3 H]galactose and GT'ase, analyzed by SDS-PAGE, and fluorographed. As shown in Fig. 1B, NF-M and NF-L were strongly radiolabeled by [ 3 H]galactose both in the total cytoskeletal pellet (Fig. 1B, lane 4) and in the purified form (Fig.  1B, lanes 1 and 2). Purified NF-H and the tail domain of NF-H could only be labeled with GT'ase if they were first denatured by treating with 0.5% SDS and boiling prior to an extended incubation with GT'ase. Even so, NF-H was only weakly la-beled in the purified form (Fig. 1B, lane 3), and the lengthy incubation resulted in several labeled bands around 50 kDa that are degradation products of NF-H (see below). The carboxyl-terminal tail domain of NF-H, consisting of 52 KSP repeats, was labeled even more weakly (data not shown). The relatively weak [ 3 H]galactose incorporation into the purified NF-H is due, at least in part, to the tendency of purified NF-H to selfaggregate during GT'ase labeling. The labeling of NF-H was improved in the total NFs fraction (Fig. 1B, lane 4). Relative to NF-M and NF-L, purified NF-H can be labeled more efficiently at 4°C or on ice than at room temperature or at 37°C because the aggregation of NF-H in the GT'ase lableling reaction is very severe at 37°C. GT'ase labeling of NF-H was also somewhat enhanced by dephosphorylation prior to labeling (data not shown), also suggesting that the phosphates may block accessibility of GT'ase. That the addition of [ 3 H]galactose to each NF subunit represented O-GlcNAcylation was demonstrated by the resistance to peptide/N-glycosidase F treatment and virtually 100% accessibility to mild alkali-induced ␤-elimination (see below).
NF-H Is Modified with Unelongated Monosaccharides, O-GlcNAc-In order to identify the sugar residues modified on NF-H, purified NF-H was labeled with GT'ase, subjected to alkaline-induced ␤-elimination, and analyzed on a Sephadex G-50 gel filtration column (Fig. 2). The ␤-elimination products chromatographed as a single Vi peak on the column (Fig. 2A). The V i peak (fractions 28 -31) was run on a high resolution gel filtration TSK Fractogel column. The labeled sugar moieties migrated as a single peak between standards of one and two GlcNAc residues, with K av of 0.845 (Fig. 2B). The radioactive peak from TSK chromatography was further analyzed on a Dionex CarboPAc-MA1 column with HPAEC-PAD. The [ 3 H]galactose-labeled sugars from NF-H exactly co-migrated with authentic Gal␤1-4GlcNAcitol (Fig. 2C), thus demonstrating that single GlcNAc moieties were O-glycosidically linked to NF-H. Similar results have previously been shown for NF-M and NF-L (33).
To determine the stoichiometry of GlcNAc on purified NF proteins, GlcNAc was released from a known amount of polypeptides by 6 M HCl acid hydrolysis (converting it to glucosamine) and analyzed on Dionex CarboPAc PA1 column with HPAEC-PAD (33). The average stoichiometry of purified NF-H from rat spinal cord was determined by this method to be 0.3 mol of GlcNAc/mol of NF-H, a molar ratio higher than the 0.15 and 0.1 mol of O-GlcNAc previously found for NF-M and NF-L, respectively (33). This establishes a minimum estimate because O-GlcNAc may be removed during NF purification (50 mM GlcNAc was always included in the homogenization buffer, but this only partially inhibits both cytosolic N-acetyl-␤-D-glucosaminidase (38) and lysosomal hexosaminidases (data not shown)). NF-L (33), repeatedly failed to identify the site(s) of O-Glc-NAcylation on this peptide, primarily because the peptide did not efficiently attach covalently to the membrane. The reason(s) for this remains unclear.

Identification of Thr 53 , Ser 54 , and Ser 56 of the Head Domain and Multiple Sites in the KSP Repeats of the Tail Domain as Major O-GlcNAc Attachment Sites of NF-H-In order to examine the natural sites of O-GlcNAcylation of NF-H protein, pu-
rified NF-H was first digested with trypsin and then labeled with UDP-[ 3 H]galactose by GT'ase as performed previously on both NF-M and NF-L (33). Radiolabeled peptides were separated by RP-HPLC on a C18 column as described under "Experimental Procedures." As seen in Fig. 4, a minimum of eight [ 3 H]galactose-labeled glycopeptide peaks were resolved. Radioactivity essentially covered the entire elution profile from 17 to 120 min. In a separate experiment, when trypsinized and GT'ase-labeled NF-H peptides were redigested with 10:1 ratio of protein to trypsin and chromatographed on the C18 column, a similar number of peaks were still observed, a finding suggesting that the multiple peaks do not represent incomplete digestion products. Sequential trypsin digestion followed by UDP-[ 3 H]galactose and GT'ase labeling of dephosphorylated NF-H also generated about a dozen tritium peaks (data not shown), further confirming the multiple O-GlcNAcylation of NF-H. By comparing the total [ 3 H]galactose incorporation of the head domain (peptide 51 ARTSVSSVSASPSRT 65 ) with that of the tryptic carboxyl domain and the amount of radioactivity in head domain (peak 4, Fig. 4, see below) with the rest of peaks (i.e., that in KSP repeats), we estimate that the ratio of O-GlcNAc on the tail domain versus the head domain is more than 10 to 1.
Peaks 1-8 were individually further chromatographed on a second dimension C18 column with a different solvent system as described. Each then generated one or more radioactive peaks (Fig. 5). Several major radioactive glycopeptide peaks from the second dimension RP-HPLC (Fig. 5) were subjected to both gas-phase sequencing and MED sequencing, and the data are shown in Fig. 6. All but Peak 4 yielded a radioactive peak eluted at cycle number 1 from MED sequencing, indicating that the first amino acid residue of these tryptic glycopeptides is either serine or threonine. Because trypsin cleaves proteins at the carboxyl side of lysine and arginine, each glycopeptide has a (K/R)(S/T) motif. Gas-phase sequencing of Peaks 3a and 3b contained a mixture of peptides, with major sequence of 114 QLEAHNT 120 and 201 FAQEA 205 both from the rod domain of NF-H (20), respectively, but also revealed minor sequences of SPA(T/S)VK and SPVTVK, which are parts of the KSP repeats. MED sequencing showed the released radioactive peak is at cycle number 1 for Peak 3b and 3c (Fig. 6) and failed to reveal the peak number for Peak 3a, suggesting that the first serines on SPA(T/S)VK and SPVTVK peptides are the ones modified by O-GlcNAc. No gas-phase sequence could be obtained for the Peaks 2a, 2b, 5, 6a, and 7a, but they all showed radioactive release at cycle number 1, indicating that the tritium is released from serine residues of other KSP repeats. Therefore, some or all serines on the KSP repeats of the tail domain of NF-H are modified with O-GlcNAc.
The peptide sequence from gas-phase sequencing of Peak 4a  (Fig. 6). MED sequencing of Peak 4b also revealed radioactivity released at cycle number 4, and two minor peaks at cycles 1 and 2 (Fig. 6), corresponding to Ser 56 , Thr 53 , and Ser 54 of the peptide 53 TSVSSVSASPSR 64 . Peak 4c released a strong single tritium peak at cycle number 4 (Fig. 6), matching to Ser 56 of the peptide. No amino acid sequence could be obtained for both peptides of Peak 4b and 4c; however, based on the very close elution profile on the first dimension RP-HPLC, a strong and single [ 3 H]galactose labeling peak of chymotryptic peptide 51 ARTSVSSVSASPSRF 64 and good match between MED sequencing data and the peptide sequence, it thus suggests that Thr 53 , Ser 54 48 and Thr 431 of NF-M, respectively. The first two tritium peaks, Peak 1 and 2, both eluted as a single radioactive peak on the second dimension. The radioactive peak of Peak 1, containing multiple peptides, was further purified on a third dimension by RP-HPLC. MED sequencing was performed to determine the sites of O-GlcNAcylation on both Peak 1 and 2. For Peak 1, radioactivity was released at the third cycle, matching Thr 19 of the tryptic peptide, 17 VPTETR 22 (Fig. 7A) (51). For Peak 2 radioactivity release peaked at cycle 7, corresponding to Ser 34 of the tryptic peptide, 28 VSGSPSSGFR 37 (Fig. 7B) (51). Although repeated efforts with gas phase sequencing failed to reveal the exact peptide sequence of the two peptides, Thr 19 (33). Peak A was further purified by a second and third dimension RP-HPLC according to "Experimental Procedures" and then subjected for gas-phase sequencing. The amino acid sequence of the purified peptide is 31 SGYSTAR 37 (52) (Fig. 8). The site of O-GlcNAc attachment was determined to be Ser 34 of NF-L by measurement of tritium counts released from each  Fig. 4 were further purified on the second dimension RP-HPLC according to the "Experimental Procedures." Tritium profiles from each 1-min fraction are shown. Each tritium peak is numbered followed by letter a, b, or c in the order of peak elution.  2a, 2b, 3b, 3c, 4a, 4b, 4c, 5, 6a, and 7a was performed as described under "Experimental Procedures." Tritium counts released from each cycle and peptide sequence of Peaks 4a, 4b, and 4c are shown. cycle of MED sequencing (Fig. 8). The amino acid sequence of Peak D, the fourth and major radioactive peak of NF-L, was revealed to be 38 SAYSSYSAPVSSSLSV 53 , but the exact O-GlcNAc site(s) was not determined previously (33), partly because of the length of the peptide and limitation of MED sequencing. In order to precisely determine the site(s) of modification, the [ 3 H]galactose-labeled peptide was digested with chymotrypsin and separated on RP-HPLC as described under "Experimental Procedures." The radioactive peak of chymotryptic 3 H-labeled peptide was shifted as expected to 45 min (Fig. 9B) from the 60 -61 min of the parental peptide (Fig. 9A), indicating removal by chymotrypsin of two tyrosine residues near the amino terminus of the peptide. Both peptides were subjected to MED sequencing, and radioactivity from each cycle was measured. As illustrated in Fig. 9, released radioactivity from the undigested peptide peaks rather weakly at cycle number 11 (Fig. 9C) and at cycle number 5 from the digested peptide (Fig. 9D), demonstrating Ser 48 , corresponding to the first serine on the carboxyl side of proline 46 , is modified by O-GlcNAc. Therefore, Thr 21 , Ser 27 , Ser 34 , and Ser 48 of head domain of rat NF-L are O-GlcNAcylated based on gas-phase sequencing, MED sequencing, and mass spectrometry. DISCUSSION Combined with our earlier effort (33), we have now shown that NF triplet proteins, NF-H, NF-M, and NF-L, are modified by O-GlcNAc in vivo to stoichiometries of at least 0.3, 0.15, and 0.1 mol GlcNAc/mol of protein, respectively. We have demonstrated that NF-H is modified by O-GlcNAc at three sites of the head domain, Thr 53 , Ser 54 , and Ser 56 , and at multiple serine residues in the KSP repeats of the carboxyl tail domain, although we would caution that the exact number of KSP repeats modified by O-GlcNAc is hard to determine (Fig. 10). In addition, four sites both in NF-M and NF-L have been identified (Fig. 10). As shown in Fig. 10B, a stretch of less than 40 amino acids of the head domain of NF-M has at least three O-GlcNAc sites and eight phosphorylation sites identified by in vitro labeling with protein kinase A or C (29), or by in vivo labeling (27). Intriguingly, the known O-GlcNAc and phosphorylation sites in the head domain are in close proximity but not overlapping. Given the proximity of the O-GlcNAc sites to the phosphorylation sites in the head domain and importance of the head domain in filament assembly (30 -32), it is plausible that modification by O-GlcNAc may also play a role in regulating the timing or extent of filament assembly. This regulation could be exerted by either direct or indirect (e.g. by affecting phosphorylation) influence on the structure of the head domain, which in turn modulates assembly of NFs.
The co-existence of both phosphate and O-GlcNAc on the same NF subunits adds to the growing list proteins that exhibit such a pattern. Nearly one hundred intracellular phosphoproteins are so far also known to bear O-GlcNAc, and more than thirty in vivo O-GlcNAc attachment sites have been mapped (33,39,53). Among all mapped O-GlcNAc sites, including all sites identified here from NF triplet proteins, most have a proline nearby, most commonly on the amino-terminal side of the modified serine or threonine residue. This has revealed a consensus of PX 0 -4 (S/T) motif (X is usually a hydrophobic residue) similar to those used by proline-directed and growth factor kinases (39). In addition, in many cases, like the identified sites on NFs, there is also a valine or glutamine residue close to the O-GlcNAc site. Therefore, although a precise consensus sequence cannot be drawn from all available sites, there is a common motif for O-GlcNAcylation: PVQX 0 -4 (S/T).
Although the stoichiometry of O-GlcNAcylation on isolated NFs is low compared with phosphorylation, it seems likely that only a subset of NF subunits is modified by O-GlcNAc, and the stoichiometry in those subunits is probably much higher. Almost all (Ͼ99%) of the assembled NFs examined here represent NFs in myelinated internodal regions known to be nearly stoichiometrically phosphorylated in the KSP repeat domain (25). In contrast, the KSP repeats of subunits in cell bodies, dendrites, and nodes of Ranvier are unphosphorylated, but account for ϳ1% of NFs in the neuron as judged by NF density and the relative volumes of the respective compartments (23)(24)(25). Thus, the NF proteins isolated here from rat spinal cord are almost certainly a mixed population of highly phosphorylated, partially phosphorylated, partially O-GlcNAcylated, or fully O-GlcNAcylated proteins. Thus, it seems plausible that for those subunits modified by O-GlcNAc, the true stoichiometry may be much higher (estimated by the measured stoichiometry divided by the fraction of protein from the unmyelinated domains). Further, the exact level of O-GlcNAcylation may be even higher in view of the dynamic nature of the modification and the presence of strong activities of both cytosolic N-acetyl-␤-D-glucosaminidase (38) and lysosomal hexosaminidases (33,38,54) that may act during the purification process. NFs are dynamic structures (55,56), the dynamic process involves the continuous turnover of NFs and lateral subunit exchange and involving incorporation of new subunits (55). Because NFs are also known to be highly phosphorylated post assembly in axons and less phosphorylated in cell bodies and proximal axons (23,25), it is possible that O-GlcNAc is added to NF proteins by O-GlcNAc transferase (37) right after their synthesis and prior to segmental assembly into filaments and transport into the distal part of axons. Subsequently, a N-acetyl-␤-D-glucosaminidase (38) removes the sugars, and phosphates are added by protein kinases (5,56). Phosphorylation of the carboxyl domains of NF-H and NF-M has been proposed to modulate interfilament spacing (21-23) by interacting with adjacent filaments and other organelles, thereby establishing a wider filament spacing, perhaps through repulsion between the adjacent filaments as a result of the highly negatively charged phosphates (17,(22)(23)(24). By replacing phosphates with O-GlcNAc, the interactions between filaments may switch from a repulsive one to an associative one, leading to close packing of filaments, for example, in nodes of Ranvier. Therefore, it seems likely that organization of NFs is regulated by kinase/phosphatase (5,56) and O-GlcNAc transferase/Nacetyl-␤-D-glucosaminidase (37,38). The dynamic O-GlcNAcylation (39,40) and phosphorylation (56) could therefore modulate proper assembly of NFs and maintain the dynamic nature of filaments, and abnormalities in either could contribute to any of the many motor neuron diseases in which NFs accumulate aberrantly (5,57).