Mitotic arrest-associated enhancement of O-linked glycosylation and phosphorylation of human keratins 8 and 18.

Arrest of the human colonic cell line HT29 at the G2/M phase of the cell cycle resulted in changes in keratin assembly that were coupled with a significant increase in the O-linked glycosylation and serine phosphorylation of keratin polypeptides 8 and 18 (K8/18). With mitotic arrest, enhanced keratin phosphorylation occurred preferentially on K8, whereas K18 showed a higher glycosylation level than K8. Removal of the arresting agent allowed cells to proceed through the cell cycle with a concomitant decrease in K8/18 glycosylation. In contrast, keratins isolated from S phase-enriched cells, obtained after synchronization with aphidicolin, did not show enhanced glycosylation. Tryptic peptide analysis of keratins in G2/M-arrested cells showed changes in the glycopeptide pattern of K8 and in the phosphopeptide patterns of K8 and K18. Labeling of K8/18 immunoprecipitates, isolated from G2/M-arrested cells, with [3H]galactose followed by beta-elimination showed that K8/18 glycosylation consisted of single N-acetylglucosamine residues. Threonine was identified as the site of glycosylation after comparing acid hydrolysis products of beta-eliminated and non-beta-eliminated K8 and K18. Specific cleavage at tryptophan residues indicated that K18 glycosylation and phosphorylation were restricted to the head and proximal rod domains, whereas K8 did not show the same restriction. Our results show a unique association of the single O-linked N-acetylglucosamine type of modification of keratins with mitotic arrest in HT29 cells. There was no reciprocal relationship between K8/18 glycosylation and phosphorylation, and each keratin showed a preferential G2/M cell cycle-associated increase in either serine phosphorylation or threonine glycosylation.

Arrest of the human colonic cell line HT29 at the G2/ M phase of the cell cycle resulted in changes in keratin assembly that were coupled with a significant increase in the 0-linked glycosylation and serine phosphorylation of keratin polypeptides 8 and 18 (K8/18). With mitotic arrest, enhanced keratin phosphorylation occurred preferentially on K8, whereas K 1 8 showed a higher glycosylation level than K8. Removal of the arresting agent allowed cells to proceed through the cell cycle with a concomitant decrease in K8/18 glycosylation. In contrast, keratins isolated from S phaseenriched cells, obtained after synchronization with aphidicolin, did not show enhanced glycosylation. Tryptic peptide analysis of keratins in G2/M-arrested cells showed changes in the glycopeptide pattern of K8 and in the phosphopeptide patterns of K 8 and K18. Labeling of K8/18 immunoprecipitates, isolated from G2/M-arrested cells, with ['H]galactose followed by @elimination showed that K8/18 glycosylation consisted of single N-acetylglucosamine residues. Threonine was identified as the site of glycosylation after comparing acid hydrolysis products of @-eliminated and n0n-Beliminated K 8 and K18. Specific cleavage at tryptophan residues indicated that K 1 8 glycosylation and phosphorylation were restricted to the head and proximal rod domains, whereas K 8 did not show the same restriction. Our results show a unique association of the single 0-linked N-acetylglucosamine type of modification of keratins with mitotic arrest in HT29 cells. There was no reciprocal relationship between K8/18 glycosylation and phosphorylation, and each keratin showed a preferential G2/M cell cycle-associated increase in either serine phosphorylation or threonine glycosylation.
Keratins are a subgroup of cytoskeletal intermediate filament (IF)' proteins that have a characteristic expression on most epithelial tissues (Moll et al., 1982(Moll et al., , 1990Quaroni et al., 1991). Keratins are the most complex of IF proteins with 21 * This work was supported by a Veterans Administration Career Development and Merit Award and the PEW Scholars Program (to M. B. 0.) and Digestive Disease Center Grant DK 38707. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisernent" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Several post-translational modifications have been described for keratins including phosphorylation (Gilmartin et al., 1984;Steinert, 1988;Omary, 1991), glycosylation (King andHounsell, 1989;Chou et al., 1992), and acetylation (Steinert and Idler, 1975). Recently, we showed that K8/18 have multiple 0-linked glycosylation sites that consist of single N-acetylglucosamine (GlcNAc) residues . This type of glycosylation appears to be characteristic of several cytoplasmic and nuclear proteins (Torres and Hart, 1984;Hart et al., 1989a;Davis and Blobel, 1987;Haltiwanger et al., 1992). K8/18 glycosylation was also found to be a dynamic process with the biosynthesis and degradation rates faster than the corresponding rates for the protein backbone, suggestive of a functional significance for this modification . The amino acid(s) involved in the single 0-GlcNAc modifications has been described for only a few proteins including the nuclear pore complex glycoproteins (Holt et aL., 1987b), a-crystallins (Roquemore et al., 1992), and the serum response transcription factor (Reason et al., 1992). In all cases, serine appears to be the predominant involved 0-linked amino acid. Since phosphorylation and 0linked glycosylation can theoretically involve identical Ser/ Thr residues, it has been tempting to hypothesize that Ser/ Thr 0-linked glycosylation may function, at least in some settings, as a functional regulator of protein phosphorylation (Hart et al., 1989b).
Keratin phosphorylation occurs primarily on serine residues (Steinert, 1988;Chou and Omary, 1991) with evidence for the involvement of several kinases (Yano et al., 1991), including CAMP-dependent protein kinase (Gilmartin et al., 1984) and protein kinase C . The function of keratin phosphorylation is unknown; however, there is evidence that hyperphosphorylation plays a role in the disassembly of IF proteins (Inagaki et al., 1987(Inagaki et al., , 1988Gonda et aL., 1990;Chou et al., 1989) and may also play a role in signal transduction as shown after epidermal growth factor binding to rat hepatocytes (Baribault et al., 1989). Increased phosphorylation was also noted in keratins obtained from mitotic amnion cells as compared with keratins obtained from interphase cells (Celis et al., 1983).
In this study, we initially confirmed serine hyperphosphorylation of K8/18 in the human colonic epithelial cell line HT29 after colcemid-induced arrest in the G2/M phase of the cell cycle. We then examined keratin glycosylation during 4465 different stages of the cell cycle and showed that it increased in G,/M colcemid-arrested cells but not in aphidicolin-synchronized S phase cells. The glycosylation during G,/M was similar to that found in asynchronously growing cells in that it consisted of single 0-GlcNAc residues, with the site of glycosylation identified as threonine. Increased glycosylation and phosphorylation of keratins occurred in parallel. However, K8 showed a preferential increase in phosphorylation, whereas K18 showed a higher level of glycosylation during G,/M arrest. In addition, the head and proximal rod domain of K18 were the primary regions of glycosylation and phosphorylation.

MATERIALS AND METHODS
Cell Culture and Synchronization-HT29 cells (American Type Culture Collection, Rockville, MD) growing at 30-60% confluence were plated in 35-mm tissue culture dishes (2 X IO6 cells in 1 ml of medium). Colcemid or nocodazole (Sigma) was used at 0.5 pg/ml. Adherent cells were removed by treatment for 10 min with phosphatebuffered saline (PBS) containing 0.5% trypsin and 1 mM EDTA for the cell cycle analysis or by scraping for the biochemical studies. Aphidicolin (Sigma, 5 pg/ml) was used to arrest cells at the G,/S boundary.
To obtain floater cells from log phase growing cells, confluent HT29 cells were split 1:4. After allowing cells to settle for 6 h, nonadherent cells were discarded followed by rinsing. Fresh prewarmed medium (37 "C) was added followed by 10% COa incubation for 1 h and then shaking in a 37 "C room for 1 h. Floater cells were then collected, placed over ice immediately (viability greater than 95%), and then used for immunoprecipitation or cell cycle analysis.
Cell Cycle Analysis and Immunofluorescence-Cells were fixed with 70% ethanol for at least 30 min and then treated with propidium iodide (20 pg/ml) containing RNase A (40 pg/ml) in PBS for 30 min, followed by cell cycle analysis using the FACScan (RFIT program, Becton Dickinson). The homogeneity of individual cell cycle populations was estimated by calculating the standard deviation (i.e. dispersion from the mean) and coefficient of variation (i.e. standard deviation divided by the mean) using the RFIT program. For indirect immunofluorescence, cells were transferred to collagen-coated slides using a Cytospin centrifuge (7,000 rpm, 6 min), fixed in acetone (-20 "C, 2 s), washed then stained with anti-K8/18 monoclonal antibody L2A1 in PBS (pH 7.4) containing 10% human serum and 0.1% sodium azide. After 45 min (22 "C), cells were washed and then incubated with Texas Red-conjugated goat anti-mouse antibody. Photographs were taken using black and white Kodak p3200 film.
Cell Labeling and Immunoprecipitation-HT29 cells were labeled with orthophosphate (125 pCi/ml, 2 h, 5 X IO6 cells/ml) in phosphatefree RPMI 1640 medium supplemented with 5% dialyzed fetal calf serum. Cells were then washed with PBS and solubilized (30 min) with 1% Nonidet P-40 in PBS containing 25 pg/ml aprotinin, 10 pM leupeptin, 10 p~ pepstatin, 0.1 mM phenylmethylsulfonyl fluoride, 5 mM sodium pyrophosphate, and 50 mM sodium fluoride. After spinning to remove nonsolubilized material, immunoprecipitation was done using L2Al-Sepharose (anti-Kd/lS-specific monoclonal antibody coupled to Sepharose ). Relative levels of K8/ 18 phosphorylation and glycosylation were determined using densitometric scanning of the corresponding bands on the radiograph (LKB Ultrascan XL enhanced laser densitometer). Galactosylution and Reductive P-Elimination-Galactosylation of K8/18 immunoprecipitates using UDP-[3H]galactose and galactosyltransferase was carried out to completion for 2 h as described . For /3-elimination, K8 and K18 were separated using preparative sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) followed by electroelution of the Coomassie-stained hands (Amicon electroelution apparatus, Davers, MA) using the manufacturer's recommendations. Eluted samples were concentrated and then precipitated with 10 volumes of acetone (-20 "C). Precipi-tates were then treated with 0.3 ml of 1 M NaBH4 in 0.1 M NaOH (37 "c, 18 h), after which NasPdClp (20 mM) was added for 1 additional h. Samples were then gradually neutralized with glacial acetic acid (18 pl) followed by acetone precipitation of the remainingprotein. Amino acid composition of the &eliminated and duplicate non-/3eliminated material was done using a Beckman 6300 amino acid analyzer. To determine the carbohydrate composition of the 3Hgalactosylated K8/18, immunoprecipitates were processed for 0-elimination followed by analysis of the released carbohydrates exactly as described .
Phosphoamino Acid Analysis, Peptide Mapping, ana' Isoelectric Focusing-Phosphorylated amino acids were analyzed by two-dimensional electrophoresis after extraction of the phosphate-labeled K8 and K18 protein bands from the SDS-polyacrylamide gels and hydrolysis in vacw with constant boiling HCl (Cooper et al., 1983). Tryptic peptide mapping of in vitro 3H-galactosylated and in vivo metabolically 32P04-labeled K8 and K18 was carried out as described . Isoelectric focusing was done using a 4:l mix of Ampholine (pH 5-7) and Ampholine (pH 3-10) (O'Farrell, 1975) in a Bio-Rad minigel system.
Dimethyl Sulfoxide-HBr Cleavage of 3H-Galuctosylated or 32P-Phosphorylated K8 ana' K18 at Tryptophunyl Peptide Bonds-Cleavage at tryptophan residues was carried out similar to what has been described (Savige and Fontana, 1977). Immunoprecipitates of K8/18 were obtained from nonlabeled colcemid-arrested HT29 cells or from cells labeled with 32P04 in vivo. Nonlabeled K8/18 immunoprecipitates were galactosylated using UDP-[3H]galactose. K8 or K18 was individually isolated using preparative SDS-PAGE as described above. K8 or K18 (10-20 pg) was then incubated with a mixture containing 24 p1 of acetic acid, 12 pl of HCl (12 M), 1 pl of dimethyl sulfoxide, and 1 pl of phenol. After 30 min (22 "C), 4 p1 of HBr and 1 pl of dimethyl sulfoxide were added (30 min, 22 "C) followed by the addition of 0.5 ml of H20. After lyophilization, samples were analyzed using SDS-PAGE. Amino acid composition of individual fragments was determined by transferring to polyvinylidene membranes followed by analysis using a Beckman 6300 analyzer.

Colcemid Arrest of HT29 Cells Results in Increased K8/18
Glycosylation and Phosphorylation-We used the colchicine analog colcemid, which arrests cells in the Gz/M phase of the cell cycle (Ashihara and Baserga, 1979), to study the effect of cell arrest on the 0-linked glycosylation and phosphorylation of K8/18 in the colonic epithelial cell line HT29. Incubation of HT29 cells growing in log phase with colcemid resulted in a progressive increase in the percentage of GZ/M-arrested cells with increasing time of colcemid incubation (not shown). As shown in Fig. lB, colcemid treatment of HT29 cells for 36 h arrested more than 90% of the cells in the Gz/M phase of the cell cycle. Coupled with this arrest, the fine evenly distributed network of keratin filaments seen when cells were predominately in Go/G1 (Fig. 1C) changed to disassembled aggregates of cytoplasmic and perinuclear dots (Fig. 1 0 ) . A similar mitosis-associated reorganization of keratins has been observed in a variety of cell types (Lane et al., 1982;Franke et al., 1982).
We examined the phosphorylation of K8/18, after various treatment intervals with colcemid, by metabolic labeling of cells with 32P04 for 2 h. K8/18 glycosylation was examined by galactosylating K8/18 immunoprecipitates using UDP- [3H]galactose and galactosyltransferase, which donates galactose to terminal 0-linked GlcNAc residues. This galactosylation technique detects only accessible GlcNAcs and may therefore underestimate the total glycosylation. For unclear reasons, labeling of intact cells with [3H]GlcNH~ interfered with the ability of colcemid to induce G2/M cell arrest (not shown). As shown in Fig. 2, there was a gradual increase in K8/18 glycosylation and phosphorylation with increasing colcemid incubation times, as the percent of Gz/M cells increased. After 36 h of colcemid treatment, there was a 10-fold (K8) and 5-fold (K18) increase in keratin phosphorylation. The faint Coomassie-stained band migrating slightly slower than K8 represents a hyperphosphorylated form of K8 (

FIG. 2. Phosphorylation and glycosylation of KS/18 in col-
cemid-treated cells. Coomassie-stained gels of the R2P0,-labeled or 3H-galactosylated K8/18 immunoprecipitates and their respective radiographs are shown. HT29 cells were seeded to grow asynchronously or in the presence of colcemid (0.5 pg/ml) for the indicated time points and then labeled with "PO4 (2 h) in phosphate-free medium. Immunoprecipitates of K8/18 were prepared from ""PO4labeled cells or from unlabeled cells using monoclonal antibody L2A1 coupled to Sepharose beads. To examine the extent of glycosylation, unlabeled immunoprecipitates were incubated with 25 milliunits of galactosyltransferase, 0.6 pCi of UDP-[SH]galactose in 25 pl of 20 mM MnCI2, 100 mM sodium cacodylate (pH 6.5) for 2 h. After washing, immunoprecipitates were analyzed using 10% SDS-polyacrylamide gels. The percent of HT29 cells in G2/M was 5, 15, 50, 89, and 92 after 0,8, 16, 24, and 36 h of treatment with colcemid, respectively.
2, 24-and 36-h time points). The band that is intermediate between K8 and K18 is seen to a variable extent with different experiments and likely represents a degradation product of K8 (Chou and Omary, 1991). In contrast with K8, which showed a more dramatic increase in phosphorylation than K18, a higher level of overall K18 glycosylation (determined by galactosylation) was noted as compared with K8 (Fig. 2). The relative increase for both K8 and K18 glycosylation accompanying G2/M arrest was 7-fold (Fig. 2), as determined by densitometric scanning.
We asked if the increased glycosylation of K8/18 associated with Gz/M arrest was reversible after removal of the arresting agent. Colcemid-arrested cells, however, could not be readily unblocked from their arrest even after incubating for more than 72 h in the absence of colcemid (not shown). We instead used nocodazole (Zieve et al., 1980), which arrested cells in a manner similar to colcemid, and resulted in a similar increase in K8/18 glycosylation after arrest (Fig. 3, lane b). Removal of nocodazole slowly returned K8/18 glycosylation to baseline levels commensurate with a decrease in the percent of cells in Gz/M (Fig. 3, lane c). Of note, the increase in K8/18 glycosylation associated with colcemid/nocodazole G2/M arrest does not occur in all cell lines tested. For example, HeLa (human cervix) and PtKl (marsupial kidney) do not show increased K8/18 glycosylation, whereas SK-CO-1 (human colon) does (not shown).  -Noy et al., 1980). After washing off the aphidicolin and incubating with fresh medium, a substantial population of S phase cells can be obtained, which, on further incubation, proceed into G2/M and then Go/GI (Fig. 4A). No difference in glycosylation was noted between the S-enriched cells, cells blocked a t Gl/S, or cells growing asynchronously ( i e . primarily Go/Gl). Furthermore, G2/M cells obtained using the aphidicolin synchronization method also showed base-line Go/Gllike glycosylation (Fig. 4.4). Multiple time points taken at 0.5h intervals on both ends of the G2/M peak also showed baseline glycosylation (not shown). A near base-line level of K8/18 glycosylation (not shown) was also noted in GJM-enriched floater cells obtained after mechanical agitation of log phase growing HT29 cells without any drug treatment (Fig. 4B, histogram c). The lack of increased K8/18 glycosylation in Gz/M-enriched cells obtained from floater cells or from aphidicolin-synchronized cells suggests that the increased glycosylation with colcemid/nocodazole mitotic arrest may be secondary to a microtubule disassembly-related effect rather than a mitosis-associated phenomenon. However, this remains unclear since nocodazole/ colcemid arrest provides a much sharper G2/M peak (i.e. more homogeneous) than observed in the G2/M peak obtained after aphidicolin synchronization or from floater cells without drugs (Fig. 4B). To this end, we consistently observe a more homogeneous G2/M population when using colcemid/nocodazole compared with non-antimicrotubule modalities. This suggests that the "G2/M" populations used can vary significantly depending on the method of isolation.

Glycosylation of K8/18
Tryptic Peptide Mapping and Isoelectric Focusing Analysis of Glycosylated and Phosphorylated K8/18 Species-Using isoelectric focusing, we asked if the phosphorylated and glycosylated K8/18 species had identical PI values, which indicates that both modifications occur on the same molecules. The major K8/18 phosphorylated species do not correspond to the major Coomassie-stained species in asynchronously growing cells (Fig. 5a). Colcemid treatment generated a second major Coomassie-stained K8 species which corresponded to one of two major phosphorylated spots with minimal change in the K18 profile (Fig. 5b). In contrast to the phosphorylated species, the major glycosylated species corresponded to the major Coomassie-stained K8 and K18 species in asynchro-nously growing and colcemid-treated cells (Fig. 5, c and d ) .
We also examined the effect of G2/M cell arrest on the phosphorylated and glycosylated tryptic peptide pattern of K8 and K18. As shown in Fig. 6, although the overall phosphorylation of K8 and K18 increased in association with colcemid treatment, there were several distinct patterns of change. For example, several K8 phosphopeptides showed a relative decrease (arrows) or an increase in labeling intensity (arrowheads). For K18 phosphopeptides, some became newly phosphorylated after G2/M arrest (arrowheads), and others were increased (dotted arrows). Similarly, two newly glycosylated peptides can be seen for K8 (arrows), whereas K18 showed a uniform increase in peptide glycosylation.
Characterization of K8/18 Glycosylation and Phosphorylatwn Sites during Mitosis-Phosphoamino acid analysis of K8 and K18, obtained from asynchronously growing cells or Gz/ "arrested cells, was carried out. As shown in Fig. 7, serine was the only phosphorylated amino acid in K8 and K18.
The in uitro galactosylation of K8/18 shown in Fig. 2 indicates that these keratins contain terminal GlcNAcs. We recently showed that p-elimination of in uitro galactosylated K8/18, isolated from asynchrously growing cells, generated the disaccharide N-acetyllactosaminitol. This indicated that K8/18 in Go/GI cells contain single 0-GlcNAc residues . Similarly, as shown in Fig. 8, the increased glycosylation in K8/18 during G2/M arrest can be accounted for by single 0-linked GlcNAcs.
We determined the site of K8/18 glycosylation using NaBH4/NaOH p-elimination. This reaction converts O-glycosylated serine and threonine residues to alanine and aaminobutyrate, respectively. If [3H]NaBH4 is used, labeled amino acid products are generated as was done for the identification of serine as the 0-linked site in nuclear pore proteins (Holt et al., 1987b). The addition of palladium(I1) cation as a catalyst dramatically enhances threonine to a-aminobutyrate conversion without affecting the generation of alanine from serine (Tanaka and Pigman, 1965). Treatment of K8/18 with [3H]NaBH4 (which is available only in the mM range of specific activity) resulted, after acid hydrolysis, in a single species which did not comigrate on TLC with alanine or aaminobutyrate (not shown). This suggested that reduction of the diene amino acid derivative was incomplete, which was confirmed by carrying out p-elimination using 1 M unlabeled NaBH4 followed by HC1 hydrolysis and amino acid analysis. As shown in Fig. 9, c and d, both K8 and K18 generated aaminobutyrate (indicated by an arrow) after reductive Belimination, consistent with threonine being a site for K8/18 0-linked glycosylation. The mol of a-aminobutyrate generated were 2.0 and l.O/mol of K8 and K18, respectively. This is likely an underestimation of the number of glycosylated threonine residues since the recovery of amino acids in the peliminated samples was low (compare a and b with c and d, respectively, Fig. 9). Since quantitative and complete recovery of hydrolyzed amino acids are not feasible using this technique, it is possible that K8/18 serine residues may also be 0glycosylated. However, it is unlikely that serine is a major glycosylated residue since p-elimination of K8 and K18 using [3H]NaBH4 generated only a single unidentified labeled product, which upon addition of 1 M NaBH4, generated a-aminobutyrate (not shown).
Distribution of K8 and K18 Glycosylatwn and Phosphorylation within the Protein Backbone-Both K8 and K18 contain a large number of Ser/Thr residues scattered throughout the entire molecules (59/21 for K8 and 37/30 for K18; Kulesh and Oshima, 1988;Yamamoto et al., 1990). Using selective cleavage at tryptophan residues (Savige and Fontana, 1977), we asked if K8/18 phosphorylation and glycosylation localized to a particular domain (i.e. head, rod, or tail domains of the K8 K18 + electrophoresis pH 1.9 -FIG. 7. K8/18 from asynchronous or G2/M-arrested cells are phosphorylated on serine residues. HT29 cells (with or without colcemid-induced arrest) were labeled metabolically with 32P04 for 2 h followed by immunoprecipitation of K8/18. Phosphorylated species were analyzed using SDS-PAGE followed by electroelution of the indicated band and phosphoamino acid analysis using two-dimensional electrophoresis as described under "Materials and Methods." PS, phosphoserine; PT, phosphothreonine; PY, phosphotyrosine. codazole-arrested cells showed a similar increase in K8/18 glycosylation. In the case of nocodazole, the mitotic arrest and increased glycosylation were reversible upon washing off the nocodazole, concomitant with continued growth of the cells and progression through the cell cycle. Our working model is that increased keratin glycosylation occurs in some but not all cells and relates to a specific perimitotic stage of the cell cycle. However, since we were unable to observe increased K8/18 glycosylation in Gz/M-enriched cells using nonmicrotubule-disrupting methods, we cannot exclude a microtubule effect for the observed glycosylation enhancement (see also below). Altered glycosylation does not depend on simple roundness of cells. For example, in colcemid/nocodazole-treated cells (36 h), the ratio of round floater to adherent cells varies depending on the confluence level when the drug is added. However, in all cases, both adherent and floater HT29 cells show nearly similar percent of G2/M cells as well as similar increased levels of glycosylation (not shown). In addition, treatment of asynchronous cells with trypsin, which results in single rounded cells, does not affect K8/18 glycosylation (not shown). The precise association of increased K8/18 glycosylation with keratin aggregation, an antimicrotubule effect, or specific cell cycle events remains to be determined. The microtubule effects of colcemid/nocodazole are also associated with intermediate filament reorganization (Fig. 1) and, in addition to mitotic arrest, may have resulted in the observed increase in K8/18 glycosylation in a mitosis-independent manner. However, keratin reorganization alone is not necessarily related to increased glycosylation. For example, disruption of keratins using cold hypotonic solution treatment of interphase cells (Tolle et al., 1987) does not affect the glycosylation level of K8/18 (not shown). Our inability to reproduce the increase in K8/18 glycosylation using aphidicolin-synchronized or log phase floater Gz/M-enriched cells suggests that increased K8/ 18 glycosylation may be more related to an antimicrotubule phenomenon than a mitosis event. However, the G2/M-enriched cells obtained as floaters from log phase growing cells or aphidicolin synchronization are not as homogeneous in their cell cycle stage as colcemid/nocodazole-arrested cells (Fig. 4B). This lack of homogeneity may account for the observed basal level of K8/18 glycosylation in Gz/M-enriched cells obtained without the use of mitosis-arresting agents.
Well defined roles for the single 0-GlcNAc modification remain to be elucidated. T o date, several cytoplasmic and nuclear proteins with this modification have been described, and the list is growing (for reviews see Hart et al., 1989a;Haltiwanger et al., 1992). One hypothesis has been that single 0-GlcNAcs play a role in the assembly and/or organization of multiprotein complexes (Roquemore et al., 1992). This is based on finding this modification in proteins involved in multimeric structures including nuclear pore proteins (Davis and Blobel, 1987;Holt et al., 1987b;Starr and Hanover, 1990), erythrocyte band 4.1 (Holt et al., 1987a), SP-1 (Jackson and Tjian, 1988), and serum response transcription factor (Reason et dl., 1992). The presence of single 0-GlcNAc in keratins (K8 and 18 ; K13 (King and Hounsell, 1989)) also falls in line with this hypothesis. For example, glycosylation may be involved in blocking or enhancing specific sites of proposed head-to-tail interactions involved in keratin filament assembly (Lu and Lane, 1990). Alternatively, glycosylation may alter keratin solubility or effect interaction with other intermediate filament-associated proteins. With regard to keratin solubility, we find that the soluble and insoluble K8/18 fractions have a similar level of glycosylation and a similar tryptic glycopeptide pattern.2 This suggests that K8/ 18 glycosylation does not play a direct role in keratin solubility.
The small number of tryptophan residues in K8/18 has allowed us to take advantage of cleavage a t these sites to define better the region(s) of glycosylation and phosphorylation. In the case of K18, most of the glycosylation and phosphorylation occurs in the N-terminal domain spanning the first 122 amino acids (Fig. 10). This domain has eight threonine potential glycosylation sites (Kulesh and Oshima, 1988) with four major labeled glycosylated tryptic peptides for the entire K18 species (Fig. 6). The same domain also has 18 serine potential phosphorylation sites, with -10 labeled phosphopeptides (Fig. 6). In the case of K8, glycosylation and phosphorylation are not restricted to the N-terminal domain C.-F. Chou Threonine is a site of K8/18 0-linked glycosylation. HT29 cells were grown in the presence of colcemid for 36 h followed by isolation of K8/18 immunoprecipitates. K8 and K18 were purified using preparative gels. Duplicate samples were analyzed directly by acid hydrolysis (6 M HCI, 110 "C, 24 h) or were treated with 1 M NaBH4, 0.1 M NaOH (37 "C, 19 h) in the presence of 20 mM NanPdCll for the last hour of incubation and then subjected to acid hydrolysis. The plots show the amino acid elution profile with time. The a-aminobutyrate standard eluted at 24.59 min, which corresponds to the position indicated by the arrow (panels c and d ) . For the elution profiles corresponding to non-0-eliminated K8 and K18, arrowheads indicate the retention time of the a-aminobutyrate standard (panels a and b).   (Fig. 10). Studies with other IF have shown that phosphorylation occurs for the most part on the head or head and tail domains (for review see Skalli and Goldman, 1991). For example, analysis of asynchronous cells showed that vimentin is phosphorylated primarily in the head domain, whereas desmin is phosphorylated in the head and tail domains (Evans, 1988). During mitosis, phosphorylation increases in both IF proteins primarily in the head domain (Evans, 1988). Several studies showed mitosis-associated increased phosphorylation of non-keratin IF including vimentin (Chou et al., 1989) and lamin (Luscher et al., 1991), and in keratins from HeLa cells (Celis et al., 1983;Tolle et al., 1987), ME-180 cells (Gilmartin et al., 1984), and the Xenopus oocyte (Klymkowsky et al., 1991). All reported studies indicate that type I1 keratin of the keratin heterodimer (e.g. K8) is the predominant species that is hyperphosphorylated. Our results extend these findings by showing that specific serine-containing peptides are phosphorylated during mitotic arrest. Furthermore, it appears that most of the glycosylation and phosphorylation of K8 and K18 during interphase and mitotic arrest occur on different molecules (Fig. 6), suggesting differential regulation and possibly differing mitosis-associated functions for the two modifications.
Given that Ser/Thr residues can be modified by phosphorylation and glycosylation, the theoretical possibility of glycosylation acting as a regulator of phosphorylation, as proposed by Hart and his colleagues, is exciting since it offers yet an additional regulation system that a cell may use (Hart et al., 1989b). In the case of K8/18 obtained from G2/Marrested HT29 cells, the major site of glycosylation is threonine (Fig. lo), whereas the major site of phosphorylation is serine (Fig. 8). Although threonine glycosylation may modulate the phosphorylation of an adjacent serine, both glycosylation and phosphorylation of K8 and K18 increased during mitotic arrest. This increase suggests a lack of a relationship between the two modifications which in turn is supported by finding the two modifications mostly on different molecules as mentioned above. To that end, inhibition of keratin phosphorylation in intact cells using staurosporine (Chou and Omary, 1991) does not alter K8/18 glycosylation (not shown).
The increase in K8/18 glycosylation associated with mitotic arrest showed differing tryptic peptide patterns of enhancement for K8 and K18. For example, K18 exhibited a uniform increase in glycosylation of the four major labeled tryptic peptides, whereas K8 showed specific peptides that became uniquely glycosylated (Fig. 10). In contrast, changes in the phosphorylation of specific tryptic peptides were noted for both K8 and K18 after G2/M arrest. Identification of the precise glycosylated and phosphorylated K8/18 residues during interphase and mitotic arrest should provide an important handle in determining their function(s). Interestingly, other forms of cell activation such as stimulation of murine T-cells with concanavalin A resulted, within 60 min, in significant alterations in the 0-linked single GlcNAc modification of several nucleoplasmic and cytoplasmic proteins, although the identity of the modified proteins was not determined (Kearse and Hart, 1991).