A Significant Soluble Keratin Fraction In

Journal of Cell Science 105, 433-444 (1993) 433 Printed in Great Britain © The Company of Biologists Limited 1993

A signiﬁcant soluble keratin fraction in ‘simple’ epithelial cells Lack of an apparent phosphorylation and glycosylation role in keratin solubility

Chih-Fong Chou*, Carrie L. Riopel, Lusijah S. Rott and M. Bishr Omary† Palo Alto Veterans Administration Medical Center and the Digestive Disease Center at Stanford University, School of Medicine, 3801 Miranda Avenue, GI 111, Palo Alto, CA 94304, USA *Author for reprint requests †Author for correspondence

SUMMARY

We studied the solubility of keratin polypeptides 8 and aments in vitro as determined by electron microscopy. 18 (K8/18), which are the predominant intermediate fil- Cross-linking of soluble K8/18 followed by immunopre- aments in the human colonic epithelial cell line HT29. cipitation resulted in dimeric and tetrameric forms, We find that asynchronously growing cells (G0/G1 stage based on migration in SDS-polyacrylamide gels. In of the cell cycle) have a substantial pool of soluble ker- addition, cross-linked and native soluble K8/18 showed atin that constitutes approx. 5% of total cellular ker- similar migration on nondenaturing gels and similar atin. This soluble keratin pool was observed after sedimentation after sucrose density gradient centrifu- immunoprecipitation of K8/18 from the cytosolic frac- gation. Our results indicate that simple epithelial ker- tion of cells disrupted using three detergent-free meth- atins are appreciably more soluble than previously rec- ods. Several other cell lines showed a similar significant ognized. The soluble keratin form is assembly competent soluble cytosolic K8/18 pool. Arrest of HT29 cells in and appears to be primarily tetrameric. Although K8/18 G2/M stage of the cell cycle was associated with a con- solubility was found to increase during mitotic arrest, current increase in keratin solubility. Comparison of glycosylation and phosphorylation did not play an obvi- K8/18 obtained from the soluble cytosolic fraction and ous role in generating the soluble fraction, suggesting the insoluble high-speed pellet fraction showed similar an alternate mechanism for keratin solubility. levels of phosphorylation and glycosylation and similar tryptic radiolabeled phospho- and glycopeptide pat- Key words: keratin glycosylation, keratin phosphorylation, keratin terns. Soluble K8/18 can form characteristic 10 nm fil- solubility

INTRODUCTION for a number of IF. For example, newly synthesized vimentin was detected in a precursor soluble pool with sub- Intermediate filaments (IF) represent one of the three major sequent incorporation into filaments (Blikstad and classes of cytoskeletal proteins that are present in most Lazarides, 1983). The soluble pool may also include pre- eukaryotic cells (Steinert and Roop, 1988). Of the three existing older molecules in addition to newly synthesized major cytoskeletal proteins, IF are felt to be the least sol- vimentin, as determined by pulse-chase labeling experi- uble; and within IF, keratins are considered insoluble in ments (Söellner et al., 1985). These observations suggested aqueous salt solutions (Steinert et al., 1982; Bershadsky and an exchange between the soluble and insoluble IF fractions, Vasiliev, 1988; Lazarides, 1982). In fact, a generally uti- which was supported by the observation that post-transla- lized method for keratin isolation makes use of their insol- tional generation of vimentin filaments from a soluble pool ubility in high-salt buffer systems (Achtstaetter et al., 1986). may occur with or without cotranslational filament forma- Also, complete solubilization of keratins generally requires tion, depending on the tissue source (Isaacs et al., 1989). a high concentration of urea, although different keratin pairs Similarly, cytosolic fractions from rat livers contained a exhibit different solubility in urea. For example, keratin keratin-like pool based on its filament-forming ability and polypeptides 5 and 14 (K5/14, nomenclature and numeri- partial peptide mapping analysis (Sahyoun et al., 1982). cal classification are based on that of Moll et al., 1982), Neurofilaments also appeared to be formed from soluble which are expressed in basal epidermal cells, can be solu- precursors that can be chased with time into the cytoskele- bilized in 2 M urea; whereas K1 and K10/11 (expressed in tal fraction (Black et al., 1986). In most cases, the soluble suprabasal cells) required 4-6 M urea (Eichner and Kahn, IF pool was not quantitated but was implied to be ‘small’. 1990). For example, a small soluble K8 and K18 pool was iden- The presence of a ‘small’ soluble pool has been shown tified in the human hepatocellular cell line PLC (Franke et 434 C.-F. Chou and others al., 1987). In addition, using immunoblot analysis, Xeno - or Eagle’s MEM supplemented with 10% fetal calf serum. Arrest pus laevis oocytes and eggs were estimated to contain 1 to of cells in G2/M using colcemid (0.5 mg/ml) and cell cycle analy- 10% of their total keratin in a soluble form (Gall and sis was carried out exactly as described (Chou and Omary, 1993). Karsenti, 1987). When quantitated, soluble vimentin was Monoclonal antibodies to K8/18 used were: L2A1 coupled to ® estimated to be less than 0.4% of total vimentin as deter- Sepharose beads (Chou et al., 1992), or to Affinica agarose beads mined by pulse labeling with [35S]methionine for 30 min as per manufacturer’s recommendation (Schleicher & Schuell, Keene, NH); and CK5 (Sigma). Other reagents used were UDP- then chasing for up to 15 h (Söellner et al., 1985). Fur- [4,5-3H]galactose (34.6 Ci/mmol), N-[6-3H]glucosamine-HCl 32 35 thermore, for some IF (e.g. neurofilaments), solubility (40.4 Ci/mmol), orthophosphate ( PO4), [ S]protein-labeling appears to be tissue dependent. For example, mice injected mix (1186 Ci/mmol), liquid and spray ENHANCE® (Du Pont- intravitreally with [35S]methionine showed a substantial New England Nuclear); catalase (Worthington Biochemical Corp., soluble pool of retinal but not optic axon neurofilament-L Freehold, NJ); human immunoglobulin G (IgG), galactosyltrans- (Nixon et al., 1989). ferase, bovine serum albumin (BSA), goat anti-mouse IgG- The factors involved in generating a soluble IF pool are Sepharose conjugate (Sigma); peroxidase-labeled goat anti-mouse not known. On the basis of in vitro studies of IF assembly, Ig (Kirkegaard & Perry Laboratories, Inc., Gaithersburg, MD); and the polar nature of the phosphate group, phosphoryla- electron microscopy (EM) grids (Ted Pella Inc., Redding, CA); tion is an attractive potential regulator of IF solubility. For disuccinimidyl suberate (DSS) crosslinking agent (Pierce, Rock- ford, IL). example, in vitro phosphorylation of filamentous vimentin (Inagaki et al., 1987; Chou et al., 1989), rat liver K8/18 Preparation of supernatant (S), pellet (P) and high- (Yano et al., 1991), desmin (Inagaki et al., 1988) and neu- salt extract (HSE) fractions rofilaments (Gonda et al., 1990) resulted in filament disassembly as determined by negative staining electron Cells grown asynchronously (primarily G0/G1 cells (G cells)) or arrested in G2/M (M cells) were washed twice with phosphate microscopy. However, analysis of the soluble vimentin buffered saline (PBS), then suspended in 1.5 ml of PBS, buffer fraction (Blikstad and Lazarides, 1983; Isaacs et al., 1989; A (50 mM Tris-HCl, pH 7.5, 5 mM MgCl2, 25 mM KCl), or Söellner et al., 1985; Lamb et al., 1989), the soluble ker- buffer B (10 mM Tris-HCl, pH 7.5, 1 mM MgCl2, 130 mM KCl, atin fraction (Gall and Karsenti, 1987) or soluble neurofil- 5 mM EDTA, 5 mM NaCl). All buffers and detergent solutions aments (Black et al., 1986; Nixon et al., 1989) indicated also contained 0.1 mM phenylmethylsulfonyl fluoride (added that phosphorylation does not appear to play a role in gen- fresh), 25 mg/ml aprotinin, 10 mM leupeptin and 10 mM pepstatin. erating the soluble or insoluble fractions. To generate the S and P fractions, cells were disrupted in buffer Recently we showed that K8/18 in HT29 cells undergo A, B or PBS (1.5 ml) using a cell disruption bomb (no. 4639, Parr 2 a dynamic O-linked glycosylation, with multiple glycosy- Instrument Company, Maline, IL) at 1000 lbf/in for 5 min, fol- lation sites consisting of single N-acetylglucosamine lowed by ultracentrifugation (300,000 g; 90 min, 4°C, SW50.1 rotor). Alternatively, cells were disrupted in buffer B using a (GlcNAc) residues (Chou et al., 1992). The dynamic nature Dounce homogenizer (100 strokes) or by freeze-thawing three of this modification was reflected by the faster rate of K8/18 times (- 80°C, 5 min, followed by rapid thawing in a 37°C water carbohydrate biosynthesis and degradation, when compared bath) followed by ultracentrifugation. with the corresponding rates for the protein backbone (Chou High-salt extraction was performed in a manner similar to that et al., 1992). Mitotic arrest of HT29 cells, using colcemid described before (Achtstaetter et al., 1986). Briefly, HT29 cells or nocodazole, resulted in a dramatic increase in both gly- (one confluent 100 mm dish) or the P fraction obtained after bomb- cosylation and phosphorylation of K8/18 (Chou and Omary, ing and then ultracentrifugation were mixed with 1 ml of 1% 1993). Threonine was a major site of glycosylation, whereas Triton X-100 (TX-100) 5 mM EDTA in PBS for 2 min (4°C) fol- serine was the primary site of phosphorylation (Chou and lowed by centrifugation (16,000 g, 10 min). The supernatant was Omary, 1993). Here we quantitate the soluble K8/18 frac- removed and used for immunoprecipitation or discarded. The pellet was homogenized in a high-salt buffer (1 ml) containing 10 tion and ask what happens to K8/18 solubility in mitotic mM Tris-HCl (pH 7.6), 140 mM NaCl, 1.5 M KCl, 5 mM EDTA, arrest, and what role does phosphorylation and glycosyla- 0.5% TX-100, followed by a 30 min incubation and then pellet- tion play in generating the soluble fraction. Our results indi- ing to yield the high salt extract (HSE; 16,000 g, 4°C, 5 min). cate that although keratin solubility increases during mitotic The HSE was rinsed with PBS to remove excess salt, then repel- arrest, neither modification plays an obvious role in gener- leted. The high-salt solution supernatant was dialyzed against PBS ating the soluble fraction. In addition, we extend our pre- containing 0.5% TX-100, then used for immunoprecipitation or vious finding that non-ionic detergent treatment of cells sol- western blotting. ubilizes an easily detectable K8/18 pool (Omary et al., 1992). Here we show that several cell lines contain a sig- Immunoprecipitation, western blotting and tryptic nificant pool of aqueous cytosolic soluble K8/18 (~5% of peptide mapping total keratin) in the absence of detergents. The soluble Immunoprecipitation was carried out on the S fraction, or on the K8/18 pool appears to be primarily tetrameric and is able detergent-solubilized pellets obtained after bombing and then to form filaments in vitro. ultracentrifugation (P fraction). Detergent solubilization of the P fraction was done in 1.5 ml of 1% Nonidet P-40 (NP-40), 1% deoxycholate, 0.1% sodium dodecyl sulfate (SDS) in buffer B using a Dounce homogenizer (100 strokes), followed by centrifu- MATERIALS AND METHODS gation (16,000 g; 15 min) to remove nonsolubilized material. To the S fraction, a 10´ detergent mixture in buffer B was added so Reagents and tissue culture that the final detergent concentration is identical to the solubilized HT29, PtK1, SK-CO-1 and HeLa cells (American Type Culture pellet detergent mix. The S or P detergent lysate (100-150 ml) was Collection, Rockville, MD) were grown in RPMI 1640 medium mixed with 15 ml of L2A1-Sepharose (or agarose) or with 2 ml Soluble keratin polypeptides 8 and 18 435 of L2A1 or CK5 mAb ascites (30-60 min, 4°C). Non-Sepharose- preparative gels, the equivalent of 12 immunoprecipitates were containing immune complexes were collected using 20 ml of goat loaded per gel. anti-mouse Ig-Sepharose conjugate (30 min, 4°C). After 3 washes, Metabolic labeling of HT29 cells (G cells) with [3H]glu- immunoprecipitates were analyzed using SDS-PAGE (Laemmli, cosamine was carried out in glucose-free RPMI 1640 medium sup- 1970). All gels were 10% acrylamide and were analyzed under plemented with 10% dialyzed FCS. Cells were labeled for 4 h (50 nonreducing conditions except when indicated. No difference in mCi/ml), then chased using normal medium for 20 h, followed by immunoprecipitation efficiency was noted when using buffer B or immunoprecipitation of K8/18 from the S and P fractions. PBS. Exposure of 32P-labeled gels was carried out using enhanc- ing screens (Laskey and Mills, 1977), and of 3H-labeled gels using Chemical cross-linking ENHANCE® solution as recommended by the manufacturer. The supernatant fraction (100 ml) obtained from G cells that were Western blotting was carried out as described (Omary et al., disrupted by freeze-thawing or bombing, was incubated with DSS 1992) except that an enhanced chemiluminescence system (ECL (1 mM, 15 min, 22°C). Cross-linking was quenched using glycine system, Amersham) was used for detection, and peroxidase- (40 mM) followed by immunoprecipitation, then analysis by SDS- labeled goat anti-mouse Ig was used as the second-stage antibody. PAGE or western blotting. Nondenaturing gel electrophoresis of L2A1 ascites (1:600 dilution) was used as the first-stage antibody. cross-linked and noncross-linked supernatant was performed using Exposure using the ECL system varied from 10 s to 10 min. Deter- buffers identical to those used in the Laemmli system but with- mination of protein concentration was performed using a kit (Bio- out SDS. Rad, Melville, NY). Isolation of K8 or K18 for tryptic peptide mapping (Chou et al., 1992) was performed by preparative SDS- Sucrose gradient centrifugation PAGE of K8/18 immunoprecipitates, followed by electroelution A discontinuous 5% to 30% sucrose gradient (in buffer B) in 5% of the individual K8 and K18 Coomassie-stained bands (Amicon increments was used. The S fraction (0.5 ml) of G cells (before electroelution apparatus, Davers, MA). Fluorography of the pep- ® or after cross-linking or a mixture of the two) was layered on top tide map plates was performed using ENHANCE spray. of the gradient, followed by ultracentrifugation (36,000 r.p.m., 18 h, SW40 rotor), then needle puncturing of the tube bottom and Estimation of the soluble keratin fraction collection of fractions (7 drops/fraction). A 4´ sample buffer solu- This was determined using two methods. In the first method, a tion was added to aliquots of the fractions, followed by analysis known number of nonlabeled HT29 cells (G and M cells) were by SDS-PAGE, transfer to nitrocellulose then western blotting divided into four fractions, which were used for: high-salt extrac- using mAb L2A1. Ultracentrifugation of the control proteins (cata- tion, disruption by nitrogen cavitation (5 min, 1000 lbf/in2, 4°C), lase, human IgG and BSA) and collection of fractions was car- solubilization in 1% Nonidet P40 and 5 mM EDTA in PBS (45 ried out in an identical manner, followed by SDS-PAGE and min), and determination of total cellular protein. K8/18 immuno- Coomassie staining to visualize the migration position of the indi- precipitates were prepared from the S fraction of bombed cells or vidual protein standards. Identical results were obtained if the pro- from the detergent-solubilized cells. The mg of K8/18 present in tein standards were mixed with the S fraction prior to ultracen- immunoprecipitates or in the HSE was determined by comparing trifugation. the densitometric scanning of the Coomassie-stained bands of K8/18 with BSA standards (LKB ultrascan XL enhanced laser Negative staining and electron microscopy (EM) densitometer). Similarly, the amount of K8/18 released by 1% TX- HT29 cells (G and M) were used to generate an S fraction by 100 during the high-salt extraction method was determined after freeze-thawing in buffer B. The S fraction was dialyzed against immunoprecipitation (~3% of total keratin), and used in calculat- 2.5 mM Tris-HCl, 1.25 mM Na2EDTA, 5 mM 2-mercaptoethanol ing the total cellular keratin (HSE + TX-100 material). The high- (pH 7.5, 4°C) for 20 h. For EM, carbon-coated 300 mesh copper salt supernatant contains minimal keratin (<1%), as determined by grids were used. Dialyzed and nondialyzed S fraction (20 ml) were dialysis followed by immunoprecipitation of the dialysate using used to coat the grids for 3 min. After rinsing with 5 drops of anti-K8/18 mAb L2A1 and, therefore, was not included in the cal- 0.1% BSA in PBS (pH 7.4), excess buffer was removed with filter culation of total cellular K8/18. To insure complete and quantita- paper strips. Grids were then stained with 0.5% uranyl acetate for tive isolation of the soluble keratin, three rounds of immunopre- 3 min, followed by removal of the staining solution, air drying, cipitation were done. In the second method, the soluble fraction then visualization by EM (´ 45,000). was determined as in the first method except that HT29 cells were labeled with [35S]methionine (100 mCi/ml) for 30 min, then chased with normal media for 13 h. Densitometry scanning of the K8/18 bands was performed on the autoradiograph. RESULTS

Radioisotope labeling Isolation of soluble K8/18 using different buffer 32 and lysis conditions G or M HT29 cells were labeled with H3 PO4 (carrier-free, 125 mCi/ml) in 99% phosphate-free medium supplemented with 5% A soluble cytosolic cellular fraction was generated after dialyzed fetal calf serum (16 h). After labeling, cells were used ultracentrifugation of a detergent-free cell homogenate. In for high-salt extraction of K8/18 or were bombed. Ultracentrifu- preliminary experiments, we tested three buffer systems in gation was then performed followed by immunoprecipitation of which cell disruption using nitrogen cavitation was carried K8/18 from the S and P fractions as described above. All buffers out. Although all three buffer systems afforded a signiﬁ- used in phosphorylation experiments also contained 5 mM sodium cant soluble keratin fraction in the absence of detergent pyrophosphate, 50 mM sodium ﬂuoride and 1 mg/ml okadaic acid. (Fig. 1A), buffer B resulted in the least amount of keratin For UDP-[3H]galactose labeling, K8/18 immunoprecipitates degradation. In addition, we reasoned that a buffer high in obtained using L2A1-agarose were incubated for 2 h (with mixing) + + with 25 ml of 20 mM MnCl2, 100 mM sodium cacodylate (pH K and low in Na mimics more closely the intracellular 6.5), 0.6 mCi UDP-[3H]galactose and 25 munits galactosyltrans- cation concentration found in colonocytes (Del Castillo et ferase. 3H-labeled immunoprecipitates were then washed once al., 1991). We compared cell disruption, using nitrogen with 0.5% NP-40 in PBS, and analyzed by SDS-PAGE. For cavitation ‘bombing’, with freeze-thawing three times, or 436 C.-F. Chou and others

- 3 Mr´ 10

Fig. 1. Isolation of soluble K8/18 using different lysis and buffer conditions. (A) Asynchronously growing HT29 cells were lysed in buffer A, PBS or buffer B using nitrogen cavitation (1000 lbf/in2, 5 min, 4°C) as described in Materials and Methods. Bombed cell homogenates were pelleted (90 min, 300,000 g, 4°C) to generate a soluble (S) fraction, followed by immunoprecipitation of K8/18 from the S fraction using mAb L2Al. Small arrows indicate degradation products of K8/18 that are variably seen. (B) Comparison of K8/18 immunoprecipitates (using mAb L2A1) obtained from the soluble fraction of HT29 cells lysed using repeated freeze-thaw, homogenization using a Dounce homogenizer or nitrogen cavitation as described in Materials and Methods. (C) Immunoprecipitates of K8/18 were prepared from the S fraction of bombed HT29 cells, using mAbs CK5 and L2A1. The small arrow between K8 and K18 corresponds to partially reduced Ig heavy chain of goat anti-mouse Sepharose (see Materials and Methods). The small arrow (Mr>200,000) indicates the antibody band used for immunoprecipitation. homogenization using a Dounce homogenizer (100 keratin pool is soluble in G cells. After mitotic arrest, the strokes). All three methods of cell disruption resulted in solubility increased (Fig. 2) so that 11.6 + 4.3% of the total nearly equivalent levels of immunoprecipitable soluble keratin was soluble. Solubilization of asynchronous HT29 K8/18 fraction after a high-speed spin (Fig. 1B). We opted cells with nonionic detergents for 45 min resulted in approx. to use nitrogen cavitation as our routine method for cell 19% of the keratin being in the detergent phase (Fig. 2). disruption, since it is less operator-dependent. Similar As noted for the S fraction, mitotic arrest of HT29 cells amounts of soluble K8/18 were also obtained when nitro- also increased keratin solubility in NP40 to approx. 32% gen cavitation pressures of 400, 600 and 800 were com- (Fig. 2). On the basis of measurement of the total cellular pared with 1000 lbf/in2 (not shown). In addition, similar protein (not shown) and our estimation of the total cellular levels of soluble K8/18 were obtained by immunoprecipi- keratin as described in Materials and Methods, nearly 5% tation from the soluble fraction of PtK1 (marsupial kidney), of the total protein in HT29 asynchronous cells is made up HeLa (human cervix) and SK-CO-1 (human colon) cell of K8/18. lines (not shown). We also estimated the % soluble keratin in asynchronous Soluble K8/18 was also obtained by immunoprecipita- HT29 cells after metabolic labeling with [35S]methionine, tion using a commercially available, previously described, as in the method used to measure soluble vimentin (Söell- anti-K8/18 antibody termed CK5 (Tölle et al., 1985; and ner et al., 1985). We find similar results of soluble Fig. 1C). Therefore, regardless of the antibody used, a high [35S]methionine-labeled keratin in the S fraction (4.4-6% capacity K8/18 immunoprecipitating antibody is able to of the total [35S]methionine-labeled K8/18) when identify a significant soluble K8/18 cytosolic fraction. [35S]methionine-labeled cells were used (not shown, see Materials and Methods). Quantitation of K8/18 solubility in HT29 cells We estimated the level of soluble K8/18 in asynchronously Comparison of the phosphorylation of soluble and growing HT29 cells (primarily G0/G1 stage of the cell cytoskeletal K8/18 from G and M cells cycle) by comparing the amount (in mg) of K8/18 immuno- We investigated whether differences in phosphorylation precipitated from the S fraction with the amount of total play any role in generating the soluble K8/18 fraction. In cellular keratin. Estimation of total cellular keratin (see doing so, we analyzed two soluble K8/18 pools: (1) the Materials and Methods) is based on the finding that high- baseline soluble pool present in asynchronous G cells; and salt extraction of cells leaves a nearly quantitative recov- (2) the soluble pool present in G2/M arrested cells (M cells), ery of the keratins, termed the HSE (Achtstaetter et al., which has an increased 5- to 10-fold overall K8/18 phos- 1986), after stripping away most of the nonkeratin proteins. phorylation (Chou and Omary, 1993) and an increased Using immunoprecipitation to isolate soluble K8/18 from K8/18 soluble fraction (Fig. 2). As shown in Fig. 3, both HT29 cells (Fig. 2), we estimate that 5.6 + 2.4% of the total S and P fractions of G cells (lanes b and c, respectively) Soluble keratin polypeptides 8 and 18 437

Fig. 2. K8/18 solubility in G2/M arrested and asynchronous HT29 cells. Asynchronously growing HT29 cells (G cells) were used directly or were treated with colcemid (0.5 mg/ml) for 36 h (M cells). Equal numbers of G and M cells were used to generate an S fraction, or to solubilize in 1% NP- 40 in PBS containing 5 mM EDTA, or to obtain a HSE as described in Materials and Methods. Aliquots from the S fraction or from NP-40 detergent-solubilized cells were then used for 3 rounds of immunoprecipitation. After SDS-PAGE and Coomassie staining, densitometric scanning of the K8/18 bands and the shown BSA standards run on the same gel were used to estimate the percentage of soluble keratin: total mg of soluble K8/18 % soluble keratin = ————————————— ´ 100. total mg of keratin in cells used Two other experiments gave similar results (not shown), and the calculated % soluble keratin was obtained from the mean of 3 experiments. have similar phosphorylation levels, and so do the S and P tryptic peptide map analysis of isolated K8 and K18 from 32 fractions of M cells (lanes e and f, respectively). Further- the S and P fractions of PO4-labeled M cells. As shown more, the mitotic-arrest-associated increase in K8/18 phos- in Fig. 4, the S and P tryptic peptide maps of K8 and K18 phorylation occurs in both the S and P fractions (Fig. 3, are essentially identical. The peptides marked by open compare lanes b, c with e, f). The K8/18 phosphorylation arrowheads (Fig. 4c,d) likely represent degradation in immunoprecipitates of the S and P fractions of G and M products of K8, based on peptide maps of mixed K8 and cells was also similar to the overall phosphorylated keratin K18 (not shown). A similar tryptic peptide pattern was pool isolated by high-salt extraction of intact G and M cells also obtained from the S and P fractions for both K8 and 32 (Fig. 3, lanes a and d). Although in different experiments K18 after isolation from G cells labeled with PO4 (not slight differences were noted between the level of K8/K18 shown). phosphorylation in the S and P fractions, we interpret these differences to be due to minimal degradation of the more Glycosylation of K8/18 in the soluble and heavily phosphorylated K8. cytoskeletal fractions of G and M cells The lack of a role for phosphorylation in generating the We asked if glycosylation plays a role in generating the soluble K8/18 pool was also confirmed by two-dimensional soluble K8/18 fraction. This was assessed by in vitro galactosylation of K8/18 immunoprecipitates using UDP- [3H]galactose and galactosyltransferase (Törres and Hart, Mr´ 10- 3 1984). In vitro labeling by galactosylation reflects labeling of available terminal GlcNAc residues, and using this technique we generally find better labeling of K18 as compared with K8 (Chou and Omary, 1993). As shown in Fig. 5a, K8/18 from the S and P fractions are glycosylated to similar levels in G or M cells, although K8 showed a trend towards lower labeling in the P fractions, which may be due to variable accessibility of terminal GlcNAc residues (see in vivo labeling below). Glycosylation of K8/18 in the S and P fractions of M cells was higher than the corresponding fractions in G cells (Fig. 5a) as was observed for K8/18 obtained from detergent-solubilized G and M cells (Chou and Omary, 1993). A similar level of K8/18 glycosylation was also noted in the S and P fractions of G cells labeled metabolically with 3 [ H]GlcNH2 (Fig. 5b). Using in vivo labeling, both K8 and K18 were labeled to similar levels (Fig. 5b). Incubation of 3 32 cells with [ H]GlcNH2 but not PO4 prevents their colcemid-induced arrest in G2/M, as described (Chou and Fig 3. Phosphorylation of the soluble and insoluble K8/18 from G Omary, 1993). Therefore, since we were not able to ade- 32 and M cells. HT29 cells were labeled with PO4 (125 mCi/ml, 16 quately study K8/18 glycosylation in M cells by in vivo h), followed by analysis of K8/18 from the HSE of labeled cells labeling, we utilized in vitro galactosylation to further eval- (lanes a,d). Alternatively, labeled cells were disrupted using nitrogen cavitation, followed by a high-speed spin and uate K8/18 glycosylation at the peptide level in the S and immunoprecipitation of K8/18 from the soluble (b,e) or pellet P fractions of M cells. Two-dimensional tryptic peptide 3 fractions (c,f) as described in Materials and Methods. The mapping of [ H]galactosylated K8 and K18 showed simi- radiograph was obtained from the Coomassie-stained gel above it. lar patterns for each keratin isolated from the S and P frac- Equal amounts of protein were loaded. tions (Fig. 6). Taken together, our results indicate that thre- 438 C.-F. Chou and others

10 nm diameter filaments form (Fig. 7b). This indicates that soluble K8/18 is able to form typical IF structures. Under the conditions we used, the filament-forming efficiency of the soluble keratin was low, such that ultracentrifugation of the dialyzed material did not pellet significant amounts of filamentous keratin (not shown). The S fraction obtained from M cells was also able to form filaments after dialysis (not shown). For unclear reasons, we noted varibility in the caliber and aggregation of the formed filaments in the S fraction of G and M cells in different experiments (not shown). We also examined the molecular form of the soluble K8/18 species using chemical crosslinking and comparison of cross-linked and noncross-linked species after sucrose gradient sedimentation. The smallest basic unit of IF that can exist in solution is a tetramer (e.g. see Steinert and Roop, 1988, for review). This basic unit consists of homopolymers in the case of type 3 IF, such as vimentin and desmin, and heteropolymers in the case of keratins (Quinlan et al., 1984). Cross-linking of the S fraction with DSS followed by immunoprecipitation of K8/18 results in formation of a major cross-linked species (Mr approx. 300,000) that likely corresponds to tetrameric K8/18 (Fig. 8A, lane 2). Although the S fraction contains the entire cytoplasmic protein milieu and cross-linking may occur between K8 and K18 as well as between K8/18 and other associated proteins, keratin-keratin cross-linking appears to be very efficient. This is supported by the strong reactivity of the 300,000 Mr band (labeled as T, Fig. 8A), as shown by western blotting using anti-K8/18 antibody, with a con- comitant decrease in SDS-resolved monomeric K8/18 (Fig. 8A, lanes 2 and 4). Furthermore, analysis of the cross-linked material using a nondenaturing gel showed that both cross- linked and noncross-linked K8/18 have similar mobilities. For unclear reasons, the cross-linked K8/18 band migrates towards the margins of the lane under nondenaturing con- 32 Fig. 4. Tryptic peptide maps of K8/18 (labeled with PO4) from ditions (Fig. 8A, lane 6), and does not form a linear sym- soluble (S) and insoluble (P) fractions of M cells. HT29 cells were metric band. Formation of the K8/18 dimer was too faint 32 arrested in G2/M using colcemid (36 h) and labeled with PO4 to see by Coomassie staining but can be seen by western during the last 16 h of colcemid treatment. Immunoprecipitates of blotting (Fig. 8A, lane 4). K8/18 were obtained from the S and P fractions followed by SDS- Supporting evidence for the tetrameric unit of soluble PAGE analysis, and then electroelution of the individual K8 and K8/18 was obtained using sucrose gradient sedimentation. K18 proteins. Purified K8 and K18 were treated with trypsin, then As shown in Fig. 8B, K8/18 sedimented in fractions that subjected to electrophoresis and chromatography. Filled nearly coincided with human IgG. This is similar to the arrowheads indicate the origin where 1000 c.p.m. were spotted per sample (4- and 6-day exposure for K18 and K8, respectively). sedimentation position reported for soluble tetrameric vimentin (Söellner et al., 1985) and supports a tetrameric unit of soluble K8/18. Analysis of a mixture of cross-linked onine/serine glycosylation and serine phosphorylation are and noncross-linked S fraction showed that K8/18 dimers unlikely to play a significant role in K8/18 solubility. and tetramers appeared nearly to coincide with K8/18 monomer sedimentation (Fig. 8B). The peak of cross- Assembly competency and molecular nature of linked keratin appears to be slightly shifted towards cata- the soluble K8/18 species lase as compared with the peak of noncross-linked keratin We studied the assembly competency of the soluble K8/18 (Fig. 8B). This cross-linking-induced effect was also species present in HT29 cells after cell disruption. Poly- observed in cross-linked soluble keratins in PLC cells merization and filament formation of purified keratins have (Franke et al., 1987). When noncross-linked or cross-linked been shown to occur in vitro upon dialysis in dilute salt S fractions were analyzed separately by sucrose gradient solutions (Steinert et al., 1976). As expected after high- sedimentation, an identical pattern to that in Fig. 8B was speed centrifugation, electron microscopy of the S fraction observed, except that only SDS-resolved monomers were obtained from G cells prior to dialysis indicates that it does noted for the noncross-linked S fraction after SDS-PAGE not contain any filament-like structures (Fig. 7a). However, and western blotting (not shown). Only one band is seen after dialysis in a low-salt solution, an extensive array of for noncross-linked K8/18 because the L2A1 antibody rec- Soluble keratin polypeptides 8 and 18 439

A B - 3 - 3 Mr´ 10 Mr´ 10

Fig. 5. Glycosylation of K8/18 in the S and P fractions of G and M cells. (A) Soluble and insoluble fractions of bombed G and M HT29 cells were prepared, followed by immunoprecipitation of K8/18. In vitro galactosylation was carried out on the immunoprecipitates using UDP[3H]galactose and galactosyltransferase as described in Materials and Methods. The radiograph was obtained from the Coomassie-stained gel above it. Equal amounts of protein were loaded. (B) Asynchronous HT29 cells were labeled with [3H]glucosamine as described in Materials and Methods, followed by immunoprecipitation of K8/18 from the S and P fractions.

ognizes K18 preferentially by western blotting (Omary et al., 1992).

DISCUSSION

K8 and K18 are the major keratins found in the human colonic epithelial cell line HT29 used in this study. Together, K8/18 make up approximately 5% of the total cellular protein. In this report we have addressed several aspects of K8/18 solubility in HT29 cells. The major findings of our study are: (1) significant levels of K8/18, which are the predominant IF in ‘simple’ single layered epithelial cells such as the intestine, liver and pan- creas, are found in a soluble form in several cell lines tested; (2) soluble K8/18 in HT29 cells is assembly competent and appears to consist primarily of tetramers; (3) K8/18 solubility increases during arrest of cells in G2/M; (4) although K8/18 solubility increases during mitotic arrest, neither phosphorylation nor glycosylation plays an obvious role in generating the soluble fraction. The significant soluble K8/18 fraction was clearly identified using high-affinity immunoprecipitating antibodies and several cell disruption methods (Figs 1 and 2).

Simple epithelial keratins have a substantial soluble fraction The level of soluble K8/18 in asynchronous cells (approx. 4-6%) is significantly greater than what was thought to be negligible solubility, although studies of K8/18 solubility in Xenopus laevis oocytes and eggs suggested that 1-10% of K8/18 may be soluble (Gall and Karsenti, 1987). Sev- eral possibilities may explain this somewhat unexpected finding. First, K8/18 isolation using high-salt extraction, a Fig. 6. Tryptic peptide maps of [3H]galactosylated K8/18 from the 3 commonly used technique, does indeed provide a nearly soluble and insoluble fractions of M cells. [ H]galactosylated quantitative recovery of K8/18 in the pellet ‘insoluble’ frac- K8/18 immunoprecipitates were obtained from the S and P tion (~96% recovery). However, this does not imply that fractions of bombed M cells as in Fig. 5, followed by purification of individual K8 and K18. Tryptic peptide maps were produced as all of the keratins are insoluble. Second, of several keratin in Fig. 4. Arrows indicate loading origin, where 10,000 c.p.m. heteropolymers tested, K8/18 are the most soluble in urea were spotted for K18 (10 day exposure) and 5,000 c.p.m. were (Franke et al., 1983; Eichner and Kahn, 1990) and, there- spotted for K8 (30-day exposure). fore, are likely to have the largest cytosolic pool. Third, 440 C.-F. Chou and others

Fig. 7. K8/18 in the S fraction can assemble into ﬁlaments. The S fraction from asynchronously growing HT29 cells was prepared for negative staining and electron microscopy as described in Materials and Methods. (A) Nondialyzed S fraction. (B) Dialyzed S fraction. Bar, 50 nm.

K8/18 solubility may appear artificially low, depending on keratin pool using the high-salt extraction method may the method of detection used. For example, the S fraction overestimate the soluble pool. from G cells showed artificially low levels of K8/18 when The fraction of soluble K8/18 was concordant as deter- analyzed by western blotting (not shown), but higher levels mined by the amount of Coomassie-stainable keratin when analyzed by immunoprecipitation. Our estimation of obtained after exhaustive immunoprecipitation of the solu- the soluble K8/18 fraction (approx. 4-6%) may actually be ble fraction (approx. 5%), or as determined by the fraction somewhat higher or lower, depending on two factors. Inef- of soluble [35S]methionine metabolically labeled keratin ficient immunoprecipitation may underestimate the soluble (approx. 4-6%). Our results of 4.4-6% soluble [35S]methio- K8/18 pool, whereas inaccuracy in determining the total nine-labeled keratin after 30 min pulse and 13 h chase dif- Soluble keratin polypeptides 8 and 18 441 fered dramatically from the 0.18-0.39% soluble vimentin within 20 min, in dispersal of the biotin spots throughout identified in cultured rat cells (Söellner et al., 1985). Sev- the cytoplasm, followed by colocalization with the endoge- eral possibilities may account for the difference: (1) nous filament network (Miller et al., 1991). Also, transfec- vimentin may indeed be less soluble than K8/18, although tion of chicken vimentin into mouse fibroblasts resulted in the high-salt extraction method is not used for vimentin iso- incorporation of newly synthesized vimentin at multiple lation due to vimentin solubility under the same conditions; sites throughout the cytoplasm (Ngai et al., 1990). The pres- (2) the isolation of soluble vimentin by binding to single- ence of a substantial soluble K8/18 fraction that increases stranded DNA (Traub et al., 1983), as carried out by Söell- during mitotic arrest supports the dynamic nature of IF, and ner et al. (1985), may only remove part of the vimentin, allows for the exchange between the filamentous and non- thereby potentially resulting in underestimation of the sol- filamentous compartments. As mentioned previously (Chou uble vimentin pool. and Omary, 1993), we cannot exclude the possibility that Although our studies strongly suggest that simple epithe- effects observed after colcemid treatment may be more lial cells contain a significant soluble K8/18 pool, the true related to a microtubule effect than a mitosis-specific phe- solubility in intact cells may be different. Factors such as nomenon. temperature, changes in specific ion concentration(s), and cellular dynamics, may become important. However, our Role of glycosylation and phosphorylation in observed values are more likely to resemble in vivo solu- K8/18 solubility bility than measurements done on isolated in vitro systems We also investigated two potential factors that may play a using purified and reconstituted IF in aqueous buffers. role in keratin solubility: namely, phosphorylation and gly- The ability of K8/18 in the S fraction to form fil a m e n t s cosylation. Both modification levels increased during and the cross-linking and sucrose gradient sedimentation mitotic arrest (Chou and Omary, 1993) as did K8/18 solu- evidence supporting its tetrameric form are similar to what bility (Fig. 2). Our results showed that the level of phos- has been shown for soluble vimentin (Söellner et al., 1985). phorylation and glycosylation of K8 and 18 obtained from Higher oligomeric forms were not observed in our cross- the S and P fractions of G cells were similar. Identical find- linking experiments. Furthermore, analysis of K8/18 in an ings were noted for the S and P fractions of M cells (Figs S fraction prepared using a low cetrifugation speed (30 min 3 and 5). Furthermore, the tryptic peptide map profiles of at 15,000 g, which pellets only insoluble fil a m e n t o u s the glycosylated and phosphorylated K8 and K18 from S K8/18) showed identical sedimentation to that of the S and P fractions were identical to the extent that no promi- fraction obtained using high-speed centrifugation (not nent new or deleted peptides, or any significant alteration s h o w n ) . in peptide intensities, were noted (Figs 4 and 6). Taken together, our data indicate that phosphorylation and glyco- K8/18 solubility increases with mitotic arrest sylation do not contribute in an all-or-none fashion or in an Our finding of increased soluble keratin in mitosis, provides obvious site-specific manner to K8/18 solubility. biochemical support for the morphological findings of Although in vitro studies support a role for phosphory- increased protofilament and altered IF structure in many, lation in inducing filament disassembly (Ingaki et al., 1987, but not all, cell types undergoing mitosis (Franke et al., 1988; Gonda et al., 1990; Chou et al., 1989; Yano et al., 1982; Lane et al., 1982). One possible role for the increased 1991), studies, in intact cells, of vimentin (Blikstad and keratin solubility during mitosis, at least in part, would be Lazarides, 1983; Isaacs et al., 1989; Söellner et al., 1985; to facilitate the redistribution of keratins among two newly Lamb et al., 1989), keratins (Gall and Karsenti, 1987) and formed cells (Skalli and Goldman, 1991). However, a sig- neurofilaments (Black et al., 1986; Nixon et al., 1989) do nificant level of soluble K8/18 is also detected in nondi- not support a solubility role for phosphorylation. Our results viding G0/G1 cells, and therefore other functions are likely extend these observations and include examining phospho- to be involved. For example, a soluble keratin pool can act rylation and glycosylation, and potential changes in specific as a precursor to the cytoskeletal fraction as has been shown peptides, not only overall phosphorylation or isoelectric pat- for vimentin in some cell systems (Blikstad and Lazarides, terns. We cannot, however, exclude the possibility that rec- 1983; Isaacs et al., 1989). Alternatively, a significant solu- iprocal changes may occur in peptides that contain more ble K8/18 pool may also be involved in the interaction and than one serine or threonine residue, which could affect sol- possibly regulation of other cellular proteins that are ubility but not result in an overall change in the labeling involved in nonmitotic events. intensity of that particular phospho- or glycopeptide. On the basis of the work of several investigators (see Steinert and Liem, 1990, for review) it has become clear Potential factors to consider in intermediate that IF are not static, but rather are dynamic, entities. One filament solubility evidence of the dynamic nature of IF is the reorganization Several possibilities remain to be explored to explain the and reversible disassembly that takes place during mitosis enhanced solubility during mitotic arrest and the presence (Lane et al., 1982; Franke et al., 1982), although IF dis- of a steady-state soluble fraction. First, the biophysical ruption is not uniformly found in all cells (Tölle et al., properties of the IF within the cellular environment may 1987). Other evidence for the dynamic nature of IF is the dictate a level of solubility that may or may not relate to exchange that occurs between newly synthesized or post-translational modification. Hence, K8/18 may have a microinjected IF and the endogenous filamentous network. solubility equilibrium constant that differs between G and For example, microinjection of biotinylated bovine tongue M cells. In addition, mitosis-associated changes in ion con- type I keratin into primary mouse epidermal cells resulted, centration may induce filament disassembly. For example, 442 C.-F. Chou and others

- 3 Mr´ 10

Fig. 8

changes in the NaHCO3 concentration in the growth an undefined specific association between IF (S or P frac- medium may effect the ability of some cell lines to undergo tion) and a cellular protein may act as a ‘solubility factor.’ IF disassembly during mitosis (Tölle et al., 1987). Second, To this end, we find that the apparent migration of K8/18 Soluble keratin polypeptides 8 and 18 443

Fig. 8. Cross-linking and sucrose-gradient sedimentation of the S REFERENCES fraction support a tetrameric soluble K8/18. (A) Cross-linking of the S fraction using DSS was carried out as described in Materials Achtstaetter, T., Hatzfeld, M., Quinlan, R. A., Parmelee, D. C. and and Methods. Lanes 1, 2: immunoprecipitates of K8/18 were Franke, W. W. (1986). Separation of cytokeratin polypeptides by gel prepared from noncross-linked and cross-linked S fractions, electrophoretic and chromatographic techniques and their identification followed by SDS-PAGE (6% acrylamide gel) and Coomassie by immunoblotting. Meth. Enzymol. 134, 355-371. staining. Lanes 3 and 4, identical immunoprecipitates to those Bershadsky, A. D. and Vasiliev, J. M. (1988) In Cytoskeleton, pp. 133- shown in lanes 1 and 2 were transferred to nitrocellulose followed 154. Plenum Press, New York. by western blotting using mAb L2Al. Lanes 5,6: cross-linked and Black, M. M., Keyser, P. and Sobel, E. (1986). Interval between the synthesis and assembly of cytoskeletal proteins in cultured neurons. J. noncross-linked S fraction (without immunoprecipitation) were Neurosci. 6, 1004-1012. directly analyzed on a 6% acrylamide nondenaturing gel, followed Blikstad, I. and Lazarides, E. (1983). Vimentin filaments are assembled by transfer to nitrocellulose and western blotting using mAb from a soluble precursor in avian erythroid cells. J. Cell Biol. 96, 1803- L2A1. T, D and M arrows indicate K8/18 cross-linked tetramers 1808. and dimers, and noncross-linked monomers, respectively, after Chou, C.-F. and Omary, M. B. (1993) Mitotic-arrest associated SDS resolution. The open arrowhead indicates the antibody used enhancement of O-linked glycosylation and phosphorylation of human during immunoprecipitation. The antibody migrates below the keratins 8 and 18. J. Biol. Chem. 268, 4465-4472. 3 Chou, C.-F., Smith, A. J. and Omary, M. B. (1992). Characterization and 200´ 10 Mr marker (in contrast with Fig. 1) due to the lower percentage of acrylamide. (B) An equal mixture of cross-linked dynamics of O-linked glycosylation of human cytokeratin 8 and 18. J. and noncross-linked S fraction of bombed G cells was pooled and Biol. Chem. 267, 3901-3906. Chou, Y.-H., Rosevear, E. and Goldman, R. D. (1989). Phosphorylation sedimented on a 5% to 30% sucrose gradient (4°C 36,000 r.p.m., and disassembly of intermediate filaments in mitotic cells. Proc. Nat. 18 h; Beckman SW40 rotor). Fractions were collected and Acad. Sci. USA 86, 1885-1889. analyzed by western blotting using mAb L2A1. Migration of the Del Castillo, J. R., Ricabarra, B. and Sulbarán-Carrasco, M. C. (1991). standards catalase (Mr ~ 250,000), IgG (Mr ~ 150,000) and BSA Intermediary metabolism and its relationship with ion transport in (Mr ~ 66,000) is indicated by arrows. Fractions 1-12, which isolated guinea pig colonic epithelial cells. Am. J. Physiol. 260, C626- correspond to the denser end of the gradient, did not contain C634. keratins (not shown). Similar sedimentation was obtained when Eichner, R. and Kahn, M. (1990). Differential extraction of keratin cross-linked and noncross-linked S fractions from M cells were subunits and filaments from normal human epidermis. J. Cell Biol. 110, used (not shown). 1149-1158. Franke, W. W., Schiller, D. L., Hatzfeld, M. and Winter, S. (1983). Protein complexes of intermediate-sized filaments: Melting of from the S and P fractions is identical when analyzed on cytokeratin complexes in urea reveals different polypeptide separation characteristics. Proc. Nat. Acad. Sci. USA 80, 7113-7117. nondenaturing gels followed by western blotting (not Franke, W. W., Schmid, E., Grund, C. and Geiger, B. (1982). shown). However, we cannot completely exclude the pos- Intermediate filament proteins in nonfilamentous structures: Transient sible involvement of a small peptide, as has been reported disintegration and inclusion of subunit proteins in granular aggregates. for the binding of a 5 kDa peptide with unpolymerized actin Cell 30, 103-113. Franke, W. W., Winter, S., Schmid, E., Söllner, P., Hämmerling, G. and (Safer et al., 1990). So, a potential role for an associated Achtstätter, T. (1987). Monoclonal cytokeratin antibody recognizing a cytosolic solubility factor still needs to be explored. Third, heterotypic complex: immunological probing of conformational states of a nonprotein mediator may be involved as was recently cytoskeletal proteins in filaments and in solution. Exp. Cell Res. 173, 17- implicated for diacylglycerol, in a non-protein kinase C- 37. dependent fashion, in the nucleation of actin at plasma Gall, L. and Karsenti, E. (1987). Soluble cytokeratins in Xenopus laevis oocytes and eggs. Biol. Cell 61, 33-38. membrane sites (Shariff and Luna, 1992). Fourth, a post- Gonda, Y., Nishizawa, K., Ando, S., Kitamura, S., Minoura, Y., Nishi, translational modification that is different from threonine Y. and Inagaki, M. (1990). Involvement of protein kinase C in the glycosylation or serine phosphorylation may be involved. regulation of assemly-disassembly of neurofilaments in vitro. Biochem. An example is the tyrosination-detyrosination of a -tubulin Biophys. Res. Commun. 167, 1316-1325. Gundersen, G. G., Kalnoski, M. H. and Bulinski, J. C. (1984). Distinct that is involved in stabilizing microtubules (Gundersen et populations of microtubules: tryosinated and nontyrosinated alpha al., 1984). tubulin are distributed differently in vivo. Cell 38, 779-789. Finally, IF glycosylation, phosphorylation or both may Inagaki, M., Gonda, Y., Matsuyama, M., Nishizawa, K., Nishi, Y. and indeed play an important role in solubility. For example, Sato, C. (1988). Intermediate filament reconstitution in vitro. J. Biol. either post-translational modification may shift the solubil- Chem. 263, 5970-5978. Inagaki, M., Nishi, Y., Nishizawa, K., Matsuyama, M. and Sato, C. ity equilibrium constant, thereby resulting in a new state (1987). Site-specific phosphorylation induces disassembly of vimentin with enhanced overall solubility as seen in colcemid- filaments in vitro. Nature 328, 649-652. induced mitotic arrest. This could account for the similar Isaacs, W. B., Cook, R. K., Van Atta, J. C., Redmond, C. M. and Fulton, phosphorylation and glycosylation levels, and tryptic A. B. (1989). Assembly of vimentin in cultured cells varies with cell type. peptide profiles, of K8/18 in the S and P fractions. The clear J. Biol. Chem. 264, 17953-17960. Laemmli, U. K. (1970). Cleavage of structural proteins during the assembly identification of a substantial K8/18 soluble fraction should of the head of bacteriophage T4. Nature 227, 680-685. make addressing the possibilities raised above feasible. Lamb, N. J. C., Fernandez, A., Feramisco, J. R. and Welch, W. J. (1989). Modulation of vimentin containing intermediate filament distribution and We are very grateful to Dr Pierre Coulombe for his advice on phosphorylation in living fibroblasts by the cAMP-dependent protein intermediate filament cross-linking and electron microscopy, kinase. J. Cell Biol. 108, 2409-2422. Nafisa Ghori for assistance with the electron microscopy and neg- Lane, E. B., Goodman, S. L. and Trejdosiewicz, L. K. (1982). Disruption ative staining, Beth Herburger and Janet Rodgers for preparing of the keratin filament network during epithelial cell division. EMBO J. 1, 1365-1372. the manuscript. This work was supported by a Veterans Admin- Laskey, R. A. and Mills, A. D. (1977). Enhanced autoradiographic istration Career Development Award and Merit Award, and the detection of 32P and 125I using intensifying screens and hypersensitized PEW Scholars Program (M.B.O.), and a Digestive Disease Center film. FEBS Lett. 82, 314-316. Grant (DK 38707). Lazarides, E. (1982). Intermediate filaments: A chemically heterogeneous, 444 C.-F. Chou and others

developmentally regulated class of proteins. Annu. Rev. Biochem. 51, Söellner, P., Quinlan, R. A. and Franke, W. W. (1985). Identification of a 219-250. distinct soluble subunit of an intermediate filament protein: Tetrameric Miller, R. K., Vikstrom, K. and Goldman, R. D. (1991). Keratin vimentin from living cells. Proc. Nat. Acad. Sci. USA 82, 7929-7933. incorporation into intermediate filament networks is a rapid process. J. Steinert, P., Idler, W. and Zimmerman, S. B. (1976). Self-assembly of Cell Biol. 113, 843-855. bovine epidermal keratin filaments in vitro. J. Mol. Biol. 108, 547-567. Moll, R., Franke, W. W., Schiller, D. L., Geiger, B. and Krepler, R. Steinert, P. M. and Liem, R. K. H. (1990). Intermediate filament (1982). The catalog of human cytokeratins: Patterns of expression in dynamics. Cell 60, 521-523. normal epithelia, tumors and cultured cells. Cell 31, 11-24. Steinert, P. M. and Roop, D. R. (1988). Molecular and cellular biology of Ngai, J., Coleman, T. R. and Lazarides, E. (1990). Localization of newly intermediate filaments. Annu. Rev. Biochem. 57, 593-625. synthesized vimentin subunits reveals a novel mechanism of intermediate Steinert, P., Zackroff, R., Aynardi-Whitman, M., and Goldman, R.D. filament assemby. Cell 60, 415-427. (1982). Isolation and characterization of intermediate filaments. Meth. Nixon, R. A., Lewis, S. E., Dahl, D., Marotta, C. A. and Drager, U. C. Cell Biol. 24, 399-419. (1989). Early posttranslational modifications of the three neurofilament Tölle, H.-G., Weber, K. and Osborn, M. (1985). Microinjection of subunits in mouse retinal ganglion cells: neuronal sites and time course in monoclonal antibodies specific for one intermediate filament protein in relation to subunit polymerization and axonal transport. Mol. Brain Res. cells containing multiple keratins allow insight into the composition of 5, 93-108. particular 10 nm filaments. Eur. J. Cell Biol. 38, 234-244. Omary, M. B., Baxter, G. T., Chou, C.-F., Riopel, C. L., Lin, W. Y. and Tölle, H.-G., Weber, K. and Osborn, M. (1987). Keratin filament Strulovici, B. (1992). PKCe-related kinase associates with and disruption in interphase and mitotic cells - how is it induced? Eur. J. Cell phosphorylates cytokeratin 8 and 18. J. Cell Biol. 117, 583-593. Biol. 43, 35-47. Quinlan, R. A., Cohlberg, J. A., Schiller, D. L., Hatzfeld, M. and Torres, C.-R. and Hart, G. W. (1984). Topography and polypeptide Franke, W. W. (1984). Heterotypic tetramer (A2D2) complexes of non- distribution of terminal N-acetylglucosamine residues on the surfaces of epidermal keratins isolated from cytoskeletons of rat hepatocytes and intact lymphocytes. J. Biol. Chem. 259, 3308-3317. hepatoma cells. J. Mol. Biol. 178, 365-388. Traub, P., Nelson, J., Kühn, S. and Vorgias, C. E. (1983). The interaction Safer, D., Golla, R. and Nachmias, V. T. (1990). Isolation of a 5-kilodalton in vitro of the intermediate filament protein vimentin with naturally actin-sequestering peptide from human blood platelets. Proc. Nat. Acad. occurring RNAs and DNAs. J. Biol. Chem. 258, 1456-1466. Sci. USA 87, 2536-2540. Yano, T., Tokui, T., Nishi, Y., Nishizawa, K., Shibata, M., Kikuchi, K., Sahyoun, N., Stenbuck, P., LeVine III, H., Bronson, D., Moncharmont, Tsuiki, S., Yamauchi, T. and Inagaki, M. (1991). Phosphorylation of B., Henderson, C. and Cuatrecasas, P. (1982). Formation and keratin intermediate filaments by protein kinase C, by calmodulin- identification of cytoskeletal components from liver cytosolic precursors. dependent protein kinase and by cAMP-dependent protein kinase. Eur. J. Proc. Nat. Acad. Sci. USA 79, 7341-7345. Biochem. 197, 281-290. Shariff, A. and Luna, E. J. (1992). Diacylglycerol-stimulated formation of actin nucleation sites at plasma membranes. Science 256, 245-247. Skalli, O. and Goldman, R. D. (1991). Recent insights into the assembly, dynamics, and function of intermediate filament networks. Cell Motil. (Received 16 November 1992 - Accepted, in revised form, Cytoskel. 19, 67-79. 1 March 1993)