INTRODUCTION

The keratin intermediate filament (IF)¹ proteins make up a large family that is represented by more than 20 unique gene products (catalogued as K1-K20) which are specifically expressed in epithelial tissues (for review, see Refs. 1-3). Among cytoplasmic IF proteins, keratins and neurofilaments are unique in that they exist in cells as obligate heteropolymers. All epithelial cells express at least one type I (i.e. one or more of K9-K20) and one type II (i.e. one or more of K1-K8) keratins as obligate noncovalent, epithelial tissue-specific, heteropolymers that coalesce to form typical 10-nm filaments. For example, glandular and secretory epithelia (also called simple type) express K8/18 as their major keratins with variable levels of K19 and K20 (1, 4-8). Similar to all other IF proteins, keratins have the characteristic structural feature of a central 310-350-amino acid coiled coil -helical domain that is flanked by NH₂- and COOH-terminal globular head and tail domains, respectively (for review, see Refs. 3 and 9). As a group, IF proteins are considered relatively insoluble, and this is particularly so for keratins. However, in the case of K8/18, ~5% of the K8/18 pool in asynchronously growing human cultured HT29 colonic epithelial cells consists of soluble tetramers (10).

Although the function of cytoplasmic IF proteins, including keratins, remains poorly understood, their importance in human disease is accumulating at a rapid pace. Mutations in 11 of the 20 keratins (K5/14, K1/10, K4/13, K2/9, K6/16, K17) have already been identified as the cause of several tissue-specific human diseases (for review, see Refs. 11-14). In addition, a mutation in K18 was described recently in a patient with cryptogenic cirrhosis (68). One approach that we have taken toward understanding the function of K8/18 is a detailed characterization of their post-translational modifications to understand their regulation and to provide a handle on their function. The post-translational modifications of K8/18 which have been studied are glycosylation (15, 16) and phosphorylation (17-22; for review, see Ref. 23). These modifications occur in the NH₂- and COOH-terminal globular regions of the keratins (23) (termed head and tail domains, respectively, for all IF proteins), which are the regions that impart most of the structural heterogeneity to IF proteins (1-3). With regard to K8/18 phosphorylation, it occurs on serine residues (17) and increases in association with mitosis (18, 22, 24) or under a variety of stress conditions (25, 26). The majority of K8/18 molecules that are phosphorylated are not glycosylated, and vice versa, which suggests that each modification plays a separate role (27). Examples of emerging functions of human K8/18 phosphorylation (for review, see Ref. 23) include: (i) filament reorganization as determined by identification then mutation of a major and highly dynamic physiologic phosphorylation site of human K18 (21); (ii) increasing K8/18 solubility in vitro (19, 25) and in vivo (28); (iii) possibly determining the localization of keratins within specific subcellular domains (22); (iv) regulating the association of K8/18 with the 14-3-3 family of proteins during cell cycle progression, and by doing so, further increasing K8/18 solubility (28); and (v) potentially protecting cells during physiologic stress (25, 26).

The goal of this study was to characterize the major in vivo phosphorylation sites of human K8 and to begin addressing their functional significance. Previous studies showed that under stimulatory conditions (29-31) such as incubation with epidermal growth factor (EGF) or pro-urokinase, phosphorylation of type II K8 increases preferentially compared with type I K18. Biochemical cleavage at tryptophan residues indicated that human K8 phosphorylation involves not only the head domain (as found for K18; Ref. 27) but the head and other non-head domain sites (27, 32). In this study, we used manual Edman degradation of ³²PO₄-labeled then protease-digested K8 peptides to localize biochemically two major phosphorylation sites of human K8. A mutational approach was then used to confirm Ser-23 and Ser-431 as two major K8 in vivo phosphorylation sites within the head and tail domains, respectively. Since Ser-23 is highly conserved in many type II keratins, we used a mutational approach to demonstrate that K6, an epidermal type II keratin, is also phosphorylated at the corresponding serine site. We also tested the in vitro phosphorylation of Ser-431 by MAP and cdc2 kinases and showed that they are likely to play an in vivo role based on the increased phosphorylation of K8 Ser-431 upon EGF stimulation or mitotic arrest. In addition, a monoclonal antibody that specifically recognizes phosphoserine 431-K8 was generated and used to examine EGF-stimulated cells.

MATERIALS AND METHODS

HT29 (human colon), NIH-3T3 (mouse fibroblast), BHK-21 (hamster kidney) cells, and human placental K8 cDNA were obtained from the American Type Culture Collection. Cells were cultured as recommended by the supplier. Monoclonal antibody (mAb) L2A1 was used, which recognizes human K18 in the context of the heteropolymer K8/18 (15). Other reagents used were: phosphoric acid (³²PO₄) and [-³²P]ATP (DuPont NEN), trypsin and chymotrypsin (Worthington Biochemical Corp.), MAP and cdc2 kinases (BioLabs, Beverly, MA); Transformer^TM mutagenesis kit (Clontech, Palo Alto, CA); and LipofectAMINE liposomes and EGF (Life Technologies, Inc.).

Site-directed mutagenesis to generate serine  alanine K8 mutants was done using a Transformer^TM kit. The cDNA for human K8 and K18 was used as templates as described (21). Mutations were confirmed by sequencing followed by subcloning into the pMRB101 mammalian expression vector, with a human cytomegalovirus promoter-directed expression. The cDNA for human K6e (33) was kindly provided by Dr. Pierre Coulombe (Johns Hopkins University). Both strands of the entire cDNA for human K8 were sequenced, and the sequence was deposited in GenBank (accession no. U76549[GenBank]). Further confirmation of the nucleotides corresponding to Ser-431 within the sequence ⁴³¹SPGLSY (not present in published human K8 sequences but present in GenBank sequence X98614[GenBank]) was obtained by sequencing the same region of an independently isolated human K8 clone kindly provided by Dr. Robert Oshima (Burnham Institute, La Jolla, CA). Transfection was done using LipofectAMINE as recommended by the manufacturer.

Transfected or nontransfected cells were labeled with ³²PO₄, by incubating in phosphate-free medium for 30 min followed by adding 250 µCi/ml ³²PO₄ (5 h, 37 °C). Labeled cells were solubilized (2 h, 4 °C) with 1% Empigen in phosphate-buffered saline containing 5 mM EDTA, 0.1 mM phenylmethylsulfonyl fluoride, 25 µg/ml aprotinin, 10 µM leupeptin, 10 µM pepstatin, 0.5 µg/ml okadaic acid (34). After pelleting to isolate the solubilized material, K8/18 were immunoprecipitated using Sepharose-conjugated L2A1 mAb. For K6/18 transfectants, high salt extraction of transfected cells to isolate the keratin fraction was done similarly, as described (10), except that the first extraction was done using 1% Empigen instead of 1% Triton X-100.

K6 and K8 (wild type or mutant) proteins were isolated using preparative SDS-polyacrylamide gel electrophoresis (PAGE) (35) of K8/18 immunoprecipitates or K6/18 high salt extracts, followed by cutting out of the keratin band (after brief staining with Coomassie Blue and destaining). Keratins were electroeluted from the gel strips and acetone precipitated. The dried keratin precipitates were digested with trypsin or chymotrypsin followed by routine peptide mapping and autoradiography (16, 36). For manual Edman degradation, cellulose-containing spots corresponding to ³²PO₄-labeled peptides were removed from the thin layer chromatography plates followed by water extraction of the peptides. After removal of water by lyophilization, the dried peptides were subjected to manual Edman degradation for 8-29 cycles as described (37). The material released after each cycle and the disks containing any residual undigested peptide were counted.

K8/18 immunoprecipitates were isolated from HT29 cells or transfected BHK cells. Immunoprecipitates were then washed in kinase buffer containing 50 mM Tris-HCl (pH 7.5), 10 mM MgCl₂, 1 mM dithiothreitol, 1 mM EGTA. Endogenous kinase activity was abolished by heating samples (90 °C, 1 min) in 15 µl of kinase buffer. After cooling, the kinase reaction was carried out by adding 5 µCi of [-³²P]ATP, 22 µM ATP, and 1 unit of cdc2 or MAP kinase (total reaction volume = 25 µl). After 10 min at 22 °C, 25 µl of 2 × sample buffer was added followed by boiling for 1 min, then analysis by analytical or preparative SDS-PAGE.

Monoclonal antibody 5B3 was generated after immunizing mice with K8/18 which was purified from okadaic acid-treated HT29 cells. The antibody is an IgG1 and recognizes phosphoserine 431 based on several criteria²: (i) it binds to phosphorylated K8 but not to nonphosphorylated K8 as determined by immunoblotting of isoelectric focused SDS-PAGE two-dimensional gels of K8/18 immunoprecipitates; (ii) it binds specifically to a peptide that contains the phosphoserine 431-K8 sequence motif but not to other unrelated peptides; and (iii) the data shown in Fig. 5. Immunofluorescence staining, including keratin and nuclear double staining, was done as described (22). Rabbit anti-K8/18 (antibody 8592) was described previously (67).

a

b

c

RESULTS

We showed previously that serine is the major phosphorylated residue of K8/18 in human HT29 cells (38). In addition, the two major tryptic phosphopeptides that are metabolically labeled with ³²PO₄ in HT29 cells (e.g. Fig. 1B, panel a) were also observed as two major K8 tryptic phosphopeptides in normal colonic human biopsies labeled ex vivo with ³²PO₄ (21). Our approach for identifying the phosphorylation sites in these two major K8 phosphopeptides was to: (i) use manual Edman degradation of phosphopeptides obtained by trypsin or chymotrypsin digestion of K8 isolated from ³²PO₄-labeled HT29 cells; (ii) generate K8 serine mutants at potential phosphoserines, with mutations chosen to match the manual sequencing results and the amino acid sequence of K8; and (iii) test if phosphorylation at a given tryptic or chymotryptic phosphopeptide is abolished in the mutants.

Toward addressing the above goals, we compared the phosphorylation of K8 in two transfected cell lines (NIH-3T3 and BHK) to phosphorylation of K8 in HT29 cells. As shown in Fig. 1A, K8 in BHK cells manifests a phosphorylation specific activity that is comparable to K8 in HT29 cells. In addition, there is significant overlap of the tryptic (T1, T2) and chymotryptic (C1-C4) phosphopeptides of K8 isolated from HT29 and BHK cells (Fig. 1B). Therefore, we used BHK cells for the transfection and characterization of K8 phosphorylation experiments as described below (K8 expressed in NIH-3T3 cells manifested a peptide map pattern similar to that of K8 isolated from HT29 cells; not shown).

We sequenced the cDNA for human K8 since assignment of specific peptides and potential serines depends on having a precise localization of Arg/Lys (for trypsin fragments), Phe/Tyr/Trp (for chymotrypsin fragments), and serines within the proteolytic fragments. The human K8 protein contains 61 potential serine phosphorylation sites (not shown). Minor differences were noted between our sequence (see "Materials and Methods") and two other available sequences, but one relevant difference was the presence of a serine (Ser-431) within the K8 peptide ⁴²⁶YGGLTSPG in contrast to the reported K8 sequences ⁴²⁶YGDLTDPG (39) and ⁴²⁶YGGSQAG (40). An identical ⁴²⁶YGGLTSPG sequence is found in mouse (41) and rat K8 (42) and in an independent human K8 cDNA clone (not shown); and notably, this sequence contains a serine that is in the fifth position after cleavage of Tyr-426 by chymotrypsin (see below).

Each of the ³²PO₄-labeled peptides T1, T2, C1-C4 shown in Fig. 1B (panels c and f) were isolated followed by manual Edman degradation. As shown in Table IA, three of the peptides (T2, C1, and C4) afforded informative degradation with release of the ³²PO₄-serine at specific cycles, whereas a significant portion of the counts of the other three peptides (T1, C2, C3) remained within the filter and were noninformative. In the case of peptide T2 Edman degradation, analysis of the K8 sequence indicates that Ser-1, 8, 23, 133, 252, and 273 are potential phosphorylation sites since they immediately follow an adjacent Arg/Lys. To determine the relationship between T2 and C1/C4, we trypsin digested individually purified C1 and C4 peptides (Table IB), which resulted in release of radioactivity of the C4 peptide at the first cycle. This, coupled with the known K8 sequence, indicates that T2 and C4 are related and that phosphorylation of T2/C4 likely occurs on Ser-23 in ¹⁹FSSRSY. Using a similar analysis, the results of Edman degradation of C1 (Table IA) and Edman degradation of trypsin-digested C1 (Table IB) suggest that Ser-290, 423, 431, and/or 441 may be phosphorylated (Table II). Although phosphorylation of Ser-14, 42 and 456 (Table II) may correspond to chymotryptic peptide C1 (i.e. release at fifth cycle), tryptic digestion of C1 followed by Edman degradation (Table IB) excludes phosphorylation at these sites. This is due to the presence of trypsin-sensitive Lys/Arg (arrows, bottom three peptides, Table II), which would be expected to change the cycle number of radioactivity release after Edman degradation of peptide C1 from the observed fifth cycle (Table IB) to a first or second cycle (Table II). The reason for the resistance of T1, C2, C3, or chymotrypsin-digested T1 peptides to Edman degradation (Table I) is unclear, but it may reflect cyclization with inaccessibility of the peptide and/or phosphorylation at a distant serine residue that is inefficiently cleaved after more than 15 Edman degradation cycles (43).

Table I. Manual Edman degradation of tryptic and chymotryptic 32PO4-labeled K8 peptides Panel A, schematic of the tryptic (T1, T2) and chymotryptic (C1-C4) peptides of K8 which were analyzed by Edman degradation. The peptides highlighted in the schematic were extracted from several cellulose plates, counted (initial cpm), then coupled to a filter and subjected to the indicated number of Edman degradation cycles (total cycles). Also shown are the cycles where the peak cpms were obtained, the cpm range for the remaining cycles other than the peak cycle, and the remaining cpm in the filter after completion of the indicated number of cycles. Each peak cpm (e.g. 7531 for peptide T2) was significantly greater than the cpm range for the other cycles and the counts remaining in the filter (e.g. 43-151 and 184, respectively, for T2). In contrast, most of the initial counts remained in the filter in cases when no peak was obtained (e.g. T1, C2, C3). Panel B, individually isolated peptides T1, C1, and C4 were redigested with the indicated protease followed by analysis by Edman degradation as in panel A.

Table II. Location, sequence of potential K8 phosphopeptides, and summary of peptide mapping results of K8 mutant constructs The location (head, rod, tail) of potential K8 phosphopeptides is shown. Single-letter abbreviations are used to show the peptide sequences. Dotted slash and arrows indicate positions of chymotrypsin and trypsin digestion sites, respectively. Asterisks indicate the potential phosphorylated serines that were mutated to alanines in the top five peptides. Individual mutant (S  A) K8 constructs were then cotransfected with wild type K18 into BHK cells, followed by metabolic labeling with 32PO4, then peptide mapping, as described in the Fig. 2 legend, to generate the T1 and T2 peptides (trypsin digestion) or C1-C4 peptides (chymotrypsin digestion). + and  correspond to the presence or absence, respectively, of the indicated phosphopeptide. N.A., not applicable. Serine numbers highlighted in bold indicate the confirmed K8 phosphorylation sites as determined by the disappearance of a tryptic and/or chymotryptic peptide.

Given the above biochemical analysis, we generated several serine mutant human K8 constructs to confirm the phosphorylation sites in peptides C4 (or the related T2) and C1. This was done by cotransfection of wild type K18 with mutant or wild type K8 into BHK cells followed by metabolic labeling with ³²PO₄ and chymotryptic peptide mapping of isolated K8. All of the mutants resulted in a "wild type" pattern (Table II) similar to that in Fig. 1, except for the Ser-23 mutant, which showed an absence of ³²PO₄-labeled peptides C4 (Fig. 2B, panels a and b), and T2 (Table II, peptide map not shown), and the Ser-431 mutant, which showed an absence of ³²PO₄-labeled peptide C1 (Fig. 2B, panels c and d). In addition, the S23A/S431A double mutant showed absence of both peptides C1 and C4 as expected (Fig. 2B, panels e and f). Only a partial decrease in the overall phosphorylation of the double mutant was noted (Fig. 2A) secondary to the presence of several other K8 peptides that are radiolabeled in BHK cells.

Analysis of head domain sequences of type II keratins (K1-K8) shows a striking degree of conservation of Ser-23 among all type II keratins (Table III, shaded residues). This led us to test if phosphorylation of this serine is conserved in other type II keratins. For this, we tested whether the equivalent serine in K6e (i.e. Ser-59) is phosphorylated, by comparing the tryptic peptide maps of wild type and Ser-59  Ala (S59A) K6e after transfection into BHK cells and then labeling with ³²PO₄. As shown in Fig. 3a, high levels of K6/18 can be coexpressed in transfected BHK cells, and tryptic peptide mapping analysis indicates that the Ser-59 of K6 is phosphorylated in vivo in BHK cells (Fig. 3, b-d). Mutation of Ser-59 of K6e resulted in the absence of a labeled phosphopeptide (Fig. 3, panel c), which was present in wild type K6 (Fig. 3, panel b) and in the mix map (Fig. 3, panel d; peptide indicated by an arrow). Similar results were obtained, in terms of the disappearance of a labeled phosphopeptide for the S59A K6e mutant, after chymotryptic peptide mapping (not shown). Although phosphorylation of the equivalent serine residue in other type II keratins remains to be confirmed, its observed phosphorylation in K6 and K8 and the conservation of its motif suggest that other type II keratins are likely to be phosphorylated at the corresponding site.

Table III. The senine 23-containing motif of K8 is conserved among type II keratin head domains The Table shows the serine 23-containing sequence motif of human (H-) K8 and corresponding sequences for Xenopus (X-), mouse (M-), rat (R-), and other type II keratins. Dots indicate amino acid residues that are identical to the corresponding H-K8 residues. The shaded serines correspond to the conserved serine 23 of K8. The table shows keratin sequences for the type II keratins: H-K8 (39, 40), X-K8 (44), M, R-K8 (41, 42), H-K1 (45), H-K2e (46), H-K2p (47), H-K3 (48), H-K4 (49), H-K5 (50), H-K6a-f (33, 51), and H-K7 (52).

Phosphopeptide mapping of wild type (WT) and Ser-59  Ala human K6e.

Ser-431 is proximal to proline within the sequence ⁴²⁹LTSPGL, which suggests that one or more proline-directed kinases, such as MAP and cdc2 kinases, is/are likely to phosphorylate this residue. Phosphorylation of K8 Ser-431 by MAP kinase is an attractive hypothesis given that EGF, a MAP kinase-simulating growth factor (53), enhanced the phosphorylation of K8 in rat hepatocytes (29). Similarly, cdc2 kinase phosphorylates vimentin (54, 55) and glial fibrillary acidic protein (56) during mitosis. To test this hypothesis, we examined the ability of K8 to serve as a substrate to MAP and cdc2 kinases in vitro. As shown in Fig. 4A, both kinases phosphorylate K8, and the site of phosphorylation is restricted to Ser-431 based on two lines of evidence. First, in vitro phosphorylation of wild type K8 or K8 S23A by MAP kinase (Fig. 4D) or cdc2 (not shown, but pattern identical to that of MAP kinase shown in Fig. 4D) is abolished if K8 S431A is used as a substrate (in the context of a K8/18 immunoprecipitate). Second, in vitro phosphorylation of wild type K8 by MAP and cdc2 kinases (Fig. 4E, panels a-d) results in selective phosphorylation of the Ser-431-containing K8 peptide.

To assess the physiologic significance of in vitro phosphorylation of Ser-431 by MAP and cdc2 kinases, we compared the chymotryptic phosphopeptide maps of basally phosphorylated K8 in HT29 cells that are asynchronously growing (not shown; but similar to Fig. 1B, panel d), serum-starved (Fig. 4E, panel b), serum-starved then EGF stimulated (Fig. 4E, panel e) or arrested at G₂/M (Fig. 4E, panel f). The basal low level phosphorylation of Ser-431 increased significantly after EGF stimulation and during mitotic arrest (Fig. 4E, compare panels e and f with b) in association with the overall increase in K8 phosphorylation upon EGF stimulation (Fig. 4B) or mitotic arrest (Fig. 4C).

Phosphorylation of Ser-431 was also assessed using mAb 5B3, which was generated after immunizing mice with hyperphosphorylated keratins. This antibody specifically recognizes the phosphoserine 431-containing motif (see "Materials and Methods"), and its reactivity is abolished upon mutation of K8 Ser-431 (Fig. 5A). Given the results shown in Fig. 4, B and E, we examined K8/18 immunoprecipitates (obtained using mAb L2A1, which recognizes phosphorylated and nonphosphorylated K8/18) by immunoblotting using mAb 5B3. As shown in Fig. 5B, EGF stimulation of HT29 cells results in increased binding of mAb 5B3 as would be expected based on Ser-431 phosphorylation. Immunofluorescence staining of HT29 cells with mAb 5B3, before and after EGF treatment, showed reorganization of the keratin filaments that are recognized by the antibody (Fig. 5C, compare panels a and b). This reorganization was manifested by a ring-like cortical enhancement of filament staining near the cell periphery. This filament reorganization is not related to a mitotic event as determined by nuclear double staining (Fig. 5C) and appears to overlap with the overall keratin staining that was visualized using rabbit antibody 8592, which recognizes K8 and K18 (Fig. 5C, compare panels c and c).

DISCUSSION

The major findings of this study are as follows. (i) Two major in vivo phosphorylation sites were identified in human K8, one in the head domain (Ser-23) and one in the tail domain (Ser-431); (ii) The K8 head domain phosphorylation site, which is highly conserved among all type II keratins, is also phosphorylated in K6e. Hence, this conserved head domain serine is likely to be a common phosphorylation site among all type II keratins. (iii) Phosphorylation of K8 tail-domain Ser-431 is likely to involve MAP and cdc2 kinases during growth factor stimulation and mitosis, respectively, based on in vitro and in vivo evidence.

In characterizing K8 phosphorylation, we utilized a combined approach of manual Edman degradation of in vivo ³²PO₄-labeled K8 and then protease-generated peptides, sequence information, and a molecular approach to confirm biochemically suggested phosphorylation sites. The protease utilized was important in that one of two major trypsin-generated phosphopeptides consists of at least three species (i.e. C1-C3) after chymotrypsin digestion (Tables I and II). The generation of C1-C3 peptides appears possible only upon chymotrypsin digestion of the intact K8 protein (e.g. Table IA) but not after chymotrypsin digestion of tryptic peptide T1 (Table IB). Manual Edman degradation was not effective in releasing the phosphoamino acid for C2 and C3 peptides and did not aid in identifying these two phosphorylation sites. The in vitro phosphorylation of K8 by MAP and cdc2 kinases exclusively at the C1 peptide indicates that C2 and C3 are more likely to be unique phosphorylation sites rather than partially digested peptides that are related to C1.

The kinase(s) that phosphorylates Ser-23 of K8 in vivo is(are) not known, and the amino acid sequence context of phosphoserine 23 does not not provide a clear structural motif for a specific kinase. Several in vitro phosphorylation sites were recently identified biochemically after phosphorylating purified K8 with cAMP-dependent protein kinase (protein kinase A) (57). One of the sites corresponds to Ser-23, which raises the possibility that protein kinase A may be a relevant physiologic kinase for Ser-23 of K8. To that end, phosphorylation of K8/18 increased in cultured human intestinal Caco-2 cells upon treatment with forskolin (58). However, activation of protein kinase A by incubating HT29 cells with 8-bromo-cAMP or forskolin did not increase K8 phosphorylation (38), and incubation of rat hepatocytes (59) or dog thyroid follicle cells (60) with dibutyryl-cAMP did not increase keratin phosphorylation. In addition, phosphorylation of purified rat K8 by protein kinase A results in the phosphorylation of several sites (19, 57) so that the pertinent kinase(s) involved in Ser-23 phosphorylation remains to be investigated. In the case of Ser-431 phosphorylation, both cdc2 and MAP kinases appear to be relevant physiologic kinases as supported by the consensus sequence of Ser-431, by the specific enhancement of Ser-431 phosphorylation upon EGF stimulation or mitotic arrest, and the phosphorylation by these kinases exclusively on Ser-431.

Phosphorylation of K8 at Ser-23 and Ser-431 indicates that both the head and tail domains of K8 are phosphorylated as shown previously for K1 (61), another type II keratin. In the case of type I keratins, the only studied example is K18, which appears to have its phosphorylation preferentially in the head domain (21, 27). The phosphorylation of type I and type II keratins appears to be independent in that mutation of a major phosphorylation site of K18 does not affect K8 phosphorylation (21), and mutation of Ser-23/431 K8 does not affect overall K18 phosphorylation or the tryptic phosphopeptide pattern of K18 phosphorylation (not shown).

Both Ser-23 of K8 and Ser-59 of K6e are contained within the highly conserved motif SRX (underlined serine is phosphorylated, and X represents an aliphatic/aromatic residue), which is found within all type II keratins and is conserved across species except for Xenopus (Table III). The in vivo phosphorylation of the serine within this motif in K8 and K6e suggests that the equivalent serines in other type II keratins are also likely to be physiologically phosphorylated. If phosphorylation at this highly conserved site is to have any physiologic importance, then it is likely that such a role will involve a function that is common to all type II keratins. Interestingly, the domain containing this phosphorylation site is implicated, in vitro, to be important for desmoplakin binding (62). However, it remains unclear if all type II keratins bind directly to desmoplakin. For example, despite colocalization of keratins with desmoplakin (63) in a manner that is regulated negatively by phosphorylation of a COOH-terminal site of desmoplakin (64), desmoplakin did not bind to K8 using an in vitro overly assay under conditions that other type II keratins bound (62). In exploring this question, desmoplakin does not coimmunoprecipitate with K8/18 after solubilizing cells with nonionic detergents (most of desmoplakin remains insoluble under these conditions in HT29 cells) or with the zwitterionic detergent Empigen (not shown), which solubilizes a significant component of the cytoskeletal/filamentous K8/18 (28, 34). However, this does not exclude a potential weak interaction between desmoplakin and K8 which becomes abolished after Empigen solubilization.

With regard to Ser-431, its phosphorylation increases upon stimulation of cells with EGF (Fig. 4), which suggests a role in mitogen-induced signaling. This site is conserved in K8 sequences of mouse (41) and rat (42), but not Xenopus (44). The ⁴²⁹LTSPG sequence of K8 is also not conserved in the tail domain of other type II keratins, which suggests a specific function for this site which is related to glandular epithelia. The significance of Ser-431 phosphorylation after EGF exposure to cells and its importance in regulating downstream events of EGF-receptor or other mitogen-induced activation remain to be determined. The known activation of MAP kinase by EGF and the exclusive phosphorylation of Ser-431 in vitro by MAP kinase suggest that several modes of MAP kinase activation maybe associated with an increase in K8 phosphorylation at Ser-431. The increase in K8 Ser-431 phosphorylation after EGF stimulation of HT29 cells (Fig. 4) supports the previously observed overall increase in K8 phosphorylation in EGF-stimulated rat hepatocytes (29). In addition, EGF stimulation of HT29 cells was associated with filament reorganization as determined using a mAb that is specific to phosphoserine 431-K8 (Fig. 5), as noted previously in cultured rat hepatocytes (29). However, the reorganized filaments were also recognized by an antibody that recognized the total keratin pool, which suggests that K8 Ser-431 phosphorylation after EGF stimulation may not be solely responsible for this reorganization. This conclusion is based on the observation that EGF stimulation of HT29 cells is also associated with increased K18 phosphorylation (Fig. 4B) and with phosphorylation of a K8 site that is different from Ser-23 or Ser-431 (not shown). Furthermore, staining with the 5B3 mAb shows a typical filamentous "nonreorganized" pattern under basal unstimulated conditions. This suggests that filament reorganization in vivo may involve several phosphorylation events on K8 and/or K18, with the stoichiometry of phosphorylation also being another important player.

Phosphorylation of Ser-431 also increased upon mitotic arrest (Fig. 4), but the biologic significance of this phosphorylation in terms of progression through mitosis or other mitosis-related events remains to be determined. Approaches such as the generation of antibodies that are unique to specific phosphoepitopes (for review, see Refs. 65 and 66) should provide additional functional insights into K8/18 phosphorylation and the relationship and sequence of events of the multiple K8 and K18 phosphorylation sites.