Keratin 8 Phosphorylation by p38 Kinase Regulates Cellular Keratin Filament Reorganization

Keratin 8 (K8) serine 73 occurs within a relatively conserved type II keratin motif (68NQSLLSPL) and becomes phosphorylated in cultured cells and organs during mitosis, cell stress, and apoptosis. Here we show that Ser-73 is exclusively phosphorylated in vitro by p38 mitogen-activated protein kinase. In cells, Ser-73 phosphorylation occurs in association with p38 kinase activation and is inhibited by SB203580 but not by PD98059. Transfection of K8 Ser-73 → Ala or K8 Ser-73 → Asp with K18 generates normal-appearing filaments. In contrast, exposure to okadaic acid results in keratin filament destabilization in cells expressing wild-type or Ser-73 → Asp K8, whereas Ser-73 → Ala K8-expressing cells maintain relatively stable filaments. p38 kinase associates with K8/18 immunoprecipitates and binds selectively with K8 using an in vitro overlay assay. Given that K1 Leu-160 → Pro (157NQSLLQPL →157NQSPLQPL) leads to epidermolytic hyperkeratosis, we tested and showed that the analogous K8 Leu-71 → Pro leads to K8 hyperphosphorylation by p38 kinase in vitro and in transfected cells, likely due to Ser-70 neo-phosphorylation, in association with significant keratin filament collapse upon cell exposure to okadaic acid. Hence, K8 Ser-73 is a physiologic phosphorylation site for p38 kinase, and its phosphorylation plays an important role in keratin filament reorganization. The Ser-73 → Ala-associated filament reorganization defect is rescued by a Ser-73 → Asp mutation. Also, disease-causing keratin mutations can modulate keratin phosphorylation and organization, which may affect disease pathogenesis.

The "soft" mucosal keratins (K) 1 make up the intermediate filament (IF) proteins that are preferentially expressed in epithelial cells that line the inner and outer surfaces of animal tissues. These mucosal keratins consist of a large family (at least 20 members termed K1 to K20) of cytoplasmic proteins that are divided into relatively acidic type I (K9 to K20, pI Ͻ 6) and relatively basic type II (K1 to K8, pI Ն 6) keratins (1)(2)(3)(4). Epithelial cells generally express two or more keratin noncovalent heteropolymers in a 1:1 molar ratio of type I to II IFs, with an epithelial cell type-specific unique keratin complement. For example, single layered "simple type" epithelia express K8 and K18, with variable levels of K19 and K20 depending on the cell type, whereas keratinocytes express K5/14 or K1/10 basally and suprabasally, respectively. The prototype structure of all IF proteins, including keratins, consists of a central coiled-coil ␣-helix domain termed the "rod" that is flanked by non-␣-helical N-terminal "head" and C-terminal "tail" domains (5,6). Notably, the head and tail domains of keratins contain most of the structural heterogeneity among IF proteins and also include the domains that undergo phosphorylation. This distribution correlation and other accumulating data (7)(8)(9)(10)(11) strongly suggest that phosphorylation plays an important role in regulating the tissue-specific functional roles of the large keratin family.
Although spectacular gains have been made in linking 14 of the more than 20 keratins to a number of skin, oral, esophageal, and liver diseases (12)(13)(14)(15)(16)(17), full appreciation of keratin and other IF protein function has been lagging. For some keratins, one clearly delineated function is to protect cells from mechanical and nonmechanical forms of injury, but how this occurs remains poorly understood (11,12,18,19). Regardless, an intact keratin filament network and how keratin filaments are organized appear to be important effectors of this ability to maintain cellular integrity. This is borne out by many in vitro studies that correlated the importance of various keratin domains to form typical-appearing filaments and by the phenotypes that have been observed in patients with keratin diseases and in animal models that express different keratin mutants (11, 13, 15-17, 19, 20). Although perturbations within the highly conserved proximal and distal ends of the rod domain (which harbor most of the described disease-causing keratin mutations but lack any evidence of phosphorylation (10,15)) have significant effects on filament organization in vivo and in vitro, keratin phosphorylation within the head and tail domains also plays a significant role in filament organization in vitro (8,21) and in vivo (9,10). In addition, keratin mutations within the head domains, which may modulate keratin phosphorylation, have been described. For example, mutations have been described that either introduce a new potential phosphorylation site (e.g. K1 157 NQSLLQP 3 157 NQSPLQP which renders Ser-159 a potential proline-directed kinase phosphorylation site (22)) or remove possible phosphorylation sites (e.g. Ref. 23).
Keratin phosphorylation has been most extensively studied in K8/18/19 (10), due in part to the relative solubility of these keratins as compared with epidermal keratins (24). These studies resulted in the identification of several phosphorylationmediated K8/18 functions. For example, K18 Ser-33 phosphorylation regulates keratin binding to the 14-3-3 family of proteins during mitosis, which in turn plays a role in keratin filament organization and solubility (25,26). A direct role for keratin phosphorylation may also occur, as noted for K19, whereby mutation of its major phosphorylation site (Ser- 35 3 Ala) altered keratin filament organization in transiently transfected cells (27). In addition, transgenic mouse studies showed that K18 Ser-52 phosphorylation facilitates a protective role against hepatotoxic injury (28), a finding that has provided direct evidence for a number of correlative data that document increased keratin phosphorylation in association with a variety of stresses in cultured cells and in intact animals (29). In the case of human K8, three major in vivo phosphorylation sites have been identified: Ser-23, Ser-431, and Ser-73. Ser-23 is a highly conserved site among all type II keratins, which suggests a common keratin function for this modification, whereas Ser-431 is a basally phosphorylated site that increases its phosphorylation specific activity during mitosis and upon exposure to epidermal growth factor in association with filament reorganization (30). In contrast K8 Ser-73 phosphorylation behaves like an on/off switch in cultured cells and in tissues, with phosphorylation being "on" during mitosis, a variety of cell stresses including heat and drug exposure, and during apoptosis (31). Although the function of K8 Ser-73 phosphorylation was unknown, our hypothesis prior to embarking on this study was that its phosphorylation is likely to be important due to its on/off property and its association with important cell processes. Here we show that the mitogen-activated protein kinase (MAPK) p38 (reviewed in Refs. 32-35) is a physiologic kinase for K8 Ser-73 phosphorylation, and we demonstrate that K8 Ser-73 phosphorylation plays a significant role in keratin filament reorganization in response to the phosphatase inhibitor okadaic acid. Since K8 Ser-73 is proximal to a human disease mutation site in epidermal K1 (NQSLLQPL 3 NQSPLQPL, Ref. 22; with K8 Ser-73 being part of the motif 68 NQSLLSPL of K8), we generated the equivalent K1 mutation in K8 (i.e. NQS-LLSPL 3 NQSPLSPL) and showed that it increased K8 phosphorylation, as compared with wild-type K8. This skin diseasecausing mutation also resulted in significant keratin filament collapse in the presence of okadaic acid. Therefore, K8 Ser-73 phosphorylation plays an important role in modulating keratin filament reorganization. In addition, this is the first demonstration that human keratin disease-causing mutations can indeed result in keratin hyperphosphorylation and that such hyperphosphorylation can affect keratin filament organization, which in turn may contribute to disease pathogenesis.
Cell Culture-HT-29 (human colon), BHK (hamster kidney), and NIH-3T3 (mouse fibroblast) cells were obtained from the American Type Culture Collection (Manassas, VA) and cultured as recommended by the supplier. To activate p38 kinase, cells were incubated with An (10 g/ml, 0 -20 h) or with MMS (0.1 or 1 mg/ml, 0 -24 h). Cells were then solubilized with 2% SDS-containing sample buffer (37) followed by shearing of the DNA with a 27-gauge needle and then boiling for 2 min to generate a total cell lysate. Alternatively, cells were processed for immunoprecipitation as described below. For the kinase inhibitors SB203580 (p38 kinase) and PD98059 (MAPK kinase), cells were preincubated with these compounds (20 and 100 M, respectively) for 1 h and then treated with An for 2 h.
Immunofluorescence Staining-Transiently transfected cells were grown on coverslips and fixed 3 days after transfection, using 100% methanol (Ϫ20°C) for 3 min. Staining was done as described (26). For okadaic acid (OA) treatment, OA (1 g/ml) was added to the transfected cells for 2 h before fixation and processing. Fluorescence was analyzed using a Bio-Rad MRC1024 confocal laser scanning and a Nikon TE300 inverted microscope. Cells co-transfected with WT K18 and one of the four K8 constructs (WT, S73A, S73D, or L71P) were scored, after treatment with OA, based on their filament organization as follows: (i) cells with residual filaments, (ii) cells with fine dots but without any residual filaments, and (iii) cells with large dots.
Cell Transfection and cDNA Constructs-The K8 mutants K8 Ser-73 3 Ala (S73A), S73D, and L71P were generated using a Transformer TM mutagenesis kit (CLONTECH Laboratories Inc., Palo Alto, CA) as recommended by the supplier. Wild-type (WT) K8, WT K18, or mutant K8 cDNAs were subcloned into the pMRB101 mammalian expression vector under control of the hCMV promoter. The FLAG-tagged ␣-isoform of WT p38 or p38 AF (kinase-inactive form due to double mutation at the phosphorylation sites, T180A and Y182F; Ref. 38) were used to overexpress the p38 proteins in BHK cells with keratin constructs. Transient transfections into NIH-3T3 or BHK cells were done using LipofectAMINE as recommended by the supplier. The NIH-3T3 cells were used for immunofluorescence experiments because they provided a well formed keratin filament-staining pattern, whereas BHK cells were used to generate keratins for the biochemical experiments since they had a higher transfection efficiency.
Biochemical Methods-Immunoprecipitation was carried out by solubilizing cells with 1% Emp (1 h, 4°C) in buffer A (phosphate-buffered saline (PBS) (pH 7.4) containing 5 mM EDTA, 0.1 mM phenylmethanesulfonyl fluoride, 10 M pepstatin, 10 M leupeptin, 25 g/ml aprotinin, and 1 g/ml OA) or by solubilizing cells with 1% Nonidet P-40 in buffer A. After pelleting (15 min; 16,000 ϫ g), keratins were immunoprecipitated from the supernatant using Sepharose-conjugated L2A1 followed by washing, analysis by SDS-PAGE (37), and then staining with Coomassie Blue. For immunoblotting, gels were transferred to membranes followed by blotting (39) with individual anti-keratin antibodies. Bound antibodies were visualized with peroxidase-conjugated secondary antibodies and enhanced chemiluminescence. Two-dimensional chymotryptic phosphopeptide mapping was carried out exactly as described (30,40) using electrophoresis in the first (horizontal) dimension and chromatography in the second (vertical) dimension.
The overlay assay was performed as described (41) with minor modifications. Briefly, total lysate and K8/18 immunoprecipitates from HT-29 cells were analyzed using SDS-PAGE followed by transfer to a polyvinylidene difluoride membrane (4°C). The membrane was blocked with 3% BSA in PBS for 2 days, followed by incubation with 1 g/ml p38 kinase in PBS with 0.05% Tween and 0.1% BSA for 2 h (22°C). After washing, the membrane was incubated with anti-p38 antibody for immunoblotting.
In Vivo and in Vitro 32 P Labeling-In vitro kinase reactions were carried out using K8/18 immunoprecipitates. For each of the kinases used (p38, p42, and Jun kinases), the buffers provided by the supplier were used as recommended. Immunoprecipitates of K8/18 were washed two times with the respective kinase buffer (in addition to the routine washings as part of immunoprecipitation) and then incubated with 5 Ci of [␥-32 P]ATP, the kinase, and 20 M ATP (10 min in a total volume of 25 l). The kinase reaction was quenched by adding 4 times the normal concentration of Laemmli sample buffer, followed by boiling for 90 s and then analysis by SDS-PAGE and autoradiography. Metabolic labeling with [ 32 P]orthophosphate was done by incubating cells (in 100-mm dishes) with 5 ml of phosphate-free Dulbecco's modified Eagle's medium containing 10% dialyzed fetal calf serum and 100 mM glutamine for 30 min followed by the addition of 50 l of normal medium and 250 Ci/ml [ 32 P]orthophosphate. After labeling for 5 h, keratins were immunoprecipitated from the detergent-solubilized cells using mAb L2A1 and then analyzed by preparative SDS-PAGE and Coomassie staining, followed by isolation of the individual keratin-stained bands for peptide mapping.

Examination of K8 Ser-73 Phosphorylation by Mutational
Analysis and by in Vitro Phosphorylation-We identified previously (31) K8 Ser-73 as a K8 phosphorylation site using what we termed a "reverse immunologic" approach. This was aided by an antibody termed LJ4, which was generated by immunizing mice with keratins that were purified from okadaic acidtreated HT-29 cells. As shown previously (31) and exemplified in Fig. 1A, mAb LJ4 selectively recognizes the hyperphosphorylated and slightly slower migrating K8 species termed HK8. The HK8 species are present in very small amounts in exponentially growing HT-29 cells as determined by immunoprecipitation with mAb L2A1, which recognizes the entire keratin pool (31), but become markedly enriched after immunoprecipitation with mAb LJ4. The LJ4 Ab recognizes HK8 exclusively (Fig. 1B, lane 1), and its reactivity is abolished if Ser-73 is mutated to an alanine (S73A) (Fig. 1B, lane 2). However, LJ4 does recognize K8 S73D weakly (Fig. 1B, lane 3), which migrates slightly faster than HK8 and a bit slower than K8, such that LJ4 has almost equal binding intensity to the barely visible Coomassie-stained HK8 as compared with the strongly staining K8 S73D species (Fig. 1B).
We compared the in vitro phosphorylation of K8 by the proline-directed MAPKs p38 and p42, given the sequence context of K8 Ser-73 ( 71 LLSPL) and the previous observation (31) that K8 Ser-73 becomes phosphorylated during heat stress and apoptosis. As shown in Fig. 1C, p38 kinase generates the radiolabeled HK8 species exclusively (a signature of Ser-73 phosphorylation), whereas p42 kinase generates phosphorylated K8 and HK8 (K8 Ͼ HK8; compare lanes 2 and 3). Mutation of the two major K8 phosphorylation sites, Ser-23 and Ser-431 (30), did not affect formation of HK8 upon in vitro phosphorylation of K8/18 precipitates with p38 kinase (Fig. 2A, lanes 2 and 4). In contrast, mutation of K8 Ser-73 abolished formation of the HK8 species and resulted in barely detectable K8 phosphorylation ( Fig. 2A, lane 3) that is likely due to Ser-431 phosphorylation (the only other K8 potential proline-directed kinase site, with the sequence 429 LTSPG). The specificity of p38 kinase toward K8 Ser-73 is evident by the minimal formation of HK8 by p42 kinase (Fig. 2B) and the nearly equal generation of phospho-K8 and HK8 species by JNK (Fig. 2C). K8 Ser-23, which is a major basally phosphorylated K8 site (30), is not phosphorylated in vitro by any of the three tested MAPKs, whereas K8 Ser-431 phosphorylation occurs by p42 and JNK but not by p38 kinase (Fig. 2). Hence, the in vitro kinase assays of WT and mutant K8 immunoprecipitates indicate that both JNK and p42 phosphorylate K8 Ser-431 and Ser-73 relatively promiscuously, albeit to varied levels, in marked contrast to the selectivity of p38 kinase to the K8 Ser-73 site. In addition, phosphorylation of K8 Ser-73 does not appear to impact on K8 Ser-431 phosphorylation and vice versa.
Evidence of in Vivo K8 Ser-73 Phosphorylation by a p38-like Kinase-Given the findings in Figs. 1 and 2, we explored the role of p38 kinase as a potential in vivo K8 kinase by utilizing known specific activators and inhibitors of p38 kinase and by comparing phosphopeptide maps of in vivo versus in vitro p38phosphorylated K8. As shown in Fig. 3A, activation of p38 kinase in cultured HT-29 cells by An (42), as determined by p38 phosphorylation, is associated with rapid K8 Ser-73 phosphorylation. Similarly, the alkylating agent MMS, a known p38 kinase and JNK activator (43), generates the HK8 species in a dose-and time-dependent fashion (Fig. 3B). Inhibition of Aninduced p38 kinase activation with the specific inhibitor compound SB203580 abrogated K8 Ser-73 phosphorylation (Fig.  3C). In contrast, inhibition of ERK1/2 kinase activation with compound PD98059 did not significantly affect K8 Ser-73 phosphorylation but did inhibit K8 Ser-431 phosphorylation as determined by blotting with mAb 5B3 (Fig. 3D).
A comparison of the chymotryptic phosphopeptide maps of K8 and HK8 that are isolated from in vivo phosphorylated cells shows that HK8 differs from K8 by the presence of peptides 2-5 and by the absence of the peptide highlighted by an unnumbered arrow (Fig. 4, a and b). Interestingly, the phosphopeptide profile of HK8 that is generated by in vitro phosphorylation of K8 with p38 kinase shows five major peptides (Fig. 4c) that co-migrate with peptides 1-5 that are isolated from in vivo labeled HK8. This is confirmed by mixing in vitro and in vivo labeled K8 (Fig. 4d) and by mixing in vivo labeled HK8 with p38-labeled K8 (not shown). The five peptides are generated by incomplete chymotryptic digestion (not shown). Taken together, these results suggest that a p38-like kinase is likely to be involved, in vivo, in K8 phosphorylation at Ser-73.
p38 Kinase Associates with K8/18 and Phosphorylates K8 Ser-73 in Vivo and Binds to K8 in Vitro-We further substantiated in vivo p38 phosphorylation of K8 Ser-73 by comparing K8 Ser-73 phosphorylation in BHK cells transfected with FLAG-tagged human WT p38 or kinase-inactive p38 AF (Fig.  5A). The overexpressed p38␣ proteins are detected with anti- FIG. 1. In vitro K8 phosphorylation by p38 or p42 kinases and mAb LJ4 reactivity with K8 Ser-73 mutants. A, K8/18 immunoprecipitates from HT-29 cells were prepared using anti-K18 mAb L2A1 or anti-K8 Ser(P)-73 mAb LJ4, followed by SDS-PAGE, and then staining with Coomassie Blue. Note that mAb LJ4 preferentially recognizes the K8 Ser-73-phosphorylated species, HK8 (residual presence of K8 in lane 2 reflects the tetrameric nature of keratins that may contain two K18, one K8, and one HK8 molecules per tetramer). B, BHK cells were co-transfected with WT K18 and WT K8 or with WT K18 and the indicated K8 phosphorylation mutants. K8/18 were precipitated with mAb L2A1 and then analyzed by SDS-PAGE and Coomassie staining. Duplicate K8/18 immunoprecipitates were also separated by SDS-PAGE and then blotted with mAb LJ4. Note that LJ4 reactivity is abolished in the S73A mutant and is limited when blotted against the S73D mutant as compared with WT K8 (see text). C, in vitro kinase assays were performed using K8/18 immunoprecipitates that were obtained from HT-29 cells. Precipitates were incubated with 5 Ci of [␥-32 P]ATP, 20 M ATP, and 1 unit of p38 or p42 kinase for 15 min followed by quenching with sample buffer then SDS-PAGE analysis, Coomassie staining, and autoradiography (autorad). i.p., immunoprecipitation.
FLAG and anti-human p38 antibodies. As anticipated, p38 AF is not recognized by phospho-p38 antibody, and K8 Ser-73 phosphorylation increases in BHK cells that overexpress WT but not AF p38 (Fig. 5A, lanes 1-3). In addition, WT and AF p38 kinases co-immunoprecipitate with K8/18 in transfected cells (Fig. 5A, lanes 5 and 6); arrowhead and arrows indicate degraded K8 or apoptotic K18 fragments (44), respectively. Coimmunoprecipitation of p38 with K8/18 was not observed in non-transfected cells (e.g. HT-29 cells), which may be related to the high levels of p38 kinase in transfected cells and the tran-sient/weak nature of the kinase-substrate interaction (not shown). The interaction of p38 kinase with keratins was also confirmed using an in vitro overlay assay. As shown in Fig. 5B, p38 kinase bound specifically to K8 but not to K18. Taken together, these results support the conclusion that p38 kinase associates with K8 and phosphorylates K8 Ser-73 in vivo.
Effect of Disease-related Keratin Mutations on Keratin Phosphorylation-K8 Ser-73 is part of the sequence 68 NQSLLSPL, a sequence that is identical in all type II keratins (except for the Ser-73-equivalent residue which is substituted by Ala in K7, FIG. 2. Phosphorylation of WT K8 and K8 mutants by p38 kinase, p42 kinase, or JNK. BHK cells were cotransfected with WT K18 and WT K8 or with WT K18 and one of three K8 phosphorylation mutants (S23A, S73A, or S431A). Three days after transfection, K8/18 immunoprecipitates were obtained and then used in an in vitro phosphorylation assay with the indicated kinases. Precipitates were analyzed by SDS-PAGE, Coomassie staining, and then autoradiography.

FIG. 3. Modulation of K8 Ser-73 phosphorylation by activation or inhibition of p38 kinase.
A, HT-29 cells were treated with 0.1% Me 2 SO (0-h time point) or with An (10 g/ml) for the indicated times. Total lysates were then prepared by solubilizing with SDS sample buffer. Lysates were separated by SDS-PAGE, transferred to membranes, and then blotted with anti-p38 and anti-phospho-p38 kinase antibodies or anti-K8 Ser(P)-73 mAb LJ4. B, HT-29 cells were incubated with MMS and then harvested after the indicated time points, solubilized with 1% Nonidet P-40, followed by immunoprecipitation with mAb L2A1. K8/18 precipitates were separated by SDS-PAGE and then stained with Coomassie Blue or immunoblotted with mAb LJ4. Asterisks in lane 7 represent degraded K8 species. C and D, HT-29 cells were preincubated for 1 h with 20 M SB203580 (p38 kinase inhibitor) or 100 M PD98059 (MAPK kinase inhibitor) followed by An treatment for 2 h. K8/18 immunoprecipitates were obtained from 1% Nonidet P-40-solubilized cells and then blotted with anti-K8 Ser(P)-73 (mAb LJ4) or anti-K8 Ser(P)-431 (mAb 5B3). K8 Ser-431 phosphorylation is used as a control since we previously showed that this site becomes phosphorylated upon epidermal growth factor stimulation and that it is likely to be phosphorylated in vivo by p42 MAPK (30).
Gln in K1-3, and Thr in K4 -6; Ref. 31). Several of the mutations that have been described for epidermal keratins result in amino acid substitutions that potentially create a new, or remove a potential, phosphorylation site. Given the known impact of phosphorylation on keratin filament organization (9 -11), it is possible that such mutations could impact significantly on keratin filament organization and disease pathogenesis, although such a possibility has not been formally tested for any such mutation. To address this, we focused on one such mutation (Leu-160 3 Pro of K1 in a family of patients with epidermolytic hyperkeratosis (22)) that occurs in the highly conserved Ser-73-containing domain of K8 (i.e. Leu-71 within 68 NQSLLSPL of K8) by using K8 as a model system (because K1 cDNA is not available). This mutation generates a potential new proline-directed kinase-related site at Ser-70 of K8 (Ser-159 of K1). As shown in Fig. 6A, the K8 L71P mutation significantly increased K8 susceptibility to in vitro phosphorylation by p38 (compare lane 1 with 2) and p42 kinases (compare lane 3 with 4) but not by JNK (compare lane 5 with 6). The K8 L71P mutation also increased K8 phosphorylation in transfected cells after exposure to okadaic acid (Fig. 6B). This was confirmed by the presence of an HK8-like species in cells transfected with the K8 L71P but not with WT K8 as determined by Coomassie staining (Fig. 6B, compare lane 1 versus 2) and confirmed by immunoblotting with antibodies that recognize the total and phospho-K8 pools (Fig. 6B, lanes 3-6). No change was noted in K18 Ser-52 phosphorylation (Fig. 6B, lanes 9 and  10), which represents the major K18 phosphorylation site (10), thereby indicating specificity of the increased phosphorylation toward the mutant K8. Of note, the L71P K8 mutation inhibits binding of the LJ4 antibody to K8 (Fig. 6B, lane 8) 4. Phosphopeptide maps of in vitro and in vivo phosphorylated K8 and HK8. HT-29 cells were metabolically labeled with 32 PO 4 (250 Ci/ml) for 5 h (in the presence or absence of 100 g/ml of MMS to generate the HK8 species) followed by immunoprecipitation of K8/18. K8 (from cells without MMS treatment, a) and HK8 (from MMS-treated cells, b) were individually isolated using preparative SDS-PAGE, followed by chymotryptic phosphopeptide mapping. Alternatively, K8/18 immunoprecipitates were obtained from untreated HT-29 cells followed by in vitro phosphorylation using [␥-32 P]ATP and p38 kinase. K8 was separated by SDS-PAGE and then subjected to chymotryptic peptide mapping (c). Equal counts of the samples shown in a and c were also mixed and analyzed (d). The x in the left lower corners indicates the origin where samples were spotted onto thin layer cellulose plates for two-dimensional separation using electrophoresis (horizontal dimension) and then chromatography (vertical dimension). Note that the bracketed spots 2-5, which are not phosphorylated in K8 in vivo, are phosphorylated in vivo in HK8 and are also generated after in vitro phosphorylation of K8 with p38 kinase. The K8 peptide highlighted by an unnumbered arrow becomes relatively dephosphorylated after MMS treatment in HK8 (compare a with b) and in K8 (not shown).

FIG. 5. Association of p38 kinase with K8/18 immunoprecipitates and specific binding of p38 kinase to K8 in vitro.
A, BHK cells were co-transfected with WT K8/18 and one of the three constructs: vector, FLAG-tagged WT, or AF p38. Transfected cells were solubilized with SDS-containing sample buffer (total lysate) or with 1% Nonidet P-40 followed by immunoprecipitation of K8/18. Total lysates and K8/18 precipitates were analyzed by SDS-PAGE and stained with Coomassie Blue or transferred to polyvinylidene difluoride membranes for immunoblotting with the indicated antibodies. B, total lysate and a K8/18 immunoprecipitate (i.p.) were obtained from HT-29 cells and then separated by SDS-PAGE and transferred to a membrane. The membrane was incubated with purified p38 kinase, washed, and then blotted with anti-p38 antibody as described under "Experimental Procedures." showed a normal appearing and an indistinguishable filament organization among the four K8 constructs (shown only for WT and L71P K8 in Fig. 7, a and e, respectively; with very similar profiles for K8 S73A and S73D (not shown)). However, exposure of the transfected cells to okadaic acid unmasked significant differences in filament reorganization when comparing WT or S73D K8 with S73A K8 or when comparing WT K8 with L71P K8 (Fig. 7). For example and as shown in Fig. 7, okadaic acid resulted in 42 and 41% of the cells maintaining residual filaments in WT and S73D K8-transfected cells, respectively, whereas 61% of the cells transfected with S73A K8 had cells with intact filaments (a total of 120 -180 cells were counted in three independent experiments, p Ͻ 0.05). Hence, S73D rescues the filament reorganization defect caused by the S73A mutation, likely due to the negative charge of the aspartate.
In the case of the L71P K8 mutant, nearly 10% of the cells with collapsed filament (after exposure to okadaic acid) had prominent large dots (Fig. 7f), whereas none of the cells transfected with any of the other K8 constructs manifested this phenotype. These large dots likely represent coalesced smaller dots since longer exposure of cells expressing WT K8 results in progression from a fine dot to a large dot pattern (not shown). Taken together, these data suggest that K8 Ser-73 phosphorylation is associated with keratin filament destabilization and likely occurs to facilitate reorganization of the keratin filaments, whereas the patient-associated K8 L71P mutation results in an exaggerated keratin hyperphosphorylation response upon okadaic acid stimulation with consequent amplified destabilization of the keratin filament network.

DISCUSSION
K8 Ser-73 Is a Physiologic Substrate for a p38 MAPK-The temporal associations of K8 Ser-73 phosphorylation, as determined by the signature formation of the HK8 species and by reactivity with mAb LJ4, suggests that a stress-induced kinase is responsible for its phosphorylation. We tested, in vitro, three candidate kinases that are members of the MAPK superfamily, namely JNK, p42 (ERK1), and p38 kinase. Of these kinases, only p38 kinase phosphorylated K8 Ser-73 exclusively, based on HK8 formation ( Fig. 1C and Fig. 2A), whereas JNK and p42 kinase resulted in preferential phosphorylation of K8 (due to K8 Ser-431 phosphorylation) with some phosphorylation of the Ser-73 site ( Fig. 1C; Fig. 2, B and C). Further support for a physiologic role of p38 kinase in Ser-73 phosphorylation includes the following: (i) association of K8 Ser-73 phosphorylation with states that activate p38 kinase (e.g. An and MMS exposure of cells, Fig. 3, A and B); (ii) generation of a chymotryptic phosphopeptide pattern, upon in vitro phosphorylation of K8 with p38 kinase, that is very similar to the pattern of HK8 but not K8 in vivo phosphorylation (Fig. 4); (iii) inhibition of K8 Ser-73 phosphorylation by the selective p38 kinase inhibitor SB203580 but not by PD98059 (Fig. 3, C and D) which inhibits Erk1/2 kinase activation by inhibiting MEK1/2 kinases; (iv) p38 kinase association with K8/18 immunoprecipitates and phosphorylation of K8 Ser-73 by p38 kinase in transfected cells (Fig. 5A); and (v) specific binding of p38 kinase with K8 using an overlay assay (Fig. 5B). Hence, our data strongly implicate K8 Ser-73 as a physiologic substrate for p38 kinase and adds K8 to the few known likely physiologic substrates of p38 kinase that include MAPK-activated protein kinase-2 and ribosomal S6 kinase-B (45).
Our assignment of K8 Ser-73 as a physiologic substrate of a p38 kinase pertains in particular to p38␣, although other p38like kinases may be involved given the growing list of related p38 kinases. The p38 kinase family (also called stress-activated protein kinase-2 (SAPK-2)) has several known members including p38␣ (SAPK-2␣), p38␤ (SAPK-2␤), p38␥ (SAPK-3), SAPK-4, and p38␦ (46). These kinases share nearly 60 -75% sequence identity and have some differences in substrate spec- FIG. 6. Effect of the L71P K8 mutation on K8 phosphorylation. A, BHK cells were co-transfected with WT K18 and WT K8 or K8 L71P. After 3 days, cells were harvested followed by immunoprecipitation of K8/18. Precipitates were subjected to an in vitro kinase assay, followed by gel analysis and autoradiography as described in Fig. 1 legend. B, cells were transfected as in A. Just before harvesting, cells were incubated with OA (1 g/ml) for 2 h followed by immunoprecipitation of K8/18. Precipitates were analyzed by SDS-PAGE and then Coomassie staining or were analyzed by immunoblotting using antibodies that recognize the total K8 pool, K8 Ser(P)-73, K8 Ser(P)-431, and K18 Ser(P)-52. Note that the K8 L71P mutation results in hyperphosphorylation of K8 in vitro, using p38 kinase, and in transfected cells as determined by formation of the HK8 species.
FIG. 7. Immunofluorescence staining of NIH-3T3 cells transfected with WT or mutant K8. NIH-3T3 cells were co-transfected with WT K18 and one of the following K8 constructs: WT, S73A, S73D, or L71P. Transfected cells were grown on coverslips, and 3 days after transfection they were further cultured in the presence (ϩOA) or absence (control) of okadaic acid (1 g/ml) for 2 h. Cells were then fixed with methanol and stained using anti-K8 mAb M20. Note that OA results in the formation of a fine punctate pattern preferentially in WT and S73D K8 transfectants but less so in S73A transfectants. Also, note that OA-treated L71P K8-transfected cells form large keratin-staining dots that are not seen in the cells transfected with the other K8 constructs. ificity and in inhibition by pyridinylimidazole compounds such as SB203580. In our case, we only tested the p38␣ kinase, which is known to be inhibited by SB203580. It is likely that more than one kinase does phosphorylate K8 Ser-73 in vivo since such phosphorylation occurs during mitosis, a variety of cell stresses, and apoptosis (10). To that end, p38 kinase activation is reported after a variety of apoptotic stimuli and can also occur upon induction of proliferation as noted for B cells (47,48). In addition, Fas receptor stimulation of HT-29 cells activates JNK selectively, rather than p38 kinase, and results in phosphorylation of K8 Ser-73 (53).
Disease-causing Keratins Mutations May Modulate Keratin Phosphorylation-Several epidermal keratin mutations have been described at sites that may potentially introduce or remove a phosphorylation site and hence may affect disease pathogenesis by modulation of keratin phosphorylation upon the appropriate cell stimulation (12,13,(15)(16)(17). However, this potential of mutation-associated modulation of phosphorylation has not been formally tested for any of these mutations. Given that one such mutation in K1 (L160P) occurs at the highly conserved K8 Ser-73-like motif (K8 Leu-71 in 68 NQS-LLSPL is the equivalent Leu (boldface letters indicate conserved residues in all type II keratins)), we tested in K8 the effect of the K1 equivalent L71P mutation. This mutation resulted in K8 hyperphosphorylation, likely due to phosphorylation at the newly generated proline-directed kinase site in 68 NQSPLSPL. Hyperphosphorylation of K8 L71P was confirmed in cultured cells after exposure to okadaic acid and in vitro by p38 kinase phosphorylation (Fig. 6) and was associated with abnormal keratin filament reorganization (Fig. 7). Hence, our results support the conclusion that disease-causing keratin mutations can indeed generate abnormally phosphorylated keratins in a fashion that will predictably depend on the context of the mutation. At least in some cases, such modulation of keratin phosphorylation can alter keratin filament organization (Fig. 8) in response to physiologic and nonphysiologic hyperphosphorylating stimuli.
We used okadaic acid as a model system for the induction of generalized hyperphosphorylation, including the K8 Ser-73 site that undergoes phosphorylation in the presence of OA (31), since we were not able to visualize with confidence enough mitotic cells in our transient transfection system (not shown). Of note, phosphatase inhibitors, such as okadaic acid and microcystin, are major hepatotoxins in animals (49 -51) and in humans (52). Therefore, despite their generalized effects, the use of such compounds in cultured cells provides a relevant and sensitive filament reorganization model system.
K8 Ser-73 Phosphorylation Plays an Essential Role in Keratin Filament Organization-One unique feature of the K8 Ser-73 phosphorylation site, as contrasted with other known K8 and K18 phosphorylation sites, is its near-absolute on/off property whereas other phosphorylation sites manifest up/ down modulation of a basal phosphorylation state depending on the stimulus (10). This on/off property and the reversible induction of this phosphorylation suggest important biologic role(s) for this modification that represents the convergence of several contexts (e.g. stress, apoptosis, and mitosis; Ref. 31) that include p38 kinase activation and subsequent K8 Ser-73 phosphorylation. One common feature for these differing biologic contexts is the observed keratin filament reorganization that is associated with these processes. The data presented herein suggest a unique function for K8 Ser-73 phosphorylation, which is to allow keratin filaments to reorganize. The evidence for this role is the absence of a keratin-assembly defect upon transient transfection of a K8 S73A mutant but the unmasking of a phenotype upon exposure to OA-mediated hyperphosphorylating conditions (Fig. 7). Furthermore, the K8 S73D mutation rescues the S73A phenotype thereby supporting the role of the phosphoserine moiety at that site. Hence the aspartate substitution mimics the phosphate of K8 Ser(P)-73 biologically by rescuing the K8 S73A phenotype (Fig. 7) and biochemically by altering the migration pattern in SDS-PAGE gels from K8 to HK8-like (Fig. 1B).
Another unique feature of the K8 Ser(P)-73 species is their distribution among various cellular compartments, as compared with other K8 and K18 species that are phosphorylated at other sites. For example, K8/18 are found in increasing abundance in the sequentially isolated cytosolic, Nonidet P-40, Emp, and then post-Emp solubilized fractions (10,25). Interestingly, the HK8 species are distributed nearly uniformly FIG. 8. Proposed model for the significance of K8 Ser-73 phosphorylation and the potential impact of disease-causing phosphorylation-modulating keratin mutations. The exchange between the basal K8/18 filaments and the progenitor soluble filament pool is likely to be K8 Ser-73-independent due to the absence of any detectable basal K8 Ser-73 phosphorylation. Stimulation of cells, as may occur during cell stress or apoptosis, results in K8 Ser-73 phosphorylation via p38 kinase and in "normal" keratin filament reorganization (with increased keratin solubility), which becomes limited upon a K8 Ser-73 3 Ala mutation. A K8 Ser-73 3 Asp mutation rescues the filament reorganization defect that is caused by blocking Ser-73 phosphorylation. However, disease-causing mutations, such as the K1-like mutation that was introduced into K8 (L71P), can cause a hyper-hyperphosphorylated keratin state (upon stimulation) with subsequent abnormal keratin filament reorganization (indicated by large dots). Disease-causing mutations may also result in abnormal filament reorganization due to a hypophosphorylated state, as would be the case for a K8 S73A-like mutation. throughout these fractions, whereas keratins that are phosphorylated on K18 Ser-52 or K18 Ser-33 are preferentially found in the cytosolic and Nonidet P-40-containing fractions (10,26). This implies that keratin species that are typically cytoskeletal and insoluble (K8 Ser-73 state) become reorganized in a fashion that is associated with p38 kinase activation (and phosphorylation at other keratin sites) to favor generation of the K8 Ser(P)-73 state and distribution within the different cellular compartments in order to facilitate filament reorganization (Fig. 8).