HSP25 Is Involved in Two Steps of the Differentiation of PAM212 Keratinocytes*

HSP25 is a member of the small heat shock protein family. This 25-kDa protein exhibits a highly specific distribution during mouse embryonic development. Although multiple functions have been proposed for HSP25, the role it plays during differentiation is still unknown. High levels of HSP25 can be detected in embryonic and adult skin. During epidermis differentiation, the concentration of HSP25 increases with the distance of keratinocytes from the basal layer, in parallel with the extent of keratinization. We used an ex vivo cellular system, PAM212 cells, to analyze quantitatively and qualitatively the dynamics of HSP25 production and phosphorylation during the differentiation of keratinocytes. Our observations suggest that HSP25 is involved in two steps of PAM212 keratinocyte differentiation. Shortly after the induction of differentiation, a transient hyperphosphorylation of HSP25 seems to be essential for the expression of differentiation markers. Later, the chaperone-active form of HSP25 is organized progressively into characteristic aggregates involved in the dynamics of keratin filament networks. HSP25 (called HSP27 human cells) small heat shock protein 1 family. All family are small and a highly conserved carbox-yl-terminal with increasing concentrations of ethanol in- cluding a 30-min treatment in 0.4% uranyl acetate in 70% ethanol. Then, a monolayer of EPON TM was deposited on the cells and incubated for 1 h at 37 °C and then at 60 °C for 2 days. Thin strip of cells embedded in EPON were cut and embedded in cross-section. Thin sections were cut and stained with uranyl acetate and lead citrate and then examined in a Philips Tecnai 12 electron microscope.

HSP25 (called HSP27 in human cells) belongs to the small heat shock protein (HSP) 1 family. All members of this family are small (20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) and contain a highly conserved carboxyl-terminal region, called the ␣-crystallin domain (1). HSP25 was the first of the 10 mammalian members of this family to be described. Its synthesis is strongly up-regulated in response to heat and chemical and oxidative stresses. Two residues from HSP25 can be phosphorylated via the p38 MAP kinase pathway: Ser-15 and Ser-86. p38 phosphorylates MAPKAP kinase 2/3 which in turn phosphorylates HSP25 (2). ␣Aand ␣Bcrystallin were initially described as eye lens structural proteins. However, ␣B-crystallin has a function and sequence similar to those of HSP25. It can be induced by heat and chemical and oxidative stresses. These proteins are well known chaperones (3,4), although no substrate specificity has been described so far.
During embryonic development, HSP25 appears to be strongly produced in specific organs or tissues such as heart, skeletal muscles, smooth muscles, epidermis, eye lens, cartilage, and bone (5)(6)(7)(8)(9). Immunohistochemical analyses have detected HSP25 in some areas of the developing central nervous system, even though global Western and Northern blotting failed to detect this protein (10,11). This specific pattern suggests that this small heat shock protein is involved in cell differentiation, which is further supported by the results obtained by a few cellular approaches.
The ex vivo differentiation of HL-60 promyelocytic leukemic cells into macrophages or granulocytes is accompanied by the transient phosphorylation of HSP27 and by an increase in the production of the protein and a decrease in proliferation (12,13). The use of antisense oligonucleotides to reduce the production of HSP27 in HL-60 cells leads to a less pronounced reduction in cell growth after induction with retinoic acid and alters some parameters of granulocytic differentiation (14). However, HSP25 phosphorylation is not essential for HL-60 cell differentiation (15). During the pluripotent differentiation of embryonic stem cells, HSP27 is dephosphorylated rapidly after induction, which precedes the transient accumulation and oligomerization of the protein. In these cells, the overexpression of HSP27 reduces the rate of cell proliferation, whereas the down-regulation of the protein by stable antisense RNA compromises the differentiation program, causing massive cell apoptosis (16). Thus, HSP27 may act as a switch between differentiation and apoptosis. The ex vivo differentiation of rat olfactory progenitors is accompanied by a transient increase in HSP27 production. A down-regulation of HSP27 leads to the failure of differentiation and a massive commitment to cell death, whereas the overexpression of this protein strongly decreases the proportion of dying cells. Thus, HSP27 is probably a key element in the control of cell death during neuronal differentiation (17). The production of HSP25 and the activation of p38 MAP kinase are essential for the differentiation of P19 cells into cardiomyocytes. However, the effects of p38 MAP kinase activation seem to be independent of HSP25 accumulation (18). The overexpression of HSP25 in C1 cells, which are able to differentiate into chondrocytes, showed that elevated HSP25 levels decrease the cellular adhesion of C1 cells and interfere with their differentiation into chondrocytes (19).
In this study, we concentrated on the role of HSP25 in the differentiation of mouse epidermis. In the adult mouse, the epidermis consists of several cell layers. The degree of differentiation increases from the basement membrane to the outside of the body: basal layer, spinous layer, granular layer, and the stratum corneum. The cells in the basal layer express mainly mouse keratins 5 (MK5) and 14 (MK14), whereas those in the suprabasal layers (from the spinous layer) express mainly keratins 1 (MK1) and 10 (MK10). HSP27 can be detected in the human epidermis from week 14 of gestation. HSP27 staining increases with the distance of the keratinocytes from the basal layer, in parallel with the extent of keratinization (40). In normal human skin, HSP27 colocalizes with keratins and other markers of the cornified layer, suggesting that it plays a role in the assembly of keratin filaments (41).
We used PAM212 cells as a model for keratinocyte differentiation. This malignantly transformed mouse cell line can proliferate in a medium with a low calcium concentration, and terminal differentiation of these keratinocytes can be induced by increasing the calcium concentration of the culture medium (42). The overall pattern of keratin synthesis in these cells is similar to that in the newborn epidermis, whereas primary cultures of newborn mouse epidermis fail to produce ex vivo the same keratins (43). Therefore, the PAM212 cell line is an appropriate model for studying the role of HSP25 in the differentiation of the epidermis.
We show that HSP25 is involved in two stages of PAM212 keratinocyte differentiation. First, a transient accumulation of the phosphorylated form of HSP25 seems to be essential for the induction of differentiation. Second, the nonphosphorylated chaperone-active form of HSP25 contributes to the reorganization of the keratin network by driving the disassembly of the MK5/MK14 network.

MATERIALS AND METHODS
Cell Culture and Differentiation-PAM212 cells were cultured in tissue culture dishes (Falcon) at 37°C in 7.5% CO 2 water-saturated atmosphere in calcium-free Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% heat-inactivated fetal calf serum, 1 mM pyruvate, 2 mM glutamine, 50 g/ml streptomycin, 50 units/ml penicillin, and 0.05 mM CaCl 2 (low calcium medium). To carry out terminal differentiation, PAM212 cells were grown until confluence and then incubated in the same medium containing 0.5 mM CaCl 2 (high calcium medium). High calcium medium was replaced every day. For p38 inhibition assays, 10 M SB203580 (Calbiochem) was added to the culture medium.
Gel Electrophoresis and Immunoblotting-Protein extracts were prepared by washing cells in 4°C PBS, scraping into Laemmli sample buffer with 5% ␤-mercaptoethanol, and then heating for 10 min at 95°C (44). Protein concentration was determined using the Bio-Rad protein assay kit, and equal amounts of protein were loaded on an SDS-polyacrylamide gel (15%). Prestained low range SDS-PAGE standards from Bio-Rad were included in each gel. Proteins were transferred onto Hybond-ECL nitrocellulose membranes (Amersham Biosciences). Immunoblotting was performed using the ECL kit (Amersham Biosciences). The primary antibodies were MK1 (Babco), MK10 (Babco), and HSP25 (Stressgen). The secondary antibody was horseradish peroxidase-conjugated anti-rabbit IgG (Promega).
Two-dimensional Gel Electrophoresis-Aliquots of protein extracts, containing equivalent amounts of HSP25, saturated with urea, and completed with 2% Nonidet P-40, were separated by isoelectric focusing in 4% acrylamide gels containing 8 M urea, 2% Nonidet P-40, 0.02% Ampholine 3-10, and 0.02% Ampholine 4 -6, in glass tubes. The upper tank was filled with 20 mM NaOH and the lower tank with 10 mM phosphoric acid. Samples were subjected to electrophoresis at 800 V overnight. First dimension gels were then removed from the tubes, soaked in Laemmli buffer, laid on the top of a 15% SDS-PAGE, and subjected to second dimension electrophoresis in classical conditions for size separation. Proteins were electrotransferred onto Hybond-ECL nitrocellulose membranes. HSP25 was detected by immunoblotting as described above.
Immunocytochemistry-Cells were grown and allowed to differentiate on gelatin-coated sterile glass coverslips. Cells at each stage were washed with PBS at 4°C and fixed in 4% paraformaldehyde for 15 min at room temperature. Coverslips were washed in PBS and stored at 4°C in PBS. The cells were subjected to immunocytochemistry in a humid chamber. To permeabilize the cells, coverslips were incubated in 0.1% Triton X-100 in PBS for 5 min. After blocking in 3% BSA in PBS for 1 h, cells were incubated with primary antibodies in 3% BSA and PBS for 1 h. The coverslips were washed three times in PBS and once in 3% BSA and PBS (5 min), then incubated with secondary antibodies, Hoechst 33342 (Sigma) and fluorescein isothiocyanate-conjugated phalloidin when necessary (Sigma) for 30 min in the dark. After four washes in PBS, coverslips were mounted on slides with Mowiol, dried overnight, and examined on a Leica DMRB microscope or a Leica TCS/SP2 confocal microscope. The primary antibodies were HSP25 (Stressgen), MK1 (Babco), MK5 (gift from Dr. Oulad, IGBMC, Strasbourg, France), and MK10 (DAKO). The secondary antibodies were Cy®3-conjugated goat anti-rabbit IgG (Jackson,) and Alexa Fluor® 488-conjugated goat antimouse IgG (Molecular Probes).
Immunohistochemistry-For immunohistochemical analyses embryos were fixed in 4% paraformaldehyde in PBS overnight at 4°C. They were embedded in Paraplast® Plus (Sherwood Medical). 10-m sections cut on a Leica microtome were laid on treated slides (Super-Frost Plus) in water, dried at 37°C, and stored at 4°C. Before immunohistochemistry, slides were cleared of Paraplast in two Histo-Clear® (National Diagnostics) baths of 15 min, then gradually rehydrated through an ethanol series and soaked in PBS for 5 min. Immunohistochemistry was performed in a humid chamber. Free aldehydes were saturated with 50 mM NH 4 Cl in PBS for 30 min. Sections were submerged for 30 min in a blocking solution (3% BSA in PBS), containing 0.5% Triton X-100 for permeabilization. Slides were washed three times for 10 min in 3% BSA, 0.1% Triton X-100 in PBS (dilution buffer), then incubated with primary antibodies in dilution buffer for 1 h. After washing three times in dilution buffer, slides were incubated with secondary antibodies in dilution buffer for 30 min. Sections were washed three times in PBS and mounted in Mowiol. Slides were examined on a Leica TCS/SP2 confocal microscope. Each picture was the result of an average performed from a series of 12 scans separated by 0.2 m. The same antibodies were used as for immunocytochemistry.
Electron Microscopy-PAM212 cells grown on Petri dishes were fixed with 2.5% glutaraldehyde and 1% tannic acid in 0.1 M cacodylate buffer for 1 h at pH 7.4, washed in the same buffer, postfixed in 1% OsO 4 for 45 min, and dehydrated with increasing concentrations of ethanol including a 30-min treatment in 0.4% uranyl acetate in 70% ethanol. Then, a monolayer of EPON TM was deposited on the cells and incubated for 1 h at 37°C and then at 60°C for 2 days. Thin strip of cells embedded in EPON were cut and embedded in cross-section. Thin sections were cut and stained with uranyl acetate and lead citrate and then examined in a Philips Tecnai 12 electron microscope.

The Morphology of PAM212 Cells Changes during Terminal
Differentiation-PAM212 keratinocytes were cultured in low calcium medium until confluence. On day 0, terminal differentiation was induced by increasing the concentration of calcium in the medium. The morphology of the cells was followed by direct observation in the culture dish. Half-confluent cells, observed 2 days before induction (day Ϫ2), were stretched and often exhibited cytoplasmic extensions (Fig. 1A, a). At day 0, PAM212 cells were apposed, but spaces were visible between the cells (Fig. 1A, b). On day 1, the cells were already organized into two layers, but spaces could still be seen between cells (Fig.  1A, c). From day 2 onward, the cells in the upper layer formed a flat surface of tightly sealed cells, making it difficult to localize cell boundaries. Only nuclei were clearly visible at these stages (Fig. 1A, d and e). The stratification of PAM212 cells could easily be visualized after staining with Hoechst. Indeed, at day 0, nuclei were well separated from each other, showing that there is only one layer of cells at this stage ( Fig.  1B, a). On day 1, the nuclei density had significantly increased, and some nuclei overlapped each other, demonstrating that cells had begun to form a second layer (Fig. 1B, b). The density of the nuclei continued to increase with time such that three and sometimes four nuclei overlapped, suggesting the presence of at least three cell layers (Fig. 1B, c). Electron microscopy analysis of transverse cell sections on day 4 was consistent with these observations. On day 4, three or four cell layers were visible. The nuclear region appeared to be rather thick, but the cells exhibit very thin cytoplasmic extensions that make them creep between neighboring cells (Fig. 1C).
The Production of HSP25 Increases during the Differentiation of PAM212 Cells-We used Western blotting to analyze the expression of specific differentiation markers to confirm the final differentiation of the keratinocytes (Fig. 2). MK1 and MK10 are the two main keratins produced by the suprabasal layers of the epidermis in vivo. In differentiating PAM212 cells, MK1 could be detected as early as day 1 and then increased 2-fold during the following steps. However, MK10 could not be detected before day 2 and increased until day 4.
The level of HSP25 changed considerably during the differentiation process (Fig. 2). Western blot analysis detected a low level of HSP25 in proliferating cells cultured in a low calcium medium (days Ϫ2 and 0). During differentiation in high calcium medium, the level of HSP25 increased (ϳ6-fold) until day 3. The amount of HSP25 decreased slightly at day 4.
These results confirm that PAM212 cells are a good model for epidermal differentiation and that this system is appropriate for studying the role of HSP25 in the differentiation of keratinocytes.
HSP25 Aggregates Progressively during the Differentiation of the Keratinocytes-We used immunocytochemistry to analyze the distribution of HSP25 in PAM212 cells before and during terminal differentiation. Actin filaments were stained in the same experiments to (i) distinguish all the cell layers regardless of whether they contain HSP25; (ii) follow changes in cell morphology; and (iii) look for possible interactions between HSP25 and actin filaments.
We found that HSP25 aggregates progressively during the differentiation of PAM212 cells (Fig. 3). At days Ϫ2 and 0, before induction, the protein was distributed homogeneously in the cytoplasm (Fig. 3, a and d), and the cells (one layer) dis-played a typical actin network (Fig. 3, b and e). The patterns of HSP25 and actin did not overlap at these stages (Fig. 3, c and f). From day 1 onward, the cells became stratified, and those in the bottom layer exhibited a bundle of stretched actin filaments crossing the cell in a preferential direction (Fig. 3t). HSP25 was no longer expressed in these cells (Fig. 3s). At day 1, the distribution of actin filaments was very characteristic in the cells in the newly formed upper layer. Indeed, actin formed granular structures, concentrated mainly at the cytoplasm periphery (Fig. 3, h and hЈ). In these cells, HSP25 had already started to form thin, elongated, tortuous aggregates. These structures were located mainly at the cell periphery and around the nuclei (Fig. 3, g and gЈ). The HSP25 and actin signals seemed to overlap partially at the plasma membrane (Fig. 3, i and iЈ). At day 2, the actin filaments in upper layer cells were mainly flattened along the cortical cytoplasm, thus revealing the borders between the cells, and granular structures were still found throughout the cytoplasm (Fig. 3k). These cells were polygonal and tightly adhered to each other, giving the whole surface a paved appearance. This pattern of actin filaments did not change during the following steps of differentiation (Fig. 3, n, q, and qЈ). At day 2, the aggregated HSP25 began to concentrate in some parts of the cell, but other parts tended to be free of HSP25 (Fig. 3j). Interestingly, high amounts of HSP25 were detected around the nuclei of these cells. At day 3, HSP25 formed large (ø 1-2 m), very compact aggregates distributed in the whole cytoplasm and in the nuclei (Fig. 3m). At day 4, the aggregates were the same shape, but in a large proportion of the cells they were located mainly at the periphery, just below the cytoplasmic membrane, and in the nucleus (Fig. 3, p and pЈ). Interestingly, these aggregates were ring-shaped. Indeed, staining revealed spherical structures in peripheral regions. These structures sometimes formed doublets ( Fig. 3rЈ) but often formed single rings (see Fig. 4, p-r). This characteristic pattern suggests that HSP25 is located at the periphery of areas that do not contain HSP25 and that may or may not contain other elements. These aggregates and actin did not overlap in any part of the cell (Fig. 3, l, o, r, and rЈ). At day 4, HSP25 was again distributed diffusely in the cytoplasm of some cells (see Fig. 4p).
Thus, HSP25 forms progressively large aggregates that tend to concentrate at the edges of the cells and in the nuclei. This highly dynamic pattern, plus the increase in the amount of protein, strongly suggested that HSP25 plays a specific role in the differentiation of PAM212 cells.  Ϫ2, 0, 1, 2, 3, and 4), proteins were extracted from PAM212 cells by treatment with Laemmli buffer. Equal amounts of proteins were used to analyze the levels of MK1, MK10, and HSP25 throughout differentiation.
we compared the pattern of HSP25 with that of each of these cytokeratins.
We compared the pattern of HSP25 with that of MK5, which is expressed in PAM212 cells before induction. At days Ϫ2 and 0, MK5 formed a typical network of cytokeratin filaments throughout the cytoplasm (Fig. 4, b and e). As described above, HSP25 was present in the whole cytoplasm at these stages (Fig. 4, a and c). The patterns of HSP25 and MK5 did not seem to overlap at these stages (Fig. 4, c and f). From day 1 onward, the cells in the bottom layer no longer contained either HSP25 or MK5. The distribution of MK5 and HSP25 overlapped in the upper layer cells from this stage, both forming similar aggregates (Fig. 4, g-r). The aggregates of HSP25 and MK5 are both ring-shaped (singlets or doublets) (Fig. 4, p-r). Immunocytochemical analysis showed that MK14 exhibits the same aggregation pattern as MK5 and thus HSP25 (data not shown). Thus, HSP25 interacts with the main keratins present in the basal layer in vivo. HSP25 seems to be involved in the reorganization of the keratin network during the differentiation of PAM212 cells.
MK1 and MK10 were not expressed in PAM212 cells before induction but appeared during differentiation. For technical reasons, we did not directly compare MK1 and HSP25 patterns (both antibodies raised in rabbit). Instead, we compared the expression patterns of MK1 and MK5, the pattern of which is similar to that of HSP25 from day 1 onward. At day 1, MK1 was expressed in cells expressing MK5 that had started to aggregate (Fig. 5A, a-c). At day 2, more cells expressed MK1, and these cells no longer expressed MK5. The cells in the upper layers were very flat and intertwined, which sometimes made it difficult to distinguish the borders between the cells. Exam-ination of a series of sections (averaged) from the upper layer showed that the cells expressing MK5 were distinct from those expressing MK1 (Fig. 5A, d-f). The same was observed at day 4, when a large proportion of the cells expressed MK1 and were still closely intertwined with cells expressing MK5, aggregated or not (Fig. 5A, g-i). We also compared the expression patterns of MK1 and MK10. Basically, cells expressing MK10 also expressed MK1, and some cells expressed MK1 without expressing MK10 (Fig. 5B). These observations are consistent with the kinetics of differentiation markers expression revealed by Western blot analysis (Fig. 2).
Thus, MK1 is initially induced in cells containing HSP25 that has started to aggregate, but at subsequent stages cells expressing MK1 no longer contain HSP25.
Interactions between HSP25 and MK5 Are Likely to Occur in Developing Epidermis, at the Beginning of Stratification-The interaction described above between HSP25 and MK5 was quite puzzling. Indeed, these two proteins are not produced in the same cells in adult epidermis; MK5 is specific for the basal layer, whereas HSP25 is expressed mainly in the suprabasal layers. We wondered whether these two proteins could be detected in the same cells during epidermis development. Therefore, we analyzed by immunohistochemistry the expression patterns of HSP25 and MK5 in mouse epidermis (dorsal side of the embryo), before and after the beginning of stratification. We observed that HSP25 and MK5 are both present in the monolayer constituting the epidermis at E12.5 (Fig. 6, a-c). At E15.5, when the epidermis is already stratified and keratinized, HSP25 was present in the suprabasal layers known to express MK1 and MK10, whereas the basal layer expressing MK5 did not contain HSP25 (Fig. 6, d-f).  Ϫ2 (a-c), day 0 (d-f), day 1 (g-i), day 2 (j-l), day 3 (m-o), and day 4 (p-r). HSP25 was detected by immunocytochemistry (Cy3, red; a, d, g, j,  m, and p), and actin microfilaments were stained with fluorescein isothiocyanateconjugated phalloidin (green; b, e, h, k, n,  and q). c, f, i, l, o, and r, overlay of the signals. For days 1 and 4, areas delimited by white squares were magnified three times (gЈ-iЈ and pЈ-rЈ, respectively). All observations were carried out on a confocal microscope (single scans). At days Ϫ2 and 0 (a-f), only a single cell layer existed. From day 1 to day 4 (g-r), several cell layers existed, and we focused on the upper layers. Cells in the bottom layer were also observed at day 1 for HSP25 (s) and actin (t). The white arrow indicates a nucleus containing HSP25. The white circle delimits a nucleus at day 4.
Thus, interactions between HSP25 and MK5, such as those observed in PAM212 cells, are likely to occur between the time when epidermis forms a monolayer and the time when it is stratified, during embryonic development.

HSP25 Is Transiently Phosphorylated a Few Hours after Induction, and the Inhibition of p38 MAP Kinase Prevents
Differentiation-We wondered whether the increase in the HSP25 level during the differentiation of PAM212 cells and its highly dynamic pattern were accompanied by a change in the phosphorylation state of the protein.
The phosphorylation of HSP25 is known to be dependent on the p38 MAP kinase pathway. p38 phosphorylates MAPKAP kinase 2/3, which subsequently phosphorylates HSP25. p38 can be specifically inhibited by the pyridyl imidazole SB203580. We tested the effects of this inhibitor on the differentiation of PAM212 cells (Fig. 7A). SB203580 was added to the culture medium on day Ϫ2, 0, 2, 4 or 6. In each case, cells were cultured until day 8, and proteins were extracted every other day (day Ϫ2, 0, 2, 4, 6, and 8). The addition of SB203580 at day Ϫ2 or 0 prevented the terminal differentiation of PAM212 cells, as revealed by the absence of MK10. When added at day 2, 4, or 6, SB203580 did not prevent the expression of MK10, suggesting that PAM212 cells differentiate properly in these conditions. Given that MK1 is expressed before MK10 in normal conditions, we analyzed the expression of MK1 in cells treated with SB203580 from day Ϫ2. The expression of MK1 in SB203580-treated cells was strongly delayed and very slight compared with that of MK1 in nontreated cells analyzed in the same conditions (Fig. 7B).
These results suggest that p38 is activated early in the differentiation process and that this activation is essential for the expression of MK1 and MK10. Given that MK1 was detected on day 1 (see Fig. 2), p38 must be involved before this stage. Thus, we studied the kinetics of the expression of this marker. MK1 was expressed faintly 10 h after induction and strongly from 16 h (Fig. 7C).
Given that the expression of MK1 and MK10 is dependent on the activation of p38 a few hours after induction, we followed the dynamics of HSP25 phosphorylation during the hours after induction. HSP25 was partially phosphorylated (one phosphorylated isoform) before induction (0 h). Between 4 and 6 h after induction, two phosphorylated isoforms of HSP25 appeared, the proportion of which increased dramatically between 6 and 12 h after induction (Fig. 7D). At day 1, the two phosphorylated isoforms were still detected, although in a lower proportion, whereas later, the phosphorylated isoforms could no longer be detected in these conditions (data not shown).
Thus, p38 is essential for the differentiation of PAM212 cells, and HSP25 is transiently hyperphosphorylated a few hours after induction, before accumulating in its nonphosphorylated form.
Specific Inhibition of p38 MAP Kinase Alters the Aggregation Pattern of HSP25-We used immunocytochemistry to analyze PAM212 cells treated with SB203580 from day Ϫ2 to see whether HSP25 aggregates when p38 is inhibited. At day 4, in the presence of SB203580, PAM212 cells were stratified, and the morphology of the actin network was similar to that observed in normal conditions. Indeed, the cells in the bottom layer exhibited typical transversal fibers (Fig. 8b), and, in the cells in the upper layers, actin was distributed below the plasma membrane and in granular structures within the cytoplasm (Fig. 8, e and eЈ). However, the pattern of HSP25 was strongly altered by the inhibition of p38. Indeed, contrary to what was observed in normal conditions at day 4, the cells in the bottom layer still expressed HSP25 (Fig. 8a), and those in the upper layers did not exhibit the typical aggregates described above (Fig. 8d). The pattern of HSP25 looked rather like that observed on day 1 (small granular aggregates), with the exception of around the cell borders ( Fig. 8dЈ; see Fig. 3, i  and iЈ). The patterns of HSP25 and actin did not overlap (Fig.  8, c, f, and fЈ). In the same conditions, the distribution of MK5 was the same as that of HSP25 (Fig. 8, g-l). As expected from the Western blot analysis (see Fig. 7B), MK1 was expressed in a few isolated cells in the upper layers, less numerous than those observed at day 1 in normal conditions (Fig. 8, m and mЈ). Thus, the inhibition of p38 MAP kinase not only impairs differentiation, but also alters the aggregation pattern of HSP25.  HSP25 (Cy3, red; a, d, g, j, m, and p) and MK5 (Alexa Fluor 488, green; b, e, h, k, n, and q) were detected by immunocytochemistry. c, f, i, l, o, and r, overlay of the signals. All observations were carried out on a confocal microscope (single scans). At days Ϫ2 and 0 (a-f), only a single cell layer existed. From day 1 to day 4 (g-r), several layers existed, and we focused on the upper layers. The insets in p, q, and r show an enlargement of an aggregate (bar, 2 m).

HSP25 and the Dynamics of Cytokeratin Filaments-We
showed that aggregates containing HSP25, MK5, and MK14 form progressively during the differentiation of PAM212 cells. These observations clearly demonstrate that HSP25 interacts specifically with the two keratins constituting the keratin network present in these cells before induction.
Several studies have already reported interactions between small HSPs and intermediate filaments. The first evidence of such interactions came from the study of Alexander's disease involving the formation of characteristic cellular inclusions (Rosenthal fibers) containing glial fibrillary acidic protein associated with ␣B-crystallin (29) plus HSP27 (34). Other diseases were subsequently shown to be characterized by the aggregation of ␣B-crystallin with intermediate filaments (30 -32). Independently of disease conditions, a study performed on eye lens extracts demonstrated that ␣A-crystallin and ␣B-crystallin interact directly with vimentin and their soluble subunits (33). The same study showed that ␣B-crystallin can inhibit intermediate filaments assembly in vitro. HSP27 was later shown to interact with glial fibrillary acidic protein and ␣B-crystallin in astrocytoma cells producing both small HSPs, and to interact with keratin 18 in epithelial cells that do not contain ␣B-crystallin (35). According to these observations, the pattern of the small HSPs perfectly overlaps the intermediate filament network. Even though several models have been proposed, the physiological significance of this interaction has not been elucidated. The discovery that a point mutation in ␣Bcrystallin causes desmin filament aggregation in certain myopathies (45) provided evidence of a physiological role for small HSPs with intermediate filaments. We did not find any correlation between the pattern of HSP25 and the intact keratin network before induction of PAM212 cells. The interaction we describe here involves keratins 5 and 14, dissociated from their initially intact network and involved in the formation of characteristic aggregates. These aggregates are totally different from those involving small HSPs and intermediate filaments in disease contexts. This is the first time that such structures containing HSP25 and intermediate filaments have been described. A previous study found an interaction between keratin 18 and HSP27 in epithelial cells (35), but these cells exhibited a stable and well established keratin network. The type of interaction we describe here may be typical of the differentiation of stratified epithelia, involving a switch from a keratin network to a new one.
Our observations suggest that HSP25 drives the disassembly of the keratin network through the sequestration of keratin 5 and 14 subunits. Moreover, we noticed that MK1 is expressed initially in cells positive for HSP25/MK5/MK14 aggregates before being limited to another kind of cell that does not contain these aggregates. Thus, it is possible that the keratin network of undifferentiated cells (MK5/MK14) has to be disassembled before the keratin network of differentiated cells can be built. The link between the aggregation of HSP25/MK5/MK14 and the expression of MK1 and MK10 is strengthened by the fact that SB203580 both inhibits differentiation and prevents aggregation. However, it is not clear whether the beginning of the aggregation process is a consequence of differentiation induction or whether it drives differentiation. Interestingly, in PAM212 cells containing MK5/MK14 starting to aggregate and newly synthesized MK1, HSP25 did not interact with MK1, which is an intermediate filament of the same type as MK5 (type I). However, in epithelial cells normally expressing keratins in interaction with HSP27, HSP27 interacts with glial fibrillary acidic protein (type III intermediate filament), the expression of which was triggered by transfection (35). This observation further confirms the high specificity of the interaction among HSP27, MK5, and MK14 in this system. Fig. 3. A, at day 1 (a-c), day 2 (d-f), and day 4 (g-i), MK5 (Alexa Fluor 488, green; a, d, and g) and MK1 (Cy3, red; b, e, and h) were detected by immunocytochemistry. c, f, and i, overlay of the signals. All observations were carried out on a confocal microscope, and the results are the mean of a series of 12 images separated by 0 .2 m. B, at day 4, MK1 (Cy3, red; a) and MK10 (Alexa Fluor 488, green; b) were detected by immunocytochemistry. c, overlay of the signals. The observations were carried out on a confocal microscope (single scan).

MK5 and MK14 Are Physiological Substrates of HSP25
Chaperone Activity-According to the dynamics of HSP25 phosphorylation throughout differentiation, the aggregation process involves the nonphosphorylated form of HSP25, which should form large oligomers known to be associated with the chaperone activity of HSP25 (46). A recent study showed that the level of HSP27 is increased during the keratinocyte differentiation of human Hacat cells and that this HSP27 accumulates in the form of large oligomers (47). However, the authors considered the accumulation of the chaperone-active form of HSP27 to be a marker of endogenous stress and claimed that it was not to be related to a particular differentiation process.
Our results clearly show that during PAM212 cells differentiation HSP25 interacts with proteins that are specifically related to the differentiation process. The chaperone-active form of HSP25 interacts with denatured proteins. Thus, the disruption of the MK5/MK14 network may involve a partial unfolding of keratin subunits that are sequestered by HSP25 to prevent undesirable interactions and aggregation. MK5 and MK14 are the first physiological substrates of the HSP25 chaperone complex described so far. This is the first report showing that HSP25 chaperone activity is involved in a specific differentiation process, with specific substrates.
However, these considerations do not explain the fate of the aggregates. We thought that the formation of aggregates might be a first step before degradation by the proteasome, but this hypothesis is not in agreement with the oriented motility of the structures. Moreover, the finding of cells exhibiting a typical MK5/MK14 network among the cells in which these keratins had aggregated suggests that sequestered keratins can still be used to build an intact network. This hypothesis is in agreement with the demonstration that the substrates interacting with the chaperone-active form of HSP25 are not irreversibly sequestered but can readopt their native state (48). The nature of the ring-shaped structures and the significance of their localization remain to be elucidated.
HSP25 Phosphorylation, p38 Activation, and Terminal Differentiation of PAM212 Cells-HSP25 is transiently hyperphosphorylated a few hours after induction of PAM212 cell terminal differentiation. Indeed, the phosphorylation of HSP25 increased dramatically between 6 and 12 h after induction. The two phosphorylated isoforms were still detected 24 h after induction, but subsequently they were much rarer than the nonphosphorylated isoform.
The phosphorylation of HSP25 is dependent on the p38 MAP kinase pathway (2). The role of MAP kinases in keratinocyte differentiation has already been addressed in other systems, generating different conclusions. For example, human foreskin terminal differentiation induced with 12-O-tetradecanoylphorbol-13-acetate depends on a MAP kinase pathway including protein kinase C, Ras, MEKK1, MEK3, p38 and activator protein 1 (49). In another system, the induction of terminal differentiation after an increase in calcium concentration was shown to be independent of protein kinase C and Ras and to involve the activation of Raf and MEK, the transient activation of ERK, and an increase in p21 associated with cell cycle arrest. However, in this system, ERK was shown not to be sufficient alone to induce differentiation, suggesting that other pathways act in concert with the Raf/MEK/ERK pathway (50). Here, we show that p38 MAP kinase activity is required early in the differentiation of PAM212 cells.
These data show that HSP25 is transiently hyperphosphorylated a few hours after induction and that the activation of p38 MAP kinase is essential for terminal differentiation. Thus, the phosphorylation of HSP25 may be involved in the differentiation process. Of course, other targets such as transcription factor (51,52) or other protein kinases (53-55) may also be affected by the inhibition of p38. Nevertheless, the fact that the phosphorylation of HSP25 (6 -12 h) is followed rapidly by the induction of MK1 (10 -16 h) strengthens the hypothesis that HSP25 phosphorylation is involved in this induction process, probably in concert with other pathways. HSP25 phosphorylation may modulate actin polymerization (22,56) in a way that could affect differentiation. This hypothesis is likely to be true because the regulation of actin dynamics is essential for epithelial cell-cell adhesion (57). Moreover, at day 1, HSP25 seemed to colocalize partially with actin at the plasma membrane. A more detailed analysis of the dynamics of HSP25 localization during the early steps of the differentiation may provide a clearer idea of what happens during these steps.
PAM212 Cells and Epidermis Differentiation-The expression patterns of the differentiation markers that we analyzed further characterize PAM212 cells as a model for epidermis differentiation.
During terminal differentiation, PAM212 cells express MK1 before MK10, which is surprising as the cytokeratin network of differentiated cells consists of a MK1/MK10 heteropolymer. Moreover, all cytokeratin filaments are made of both a type I cytokeratin and a type II cytokeratin. Thus, the filaments containing MK1 at day 1 must contain another type II cytokeratin.
Given that PAM212 cells stratify, one could expect a parallel between this pattern of stratification and the organization of the cell layers in the epidermis: lower cells constituting the basal layer and upper cells the suprabasal layers. In fact, even though PAM212 cells express the MK5/MK14 network before induction, the bottom cells no longer expressed either MK5/ MK14 or MK1/MK10 on day 1, suggesting that they do not correspond to either basal or suprabasal layers, respectively. These cells may serve as a matrix for the upper keratinocytes involved in the differentiation process, separating them from the surface of the culture dish.
What particularly puzzled us was the lack of correlation between the distribution of HSP25 in the different layers of the skin and the distribution of HSP25 in undifferentiated and differentiated PAM212 cells. On the one hand, in vivo, basal cells mainly expressing MK5/MK14 keratins do not contain HSP25, whereas suprabasal cells expressing mainly MK1/ MK10 keratins contain high levels of HSP25. On the other hand, we showed that HSP25 is present in PAM212 cells containing MK5/MK14 and closely interacts with these keratins and that it is present in differentiated cells (MK1/MK10) only when they emerge. Thus, it is difficult to consider that the observations made in PAM212 cells correspond to a physiological phenomenon taking place in skin at a time when it is already composed of several layers. The analysis of the expression patterns of HSP25 and MK5 in developing mouse epidermis shows that the two proteins are present in the same cells in the monolayered epidermis, before stratification. Thus, the interaction between HSP25 and MK5 observed in PAM212 cells is likely to correspond to a phenomenon happening at the beginning of epidermis stratification and not at a time when the epidermis already consists of several cell layers.
These observations suggest that what we observed with PAM212 cells is not really a model of what happens in newborn or adult epidermis, which is involved in a continuous renewal process. Instead, this model reflects what happens at the beginning of stratification during embryonic development. These results also show that HSP25 has distinct roles in the suprabasal layers of the epidermis (from E15.5, in newborn and adult mice), where it acts with other partners.
HSP25 and Cellular Differentiation-Studies on HL-60 promyelocytic leukemic cell differentiation have shown that cell proliferation decreases as HSP27 levels increase (12,13) and that HSP27 is essential for the correct differentiation of HL-60 cells (14). Studies on the pluripotent differentiation of embryonic stem cells (16) and on rat olfactory neurons differentiation (17) have suggested that HSP27 acts as a switch between differentiation and apoptosis. Other studies have demonstrated the importance of HSP25 in cell differentiation, without elucidating the function of the protein in the processes analyzed. Thus, apart from possibly being involved in the balance between cell proliferation, apoptosis, and differentiation, the specific role of HSP25 in the organs where the protein is produced in vivo remains unknown.
Here we show that HSP25 chaperone complexes drive the disassembly of the MK5/MK14 network during PAM212 keratinocytes differentiation. This is obviously specific to epithelial cell differentiation. Thus, we believe that HSP25 has different functions in cell differentiation according to the system studied. This protein probably interacts with different specific partners in each tissue producing HSP25 during embryo development. Moreover, in vivo observations clearly suggest that HSP25 is involved in the differentiation of a subset of tissues, but not in all differentiation processes. It is therefore difficult to draw general conclusions about the role of HSP25 in cell differentiation.