The cytokeratin filament network is intrinsically dynamic, continuously exchanging subunits over its entire surface, while conferring structural stability on epithelial cells. However, it is not known how cytokeratin filaments are remodeled in situations where the network is temporarily and spatially restricted. Using the tyrosine phosphatase inhibitor orthovanadate we observed rapid and reversible restructuring in living cells, which may provide the basis for such dynamics. By examining cells stably expressing fluorescent cytokeratin chimeras, we found that cytokeratin filaments were broken down and then formed into granular aggregates within a few minutes of orthovanadate addition. After drug removal, gradual reincorporation of granules into the filament network was observed for aggregates that were either part of residual filaments or stayed in close apposition to remaining filaments. Even when cytokeratin filaments were no longer detectable, granules with low mobility were still able to reestablish a cytokeratin filament network. This process took less than 30 minutes and occurred at multiple foci throughout the cytoplasm without apparent correlation to alterations in the actin- and tubulin-based systems. Interestingly, the short-lived and rather small orthovanadate-induced cytokeratin granules contained the cytoskeletal crosslinker plectin but lacked the cytokeratin-solubilising 14-3-3 proteins. By contrast, the long-lived and larger cytokeratin aggregates generated after treatment with the serine/threonine phosphatase inhibitor okadaic acid were negative for plectin but positive for 14-3-3 proteins. Taken together, our observations in living orthovanadate-treated interphase cells revealed modes of cytokeratin remodeling that qualify as basic mechanisms capable of rapidly adapting the cytokeratin filament cytoskeleton to specific requirements.
Intermediate filaments (IFs) of the cytokeratin (CK) type are major stabilising elements of the epithelial cytoskeleton(Fuchs and Cleveland, 1998;Coulombe and Omary, 2002). The numerous genodermatoses caused by CK gene mutations are proof of this important property of the CK filament (CKF) system(Irvine and McLean, 1999). This view is further supported by analyses of knockout mice, which present variable degrees of epithelial deficiencies and increased sensitivity to various forms of stress (Galou et al.,1997; Magin,1998). It is evident, however, that CK functions extend beyond the mere structural maintenance of epithelial integrity. Thus, CKs influence the availability of regulatory molecules such as apoptosis-inducing factors, heat shock proteins or signalling molecules(Omary et al., 1998;Perng et al., 1999;Inada et al., 2001;Toivola et al., 2001). Therefore, CKs affect the sensitivity of cells to proliferative and apoptotic stimuli (Paramio et al., 1999;Paramio et al., 2001;Gilbert et al., 2001;Inada et al., 2001;Toivola et al., 2001) and play a role in cellular stress responses (Liao et al., 1995; Liao et al.,1997; Omary et al.,1998; Gilbert et al.,2001; Marceau et al.,2001; Coulombe and Omary,2002) and to drug resistance(Bauman et al., 1994).
In terms of molecular diversity, the IF polypeptides of the CK type are certainly the most complex, comprising more than 50 members, which are expressed as pairs of type I and type II isoforms in a differentiation-dependent manner (Moll,1998; Hesse et al.,2001; Coulombe and Omary,2002). In contrast to the other major cytoskeletal filaments, IFs form spontaneously and rapidly in vitro without energy consumption, auxiliary proteins or other factors (Hofmann,1998; Herrmann et al.,1999; Herrmann and Aebi,2000; Coulombe and Omary,2002). In the case of CKs, parallel heterodimers assemble in an antiparallel fashion into tetramers, which associate longitudinally and laterally into protofilaments to form the bona fide 8-12 nm IFs(Herrmann et al., 1999;Herrmann and Aebi, 2000;Parry and Steinert, 1999;Coulombe and Omary, 2002). In vivo, soluble tetrameric and larger oligomeric precursors have been identified which presumably incorporate directly into pre-existing CKFs (e.g.,Chou et al., 1993;Bachant and Klymkowsky, 1996). The molecular interactions of IF polypeptides are very strong, thus favouring the filamentous over the soluble state by far. It takes saturated urea solutions to completely break up CKFs into their monomeric subunits(Franke et al., 1983). In spite of these extreme biochemical properties, CKs must be in an adjustable equilibrium between the soluble and filamentous state in vivo to allow processes that require a less rigid and more pliable cytoskeleton.
Consequently, CKFs can not be viewed simply as rigid and immobile scaffoldings around which the cell body is arranged. On the contrary, the arrangement and dynamic properties of the CKF cytoskeleton must be in constant interplay with cellular requirements. Probably the best examined physiologically occurring process of major IF rearrangement takes place during mitosis, when CKFs either disassemble completely into soluble subunits and granular aggregates or collapse into cage-like thick filament bundles(Franke et al., 1982;Lane et al., 1982;Jones et al., 1985;Kitajima et al., 1985;Tölle et al., 1987;Windoffer and Leube, 1999;Windoffer and Leube, 2001;Windoffer et al., 2002). Protein modification and altered interactions with IF-associated proteins(IFAPs) are supposed to be major factors contributing to such CKF network remodeling. Accordingly, increased levels of phosphorylated CK polypeptides were detected in mitotic cells (Celis et al., 1983; Chou and Omary,1993; Liao et al.,1997; Omary et al.,1998) and also during meiosis(Klymkowsky et al., 1991). Phosphorylation may affect CK organization in a number of different ways: it leads to a shift in the equilibrium between the soluble and filamentous state toward the soluble form (Chou and Omary,1993); it may directly induce filament disassembly as suggested by in vitro experiments (Yano et al.,1991); and it alters the interaction with IFAPs, as demonstrated for the binding to the signaling 14-3-3 proteins, which act as solubility factors for CKs (Liao and Omary,1996; Ku et al.,1998). Furthermore, phosphorylation of the cytoskeletal crosslinker plectin may affect CK organisation, as has been shown for plectin—lamin-B and plectin-vimentin interactions(Foisner et al., 1991;Foisner et al., 1996).
Among the non-filamentous assemblies of CKs, granular aggregates are of particular relevance for the understanding of disease pathologies. For example, during chronic liver disease Mallory bodies are formed, which contain large amounts of hyperphosphorylated CK polypeptides(Franke et al., 1979;Cadrin and Martinoli, 1995;Stumptner et al., 2000). In this case, phoshorylation itself may be the cause of aggregation(Yuan et al., 1998), possibly by preventing ubiquitin-dependent degradation(Ku and Omary, 2000;Coulombe and Omary, 2002). Granular aggregates are also a hallmark of various genodermatoses(Anton-Lamprecht, 1983;Coulombe et al., 1991;Cadrin and Martinoli, 1995;Kobayashi et al., 1999). Furthermore, granules are formed in vitro in response to temperature stress or to drugs that affect the state of cellular phosphorylation(Schliwa and Euteneuer, 1979;Shyy et al., 1989;Falconer and Yeung, 1992;Kasahara et al., 1993;Liao et al., 1995;Blankson et al., 1995;Toivola et al., 1997;Strnad et al., 2001). Elucidation of the mechanisms of granule formation may reveal ways to influence their formation and thereby positively affect disease outcome.
Very little is known about the molecular principles that determine the dynamic CKF organisation in interphase cells beyond the continuous and non-selective exchange of subunits throughout the entire CKF network(Franke et al., 1984;Miller et al., 1991;Miller et al., 1993). One would expect that filaments are broken down and reassemble in various cellular domains to support ongoing cellular functions such as movement of vesicular carriers and organelles within the cytoplasm and migration of cells or their rearrangement within a complex tissue. This type of plasticity should fulfil certain requirements: it must be rapid to respond immediately to dynamic requirements; it must be reversible to prevent interference with basic CK functions; and it must be restricted temporospatially to prevent catastrophic disruption of the IF cytoskeleton and interference with its basic housekeeping functions. As a first step in the identification of such dynamic principles of CK organisation, we now show by in vivo fluorescence microscopy that the tyrosine phosphatase inhibitor orthovanadate (OV) induces alterations in the CKF network that fulfil the above-mentioned requirements. These observations give rise to the exciting possibility that transient CKF disruption is regulated through signaling pathways that involve phosphorylation.
Materials and Methods
Construction of PLC cells stably expressing fusion protein HK18-YFP
To generate a cDNA construct coding for a fluorescent CK18 chimera, we amplified a 270 bp fragment of the 3′ end of the coding region of cDNA clone pHK18-P-7 (Bader et al.,1991) by polymerase chain reaction (PCR) with amplimers 99-16 5′-CTC AAC GGG ATC CTG CTG CA-3′ and 99-17 5′-TTT GGT ACC CCA TGC CTC AGA ACT TTG GTG T-3′. The BamHI/Asp718-cleaved PCR product was cloned into pBluescript KS+ (Stratagene, La Jolla, CA) and complemented by the 1 kb BamHI 5′-fragment of pHK18-P-7, thereby generating clone pHK18Δstop. The EcoRI/Asp718 insert of this plasmid was transferred into vector pEYFP-N1 (Clontech Laboratories, Palo Alto, CA). The resulting recombinant plasmid HK18-YFP1 confers neomycin resistance and codes for the hybrid HK18-YFP, which consists of complete human CK18 that is connected to the enhanced yellow fluorescent protein by the short linker sequence GDPPVAT. The expression of this chimera is under the control of the CMV promoter.
To obtain cell lines stably expressing HK18-YFP, we used a modified calcium phosphate precipitation method (Leube et al., 1989). Transfected cells were selected using geneticin sulphate (Invitrogen Life Technologies, Karlsruhe, Germany). Surviving colonies were picked, transferred into MicrotestM Tissue Culture Plates (Becton Dickinson Labware, Franklin Lakes, NJ) and amplified for further analysis.
AK13-1 cells expressing the chimera HK13-EGFP(Windoffer and Leube, 1999)and hepatocellular-carcinoma-derived PLC cells stably expressing fusion protein HK18-YFP were grown in high glucose DMEM (PAA Laboratories,Cölbe, Germany) at 37°C with 5% CO2. In some instances,okadaic acid (Sigma, St. Louis, MO) was dissolved in DMSO at 10 μg/ml and added to cells at final concentrations between 0.1 μg/ml and 1 μg/ml. In other instances, sodium orthovanadate (OV; Aldrich Chemical Corporation,Milwaukee, WI) was dissolved in distilled water and stored as a 1 M stock. It was either used directly or incubated with H2O2 prior to usage to generate pervanadate (Feng et al., 1999). In some cases, especially for timelapse fluorescence microscopy, phenol-red-free Hanks medium was used (cf.Strnad et al., 2001). To enrich for mitotic cells, cultures were treated for 6 hours with demecolcine(Sigma) at a final concentration of 20 μM.
Cells grown on 18 mm glass coverslips to the desired density were fixed with pre-cooled methanol (5 minutes, -20°C) and acetone (15 seconds,-20°C), and mounted either directly in elvanol or subjected to indirect immunofluorescence prior to embedding as described previously(Windoffer and Leube, 1999;Strnad et al., 2001). Fluorescence was viewed in an epifluorescence microscope (Axiophot, Carl Zeiss, Jena, Germany) and was recorded with a digital camera (Hamamatsu 4742-95, Hamamatsu, Herrsching, Germany).
Murine monoclonal antibodies were used for detection of plectin (clone 10F6; kindly provided by Gerhard Wiche, Biocenter, Vienna, Austria)(Foisner et al., 1994) andα-tubulin (Amersham Pharmacia Biotech, Freiburg, Germany). Monoclonal CK epitope antibodies were generously provided by Bishr Omary and Nam-On Ku(Stanford University, Palo Alto, CA), reacting with CKs 8 and 18 (L2A1)(Chou et al., 1993),phospho-S73 of CK8 (LJ4) (Liao et al.,1997), phospho-S431 of CK8 (5B3)(Ku and Omary, 1997) and phospho-S33 of CK 18 (IB4) (Ku et al.,1998). Polyclonal antibodies from rabbits against CK5 were from Regina Reichelt and Thomas Magin (Department of Biochemistry, Bonn University,Germany) (Peters et al.,2001), against plectin from Harald Herrmann (German Cancer Research Center, Heidelberg, Germany)(Schröder et al., 1999). For detection of protein 14-3-3, isoform-ζ-specific and broad-reactive rabbit antibodies (antibodies sc-1019 and sc-629, respectively) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). To visualise bound primary antibodies, Texas-Red-conjugated goat anti-rabbit antibodies (Jackson ImmunoResearch Laboratories, West Grove, PA) and Cy3-conjugated goat anti-mouse IgGs (Rockland Laboratories, Gilbertsville, PA) were used.
For actin staining, cells were briefly washed in PBS prior to incubation with Texas-Red-coupled phalloidin (Molecular Probes, Eugene, OR) at room temperature for 30 minutes, which was followed by washing in PBS (10 minutes)and distilled water (2 minutes) before mounting in elvanol.
For statistic analysis, digital images were imported into Image-Pro Plus 4.5 (Media Cybernetics, Silver Spring, CA) to calculate Pearson's correlation coefficients. The resulting data were further analysed for statistically significant differences between groups with two-tailed Wilcoxon-W-test using SPSS software (version 9; SPSS Incorporation, Chicago, IL). The determined Pearson's coefficients were transferred to SigmaPlot 2001 (Jandel Scientific GmbH, Erkrath, Germany) to prepare Whisker-box-blots.
Cells were grown on glass slides and washed briefly in PBS prior to a 2 hour fixation in a freshly prepared solution of 2% (w/v) paraformaldehyde and 2.5% (w/v) glutaraldehyde in phosphate buffer (0.05 M Na2HPO4, 0.05 M NaH2PO4). Subsequently, cells were washed three times for 5 minutes in phosphate buffer containing 0.1 M sucrose and were postfixed for 30 minutes with osmium tetroxide [2% (w/v) in PBS]. After two 5 minute washes in distilled water,cells were dehydrated in an ascending ethanol series. Embedding was done in Epon 812 (Serva, Heidelberg, Germany). Ultrathin sections were stained with 8%(w/v) uranyl acetate for 10 minutes and contrasted with lead citrate for 5 minutes. Sections were viewed in an EM 906 (Zeiss, Oberkochen, Germany). Immunoelectron microscopy was done as described previously(Windoffer and Leube, 1999;Strnad et al., 2001) using antibodies against green fluorescent protein that crossreact with enhanced green fluorescent protein (Molecular Probes, Eugene, OR) in combination with gold-labelled secondary antibodies and the silver amplification technique.
Time-lapse fluorescence microscopy
To view living cells, Petri dishes with glass bottoms (MatTek Corporation,Ashland, MA) were used. Images were recorded by epifluorescence microscopy using inverse optics from Olympus (Hamburg, Germany) and an attached IMAGO slow scan CCD camera. The microscope was placed in a heated chamber(37°C). The whole system was controlled by TILLvisION software. Excitation was adjusted with the monochromator to 496 nm for the detection of EGFP and to 498 nm for EYFP. Image sequences were imported into Image-Pro Plus 4.5 (Media Cybernetics) and converted into movies (available at jcs.biologists.org/supplemental). Photoshop software (Adobe Photoshop 5.0) was used to edit single pictures and to assemble figures.
Gel electrophoresis and immunoblotting
High salt pellet fractions containing enriched CKs and corresponding supernatant fractions were prepared in the continued presence of OV (5 mM)prior to one-dimensional SDS PAGE or two-dimensional gel electrophoresis employing isoelectric focusing in the first dimension using established procedures (Achtstaetter et al.,1986). Separated polypeptides were either stained with Coomassie Brilliant Blue (Serva, Heidelberg, Germany) or were transferred onto nitrocellulose membranes for immunoblotting. In addition to the above-mentioned CK antibodies monoclonal antibodies against phospho-tyrosine epitopes (P-Tyr-100; New England Biolabs GmbH, Frankfurt, Germany) and CK13(Ks13.1, Progen Biotechnics, Heidelberg, Germany) were used. Detection of bound antibodies was accomplished with horseradish-peroxidase-coupled secondary antibodies (Jackson ImmunoResearch Laboratories) and an enhanced chemiluminescence system (Amersham Biosciences Europe, Freiburg, Germany).
Orthovanadate induces multifocal and reversible cytokeratin filament network disruption and granule formation
To examine the effect of the tyrosine phosphatase inhibitor orthovanadate(OV) on CKF network organization, vulva-carcinoma-derived A-431 subclone AK13-1 stably expressing fluorescent human CK13 chimera HK13-EGFP was chosen,because it is well suited for multidimensional monitoring of CKs in living cells (Windoffer and Leube,1999; Windoffer and Leube,2001; Strnad et al.,2001). Remarkably, most of the CKF network disappeared within less than 10 minutes of addition of OV, and the entire cytoplasm became filled with numerous small granules. After 2 minutes, the majority of emerging granules was in close apposition to or in direct continuity with remaining filaments,connecting them like pearls on a string(Fig. 1B,Fig. 2A,B). At this stage,cells were still outspread with only minor retractions at the cell surface as judged by phase-contrast microscopy (data not shown). Chromatin staining with Hoechst 33258 was indistinguishable from that of untreated cells (data not shown). Subsequently, fragmented structures were occasionally observed(Fig. 2C), and after 10 minutes of OV treatment, practically all cells were rounded and CKFs had completely disappeared (Fig. 1C).
In the corresponding electron micrographs, bundled filaments with local thickenings appeared (Fig. 2E). Some of the thickenings contained an amorphous core, probably representing an early state of non-filamentous granular structures (+ inFig. 2E). Many of the emerging spheroidal granules were in direct continuity with residual filaments(Fig. 2E and inset). Practically all filaments were gone in the final stages and, instead, multiple non-filamentous granules ranging in diameter from less than 200 nm to 500 nm were spread throughout the cytoplasm without apparent linkage(Fig. 2F). Immunoelectron microscopy using antibodies against EGFP confirmed that the dense bundles with local enlargements and the newly formed aggregated material contained HK13-EGFP (Fig. 2G,H).
To examine whether these drastic alterations were reversible, OV was washed out after 10 minutes. Indeed, cells restored their typical outspread morphology within 50 minutes of removal of the drug. Remarkably, all granules were gone by this time, and an extensive CKF network had been re-established(Fig. 1D). This network was finer than the typical interphase web, and the morphological appearance remained practically unchanged for the next 5 hours (data not shown).
To demonstrate that the changes in HK13-EGFP-containing filaments reflected the behaviour of the entire CKF system, the endogenous network was examined by indirect immunofluorescence in wild-type A-431 cells and AK13-1 cells. Direct comparison of CK5 immunofluorescence with HK13-EGFP fluorescence before OV-treatment, 2 and 10 minutes after addition of the drug and 50 minutes after washout showed that both patterns were practically identical(Fig. 1A′-D′). Similar results were also obtained, when antibodies against CKs 8 and 18 were used (data not shown). In most instances, however, staining of granules with CK antibodies was considerably weaker than the corresponding HK13-EGFP fluorescence. In particular, large granules were labelled only on their surface by indirect fluorescence microscopy, indicating that epitopes within granules were either altered or, more likely, not accessible to the antibodies, as it was also evident from immunoelectron microscopy(Fig. 2G,H).
The reaction of all cells in a given experiment was exceptionally uniform,and the kinetics of granule formation were similar for OV concentrations between 2 and 10 mM. However, considerably reduced OV sensitivity was often noted in cells grown in phenol-red-free Hanks medium in comparison to those maintained in high glucose DMEM [for medium dependency of OV see also(Huyer et al., 1997)]. Furthermore, OV activation by H2O2 did not affect the observed pattern of CKF reorganisation, although approximately 10-times lower concentrations of pervanadate were sufficient to cause comparable alterations. For the following experiments the non-activated form of OV was used. In some instances, cell morphology normalised even without removal of the drug,probably indicating inactivation of OV and/or acquisition of insensitivity to the drug (see also Huyer et al.,1997). Concurrently, granules disappeared to a large extent and CKFs were restored.
Orthovanadate effects on the cytokeratin filament network show no apparent correlation to alterations in the microtubule and actin system
Next, we examined whether OV-induced CKF network changes correlate with alterations of other cytoskeletal components. After 2 and 10 minutes of OV treatment, the microtubule network was still visible(Fig. 3B′,C′; see also arrowheads in Fig. 2E). However, the number of microtubules was slightly reduced, and microtubules were distributed differently, sometimes in proximity to CK granules(Fig. 2E,Fig. 3B,B′), which probably reflected only the cell shape change and not a specific association. After 2 minutes of OV incubation, an increase in actin stress fibers was noted(Fig. 3F′). Later on,actin was further concentrated in the cell cortex(Fig. 3G′). After washout of OV, both microtubules and the actin system were largely restored within 50 minutes (Fig. 3D′,H′). Taken together, these observations show that alterations in these systems do not correlate with CKF network remodeling.
Orthovanadate-induced changes in plectin distribution reflect altered cytokeratin filament network organisation and resemble alterations observed during mitosis
The distribution of the IF-binding protein plectin, which mediates interactions among all three major cytoskeletal filament networks(Foisner, 1997;Wiche, 1998), was examined next. Plectin immunofluorescence partially co-distributed with HK13-EGFP fluorescence in interphase AK13-1 cells when either monoclonal(Fig. 4A,A′) or polyclonal antibodies were used (data not shown). A significant increase in colocalisation was apparent after 2 minutes of OV addition(Fig. 4B,B′) and reached near identity after 10 minutes (Fig. 4C,C′). After removal of OV, newly formed CKFs still showed a somewhat increased co-distribution with plectin (data not shown). To quantify the changes in plectin-CK co-distribution, Pearson's correlation coefficients were determined between HK13-EGFP fluorescence and plectin immunofluorescence in untreated cells, in cells with completely disrupted CKF networks (10 minutes of OV treatment) and in partially recovered cells 50 minutes after removal of OV. For each situation, eight typical regions were chosen on the basis of DNA staining without prior knowledge of the plectin distribution. The results (Fig. 5) were fully in support of the visual evaluation, revealing a significant increase in colocalisation between plectin and HK13-EGFP both after 10 minutes of OV treatment and after an additional 50 minutes recovery period in comparison to untreated cells with a P value of <0.005. Furthermore, co-distribution was slightly lower after recovery in comparison to the time of maximum CKF disruption, although the difference was statistically insignificant (P=0.08).
To find out whether the increased colocalisation of plectin and CKs is a common feature of CKF disruption, we examined two other situations in which granular aggregates are formed. First, cells were treated with the serine/threonine-phosphatase inhibitor okadaic acid (see alsoStrnad et al., 2001). In this instance, granules did not significantly co-distribute with plectin at any time point during CKF disruption (Fig. 4D,D′), which was also evident from Pearson's coefficient analysis (Fig. 5). The P value of the analysis for okadaic-acid-treated and untreated control cells was >0.8, suggesting no significant difference between both groups, whereas the P value of the Pearson's coefficient analysis for okadaic-acid-treated cells and cells incubated for 10 minutes with OV was<0.0002, demonstrating the significant difference between both granule types. Second, double fluorescence microscopy was performed during mitosis when granular and rod-like aggregates are formed and additional strong diffuse fluorescence occurs as a result of increased soluble subunits in the cytoplasm of AK13-1 cells (Windoffer and Leube,2001). Despite the obscuring effect of the diffuse HK13-EGFP staining, a strong correlation was observed between fluorescent HK13-EGFP aggregates and plectin immunofluorescence during mitosis(Fig. 4E,E′). This was also confirmed by Pearson's coefficient analysis(Fig. 5), which produced a P value of <0.0007 for comparison with untreated cells.
Interestingly, the plectin-negative granules formed during okadaic acid treatment were positive for 14-3-3 proteins(Fig. 4I,I′), IFAPs that preferentially associate with soluble and phosphorylated CK subunits(Liao and Omary, 1996). By contrast, neither OV-induced granules nor those generated during mitosis contained significant amounts of 14-3-3 proteins(Fig. 4G,G′,H,H′). Taken together, our results show that aggregates formed during okadaic acid treatment differ from those formed during OV incubation and mitosis.
Cytokeratin granules are generated directly from cytokeratin filaments and cytokeratin filament fragments during orthovanadate treatment
Time-lapse fluorescence microscopy was performed in AK13-1 cells to find out how CK granules are formed during OV treatment. Using this method, small granules were seen to rapidly emerge from filaments(Fig. 6A). Further details of this process were difficult to image owing to its speed and the simultaneous rounding of cells, which resulted in focal shifts of fluorescent structures. For better resolution we therefore established hepatocellular-carcinoma-derived PLC cells expressing fluorescent CKFs. The cytoplasm of these cells is much more outspread than it is in A-431 cells, and the CKF network has a larger mesh size with thicker appearing filaments. A HK18-YFP chimera was stably introduced into PLC cells. The fluorescent chimeras were incorporated into a typical CKF network(Fig. 6B) and co-distributed with the endogenous CKs (data not shown). Although PLC cells were less sensitive to OV than A-431 cells, similar alterations were induced(Fig. 6C). The time until filament breakdown began was usually longer in PLC than in A-431 cells even when high concentrations of OV were used. Yet, the breakdown itself was remarkably fast once it had started and occurred homogeneously throughout the entire cell. High magnification images of distinct CKF bundles showed that they fragmented and formed small rodlets and/or granules(Fig. 6D). Interestingly, CKFs straightened prior to disruption (arrowsFig. 6D). The dynamic aspects of these processes are best appreciated in Movie 1 (available at jcs.biologists.org/supplemental),in which precursor-product relationships are unambiguously resolved.
Orthovanadate-induced cytokeratin filament network reorganisation is rapidly reversible
In a few instances, OV treatment resulted only in incomplete disassembly of the CKF network. In this case, it was possible to trace small CK granules,which disappeared again after a short time (usually within less than 30 minutes; Fig. 7). Time-lapse fluorescence imaging suggested that the granules formed as local thickenings within or next to CKFs and that they were re-integrated directly into the original filaments afterwards (arrows inFig. 7). After re-integration,the CKF network looked similar to that prior to treatment, although the cells were still slightly contracted (see Fig. 7, 38 minutes). Movie sequences taken at high frequency provided further evidence for the reversible nature of OV-induced granule formation by direct transformation into and from CKFs (Movie 2; available at jcs.biologists.org/supplemental). Furthermore, the movies demonstrated that the granules stayed in close apposition to the filaments during the entire process, suggesting a remaining connection. Movie 2 also revealed peripheral CKFs that were most probably anchored to desmosomal contact sites (arrowheads inFig. 7; see also arrowheads inFig. 6A). Loss of cell adhesion upon OV treatment resulted in retraction of these filaments, which appeared to remain in contact with the plasma membrane.
To investigate the re-formation of a CKF network from OV-induced granules in a more controlled fashion, OV was washed out after short incubation periods prior to time-lapse fluorescence microscopy. In general, the same phenomena were seen as in cells with spontaneously occurring reversion of CK granule formation. Even when most of the CKF network was disrupted and the entire cytoplasm was filled with small granules, fast reformation of an extended CKF network took place after OV washout (Fig. 8 and corresponding Movies 3,4, available at jcs.biologists.org/supplemental). Although no continuous connection between such granules was noticeable, they still appeared to be anchored either to CKFs that were below the detection limit and/or to other parts of the cytoskeleton. The CK granules completely disappeared within 25 minutes of removal of OV, and a very delicate but extensive and dense network was re-established(Fig. 8). Reformation of the network was not restricted to certain cellular subdomains but occurred at multiple sites throughout the cytoplasm. High magnification revealed further details of this process (Movie 4). Granules were seen to elongate and to fuse occasionally. They decreased in size and disappeared gradually whereas thin filaments were formed, often extending from the vanishing granules. The first visible filaments were just above background fluorescence, subsequently gaining intensity and generating a finely woven network. This newly formed network differed in appearance from the coarser network in most interphase cells (e.g., Fig. 1A,Fig. 3A,E,Fig. 4A,F). Furthermore, the network was remarkably homogeneous, lacking, most notably, the thick perinuclear filament bundles and prominent desmosome-anchored filaments. However, the peripheral part of the newly formed CKF network exhibited the typical inward-directed movement of CK fluorescence [compare Movie 3 with movies shown in (Windoffer and Leube,1999)].
The solubility and phosphorylation of the majority of cytokeratin polypeptides is not significantly altered in orthovanadate-treated cells
To examine if and to what degree CKs were altered by OV, we looked for biochemical changes in cells with completely disrupted CKF networks, as assessed by fluorescence microscopy. We found no significant increase in CK solubility in OV-treated cells in comparison with untreated cells(Fig. 9A). This was in contrast to elevated levels of soluble CKs in okadaic-acid-treated and enriched mitotic cells (Fig. 9A). Next,two-dimensional gel electrophoresis was performed to search for altered patterns of CK modification. No major differences between OV-treated and untreated cells were apparent, although some minor differences were seen, most notably for CK 5 (Fig. 9B,C). Furthermore, commercially available phospho-tyrosine antibodies did not reveal any significant reactivity of cytoskeletal fractions before or after OV incubation (data not shown). Similarly, antibodies reacting with specific phosphoserine epitopes of CKs 8 and 18 did not pick up increased phosphorylation, although some of the sites were clearly elevated after okadaic acid treatment (Fig. 9D). These observations suggest that the percentage of modified CKs in OV-treated cells is either below the detection limit of the methods used so far and/or that other factors are responsible for the observed effects.
Orthovanadate-induced changes in cytokeratin organisation serve as a paradigm for phosphorylation-dependent cytokeratin filament network regulation
We detected, for the first time, reversible CKF network breakdown in living cells as a result of transient treatment with the tyrosine phosphatase inhibitor OV. Pervanadate induced the same reversible process with identical kinetics, albeit at lower concentrations, suggesting that it acts in a similar fashion. We found that the phosphorylation and solubility of the majority of CKs was not altered at times of complete CK network breakdown. Yet, using considerably longer incubation periods of 60-90 minutes, it has been demonstrated that pervanadate leads to an increase in tyrosine phsophorylation, serine phosphorylation and solubility of CKs 8 and 19(Feng et al., 1999). It therefore remains a possibility that a minor fraction of CKs was also modified in our cells, which was, however, below the detection limit. In fact, the observed fragmentation of CKFs suggests that the modification of a very small proportion of CK polypeptides in restricted hot spots would be sufficient to disrupt the CKF network. Furthermore, phosphorylation of CKs appears to be the major principle by which CKF network organisation is regulated in vivo (e.g. Celis et al., 1993; Klymkowsky et al.,1991; Chou and Omary,1993; Stumptner et al.,2000). On the other hand, alternative explanations need to be considered given the non-specific nature of OV, which may exert its effects indirectly via IFAPs and/or other cytoskeletal components. Either way, the examination of OV-treated cells revealed physiologically relevant principles of CK dynamics, which are directly or indirectly regulated by phosphorylation. The speed of the observed granule formation and the reversibility of this process are features that would be needed for rapid adaptation of cells not only to gross cell shape changes as they occur during mitosis or migration but also to local alterations that take place throughout the lifetime of each interphase cell. Furthermore, epidermal growth factor and stress-activated protein kinases such as p38, which have been shown to affect CK phosphorylation (Aoyagi et al.,1985; Baribault et al.,1989; Liao et al.,1997; Feng et al.,1999; Ku et al.,2002), are potential and well suited regulators of restricted CKF disassembly and assembly. The dynamic on/off property of phosphorylation of CK 8 S73 and of tyrosine residues of CKs 8/19 by such enzymes(Feng et al., 1999;Ku et al., 2002) adds further support to the suitability of these mechanisms. In this way, local phosphorylation of CKs and/or associated mediators could result in a disassembly that would be reversed after removal of the phosphate group,thereby inducing temporally and spatially restricted remodeling of the CKF cytoskeleton.
Orthovanadate-induced cytokeratin filament remodeling is similar to that occurring during mitosis but is profoundly different from that mediated by okadaic acid
The effect of various modulators of phosphorylation on the dynamics and organization of the CKF cytoskeleton has been the subject of many studies(Baribault et al., 1989;Eckert and Yeagle, 1990;Cadrin et al., 1992;Falconer and Yeung, 1992;Ohta et al., 1992;Deery, 1993;Kasahara et al., 1993;Yatsunami et al., 1993;Baricault et al., 1994;Blankson et al., 1995;Toivola et al., 1997;Toivola et al., 1998;Yuan et al., 1998;Feng et al., 1999; Paramio,1999; Sanhai et al., 1999;Negron and Eckert, 2000;Strnad et al., 2001). One of the best examined drugs in this context is the serine/threonine phosphatase inhibitor okadaic acid, which induces complete disruption of the CKF system and results in formation of granular aggregates(Kasahara et al., 1993;Yatsunami et al., 1993;Chou and Omary, 1994;Blankson et al., 1995;Strnad et al., 2001). The effects of okadaic acid, however, differ in several respects from those induced by OV as reported in this communication. The okadaic-acid-induced disruption takes several hours to complete, starting only after a lag period of 1-2 hours in AK13-1 cells, and cells did not recover for several hours after removal of the drug (Strnad et al.,2001) (see also Lee et al.,1992) (P.S., R.W. and R.E.L., unpublished). Furthermore, increased levels of soluble CKs and CK phosphoepitopes were readily identified after okadaic acid treatment but not after OV. In addition, okadaic-acid-induced granules were, on the average, more than twice the size of OV granules[compare Fig. 2E-H with(Strnad et al., 2001)]. They also differ compositionally by the presence of 14-3-3 proteins and the absence of plectin (this study). Finally, in okadaic-acid-treated cells, CKFs disappear first in the cell periphery and only later in the perinuclear region(Strnad et al., 2001). Although okadaic acid is also known to induce p38(Westermarck et al., 1998;Chen et al., 2000) and to affect p38-mediated phosphorylation of IFs and IFAPs(Cheng and Lai, 1998;Chen et al., 2000;Ku et al., 2002), it must act differently from OV. We recently proposed that the effect of okadaic acid is mainly caused by prevention of integration of soluble CK subunits into CKFs,which occurs preferentially in the cell periphery(Strnad et al., 2001). The observed colocalisation of CK aggregates with 14-3-3 protein, which may keep CKs `soluble, that is, non-filamentous, further supports this notion(Liao et al., 1997). This also explains the centripetal decrease in filaments in okadaic-acid-treated cells,where cortical filament reformation(Windoffer and Leube, 1999;Windoffer and Leube, 2001)appears to be blocked, thereby resulting in a filament-free peripheral cytoplasm while perinuclear filament bundles are still intact. However, this process is rather slow and therefore not sufficient to account for the rapid CKF disruption observed in OV-treated cells and at the beginning of mitosis(Windoffer and Leube, 2001). The morphological similarities of mitotic CKF breakdown to OV-induced changes are striking, since small, filament-associated granules were found in both instances during the rapid disruption process which takes less than 10 minutes(Windoffer and Leube, 2001). Furthermore, we have shown here that mitotic granules contain, similar to OV-induced granules, plectin and lack 14-3-3 protein. Whether the underlying molecular mechanisms of granule formation and disassembly are identical in both situations remains to be determined.
Plectin is important for cytokeratin filament network reformation
We were particularly intrigued by the observed association of plectin with OV-induced CK granules. Among the various linker proteins of the plakin type that connect the cytoskeletal filament systems to each other and to certain plasma membrane domains (Ruhrberg and Watt, 1997; Fuchs and Karakesisoglou, 2001), plectin is certainly the most prominent crosslinker for the IF system (reviewed inFoisner, 1997;Wiche, 1998). Interestingly,plectin affects assembly properties of both unassembled and assembled IF polypeptides, including CKs, in a complex fashion(Steinböck et al., 2000). Furthermore, plectin is subject to dynamic phosphorylation, most notably during M phase, when increased p34cdc2 kinase-mediated phosphorylation of plectin is accompanied by its redistribution from a mostly filament-associated to a diffuse state(Foisner et al., 1996). The observed colocalisation of plectin with OV-induced CK granules was specific,since other CK-associated proteins, such as the desmosomal proteins desmoplakin and plakophilin (data not shown) and the signalling molecules of the 14-3-3 type did not co-distribute. It therefore appears reasonable to assume that the particular composition of granular CK aggregates determines their dynamic properties. Hence, given the compositional differences between okadaic-acid-induced and OV-induced granules we propose that plectin defines,in a yet unknown manner, CK aggregates that can be rapidly re-integrated into filaments. A possibility is that plectin maintains anchorage of granules to the cytoskeleton, since plectin-positive granules were rather immobile in contrast to those seen during okadaic-acid treatment(Strnad et al., 2001). On the other hand, loss of plectin and association with 14-3-3 proteins abolishes the capacity of CK aggregates to re-establish a CKF system.
Cytokeratin network formation is determined by two mechanistically different pathways
We have shown here that the CKF network recuperates after OV-induced breakdown within a short time. However, two alternative modes of CKF re-formation were noted. Most frequently, multiple cytoplasmic CK granules were seen to vanish by integrating into a fine filamentous network. This multifocal CKF assembly mechanism is highly reminiscent of experiments in which CKs were introduced into epithelial and non-epithelial cells by microinjection of epidermal poly-A+-RNA or by transfection with CK cDNA-constructs, in which multiple sites were shown to be involved in CKF network formation (Kreis et al.,1983; Franke et al.,1984; Magin et al.,1990; Bader et al.,1991). The relevance of this mechanism was further strengthened by the experiments of Miller et al. that demonstrated that microinjected CK polypeptides first formed granules together with their endogenously expressed partner polypeptides and integrated subsequently at multiple sites into the IF cytoskeleton of epithelial cells (Miller et al., 1991; Miller et al.,1993). Their images are practically indistinguishable from our photomicrographs of OV-treated cells [compare, for example,Fig. 2A-C andFig. 7 withFig. 7 of(Miller et al., 1993)]. Furthermore, ultrastructural analyses showed in both instances that the granules were composed of non-filamentous material and were in direct continuity with IF bundles. Finally, the speed of integration of granular material into filaments is comparable.
The other mode of CKF reformation after OV treatment was observed only in a minority of cells and will be presented in detail elsewhere. In this case,novel CKFs were exclusively detected in the cell periphery, which was similar to the recently described CKF network reformation after mitosis in AK13-1 cells (Windoffer and Leube,2001). This cortex-restricted mode is significantly slower than multifocal remodeling. Apparently, factors that are needed for successful cytoplasmic CKF network formation were inactivated in some OV-treated cells and are lacking after mitosis. Anchorage of CK granules to the cytoskeleton either by direct linkage to residual CKFs that are below the detection limit or to other, yet unknown, cytoskeletal components may be necessary for CKF network formation. In support, cells with multiple cytoplasmic CKF-forming foci presented granules with low mobility throughout the cytoplasm. By contrast, extremely mobile granules that were observed in the central cytoplasm during mitosis (Windoffer and Leube, 2001) and sometimes after OV treatment (data not shown)lost the capacity for multifocal CKF network reformation and were only able to contribute to cortical CKF formation. We therefore propose that anchorage of CK granules in the cytoplasm, which may be mediated by plectin, is a prerequisite for successful integration into a filamentous network. Probably cortex-dependent and multifocal cytoplasmic CK-reorganization occur simultaneously but fulfil different functions. The multifocal principle would be expected to contribute to local and rapid changes of the cytoskeleton and can be induced by specific requirements in a temporospatially restricted fashion. Any out of balance situation, such as the elevated presence of CKF polypeptides in cells that were microinjected with polypeptides or mRNA(Miller et al., 1991;Miller et al., 1993;Franke et al., 1984) or the OV-induced CKF-disruption, allows the detection of this mechanism that is otherwise difficult to see in sessile epithelial cell assemblies. The cortical principle, by contrast, may be a basic housekeeping function of epithelial cells, sustaining the continuous CKF network replacement. Accordingly, a continuous inward-directed movement of CK fluorescence has been reported in living epithelial cells (Windoffer and Leube, 1999; Yoon et al.,2001). This system would also be invoked in situations of complete disruption of the CKF network, because its associated structures are able to initiate de novo filament formation as it occurs after mitosis and after extended OV treatment (for details, seeWindoffer and Leube,2001).
Movies available on-line
The expert technical help of Antje Leibold, Sabine Thomas and Ursula Wilhelm is gratefully acknowledged. The authors thank Gerhard Wiche (Vienna Biocenter, Vienna, Austria), Harald Herrmann (German Cancer Research Center,Heidelberg, Germany), Thomas Magin, Bettina Peters (both University Bonn,Germany), Bishr Omary and Nam-On Ku (both Stanford University, Palo Alto, CA)for generous gifts of antibodies. We also thank Thomas Magin for helpful comments. The work was supported by the `Stiftung Rheinland-Pfalz für Innovation', the German Research Council (Le566/7-1) and a student grant to P.S. by the state of Rheinland-Pfalz.