|Home | About | Journals | Submit | Contact Us | Français|
Focal adhesions (FAs) located at the ends of actin/myosin-containing contractile stress fibers form tight connections between fibroblasts and their underlying extracellular matrix. We show here that mature FAs and their derivative fibronectin fibril-aligned fibrillar adhesions (FbAs) serve as docking sites for vimentin intermediate filaments (IFs) in a plectin isoform 1f (P1f)-dependent manner. Time-lapse video microscopy revealed that FA-associated P1f captures mobile vimentin filament precursors, which then serve as seeds for de novo IF network formation via end-to-end fusion with other mobile precursors. As a consequence of IF association, the turnover of FAs is reduced. P1f-mediated IF network formation at FbAs creates a resilient cage-like core structure that encases and positions the nucleus while being stably connected to the exterior of the cell. We show that the formation of this structure affects cell shape with consequences for cell polarization.
Focal adhesions (FAs), elongated and streak-like protein interfaces located at both ends of actin/myosin-containing contractile stress fibers, form tight connections between cultured cells and their underlying extracellular matrix (Abercrombie and Dunn, 1975 ; Burridge and Chrzanowska-Wodnicka, 1996 ). Aggregation of integrin heterodimeric transmembrane receptors leads to the formation of focal complexes subsequently differentiating into FAs (Rottner et al., 1999 ). In fibroblasts, focal adhesions may further evolve into fibrillar adhesions (FbAs), which are characterized by their more elongated structure, central cellular localization, enrichment in tensin and integrin α5β1, lack of tyrosine phosphorylation, and association with fibronectin fibrils (Zamir et al., 1999 , 2000 ). Over the past years, more than 150 FA components have been identified that are linked together in multiple ways to form a complex “integrin adhesome network” (Zaidel-Bar et al., 2007 ).
Plectin, a prototype cytolinker protein and classified adhesome network component (Zaidel-Bar et al., 2007 ) of enormous size (>500 kDa), was identified as an FA constituent protein nearly two decades ago (Seifert et al., 1992 ). In fact, the phenotypic analysis of plectin-deficient (plectin−/−) dermal mouse fibroblasts showed an increase in the numbers of FAs and stress fibers with consequences for cell motility (Andrä et al., 1998 ), suggesting that plectin is important for the proper functioning of FAs. However, the specific tasks of FA-associated plectin and the molecular mechanism(s) underlying its docking and FA regulation remained obscure. A peculiarity of plectin is its isoform diversity based on differential splicing of over a dozen alternative first exons into a common exon 2, giving rise to variants with differing N-terminal sequences and functions (Fuchs et al., 1999 ; Rezniczek et al., 2003 ).
Although the role of actin filaments and microtubules (MTs) in regulating FA dynamics has been addressed in a number of studies, the relevance of intermediate filaments (IFs) for FA formation and dynamics has remained largely elusive (for reviews see Geiger et al., 2001 , 2009 ; Chang and Goldman, 2004 ), except for studies indicating that in epithelial cells FAs serve as sites of keratin IF precursor formation (Windoffer et al., 2006 ). A role of plectin in mediating the interaction of vimentin IFs with FAs in fibroblasts and endothelial cells has been anticipated (Bershadsky et al., 1987 ; Seifert et al., 1992 ; Gonzales et al., 2001 ; Tsuruta and Jones, 2003 ; Kreis et al., 2005 ), albeit convincing experimental evidence is still missing.
Here we assessed the role of plectin at FAs of fibroblasts with focus on IF network–plasma membrane linkage, IF assembly dynamics, and FA turnover, using immunolocalization and live imaging techniques. We found that plectin 1f (P1f), one of several isoforms expressed in fibroblasts, preferentially associates with a subpopulation of FAs, turning them into recruitment sites for motile vimentin filament intermediates. Our analysis suggests a model for IF (vimentin) network formation where tandem fusion of FA-immobilized filament intermediates leads to IF assembly and eventual integration into a central IF network that encases and positions the nucleus and is stably anchored into centrally located FbAs.
For immunofluorescence microscopy the following primary (1) and secondary (2) antibodies were used: 1) anti-plectin isoform 1f (Rezniczek et al., 2007 ), anti-plectin isoform 1 (Abrahamsberg et al., 2005 ), anti-plectin isoform 1c (Andrä et al., 2003 ; Fuchs et al., 2009 ), and goat anti-vimentin antisera (all affinity-purified); mouse mAbs to actin (AC-40), talin, and vinculin (VIN-11-5; all from Sigma-Aldrich, St. Louis, MO), and tensin (BD Transduction Laboratories, Lexington, KY), and rat mAbs to integrin β1 (Millipore, Bedford, MA); and 2) donkey anti-goat IgG Alexa Fluor 488 (Invitrogen), donkey anti-mouse IgG Rhodamine Red-X–conjugated (Jackson ImmunoResearch Laboratories, West Grove, PA), and donkey anti-rabbit Cy5 (Jackson ImmunoResearch Laboratories). For immunoblotting anti-plectin antiserum no. 9 (Andrä et al., 2003 ) and affinity-purified anti-plectin isoform 1f antiserum (see above) were used as primary, and goat anti-rabbit IgG and goat anti-mouse IgG (both conjugated to horseradish peroxidase, Jackson ImmunoResearch Laboratories) as secondary antibodies.
Full-length mouse P1f as well as truncated P1f-8 cDNA constructs (including 5′ untranslated regions) encoding fusion proteins with C-terminal GFP have been described previously (Rezniczek et al., 2003 ). P1f and P1f-8 cDNAs were excised from pGR258 and pGR240 (Rezniczek et al., 2003 ) via EcoRI sites and subcloned into vector mCherry-N2 (pGB7; Kostan et al., 2009 ). Vector mCherry-N2 was prepared from vector mCherry-C1 (a kind gift of M. Nemethova and V. Small, IMBA, Vienna, who originally received mCherry cDNA from R. Y. Tsien, University of California, San Diego, CA). A truncated P1f-8 cDNA construct encoding fusion proteins with an N-terminal His-tag (His-P1f-8, pGR150) was generated by subcloning P1f-8 cDNA from pGR240 via EcoRI sites into vector pBN120 (Nikolic et al., 1996 ). Substitution mutagenesis of His-P1f-8 (Y20F, Y20D, and Y20E) was accomplished using the QuikChange mutagenesis system (Stratagene, La Jolla, CA) according to the manufacturer's instructions, resulting in the bacterial expression plasmids pMG19 (His-P1f-8F), pMG20 (His-P1f-8D), and pMG21 (His-P1f-8E), respectively. For expression in mouse fibroblasts, mutated cDNAs were excised via EcoRI sites and subcloned into pGR240, resulting in enhanced green fluorescent protein (EGFP)-tagged truncated versions P1f-8F-EGFP (pGB24), P1f-8D-EGFP (pMG22), and P1f-8E-EGFP (pMG23). pMG3 encoding vimentin-EGFP has previously been described (Spurny et al., 2008 ). pGB20 encoding vimentin-mCherry was obtained by amplifying vimentin-EGFP cDNA via PCR. The expression plasmids for EGFP-tensin and EGFP-zyxin (Rottner et al., 1999 ) were generous gifts of K. M. Yamada (NIH/NIDCR, Bethesda, MD) and M. Gimona (University of Salzburg, Austria), respectively. All plasmids were verified by sequencing.
Recombinant protein fragments were expressed in Escherichia coli BL21 (DE3) and purified as described previously (Rezniczek et al., 2004 ). To remove the His-tag, purified protein fragments were treated with thrombin. Actin cosedimentation assays using rabbit skeletal muscle F-actin were performed following the protocol of Kostan et al. (2009) . Proteins were incubated in 5 mM Tris, pH 8, 50 mM KCl, 2 mM MgCl2, and 2 mM ATP at room temperature (RT) for 30 min, and filaments were sedimented by centrifugation at 220,000 × g for 40 min at RT. Equivalent volumes of the supernatant and pellet fractions were analyzed by SDS-12.5% PAGE. Coomassie Brilliant Blue–stained protein bands were quantified using ImageQuant 5.1 software package (Molecular Dynamics, Sunnyvale, CA).
Immortal fibroblasts were derived from plectin wild-type (plectin+/+) and plectin-deficient (plectin−/−) p53-deficient (p53−/−) mice as previously described (Andrä et al., 2003 ) and used at passage numbers not higher than 15. Primary cells were derived in a similar manner from plectin+/+ and plectin−/− mice. Cells were cultivated and passaged under standard conditions (Spurny et al., 2008 ). For immunofluorescence and videomicroscopy, cells were spread onto 11- or 30-mm coverslips, respectively. Cells were transiently transfected in serum-free medium using FuGENE6 (Roche, Indianapolis, IN) according to the manufacturer's manual. Experiments were carried out 16–24 h after transfection. Coverslips (11 mm) were coated with Alexa 488 (Molecular Probes)-labeled fibronectin (Sigma-Aldrich) dissolved in 2 M urea/PBS at a final concentration of 50 μg/ml.
Mouse fibroblasts grown in plastic dishes (10 cm) to 70–80% confluency were washed one time with ice-cold PBS and lysed in 500 μl of ice-cold 50 mM Tris, pH 7.5, 150 mM NaCl, 2 mM EDTA, 10 mM MgCl2, 10% glycerol, 2% NP-40, 1 mM NaVO4, 0.1 mM DTT, 50 μg/ml DNAse, 20 μg/ml RNAse, and complete miniprotease inhibitor cocktail (Roche). Lysed cells were harvested using a cell scraper and incubated for 30 min on ice, and lysates were passed five times through a 22-gauge needle. Samples were mixed with 50 mM Tris-HCl, pH 6.8, 100 mM DTT, 2% SDS, 1% bromphenol blue, and 10% glycerol, and proteins were separated using standard SDS-15% PAGE. For immunoblotting, proteins were transferred to nitrocellulose membranes, which were blocked with 5% BSA in TBST (0.1% Tween 20/TBS) and incubated with primary, followed by HRP-conjugated secondary antibodies.
Cells grown overnight (16 h) were methanol-fixed, washed with PBS, mounted in Mowiol, and viewed in a laser-scanning microscope (LSM 510; Carl Zeiss MicroImaging, Thornwood, NY) at RT. Images were obtained with either a Plan-Apochromat 63×, 1.4 NA or a Plan-Apochromat 100×, 1.4 NA objective lens (Carl Zeiss MicroImaging) using LSM software.
Colocalization was evaluated by visual inspection of signal overlap on merged RGB confocal microscopy images, obtained using a 63× objective under constant acquisition settings avoiding pixel saturation. To eliminate noise and background artifacts and to smooth sharp edges of stained structures, images were processed with a 2 × 2 median filter. Structures corresponding to FAs or FbAs, displaying predominance of yellow pixels, were scored as colocalizing. To quantify coalignment of P1f with actin and fibronectin, image acquisition was performed sequentially (multitrack mode option in the LSM 510 Meta) in single plane (z = 0.3 μm) from all samples on the same day under identical conditions (laser power, gain, pinhole ~1 Airy Unit for all three channels). Acquired eight-bit images were processed as described above, and colocalization was first quantified by visual inspection of individually selected FbAs. To measure areas covered with P1f-positive structures, two best-fit lower thresholds were determined from eight-bit images of the green channel, using the threshold tool of open source software ImageJ (http://rsb.info.nih.gov/ij/; W. S. Rasband, NIH, National Institutes of Health, Bethesda, MD) and confirmed by visual inspection. For statistical evaluation of actin colocalization with P1f-positive FbAs we evaluated merged eight-bit images using the Colocalization Highlighter plug-in of ImageJ. Before colocalization analyses threshold settings for each eight-bit image were determined by careful manual thresholding and then assigned to the input window of the Colocalization Highlighter plug-in. The intensity ratio of colocalized pixels was set at 50%. Highlighted areas of maximal colocalization overlapping with previously determined FbAs (see above), were scored as colocalizing and calculated as a percentage of FbAs areas. A similar procedure was used to measure areas of FbAs (actin- and P1f-positive) overlapping with fibronectin.
Time-lapse video microscopy was implemented on an Axiovert S100TV microscope (Carl Zeiss MicroImaging) equipped with phase contrast and epi-illumination optics. Cells were spread either onto uncoated, or fibronectin-coated glass coverslips and kept in phenol red–free DMEM supplemented with 10% FCS during the whole period of observation. Observation of the cells was performed in a closed POCmini cultivation system (Carl Zeiss MicroImaging) supplemented with 5% CO2 at 37°C. Recordings started ~16 h after plating, and frames were taken with the objective lenses described above in intervals as indicated. Images were obtained using a back-illuminated, cooled CCD camera (Princeton Research Instruments, Princeton, NJ) driven by a 16-bit controller. The videomicroscopy system was automated by Metamorph 6.3 (Molecular Devices).
Plectin+/+ mouse fibroblasts were cotransfected with zyxin-EGFP and vimentin-mCherry expression plasmids. Time-lapse movies were taken over a period of 16–40 h after cell spreading with 2-min intervals between frames. ImageJ software was used to measure the fluorescent intensity of individual FAs over time. Only FAs showing a clear colocalization of zyxin-EGFP and vimentin-mCherry were evaluated. A similar number of FAs that were positive for zyxin but lacked vimentin were assessed in the same cells under similar conditions.
Cell contours of primary fibroblasts were measured 16–24 h after spreading, to determine the mean cell area, cell perimeter, shape factor, and aspect ratio. The shape factor is defined as 4π × area/perimeter2; thus, perfectly round cells have a shape factor of 1, whereas more elongated or stellate cells exhibit lower values (Soll, 1988 ). The aspect ratio is defined as the ratio of the largest and the smallest (orthogonal) diameter (Samet and Lelkes, 1994 ).
Postacquisition processing of either time-lapse movies or confocal images was performed with LSM5 Image Browser (Carl Zeiss MicroImaging), Photoshop CS2 (Adobe Systems, San Jose, CA) software package, freeware ImageJ 1.32 software and Metamorph 6.3 (Molecular Devices).
Using P1f isoform–specific antibodies and talin (or vinculin) as marker for FAs, in fully spread fibroblasts two populations of FAs could be distinguished, only one of which was associated with P1f (Figure 1A). The P1f-positive population was found mostly in central areas of the cells, whereas P1f-negative FAs were typically residing in peripheral areas. Interestingly, colocalization of P1f with many of the most prominent streak-like FAs was only partial (Figure 1B, arrows), reminiscent of “mosaic adhesions,” which were suggested to represent early stages of FbAs, before their complete segregation from FAs (Zamir et al., 2000 ).
The mosaic P1f-specific staining pattern of FAs and FbAs was confirmed by triple-staining of fibroblasts using antibodies to P1f, talin, and the FbA marker protein integrin β1 (Zamir et al., 1999 ). Integrin β1 was predominantly detected at centrally located FbAs (Figure 1C, arrows), whereas FAs at the cell periphery generally were not or less intensively stained (Figure 1C, arrowheads). As all FbAs were found to be talin-positive (Figure 1, B and C), talin is to be considered a marker protein of FbAs (in addition to integrin β1). For a statistical analysis of P1f's topological codistribution with talin, two subpopulations of FAs and FbAs were defined. Adhesions found in the core regions of cells (where also the bulk of the vimentin filament network was located; see e.g., Figure 6C), were classified as central and the rest (found in peripheral, mostly vimentin-free regions) as peripheral. Strikingly, nearly all (~98%) of the centrally located talin-positive adhesions (presumably FbAs) were found to be P1f-positive, whereas this was true for only ~12% of the adhesions classified as peripheral (Figure 1D). A quantitation of integrin β1–positive structures colocalizing with P1f yielded a value of 97.4 ± 0.7%, comparable to the value obtained for centrally located P1f-positive FbAs/FAs. These data confirm that P1f is an integral component of centrally located (integrin β1–positive) FbAs.
In FbAs, integrin α5β1 and tensin transmit actomyosin-produced tension across the cell membrane resulting in fibrillogenesis of fibronectin and rearrangement of the extracellular matrix (Pankov et al., 2000 ). By culturing wild-type (plectin+/+) mouse fibroblasts on surfaces coated with Alexa 488–labeled fibronectin, we observed a massive rearrangement of the fibronectin matrix within 16 h after seeding, with rearranged fibronectin fibrils generally codistributing with actin stress fibers and P1f spots (Figure 2A, arrowheads). Statistical analyses showed that more than 80% of P1f-positive FbAs (scored either individually after selection or as total FbA signal within an area; for details see Materials and Methods) codistribute with actin stress fibers (Figure 2B). Furthermore, 70–80% of the FbAs that were positive for both P1f and actin colocalized with Alexa 488–labeled fibronectin fibrils. To assess whether P1f affects fibrillogenesis, we subjected fibroblast cultures derived from wild-type (plectin+/+) and plectin-deficient (plectin−/−) mouse lines to a comparative quantitative analysis of external fibronectin fibril formation. However, as shown in Figure 2C, both types of cells formed similar numbers of fibronectin fibrils, suggesting that P1f is dispensable for fibronectin fibrillogenesis per se.
Triple-immunofluorescence microscopy of subconfluent mouse fibroblasts, using antibodies to vimentin, talin, and P1f, clearly revealed vimentin filaments that were targeting P1f/talin-positive FAs (Figure 3A, large arrows). A statistical evaluation revealed 27% of FAs to be vimentin-positive (our own unpublished data). In addition, in peripheral regions of the cells we observed short vimentin filament intermediates, also referred to as squiggles (Prahlad et al., 1998 ; Yoon et al., 1998 ; Spurny et al., 2008 ). Three types of such intermediates were observed. Type 1 was P1f-negative (Figure 3A, small arrows), type 2 (not shown) was partially, and type 3 heavily (all along the filaments) decorated with antibodies to P1f (Figure 3B, arrowheads). Only type 3 intermediates colocalized with talin-positive FAs (Figure 3B, arrowheads).
To monitor vimentin filament dynamics and P1f-FA association, we performed time-lapse video microscopy of mouse fibroblasts cotransfected with cDNA encoding vimentin-EGFP and full-length (C-terminally tagged) P1f-mCherry. Similar to endogenous P1f (see Figure 3A, large arrows), overexpressed P1f-mCherry colocalized with FA-like structures (Figure 3C, arrowheads within magnified images, which represent frames 0′, 7′, and 11′ taken from Supplementary Movie 1 in the Supplementary Materials). In fact, the videomicroscopy data conveyed the impression that vimentin filaments were attached (and thus immobilized) to peripheral P1f-mCherry–positive FAs, whereas those parts of the filaments that interconnected these FAs (Figure 3C, arrows) were flexible, swaying in a wobbling manner between their points of anchorage (see Supplementary Movie 1 in the Supplementary Materials). These observations suggested that the vimentin network, albeit being pinned down at FAs, retained its ability to form flexible links between its anchoring sites.
In contrast to plectin+/+ cells, in plectin−/− fibroblasts, we could not observe FA-targeting of vimentin filaments or any physical connection between filament bundles and FAs (arrowheads in Figure 4A). Moreover, the vimentin network was not restricted to the central part of the cells like in plectin+/+ cells (compare with Figure 6C), but filaments were extending to the outermost boundary of the cells, where they often were found to bend backward toward the cell interior (Figure 4A, arrows). A statistical evaluation showed that ~85% of lamellipodial and filopodial areas of plectin−/− cells to be intruded by vimentin filaments, compared with only ~10% in plectin+/+ fibroblasts (Figure 1E).
To assess whether the loss of vimentin filament attachment to FAs in plectin−/− cells could be rescued by forced expression of P1f, cells were cotransfected with cDNA constructs encoding C-terminally EGFP-tagged full-length P1f (P1f-EGFP) and vimentin-mCherry. As shown in Figure 4B, P1f-EGFP was targeted to FAs where it showed colocalization with vimentin filaments (arrows). In general, vimentin filaments appeared to coalign with elongated patches of P1f that presumably were oriented along actin stress fibers (Figure 4B, arrowheads). These patches resembled those of endogenous P1f localized at FbAs, as visualized by immunofluorescence microscopy. Because overexpressed vimentin and P1f clearly showed colocalization and thus appeared to be physically connected, we concluded that forced expression of P1f in plectin-negative cells results in the reattachment of vimentin filaments to plectin-positive FAs.
Examining whether vimentin filament reattachment to FAs required the expression of full-length P1f, in particular of its C-terminal IF-binding domain (Nikolic et al., 1996 ), we transfected plectin−/− fibroblasts with a cDNA construct that encoded an N-terminal fragment of P1f, comprising just its isoform specific (exon 1f–encoded) sequence and the succeeding actin-binding domain (ABD), fused to EGFP (P1f-8-EGFP; Rezniczek et al., 2003 ). Confocal microscopy of P1f-8-EGFP–expressing cells revealed vimentin filaments similar in appearance to those of nontransfected cells (Figure 4C, arrows; compare with Figure 4A). In peripheral regions of the cell these filaments were clearly overshooting P1f-8–positive FAs (Figure 4C, arrowheads), and the only hindrance for their extension appeared to be the boundaries provided by the plasma membrane. This observation was confirmed by videomicroscopy of P1f-8-EGFP/vimentin-mCherry coexpressing fibroblasts. Partially looped, arced, and bundled vimentin filaments were entering filopodia, freely extending and retracting from its distal tip (Supplementary Figure 1 and Supplementary Movie 2 in the Supplementary Materials, uppermost arrow). These data indicated that P1f-8-EGFP, albeit being targeted to FAs, was unable to mediate the physical linkage of vimentin filaments to FAs. In all, these rescue experiments provided strong evidence for P1f physically joining parts of the vimentin network to FAs and FbAs.
Previous cell transfection experiments using cDNA expression plasmids encoding an EGFP-tagged version of plectin's ABD without preceding isoform-specific sequences indicated a rather diffuse distribution of the fusion protein with weak stress fiber association (Rezniczek et al., 2003 ), but no association with FAs. An EGFP fusion protein containing just the isoform 1f–specific sequence without the ABD (P1f-1st exon-EGFP) showed a similar distribution upon expression in fibroblasts, and only when both domains were combined (P1f-8-EGFP) association with FAs was observed (Figure 5B).
It has been shown that tyrosine phosphorylation of the sequence preceding the ABD of α-actinin results in dissociation of the protein from FAs and prevents its binding to actin (Izaguirre et al., 2001 ). Investigating similar mechanisms for P1f, we assessed whether tyrosine residues residing in P1f's isoform-specific (first exon–encoded) sequence could influence the binding properties of its ABD. On the basis of in silico predictions (Scansite) of putative phosphorylation sites, we replaced the tyrosine residue at position 20 (Tyr 20) in P1f's first exon-encoded sequence by either phenylalanine, glutamic acid, or aspartic acid (Figure 5A) and transfected the corresponding N-terminal cDNA constructs (P1f-8F-EGFP, P1f-8D-EGFP, and P1f-8E-EGFP) into fibroblasts. Although we never found Glu and Asp mutants (mimicking phosphorylated tyrosine) to be targeted to FAs, the Phe mutant clearly was confined to FAs (Figure 5C).
To measure actin-binding activities of the mutated P1f-8 variants, recombinant versions of the proteins were purified from bacterial cell lysates and subjected to actin cosedimentation assays (Figure 5, D and E). In accordance with the data obtained by forced expression in fibroblasts (Figure 5C), Glu and Asp mutants displayed significantly reduced actin-binding capacities, whereas the Phe mutant protein showed similar, or even higher binding compared with wild-type P1f-8 (Figure 5, D and E). In vitro binding and overexpression data combined would support a model where P1f is recruited to FAs via direct binding to actin with the preceding P1f isoform–specific sequence regulating or fine tuning its binding affinity.
We recently found that vimentin filament assembly during cell spreading of unpolarized plectin−/− fibroblasts proceeds without the formation of granular and short filamentous intermediates that are typical for plectin+/+ cells (Spurny et al., 2008 ), and we made a similar observation in migrating (polarized) fibroblasts (our own unpublished data). As shown in Figure 6A, the forced expression of P1f in polarized plectin−/− cells triggered the formation of filamentous intermediates in their lamellipodia, thus reversing the phenotype. Interestingly, just as in plectin+/+ cells (compare to Figure 3A), these structures were either positive or negative for P1f (Figure 6A, region of interest [ROI], arrows and arrowheads, respectively).
We then assessed whether the association of P1f with vimentin filament intermediates was isoform-specific. Immunoblotting of cell lysates from primary mouse fibroblasts using isoform-specific antibodies revealed expression of plectin isoforms 1c (P1c) and 1 (P1), in addition to that of P1f (Figure 6B), and as previously reported (Winter et al., 2008 ) these cells also express plectin 1b (P1b), a mitochondrion-associated variant. When cells were triple-immunostained for talin, vimentin, and either P1 or P1c, both isoforms showed colocalization with the filamentous vimentin network in the central part of the cell. However, in peripheral regions of the cells only P1c colocalized with filament intermediates, whereas P1 showed a diffuse distribution (our own unpublished data). P1c staining was found to partially overlap with FAs (Figure 6C, arrows in magnified ROI). Statistical analyses showed that on average, 47% of squiggle-like intermediates observed in a cell were associated with P1c, but only 14% with P1f (Figure 6D, first pair of bars). Strikingly, among the FAs that did not colocalize with filament intermediates, 10% were positively stained for P1f, but only 1% for P1c (Figure 6E). The proportions of P1f- or P1c-positive filament intermediates found associated with FAs were similar, i.e., 9 and 13%, respectively (Figure 6D, second pair of bars). Coinciding triple-staining of filament intermediates (vimentin), FAs (talin), and plectin isoforms was slightly biased toward P1f (14%) compared with P1c (9%; Figure 6D, third pair of bars). These data would be consistent with a model where P1c is targeted to FAs in conjunction with vimentin filament intermediates, whereas P1f can associate with FAs independently of vimentin.
The association of P1c with filament intermediates raised the question whether it too, could induce their formation or whether this was a P1f-specific trait. When full-length P1c was expressed in plectin−/− cells, similar to the case of P1f (Figure 6A), filament intermediates became clearly visible (Figure 6F). P1c's potential to rescue this phenotype appeared to be lower (~50%) than that of P1f (Figure 6G), but whether this has any biological significance remains to be shown. However, in contrast to P1f, P1c did not restore vimentin filament targeting to FAs (Figure 6G). In fact reminiscent of plectin-null cells expressing P1f-8 (Figure 4C), the P1c-expressing cells displayed partially bundled filaments that extended to the periphery of the cell (Figure 6G, arrowheads) without being captured at FAs (Figure 6G, arrows). This lends further support to the notion that FA anchorage of IFs is a unique feature of P1f.
To monitor the dynamics of vimentin filament intermediates and the accumulation of P1f at FAs in live cells, we cotransfected plectin+/+ fibroblasts with plasmids encoding full-length P1f and vimentin EGFP- or mCherry-fusion proteins. Consistent with published data (Yoon et al., 1998 ), time-lapse video microscopy revealed that the majority of traceable intermediates was motile and moving in an anterograde manner toward the cell periphery. Out of five movies inspected, immobilized intermediates were generally found associated with P1f-positive FAs (Figure 7A, arrowheads, and Figure 6D), whereas motile intermediates (Figure 7B, short arrow) were never found associated with P1f. However, filament intermediates moving toward the cell periphery were often seen to eventually fuse with larger vimentin filaments that were already immobilized at FAs (Figure 7B, long arrow; see also Supplementary Movie 3 in the Supplementary Materials). Forced expression of EGFP-tagged vimentin occasionally led to polymeric structures resembling granules rather than filaments. Intriguingly, many of these nonfilamentous structures were found aggregated and associated with P1f, presumably residing at FAs (Supplementary Figure 2, arrows; see also Supplementary Movie 4 in the Supplementary Materials). On the other hand, vimentin granules that were not associated with P1f appeared to be distant from FAs (Supplementary Figure 2, arrowheads; see also Supplementary Movie 4 in the Supplementary Materials). In general we noticed that the P1f- and the vimentin-specific signals matched each other in intensities whenever vimentin granules were found in association with FAs (Supplementary Figure 2, see also Supplementary Movie 4 in the Supplementary Materials) and were therefore scored as colocalizing (for details see Spurny et al., 2008 ). Taken together these observations indicated that vimentin filament intermediates became immobilized at FAs, with immobilization being P1f dependent.
Videomicroscopy of filament intermediates in plectin+/+ fibroblasts expressing EGFP-vimentin, revealed end-to-end fusion of short vimentin filament intermediates in tandem, resulting in filament elongation (see Supplementary Figure 3 and Supplementary Movie 5 in the Supplementary Materials). Intriguingly, such fusions invariably required one immobilized short filament acting as a docking site for subsequently delivered motile intermediates. Albeit fusion of small vimentin (and peripherin) filaments in tandem has been hypothesized before (Helfand et al., 2003 ; Chang et al., 2006 ), only recently has it been demonstrated experimentally (Colakoglu and Brown, 2009 ).
Next we assessed whether vimentin filament association of FAs affected their dynamic properties. Monitoring the lifetimes of FAs with or without associated vimentin filaments within single cells cotransfected with vimentin-mCherry and EGFP-tagged zyxin, a FA marker protein (Webb et al., 2004 ), we observed a striking difference. Although for FAs that were associated with IFs the average lifetime was 240 min, it was only 87 min for those that were not (Figure 8A). Thus, the recruitment of vimentin filaments to FAs through P1f clearly led to a reduction in the turnover and thus stabilization of FAs. Interestingly, the structurally related cytolinker protein ACF7 has recently been reported to exert the opposite effect, i.e., destabilization of FAs, by mediating MT-targeting to FAs in keratinocytes (Wu et al., 2008 ).
To determine whether the alterations in IF network organization, including its detachment from FAs affects the cellular geometry of fibroblasts, we measured the area and perimeter of primary fully spread plectin+/+ and plectin−/− cell cultures (Figure 8B). Although there was no difference observed for the average cell area of either cell type (~2300 μm2; Figure 8Bi), surprisingly, the average perimeter of plectin+/+ fibroblasts was significantly larger then that of plectin−/− cells (348 and 299 μm, respectively; Figure 8Bii), suggesting that plectin+/+ cells had more protrusions. When the polarization of cell bodies was quantified by calculating the shape factor (SF; see Materials and Methods) and the aspect ratio (AR, defined as the ratio of the largest diameter and the smallest diameter orthogonal to it), both parameters indicated a significantly more polarized shape of plectin+/+ (SF 0.28, AR 0.33), compared with the more rounded plectin−/− cells (SF 0.4, AR 0.45; Figure 8, iii and iv, respectively). Similar results were obtained when immortal (p53-deficient) plectin+/+ and plectin−/− fibroblasts (instead of primary cultures) were subjected to similar morphometric analyses (Figure 8C, i–iv). This enabled us to use transient transfection of plectin−/− fibroblasts with full-length plectin-encoding expression plasmids for examining whether the observed changes in cell geometry were directly connected to the deficiency in a particular isoform. On forced expression of P1f, P1c, and P1, the three major isoforms normally present in this type of cells (Figure 6B), only P1f-expressing cells were assuming a more polarized shape (SF 0.21, AR 0.24), compared with the nontransfected (rounder) cells (SF 0.45, AR 0.64). The restored SF and AR values in P1f-transfected cells both came very close to those determined for wild-type fibroblasts (SF 0.23, AR 0.26; Figure 8C, iii and iv, respectively). Cells expressing P1c or P1 showed hardly any restoration. This strong and specific rescue potential of P1f clearly demonstrated a direct link between the absence of this isoform and the alterations observed in fibroblast geometry.
We report here that mature FAs and FbAs act as anchorage sites for vimentin IFs, with P1f being the crucial linker protein. Moreover, we show that P1f serves as a docking site and assembly center for the de novo formation of IF networks and that IF-linked FAs have a longer half-life (and thus are more stable), compared with FAs without IF contact.
The molecular mechanism responsible for the recruitment of P1f to mature FAs of fibroblasts remains elusive. Based on knockdown and MnCl2-FA modulation experiments, recently integrin β3 has been proposed to play a role in the recruitment of plectin and vimentin IFs to FAs in endothelial cells (Bhattacharya et al., 2009 ). As one of the well established FAs markers (Cukierman et al., 2002 ), integrin β3 could play a similar role in P1f-FA targeting in fibroblasts. However, the situation in fibroblasts must be different, because our results clearly show that P1f is essential for the association of IFs with FAs and FbAs, whereas plectin in endothelial cells was found to be without apparent impact on IF–FA interaction (Bhattacharya et al., 2009 ). Moreover, as FbAs evolve from FAs, integrin αvβ3 complexes are excluded and remain with FAs (Pankov et al., 2000 ). Thus it seems that in fibroblasts integrin β3 is not involved in P1f or IF docking at FbAs.
In previous attempts we were unable to demonstrate direct binding of P1f to a number of typical FA components other than actin. Therefore we consider it a realistic option that P1f is targeted to FA-associated actin via its multifunctional N-terminal ABD (Sevcik et al., 2004 ). However, as shown here and in part in a previous study (Rezniczek et al., 2003 ), fragments corresponding to just the isoform-specific sequence, or the ABD alone, are not targeted to FAs when expressed in fibroblasts. FA targeting occurs only when the isoform-specific sequence and the ABD are combined, suggesting that they act cooperatively. Consistent with functional cooperativity is also our finding that a single amino acid substitution within the exon 1f–encoded sequence not only can have dramatic effects on the cellular localization (in particular FA association) of P1f's N-terminal fragment, but also affects its actin-binding capacity. Plectin's ABD gaining a specific function when combined with exon 1f–encoded sequences would have a precedent in the case of plectin isoform 1a (P1a). Isoform-dependent binding of the Ca2+-sensing protein calmodulin to P1a's ABD has recently been shown to not only cause P1a's dissociation from integrin β4 clusters, but also prevent its actin binding (Kostan et al., 2009 ).
Thus it is an intriguing possibility that P1f, being part of an intricate mechanoresponsive machinery of the cell, might act either alone or in combination with yet unidentified binding partners, as a sensor for distinct F-actin conformations within FAs. This would explain the frequently observed uneven or patchy distribution of P1f at FAs and FbAs. Once bound, P1f may stabilize the actin molecule in its particular conformation and thereby conserve a tensional state of actomyosin required for keeping up FA stability. A stabilizing function such as this would be very much in line with the observation that P1f/vimentin-associated FAs displaying a longer life time than P1f/vimentin-negative FAs, and it would confirm previous studies showing that actin-binding proteins can influence the twist and tilt of F-actin with consequences for actin filament stability (Galkin et al., 2001 ). Even if the molecular mechanisms underlying plectin targeting and IF anchorage in fibroblast and endothelial cells may turn out to be different, our findings are consistent with the observation of reduced dynamics and increased numbers of stable FAs correlating with increased vimentin-FA association in endothelial cells exposed to shear stress (Tsuruta and Jones, 2003 ), and the decreased shear stress resistance of endothelial cells after vimentin knockdown (Bhattacharya et al., 2009 ).
Associating with vimentin from the early stages of filament assembly, plectin is required for the formation of filament intermediates and their directional movement toward the cell periphery, as well as the stepwise formation of robust filament networks in fibroblasts (Spurny et al., 2008 ). The present live cell imaging data clearly demonstrated that the movement of vimentin intermediates came to a stop at P1f-positive, but not P1f-negative FA, and that subsequent end-to-end fusion of such already immobilized with newly approaching (mobile) filament intermediates led to the formation of immobile longer filaments that eventually became fused to the already existing vimentin network. IF precursor formation next to lamellipodial FAs of epidermal cells has previously been shown also for keratin filaments (Windoffer et al., 2006 ), albeit in this case subsequent filament formation was suggested to be dependent on the actin filament system rather than MTs (Kölsch et al., 2009 ).
As we found fibronectin fibrillogenesis and FbA formation not to be compromised in plectin−/− fibroblasts, we suggest that in plectin+/+ fibroblasts, P1f and its attached de novo formed vimentin filaments are translocated together with integrin α5β1 and tensin to centrally located FbAs where they become integrated into the centrally located IF network surrounding the nucleus. A model depicting the putative role of P1f at the various stages of this process is shown in Figure 9A. IF intermediates transported toward FAs along MT tracks are unrestrained by FAs in their movement as long as they lack P1f (Figure 9Ai). A “conformational switch” (Geiger et al., 2001 ) in the multiprotein FA complex itself could be the physiological trigger for the recruitment of P1f to FAs (Figure 9Aii). This switch could be the result of intrinsic forces triggered by molecular components of the Rho pathway, or it might happen in response to extracellular cues such as ECM rigidity or osmotic stress. Once targeted to FAs, P1f blocks the movement of vimentin filament intermediates along MTs and thus provides an assembly platform for the de novo assembly of longer filaments by end-to-end fusion of intermediates (Figure 9Aiii). Translocation of P1f together with its attached vimentin filaments during FbAs formation and incorporation of the filaments into the perinuclear IF network (Figure 9Aiv) then leads to the formation of a central cagelike core structure stabilized by the vimentin network (Figure 9B).
Contrary to plectin+/+ cells, in plectin−/− fibroblasts, we found vimentin filaments to overshoot peripheral FAs without being captured and to reach far into lamellipodial protrusions and even to the very periphery of the cells. As we show here, this more extended and less compact IF network correlates with a more rounded cell shape and fewer protrusions of the plectin−/− cells. Interestingly, the collapse of the IF network in fibroblasts after induced disruption (Goldman et al., 1996 ) or vimentin silencing with shRNA (Mendez et al., 2010 ) led to similar phenotypes. Similarly, the reduction in vimentin and GFAP (glial fibrillary acidic protein) expression led to suppression of process formation and reduced polarization of glial cells (Lepekhin et al., 2001 ). Although the roles of the actin and MT cytoskeleton in cell polarization are relatively well established (for a review see Ridley et al., 2003 ), the importance of the IF system in this process is less clear. We postulate that a well-anchored and stable perinuclear IF cage is required for front–rear polarization, similar to stable MTs (Pegtel et al., 2007 ). As the shape of cells is mechanically stabilized by balanced forces within the cytoskeleton and between cytoskeletal components and the extracellular matrix (Ingber, 2003 ; Polte et al., 2004 ), the recently described reduced stiffness and cytoskeletal tension in plectin-negative fibroblasts (Na et al., 2009 ) could be a consequence, at least in part, of unanchored vimentin IFs compromising cell shape and polarity. Furthermore, vimentin IFs were only recently shown to interact with Scrib, a well-established regulator of cell polarity, and silencing of either vimentin, Scrib, or both, were shown to affect polarization of endothelial cells (Phua et al., 2009 ). These data and the ones reported here lend strong support to an emerging new role of IFs in cellular shape determination and polarization (Chang and Goldman, 2004 ; Kim and Coulombe, 2007 ). As altered cytoarchitecture in plectin−/− cells is paralleled by perturbations of signaling pathways (Osmanagic-Myers and Wiche, 2004 ; Osmanagic-Myers et al., 2006 ), it will be a challenging task to unravel the mechanistic links between plectin-dependent cell shape changes and polarity-related processes such as cell migration.
We thank M. Nemethova and V. Small (IMBA, Vienna), R. Y. Tsien (University of California, San Diego, CA), B. Geiger (Weizmann Institute of Science, Rehovot, Israel), K. M. Yamada (NIH/NIDCR, Bethesda, MD), M. Gimona (University of Salzburg, Austria), and P. Traub (University Bonn, Germany) for generously providing materials. This work was supported by Austrian Science Research Fund Grants P20744-B11 and I 413-B09 (a subproject of Multilocation DFG-Research Unit 1228).
This article was published online ahead of print in MBoC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E10-02-0094) on August 4, 2010.