Intermediate filaments: dynamic processes regulating their assembly, motility, and interactions with other cytoskeletal systems

Intermediate filaments (IF) are major cytoskeletal proteins of animal cells. They are comprised of a large family of >50 gene products that are capable of polymerizing into ∼10 nm diameter IF. The rules for IF polymerization from these protein subunits are complex. Some can form IF from a single type of protein chain (homopolymers), while others can form IF only when they are combined with more than one type of protein chain (heteropolymers). Within the cytoplasm of cells, IF form elaborate and complex networks. These are usually concentrated in the perinuclear region and radiate throughout the cytoplasm, where they are frequently associated with the cell surface (1, 26). The extent of sequence homology, the pattern of cell type specific expression, and the similarity of intron positions of their genes are properties that have been used to classify IF proteins into six different types (1–3). The largest number of IF proteins are categorized as the type I and II keratins, which are expressed mainly in epithelial cells. The keratins are obligatory copolymers that require both a type I and a type II protein to form IF. The expression of specific keratin pairs is precisely regulated during development, and this unique feature has led to their use as differentiation markers for epithelial cells. Vimentin, desmin, glial fibrillary acidic protein (GFAP), and peripherin are the four known type III IF proteins. Each of these type III proteins is capable of assembling into homopolymer IF. Among them, vimentin is unique in that it often forms a scaffold IF network before the expression and assembly of differentiation-specific IF proteins such as desmin, GFAP, and peripherin. Type IV IF proteins consist of the three neurofilament subunit proteins, NF-L, NF-M, and NF-H, as well as α-intenexin. Neurofilaments are obligatory heteropolymers, while α-internexin is self-assembly competent (4). Although the majority of IF proteins are expressed in the cytoplasm, the type V IF proteins (the nuclear lamins) are localized in the nucleus where they polymerize to form the nuclear lamina. The lamina is located at the interface between chromatin and the inner face of the nuclear envelope membrane. Recent observations demonstrate that the nuclear lamina is not only essential for the structural integrity of the nuclear envelope but is also involved in the process of DNA replication during S phase of the cell cycle (5). The type VI IF protein nestin has not been characterized extensively. It was discovered as an early developmental marker of neuroepithelial cells (3), which cannot form IF on its own but is capable of coassembling with vimentin to form IF (28). The IF proteins possess both well-conserved and nonconserved domains. The most conserved region is the central α-helical rod domain consisting of ∼310 amino acids, known to form coiled-coils in assembled IF. This central rod is flanked at both ends by NH2- and carboxyl-terminal non-α-helical regions that are not as well conserved among the various types of IF proteins. It has often been speculated that these less conserved end domains may be involved in the cell type specific functions of IF, as well as their higher order structure. Within the rod domain, two stretches of sequence are highly conserved that are known to play important roles in IF assembly (1). They are located at the beginning (the IA region) and at the end (the 2B region) of the rod. Although the details of IF structure remain to be determined, the most widely accepted model is that IF are comprised of four protofibrils. Individual protofibrils contain two protofilaments, each of which is constructed from linear arrays of antiparallel and staggered dimers. The basic building block is the dimer, consisting of two parallel and in register protein chains (6). IF are known to provide mechanical integrity to cells. In vitro they exhibit unusual viscoelastic properties making them much more resistant to mechanical stress when compared with microtubules and actin-containing microfilaments (7). Their mechanical role has been demonstrated at the tissue level in vivo by the finding that numerous human blistering diseases are caused by point mutations in epidermal keratin genes (8). This is also supported by studies in which targeted expression of dominant negative keratin mutations in mice results in severe blistering of the skin. The latter lesions appear to be caused by the weakening of the mechanical properties of keratin IF in epidermal keratinocytes, with concomitant loss of cell shape and cell lysis (8). In the case of type III IF, it has been shown in mice that ablation of the desmin gene leads to the degeneration of cardiac and skeletal muscle, suggesting that desmin plays an important role in maintaining the mechanical integrity of muscle cells (9). In a similar fashion, defective expression of neurofilaments in animal models results in the aberrant organization of neurofilament networks, which is sufficient to generate phenotypes reminiscent of neurological diseases (10). IF are closely associated with various types of junctional complexes at cell surfaces, including desmosomes and hemidesmosomes in epidermal cells (11) and focal adhesions in fibroblasts (12). These associations are mediated by a growing family of proteins known as intermediate filament associated proteins (IFAPS; ref 13). These include plectin, desmoplakin, and BPAG1n/dystonin (13). IFAPs have also been shown to play an important role in the interactions among cytoskeletal systems. For example, plectin (14) BPAG1n/dystonin (15) and fimbrin (32) possess both IF and actin binding sites (14, 15). The importance of IFAPs in cytoskeletal integrity has been highlighted by studies of BPAG 1-null mice and patients afflicted with muscular dystrophy associated with epidermolysis bullosa simplex (MD-EBS). In the case of MD-EBS, the effects are linked to the expression of truncated plectin molecules (14, 16). Based on their biochemical properties in vitro and their known roles in contributing to the mechanical properties of cells, many researchers have assumed that IF are static elements of a cell's cytoskeletal repertoire. However, it has become increasingly obvious over the past decade that IF are dynamic structures. For example, dramatic changes in IF network organization occur in mitotic cells and in spreading cells after cytokinesis or, in the case of cultured cells, after trypsinization and replating (25). In addition, the nuclear lamina is dismantled and dispersed throughout the cytoplasm during mitosis and is reassembled during the reformation of daughter cell nuclei (5). Studies designed to determine the properties of IF during interphase have been carried out using microinjection techniques. Experiments involving the microinjection of soluble biotinylated vimentin and type I keratin subunits have revealed that endogenous polymerized IF can incorporate exogenous subunits (17, 18). However, if the amount of micro-injected type I keratin is above a certain concentration, there is a rapid induction of the disassembly of endogenous keratin tonofibrils. Presumably, this is because of the formation of types I and II heterodimers, which are required for keratin IF polymerization. Based on these findings, we have speculated that an excess of type I protein could effectively remove type II from an exchangeable pool of subunits, thereby displacing additional subunit proteins from keratin IF, ultimately resulting in their disassembly (17). These microinjection results suggest that there is an equilibrium state between exchangeable subunits and polymerized IF. By far the most convincing evidence supporting the dynamic properties of IF in vivo has been derived from fluorescence recovery after photobleaching (FRAP) experiments. These experiments on IF in living cells were initially carried out following the microinjection of rhodamine-conjugated vimentin (19). The results show that fluorescent vimentin fibrils are capable of recovering their fluorescence after photobleaching. This fluorescence recovery appears to be uniform all along the bleach zones, indicating that in vivo IF are apolar filaments. Unfortunately, in these studies only the initial phases of recovery could be monitored because of the low level of fluorescence emission of the rhodamine-tagged vimentin fibrils. This made it impossible to measure the t1/2 for full recovery (19). More recently, we have been able to track total recovery of fluorescence in bleach zones across vimentin fibrils tagged with green fluorescent protein (GFP)-vimentin (see below; refs 20, 21). To further our understanding of the properties of IF in vivo, we found it essential to develop a technique that would allow us to study IF for prolonged time periods. This has involved the use of GFP-tagged vimentin. Live cells transfected with GFP-vimentin contain IF networks that are indistinguishable from fixed/stained cells studied by immunofluorescence. Furthermore, we have found that these transfected cells can be observed for intervals of more than 1 h (21). The results of our initial studies using GFP-vimentin reveal that the constituents of IF networks are much more dynamic than previously thought. This has been shown by time-lapse confocal microscopy, which demonstrates that individual vimentin fibrils alter their shape, change their length, and either appear or disappear in various regions of the cytoplasm. IF shorten at mean rates of 0.42 µM/min and elongate or extend at average rates of 0.4 µM/min (Fig. 1a–f). Frequently, neighboring vimentin fibrils translocate at different rates and sometimes in opposite directions. This is indicated by the different rates of movements of photobleach marks that average 0.24 µm/min (Fig. 1g–i; ref 21). When GFP-vimentin expressing cells are treated with nocodazole, the average rate of these movements is reduced by 60%, and the same is true of cytochalasin B. These results indicate that the translocation of IF is somewhat dependent on both microtubules and microfilaments (21). Our measured rates of IF bleach zone translocation compare favorably with those that have been measured for microtubules and microfilaments using photobleaching and photoactivation, although the underlying mechanisms are probably different (e.g., treadmilling for microtubules and microfilaments). These techniques have revealed that the rates for translocation along kinetochore microtubules range from ∼0.12 to 0.7 µm/min (22, 23). In the case of actin-containing stress fibers, average rates of 0.29 µm/min have been recorded (24). As for actin-containing fibrils in la-mellipodia, a faster speed of 0.79 µm/min has been reported (27). In addition to the above, rapid movements of short vimentin IF, termed ‘squiggles’, are observed. These are seen primarily in the peripheral regions of the cytoplasm where they translocate at an average rate of 3.3 ± 1.9 µm/min with speeds ranging from 1.3–11.4 µm/min (Fig. 1j–l). These latter movements are almost completely inhibited when microtubules are disassembled with nocodazole (21), indicating the involvement of microtubule-based motors (see below). The opportunity to study individual vimentin fibrils in cells transfected with GFP-vimentin for longer time intervals in vivo has also permitted us to calculate the t1/2 for full recovery of fluorescence after photobleaching. FRAP analyses of GFP-vimentin fibrils yield a t1/2 = 5 ± 3 min. This rate is similar to those obtained for microtubules and microfilaments in vivo (21), indicating that the speeds of subunit/polymer exchange for all three cytoskeletal elements are comparable. The assembly of IF in vitro has been studied extensively, and the accumulated data suggest that the process requires no energy and no accessory proteins. However, details of the in vivo assembly process remain largely unknown. To begin to investigate this process, we have monitored the behavior of GFP-vimentin by time-lapse confocal microscopy in spreading BHK-21 cells following trypsinization and replating. During the first 30–45 min of spreading, the majority of the GFP-vimentin is localized as a filamentous aggregate in the perinuclear area (25). However, in the peripheral regions of the cell, most of the GFP-vimentin is present as nonfilamentous vimentin particles (also called ‘dots’ (20); Fig. 2a, b). Within 1–1.5 h after replating, the vimentin particles in the periphery appear to be transformed into the short fibrous squiggles (see above; ref 21). After 3 h, as cells spread even further, the number of particles and squiggles decreases, apparently replaced by the long vimentin fibrils that typify the IF networks of spread cells. From these time-lapse observations, it appears that the assembly of the vimentin IF network involves three morphologically distinct steps: non-filamentous particles, short fibrous squiggles, and long fibrils (20). Time-lapse observations have also provided us with remarkable new insights into the behavior of the vimentin particles. At any given time during the early stages of cell spreading, these particles can be seen to undergo either vibrational movements with no net translocation or rapid and unidirectional movements. The same particle can switch back and forth between these two types of movements. Time-lapse measurements of these rapidly moving particles give an average of 0.55 ± 0.24 µm/s with peak velocities of 1 µm/s. The movement of the particles is along relatively straight paths primarily oriented toward the peripheral regions of the cell. In a few cases, vimentin particles move away from the cell surface toward the nucleus (Fig. 2c–e). Furthermore, the rapid translocation is abolished after the treatment of cells with nocodazole, suggesting that particle motility depends on microtubules. Taken together, these observations suggest the involvement of a plus-end directed microtubule-based motor in the transport of nonfilamentous vimentin from the perinuclear to the peripheral region of the cell. Indeed, immunofluorescence observations using antibodies specific for conventional kinesin indicate that the majority of vimentin particles in the peripheral regions of the cell colocalize with conventional kinesin (20). Overall, our observations of GFP-vimentin in live spreading cells suggest that IF assembly in vivo is a dynamic process involving kinesin, microtubules, and an energy source such as ATP (20, 21). From the above described studies, it can be inferred that the organizational state of IF in vivo is regulated by an equilibrium between IF subunits and polymerized IF. This led us to the development of specific inhibitors of IF polymerization for use in studies of their function in vivo (26). The inhibitors are mimetic peptides corresponding to the conserved amino acid sequence of the helix 1A regions of IF proteins that are known to be potent and specific disruptors of IF assembly in vitro. On microinjection into fibroblasts, these peptides induce rapid disruption of IF networks, followed by the disassembly of both microtubule and microfilament networks, as the cells alter their shape from a flattened fibroblast configuration to a rounded morphology (Fig. 2f–i). The underlying reason for these dramatic effects is not clear. However, it is possible that IF, together with IFAPs, play pivotal roles in linking the three cytoskeletal elements into interdependent functional units. Therefore, it is likely that the disruption of IF networks alters the functional states of IFAPs, which in turn leads to the profound alterations in the organization of all three cytoskeletal systems that is seen following the injection of the 1A peptide (13, 26). These observations also lend further support to the role of IF in maintaining the integrity and the mechanical properties of the cytoplasm. The results from this laboratory as well as the work of others demonstrate that IF are dynamic structures in living cells. Their dynamic properties are regulated by complex and largely unknown mechanisms involving linkages to other cytoskeletal elements, such as microtubules and microfilaments, and with molecular motors, such as kinesin. It appears likely that protein phosphorylation plays a role in the subunit/polymer exchange process underlying many of their dynamic properties. For example, global and localized phosphorylation of IF proteins has been correlated with the dramatic changes in their organizational and assembly states in mitotic cells (29) and in the regional disassembly of IF networks in cleavage furrows (30). Furthermore, during interphase, it has been suggested that the state of IF assembly is regulated by a balance between the activities of protein kinases and phosphatases (31). Clearly, the properties of IF suggest that they play important roles in many normal physiological activities ranging from cellular mechanics to signal transduction (13). This work has been supported by a MERIT award from the National Institute of General Medical Sciences.