Unlike plants and fungi, animal cells lack cell walls and, therefore, animals require other ways to stabilize cells and tissues. Furthermore, animals require muscles for various essential activities such as breathing, the circulation of blood, peristaltic activities during the ingestion of food and digestion, and locomotion. These abilities of autonomous movement constitute a severe challenge to the integrity of tissues and generate the need for mechanisms to cope with mechanical stress. Whereas arthropods use exoskeletons for the stabilization of their body parts, most other animals have evolved various components to stabilize multicellular ensembles and tissues. One hallmark of animals is the existence of cell–cell junctions, such as desmosomes, adherens junctions, gap junctions and tight junctions. In conjunction with the intermediate filaments (IFs), a metazoan-specific cytoskeletal system, these junctions generate transcellular networks of both high rigidity and flexibility that integrate individual cells both dynamically and functionally into tissues1, 2, 3. Therefore, we must consider the specific cellular IF systems as a tool for cells to functionally integrate the corresponding cytoskeletal systems with the physiological requirements of individual tissues and, eventually, entire organs.
In humans, IF proteins are encoded by at least 65 genes, giving rise to a large protein family with limited sequence identity3, 4. This constitutes the greatest difference between the IF system and both the microtubule (MT) and microfilament (MF) systems — the two principal cytoskeletal elements of eukaryotic cells. These two systems are engaged in many basic cellular functions such that mutations in their subunit proteins, tubulin and actin, are much less tolerated than those in IF proteins. Yet, recent work has revealed a multitude of disease mutations in various IF proteins, leading to complex diseases that directly reflect the intricate expression patterns of IF genes5, 6.
In contrast to a widely held assumption that individual IFs have more or less similar, or identical, functions and properties, we will emphasize in this review that IFs exhibit, in addition to their cell-type-specific expression, a significant non-equivalence in primary sequence. We will further attempt to elaborate on what is known about the molecular mechanisms that underlie the nanomechanical properties of IFs and how these might influence tissue architecture and function. Because of the cell-type-specific properties and the high number of different IF systems, we use as a paradigm the mesenchymal protein vimentin and the muscle IF system, which is represented by desmin. However, other IF proteins such as synemin, syncoilin, nestin and, to some extent, keratins are also expressed in specific muscles in different amounts during distinct phases of life.
Cytomatrices work together
One of the major 'skeletons' in animals is the extracellular matrix (ECM), which comprises a complex three-dimensional (3D) scaffold of fibrous proteins and is made up mostly of collagens. The collagen fibrils of the ECM are linked to the interior of cells by hemidesmosomes and focal adhesions7, 8. The principal molecular components for this interaction are integrins, which can connect to IFs, MFs and membrane-associated collagens. Therefore, both the shape and functional compartmentalization of metazoan cells strongly depend on the coordinated interplay between the ECM and the cytoskeleton.
Whereas both MTs and MFs are confined by and large to the cytoplasm, most metazoan cells contain two principally different IF systems: one inside the nucleus attached to the inner nuclear membrane, and one that is cytoplasmic, which connects intercellular junctional complexes situated at the plasma membrane with the outer nuclear membrane9. The cytoplasmic IF system is a major factor in stabilizing the shape of cells, as has been demonstrated by the microinjection of peptides that destroy individual IFs10. In the nucleus, the IF system is assembled from lamins, which together with an ever increasing number of associated transmembrane and chromatin-binding proteins constitute the nuclear lamina11. Notably, simple sessile animals, such as Hydra, and arthropods do not appear to express cytoplasmic IF proteins.
The lamina is engaged in the organization of heterochromatin and provides a platform for the assembly of various nuclear protein complexes. This group of ever growing networking elements includes emerin, the lamina-associated proteins (LAPs), the lamin B receptor (LBR), the heterochromatin protein-1 (HP1) family and — through MAN1 — even signalling molecules such as the SMAD proteins, which can interact with transcription factors11. In addition, according to recent findings, the lamina connects through SUN-domain proteins to a set of outer nuclear membrane proteins from the nesprin family, which themselves bind to MTs, MFs and IFs, either directly or with the help of proteins, such as plectin or ACF7, from the spectraplakin family12, 13, 14 (Fig. 1). The interaction of the three cytoplasmic filament systems with both these multifunctional 'cytolinker' proteins and with molecular motors, as well as the regulation of their interaction by protein kinases and phosphatases, generates a dynamic multicomponent system that mediates, among other activities, the positioning of the nucleus and various cellular organelles, including mitochondria15. So, both these interconnected protein scaffolds (the nuclear lamina and the cytoskeleton), contribute significantly to the dynamics and structural integrity of cells. Using micromanipulation techniques, it has been directly demonstrated that the ECM is mechanically connected to the nuclear matrix and to nucleoli through the cytoskeleton and cell-adhesion structures16.
Figure 1 | Intermediate filament organization in metazoan cells.
In the hypothetical epithelial cell depicted, the three key filament systems of the cytoskeleton, microfilaments (MFs), microtubules (MTs) and intermediate filaments (IFs), are connected to each other by dimeric complexes of plakin-type molecules such as plectin and BPAG1. In addition, a multitude of MT-associated proteins and actin-binding proteins, including motor proteins, are thought to increase the complexity of these interactions. IFs are coupled to IF-anchoring plaques of cell–cell junctions (desmosomes) by desmoplakin, a prototype plaque molecule (plakin), and to those of cell–matrix junctions (hemidesmosomes) by plectin and BPAG1. The transmembrane proteins that mediate the contact with the neighbouring cells and with the extracellular matrix (ECM) are desmosomal cadherins and integrins, respectively. IFs are furthermore coupled to the outer nuclear membrane (ONM) by plectin and nesprin-3, whereas nesprin-2 anchors the MF system to the nucleus. On the inner side of the nuclear envelope, a layer of nuclear IF proteins (lamins) is attached to pores and inner nuclear membrane (INM) proteins as well as to chromatin. The membrane proteins of the INM might be linked to those of the ONM and thereby provide a mechanical continuum reaching from the ECM to chromatin. The number of newly identified INM and ONM proteins is increasing steadily and is represented here only in a schematic manner. ER, endoplasmic reticulum; MTOC, microtubule-organizing centre; NPC, nuclear pore complex.
Tissue specificity and development
In accordance with their role in tissue integrity and cell-shape determination in the adult organism, IFs are also thought to have an important role in coordinating mechanical forces in embryonic development, growth and maturation of specific tissues17. Whereas B-type lamins are expressed during all embryonic stages, the expression of A-type lamins is turned on only during differentiation. By contrast, the expression of cytoplasmic IF proteins is much more complex and proceeds in parallel to specific routes of embryogenesis and differentiation. In particular, muscle cells express desmin as the main IF protein and neuronal cells synthesize neurofilament triplet proteins as well as 
-internexin and nestin, whereas the precursor cells of both of these tissues express the mesenchymal IF protein vimentin. Glia cells synthesize glial fibrillary acid protein (GFAP), the expression of which is often preceded by the expression of vimentin. Last, epithelia express a multitude of different keratins. An impressive example of a complex fine-tuning of expression programmes during differentiation is that of keratins in the various segments of the eccrine sweat glands. In the secretory portion, four distinct keratins are expressed in the myoepithelial gland cells, six different keratins are synthesized in the secretory gland cells and one keratin is found in both cell layers. The cells of the luminal cell layer express a total of nine different keratins, which represents one of the highest complexities found in a single epithelial cell layer18.
The interaction between cells in cell layers in tissues or organs, such as the epidermis and the heart, is mediated in part by desmosomes. These cell–cell junctions use desmosome-specific calcium-dependent adhesion molecules, such as desmogleins and desmocollins, and thereby anchor different IFs in a cell-type-specific manner; they anchor to keratins in epithelia, desmin in cardiomyocytes and vimentin in the arachnoid mater and pia mater cells of the membranes that envelop the central nervous system (meninges) as well as in specialized endothelial cells19. IFs are distinctly separated from and organized in parallel to MF-anchoring structures of the adherens junction type (Fig. 1). The coordination of the function of both systems — for example, in the intercalated discs of the heart — is at present largely elusive but involves the plaque proteins plakoglobin and plakophilin, although it is becoming clear now that the composition of cell–cell junctional structures is much more complex than was previously expected20. Corresponding to their central function in tissue homeostasis, mutations both in desmosomal and in IF proteins have been discovered that lead to severe malfunctions in several tissues, especially in the heart21, 22. So, the fine-tuning of the interaction of these various elements might be a prerequisite for optimal tissue function, remodelling and repair23. To understand such functions, it is important to gain more insight into the mechanical properties of individual types of IFs.
Structure of IF proteins
IF proteins have been grouped into five types, or sequence homology classes (SHC), on the basis of amino-acid-sequence identity3. The acidic and basic keratins are grouped into type 1 and 2, respectively. Vimentin, desmin and GFAP are designated type 3, the neurofilament proteins are type 4, and the nuclear lamins are type 5. Using a functional criterion for classification, IF proteins can alternatively be subdivided into three independent groups according to their mode of assembly: keratins, vimentin-like proteins and lamins. As we want to concentrate on IF systems in living cells, we will not cover the complex group of 'hard' keratins that are found in hair, wool, hoof, nails and feathers.
Despite the large diversity among IF proteins, they all share a similar structural building plan, with an 
45-nm-long central -helical 'rod' domain that is flanked by non--helical N- and C-terminal end domains called 'head' and 'tail', respectively. The structural organization of coil 2 is highly conserved; yet, distinct differences exist in the building plan of coil 1 and the tail domain of cytoplasmic (Fig. 2, upper model) and nuclear (Fig. 2, lower model) IF proteins, as exemplified in Fig. 2 for vimentin and lamin A. In particular, lamins exhibit 42 extra amino acids in coil 1B and a highly conserved immunoglobulin-fold structure of 108 amino acids in the centre of the tail (Fig. 2).
Figure 2 | Structural model of cytoplasmic and nuclear intermediate filament protein dimers.
Modelling of the human vimentin and the lamin A dimers, on the basis of structural data and structure prediction, revealed that the central -helical rod domain of the individual molecules is subdivided into the coil segments 1A, 1B, 2A, 2B1 and 2B2. The vimentin coil 1A is preceded by an -helical pre-coil domain (PCD), which is probably not engaged in coiled-coil formation. Linker segments that connect the individual -helical segments are indicated: L1, L12 and L2. Left-handed coiled-coil segments are shown in green. Regions that are predicted to form nearly parallel -helical bundles as well as the so-called stutter (stu) region in the heptad repeat pattern are represented in yellow. Non--helical linkers are shown in grey. The non--helical N- (head) and C-terminal (tail) domains are coloured blue and red, respectively. Parts of the -helical coiled coils of vimentin and lamin A have been solved by X-ray crystallography9. The structure of the immunoglobulin-fold domain in the tail domain of lamin A (red wide arrows) has been solved both by X-ray crystallography and by NMR9. The numbers in brackets refer to the number of amino acids in each respective domain. Scale bar, 5 nm. NLS, nuclear localization signal.
At the first level of assembly, two individual polypeptide chains associate in parallel and in register to form a coiled coil, as was demonstrated by Crick for keratins over 50 years ago (Box 1; Fig. 2). These coiled-coil dimers are the basic building block of IF assembly. Cytoplasmic IF proteins form, at low ionic strength and physiological pH values, anti-parallel, half-staggered tetramers. In the mature filament, these tetramers are roughly aligned along the filament axis. As a consequence of the anti-parallel association of the polar dimers, IFs exhibit no polarity, as opposed to both MTs and MFs. By contrast, the solubility properties of lamins are much more complex. Although stable dimers are obtained at high pH and salt concentrations of about 250 mM NaCl, the formation of higher-order complexes begins as soon as more physiological conditions are established24, 25.
-helical coiled-coil structure of the intermediate fibre dimerThe mechanical properties of IFs are to a certain extent defined by those of the coiled coils. At the same time, the cohesive forces between adjacent dimers are also important in the nanomechanical behaviour of IFs. Although chemical crosslinking studies have indicated the existence of several specific modes of lateral dimer–dimer alignment26, it is possible that the individual dimers can to some extent slide relative to each other. So, it is both the properties of the coiled-coil dimers and lateral interactions with each other that specify the nanomechanical behaviour of individual IFs in terms of plasticity, fragility and flexural rigidity (Box 2). Ultimately, these characteristics translate into the unique properties of the complete IF network that brings about the suggested shock-absorbing function.
Filament assembly and dynamics
One of the fundamental differences between IFs and both MTs and MFs is the fact that the subunit proteins of MTs and MFs (tubulin and actin, respectively) are globular proteins with bound nucleotides, which they can hydrolyse after assembly has occurred. Nucleotide hydrolysis leads to conformational changes and, therefore, the conformational status of MTs and MFs is linked to the chemical load of a cell; that is, the concentration of available nucleotide triphosphates.
IF assembly. In vitro IF assembly is not directly dependent on cofactors, but in vivo their remodelling and structural performance as 'stress absorbers' is functionally dependent on the combined action of kinases, phosphatases and chaperones7, 27. Moreover, IFs are resistant to challenges such as cold or high concentrations of salt, as they do not dissociate even in buffers of high ionic strength (1.5 M KCl). Only in buffers of low ionic strength do they disintegrate into soluble complexes. The biochemical properties of cytoplasmic and nuclear IFs differ significantly, and this is probably the basis for their principally different ways of generating filamentous structures28, 29, 30. Further evidence from in vitro studies indicates how the assembly pathways of nuclear and cytoplasmic IF proteins differ25. Among cytoplasmic IF proteins, two types of assembly can occur: whereas keratins represent obligatory heteropolymeric dimers of one basic and one acidic partner, desmin and vimentin IFs can form homopolymers, although in many situations they form mixed dimers with proteins from the same assembly group, side by side with homodimers. IF properties might thereby be modulated extensively even in or along one filament. For example, complex co-assembly patterns allow the incorporation of four different neurofilament proteins, NF-L, NF-M, NF-H and -internexin, into neurofilaments. In peripheral nerves, the SHC3 IF protein peripherin incorporates into the neurofilaments in varying ratios, thereby complementing the neurofilament triplet proteins.
A similar level of complexity is introduced into muscle IFs through IF proteins, such as synemin and syncoilin, that, like NF-M and NF-H, have a long non--helical C-terminal tail domain. Synemin and syncoilin do not form IFs on their own, but integrate into IFs through the dimerization of their -helical rod domain with that of vimentin, desmin, -internexin or NF-L. Synemins are expressed in all types of muscle cells and provide IFs with the ability to interact with costameres31 through dystrophin and utrophin32. Moreover, their ability to associate with the actin-binding protein -actinin and with vinculin enables IFs to directly connect to focal adhesions and thereby to the MF system33. In summary, mixed IFs can generate an enormous complexity, even varying along one filament, which in turn makes IFs one of the most variable biochemical 'platforms'.
In addition, IFs from members of the three assembly groups — keratins, vimentin-like IF proteins and lamins — do not form copolymers but can reside as distinct filament systems in one type of cell. There, they fulfil distinct functions side by side and can at the same time enforce each other. So, if one envisions a cell as a complex material, the contribution of one element alone is surely insufficient to explain integrative parameters such as viscoelastic properties and resistance to mechanical stress34. Last but not least, the surface of individual types of IF can vary considerably owing to the low amino-acid sequence identity between individual IF proteins in those parts of the coiled coil that are exposed to the surface as well as in the entire head and tail domains. So, unlike MTs and MFs, every single type of IF differs significantly from others with respect to its chemical surface properties.
Cytoplasmic unit-length filament formation. Unlike actin and tubulin, cytoplasmic IF proteins do not form seeds to which individual subunits such as monomers and dimers add, but they laterally associate into full-width 60-nm-long IFs, also known as unit-length filaments (ULFs), in a process that is complete in seconds. Moreover, this lateral interaction is so strong that it even takes place at high pH following the addition of salt28, 35. Subsequently, a much slower elongation phase, driven by longitudinal annealing of individual ULFs, takes over and probably involves molecular rearrangements in individual ULFs. So, ULFs both serve as nuclei for IF formation and constitute the building blocks for filament growth. In addition, growing IFs can still fuse end to end. In a third, cooperative phase, filament diameters are reduced, which indicates a further intrafilamentous subunit reorganization36. This 'radial compaction' step occurs to a similar extent under various conditions of assembly, indicating that it represents an essential general step in the conversion of assembly intermediates to mature IFs. Recently, a mathematical model that describes the kinetics of this assembly process has been published37.
Dynamics of nuclear lamins. In contrast to cytoplasmic IF proteins, the in vitro assembly of lamins from dimers involves the simultaneous lateral and longitudinal association of dimers25. So, already 5 seconds after the initiation of assembly, interconnected fibrillar strands of varying length and thickness are observed. The variation of the diameter (2–16 nm) is seen along individual fibres, the thicker parts exhibiting a knob-like surface that probably represents the globular immunoglobulin-fold of the tail domain (Fig. 2). It is easy to predict that measurements based on high concentrations of protein for assembly and using bulk assemblies, such as in rheology, will yield very complex results owing to the heterogeneity of the structures generated38. In contrast to the rapid in vitro assembly scenario, lamin structures formed in vivo appear to be much more regular (when they can be observed, such as in the lamina of the Xenopus laevis oocyte24).
In vivo, the dynamics of IFs have been followed extensively by the microinjection of fluorescently labelled IF proteins or by the transfection of chimeras of IF proteins and green fluorescent proteins30, 39. Whereas cytoplasmic IFs appear to be very dynamic, nuclear lamins have been demonstrated to stay more or less in place as soon as they have been integrated into the lamina, indicating that they are part of a stable molecular supernetwork or matrix40, 41.
Single IF mechanics and beyond
At the single-filament level, not much information is available on the mechanical properties of the three components of the cytoskeleton. Using atomic force microscopy (AFM), it is now possible to perform time-lapse imaging and to mechanically stress single filaments, including IFs, in various ways (Box 2). These techniques might be further developed for use in assays to analyse the effects of mutations, as well as mutations of associated proteins, on the filament properties.
Soft, extensible and nearly unbreakable. In vitro assembly of both recombinant and authentic IF proteins yields smooth-looking, flexible filaments by electron microscopy (EM) and AFM42. From such images, a persistence length of 1 
m has been estimated for vimentin IFs43, which in turn gives rise to a dynamic shear modulus of a few Pa for a dilute suspension (0.1–1 mg per ml) of entangled IFs. This value is significantly smaller than that of MFs assembled at the same protein concentration44. The keratin-rich cornified epidermal layer of skin contains an IF network that is 100- to 1000-fold more concentrated and therefore has an elastic modulus in the MPa range45. Further alignment of keratin IFs and crosslinking through disulphide bonds gives rise to mammalian appendages such as hoof, nail, quill and hair. These materials have an elastic modulus in the GPa range46, which can be decreased by at least a factor of 10 by using reducing agents such as dithiothreitol47.
IFs are not simply flexible filaments, they also have an unusual extensibility compared with MFs and MTs. In a recent AFM study, it was demonstrated that single neurofilaments, desmin and keratin IFs can be stretched up to 3.5-fold48 (250% tensile strain; Box 2). This is in agreement with rheological measurements performed with entangled IFs that can bear 300–400% shear strain before the network breaks49. Similarly, hagfish slime threads, which are extruded, macroscopically visible bundles of aligned keratin-like IFs, can bear 220% tensile strain before breaking. By contrast, wool and hair can only be stretched up to 50–60% strain in water due to their extensive crosslinking by disulphide bonds46.
IFs combine an unusual extensibility with a strong resistance to breakage50. Preliminary AFM data indicate that a single desmin filament can bear 1–2 nN before breaking at 250% tensile strain (L.K., unpublished observations). For comparison, MFs break above 0.6 nN at a low level of tensile strain51. Following the single-filament behaviour, keratin-rich fibres break at large stresses between 150 and 180 MPa. This is achieved by a spectacular hardening above a strain threshold that is different for each fibre type. Although the so-called strain-hardening is a common feature of all IF assemblies, it is not observed with MFs and MTs. Even dilute suspensions of filaments show a nonlinear increase of their dynamic shear modulus for large shear strains (of 50% or more)44.
Mechanical properties of the cytoskeletal network. In the cytoskeleton, IFs seem to work synergistically with the MF and MT networks. On the basis of in vivo measurements of MT buckling, it has recently been proposed that MTs might be more resistant to compressive forces than expected from in vitro measurements of MTs52. The mechanism proposed in this study is that IFs are reinforcing MTs, which in turn reduces the ability of IFs to bend. Along the same lines, it has been demonstrated in a concomitant study that a mixed suspension of entangled vimentin IFs and actin filaments has a significantly greater dynamic shear modulus compared with each individual suspension at the same total protein concentration53. As a possible mechanism, these authors suggest that the tail domain of vimentin can directly bind to MFs, thereby yielding a crosslinked network instead of an entangled suspension. In fact, direct binding might not even be necessary to explain the cooperative behaviour of the two filament systems. Instead, we propose that most of the tail domain of vimentin protrudes from the filament surface, as previously shown for the tail domain of the neurofilament triplet protein NF-H, thereby yielding an hydrodynamic radius that is higher than the physical radius of 5 nm54. Therefore, the rigid MFs, when embedded in a vimentin IF matrix, would be more constrained in motion than if they were surrounded by other MFs. Just as in the case of MTs, this lateral reinforcement would give rise to a stiffer gel.
These independent studies highlight the fact that IFs might mechanically integrate into the MF and MT cytoskeleton to yield a scaffold with unique properties. It is interesting to note that the contribution of IFs to the mechanical properties of cells and tissues has been completely neglected by a large part of the research community. This is clearly not due to the lack of suitable experimental approaches, as several are available (reviewed in Ref. 55). Instead, most researchers exploring the mechanical properties of cells and tissues try to correlate them only with changes in the architecture of the MF and MT networks, despite the presence of significant IF systems in these specimens55. So, it is obvious that a change in paradigm is needed.
Mechanotransduction
The function of a stress-bearing structural continuum, such as the IF system, in cellular homeostasis is not yet understood at a mechanistic level, but it might constitute an important platform to mediate cellular mechanotransduction processes17. Early on, studies of the interaction of ECM receptors with cytoskeletal elements pointed to a direct mechanical coupling of cell-surface structures with the nucleoskeleton56. Moreover, it was demonstrated that stretching of cells, such as cardiac myocytes, causes the induction of immediate–early genes followed by a strong growth response57. The importance of transcellular IF networks for tissue integrity became evident after the discovery of disease-causing keratin mutations, which lead to severe cell fragility in the skin of affected patients upon mechanical trauma. Furthermore, recent evidence from various rare diseases indicates that besides its structural functions, the IF cytoskeleton is also involved in cell signalling. Indeed, these cell-type-specific multicomponent protein assemblies are all substrates for multiple phosphorylation reactions58, 59. For this reason, one might assume that the number of effective interactions is high and probably beyond our ability to be appropriately described60.
How does mechanical stress affect tissue physiology? Gene targeting is a powerful tool that can be used to analyse the physiological role of IF proteins. The vimentin gene is one of the first IF genes that was knocked out in mice61. Although embryonic and post-natal development was apparently not significantly affected, drastic effects were observed in experimental situations that challenged physiological properties of the transgenic animals. For example, the ablation of three quarters of the renal mass was lethal in mice that lacked vimentin because of end-stage renal failure within 72 hours, whereas control mice survived by adjusting the flow properties of their blood vessels62. The balance in the endothelial production of nitric oxide and endothelin was disturbed in knockout mice because they synthesized more endothelin than nitric oxide, and death was a consequence of the lack of vascular adaptation to nephron reduction. However, the perfusion of nephrectomized mice with an endothelin-receptor antagonist enabled the vimentin-null mice to survive. Various experimental approaches demonstrated that vimentin modulates the structural responses of arteries to changes in blood flow and pressure, and so plays a crucial role in the mechanotransduction of shear stress63, 64 (that is, in pathological conditions that require vascular adaptations). It was furthermore documented that in a regeneration situation after induced bilateral renal ischaemia, vimentin is essential to mediate Na–glucose cotransporter I localization in brush border membranes, thereby preventing glucosuria in post-ischaemic mice65.
In a different physiological context, loss of vimentin appears to cause impaired motor coordination, as revealed by behavioural tests of the same knockout mice. Morphological analysis of brains from vimentin-null mice revealed poorly developed and highly abnormal Bergmann glia as well as developmental defects in Purkinje cells66. More recent experiments showed that vimentin is involved in cellular processes such as retrograde signalling following injury in nerves and the migration of leukocytes through the endothelium, also termed diapedesis67, 68. In injured peripheral nerves, the local synthesis in axons of carrier proteins, such as vimentin, provides molecules that incorporate potential signalling molecules, such as transcription factors and mitogen-activated protein (MAP) kinases, into the dynein retrograde motor complex. Most importantly, the regeneration of injured dorsal root ganglion neurons is delayed in vimentin-null mice67. In diapedesis, the presence of vimentin was shown to be important for peripheral blood mononuclear cells (PBMCs), as these cells have a markedly reduced capacity to home to mesenteric lymph nodes and spleen in vimentin-knockout mice. Moreover, surface receptors that are crucial for the homing of lymphocytes, such as intracellular adhesion molecule-1 (ICAM1) and vascular cell adhesion molecule-1 (VCAM1) on endothelial cells as well as integrin-
1 on PBMCs, were aberrantly expressed and distributed in the absence of vimentin68. Consequently, it is evident that IFs are active in lymphocyte adhesion and transmigration.
These few examples amply show that although vimentin is not essential to generate a mouse, its expression is probably essential for mice to survive in a natural habitat where the performance and health of animals is challenged by infectious microbes, parasites and predators.
IFs and disease
As mentioned above, IFs took centre stage when it was discovered that point mutations in keratin genes give rise to severe human blistering diseases (reviewed in Ref. 69). The most obvious explanation for the disease mechanism involved a mechanical stress model, whereby exposure of the skin to mechanical stress would lead to the rupture of a large part of the epidermis in the absence of a proper keratin network. Following this discovery, mutations in desmin were demonstrated to cause muscular dystrophy (reviewed in Ref. 70; see also below). Shortly afterwards, mutations in lamin A were also found to cause muscular dystrophies (reviewed in Ref. 71). This latter finding led to the identification of more than 230 mutations in lamin A that cause a complex set of at least 13 different human diseases6, 72. Among them, severe diseases that lead to premature ageing, such as the Hutchinson–Gilford progeria syndrome and atypical Werner syndrome, are observed73. Most interestingly, mutations in desmin and lamin A can both cause muscular dystrophies and cardiomyopathy. Although the disease mechanism is not at all clear in either case, several models including stress, cell fate and gene-expression models have been proposed74.
One of the most dramatic disease phenotypes evoked by mutations in an IF protein are those of GFAP. These mutations cause Alexander disease, a fatal disorder of the central nervous system that is characterized by devastating disturbances in the normal development of the brain and skull75. As part of the pathomechanism of Alexander disease, it has been assumed that the corresponding GFAP mutations might compromise the astrocyte stress response76. In addition, disturbances in signalling pathways, as caused by mutations in IF proteins, could be responsible for some aspects of IF-related diseases in general77.
Manifestation of IF-related diseases usually occurs at distinct times5. Whereas keratin mutations can manifest themselves during birth, heart diseases caused by mutations in desmin or lamin A have a comparatively late age of onset in the second or third decade of life. By contrast, symptoms of epidermolytic diseases, caused by mutations in keratins, can actually improve with age. This suggests that the complement of proteins that are involved in the generation of these diseases changes during development and with age. Moreover, the balance between the functional and the diseased state is dependent on subtle changes in the IF cytoskeleton.
IFs — a dispensable part of muscle architecture?
In both vertebrate skeletal and cardiac muscle, the IF protein desmin is abundantly found in structures that surround the sarcomeres at the position of the Z-discs and connect sarcomeres to costameres (in the case of skeletal muscle) or desmosomes (in the case of cardiac muscle). In addition, desmin IFs structurally integrate nuclei and mitochondria into the myocyte cytoskeleton78. Although the presence of desmin is not essential for proper muscle formation during embryonic development, as demonstrated by gene targeting in the mouse, its absence has severe consequences when the mouse is challenged to exercise, as will be outlined below.
Towards an understanding of desminopathies. Myofibrillar myopathy (MFM) is histologically characterized by the disintegration of Z-discs and myofibrils as well as by ectopic subsarcolemmal and intrasarcoplasmic accumulation and aggregation of desmin, B-crystallin, plectin, ubiquitin, titin and other proteins (Fig. 3). Usually becoming symptomatic in the second or third decade of life, this devastating disease can affect striated as well as smooth muscle, leading to slow progressive myopathy. However, affection of cardiac muscle resulting in dilated cardiomyopathy (DCM), restrictive cardiomyopathy (RCM) or hypertrophic cardiomyopathy (HCM) and the characteristic early occurrence of arrhythmia is the major cause of death in these patients. So far, the pathomechanism that underlies the development of MFM is only partly understood. Most investigations have focused on desminopathy and B-crystallinopathy, which are caused by mutations in desmin and B-crystallin, respectively. For desminopathy, recent investigations have concentrated on the hypothesis that mutations in the desmin gene lead to defective IF assembly and that this results in aggregation of the misfolded protein70. Accordingly, mutations in the rod domain of desmin were shown to give rise to distinct assembly defects: either they arrested the normal in vitro assembly process at specific stages or they led to disassembly of irregular precursor structures79 (Fig. 4). In contrast to prior notions, many of the mutations allowed filament formation to take place, although these filaments had distinct alterations of filament architecture, including a change in the number of subunits per cross-section as compared to wild-type desmin IFs80. So, the nanomechanical properties of these filaments seem to be severely compromised.
Figure 3 | Destruction of muscle architecture in desminopathy.
a | Immunofluorescence microscopy of desmin in an isolated myofibre of a patient who suffers from desminopathy. Desmin (red) aggregates are deposited in large aggregates, whereas the typical staining pattern of desmin intermediate filaments (IFs) intersecting individual myocytes at the level of their Z-discs is preserved82. Blue, DNA stain (DAPI). Scale bar, 50 m. b | Ultrastructural analysis of skeletal muscle from a patient affected by the DesR350P mutation. Note the massive accumulation of granulofilamentous material on the left side of the image. Scale bar, 500 nm. c | Immunogold electron microscopy with the monoclonal anti-desmin antibody (mab-D33) and a secondary antibody that is coupled to 10-nm gold particles shows a dense labelling of pathological protein aggregates in the subsarcolemmal region of skeletal muscle from the same biopsy shown in panel b. Scale bar, 200 nm. Panels b and c are reproduced with permission from Ref. 106 © (2005) Oxford University Press.
Figure 4 | Overview of the assembly and decay pathway of various desmin mutants.
a | When soluble tetrameric complexes of wild-type (WT) desmin are induced to assemble, initially eight tetramers associate laterally to form a unit-length filament (ULF). Next, these ULFs anneal longitudinally and give rise to short, 'open' (that is, less compact) filaments. Last, elongated filaments radially compact to yield mature intermediate filaments (IFs). An example of mature desmin IFs is depicted in the electron microscopy image of negatively stained preparations of in vitro assembled mouse recombinant desmin. b | Mutant desmins assemble into various types of structures that can be classified into four types, exemplified here by: DesA360P, forming IFs of seemingly normal appearance (panel 1); DesR406W, being arrested during elongation and thereby exhibiting short, still segmented filaments (panel 2); DesN342D, in addition to extensive elongation, individual filaments have opened along their length and generated meshworks of protofilamentous masses (panel 3); DesL370P, initially successfully assembles into ULF-like structures and short regular filaments, but within 20 seconds reorganizes into relatively regular, round aggregates that are 30 nm in diameter (panel 4). All micrographs represent negatively stained specimens prepared and recorded under identical conditions. For more details, see Ref. 79. Scale bars, 100 nm.
The situation in myocytes is even more complex owing to heterozygosity, as affected patients harbour both wild-type and mutant alleles. Keeping this in mind, it was shown that filament formation by assembly-deficient mutant desmins can be rescued in some cases by the presence of wild-type desmin81. In many cases, however, the mutant protein drives the wild-type protein into non-IF structures82. These in vitro analyses were corroborated by transfection studies, which revealed that assembly-incompetent desmin mutants formed cytoplasmic aggregates, whereas filament-forming mutants assembled into filamentous networks79, 83.
A potential disease-causing mechanism that is induced by the filament-forming mutants might be that protein misfolding and/or alterations of surface-charge patterns interfere with proper binding to IF-associated proteins. Alternatively, an alteration in the intrinsic viscoelastic properties of single desmin filaments might cause a failure in the mechanical coordination of the positioning of individual myofibres. Here, detailed binding studies with IF-associated proteins and analyses of biophysical properties at the single-filament level should help to gain more insight. For example, misfolded desmin or B-crystallin mutants might override protein-quality-control mechanisms, as provided by the ubiquitin–proteasome pathway, and bring about aggregate formation84, 85. The formation of aggregates might actually protect the myocyte, as potentially toxic soluble protein complexes are thereby removed86. Indeed, it has been demonstrated in some desmin-related myopathies that the elevated concentration in cells of soluble misfolded proteins causes mitochondrial dysfunction and activates the mitochondrial apoptosis cascade87, 88.
In summary, desmin mutations can affect myocyte and muscle homeostasis in different ways that are not mutually exclusive. Accordingly, different hypotheses have been put forward with respect to the pathomechanism in attempts to explain how, and at what level of organismic organization, the corresponding mutation might take effect (Fig. 5). However, a more rational understanding of the pathogenesis of desminopathy will require more insight into the fundamental principles of muscle function. This, in turn, will help us to understand the pathogenesis of this orphan disease and also that of other, more common, degenerative muscle diseases89.
Figure 5 | Hypothetical scheme for the disease mechanism caused by desmin mutations.
Individual mutations can affect different biophysical properties of the filaments (green boxes), which then can interfere with distinct cellular activities (blue boxes). As a consequence, different physiological responses can take place at the cellular level (yellow boxes). Ultimately, these cellular events cause various tissue-wide pathogenic alterations from apoptosis to heart failure (orange boxes). IF, intermediate filaments; WT, wild type.
Desmin at work
Although 'plastic dish' cell biology has elucidated many interesting features of desmin IFs, a more profound understanding has been obtained in studies that have involved isolated muscle fibres. More specifically, it has been demonstrated that the degree of structural damage of muscle in an experimental situation — after forced stretching, for example — correlates with the disappearance of desmin immunoreactivity from muscle during the first minutes of eccentric contraction (which is defined as lengthening of an activated muscle). This is probably due to masking of the antibody epitope after the structural reorganization of desmin IFs or to proteolytic digestion of desmin90. After prolonged exercise (30 minutes), depending on the muscle type, 8–24% of the muscle fibres were desmin negative. The magnitude of apparent desmin loss correlated well with the loss of contractile force. Although the sarcomeric organization was not significantly affected, the distribution of titin was drastically altered in cells that had lost immune reactivity to desmin. This indicated that the extrasarcomeric cytoskeleton, which primarily consists of desmin, -actinin and plectin, stabilizes the intrasarcomeric cytoskeleton that harbours titin and nebulin as its major components. So, the two systems function together to laterally integrate mechanical work in the individual muscle fibre and through costameres in the whole tissue. Moreover, these experiments have shown that desmin has a major role in mediating proper force transduction and propagation in muscle.
Moreover, the mechanical interactions between desmin IFs and costameres, Z-disks and nuclei have been followed directly during passive deformation in single muscle cells. In particular, the connectivity between these structures was quantified by integrative experimental and computational analysis, from myofibres of both wild-type and desmin-null mice91. Similar to vimentin-null mice, desmin-null mice develop normally until birth. Soon after birth, their hearts exhibit extensive structural defects, including myocyte cell death and calcific fibrosis, which suggest a major malfunctioning of the working muscle92, 93, 94. The earliest ultrastructural defects observed affected mitochondria, and these defects could be partially restored by overexpressing the anti-apoptotic regulator BCL2 in the desmin-null mice95, 96. As a consequence of the desmin knockout, voluntary and forced running performances were adversely affected in null mice compared with wild-type mice, and so normal levels of desmin are a necessary component of exercise performance97. This is another example in which structural and physiological functions cannot be separated. In summary, these experiments indicate that loss of desmin makes mice 'lazy', which is surely of importance in an evolutionary context for animals whose survival, as individuals and as a species, crucially depends on their ability to escape.
Conclusions
Evidently, IFs are among the most versatile structures of metazoan cell architecture. They exist in two separate moieties that interact via the nuclear envelope by a complex system of inner and outer nuclear membrane proteins. These moieties are the nuclear lamins (which form a planar network that interlaces the inner nuclear membrane proteome and the interphase chromosome surface) and cell-type-specific cytoplasmic IFs (which form a flexible system of long individual filament arrays that integrate multiple cellular components, including MTs and MFs, into a dynamic, stress-buffering cytoskeleton).
Through a multitude of associated proteins, IFs connect the cytoskeletons of a cell to cell–cell and cell–matrix junctions, thereby establishing transcellular networks. At these mechanically coupled interfaces, IFs interact with multiple supramolecular complexes that are part of regulatory and signalling chains, including receptor tyrosine kinases, such as integrins, and structural components of adhesion-plaque proteins, such as plakophilins. IFs are therefore a crucial part of the 'signalosome' of cells and tissues that translate changes of environmental conditions into alterations of gene expression at the cellular level. IFs also provide an extensive and biochemically versatile interface surface that can be tailored by individual cells to serve as a dynamic platform for the binding of protein complexes, organelles and 'receptors' that tether internal membranes to the cytoskeleton.
The complex clinical phenotypes that are exhibited in humans as a consequence of mutations in IF proteins amply show how intimately IFs are linked to developmental processes of humans and animals. The most dramatic examples are lamin A mutations that lead to premature ageing, desmin mutations that destroy an entire organ (the heart) and GFAP mutations that cause Alexander disease. Therefore, as mutations in IF proteins, such as lamin A, affect the execution of genetic developmental programmes as well as ageing, the 'engineering' of IF proteins by evolution was and is of utmost importance for the successful development of vertebrates and probably animals in general.
