Intermediate Filament Networks: Organization And Possible Functions Of A Diverse Group Of Cytoskeletal Elements

Intermediate filaments (IF) are cytoskeletal components found in many types of cells. They are usually ≈10nm in diameter, and contain alpha-type proteins in a coiled-coil configuration. Despite their morphological, physical and chemical similarities, they in fact represent a very large and diverse family of cytoskeletal elements (Zackroff et al. 1981; Steinert et al. 1984; Wang et al. 1985). The diversity of IF is especially evident when their polypeptide composition is compared between different cells from different tissues of the same organism. This has led numerous investigators to attempt to categorize cells on the basis of differences in the immunological, molecular and biochemical properties of the structural subunits of their IF. To date, the largest subgroup of IF proteins to be described are the keratins. In mammalian cell systems, there may be 30 or even more keratins (Moll et al. 1982; Steinert et al. 1984; Sun et al. 1986). This large group of IF proteins is distinguished by their relative insolubility, their diversity and their abundance in epithelial cells. The type or types of keratin family members expressed in any individual epithelial cell type varies with the state of cellular differentiation and the location of the cell (Zackroff et al. 1981; Steinert et al. 1984; Wangí al. 1986; Sun et al. 1985).

In contrast to these differences within the keratin sub-group, there are also obvious similarities amongst the keratin subunits. For example, there are antibodies that react with a variety of keratins, indicating that they possess antigenic similarities. On comparative biochemical grounds the keratins are known to be serine- and glycine-rich and more recent sequencing data have suggested that these serine + glycine-rich residues are located in terminal domains (Steinert et al. 1984; Wanget al. 1986).

There are also other types of IF proteins, the most extensively studied being vimentin, which is the major structural protein found in mesenchymal cells such as fibroblasts (Zackroff et al. 1981; Steinert et al. 1984; Wanget al. 1986). In addition, the major building blocks of neuronal IF (the neurofilament triplet proteins), the glial IF proteins, and the major protein of muscle IF, which has been called either desmin or skeletin, have also been studied intensively in recent years (Zackroff et al. 1981; Steinert et al. 1984; Wanget al. 1985).

Besides the major structural subunits comprising IF in different cell types, there is also some evidence for the existence of IF-associated proteins (IFAPs). These are diverse and appear to be cell-and-tissue type specific. For example, filaggrin is a specific IF-IF cross-bridging protein found in skin and related tissues (Steinert et al. 1981). Filaggrin is a very basic protein, which may act to cross-link individual keratin-containing IF into bundles (tonofilaments) via their serine + glycine-rich terminal domains. Another very different IFAP has been described in BHK cells, which appears to form cross-bridges between neighbouring BHK IF. In situ, these bridges appear to be intermittent and may be involved in the formation of cytoplasmic IF networks (Yang et al. 1985; Lieska et al. 1985). In epidermal cells, there is some evidence that a protein called desmoplakin, a major component of desmosomes, may also be an IFAP (Jones & Goldman, 1985).

Recently, we have demonstrated that there are keratin-like proteins in mesenchymal cells, such as cultured BHK and 3T3 cells, which do not form IF under normal conditions. These proteins appear to be localized at the nuclear surface where they form an insoluble polymer system known as the nuclear lamina (Aaronson & Blobel, 1975; Gerace et al. 1978, 1984; Zackroff et al. 1984; Goldman et al. 1984, 1985, 1986). Furthermore, Linck et al. (1985) have suggested that the tektins, components of the outer doublet microtubules of cilia, are also intermediate filament-like proteins. These findings are intriguing, as they demonstrate that not all members of the IF protein family form 10 nm diameter filaments. In the case of the nuclear lamina proteins, they appear to possess several keratin-1F-like properties including their high content of glycine and serine, their alpha-helical content, their cross-reactivity with antibodies directed against IF-forming epidermal keratins and their primary sequence derived from cDNA clones (Zackroff et al. 1984; Goldman etal. 1984, 1985, 1986; McKeon et al. 1986).

The fact that there is remarkable diversity amongst IF structural and associated proteins, and the fact that there are additional IF protein family members that can form polymeric systems other than IF, must presumably reflect functional significance. In this report, we review some of our work that has been aimed at determining the functions of the IF system.

Immunofluorescence observations of various types of cultured cells indicate that IF are generally distributed between the nuclear and cell surfaces in cultured fibroblastic cells, such as mouse 3T3 and baby hamster kidney (BHK). When BHK cells are observed by indirect immunofluorescence; using an antibody preparation directed against the major BHK IF structural protein of 55 000 molecular weight (55 K, termed vimentin), a complex network of fibrils is seen to radiate from a juxtanuclear mass towards the cell surface (Fig. 1) and individual IF frequently appear to approach the cell surface very closely (Green & Goldman, 1986). A more dramatic way in which to show the association between IF and the nuclear and cell surfaces is to observe cells during attachment, spreading and shape formation. Following trypsinization and replating, rounded up BHK cells contain a large juxtanuclear cap enriched in IF structural protein (Fig. 2; and Goldman & Follett, 1970; Goldman & Knipe, 1973). As cell spreading progresses, this cap gives rise to fibres that radiate away from the nucleus towards the cell surface (Fig. 3; and Goldman & Follett, 1970; Goldman & Knipe, 1973). Similar observations have been made on all fibroblasts studied to date, on cultured nerve cells such as neuroblastoma (Fig. 4; and Goldman et al. 1981), and on several types of epithelial cells (Goldman et al. 1973; Jones et al. 1982).

The most extensive and dramatic IF networks that can be resolved in the fluorescence microscope are seen in many types of epithelial cells with polyclonal antibodies directed against epidermal keratin. For example, in cultured mouse keratinocytes, these networks are comprised of tonofibrils (IF bundles). These form a complex cage surrounding the nucleus, from which they radiate towards the cell surface (Fig. 5; and Jones et al. 1982). During the early stages of attachment and spreading of such cells, the network of tonofibrils can be seen to surround the nucleus. As cells begin to spread upon a substrate, individual tonofibrils are seen to extend into the expanding regions of cytoplasm from the concentrated juxtanuclear mass and move towards the cell surface (Fig. 6).

We have observed another example of nuclear-cell surface reorganization of tonofibrils in cultures of primary newborn mouse keratinocytes (PME) (Jones et al. 1982; Jones & Goldman, 1985). These cells are capable of attaching, spreading and dividing in low Ca²⁺ conditions (Hennings et al. 1980; Jones et al. 1982; Hennings & Holbrook, 1983; Jones & Goldman, 1985). However, even though the cells are extensively spread under these culture conditions, tonofibrils are seen primarily in large perinuclear accumulations (Fig. 7; and Jones et al. 1982; Jones & Goldman, 1985). When the Ca²⁺ concentration is increased to normal levels, the tonofibrils move rapidly towards regions of the cell surface in contact with other cells (Fig. 8; and Jones et al. 1982; Jones & Goldman, 1985). When IF associate with the cell surface, desmosome assembly takes place. Desmosomes can be distinguished in this culture system by using immunofluorescence methods with antibodies directed against desmosomal components, such as desmoplakin (Figs 9, 10; and Mueller & Franke, 1982; Jones et al. 1985).

The IF patterns seen in the different cell types described above have some general features in common, including the fact that they appear to have radiating networks between the nuclear and cell surfaces. These observations with the light microscope are readily confirmed at higher resolution with the electron microscope. In the case of fibroblasts such as BHK or 3T3 cells, arrays of 10 nm diameter IF can be seen in the region immediately adjacent to the nuclear surface. These frequently appear to terminate at the nuclear surface (Fig. 11; Goldman et al. 1984, 1985, 1986). In mouse epidermal cells an elaborate cage composed of IF bundles can be seen to surround the nucleus, which readily accounts for the images resolved by light microscopy described above (Fig. 12; and Jones et al. 1985). Frequently, IF can be seen apparently terminating in close proximity to the nuclear pore complexes (Jones et al. 1982, 1985; Goldman et al. 1985).

At the level of the cell surface there are numerous types of IF-plasma membrane interactions as revealed by electron-microscopic observations. IF can be found in close association with the cell surface in cultured BHK and chicken embryo (CEF) fibroblasts (Fig. 13; and Green & Goldman, 1986). In the case of CEF, plasma membrane-associated IF are frequently seen in regions in which the outer cell surface contains an accumulation of extracellular matrix material, which by immunofluorescence and immunogold-labelling criteria contains fibronectin (Green & Goldman, 1986). IF are also seen in close proximity to the microfilament bundles that accumulate subjacent to the plasma membrane in the specialized regions known as the ‘fibronexus’ (Fig. 14; and Green & Goldman, 1986). These latter regions are specialized membrane domains that are thought to represent sites of interactions between submembranous microfilament bundles and extracellular matrix components, especially fibronectin (Singer, 1979).

In addition to these ultrastructural observations, double-label immunofluorescence studies have been carried out in CEF, which indicate that, although there is not an extensive degree of overlap between the patterns generated by antibodies directed against CEF IF structural protein and fibronectin, there are regions in every cell in which a close association is seen between the two patterns (Figs 15, 16; and Green & Goldman, 1986). Similar observations have been made using fibronectin antibody in conjunction with the immunogold-labelling procedure (Fig. 17; and Green & Goldman, 1986). Additional evidence to support the possibility that IF associate with a ‘cell membrane complex’, which acts in some unknown fashion to link IF with the outer cell surface, stems from the observation that trypsinization of cells, which removes surface-bound fibronectin, induces the retraction of IF away from the plasma membrane. However, when CEF are removed from their growth substrates in the absence of trypsin, IF-cell surface associations are retained (Green & Goldman, 1986).seen coursing parallel to stress fibres in fully spread CEF by immunofluorescence (Green & Goldman, 1986). Electron-microscopic examination of the nodules supports the light-microscopic findings; they appear as an outpocketing of a microfilament bundle that contains electron-dense fine fibrillar material closely associated

Epithelial cells also exhibit obvious IF-cell surface interactions at the level of the desmosome – a double-membrane-containing structure, each half of which is associated with a cell–cell adhesion site (Arnn & Staehelin, 1981). Desmosomes are especially abundant in epidermal cells, where they are associated with keratin- containing IF. Indeed, IF appear to be attached to the innermost region of the submembranous plaque, which is associated with the cytoplasmic face of desmosomes (Fig. 18; and Arnn & Staehelin, 1981; Jones etal. 1987). As the bundles of IF approach the desmosomal plaque they appear to loop in and out of the fine fibrillar material associated with the plaque (Kelly, 1966). Recently, we have provided evidence that suggests that at least one of the desmosomal plaque proteins, desmoplakin 1, is associated with IF prior to the formation of desmosomes (Jones & Goldman, 1985).

The results of these morphological studies support the hypothesis that the IF system forms a cytoskeletal network that links the nuclear surface with the cell surface in cultured fibroblasts and epithelial cells (Goldman et al. 1985). We have also been able to show that IF interact with the two other major cytoskeletal elements, microtubules and microfilaments.

One of the most obvious cases of IF interaction with microtubules (MT) can be seen in the cytoplasmic region between the nucleus and the cell surface of cultured fibroblasts that have been permitted to flatten on a solid substrate, such as a plastic Petri dish or a glass coverslip. Numerous IF can be seen in this region by electron- microscopic examination of thin-sectioned cells. In these areas, parallel arrays of IF and MT can be found (Fig. 19). Possible cross-bridging elements between these two cytoskeletal components have been reported elsewhere (Goldman & Knipe, 1973). Further support for this interaction also stems from the finding that in the presence of microtubule disrupting agents such as vinblastine or colchicine, the majority but not all of the IF visualized by immunofluorescence collapse around the nuclear surface (Fig. 20; and Starger etal. 1978; Green & Goldman, 1983, 1986). These latter arrays of IF are similar to those seen in cells observed during the early stages of attachment and spreading (see Figs 2, 3, above). Similar MT–IF configurations have been reported in nerve cells (Goldman et al. 1981).

IF also appear to be closely associated with submembranous microfilaments. In fibroblasts, we have been able to demonstrate, utilizing both immunofluorescence and electron microscopy, two types of MF–IF associations. One of these is apparent in the early stages of spreading in cultured CEF. In about 50% of these cells, one or more phase-dense nodules are apparent within the cytoplasm in close association with stress fibres (microfilament bundles). These structures stain intensely with rhodamine-labelled phalloidin, indicating the presence of F-actin. When double stained with an antibody preparation directed against the CEF IF structural protein, these actin-rich nodules appear as focal centres from which IF-containing fibres radiate (Figs 21, 22; and Green & Goldman, 1986). In addition, numerous IF can be with IF (Figs 23, 24; and Green & Goldman, 1986). The fine fibrillar material probably represents the F-actin-enriched region of the nodule as seen by fluorescence microscopy. In addition, as mentioned above for fluorescence, we have seen close associations between MF bundles and IF by electron microscopy in spread CEF (Fig. 25; and Green & Goldman, 1986).

IF also interact with IF as determined ultrastructurally. For example, parallel arrays of IF are seen both in the juxtanuclear caps in the early stages of spreading in BHK cells and, during spreading, in the cytoplasm between the nucleus and cell surface (Figs 26, 27; and Goldman & Follett, 1970; Goldman & Knipe, 1973; Goldman et al. 1973). In epithelial cells, IF bundles can be found that form the tonofibrils resolved by light optical methods (Jones et al. 1982). These consist of tightly packed parallel arrays of IF (Fig. 28). In the case of epidermal cells, these latter arrays are cross-linked by a protein called filaggrin (Steinert et al. 1981).

When cultured cells are extracted with high ionic strength detergent-containing solutions (0&6m-KCl, 1% Triton X-100 in phosphate-buffered saline) and subsequently treated with DNase I (Starger & Goldman, 1977; Zackroff & Goldman, 1979), a so-called cytoskeletal preparation is formed that is greatly enriched in IF. Despite these rather drastic lysis conditions, the IF-enriched cytoskeletons appear to retain their overall shape in both fibroblasts and epithelial cells, as determined by whole-mount electron-microscopic observations (Figs 29, 30; Jones et al. 1982; Green & Goldman, 1986). In addition, these cytoskeletons are virtually devoid of other major cytoplasmic structural proteins such as tubulin and actin. In the case of fibroblast cytoskeletons, elements of the extracellular matrix can also be seen in their normal location (Fig. 30; and Green & Goldman, 1986). In the cell centre lies a nuclear remnant and between this latter structure and the cell surface there are large numbers of fibrous components that represent IF networks (Figs 29, 30; and Jones et al. 1982; Green & Goldman, 1986).

When such preparations are fixed and embedded for conventional electron- microscopic observations, more detailed views of the morphological properties of these cytoskeletons can be obtained. Beginning at the level of the nuclear surface, an obvious ‘nuclear ghost’ is present that is decorated on its cytoplasmic face with large numbers of IF (Figs 31–34; and Jones et al. 1982; Goldman et al. 1985). The nuclear ghost contains fine fibrillar material that represents the nuclear lamina region (Figs 31, 32, 34; Jones et al. 1982; Goldman et al. 1985). Nuclear pore complexes are frequently seen embedded in this latter region (Figs 31, 32, 34; and Jones et al. 1982; Goldman et al. 1985). In the case of fibroblast-derived cytoskeletons, IF at the cytoplasmic surface of the nuclear ghost appear to be anchored to a network of finely fibrous material (Figs 31, 32). In epidermal cell preparations, IFB are frequently seen to splay apart as they approximate the nuclear ghost and individual IF appear to terminate at the level of the nuclear pore complexes (Fig. 34; and Jones et al. 1982). The cytoplasmic networks of IF appear to be in their usual organizational states.

Very few other cytoplasmic constituents can be detected (Figs 35, 36; and Jones et al. 1982; Jones & Goldman, 1985; Green & Goldman, 1986). At the level of the cell surface, close associations between IF and the extracellular matrix are retained (Fig. 35; and Green & Goldman, 1986). Frequently, in these regions, there appear to be residual elements of the plasma membrane (membrane ghost material) (Fig. 30; and Green & Goldman, 1986). In the case of cultured keratinocytes, desmosomes as well as their connections with IF are retained (Fig. 36; and Jones et al. 1982; Jones & Goldman, 1985).

The biochemical properties of these IF-enriched cytoskeletal preparations have been analysed in some detail (e.g. see Starger & Goldman, 1977; Zackroff & Goldman, 1979; Steinert et al. 1982; Jones et al. 1982). SDS-PAGE (sodium dodecyl sulphate-polyacrylamide gel electrophoresis) analyses indicate the presence of relatively few polypeptides; the major components being the IF structural proteins. Since these vary from cell type to cell type, this is reflected in differences in the major proteins comprising their cytoskeletons. For example, in BHK preparations, two major polypeptides are apparent, of molecular weights 54 K and 55 K (the 54 K component is frequently termed desmin and the 55 K, vimentin) (Fig. 37; and Starger & Goldman, 1977), whereas in the case of PME cells, several bands of the keratin-type of IF structural protein are resolved (Fig. 38; and Jones et al. 1982; Jones & Goldman, 1985).

Besides the major building-block proteins that make up the walls of IF, there are numerous other proteins that are consistently found in these IF-enriched cytoskeletons. In the majority of such preparations studied to date, which include several types of fibroblasts, epithelial cells and nerve cells, we have identified several proteins in the 60-70 K range (Fig. 37). In the case of BHK cytoskeletal preparations, there are four polypeptides of 60 K, 65 K, 67 K and 70 K (Fig. 37; and Zackroff et al. 1984). We have been able to purify these proteins as paracrystalline arrays (Figs 39, 40) and to demonstrate their keratin-like nature on the basis of their amino acid composition, their alpha-helix content and their cross-reactivity with antibodies directed against keratin (Zackroff etal. 1984; Goldman etal. 1986). Antibodies directed against these proteins have been prepared and when used in indirect immunofluorescence they generate a bright ring around the edge of the nucleus and diffuse fluorescence throughout the nuclear matrix (Fig. 41; and Goldman etal. 1985, 1986), indicating that this set of IF-like proteins is concentrated at the nuclear surface. From their size and cellular location, these proteins appear to be very similar to a set of nuclear matrix-derived proteins, which have been termed the nuclear lamins (Gerace et al. 1984), and which are thought to comprise the nuclear lamina. Indeed, we have been able to demonstrate that bona fide nuclear lamins derived from isolated BHK nuclei (Fig. 40) form identical paracrystalline arrays as described for the 60–70 K proteins that co-isolate in the IF-enriched cytoskeletal preparations (Goldman et al. 1986). Furthermore, antibodies directed against the nuclear lamins recognize the 60–70 K cytoskeletal polypeptides and vice versa (Goldman et al. 1986). Moreover, it has recently been shown that nuclear lamin sequences derived from cDNA clones are homologous to keratin sequences (McKeon et al. 1986).

Therefore it appears that the peripheral region of the nucleus contains a set of IF- like proteins. This latter term is used to describe these proteins, which have numerous biochemical and immunological properties indicating their similarities to the IF protein family, yet they appear to be incapable of forming IF either in situ or in vitro. These 60–70K keratin-like proteins are probably major structural constituents of the fine fibrillar material associated with the ends of the IF as seen in Fig. 32. We find these results intriguing as they suggest, but by no means prove, that the 60—70K proteins may comprise a binding domain to which the cytoskeletal IF may be anchored, albeit indirectly (see below; Goldman etal. 1985, 1986). These results are even more provocative if one considers the possibility that the nuclear lamina and its constituent proteins are thought to be involved in the attachment of chromatin to the nuclear surface, and in this fashion may be involved in regulating gene activity (Gerace et al. 1984). It is conceivable, therefore, that an insoluble, relatively amorphous IF-like protein system localized at the nuclear surface may interact with the cytoplasmic IF system, providing a skeletal continuum involved in nuclear-cytoplasmic communication. However, it should be emphasized that this is speculative as there are no known molecular connecting links between cytoplasmic IF and the nuclear IF-like system. If such connecting links exist, they would probably be in the form of complex transmembranous elements or transnuclear porecomplex components (Goldman et al. 1986).

There are also higher molecular weight components in the IF-enriched cytoskeletal preparations, some of which have been investigated using biochemical and immunochemical approaches. For example, in fibroblasts such as BHK and CEF there is a doublet or triplet of polypeptides at approximately 220 K, which by immunoblotting analyses are recognized by fibronectin antibody preparations (Fig. 42; and Green & Goldman, 1986). Furthermore, fibronectin is absent in the cytoskeletal preparations if they are prepared from freshly trypsinized cells, lending support to the idea that IF are part of a ‘cell membrane complex’, as mentioned above (Green & Goldman, 1986).

In the case of BHK cytoskeletal preparations we have initiated studies on a 300 K protein, which we have termed the 300 K IFAP (IF-associated protein) (Yang et al.1985). A monoclonal antibody has been prepared that reacts specifically with this protein and in double-label immunofluorescence assays, it appears to co-localize with IF (Figs 43, 44; and Yang et al. 1985). Observations made using the immunogold method at the ultrastructural level, indicate that this protein is localized in areas of close association between IF, indicating an involvement in cross-linking IF to IF (Fig. 45). More recently, we have been able to purify this protein and to demonstrate that it forms complexes with in vitro-reconstituted IF (Lieska et al. 1985).

We have also spent considerable time studying some of the high molecular weight proteins associated with IF in cultured epidermal cells (Jones & Goldman, 1985). One of these turns out to be desmoplakin (Franke etal. 1983), a proposed component of the desmosomal plaque (see above). Indeed, the possibility that desmoplakin might represent another species of IFAP is suggested by its association with bundles of IF found in juxtanuclear complexes long before desmosome assembly occurs in PME cells grown in medium containing low levels of calcium (Figs 46, 47; and Jones & Goldman, 1985). This association has also been shown by immunogold localization (Jones & Goldman, 1985) and by immunoblotting analyses of IF- enriched cytoskeletal preparations made before and at various times after the initiation of desmosome assembly, following the switch to higher levels of calcium in the growth medium (Fig. 48; and Jones & Goldman, 1985). Moreover, once desmosome assembly has occurred and normal IF-desmosome complexes have formed along the borders of contiguous epidermal cells, cytoskeletons can be prepared using the method described above. We have found that the cytoskeletal preparations of these cells retain their intercellular desmosomal contacts by morphological criteria (see Fig. 36; and Jones et al. 1982; Jones & Goldman, 1985), and that desmoplakins 1 and 2 are both major constituents of the IF-enriched cytoskeletons (Fig. 48; and Jones & Goldman, 1985).

At this juncture it is worth reviewing briefly some of our work on the desmosomal system, specifically with regard to its subfractionation into constituent cytoplasmic and cell surface domains. With regard to the cell surface domains, we have been able

Immunofluorescence analyses have indicated that at least some of the antibodies found in the serum of patients with this autoimmune disease are directed against desmosomal components (Figs 49, 50; Jones et al. 1984, 1986a), and that these sera induce the separation of adherent keratinocytes in culture (Jones et al. 1984). Furthermore, immunoblotting analyses using cell-free preparations of bovine des- mosomes (Fig. 51; and Jones et al. 1986a) indicate that there are several major antigens with which pemphigus serum samples react. In the case of pemphigus vulgaris, the majority of patients exhibit a reaction with a 140 K glycoprotein found primarily in bovine tongue mucosa desmosomes (Jones et al. 1986a), while in some pemphigus foliaceus patients, the most obvious consistent reaction is with a doublet of proteins in the 160-165 K range (Fig. 51; and Koulu et al. 1984; Jones et al. l986a). The pemphigus vulgaris reaction is not detectable in desmosomes derived from a closely related stratified squamous epithelial tissue obtained from the epidermis of cow snouts (Jones et al. 1986a). Since these initial observations, we have prepared rabbit antibodies directed against this 140 K protein. We have been able to determine that this antibody preparation induces cell separation in living cultured mouse keratinocytes possessing desmosomal junctions, and furthermore recognizes a cell surface component (Jones et al. 19866). More recently we have been able to demonstrate that this glycoprotein is present in membrane fractions obtained from bovine tongue desmosomes following extraction with urea (Fig. 52; and Jones et al. unpublished). Indeed, the 140K protein possesses the characteristics of a cell adhesion molecule, i.e. it is a glycoprotein located on the cell surface, and an antibody directed against it disrupts cell-cell interactions (Edelman, 1982).

We have also been attempting to determine the nature of the cytoplasmic domains of isolated desmosomes (Fig. 53). In the urea-soluble fraction of desmosome preparations there is enrichment for desmoplakin and several polypeptides in the 50-65 K range that react with keratin antibodies. However, we have been unable to assemble IF from this urea-soluble fraction when it is placed into keratin assembly buffers. Indeed, under these conditions, a precipitate is formed that contains fibrillar material with no obvious IF (Fig. 54; and Jones et al. unpublished). This material appears to be morphologically similar to the fine fibrillar region that comprises the innermost (cytoplasmic) portion of the desmosome to which IF appear to attach in situ (Jones et al. 19866). We are tempted to speculate that the interaction of desmoplakin with these keratins may reflect a mechanism by which IF associate with the desmosomal plaque. On the basis of these observations we hope to be in a position to study the ‘chain’ of molecules that is involved in interactions between the IF bundles of neighbouring cells via their desmosomal junctions.

The data reviewed in this paper provide support for the hypothesis that the IF system and its associated proteins may represent a chain of molecular connecting links between the nucleus and the cell surface. To date, we believe that we have characterized a number of links in this chain that has suggested to us that the IF to determine that there are glycoproteins present that represent cell-type-specific adhesion molecules. The first clues that suggested that desmosomes might contain specific cell—cell adhesion (CAM) molecules came from studies of human autoantibodies found in patients with pemphigus, a blistering disease of the skin (Beutner et al. 1970; Lever & Schaumberg-Lever, 1983). In this disease, blister formation is coincident with a loss of cell-cell adhesion in the epidermis, followed by cell death (Beutner et al. 1970; Lever & Schaumberg-Lever, 1983). This process is known as acantholysis. The acantholytic process takes place either in the suprabasal cell layer or in the upper cell layers of the epidermis. In the former case, the disease subclass is termed pemphigus vulgaris, while in the latter case it is called pemphigus foliaceus (Beutner et al. 1970; Lever & Schaumberg-Lever, 1983). networks in cells may represent a type of variable cytoplasmic infrastructure upon which other reactions or interactions take place. These may include interactions with the plasma membrane—extracellular matrix, interactions with other cytoskeletal components such as microtubules and microfilaments and the nuclear surface-nuclear matrix. Such a system may function in many aspects of cell surface-nuclear interactions, cell-cell adhesion and intercellular cytoskeletal interactions that are important in tissue structure and homeostasis.

We thank Drs James Arnn and Andrew Staehelin for the gift of the desmoplakin antiserum. Dr Ken Yamada kindly provided an antibody directed against fibronectin. We also thank Ms Laura Davis, Manette McReynolds and Jack Gibbons for their assistance in the preparation of the manuscript.

Research support has been provided by grants to Dr Robert D. Goldman from the National Cancer Institute and the National Institutes of General Medical Sciences.