Keratin protein domains within the human epidermis

Three species of human keratins are shown to have specific localizations within the epidermis. Using an immunohistochemical technique with rabbit antisera prepared against purified human keratins, two distinct epidermal domains were defined. The 45K and 46K MW keratins occur predominantly in the basal epidermal layer, whereas 55K keratin protein occurs chiefly in the suprabasal, differentiated squamous cells. Commitment of proliferating basal cells to terminal differentiation is accompanied by changes in the proportions of keratin species.     

The use of aIF, AE1, and AE3 monoclonal antibodies for the identification and classification of mammalian epithelial keratins

Recent data have indicated that specific keratin molecules can provide useful markers for studying different types and stages of epithelial differentiation. To utilize these protein markers, however, it is important to establish the keratin nature of the molecules and identify unambiguously the individual keratin species. In this paper, we show that this can be done relatively easily by one- and two-dimensional gel electrophoresis combined with immunoblotting using three monoclonal antibodies (aIF, AE1, and AE3). The aIF antibody has previously been shown to crossreact with all classes of intermediate-filament proteins. Using one- and two-dimensional immunoblotting, we establish that this antibody recognizes all known epithelial keratins of human and rabbit, although the reaction is relatively strong for the larger, basic keratins and is relatively weak for some of the smaller, acidic keratins. In contrast, AE1 and AE3 monoclonal antibodies have previously been shown to be highly specific for the acidic and basic subfamilies of the keratins, respectively. The combined use of the broadly reacting aIF antibody and the subfamily-specific AE1 and AE3 monoclonal antikeratin antibodies should facilitate the immunological definition, identification, and classification of mammalian epithelial keratins.     

Tissue distribution of keratin 7 as monitored by a monoclonal antibody

Monoclonal antibody (RCK 105) directed against keratin 7 was obtained after immunization of BALB/c mice with cytoskeletal preparations from T24 cells and characterized by one- (1D) and two-dimensional (2D) immunoblotting. In cultured epithelial cells, known from gel electrophoretic studies to contain keratin 7, this antibody gives a typical keratin intermediate filament staining pattern, comparable to that obtained with polyclonal rabbit antisera to skin keratins or with other monoclonal antibodies, recognizing for example keratins 5 and 8 or keratin 18. Using RCK 105, the distribution of keratin 7 throughout human epithelial tissues was examined and correlated with expression patterns of other keratins. Keratin 7 was found to occur in the columnar and glandular epithelium of the lung, cervix, breast, in bile ducts, collecting ducts in the kidney and in mesothelium, but to be absent from gastrointestinal epithelium, hepatocytes, proximal and distal tubules of the kidney and myoepithelium. Nor could it be detected in the stratified epithelia of the skin, tongue, esophagus, or cervix but strongly stained all cell layers of the urinary bladder transitional epithelium. When applied to carcinomas derived from these different tissue types it became obvious that an antibody to keratin 7 may allow an immunohistochemical distinction between certain types of adenocarcinomas.     

Use of monospecific antisera and cRNA probes to localize the major changes in keratin expression during normal and abnormal epidermal differentiation

We report here the isolation and characterization of three antisera, each of which is specific for a single keratin from one of the three different pairs (K1/K10, K14/K5, K16/K6) that are differentially expressed in normal human epidermis and in epidermal diseases of hyperproliferation. We have used these antisera in conjunction with monospecific cRNA probes for epidermal keratin mRNAs to investigate pathways of differentiation in human epidermis and epidermal diseases in vivo and in epidermal cells cultured from normal skin and from squamous cell carcinomas in vitro. Specifically, our results suggest that: (a) the basal-specific keratin mRNAs are down-regulated upon commitment to terminal differentiation, but their encoded proteins are stable, and can be detected throughout the spinous layers; (b) the hyperproliferation-associated keratin mRNAs are expressed at a low level throughout normal epidermis when their encoded proteins are not expressed, but are synthesized at high levels in the suprabasal layers of hyperproliferating epidermis, coincident with the induced expression of the hyperproliferation-associated keratins in these cells; and (c) concomitantly with the induction of the hyperproliferation-associated keratins in the suprabasal layers of the epidermis is the down-regulation of the expression of the terminal differentiation-specific keratins. These data have important implications for our understanding of normal epidermal differentiation and the deviations from this process in the course of epidermal diseases of hyperproliferation.     

Embryogenesis of stratified squamous epithelium. II. Developmental stage-specific immunoreactivity of keratins

Throughout embryogenesis of the rat palate, from the early fetal to adult stages, a consistent subset of keratin proteins is synthesized in the epithelial lining cells. Although the relative abundance of particular keratins has been shown to vary with ongoing palatogenesis, the expression of finite keratins appears to be genetically predetermined. In order to preliminarily ascertain whether conformational changes accompanied intermediate filament 'maturation' from monomeric to polymeric keratin formation, we screened cytokeratins with polyclonal and monoclonal antibodies generated against adult-type keratins. Until epithelial stratification occurred on the 16th day of gestation, the keratin proteins were weakly immunoreactive. On the other hand, subsequent to epithelial thickening, adult-type immunoreactivity was initiated and progressed concomitantly with ongoing palatal development. These findings suggest that the cytokeratin intermediate filaments may progress through conformational 'maturation' during polymerization, and play a role in the eventual acquisition of the adult-type epithelial structure and function.     

Phosphorylation of keratin and vimentin polypeptides in normal and transformed mitotic human epithelial amnion cells: behavior of keratin and vimentin filaments during mitosis

Analysis by means of two-dimensional gel electrophoresis (IEF) of [32P]orthophosphate-labeled proteins from mitotic and interphase transformed amnion cells (AMA) has shown that keratins IEF 31 (Mr = 50,000; Hela protein catalogue number), 36 (Mr = 48,500), 44 (Mr = 44,000), 46 (Mr = 43,500), as well as vimentin (IEF 26; Mr = 54,000) are phosphorylated above their interphase level during mitosis. Similar studies of normal human amnion epithelial cells (AF type) confirmed the above observations except in the case of keratin IEF 44 whose relative proportion was too low to be analyzed. Immunofluorescent staining of methanol/acetone-treated mitotic transformed amnion cells with a mouse polyclonal antibody elicited against human keratin IEF 31 showed a dotted staining (with a fibrillar background) in all of the cells in late anaphase/early telophase (characteristic "domino" pattern) and in a sizeable proportion of the cells in other stages of mitosis. Normal mitotic amnion cells on the other hand showed a fine fibrillar staining of keratins at all stages of mitosis. Similar immunofluorescent staining of normal and transformed mitotic cells with vimentin antibodies revealed a fibrillar distribution of vimentin in both cell types. Taken together the results indicate that the transformed amnion cells may contain a factor(s) that modulates the organization of keratin filaments during mitosis. This putative factor(s), however, is most likely not a protein kinase as transformed amnion cells and amnion keratins are modified to similar extents. It is suggested that in general the preferential phosphorylation of intermediate-sized filament proteins during mitosis may play a role in modulating the various proposed associations of these filaments with organelles and other cellular structures.     

Monoclonal antibody analysis of bovine epithelial keratins. Specific pairs as defined by coexpression

We have characterized the keratin proteins of various bovine epithelial tissues by one- and two-dimensional gel electrophoresis, coupled with the immunoblot technique using AE1, AE2, AE3, AE5, CA20, BE14, and 6.11 monoclonal antikeratin antibodies. The results indicate that all known bovine keratins can be divided into two subfamilies. The "acidic" (Type I) subfamily consists of 41-, 43-, 45-, 46-, 50-, 54-, 56-, and 56.5-kDa keratins, all of which have a pI of less than 5.6, and most of them are recognized by our AE1 antibody, whereas the "neutral-to-basic" (Type II) subfamily consists of 55-, 57-, 58-, 62-65-, 66-, and 67-kDa keratins, all of which have a pI of greater than 6.0 and are recognized by our AE3 antibody. Tissue distribution data and cell culture studies show that, within the two subfamilies, keratins with similar "size ranks" form a "pair" as defined by frequent co-expression. Furthermore, within most "keratin pairs," the basic keratin is larger than the acidic one by 8-10 kDa. These results provide further support for the concepts of "keratin subfamilies" and keratin pairs and are consistent with the possibility that the acidic and basic members of at least some keratin pairs may interact specifically during in vivo tonofilament assembly and/or function. Immunoblotting data derived from the use of several monospecific antibodies show that although the size, charge, and pattern of expression of most bovine keratins are similar to those of the human counterparts, there are important exceptions to this rule.     

Classification of epidermal keratins according to their immunoreactivity, isoelectric point, and mode of expression

Human epidermal keratinocytes express under various growth conditions a total of at least nine keratins that can be divided into two subfamilies. Subfamily A comprises 40-, 46-, 48-, 50-/50'-, and 56.5-kilodalton (kd) keratins which are relatively acidic (pI less than 5.5) and, with the exception of 46-kd keratin, are recognized by AE1 monoclonal antibody. Subfamily B comprises 52-, 56-, 58-, and 65-67-kd keratins which are relatively basic (pI greater than 6) and are recognized by AE3 monoclonal antibody. Within each keratin subfamily, there is a constant member (50-/50'- and 58-kd keratins of the subfamilies A and B, respectively) that is always expressed. The other seven keratins of both subfamilies are variable members whose expression depends upon the cellular differentiated state, which is in turn modulated by the growth environment. The 56.5-kd keratin (subfamily A) and the 65-67-kd keratins (subfamily B) are coordinately expressed during keratinization. In contrast, the 40-, 46-, and 48-kd keratins (subfamily A) and the 52- and 56-kd keratins (subfamily B) are characteristic of cultured epidermal cells forming nonkeratinized colonies. These results demonstrate that human epidermal keratins can be classified according to their reactivity with monoclonal antikeratin antibodies, isoelectric point, and mode of expression. The classification of keratins into various subgroups may have important implications for the mechanisms of epidermal differentiation, the evolution of keratin heterogeneity, and the use of keratin markers for tumor diagnosis.     

A new method for induction of specific polyclonal antibodies using immunostained proteins on nitrocellulose membrane, and HIV 1 core antigen and Escherichia coli verotoxin as examples

Abstract Two polyclonal antibodies against verotoxin 1 and the core protein p24 of HIV 1 were raised in mice by a new immunization procedure. Both proteins were transferred to nitrocellulose, reacted with polyspecific antisera and the antigen-antibody complexes were then visualized by immunostaining. For preparation of antisera the stained protein bands were cut from the nitrocellulose sheets and implanted subcutaneously into the backs of BALB/c mice, without any adjuvant. A single booster was given 4 weeks later by implanting a second strip. All mice produced high titers of antibody directed against the antigen used for immunization. Thus, antibodies of high specificity can be elicited against protein bands in stained immunoblots.     

A new method for induction of specific polyclonal antibodies using immunostained proteins on nitrocellulose membrane, and HIV 1 core antigen and Escherichia coli verotoxin as examples

Two polyclonal antibodies against verotoxin 1 and the core protein p24 of HIV 1 were raised in mice by a new immunization procedure. Both proteins were transferred to nitrocellulose, reacted with polyspecific antisera and the antigen-antibody complexes were then visualized by immunostaining. For preparation of antisera the stained protein bands were cut from the nitrocellulose sheets and implanted subcutaneously into the backs of BALB/c mice, without any adjuvant. A single booster was given 4 weeks later by implanting a second strip. All mice produced high titers of antibody directed against the antigen used for immunization. Thus, antibodies of high specificity can be elicited against protein bands in stained immunoblots.     

The 50- and 58-kdalton keratin classes as molecular markers for stratified squamous epithelia: cell culture studies

The keratins are a highly heterogeneous group of proteins that form intermediate filaments in a wide variety of epithelial cells. These proteins can be divided into at least seven major classes according to their molecular weight and their immunological reactivity with monoclonal antibodies. Tissue-distribution studies have revealed a correlation between the expression of specific keratin classes and different morphological features of in vivo epithelial differentiation (simple vs. stratified; keratinized vs. nonkeratinized). Specifically, a 50,000- and a 58,000-dalton keratin class were found in all stratified epithelia but not in simple epithelia, and a 56,500- and a 65-67,000-dalton keratin class were found only in keratinized epidermis. To determine whether these keratin classes can serve as markers for identifying epithelial cells in culture, we analyzed cytoskeletal proteins from various cultured human cells by the immunoblot technique using AE1 and AE3 monoclonal antikeratin antibodies. The 56,500- and 65-67,000-dalton keratins were not expressed in any cultured epithelial cells examined so far, reflecting the fact that none of them underwent morphological keratinization. The 50,000- and 58,000-dalton keratin classes were detected in all cultured cells that originated from stratified squamous epithelia, but not in cells that originated from simple epithelia. Furthermore, human epidermal cells growing as a monolayer in low calcium medium continued to express the 50,000- and 58,000-dalton keratin classes. These findings suggest that the 50,000- and 58,000-dalton keratin classes may be regarded as "permanent" markers for stratified squamous epithelial cells (keratinocytes), and that the expression of these keratin markers does not depend on the process of cellular stratification. The selective expression of the 50,000- and 58,000-dalton keratin classes, which are synthesized in large quantities on a per cell basis, may explain the high keratin content of cultured keratinocytes.     

The catalog of human hair keratins. II. Expression of the six type II members in the hair follicle and the combined catalog of human type I and II keratins.

The human type II hair keratin subfamily consists of six individual members and can be divided into two groups. The group A members hHb1, hHb3, and hHb6 are structurally related, whereas group C members hHb2, hHb4, and hHb5 are rather distinct. Specific antisera against the individual hair keratins were used to establish the two-dimensional catalog of human type II hair keratins. In this catalog, hHb5 showed up as a series of isoelectric variants, well separated from a lower, more acidic, and complex protein streak containing isoelectric variants of hair keratins hHb1, hHb2, hHb3, and hHb6. Both in situ hybridization and immunohistochemistry on anagen hair follicles showed that hHb5 and hHb2 defined early stages of hair differentiation in the matrix (hHb5) and cuticle (hHb5 and hHb2), respectively. Although cuticular differentiation proceeded without the expression of further type II hair keratins, cortex cells simultaneously expressed hHb1, hHb3, and hHb6 at an advanced stage of differentiation. In contrast, hHb4, which is undetectable in hair follicle extracts and sections, could be identified as the largest and most alkaline member of this subfamily in cytoskeletal extracts of dorsal tongue. This hair keratin was localized in the posterior compartment of the tongue filiform papillae. Comparative analysis of type II with the previously published type I hair keratin expression profiles suggested specific, but more likely, random keratin-pairing principles during trichocyte differentiation. Finally, by combining the previously published type I hair keratin catalog with the type II hair keratin catalog and integrating both into the existing catalog of human epithelial keratins, we present a two-dimensional compilation of the presently known human keratins.     

Expression of keratins during experimentally induced carcinogenesis in hamster cheek pouch visualized polyclonal and monoclonal antibodies

Summary We obtained immnohistochemical profiles of several keratin proteins during experimentally induced carcinogenesis in hamster cheek-pouch mucosa using a polyclonal antibody (TK; detecting keratins with molecular masses of 41 65 kilodalton) and two monoclonal antibodies (KL1, 55- to 57-kilodalton keratins; PKK1; 40-, 45- and 52.5-kilodalton keratins). The squamous epithelium of normal pouch mucosa exhibited positive TK staining in all layers. KL1 staining in the spinous layer and PKK1 staining in the basal layer, thus indicating a regional or zonal distribution pattern. Epithelia undergoing basal hyperplasia showed irregular localization of PKK1 binding, while hyperkeratinized lesions exhibited the binding pattern found in normal epithelium. In case of epithelial dysplasia, there was reduced KL1 staining in spinous cells and decreased PKK1 staining in the basal and parabasal layers. Papillomas exhibited a rather zonal distribution of keratin staining. All squamous-cell carcinomas, irrespective of their degree of keratinization and infiltration pattern, showed slight or no PKK1 staining. Such lesions were only positive for KL1-detectable keratins in keratinizing tumour cells and exhibited an irregular distribution of TK binding. The expression of keratin proteins during carcinogenesis in hamster cheekpouch mucosa may parallel that of keratins in human squamous-cell carcinomas originating in the oral mucosa.     

Monoclonal antibody analysis of keratin expression in epidermal diseases: a 48- and 56-kdalton keratin as molecular markers for hyperproliferative keratinocytes

The polypeptide composition of epidermal keratin varies in disease. To better understand the biological meaning of these variations, we have analyzed keratins from a number of human epidermal diseases by the immunoblot technique using AE1 and AE3 monoclonal antikeratin antibodies. The results reveal a continuous spectrum of keratin expression ranging from one closely resembling the normal in vivo pattern to one almost identical to cultured epidermal keratinocytes. Specifically, a 50-kilodalton (kd) (AE1-positive) and a 58-kd (AE3-positive) keratin are present in all diseases, supporting the concept that they represent "permanent" markers for keratinocytes. A 56.5-kd (AE1) and a 65-67-kd (AE3) keratin, previously shown to be markers for keratinization, are expressed only by lesions retaining a keratinized morphology. A 48-kd (AE1) and a 56-kd (AE3) keratin are present in all hyperproliferative (para- or nonkeratinized) disorders, but not in normal abdominal epidermis or in ichthyosis vulgaris which is a nonhyperproliferative disease. These two keratins have previously been found in various nonepidermal keratinocytes undergoing hyperproliferation, suggesting that these keratins are not epidermis-specific and may represent markers for hyperproliferative keratinocytes in general. In various epidermal diseases, there is a reciprocal expression of the (keratin) markers for hyperproliferation and keratinization, supporting the mutual exclusiveness of the two cellular events. Moreover, our results indicate that, as far as keratin expression is concerned, cultured human epidermal cells resemble and thus may be regarded as a model for epidermal hyperplasia. Finally, the apparent lack of any major, disease-specific keratin changes in the epidermal disorders studied so far implies that keratin abnormalities probably represent the consequence, rather than the cause, of these diseases.     

Various keratin antibodies produce immunohistochemical staining of human myocardium and myometrium

Various polyclonal and monoclonal antibodies to keratins were used to stain different human muscle tissues by paired immunofluorescence and the unlabelled antibody peroxidase-anti-peroxidase method. In the myocardium, distinct coloration of the intercalated discs was produced by two polyclonal reagents to human epidermal keratins but not by two monoclonal antibodies to cytokeratins from pig renal tubular cells. In the myometrium--mainly in the middle layer of the uterine wall--cytoplasmic coloration of a varying fraction of the smooth muscle bundles was produced, especially by one of the polyclonal and by both monoclonal reagents. The staining was often confined to the perinuclear region. The keratin-positive myometrial cells usually coexpressed vimentin and actin in various proportions. These findings indicated that intermediate filaments of the keratin type, or antigenically similar elements, are not restricted to cells of epithelial origin. Other types of muscle cells did not react with keratin antibodies, but keratin-positive macrophages were occasionally found in tongue musculature and in inflamed epicardium. Altogether, our observations emphasize that keratin reactivity cannot be considered specific for epithelial (or mesothelial) cells without reservation.     

Expression of keratins during experimentally induced carcinogenesis in hamster cheek pouch visualized polyclonal and monoclonal antibodies

We obtained immunohistochemical profiles of several keratin proteins during experimentally induced carcinogenesis in hamster cheek-pouch mucosa using a polyclonal antibody (TK; detecting keratins with molecular masses of 41-65 kilodalton) and two monoclonal antibodies (KL1, 55- to 57-kilodalton keratins; PKK1; 40-, 45- and 52.5-kilodalton keratins). The squamous epithelium of normal pouch mucosa exhibited positive TK staining in all layers, KL1 staining in the spinous layer and PKK1 staining in the basal layer, thus indicating a regional or zonal distribution pattern. Epithelia undergoing basal hyperplasia showed irregular localization of PKK1 binding, while hyperkeratinized lesions exhibited the binding pattern found in normal epithelium. In case of epithelial dysplasia, there was reduced KL1 staining in spinous cells and decreased PKK1 staining in the basal and parabasal layers. Papillomas exhibited a rather zonal distribution of keratin staining. All squamous-cell carcinomas, irrespective of their degree of keratinization and infiltration pattern, showed slight or no PKK1 staining. Such lesions were only positive for KL1-detectable keratins in keratinizing tumour cells and exhibited an irregular distribution of TK binding. The expression of keratin proteins during carcinogenesis in hamster cheek-pouch mucosa may parallel that of keratins in human squamous-cell carcinomas originating in the oral mucosa.     

Linkage of human keratin genes

Two families of keratins, type I and type II, can be distinguished within the intermediate filament family of proteins, and at least 20 genes in the human genome code for the 20 known keratin proteins. In epithelial intermediate filaments, keratins from both families appear to be coordinately expressed. We have screened a library of human genomic DNA and have identified several cases of linkage among homologous and heterologous pairs of keratin genes. Genes coding for type I keratins were found linked to those coding for type II keratins. Linkage was discovered also among homologous genes coding for type I keratins and among genes encoding type II keratins. In addition, we found genes coding for glycine-rich keratins linked to genes coding for those that do not contain glycine-rich regions. Our results raise the possibility that all keratin genes are linked in a single region of the human genome.     

The cDNA sequence of a human epidermal keratin: divergence of sequence but conservation of structure among intermediate filament proteins

We have determined the DNA sequence of a cloned cDNA that is complementary to the mRNA for the 50 kilodalton (kd) human epidermal keratin. This provides the first amino acid sequence for a cytoskeletal keratin. Comparison of this sequence with those of other keratins reveals an evolutionary relationship between the cytoskeletal and the microfibrillar keratins, but shows no homology to matrix or feather keratins. The 50 kd keratin shares 28%-30% homology with partial sequences of other intermediate filament proteins, which suggests that keratins may be the most distantly related members of this class of fibrous proteins. Our computer analyses predict that the 50 kd keratin contains two long alpha-helical domains separated by a cluster of helix-inhibitory residues in the middle of the protein. These findings indicate that despite major sequence divergence among intermediate filament proteins, they retain sequences compatible with secondary structural features that appear to be common to all of them.     

Expression of keratins and other cytoskeletal proteins in mouse mammary epithelium during the normal developmental cycle and primary culture

Mammary epithelium is composed of ductal, alveolar, and myoepithelial cells, and undergoes dramatic responses in growth, differentiation, and function to hormonal stimuli during the four stages of the mammary developmental cycle represented in virgin, pregnant, lactating, and involuting animals. To determine if progression of the epithelium through the cycle is accompanied by changes in cytoskeletal composition, particularly the keratins, the polypeptides in cytoskeletal extracts from BALB/c mouse mammary tissues were analyzed by one- and two-dimensional gel electrophoresis combined with immunoblots using polyclonal and monoclonal antikeratin antibodies. The major polypeptides in cytoskeletal fractions enriched in intermediate filaments included seven acidic and three basic components ranging in molecular weight from 40,000 to 90,000. Two major polypeptides of Mr 50,000 and 40,000, along with two minor components of Mr 57,000 and 55,000 were identified as keratins. The polypeptide profiles of mammary glands from virgin, pregnant, lactating, and involuting mice were very similar, indicating a remarkable stability of cytoskeletal composition during hormonal shifts and periods of minimal or maximal cell growth and differentiated function. The data also suggest that ductal and alveolar cells express the same set of cytoskeletal polypeptides, including keratins. Mammary cells grown in primary culture exhibited a loss or reduction in most of the basic polypeptides, a large increase in an acidic Mr 55,000 keratin, and the appearance of a prominent acidic polypeptide of Mr 46,000. The latter results demonstrate that keratin expression in mouse mammary epithelial cells is subject to regulation by certain environmental factors.     

Microinjection of monoclonal antibodies specific for one intermediate filament protein in cells containing multiple keratins allow insight into the composition of particular 10 nm filaments

Monoclonal antibodies specific for vimentin (V9), keratin 7 (CK 7) and keratin 18 (CK5) have been microinjected into three human epithelial cell lines: HeLa, MCF-7 and RT-4. The effect of the injection on other keratin polypeptides and vimentin filaments has been observed by double label immunofluorescence and in some instances by immunoelectron microscopy using gold labels of different sizes. Microinjection of V9 into HeLa cells causes the vimentin to collapse into a perinuclear cap leaving the keratin filaments unaffected. Injection of CK5 does not affect the vimentin filaments but disrupts the keratin filaments revealing keratin aggregates similar to those seen in some epithelial cell lines during mitosis. The keratin aggregates obtained after microinjection in HeLa contain the keratins 8 and 18 and probably also other keratins, as no residual keratin filaments are observed with a keratin polyclonal antibody of broad specificity. Aggregates in mitotic HeLa cells contain at least the keratins 7, 8, and 18. In MCF-7 cells keratins 8, 18, and 19 are observed in the aggregates seen 3 h after microinjection which, however, show a different morphology from those seen in HeLa cells. In MCF-7 cells a new keratin filament is built within 6 h after the injection which is composed mainly of keratin 8 and 19. The antibody-complexed keratin 18 remains in spherical aggregates of different size. The results suggest that in HeLa cells vimentin and keratin form independent networks, and that individual 10 nm filaments in epithelial cell lines can contain more than two keratins.