Vimentin and keratin intermediate filament systems in cultured PtK2 epithelial cells are interrelated.

Certain cultured epithelial cells contain separate vimentin and keratin-type intermediate filament networks. The intracellular injection of monoclonal antibodies directed against either vimentin or keratin filaments into PtK2 cultured epithelial cells specifically disrupted the organization of both filament types. Neither antibody had any effect when injected into cells which, while containing vimentin or keratin filaments, lacked the specific filament type which that antibody recognized. These experiments suggest that keratin and vimentin filament networks are associated in some way with one another.     

Monoclonal antibodies provide specific intramolecular markers for the study of epithelial tonofilament organization

The tonofilament-associated protein antigens recognized in epithelial cells by a group of six monoclonal antibodies have been studied by immunofluorescence and gel immunoautoradiography. The monoclonal antibodies were generated against detergent insoluble cytoskeleton extracts from a cultured simple epithelium derived cell line, Ptk1 cells. They show various tissue specificities, and while they all recognize components at the low end of the molecular weight range for intermediate filament proteins, they confirm that single antibody species can react with multiple polypeptides of different molecular weights in the tonofilament complex. The monoclonal antibodies described here demonstrate the presence of a simple epithelium antigenic determinant associated with intermediate filaments that is not detectable in the specialized cells of squamous and keratinizing epithelia but can reappear in such cells after transformation.     

Intermediate filaments in the developing type II cell of fetal monkey lung

Anticytokeratin monoclonal antibody was used to study epithelial cell development in fetal monkey lungs taken from animals of different ages. It is well established that the overall maturity of fetal lung depends greatly on the maturation of type II epithelial cells in the alveolus. In this study, we have correlated the cytokeratin phenotype of mammalian epithelial cells with pneumocyte maturation. We show that differentiation and maturation of the type II cell is related to intermediate filament expression. Twenty-four fetal monkeys (Macaca nemestrina) were delivered by cesarean section at a gestational age of 135-140 days (term = 168 days) and divided into two groups. One group of animals was sacrificed during the first 3 hr of life, and the other group was maintained in incubators for 92-120 hr. Anticytokeratin monoclonal antibody recognizes only alveolar type I and type II epithelial cells. In the first 3 hr of life, the cytokeratin was localized only at the alveolar surface and at the cytoplasmic periphery of the type II cells of these premature animals. However, at the age of 92-120 hr, the epithelia in the lungs reacted more intensely than they did during the first 3 hr. Electron microscopy revealed and confirmed that the type II cells were matured and abundant intermediate filaments appeared in the cytoplasm. The filaments appeared to form either aggregates or parallel filament bundles and few were closely associated with the lamellar bodies. In the immature type II cells at 0-3 hr of life, few intermediate filaments could be localized in the cytoplasm, and no parallel filament bundle was observed, though many appeared in the 92-120 hr lungs. This suggests that the intermediate filaments have a functional significance in the development and maturation of the type II cell. The location and stability of keratin filaments in type II cells may confer the structural strength necessary for cells covering a free surface in the alveoli during lung maturation.     

Intermediate filament proteins in nonfilamentous structures: Transient disintegration and inclusion of subunit proteins in granular aggregates

The intermediate filament cytoskeleton of cultured bovine kidney epithelial cells and human HeLa cells changes dramatically during mitosis. The bundles of cytokeratin and vimentin filaments progressively unravel into protofilament-like threads of 2–4 nm diameter, and intermediate filament protein is included in numerous, variously sized (2–15 μm) spheroidal aggregates containing densely stained granular particles of 5–16 nm diameter. We describe these mitotic bodies in intact cells and in isolated cytoskeletons. In metaphase to anaphase of normal mitosis and after colcemid arrest of mitotic stages, many cells contain all their detectable cytokeratin and vimentin material in the form of such spheroidal aggregate bodies, whereas in other mitotic cells such bodies occur simultaneously with bundles of residual intermediate filaments. In telophase, the extended normal arrays of intermediate filament bundles are gradually reestablished. We find that vimentin and cytokeratins can be organized in structures other than intermediate filaments. Thus, at least during mitosis of some cell types, factors occur that promote unraveling of intermediate filaments into protofilament-like threads and organization of intermediate filament proteins into distinct granules that form large aggregate bodies. Some cells, at least certain epithelial and carcinoma cells, may contain factors effective in structural modulation and reorganization of intermediate filaments.     

Immunofluorescent and immunogold replica studies of desmin distribution in cultured normal and cardiomyopathic hamster heart cells.

The architecture of desmin intermediate filament arrangements in cultured cardiomyocytes from heart of normal and cardiomyopathic hamsters was studied by immunofluorescent light microscopy and immunogold replica electron microscopy. Both polyclonal and monoclonal antidesmin antibodies were used in a biotin-streptavidin system. Immunofluorescent staining of normal and cardiomyopathic myocytes for desmin at 5 days in culture exhibited filamentous staining patterns with polyclonal antidesmin and a coarse punctate staining pattern with the monoclonal antibody. At 9 days in culture, most normal myocytes showed filamentous staining with the polyclonal antibody; many of the stained filaments were associated with Z lines. With the monoclonal antidesmin, these same cells exhibited a very fine 'spotty' staining pattern. These results suggest that the arrangements and immunoreactivities of intermediate filaments change during normal cardiac myocyte development. In cardiomyopathic cells, this pattern of rearrangement and immunoreactivity appears to be delayed or possibly nonexistent. The three-dimensional electron-microscopic observation of immunogold localization of desmin achieved by a deep-etching replica technique is made on both normal and cardiomyopathic cultured heart cells. Abnormalities of desmin filament arrangements in cardiomyopathic cells are confirmed.     

Transient change of organization of vimentin filaments during mitosis as demonstrated by a monoclonal antibody

A monoclonal antibody specific for vimentin is described which, by immunofluorescence and immunoelectron microscopy, decorates fibrillar and/or granular structures in mitotic and early postmitotic cells but does not react with vimentin filaments of interphase stages of various cultured cells (rat vascular smooth muscle-derived cell line RVF-SM; SV40-transformed human fibroblasts; bovine kidney epithelial cells of line MDBK). These observations indicate that the organization of vimentin filaments varies during the cell cycle, undergoing a perimitotic change of filament organization. These changes of vimentin filaments are described in relation to those reported for cytokeratin filaments of various epithelial and carcinoma cells. The possible functional implications of filament protein rearrangements both during the cell cycle and in cell differentiation processes are discussed.     

Cytoplasmic network arrays demonstrated by immunolocalization using antibodies to a high molecular weight protein present in cytoskeletal preparations from cultured cells

Antisera were raised in rabbits against the two major components of intermediate filament preparations from glia-derived C6 cells, polypeptides of M_r around 300 000 and 58 000 (vimentin). These, and a third antiserum raised against microtubule proteins from hog brain, were shown to be specific for their respective immunogens. The assay employed involved the separation of components of crude cell extracts or filament preparations by SDS-polyacrylamide gel electrophoresis, and their subsequent transfer to and immobilization on nitrocellulose sheets. Cross-reacting counterparts of the immunogens were found in various cell lines, including C6, BALB/c 3T3, SV101, CHO, HeLa and PtK2 cells. In indirect immunofluorescence studies, antibodies to the high-M_r polypeptide component stained dense cytoplasmic network arrays of seemingly short, irregularly oriented fibres and lines of dots, in fibroblasts and in HeLa cells, but not in PtK2 cells. In well spread cells these networks were clearly distinguishable in morphology from the fibres decorated by antibodies to either microtubule protein or vimentin. The network arrays were resistant towards treatments with Triton X-100 and colcemid. By double immunofluorescence microscopy of single cells, using an additional antibody preparation to vimentin raised in guinea pigs, it was shown that after prolonged colcemid treatment of BALB/c 3T3 cells both; vimentin filaments and the structures stained by antibodies to the high-M_r component, accumulated in corresponding areas of the cytoplasm. The possibilities are discussed that this novel network-like structure is of the intermediate filament type and that it might function as a cross-linker of cytoplasmic—in particular cytoskeletal—elements. To signify its fluorescent localization and its possible linking role it is proposed to call the high-M_r component of intermediate filament preparations from cultured cells ‘plectin’.     

Morphology, behavior, and interaction of cultured epithelial cells after the antibody-induced disruption of keratin filament organization

The organization of intermediate filaments in cultured epithelial cells was rapidly and radically affected by intracellularly injected monoclonal antikeratin filament antibodies. Different antibodies had different effects, ranging from an apparent splaying apart of keratin filament bundles to the complete disruption of the keratin filament network. Antibodies were detectable within cells for more than four days after injection. The antibody-induced disruption of keratin filament organization had no light-microscopically discernible effect on microfilament or microtubule organization, cellular morphology, mitosis, the integrity of epithelial sheets, mitotic rate, or cellular reintegration after mitosis. Cell-to-cell adhesion junctions survived keratin filament disruption. However, antibody injected into a keratinocyte-derived cell line, rich in desmosomes, brought on a superfasciculation of keratin filament bundles, which appeared to pull desmosomal junctions together, suggesting that desmosomes can move in the plane of the plasma membrane and may only be 'fixed' by their anchoring to the cytoplasmic filament network. Our observations suggest that keratin filaments are not involved in the establishment or maintenance of cell shape in cultured cells.     

Disruption of the in vivo distribution of the intermediate filaments in fibroblasts through the microinjection of a specific monoclonal antibody

Monoclonal antibodies (JLB1 and JLB7) that recognize minor components of the intermediate filament system of cultured cells were introduced into living fibroblasts by microinjection. Several minutes after injection of the JLB7 antibody virtually all of the intermediate filaments of the cells were found to be aggregated into tight bundles near or around the nucleus. In contrast, injection of the JLB1 antibody caused little or no aggregation of the intermediate filaments. Electron microscopy showed that the perinuclear bundles that formed after injection of the JLB7 antibody each consisted of ten or more intermediate filaments apparently crosslinked together. Double-label immunofluorescence microscopy showed that virtually all of the vimentin-containing intermediate filaments in the JLB7 antibody-injected cells were redistributed to the perinuclear region and remained there for at least 24 hr. The distributions of actin microfilaments and microtubules were seemingly undisturbed following microinjection. No obvious changes in cell morphology or behavior were apparent in the cells injected with JLB7 antibody; the cells displayed a flat appearance, showed a polarity, were able to ruffle and bleb and even appeared to show the normal saltatory movements of intracellular vesicles, granules and mitochondria, suggesting that intermediate filaments are not involved in these activities. The microinjection of highly specific monoclonal antibodies that recognize and alter components of the cell provides an additional approach to determine the in vivo functions of intracellular elements.     

On the relation between distinct components of the cytoskeleton: an epitope shared by intermediate filaments, microfilaments and cytoplasmic foci.

Monoclonal antibodies were generated against detergent-insoluble cytoskeletal proteins isolated from low-density membrane fractions of rat liver. By immunofluorescence, one of the antibodies stains three distinct structures in cultured rat fibroblast and hepatocyte lines as well as the PtK2 rat-kangaroo kidney epithelial line. These structures are: i) many tangled filaments similar to intermediate filaments (IFs), ii) fewer and variable numbers of straight filaments, and iii) punctate cytoplasmic foci, often most intense around the nucleus. All three of these structures are resistant to extraction by non-ionic detergent. Close examination reveals that the tangled and straight filaments are not stained uniformly, but as a series of bright patches. In cells treated with nocodazole, the antibody reacts strongly with a perinuclear filamentous cage. Very few tangled filaments are detected in these cells, however, the straight filaments and punctate cytoplasmic staining are resistant to nocodazole treatment. Double-label immunofluorescence shows that, even though tangled filament distribution and nocodazole sensitivity are similar to the behavior of vimentin IFs, there is only partial coincidence of staining with either vimentin or cytokeratin IFs. The straight filaments coincide with some actin stress fibers, but the punctate cytoplasmic staining is not related to IFs, actin, or tubulin. Thus, this monoclonal antibody stains a novel group of three seemingly unrelated cytoskeletal structures, including a previously undescribed insoluble nonfilamentous pool. Taken as a whole, two hypotheses are consistent with these data. i) The antigen recognized may be a protein which has a large insoluble cytoplasmic pool and binds both IFs and actin, but only binds to a subset of each class of filaments.(ABSTRACT TRUNCATED AT 250 WORDS)     

Immunofluorescent localisation of cytokeratin antigens in mitotic HeLa cells using monoclonal antibodies

The organisation of cytokeratin filaments in mitotic HeLa cells has been analysed by immunofluorescence microscopy using a monoclonal antibody which recognises proteins with apparent subunit molecular weights of 52 kDa and 57 kDa and which binds exclusively to cytokeratin-type filaments. Mitotic cells were prepared for microscopic analysis by hypotonic swelling, centrifugation onto glass slides, brief pre-extraction with 0.1% Triton X-100 and fixation in 80% ethanol. This procedure gave particularly good resolution of intermediate filaments and preservation of chromosome morphology. In prometaphase-metaphase cells the antigen was present in an anastomosing filament network which completely or partially enclosed the chromosomes, in filament fragments and in cytoplasmic aggregates. The epichromosomal filament network was absent from cells in anaphase or later stages of mitosis. In these cells non-filamentous antigen was often located in a narrow band defining the periphery of individual chromosomes and in variable numbers of cytoplasmic filaments or fragments. The results suggest that extensive disaggregation and reformation of cytokeratin filaments occurs during mitosis and that disaggregated cytokeratin proteins are frequently located adjacent to mitotic chromosomes.     

Intermediate filament protein partnership in astrocytes.

Intermediate filaments are general constituents of the cytoskeleton. The function of these structures and the requirement for different types of intermediate filament proteins by individual cells are only partly understood. Here we have addressed the role of specific intermediate filament protein partnerships in the formation of intermediate filaments in astrocytes. Astrocytes may express three types of intermediate filament proteins: glial fibrillary acidic protein (GFAP), vimentin, and nestin. We used mice with targeted mutations in the GFAP or vimentin genes, or both, to study the impact of loss of either or both of these proteins on intermediate filament formation in cultured astrocytes and in normal or reactive astrocytes in vivo. We report that nestin cannot form intermediate filaments on its own, that vimentin may form intermediate filaments with either nestin or GFAP as obligatory partners, and that GFAP is the only intermediate filament protein of the three that may form filaments on its own. However, such filaments show abnormal organization. Aberrant intermediate filament formation is linked to diseases affecting epithelial, neuronal, and muscle cells. Here we present models by which the normal and pathogenic functions of intermediate filaments may be elucidated in astrocytes.     

Assembly of type IV neuronal intermediate filaments in nonneuronal cells in the absence of preexisting cytoplasmic intermediate filaments

We report here on the in vivo assembly of alpha-internexin, a type IV neuronal intermediate filament protein, in transfected cultured cells, comparing its assembly properties with those of the neurofilament triplet proteins (NF-L, NF-M, and NF-H). Like the neurofilament triplet proteins, alpha-internexin coassembles with vimentin into filaments. To study the assembly characteristics of these proteins in the absence of a preexisting filament network, transient transfection experiments were performed with a non-neuronal cell line lacking cytoplasmic intermediate filaments. The results showed that only alpha-internexin was able to self-assemble into extensive filamentous networks. In contrast, the neurofilament triplet proteins were incapable of homopolymeric assembly into filamentous arrays in vivo. NF-L coassembled with either NF-M or NF-H into filamentous structures in the transfected cells, but NF-M could not form filaments with NF-H. alpha- internexin could coassemble with each of the neurofilament triplet proteins in the transfected cells to form filaments. When all but 2 and 10 amino acid residues were removed from the tail domains of NF-L and NF-M, respectively, the resulting NF-L and NF-M deletion mutants retained the ability to coassemble with alpha-internexin into filamentous networks. These mutants were also capable of forming filaments with other wild-type neurofilament triplet protein subunits. These results suggest that the tail domains of NF-L and NF-M are dispensable for normal coassembly of each of these proteins with other type IV intermediate filament proteins to form filaments.     

The enhanced association of keratin with hepatoma cell nuclei

A monoclonal antibody to rat hepatoma keratin demonstrates a close association of intermediate filaments with the nucleus in hepatoma cells. Immunoblot analysis of nuclear fractions and immunofluorescence of nuclei both prepared by standard procedures, indicate that intermediate filament proteins are consistently present. Sodium citrate extraction of these preparations diminishes the amount of intermediate filament proteins but does not totally remove the antigenic moieties, suggesting a tight association of intermediate filaments with nuclei. The results from both immunoblot analysis and immunofluorescent localization demonstrate the increased amount of keratins associated with hepatoma cell nuclei.     

Monoclonal antibodies against plant proteins recognise animal intermediate filaments

Four monoclonal antibodies were raised against polypeptides present in a high-salt detergent-insoluble fraction from cells of Chlamydomonas reinhardtii. Indirect immunofluorescence microscopy of fibroblasts and epithelial cells grown in culture using these plant antibodies revealed staining arrays identical to those obtained with well characterised antibodies to animal intermediate filaments. Immunofluorescence microscopy of Chlamydomonas with these monoclonal antibodies and a monoclonal antibody that recognises all animal intermediate filaments (anti-IFA) gave a diffuse, patchy cytoplasmic staining pattern. Both the plant antibodies and anti-IFA stained interphase onion root tip cells in a diffuse perinuclear pattern. In metaphase through to telophase, the labelling patterns colocalised with those of microtubules. Labelling of the phragmoplast was also detected but not staining of the preprophase band. On Western blots of various animal cell lines and tissues, all the antibodies labelled known intermediate filament proteins. On Western blots of whole Chlamydomonas proteins, all the antibodies labelled a broad band in the 57,000 Mr range, and three antibodies labelled bands around 66,000 and 140,000 Mr but with variable intensities. On Western blots of whole onion root tip proteins, all the antibodies labelled 50,000 Mr (two to three bands) polypeptides and a diffuse band around 60,000 Mr and three of the antibodies also labelled several polypeptides in the 90,000-200,000 Mr range. The consistent labelling of these different bands by several different monoclonal antibodies recognising animal intermediate filaments makes these polypeptides putative plant intermediate filament proteins.     

Vimentin: the conundrum of the intermediate filament gene family

Intermediate filaments are a major component of the “cytoskeleton” of “higher” eukaryotes. These filaments are composed of a number of different, although structurally related, proteins. Different intermediate filament protein genes are expressed in different tissues. Spontaneous and experimentally produced mutations in the intermediate filament genes indicate that these filaments function to enhance the mechanical stability of epidermal and muscle cells. As a result, the use of transgenic mice with “knockout” or dominant negative mutations in IF genes has become an important approach for investigating the significance of IFs in other cell types. However, a knockout mutation of vimentin (^-/-), the intermediate filament protein characteristically expressed in cells of mesenchymal origin, results in very subtle phenotypes that are not obviously related to cell fragility. Although experiments with cultured cells have described a variety of discrete changes in cell properties that are associated with vimentin expression or organization, there is no evidence yet that any of these properties are affected in the vimentin^-/- mouse. BioEssays 20:79–86, 1998. © 1998 John Wiley & Sons, Inc.     

Induction of cytokeratin expression in human mesenchymal cells

We studied the phenotypic features of some typical human mesenchymal cells, including decidual stromal cells and adult and fetal fibroblasts under different cell culture conditions by using antibodies to intermediate filament proteins and desmoplakins. In cell culture, the decidual stromal cells rapidly acquired typical fibroblastoid appearance with abundant arrays of vimentin filaments while the cytokeratin-positive epithelial cells, occasionally found in typical epithelioid colonies, lacked vimentin positivity and showed desmoplakin positivity. Within a few days, many of the stromal cells started to present cytokeratin positivity when cultured either in Condimed or in Chang medium. The cytokeratin positivity was first detected in small, scattered cytoplasmic dotted fibrils or in perinuclear dotlike aggregates with fibrillar projections. Later, denser cytokeratin-positive fibrillar arrays could also be seen in stromal cells, which lacked desmoplakin positivity as judged by two monoclonal antibodies. Decidual stromal cells were also cloned and in five out of ten clones some of the cells acquired a similar cytokeratin positivity when transferred into Chang or Condimed medium. Immunoblotting results indicated that cytokeratins 8, 18, and 19 can be found in these cultures. Similar cytokeratin positivity could also be seen in the same culture conditions in cultured fetal fibroblasts from skin, chorionic villi, and lung but not in young or adult skin fibroblast cultures. The present results suggest that decidual stromal cells as well as some embryonal mesenchymal cells can acquire epithelial differentiation in vitro as judged by the emergence of cytokeratin proteins. This ability appears to be lost in the corresponding adult cell. The results furthermore suggest that cytokeratin fibrils can be organized in the cytoplasm without an apparent organization center and that neither the appearance of desmoplakins nor the formation of cell-to-cell contacts are required for cytokeratin filament assembly.     

Visualization of intermediate filaments in living cells using fluorescently labeled desmin

Fluorescently labeled desmin was incorporated into intermediate filaments when microinjected into living tissue culture cells. The desmin, purified from chicken gizzard smooth muscle and labeled with the fluorescent dye iodoacetamido rhodamine, was capable of forming a network of 10-nm filaments in solution. The labeled protein associated specifically with the native vimentin filaments in permeabilized, unfixed interphase and mitotic PtK2 cells. The labeled desmin was microinjected into living, cultured embryonic skeletal myotubes, where it became incorporated in straight fibers aligned along the long axis of the myotubes. Upon exposure to nocodazole, microinjected myotubes exhibited wavy, fluorescent filament bundles around the muscle nuclei. In PtK2 cells, an epithelial cell line, injected desmin formed a filamentous network, which colocalized with the native vimentin intermediate filaments but not with the cytokeratin networks and microtubular arrays. Exposure of the injected cells to nocadazole or acrylamide caused the desmin network to collapse and form a perinuclear cap that was indistinguishable from vimentin caps in the same cells. During mitosis, labeled desmin filaments were excluded from the spindle area, forming a cage around it. The filaments were partitioned into two groups either during anaphase or at the completion of cytokinesis. In the former case, the perispindle desmin filaments appeared to be stretched into two parts by the elongating spindle. In the latter case, a continuous bundle of filaments extended along the length of the spindle and appeared to be pinched in two by the contracting cleavage furrow. In these cells, desmin filaments were present in the midbody where they gradually were removed as the desmin filament network became redistributed throughout the cytoplasm of the spreading daughter cells.     

Identification of an intracellular postsynaptic antigen at the frog neuromuscular junction

A layer of amorphous, electron-dense material is situated at the cytoplasmic surface of the postsynaptic membrane of vertebrate neuromuscular synapses. The function of this structure is not clear, but its location suggests that it may have an important role in the formation and/or maintenance of the synapse. This paper demonstrates that a monoclonal antibody raised against antigens from Torpedo electric organ binds to an intracellular, postsynaptic protein at the frog neuromuscular synapse. Indirect immunofluorescence on frozen sections of frog muscle was used to demonstrate that the antigen is concentrated at synaptic sites in normal muscle. In denervated muscle, the antigen remains concentrated at synaptic sites, but is also present at extrasynaptic regions of denervated myofibers. The antigen cannot be labeled in intact, whole muscle, but only in whole muscle that has been permeabilized with nonionic detergents. The antibody staining pattern in Triton X-100-permeabilized whole-mounts of the frog neuromuscular synapse is arranged in elongate, arborized areas which are characteristic of the frog neuromuscular synapse. The stained areas are striated and the striations occur with a periodicity that corresponds to the regular folding of the postsynaptic membrane. Immunoferritin labeling of fixed, saponin-permeabilized muscle demonstrates that the antigen is associated with amorphous material that is situated between the postsynaptic membrane and an underlying layer of intermediate filaments. The antigen, solubilized from membrane and an underlying layer of intermediate filaments. The antigen, solubilized from Torpedo electric organ by high ionic strength, was identified by antibody binding to nitrocellulose replicas of SDS gels of Torpedo tissue. In Torpedo tissue, the antibody binds to a single protein band at 51,000 daltons (51 kd). The 51-kd protein shares an antigenic determinant with intermediate filament proteins, since a monoclonal antibody to all intermediate filaments reacts with the same 51-kd protein. The monoclonal antibody also reacts with a 55-kd protein in frog skin which is localized to the perinuclear region of the epithelial cells.     

Keratin and carcinoembryonic antigen (CEA) in human melanoma cells.

Human melanomas are known to contain vimentin intermediate filaments but there has been some dispute about their expression of cytokeratins. The cytoplasm of human M21 melanoma cells maintained in culture reacted with a rabbit anti-keratin antibody and two monoclonal anti-keratin antibodies AE1 and AE2. Cells derived directly from subcutaneous xenografts of M21 melanoma in nude mice, however, failed to express cytokeratins. The presence of keratin filaments in cultured M21 cells was confirmed by electronmicroscopic and immuno-electronmicroscopic examinations of cell extracts. Polyacrylamide gel electrophoresis (PAGE), revealed 46 KD keratin proteins in cultured M21 cells. Small amounts of these low molecular weight keratins were detected by PAGE in M21 melanoma xenografts even though immunofluorescence and immunoperoxidase assays failed to demonstrate keratin at the light microscopic level. Immunofluorescence revealed keratin and carcinoembryonic antigen (hitherto undetected in human melanomas) first on the 9th day of culture of xenograft-derived M21 cells. The appearance of keratin and CEA in M21 melanoma cells in vitro was not affected by inhibition of cellular proliferation or as a result of exposure to methotrexate or adriamycin. However, adriamycin altered the cytoplasmic distribution of keratin.