Phorbol 12-myristate 13-acetate induces protein kinase ceta-specific proliferative response in astrocytic tumor cells

Protein kinase C (PKC) activation has been implicated in cellular proliferation in neoplastic astrocytes. The roles for specific PKC isozymes in regulating this glial response, however, are not well understood. The aim of this study was to characterize the expression of PKC isozymes and the role of PKC-eta expression in regulating cellular proliferation in two well characterized astrocytic tumor cell lines (U-1242 MG and U-251 MG) with different properties of growth in cell culture. Both cell lines expressed an array of conventional (alpha, betaI, betaII, and gamma) and novel (theta and epsilon) PKC isozymes that can be activated by phorbol myristate acetate (PMA). Another novel PKC isozyme, PKC-eta, was only expressed by U-251 MG cells. In contrast, PKC-delta was readily detected in U-1242 MG cells but was present only at low levels in U-251 MG cells. PMA (100 nm) treatment for 24 h increased cell proliferation by over 2-fold in the U-251 MG cells, whereas it decreased the mitogenic response in the U-1242 MG cells by over 90%. When PKC-eta was stably transfected into U-1242 MG cells, PMA increased cell proliferation by 2.2-fold, similar to the response of U-251 MG cells. The cell proliferation induced by PMA in both the U-251 MG and U-1242-PKC-eta cells was blocked by the PKC inhibitor bisindolylmaleimide (0.5 micrometer) and the MEK inhibitor, PD 98059 (50 micrometer). Transient transfection of wild type U-251 with PKC-eta antisense oligonucleotide (1 micrometer) also blocked the PMA-induced increase in [(3)H]thymidine incorporation. The data demonstrate that two glioblastoma lines, with functionally distinct proliferative responses to PMA, express different novel PKC isozymes and that the differential expression of PKC-eta plays a determining role in the different proliferative capacity.     

Structural peculiarities of Na+,K+-ATPase isozymes from the calf brain

Functionally active preparations of Na+,K(+)-ATPase isozymes from calf brain that contain catalytic subunits of three types (alpha 1, alpha 2, and alpha 3) were obtained using two approaches: a selective removal of contaminating proteins by the Jorgensen method and a selective solubilization of the enzyme with subsequent reconstitution of the membrane structure by the Esmann method. The ouabain inhibition constants were determined for the isozymes. The real isozyme composition of the Na+ pump from the grey matter containing glial cells and the brain stem containing neurons was determined. The plasma membranes of glial cells were shown to contain mainly Na+,K(+)-ATPase of the alpha 1 beta 1 type and minor amounts of isozymes of the alpha 2 beta 2 (beta 1) and the alpha 3 beta 1 (beta 2) type. The axolemma contains alpha 2 beta 1- and alpha 3 beta 1 isozymes. A carbohydrate analysis indicated that alpha 1 beta 1 enzyme preparations from the brain grey matter substantially differ from the renal enzymes of the same composition in the glycosylation of the beta 1 isoform. An enhanced sensitivity of the alpha 3 catalytic subunit of Na+,K(+)-ATPase from neurons to endogenous proteolysis was found. A point of specific proteolysis in the amino acid sequence PNDNR492 decreases Y493 was localized (residue numbering is that of the human alpha 3 subunit). This sequence corresponds to one of the regions of the greatest variability in alpha 1, alpha 2, alpha 3, and alpha 4-subunits, but at the same time, it is characteristic of the alpha 3 isoforms of various species. The presence of the beta 3 isoform of tubulin (cytoskeletal protein) was found for the first time in the high-molecular-mass Na+,K(+)-ATPase alpha 3 beta 1 isozyme complex isolated from the axolemma of brain stem neurons, and its binding to the alpha 3 catalytic subunit was shown.     

Pharmacokinetic studies of radiolabelled rat monoclonal antibodies recognising syngeneic sarcoma antigens

In the preceding paper it was suggested that the tumour localisation of 125I-labelled syngeneic rat monoclonal antibodies (mAbs) may be limited in immunocompetent hosts by the presence of competing endogenous serum antibodies. In syngeneic congenitally athymic (nu/nu) and cyclosporin-A-treated rats (both of which fail to mount immune responses to tumour antigens) increased uptake of mAbs in tumour tissue was obtained compared with that in immunocompetent animals. However, in the case of IgG2b and IgG1 mAbs, this appeared to be due primarily to enhanced "non-specific" localisation mediated by Fc binding, since it was abolished by the use of F(ab')2 fragments with two out of three mAbs tested. Normal tissue distribution was also influenced by host immune status: in nu/nu rats the uptake of IgG2b mAbs in the spleen was up to fivefold higher than that previously found in normal animals and the levels in liver were also increased. This effect was not seen in cyclosporin-A-treated hosts, suggesting that the reticuloendothelial system of congenitally athymic animals contains cells with enhanced IgG2b-FcR activity. This hypothesis was strengthened by the observation that splenic uptake was reduced by either the use of F(ab')2 fragments, or prior "blockade" of Fc receptors by "cold competition" with excess unlabelled IgG2b mAbs. This blockade could not be effected by mAbs of any other isotype or by IgG2b F(ab')2 fragments. The former manoeuvre resulted in higher tumour specificity ratios but usually at the expense of reduced levels of tumour associated radiolabelled mAb. The latter was found to increase "absolute" tumour localisation by up to 35%. In an attempt to characterise further and compare the Fc receptor activity of intratumour and intrasplenic host cells. The distribution of IgG2b mAbs was assayed in 3-week, 8-week and 12-week-old rats. We were able operationally to distinguish the activity of these two categories of cells, suggesting that they represent either different lineages or differentially activated subpopulations: the splenic IgG2b binding was fully expressed in weanling nu/nu rats whereas the FcR activity of cells infiltrating MC24 sarcoma was limited in 3-week-old compared with 8-12-week-old hosts. A further difference was apparent in the subclass "preference" of FcR binding: in immunodeprived rats both IgG1 and IgG2b mAbs were able to bind to tumour-infiltrating host cells, but uptake of IgG1 mAbs in the spleen was always low and not reduced further by the use of F(ab')2 fragments.(ABSTRACT TRUNCATED AT 400 WORDS)     

Epidermal growth factor receptor expression in human gliomas

Summary The expression of epidermal growth factor receptor (EGFR) was determined in cryosections of 42 human gliomas using biotinylated epidermal growth factor (B-EGF) and two monoclonal antibodies (mAb) against EGFR. All gliomas were found to express EGFR when examined with B-EGF, whereas 33 expressed EGFR when examined with the two mAbs. The highly malignant gliomas (glioblastomas and anaplastic astrocytomas) had a more heterogeneous staining pattern and a larger proportion of tumour cells staining strongly with B-EGF than did the low-grade gliomas (astrocytomas, oligodendrogliomas, mixed gliomas, and ependymomas). This indicates that high-grade gliomas contain more tumour cells rich in EGFR than do the low-grade gliomas. Reactive astrocytes, ependymal cells, and many types of nerve cells (cerebral cortical pyramidal cells, pyramidal and granular hippocampal cells, Purkinje cells, cerebellar granular cells and neurons in the molecular layer of the cerebellum) expressed EGFR, whereas small neurons and normal glial cells were not found to express EGFR.     

Evidence that type II 5'-deiodinase is not a selenoprotein.

Brain type II 5'-iodothyronine deiodinase and liver type I 5'-iodothyronine deiodinase activities are decreased in rats fed a Se(2+)-deficient diet suggesting that both enzymes are Se(2+)-dependent proteins. Since serum thyroxine (T4) concentrations are twice normal in the Se(2+)-deficient animals, it is unclear whether the Se2+ deficiency or the increased circulating T4 account for the decrease in the brain enzyme. In order to separate these two possibilities, the effects of Se2+ on 5'-deiodinase in glial cells (type II) and LLC-PK1 cells (type I) were examined. LLC-PK1 and glial cells were grown in serum-free defined medium containing 0, 1 pM, 10 nM, and 40 nM Se2+ for 3-5 days or in medium containing 75Se2+ for 24 h. Deiodinase isozymes were determined by measuring catalytic activity and by quantification of the BrAc[125I]T4 affinity-labeled substrate binding subunits. Se2+ deficiency was confirmed by measuring the activity of the selenoprotein, glutathione peroxidase. Se2+ caused a concentration-dependent increase in glutathione peroxidase activity in both cell types, as well as in the type I enzyme, but had no effect on the type II enzyme. LLC-PK1 cells contained multiple 75Se(2+)-labeled proteins including the 27-kDa substrate binding subunit of the type I 5'-deiodinase. Glial cells contained seven 75Se(2+)-labeled proteins ranging in size from 12 to 62 kDa, none of which corresponded to the type II substrate binding subunit. these data show that, unlike the type I enzyme, the type II enzyme does not contain a selenocysteine or selenomethionine, further emphasizing the differences between these two isozymes.     

The immunohistolocalization of carbonic anhydrase in rodent tissues.

Carbonic anhydrase has been localized with an immunoenzyme bridge technique in the following sites in paraffin sections of fixed rodent tissues: gastric parietal cells, the brush border of enterocytes in the small intestine, superficial nongoblet cells of the colon, selective segments of the nephron, glial cells, erythrocytes and adipose cells. Immunocytochemical localizations of carbonic anhydrase isozymes I and II in different histologic sites, by means of affinity column purified antibodies, agreed with the distribution of these enzymes in the various sites, as indicated by immunologic assays. The immunocytochemical results are compared with those reported for the cobalt-bicarbonate cytochemical method and with biochemical knowledge of the occurence of carbonic anhydrase.     

Diacylglycerol kinase in the central nervous system--molecular heterogeneity and gene expression.

Diacylglycerol (DAG) is one of the important second messengers, which serves as an activator of protein kinase C (PKC). DAG kinase (DGK) phosphorylates DAG to generate phosphatidic acid, thus DGK is considered to be a regulator of PKC activity through attenuation of DAG. Recent studies have revealed molecular structures of several DGK isozymes from mammalian species, and showed that most of the isozymes are expressed in the brain in various amounts. We have cloned four DGK isozyme cDNAs from rat brain library (DGK alpha, -beta, -gamma, and -zeta) (previously also designated DGK-I, -II, -III, and -IV, respectively) and examined their mRNA expressions in rat brain by in situ hybridization histochemistry. Interestingly, it is revealed that the mRNA for each isozyme is expressed in a distinct pattern in the brain; DGK alpha is expressed in oligodendrocytes, glial cells that form myelin; DGK beta in neurons of the caudate-putamen; DGK gamma predominantly in the cerebellar Purkinje cells; and DGK zeta in the cerebellar and cerebral cortices. Molecular diversity and distinct expression patterns of DGK isozymes suggest a physiological importance for the enzyme in brain function. Furthermore, functional implications of these DGK isozymes are briefly discussed.     

DISC ELECTROPHORETIC SEPARATION OF PROTEINS IN NEURONAL, GLIAL AND SUBCELLULAR FRACTIONS FROM CEREBRAL CORTEX

Abstract— Soluble proteins were studied in preparations from rabbit brain cortex enriched in neuronal or glial cells and in subcellular cortical fractions. Analytical polyacrylamide gels were used for acidic (pH 9-5) and basic (pH 4-3) proteins and qualitative and quantitative differences are described. The isozymes of lactic dehydrogenase, brain specific proteins and radioactive labelling patterns were used to characterize some soluble proteins.     

Y盒结合蛋白1单克隆抗体的研制、表位测定及其免疫学应用

建立稳定分泌抗人Y盒结合蛋白1单克隆抗体(anti-YB-1 mAb)的杂交瘤细胞株,鉴定其表位与免疫学应用。将重组YB-1蛋白免疫BALB/c小鼠,取脾细胞与Sp2/0骨髓瘤细胞融合。经ELISA法筛选鉴定、定株后采用腹水诱生法制备anti-YB-1 mAb;Protein G亲和层析法纯化mAb,ELISA法测定mAb效价、亚型及相对亲和力。采用抗原表位预测法鉴定anti-YB-1 mAb识别表位所在区域。Western blot和免疫组化鉴定mAb识别内源性YB-1的特异性。经筛选鉴定获得2株稳定分泌anti-YB-1 mAb的杂交瘤细胞(1-D9,3-E8);腹水抗体效价均≥1×10-6,亚型均为IgGl;1-D9和3-E8单抗识别表位分别位于(134-160aa)与(266-303aa)肽段。Western blot、免疫组化结果证实anti-YB-1 mAb能特异性识别内源性YB-1。该研究为YB-1免疫学定性、定量检测方法的建立、肿瘤靶向抗体治疗及进一步探讨YB-1的生物学功能奠定了基础。     

Immunodetection and localization of protein(s) related to retinal S-antigen (arrestin) in kidney.

S-antigen (arrestin) is a cytosolic protein which regulates phototransduction in retinal rods. A protein immunologically related to S-antigen was identified in fractions from soluble extract of bovine kidney enriched by gel filtration or by immunoaffinity chromatography using a polyclonal antibody to retinal S-antigen. On immunoblots, this protein was recognized by a panel of monoclonal antibodies (mAbs S2D2, S1A3 and S9E2) directed against different S-antigen epitopes and displayed the same apparent molecular mass (48 kDa) as retinal S-antigen. All three mAbs revealed a specific immunoreactivity by indirect immunocytochemical technique on rat kidney sections. The three mAbs recognized some but not all glomerular cells, identified as epithelial cells by immunoelectron microscopy using the mAb S9E2. Both mAbs S2D2 and S1A3 gave a diffuse cytoplasmic staining in all tubule cells. Proximal tubule cells exhibited a weak immunoreactivity, whereas distal and collecting tubule cells were strongly labeled. In contrast, the mAb S9E2 immunoreaction was restricted to a cell subpopulation from distal and collecting tubules corresponding to intercalated cells identified by immunoelectron microscopy. With the mAb S9E2, the labeling of proximal tubule cells was localized in the apical region of the cytoplasm. These results suggest that two or more 48-kDa proteins immunologically cross-reactive with retinal S-antigen are present in kidney. The observed pattern of distribution is in keeping with the hypothesis that such proteins could play a role in the regulation of G-protein-related receptors present in renal glomerulus and tubule epithelial cells.     

The Specificity of 130-kDa Leech Sensory Afferent Proteins Is Encoded by Their Carbohydrate Epitopes

From early development through adulthood in the leech, sensory afferents, glial cells, and connective tissue express different epitopes located on a group of 130-kDa glycoproteins. The sensory epitope [reactive with monoclonal antibody (mAb) Lan3-2] is shared by the peripheral sensory afferents of different sensory modalities. In contrast, three other immunocytochemically distinct epitopes (reactive with mAbs Laz2-369, Laz7-79, and Laz6-212) differentiate these sensory afferents according to their sensory modalities. The glial epitope (mAb Laz6-297) is expressed on all macroglial processes, and the connective tissue epitope (mAb Laz9-84) is located on connective tissue surrounding the CNS, as well as in the peripheral tissues. The hydrophilic-hydrophobic nature of the 130-kDa sensory afferent and glial proteins was determined by phase separation with Triton X-114 and hypoosmotic extraction. They behave as peripheral membrane proteins. Deglycosylation of 130-kDa glycoproteins with N-Glycanase or preincubation of their respective mAbs with alpha-methylmannoside showed that the sensory epitope contains mannose, whereas the modality epitopes are of an undefined carbohydrate character. Immunoprecipitation and a peptide mapping experiment confirmed the existence of four distinct sensory afferent epitopes. Previous studies provided evidence that the mannose-containing Lan3-2 epitope mediates normal sensory afferent growth in the synaptic neuropile. We, therefore, postulate that the carbohydrate epitopes on sensory afferent glycoproteins participate in synapse formation.     

Organ-specific distribution of isozymes of 5'-nucleotide phosphodiesterase in mouse

1. The distribution of isozymes of 5'-nucleotide phosphodiesterase (E.C.3.1.4.1) was examined in various organs of mouse, including liver, spleen, pancreas, heart, lung, kidney, brain and blood. 2. Five isozymes were identified and designated as isozymes I through V. 3. These isozymes are distributed unevenly with respect to the various organs and clear differences were observed in the patterns of distribution among the organs examined. 4. The level of these isozymes was compared in serum of neonate and adult mice, and a higher level of isozyme I and a lower level of isozyme IV were found in neonates compared to adults. 5. These results suggest that each isozyme has different functional roles in individual organs and that these isozymes may be involved in proliferation and development of cells.     

Generation of FX−/− and Gmds−/− CHOZN host cell lines for the production of afucosylated therapeutic antibodies

Antibody-dependent cellular cytotoxicity (ADCC) is the primary mechanism of actions for several marketed therapeutic antibodies (mAbs) and for many more in clinical trials. The ADCC efficacy is highly dependent on the ability of therapeutic mAbs to recruit effector cells such as n atural k iller cells, which induce the apoptosis of targeted cells. The recruitment of effector cells by mAbs is negatively affected by fucose modification of N-Glycans on the Fc; thus, utilization of afucosylated mAbs has been a trend for enhanced ADCC therapeutics. Most of afucosylated mAbs in clinical or commercial manufacturing were produced from Fut8^−/− Chinese hamster ovary cells (CHO) host cells, generally generating low yields compared to wildtype CHO host. This study details the generation and characterization of two engineered CHOZN® cell lines, in which the enzyme involved in guanosine diphosphate (GDP)-fucose synthesis, GDP mannose-4,6-dehydratase (Gmds) and GDP-L-fucose synthase (FX), was knocked out. The top host cell lines for each of the knockouts, FX−/− and Gmds−/−, were selected based on growth robustness, bulk MSX selection tolerance, production titer, fucosylation level, and cell stability. We tested the production of two proprietary IgG1 mAbs in the engineered host cells, and found that the titers were comparable to CHOZN® cells. The mAbs generated from either KO cell line exhibited loss of fucose modification, leading to significantly boosted FcγRIIIa binding and ADCC effects. Our data demonstrated that both FX−/− and Gmds−/− host cells could replace Fut8−/− CHO cells for clinical manufacturing of antibody therapeutics.     

Monoclonal antibodies to the human laminin receptor recognize structurally distinct sites

Laminin, a glycoprotein of basement membranes, binds to a specific receptor on the surface of neoplastic and non-neoplastic cells. The laminin receptor purified from human breast carcinoma plasma membranes was used as an antigen to generate two types of monoclonal antibodies (mAbs). Both types of mAbs bind to (a) the purified receptor coated on a solid phase; (b) isolated breast carcinoma plasma membranes; and (c) the surface of cultured MCF-7 human breast carcinoma cells by immunohistology. Using immunoblotting, both types of mAbs recognize a single 67 000 Dalton protein among all the proteins extracted from breast carcinoma plasma membranes. The mAbs differed in their ability to block binding of laminin to the plasma membrane receptor. Antibody LR1 inhibited virtually 100% of the specific binding of laminin to both the isolated human breast carcinoma plasma membranes or the living MCF-7 cells. In contrast, antibody LR2 had no effect on laminin binding under identical conditions. Thus, the two types of mAbs may recognize structurally distinct sites on the laminin receptor. These mAbs should be useful to dissect the biology and the molecular genetics of the laminin receptor.     

Identification of the hyaluronan receptor for endocytosis (HARE)

Rat liver sinusoidal endothelial cells (LECs) express two hyaluronan (HA) receptors, of 175 and 300 kDa, responsible for the endocytic clearance of HA. We have characterized eight monoclonal antibodies (mAbs) raised against the 175-kDa HA receptor partially purified from rat LECs. These mAbs also cross-react with the 300-kDa HA receptor. The 175-kDa HA receptor is a single protein, whereas the 300-kDa species contains three subunits, alpha, beta, and gamma at 260, 230, and 97 kDa, respectively (Zhou, B., Oka, J. A., and Weigel, P. H. (1999) J. Biol. Chem. 274, 33831-33834). The 97-kDa subunit was not recognized by any of the mAbs in Western blots. Based on their cross-reactivity with these mAbs, the 175-, 230-, and 260-kDa proteins appear to be related. Two of the mAbs inhibit (125)I-HA binding and endocytosis by LECs at 37 degrees C. All of these results confirm that the mAbs recognize the bone fide LEC HA receptor. Indirect immunofluoresence shows high protein expression in liver sinusoids, the venous sinuses of the red pulp in spleen, and the medullary sinuses of lymph nodes. Because the tissue distribution for this endocytic HA receptor is not unique to liver, we propose the name HARE (HA receptor for endocytosis).     

Glial fibrillary acidic protein and desmin in salivary neoplasms

The presence of intermediate filament proteins (IFP) in normal salivary gland tissue and in a number of salivary gland neoplasms has been investigated by immunohistochemical techniques on frozen sections. Cytokeratins (CKs) were seen in almost all normal epithelial cells. In the parotid gland and in palatal gland tissue, a co-expression of cytokeratin and glial fibrillary acidic protein (GFAP) was seen in some myoepithelial cells, but this was not apparent in the submandibular gland. In some pleomorphic adenomas, carcinomas in pleomorphic adenomas, one mucoepidermoid carcinoma, one mucus-producing adenopapillary carcinoma and one adenoid cystic carcinoma, cells expressing three different IFP classes were found (CKs, vimentin, GFAP). These cells were most often situated peripherally in the tumour cords or ducts. The cytokeratin pattern in these cells, as revealed by mAbs PKK1-3, was similar to that in normal myoepithelial cells. Furthermore, reactivity for a fourth class of IFP, desmin, could be seen in this cell type in two carcinomas in pleomorphic adenomas, and also in a few cells in a pleomorphic adenoma and an adenoid cystic carcinoma. Thus the pattern of IFP expression in salivary gland neoplasms, is very complex, and cannot always be related to the normal tissue.     

PYRUVATE KINASE ISOZYMES IN NEURONS, GLIA, NEUROBLASTOMA, AND GLIOBLASTOMA

Abstract– The distribution of pyruvate kinase isozymes (EC 2.7.1.40) was examined in cells and tissues from the central and peripheral nervous system of the rat. Most tissues contain significant quantities of both the K₄ (fetal type) and M₄ (skeletal muscle type) isozymes plus tetrameric hybrids comprised of various combination of the type M and type K subunits. Retina, for example, contains a five-mem-bered hybrid set weighted toward K₄, while sciatic nerve and spinal cord have patterns very similar to that of adult brain, consisting predominantly of M₄ with small amounts of K₄ and K-M hybrids. This adult pattern is achieved by a gradual shift from a hybrid set dominated by K₄ in fetal life, to the pattern at birth at which time the two most prominent bands were M₄ and K₂M₂, and finally to the adult pattern by about 28 days after birth. Neurons and glial cells were isolated from rat and mouse brains at the various developmental levels. The pyruvate kinase isozyme patterns in the two cell types were similar to each other and to the patterns seen in whole brain homogenates at all ages, indicating similar rates of isozymic maturation in the two cell types. The correlation of maturation with pyruvate kinase isozyme patterns was further tested in cultures of malignant cell lines. A K-M hybrid set, weighted toward K₄, was seen in two clonal lines of mouse neuroblastoma under normal culture conditions. However, lowering the serum concentration in the culture medium or adding bromodeoxyuridine caused a shift in the patterns toward type M as the cells differentiated, mimicking in part the in vivo maturation of normal cells. On the other hand, a rapidly growing and poorly differentiated line of rat glioblastoma had only K₄ under all conditions examined.     

Anatomical arrangement of hypercapnia-activated cells in the superficial ventral medulla of rats.

The anatomical structure of central respiratory chemoreceptors in the superficial ventral medulla of rats was studied by using hypercapnia-induced c-fos labeling to identify cells directly stimulated by extracellular pH or PCO(2). The distribution of c-fos-positive cells was found to be predominantly perivascular to surface vessels. In the superficial ventral medullary midline, parapyramidal, and ventrolateral regions where c-fos-positive cells were concentrated, we found a common, characteristic, anatomical architecture. The medullary surface showed an indentation covered by a surface vessel, and the marginal glial layer was thickened. We classified c-fos-positive cells into two types. One (type I cell) was small, was located inside the marginal glial layer and close to the medullary surface, and surrounded fine vessels. The other (type II cell) was large and located dorsal to the marginal glial layer. c-fos Expression under synaptic blockade suggested that type I cells are intrinsically chemosensitive. The chemosensitivity of surface cells (possible type I cells) surrounding vessels was confirmed electrophysiologically in slice preparations. We suggest that this characteristic anatomical structure may be the central chemoreceptor.     

Glial fibrillary acidic protein and desmin in salivary neoplasms. Expression of four different types of intermediate filament proteins within the same cell type

The presence of intermediate filament proteins (IFP) in normal salivary gland tissue and investigated by immunohistochemical techniques on frozen sections. Cytokeratins (CKs) were seen in almost all normal epithelial cells. In the parotid gland and in palatal gland tissue, a co-expression of cytokeratin and glial fibrillary acidic protein (GFAP) was seen in some myoepithelial cells, but this was not apparent in the submandibular gland. In some pleomorphic adenomas, carcinomas in pleomorphic adenomas, one mucoepidermoid carcinoma, one mucus-producing adenopapillary carcinoma and one adenoid cystic carcinoma, cells expressing three different IFP classes were found (CKs, vimentin, GFAP). These cells were most often situated peripherally in the tumour cords or ducts. The cytokeratin pattern in these cells, as revealed by mAbs PKK1-3, was similar to that in normal myoepithelial cells. Furthermore, reactivity for a fourth class of IFP, desmin, could be seen in this cell type in two carcinomas in pleomorphic adenomas, and also in a few cells in a pleomorphic adenoma and an adenoid cystic carcinoma. Thus the pattern of IFP expression in salivary gland neoplasms, is very complex, and cannot always be related to the normal tissue.