Journal of Cell Science 102, 329-340 (1992) 329 Printed in Great Britain © The Company of Biologists Limited 1992

Cytokeratins and retinal epithelial cell behaviour

HELEN L. ROBEY*, PAUL S. HTSCOTT and IAN GRIERSON

Department of Clinical Science, Institute of Ophthalmology, 17/25 Cayton St, London EC1V 9AT, UK *Author for correspondence

Summary

The expression of 18 and 19 by human Micro-chemotaxis chambers were used to study migrat- retinal pigment epithelial cells (HRPE) has been sus- ing cells and immunohistochemical staining for cyto- pected of being associated with HRPE proliferation. We 18 and 19 revealed that actively migrating cells have investigated the involvement of these always expressed these two cytokeratins, whereas subtypes in the proliferative and migratory behaviour of stationary cells did not label for these cytokeratin cultured HRPE. Cell proliferation markers (bromo- subtypes. It was apparent that cytokeratins 18 and 19 deoxyuridine and proliferating cell nuclear antigen) and were not markers of proliferation, but were involved in the cytokeratins were identified using immunohisto- the mobility of HRPE in vitro. Cytokeratins 18 and 19 chemical techniques. In vitro, cytokeratins 18 and 19, as may be useful indicators of simple epithelial cell detected by the monoclonal antibodies RGE 53 and migration in tissues. K4.62, were expressed in a subset of HRPE and this subset was significantly less likely to be proliferating. Key words: cytokeratin, migration, .

Introduction that forms is rich in mobile proliferating RPE (Grierson et al. 1987) and many of these cells stain prominently Retinal pigment epithelial cells (RPE) of mammals and with antibodies to cytokeratin 18 (McKechnie et al. amphibians express a range of cytokeratins in vivo and 1988; Hiscott et al. 1991). It would appear that in vitro (Hitchins and Grierson, 1988; McKechnie et al. immunohistopathological observations are consistent 1988; Kaspar et al. 1988; Owaribe et al. 1988) that do with the concept that the expression of acidic low not seem to be present in the chick (Docherty et al. molecular mass cytokeratins is a feature of subsets of 1984). The cytokeratins, of the human RPE (HRPE) at RPE involved in altered behavioral activity. We least, are generally a selection of the intermediate and undertook the present study to obtain further evidence low molecular mass forms and, on the basis of the concerning proliferative and migratory activity of classification scheme introduced by Moll (Moll et al. human RPE and the expression of cytokeratins 18 and 1982), cytokeratins 5,6,7,8,14,15,16,17,18 and 19 are 19, employing the two monoclonal antibodies RGE 53 either suspected or have been shown to be present. and K4.62. Cytokeratin intermediate filaments may also be co- expressed with in HRPE (McKechnie et al. 1988). Materials and methods There seem to be differences between immunohisto- chemical staining patterns in the intact tissue monolayer Tissue culture and those in culture (McKechnie et al. 1988). Cytokera- Eyes up to but not exceeding 48 hours post-mortem were tin 18 is faintly positive if present at all and cytokeratin obtained from the local Eye Bank. Human RPE were isolated 19 is absent from the intact monolayer in situ, but in following the procedures outlined by Edwards (1981) and tissue culture subpopulations of RPE are identified that modified by Boulton et al. (1982). Briefly, the eyes were exhibit intense immunoreactivity for both cytokeratins soaked for 10 minutes in phosphate buffered saline (PBS) containing 100 units of penicillin and streptomycin. The while adjacent cells can be completely negative. It is anterior segment was removed and discarded and thereafter suspected that these cytokeratins may be markers for the vitreous humour was dislodged by gently shaking the proliferating RPE. RPE normally do not proliferate in posterior eye cup upside down. The neural retina was teased the adult eye where they function as sedentary away from the RPE and the exposed RPE was washed three phagocytes but they can be stimulated to do so in times in PBS with 0.02% EDTA to remove adherent pathological conditions such as complex retinal detach- photoreceptor debris. ment. In complex detachment, the retinal scar tissue The posterior eye cup was divided into three segments and 330 H. L. Robey and others areas of the RPE monolayer were isolated by brass cloning nuclear incorporated bromodeoxyuridine or of proliferating rings (Boulton et al. 1982). The wells formed by the cloning cell nuclear antigen/cyclin. BrdU is a thymidine analogue that rings were filled with a mixture of 0.25% trypsin, 0.02% is incorporated at S-phase of the cell cycle, whereas PCNA EDTA and PBS, and incubated at 37°C for not more than 1.5 peaks in concentration during S-phase but is a more general hours. Dissociated RPE cells were aspirated from Bruch's indicator of cell cycle activity (Kurki et al. 1988). membrane and seeded into 25 cm3 tissue culture flasks (Gibco For BrdU labelling, the culture medium was exchanged for Europe Ltd, Scotland). The RPE were fed with Ham's F-10 fresh medium containing 1:1,000 BrdU (Amersham UK). The culture medium, 20% foetal calf serum, 0.003 g/ml glucose, cells were then incubated at 37°C in 5% CO2 for a range of 2.5 mg/ml amphotericin B and 100 units/ml of penicillin and time periods between 15 minutes and 6 hours to identify the streptomycin (Gibco Europe Ltd, Scotland). The cultures optimum labelling period. After incubation, the cells were were maintained at 37°C in an atmosphere of 5% CO2 and air. washed in PBS, fixed in methanol (—20°C) for 5 minutes, Primary cultures were fed once a week until cellular acetone (—20°C) for 2 minutes and air dried. The S-phase outgrowth was stabilized, at which point feeding was nuclei having incorporated BrdU were detected with an anti- increased to twice a week. The RPE reached confluence BrdU monoclonal antibody (Amersham, UK). The reaction within 2-3 weeks, were split at a ratio of 1:3 and subcultures product was dark brown as revealed using the 'Biotin-Avidin' between 5th and 7th passage were used in the present study. peroxidase procedure (Giorno, 1984) and diamine benzidine Cells were either grown on 25 cm3 flasks or in 8-chamber (DAB) substrate (BDH, UK). tissue culture slides (LabTek, NUNC Inc, USA). Human RPE cells prepared and fixed in the same way as with BrdU were treated with an anti-PCNA monoclonal Cytokeratin staining of slides antibody (Boehringer Mannheim Biochemica, Germany), Human RPE grown on LabTek slides were fixed in pre-cooled again the 'Biotin-Avidin' procedure was adopted and the (—20°C) methanol (5 minutes) and acetone (2 minutes), then labelling was visualised as either a brown (DAB) or a red (3- incubated with monoclonal antibodies: anti-cytokeratin 8.13 amino-9-ethylcarbazole, AEC) reaction product. (clone K8.13, 1CN Biomedicals Ltd., UK), which labels The cells and their stained or unstained nuclei were cytokeratins 1,5,6,7,8,10,11 and 18 (Gigi et al. 1982); visualised using DIC optics (Polyvar, Reichert-Jung, Aus- RGE 53 (clone RGE 53, Bio-nuclear Services Ltd., UK), tria). Cells in the proliferation experiments were counted in shown to be exclusive to cytokeratin 18 (Raemaekers et al. 15 random fields (x400) per well. 1983); and anti-cytokeratin 4.62 (clone K4.62, ICN Biomedi- A double immunoperoxidase staining technique was used cals Ltd., UK), a cytokeratin 19 marker. The monoclonal to identify both cell cycle activity and expression of either antibodies were used at a 1:20 dilution, for broad-spectrum cytokeratin 18 or 19. HRPE were grown on LabTeks and cytokeratin markers and at a 1:10 dilution for the single treated as described above for labelling with either BrdU or keratin markers. Following washing in PBS, the cells were PCNA. However, nuclei were visualised using DAB. Then exposed to the particular antibody for 45 minutes. The cells either cytokeratin 18 or 19 was identified as described earlier were then washed in PBS and immersed in either anti-mouse and visualised using AEC. peroxidase conjugate (1:50, DAKO) or anti-mouse FITC conjugate (1:40, SIGMA) for 45 minutes. The peroxidase was Migration assay developed in AEC, giving a red reaction product to the RPE migration was conducted in 48-well micro-chemotaxis cytokeratins. Slides were mounted in glycerol (peroxidase) or chambers (Neuro Probe, Cabin John, USA). Bovine fibro- Fluorostab (fluorescence; Eurodiagnostics BV Holland) and nectin (SIGMA) was used as the chemoattractant because viewed by bright-field and DIC microscopy or observed under previous studies had shown that it was a potent inducer of epi-illumination optics (Polyvar, Reichert-Jung). human RPE migratory activity (Campochiaro et al. 1984). Staining for both cytokeratins 18 and 19 was done using a Bovine fibronectin was checked for purity by gel electrophor- cocktail of the two monoclonals (1:10). Combined cytokeratin esis. Contaminants were at the limit of resolution with (18 or 19) and anti-BrdU (bromodeoxyuridine) or anti-PCNA Coomassie blue staining and one ran concurrently with (proliferating cell nuclear antigen) staining was done on albumin. Since the contaminants were at such a low level, appropriate slides and counted in the manner described further purification was not needed. The attractant was made above. up to the appropriate concentration (see Results) in Ham's F- 10 medium and pipetted in 25 ^1 samples into the lower wells. Cytokeratin staining of tissue culture flasks A polycarbonate membrane with 10 ftm diameter pores Human RPE grown on tissue culture flasks were fixed for 10 (Nucleopore, Pleasanton, USA) and a gasket were placed on minutes with pre-cooled methanol (—20°C) and then stained top and the upper well section of the chamber and secured in for cytokeratin 18 as described above. The RPE were counter- place. Thus the permeable membrane, which had previously stained with haematoxylin for 15 minutes, washed in tap been coated with 300 bloom type 1 porcine gelatin (SIGMA, water, mounted in glycerol and viewed by bright-field UK), was sandwiched between the lower and upper wells of microscopy. the system. At confluence RPE cells were removed from the culture Vimentin staining of HRPE flask with trypsin (0.25%) and EDTA (0.02%) in Ham's F-10. Human RPE grown on LabTek slides were fixed in methanol The treatment was terminated by a 50:50 (v/v) mixture of and acetone as described above (see Cytokeratin staining of foetal calf serum and F-10 and then the cells were spun down slides). Using the same technique used to stain for cytokera- and resuspended in serum-free medium. The washing pro- tins, cells were stained for vimentin using the monoclonal cedure was repeated several times until finally the RPE were VIM 13.2 (1:200 SIGMA). counted in a Coulter counter (Coulter Electronics, UK). A volume of 50 /J of serum-free F-10 containing 80,000 RPE was Proliferation evaluations pipetted into the upper wells of the chamber. The whole Human RPE were seeded onto LabTek slides at a concen- apparatus was subsequently incubated in 5% CO2 at 37°C for tration of 7.5 xlO3 cells/cm2. Immunohistochemical localiz- 5 hours, following which it was dismantled and the membrane ation of cell cycle activity involved the identification of removed. The membranes were fixed in accordance with their Cytokeratins and epithelial cell behaviour 331 subsequent treatments as outlined below. For routine count- cells stained with this antibody, this confirmed the ing, the membranes were placed in ethanol for 15 seconds and purity of the cultures under investigation. Possible then air dried. Membranes were stained with haematoxylin contaminants include Muller glia, astrocytes, vascular for 30 minutes, washed in cold water and mounted on glass endothelium, pericytes and fibroblasts, none of which slides. Attached cells on the upper surface and migrated cells on the lower surface of the membrane were counted in 20 exhibits positive staining for cytokeratin markers. The high-power fields (x 1,000) per well using conventional bright- wide-spectrum antibody (K8.13) delineated a fine field microscopy (Zeiss, Germany). filamentous pattern in the cytoplasm of well-spread cells, with more-pronounced staining in the perikaryon Scanning electron microscopy (SEM) of migration than in the periphery of the cells (Fig. 1C). membranes Within the same culture not all HRPE labelled with Membranes at the end of a migration run were washed in PBS the cytokeratin 18 (K18) and cytokeratin 19 (K19) in the usual manner, then the cells were fixed in 2.5% antibodies. Indeed, at confluence, the K18 and K19 (as glutaraldehyde in Sorenson's phosphate buffer for 4 hours. recognised by RGE 53 and K4.62) populations rep- The specimens were postfixed in 1% buffered osmium resented 47% and 15% of the total number of cells tetroxide, dehydrated through graded ethanol, critical point present, respectively, and double labelling for K18 and dried (Polaron, UK) and sputter coated with gold (Polaron, K19 did not increase the labelling index above 47%, UK). After drying, the membranes were orientated on stubs indicating that the K19 population was not a separate so that both the upper and lower surfaces could be examined on the same specimen. Examination was in a Hitachi S520 group of cells but a subset of the K18 population (Fig. scanning electron microscope to identify the phenotypes of 2). The K18 and K19 staining pattern in the HRPE were motile cells. distinguishable from that associated with the more broad spectrum cytokeratin antibody by the presence of Cytokeratin staining of migration membranes distinct intensely stained filaments curled around the For routine evaluation, cytokeratin 18 and 19 intermediate nucleus with finer filaments extending into the peri- filaments were visualised using indirect immunofluorescence karyon. In recently settled cells that were still in the with the monoclonal antibodies RGE 53 and K4.62. The process of spreading (up to 4 h post-seeding), the K18 methanol-fixed membrane was rehydrated in PBS for 5 and K19 pattern was one of an intense-staining non- minutes, immersed in 5% normal goat serum in PBS for 10 filamentous aggregate in the cytoplasm, usually adjac- minutes, floated onto primary antibody and 1 ml of primary ent to the nucleus (Fig. ID). antibody was added to the upper surface. After 30 minutes the membrane was turned over and fresh antibody was applied for a further 30 minutes. The membrane was rinsed in PBS and Effects of post-seeding time and seeding density on the then incubated with the secondary antibody, goat anti-mouse K18 population FITC (1:40, SIGMA), for 30 minutes on each side. Finally the The proportion of K18 positive (K18+ve) cells was membrane was washed, mounted under a coverslip with noted to vary with post-seeding time and seeding Fluorostab and observed under epi-illumination optics (Poly- density. At 4 hours, the K18+ve cells represented over var, Reichert-Jung). 50% of the total. Thereafter, there was a steady decrease in the percentage of K18+ve cells up to post- staining of migration membranes confluence at 72 h by which time the population had The actin-staining pattern of HRPE on migration membranes decreased to 10% or less. From the raw data it was clear was examined in order to identify cells with patterns that the decrease in the percentage of K18+ve cells associated with motility. Actin was identified using an anti- actin monoclonal antibody (clone C4, ICN Biomedicals Ltd., resulted from the K18+ve cell population remaining UK) at a 1:100 dilution. Cells on the membrane were stained more or less stable while the overall population was for actin in a similar manner to that outlined above. proliferating (Fig. 3). These findingswer e true for both flask and LabTek cultures. Likewise, at the higher Statistical analysis seeding densities needed for subsequent migration experiments (5 x 105 cells/cm2 for membranes com- The data were evaluated using the paired Mest, Mann 3 2 Whitney U-test and chi-squared analysis from the Microstat- pared to 7.5 x 10 cells/cm for flasks), the incidence of 11 statistical software system (Ecosoft, Indianapolis, USA). K18+ve cells at 2 hours was 24% and at 5 hours was only 11%.

Results Immunoreactivity for cytokeratins 18 and 19 is not associated with RPE proliferation General cytokeratin staining characteristics in cultured The apparent stability of the K18+ve population in HRPE culture with time suggested that K18^was not, after all, In general, the cytokeratin staining characteristics were associated with HRPE proliferation and this was similar in HRPE grown on glass and plastic at further supported by proliferation experiments. equivalent culture times and seeding densities. All cells Between 25 and 35% of the cells labelled with the S stained positively for vimentin with monoclonal VTM phase marker BrdU following an optimal period of 13.2 (Fig. 1A). exposure of 120 minutes and approximately 60% The fifth to seventh passage human RPE exhibited a expressed PCNA (Fig. 4). Labelling preconfluent positive staining response for the monoclonal antibody cultures on LabTek slides with BrdU and K18 showed K8.13 (labels 1,5,6,7,8,10,11 and 18) (Fig. IB). Since all that the BrdU+ve/K18+ve subset was extremely sparse 332 H. L. Robey and others

Fig. 1. Immunofluorescence (A) and immunoperoxidase (B, C and D) staining of cytokeratin intermediate filaments in cultured HRPE. (A) All cells stained positively for vimentin intermediate filaments. (B) All cells stained with the wide- spectrum cytokeratin monoclonal antibody K8.13. (C) Cytokeratin staining with K8.13 reveals a fine filamentous network within the perikaryon that extends to the periphery of the cell. (D) staining appeared as an aggregate, rather than filamentous, in recently settled cells in the process of spreading. Bars: A,D, 30 ,um; B, 50 /xm; C, 25 fsm. and by far the largest group was BrdU—ve/K18-ve, solution of FCS with over 5% of the population which would be expected from the previous population migrating through the permeable membrane over the analysis. The incidence of BrdU+ve cells in the period of the assay. As a consequence, fibronectin at 10 K18+ve group was less than 11% whereas the jUg/ml was used as the positive chemoattractive stimulus BrdU+ve cells in the K18-ve population was almost in all subsequent experiments. 35%. Chi-squared analysis showed the difference to be On the upper 'settlement' side of the membrane the highly significant (P<0.001). Similarly, with BrdU and cells varied from a rounded to a flattened appearance as K19 labelling a similar trend was evident. The seen by SEM (Fig. 6A). The flattened cells had obvious BrdU+ve cells were over 14% in the K19+ve group, surface microvilli and well-developed cytoplasmic pro- compared to almost 33% in the K19-ve group and cesses (Fig. 6B). Cells were seen with processes again the difference was highly significant (P<0.001) extending into pores (Fig. 6C) and in more advanced (Fig. 5). PCNA also had a much lower K18 and K19 stages of migration only a rounded knot of cytoplasm labelling index in the +ve compared to the —ve cells, remained on the upper surface of the membrane (Fig. there being three to four times as many K18 and 19+ve 6D). Examination of the pores on the lower 'migration' cells in the PCNA—ve than the PCNA+ve group. side of the membrane showed some that contained the cytoplasm of RPE in transit through the membrane. The morphology of migrating HRPE cells The earliest stage was a small bulbous projection of Dose response curves (5 /xg/ml to 50 ^g/ml) showed that cytoplasm, which made contact with the membrane human RPE cells migrated optimally to a 10 /xg/ml surface via numerous microspikes (Fig. 7A). Other cells concentration of fibronectin, and at this level fibronec- exhibited more advanced extension out of the pore tin was almost four times more effective than a 1% (Fig. 7B), until eventually the perikaryon was evident Cytokeratins and epithelial cell behaviour 333

40-1

m so-

+ 20-

m

Keratin Keratin Keratin 18+ve 19+ve 18 & 19+ve 1 Keratin Keratin Keratin Keratin 19+ve Fig. 2. A comparison of the keratin 18 and keratin 19 18+ve 18-ve 19-ve populations in cultured HRPE. Cultures of HRPE were Fig. 5. The expression of keratins 18 and 19 is not related stained for keratin 18 alone, keratin 19 alone and both to proliferation in HRPE. Using indirect double keratin 18 and 19. Cells that stained were counted and immunoperoxidase staining the subsets of keratin 18+ve represented as a percentage of the total cells in each and 19+ve cells were identified as well as proliferating culture. Cells expressing keratin 19 were found to be a cells. The incidence of BrdU+ve cells in the keratin 18+ve subset of the keratin 18 cell population. subset was lower than in the keratin 18—ve cells. Similarly, the keratin 19+ve subset had far less BrdU+ve cells than the keratin 19—ve cells. 5,000 Total no cells No of cells taxtttln 18+ve 4,000 - (Fig. 7C). The cells coming out and just leaving the pores had an elongated kite shape and were smooth

d 2,000- The RPE that had completed the migration through z the pores were extremely flattened so that their nuclei were prominent. Processes and surface microvilli were 1,000 - more abundant than on those cells that were still obviously in transit. The cells were larger, uniform in 0 12 24 72 shape and more epithelioid than those seen on the Time (h) upper surface (Fig. 7A and D). Actin in migrating HRPE Fig. 3. The expression of keratin 18 in cultured HRPE with Actin staining of cells on the settlement side of the post-seeding time. Cells were seeded and cultured up to 72 hours. During this time the number of cells increased, but membrane revealed a variable pattern. Rounded and the keratin 18 subset remained constant throughout. poorly spread cells had diffuse cytoplasmic staining as did those in passage through pores. Spread cells had either a reticular or a filamentous pattern but well- developed stress fibres were rare (Fig. 8A). Actin BrdU+ve PCNA+ve staining of cells on the migrated side showed up clear differences. The diffuse staining of cytoplasm in pores was again evident. A reticular pattern was present in the elongated kite-shaped RPE (Fig. 8B) but the most obvious feature was abundant stress fibres in the thin extended epithelioid phenotype (Fig. 8C). Cytokeratins 18 and 19 are associated with migration in HRPE

Brdll PCNA Subsets of RPE stained up for K18 and K19 on both sides of the membrane and, in addition, +ve cells were identified in transit through the pores. It was apparent Fig. 4. Identification of proliferating HRPE in pre- even from cursory examination that there were more confluent cultures using BrdU and PCNA. Cells were +ve cells on the migrated than on the settlement side of labelled with two equally good markers of proliferation. the membrane. Subsequent counts, totalling 28,976 BrdU, an analogue of thymidine, specifically labels cells in the S-phase of the cell cycle, whereas PCNA, shown to be cells, showed that just over 8% of these were K18+ve identical to cyclin and the auxiliary of DNA on the settlement side whereas 45% were +ve on the polymerase-p, accumulates during G1; peaks in S-phase migrated side (total of 1144 migrated cells counted). and decreases during G2 and M phase. Therefore more Only 2.5%, from a total of 468,215 cells, were K19+ve cells in the cell cycle are labelled with PCNA than with on the settled side while over 21% were positive on the BrdU. migrated side (total of 13% migrated cells counted; 334 H. L. Robey and others

Fig. 6. SEM of migrating HRPE on the upper surface of a migration membrane. (A) Cells on the upper surface of the membrane had phenotypes varying from rounded (short arrow) to flattened (long arrow). (B) The flattened cells had well- developed cytoplasmic processes and surface microvilli. (C) Some cells appeared to be in the initial stages of migration with cell processes extending into a pore. (D) Other cells were in a more advanced stage of migration where only a remnant of cytoplasm remained on the upper surface of the membrane. Bars: A, 10 /zm; B, 5 ^m; C, 2 fm\; D, 1 fxm. Cytokeratins and epithelial cell behaviour 335

Fig. 7. SEM of HRPE migrating onto the lower side of the membrane. (A) Small bulbous projections of cytoplasm protruding from a pore indicated a cell emerging onto the lower side of the membrane (arrow). (B) Other cells showed more advanced stages of emergence with spread cytoplasm extending onto the lower surface. (C) In the final stages of migration the perikaryon was evident on the lower surface (arrow). (D) Cells that had completed migration were flattened epithelioid cells with abundant surface microvilli (see also foreground of A). Bars: A,D, 2 pm; B,C, 1 ^m. 336 H. L. Robey and others

Fig. 8. HRPE on the upper and lower side of a membrane were stained for actin and visualised with immunofluorescence. (A) On the upper side of the membrane rounded poorly spread cells had a diffuse actin staining pattern. Whereas, spread cells had a filamentous staining pattern. (B) The filamentous staining pattern was also seen in cells on the lower side of the membrane. (C) The flattered epithelioid cells had abundant actin stress cables radiating throughout the cytoplasm. Bars: A,B, 50 fan; C, 20 /jm.

Fig. 9). The differences were significant in both cases using chi-squared analysis and the Mann-Whitney U- test (P<0.001). (The value of 8.2% K18+ve cells on the settlement side of the membrane may appear to be so considerably lower than that of 47% in a confluent • Keratin 18+ve culture. Indeed, it could be predicted that the fre- J8 4O B Keratin 19+ve quency of K18+ve cells should be over 40% not 8% 8 (Fig. 3). However, this discrepancy is due to the + differences in keratin expression during time in culture and seeding densities (see above).) S52O- A field-by-field examination of both sides of the membranes was conducted under the x 100 oil-immer- 8*10- sion objective of the microscope. Initially we used fluorescence; however, only the keratin +ve cells were Settled Migrated Settled Migrated visualised. So subsequently, we concentrated on the membranes that had been stained using the peroxidase method, because with haematoxylin counter-staining Fig. 9. The expression of keratins 18 and 19 in migrating non-decorated (keratin —ve) cells could be seen clearly. HRPE. Keratins 18 and 19 were visualised in cells on the On the upper surface of the membrane, the rounded upper and lower sides of the migration membrane using the immunofluorescence staining technique. The subset of cells were K18 and K19 negative but some of the keratin positive cells on either the upper or lower side of partially spread cells had foci of positive staining the membrane was expressed as a percentage of the total adjacent to the nucleus (though they were relatively number of cells on that side of the membrane. The keratin infrequent). On the other hand, the flattened RPE that 18 and 19 subset of cells was larger on the migrated were positive for keratin exhibited a fibriUar staining (lower) side of the membrane than on the settled (upper) pattern and, unlike their corresponding keratin nega- side of the membrane.

Fig. 10. Immunoperoxidase staining of keratin 18 in HRPE on the upper (A) and lower (B) sides of a migration membrane. Cells were counter-stained with haematoxylin. (A) Upper surface: rounded cells were keratin 18 negative (short white arrow) and spreading cells had aggregates of keratin 18 positive staining adjacent to the nucleus (long white arrow). Cells in the initial stages of migration were keratin 18 positive and had a fibrillar staining pattern. These cells often had a cell process extending into a pore (short black arrow). Spread stationary cells were keratin 18 negative (long black arrow). (B) Lower surface: keratin 18 positive cells were small polarised cells often with some of their cytoplasm still within a pore (short white arrow). Keratin negative cells were large flat non-polarised epithelioid cells (long white arrow). Keratin 18 positive cytoplasm could be seen extending out of a pores and spreading onto the lower surface (short black arrow) (comparable to Fig. 7B). This spreading cytoplasm comes from keratin 18 positive cells on the upper surface, which, because the micrograph is focused on cells on the lower side of the membrane, appears as a diffuse red shadow (long black arrow). N.B. the density of cells on the lower side of the membrane is the same as that in Fig. 8B. Cytokeratins and epithelial cell behaviour 337

Fig. 11. (A and B) A migrating HRPE stained for keratin 18 and captured on film at two levels of focus. (A) On the lower side of the membrane a filamentous array of keratin filaments has emerged from the pore (arrow) and spread onto the lower surface (comparable to Fig. 7B and C). (B) While on the upper surface the cytoplasm within the pore, that remaining on the upper surface and a cell process stain diffusely for keratin 18 (comparable to Fig. 6B). tive cells, they all had cell processes extending into an McKechnie et al. (1988) identified a subset of cultured adjacent pore (Fig. 10A). The large flattened non- human RPE that expressed cytokeratin 18 with the polarised epithelioid cells on the migration side of the monoclonal antibody RGE 53 and suggested that the membrane were negative in the main. The vast majority expression may be cell cycle-related. Indeed the of the positive cells were smaller bipolar cells, often possibility arose that K18 could be a specific marker for with cytoplasm still within a pore (Fig. 10B). the detection of proliferating RPE in culture and in All cells in passage through the pores were K18+ve pathological tissues such as retinal scars. RPE play a and K19+ve and no exceptions were found in over 300 key role in retinal scar formation but they are difficult to wells examined. The pattern was particularly clear by identify on phenotype alone (Kampik et al. 1981). fluorescence microscopy and when the cells were However, our investigations, far from establishing an predominantly on the settlement side the cytoplasm on association between K18 expression and replication, that side had a fibrillar pattern whereas the staining of demonstrated that K18 (and for that matter K19) the cytoplasmic processes in and through the pores was positive HRPE were less likely to be replicating than diffuse (Fig. 11A). With cells that had a major K18 negative HRPE. Indeed, there are implications component of their cytoplasm on the migration surface from recent literature that K19 is not functionally of the membrane (see Figs 6C, 7B and C) the staining involved in cultured keratinocyte cell proliferation pattern was reversed. The cytoplasm spread on the (Lindberg and Rheinwald, 1991). membrane had a fibrillar or filamentous pattern while From the present study it appeared that HRPE the remnant in the pore and on the settlement side was expressing K19 were a subset of the K18 population and diffuse (Fig. 11B). the incidence of K18 steadily decreased in the culture Using BrdU-labelled cell proliferation on the upper flask as HRPE numbers increased from sparse to and lower sides of the migration membrane was confluent. This would seem on the face of it to have investigated. Probably because of the short incubation been an anomalous finding, given that their expression period of the migration experiments (i.e. 4 hours) no was not proliferation related. On the other hand, as positively BrdU-labelled nuclei were found. cells approach confluence not only does their division rate decrease but their mobility is impaired. Mobility is also compromised at high seeding density and in this Discussion circumstance K18 positive cells made up a significantly lower percentage of the population than in cultures For over a decade it has been known that the seeded at lower density when comparable time periods cytokeratin organisation in epithelial cells alters with were evaluated. general changes in cell behaviour but is particularly Much stronger evidence was provided by our che- evident during cell division (Aubin et al. 1980; Horwitz moattraction assays, which clearly showed that a five et al. 1981; Lane et al. 1982; Franke et al. 1982, 1983). times greater percentage of K18 positive HRPE were The suggestion has been put forward that cytokeratins on the lower, migrated side of the permeable mem- may be modulated to permit radical shape change branes than on the upper (settlement or non-migrated) during mitosis (Horwitz et al. 1981; Lane et al. 1982). side. Furthermore, there was an even greater (10-fold) 338 H. L. Robey and others

1 except that they were on the bottom of the membrane rather than on the top. Stage 6 cells were extremely V//////////A Y///////////A large and flat with well-developed actin cables or stress fibres in their cytoplasm. These cells were negative for both cytokeratins. Clearly HRPE committing (stage 3) and committed (stage 4) to migration became more Y///////////A W/////////7A spread and mobile, and exhibited K18 and K19 staining whereas HRPE on the upper surface, which were not yet involved in migration, were negative. In the post- migrational phase on the lower surface of the mem- Y//////////A MY////. brane, staining was progressively lost (stages 5 and 6). The post-migrational cells (stage 6) were well spread and flat but were probably immobile, given the presence of stress fibres in their cytoplasm (Herman et '//A al. 1981). Stress fibres are found in many tissue culture sm%mm^mm systems and these large actin bundles are ESS Denotes +ve K18 & K19 staining thought to be a feature of stationary cells (Herman et al. 1981). Stress fibres abound in cells with strong Fig. 12. A model for the involvement of actin, keratin 18 adhesion to their substrata (Willingham et al. 1977). and keratin 19 in the migration of HRPE in vitro. (1) A They are contractile (Kreis and Birchmeier, 1980) and rounded cell with diffuse actin staining is keratin 18 and 19 may maintain tension between the cells and their rigid negative. (2) A spread cell with diffuse or filamentous substratum (isometric forces) (Grierson et al. 1988; actin, but is still keratin 18 and 19 negative. (3) A well- Heath and Dunn, 1978). It has been said that spread flattened cell with a cell process placed in a pore, has filamentous actin and is always keratin 18 and 19 cytokeratins help cultured epithelium to remain flatan d positive. (4) A cell actively migrating through a pore has spread (Aubin et al. 1980), but this would not be the filamentous actin and is keratin 18 and 19 positive. (5) This case with K18 and K19, which are present during is identical to stage 3 although the cell is on the lower spreading, which involves locomotion, but cannot be surface rather than the upper surface and has a foot detected in spread or sedentary HRPE. process within the pore. (6) A large flattened migrated cell The expression of vimentin intermediate filaments in with actin stress cables and negative for keratin 18 and 19. HRPE would not interfere with their mobility. Vimen- tin, a rigid cytoskeletal element, has recently been shown to have unique viscoelastic properties, demon- increase in the K19 population on the migrated side of strating a flexible nature under low stress; this suggests the membranes. The enhancement of the K18 and K19 a role that vimentin may play in maintaining cell populations from upper to lower sides of the mem- integrity during events such as locomotion (Janmey et branes occurred because all cells caught in transit al. 1991). through the pores in the membrane were positive for A current area of research interest has been the these two cytokeratins. In other words, K18 and 19 altered expression of cytokeratin subtypes with the were expressed by HRPE involved in active migration. emergence of premalignant or malignant epithelial Six stages in the migration process were identified lesions (Smedtds et al. 1990; Cintorino et al. 1990; and summarised in Fig. 12. Stage 1 consisted of the Schaafsma et al. 1990; Nagle et al. 1991; Markey et al. rounded cells on the upper surface of the membrane, 1991). It has been suggested that the expression of K8, which were settled but not spread. These exhibited a K18 or K19 may be upregulated in premalignant or diffuse actin staining pattern and were invariably K18 malignant lesions of various tissues including the cervix, and 19 negative. Partially spread cells, which by SEM prostate, tongue or skin. In addition, infiltrating cells of had numerous microvilli, were called stage 2 and these transitional cell carcinomas were shown to be strongly were mostly negative for the two cytokeratins, although immunoreactive for K18 (Schaafsma et al. 1990). some cells had the intense staining non-filamentous Furthermore, it has been shown in transfected mouse L aggregate adjacent to the nucleus seen in recently cells that expression of K8 and K18 significantly settled cells (see Fig. IE). A third stage consisted of increased their invasive ability (Chu et al. 1990) and well-spread cells with cytoplasm invading the pores. that K8 and K18 are characteristic of invasive squamous These cells had prominent processes seen by both light cell carcinomas (Markey et al. 1991). Invasive epithelia microscopy and SEM, a filamentous network of actin in would be expected to be highly motile, but in addition the cytoplasm, were always K18 positive and were an association between cell flexibility and K18 ex- usually K19 positive. HRPE in transit through pores pression has been shown (Schaafsma et al. 1990; were invariably K18 and K19 positive and these we Markey et al. 1991). Undoubtedly, HRPE in transit called stage 4. Stage 4 cells had a distinctive actin through pores would need to be highly flexible and to network in the cytoplasm spread on the under side of some extent this was born out by our SEM studies, the membrane but the cytoplasm in the pore and the which demonstrated the complex distortion of cell remnant on the upper surface exhibited a diffuse shape required to negotiate successfully the pore staining pattern. Stage 5 cells were identical to stage 3 channels through the membrane. Cytokeratins and epithelial cell behaviour 339

Although this study has pinpointed as association of the formation of epiretinal membranes? Seminars Ophthalmol. 2, keratins 18 and 19 with RPE cell migration it would be 99-109. Grierson, I., Joseph, J., Miller, M. and Day, J. E. (1988). Wound interesting to investigate the expression of keratins 8 Repair: The fibroblast and the inhibition of scar formation. Eye 2, and 7 in this migration model. 135-148. Our observation of an apparent specific association Heath, J. P. and Dunn, G. A. (1978). Cell to substratum contacts of between RPE locomotion and immunoreactivity of chick fibroblasts and their relation to the microfilament system. A RPE cytoskeletal for particular monoclonal correlated interference-reflexion and high-voltage electron- antibody(s) in vitro suggests that the migratory behav- microscope study. J. Cell Sci. 29, 197-212. iour of RPE may be detected by immunocytochemical Herman, I. M., Crisona, N. J. and PoUard, T. D. (1981). Relation between cell activity and the disruption of cytoplasmic actin and means in tissue sections. Indeed, we have already . J. Cell Biol. 90, 84-91. documented the presence of a subset of K18 positive Hiscott, P., Morino, I., McKechnie, N. M., Grierson, I. and Gregor, RPE in situ in some retinal scars (Hiscott et al. 1991) Z. (1991). Proliferating retinal pigment epithelial cells in subretinal and it is also possible that K18 may be useful in the membrane. In Proc. XXVI Int. Congr. Ophthalmol. (ed. Yew detection of migratory activities by other simple K.C., Chong A.B., Mun C.W., Chew P. and Urn A.S.M.) epithelia. Amsterdam: Elsevier Science Pubs. Hitchins, C. A. and Grierson, I. (1988). Intravitreal injection of fibroblasts: the pathological effects on the ocular tissues of the rabbit following an intravitreal injection of autologous skin This work was supported by Wellcome Grant:17095/1.4R. fibroblasts. Br. J. Ophthalmol. 72, 498-510. Horwltz, B., Kupfer, H., Eshar, Z. and Geiger, B. (1981). Reorganization of arrays of prekeratin filaments during mitosis. Immunofluorescence microscopy with multiclonal and monoclonal References prekeratin antibodies. Exp. Cell Res. 134, 281-290. Janmey, P. A., Euteneuer, U., Traub, P. and Schliwa, M. (1991). Aubin, J. E., Osborn, M., Franke, W. W. and Weber, K. (1980). Viscoelastic properties of vimentin compared with other Intermediate filaments of the vimentin-type and the cytokeratin- filamentous biopolymer networks. J. Cell Biol. 113, 155-160. type are distributed differently during mitosis. Exp. Cell Res. 129, Kampik, A., Kenyon, K. R., Michels, R. G., Green, W. R. and De la 149-165. Cruz, Z. C. (1981). Epiretinal and vitreous membranes Boulton, M. E., Marshall, J. and Mellerio, J. (1982). Human retinal comparative study of 56 cases. Arch. Ophthalmol. 99, 1445-54. pigment epithelial cells in tissue culture: a means of studying Kasper, M., MoU, R., Stosiek, P. and Karsten, U. (1988). Patterns of inherited retinal diseases. Birth Defects: Original Articles Series 18, cytokeratin and vimentin expression in the human eye. 101-118. Histochemistry 89, 369-377. Campochiaro, P. A., Jerdan, J. A. and Glaser, B. M. (1984). Serum Kreis, T. E. and Birchmeier, W. (1980). Stress fibre sarcomeres of contains chemoattractants for human retinal pigment epithelial fibroblasts are contractile. Cell 22, 555-561. cells. Arch. Ophthalmol. 102, 1830-1833. Kurki, P., Ogata, K. and Tan, E. M. (1988). Monoclonal antibodies Chu, Y-W., Oshima, R. G. and Hendrix, M. J. C. (1990). Expression to proliferating cell nuclear antigen (PCNA)/cyclin as probes for of keratins 8 and 18 in mouse L cells results in an increase in proliferating cells by immunofluorescence microscopy and flow migratory and invasive ability. J. Cell Biol. 3 (ASCB abstracts, 30th cytometry. J. Immunol. Meth. 109, 49-59. Ann. Meeting.), 41a. Lane, E. B., Goodman, S. L. and Trejdosiewicz, L. K. (1982). Cintorino, M., Petracca, R., Vindigni, C, Tripodi, S. A. and Disruption of the keratin filament network during epithelial cell Leoncini, P. (1990). Topography-related expression of individual cytokeratins in normal and pathological (non-ncoplastic and division. EMBO J. 1, 1365-1372. neoplastic) human oral mucosa. Virchows Arch. Pathol. Anal. Lindberg, K. and Rheinwald, J. G. (1991). Suprabasal 40 kd keratin Physiol. 417, 419-426. (K19) expression as an immunohistologic marker of premalignancy Docherty, R. J., Edwards, J. G., Garrod, D. R. and Mattey, D. L. in oral epithelium. Am. Pathol. 134, 89-98. (1984). Chick embryonic pigmented retina is one of the group of Markey, A. C, Lane, E. B., Churchill, L. J., MacDonald, D. M. and epithelioid tissues that lack cytokeratins and desmosomes and have Leigh, I. M. (1991). Expression of simple epithelial keratins 8 and intermediate filaments composed of vimentin../. CellSci. 71,61-74. 18 in epidermal neoplasia. J. Invest. Dermatol. 91, 763-770. Edwards, R. B. (1981). Culture of mammalian retinal pigment McKechnie, N. M., Boulton, M., Robey, H. L., Savage, F. J. and epithelium and neural retina. In Visual Pigments and Purple Grierson, I. (1988). The of human retinal pigment Membranes. Methods of Enzymology, vol. 81 (ed. L. Packer), pp. epithelium: in vitro and in vivo. J. Cell Sci. 91, 303-312. 39-43. New York: Academic Press. MoU, R., Franke, W. W. and SchiUier, D. L. (1982). The catalog of Franke, W. W., Schmid, E., Grund, C. and Geiger, B. (1982). human cytokeratins; Patterns of expression in normal epithelia, proteins in nonfilamentous structures: tumours and cultured cells. Cell 31, 11-24. Transient disintegration and inclusion of subunit proteins in Nagle, R. B., Brawer, M. K., Kittelson, J. and Clark, V. (1991). gTanular aggregates. Cell 30, 103-113. Phenotypic relationships of prostatic intraepithelial neoplasia to Franke, W. W., Schmid, E., Wellsteed, J., Grand, C, Gigi, C, Gigo, invasive prostatic carcinoma. Am. J. Pathol. 138, 119-128. O. and Geiger, B. (1983). Changes of cytokeratin filament Owaribe, K., Kartenbeck, J., Rungger-Brandle, E. and Franke, W. organization during the cell cycle: selective masking of an W. (1988). of retinal pigment epithelial cells: immunologic determinant in interphase PtK2 cells. J. Cell Biol. 97, 1255-1260. Interspecies differences of expression patterns indicate independence of cell function from the specific complement of Gigi, O., Geiger, B., Eshhar, Z., Moll, R., Schmid, E., Winter, S., Schiller, D. L. and Franke, W. W. (1982). Detection of a cytoskeletal proteins. Cell Tiss. Res. 254, 301-315. cytokeratin determinant common to diverse epithelial cells by a Ramaekers, F. C. S., Huysmans, A., Moesker, O., Kant, A., Jap, P., broadly cross-reacting monoclonal antibody. EMBO J. 1, 1429- Herman, C. and Vooijs, P. (1983). Monoclonal antibody to 1437. keratin filaments specific for glandular epithelia and their tumours: Giorno, R. (1984). A comparison of two immunoperoxidase staining Use in surgical pathology. Lab. Invest. 49, 353-361. methods based on the avidin-biotin interaction. Diagnos. Schaafsma, H. E., Ramaekers, F. C. S., Van Muljen, G. N. P., Lane, Immunol. 2, 161-166. E. B., Leigh, I. M., Robben, H., Huijsmans, A., Ooms, E. C. M. Grierson, I., Hiscott, P. S., Hitching, C. A., McKechnie, N. M., and Rultcr, D. J. (1990). Distribution of cytokeratin polypeptides White, V. A. and McLeod, D. (1987). Which cells are involved in in human transitional cell carcinomas, with special emphasis on 340 H. L. Robey and others

changing expression patterns during tumour progression. Am. J. Willingham, M. C, Yamada, K. M., Yamada, S. S., Pouyssegur, J. Pathol. 136, 329-343. and Pastan, I. (1977). Microfilament bundles and cell shape are Smedtds, F., Ramaekers, F. C. S., Robben, H., Pruszczynski, M., related to adhesiveness to substratum and dissociable from growth Van mujjen, G. N. P., Lane, E. B., Leigh, I. M. and VooUs, P. control in cultured fibroblasts. Cell 10, 375-3S0. (1990). Changing patterns of expression during progression of cervical intraepithelial neoplasia. Am. J. Pathol. 136, 657-667. {Received 9 January 1992 - Accepted 9 March 1992)