The cytoskeletal elements of human retinal pigment : in vitro and in vivo

NICOL M. McKECHNlE1*, MIKE BOULTON2, HELEN L. ROBEY1, FELICITY J. SAVAGE1 and IAN GRIERSON1

^Departments of and ^Clinical Ophthalmology, Institute of Ophthalmology, 17-25 Cayton St, London ECIV9AT and Judd St, London WCI, UK

* Author for correspondence

Summary

The cytoskeletal elements of normal (in situ) and identifiable. 18 positive cells were cultured human retinal pigment epithelium (RPE) identified in all RPE cultures (other than early were studied by a variety of immunocytochemical primaries), regardless of passage number, age or techniques. Primary to and sex of the donor. In post-confluent cultures cyto- were used. Positive immunoreactivity 18 cells were identified growing over cyto- for vimentin was obtained with in situ and cultured negative cells, suggesting an association material. The pattern of reactivity obtained with of cytokeratin 18 immunoreactivity with pro- antisera and monoclonals to cytokeratins was more liferation. Immunofluorescence studies of retinal complex. Cytokeratin immunoreactivity could be scar tissue from two individuals revealed the pres- demonstrated in situ and in cultured cells. The ence of numerous cytokeratin 18 positive cells. pattern of cytokeratin expression was similar to These findings indicate that RPE cells can be ident- that of simple or glandular epithelia. A monoclonal ified by their cytokeratin immunoreactivity and that specifically recognizes cytokeratin 18 that the overt expression of cytokeratin 18 may be identified a population of cultured RPE cells that associated with proliferation of human RPE both in had particularly well-defined filamentous networks vitro and in vivo. within their . Freshly isolated RPE was cytokeratin 18 negative by immunofluorescence, Key words: RPE, , keratin, cytokeratin, but upon culture cytokeratin 18 positive cells were proliferation.

Introduction cells of myogenic origin, in neurones, glial fibrillary acid (GFAP) in glia of Proliferation of the RPE has been suspected in a variety the central nervous system and the cytokeratins in of ocular diseases including proliferative diabetic retino- epithelial cells. The cytokeratins, a related family of pathy and is known to complicate ocular trauma and about 19 polypeptides identified by two-dimensional gel retinal detachment (Grierson et al. 1987). Therefore electrophoresis, occur in varying combination in different immunocytochemical markers of the RPE would be epithelia (Moll et al. 1982). Therefore epithelia can be powerful tools in tissue culture studies and in the characterized by the cytokeratins they express. Because identification of RPE cells in pathological material. In an of their usefulness in tumour identification (Osborn & attempt to identify such markers, we have studied the Weber, 1983; Ramaekers et al. 1983a), antisera and cytoskeletal proteins expressed by cultured and patho- monoclonal antibodies to intermediate filaments are logical RPE, using antibodies to vimentin and the cyto- readily available from commercial sources. . As an epithelium, the RPE would be expected to and occur in most cell express cytokeratins, or at least to show co-expression of types and are therefore of little potential use as cell cytokeratins with vimentin in vitro as has been demon- markers, but they can be informative with regard to strated for some other epithelial cell types (Lane et al. cellular activities such as locomotion and phagocytosis. 1983; LaRocca & Rheinwald, 1984). However, following Intermediate filaments, five types of which have been studies of chick RPE, Docherty et al. (1984) stated identified, are tissue specific (Lazarides, 1980). In gen- "although RPE is an epithelium according to all other eral vimentin occurs in mesenchymal cells, in criteria, it belongs to a group of tissues (including Journal of Cell Science 91, 303-312 (1988) Printed in Great Britain © The Company of Biologists Limited 1988 303 vascular , iris and lens-forming epithelium) of this antibody. It failed to react with the but showed that have intermediate filaments composed of vimentin strong of all cells in the dermis. The specificity of the and possess neither cytokeratins nor desmosomes". Other monoclonal to vimentin was also confirmed by immunoblotting. Cytokeratins. An antiserum to human epidermal keratins of authors have demonstrated cytokeratin-positive immuno- 3 3 reactivity of normal and pathological RPE in human, 56X10 and 64X10 molecular weight (Schlegel et al. 1980), monkey, bovine and rabbit material (HiscotteJ al. 1984). and an antiserum of wide cross-reactivity raised to bovine muzzle epidermal keratins (Franke et al. 1978; Steinert, 1975), The present study was undertaken to determine the were obtained from Dakopattis, Denmark. These two antisera distribution and the nature of the cytoskeletal proteins are commonly used for immunodiagnostic purposes. There expressed by the RPE in vitro and in vivo and to establish reactivity has been published (Franke et al. 1978; Pinkus et al. if cytokeratin immunoreactivity could be used as a 1985; Taxy et al. 1985; Battifora & Kopinski, 1985; Makai et al. specific marker of RPE cells. The controversy concerning 1985). the expression of cytokeratins by the RPE, and our The reactivity, according to the suppliers, of the monoclonal interest in the culture and pathology of human RPE, antibodies to the cytokeratins is most easily described using the prompted us to investigate the cytokeratin immunoreac- classification system of Moll et al. (1982). Monoclonals, RSK tivity of cultured human RPE in some detail. 102 (cytokeratins 5 and 8) described by Broers et al. (1986), RSK 105 (cytokeratin 7), NCL 5D3 (cytokeratins 8, 18 and 19) described by Angus et al. (1987) and RGE 53 (cytokeratin 18) Materials and methods described by Raemaekers et al. (19836), were obtained from Eurodiagnostics. A further five monoclonal antibodies were obtained from ICN: immunobiologicals, AE1 (cytokeratins 10, 14, 15, 16 and 18) and AE3 (cytokeratins 1, 2, 3, 4, 5, 6, 7 and Post-mortem eyes from 11 individuals with no known ophthal- 8) described by Woodcock-Mitchell et al. (1982), KG8.13 mic disease were obtained from various hospitals in London. (cytokeratins 1, 5, 6, 7, 8, 10, 11 and 18) described by Gigie? al. Age of donors ranged from 7 years to 95 years of age. Post- (1982), KG8.12 (cytokeratins 13 and 16) and KG4.62 (cyto- mortem times ranged between 24 and 48 h. ). Immunoblotting experiments using RPE cell Human RPE cells were isolated as described (Boulton et al. cytoskeletal extracts and the antibodies to the cytokeratins were 1983). In some instances cells were retained for immunocyto- conducted to confirm the specificity of the antibodies. AE1, chemical investigation (see below). Cells were grown in Ham's AE3 and RSK 105 were used for immunoblotting but not in the F10 medium supplemented with 20% foetal calf serum, 0-4% remainder of this study. glucose and antibiotics. Cultures were maintained at 37°C in a CO2 incubator (95 % air and 5 % CO2). Cells were either grown in 25 cm2 Falcon flasks or in 8-chamber tissue culture slides Gel electrophoresis and immunoblotting Cultured RPE cells were harvested from 75 cm2 flasks using (Miles Scientific). Subcultures were obtained by detachment of - EDTA (20mM in calcium/magnesium-free PBS). Cell pellets cells from their substratum with 0 25 % trypsin and then plated were extracted with 1 % Triton X-100 in PBS for 2min. The out in growth medium at an appropriate split ratio. insoluble residue was dissolved by boiling for 5 min in sodium Prior to immunostaining cultures were washed twice with dodecyl sulphate (SDS) sample buffer (Laemmli, 1970). Gel phosphate-buffered saline (PBS) and fixed with cold (-20°C) electrophoresis was performed in 12-5% polyacrylamide slab methanol and acetone for 4 and 2min, respectively. For gels, as described by Laemmli (1970). cultures grown in flasks the acetone fixation was omitted. Electrophoretically separated polypeptides were transferred Histopathological material to a nitrocellulose sheet by overnight blotting at 100 mA following the method of Towbin et al. (1979). Immunostaining Pathological material (retinal scar tissue from two individuals) of the transferred proteins was conducted as described by was obtained from Moorfields Eye Hospital following vitrec- Broers et al. (1986). Anti-mouse and anti-rabbit peroxidase tomy surgery. One specimen was removed in the treatment of conjugates were obtained from Sigma. macular pucker, the other from an eye with massive periretinal proliferation following retinal detachment. Normal human eyes were obtained following their use for corneal grafting. Material Secondary! antibodies for histological examination was frozen in liquid nitrogen, FITC- and TRITC-labelled anti-mouse IgG and goat anti- embedded in OCT Compound (Miles Scientific). Cryostat rabbit IgG and peroxidase-labelled goat anti-mouse IgG were sections mounted on glass slides were fixed in cold ( — 20 °C) obtained from SIGMA. Gold-labelled goat anti-mouse IgG and methanol (4min) and acetone (2min) prior to immunostaining. chemicals for the silver enhancement of immuno-gold staining In some cases when RPE cells were being isolated for culture were obtained from Janssen Pharmaceuticals. (from a 16- and a 71-year-old donor) a small volume of the cells was pipetted onto glass slides, air dried and then fixed in Immunostaining methanol and acetone. Following fixation, sections and cultures were treated in an identical manner. Samples were incubated in 1 % normal goat Primary antibodies serum (NGS) for 10 min to block non-specific binding. All The antibodies used in this study were obtained from a variety subsequent dilutions of antibodies were also made up in 1 % of suppliers. NGS in PBS. For immunofluorescence, samples were incu- Vimentin. A monoclonal antibody (MAb), and a rabbit bated in a 1-10 dilution of the primary antibody. For immuno- antiserum to vimentin were obtained from Eurodiagnostics, peroxidase staining the working dilution of the primary anti- Holland. The specificity of this monoclonal antibody has been bodies was increased to a 1-40 dilution. When using the established by immunodiffusion (Ouchterlony), immunoblot- immunoperoxidase method, endogenous peroxidase was ting, immunoelectrophoresis and immunofluorescence of tissue blocked by a 30-min incubation in methanol containing 0-05 % sections (suppliers data). Immunofluorescent staining of cryo- H2O2. Incubation in the primary antibody was for 45-60 min. stat sections of bovine and human confirmed the specificity Samples were then washed in three changes of PBS over 20 min.

304 N. M. McKechnie et al. -3 Mrx10 200 V, ft ; -7*

92-5

69 •*;;• # 5 6 7,8 * m ~* 14,15 46 18" 19 1

2'3 4 10 11 12 .13 14 15 : la

Fig. 1. Immunoblotting data obtained using cytoskeletal extracts from cultured human RPE cells and different antibodies to the cytokeratins and vimentin. Lane 1, Amido Black. Lane 2, antiserum to human epidermal keratins, 1: 100 dilution. Lane 3, antiserum to bovine muzzle keratins, 1:100 dilution. Lane 4, KG8.13. Lane 5, RGE 53. Lane 6, NCL SD3. Lane 7, RSK 102. Lane 8, AE1. Lane 9, AE3. Lane 10, monoclonal to vimentin. Lane 11, KQ4.62. Lane 12, KG8.12. Lane 13, antiserum to human epidermal keratins, 1:200 dilution. Lane 14, antiserum to bovine muzzle keratins, 1:200 dilution. Lane 15, RSK 105. Lane 16, Amido Black. The expected positions of some of the cytokeratins to which the various antibodies are reactive are given. The faint staining obtained with RGE 53 is indicated by arrowheads.

Samples were labelled with a 1-40 dilution of FITC and Quantification of cytokeratin 18 staining in pre- TRITC conjugates, or a 1-10 dilution of gold-labelled anti- confluent, confluent and post-confluent cultures moiuse or anti-rabbit IgG, or a 1-100 dilution of peroxidase- Cultures derived from donors of 17, 26, 66 and 85 years of age at labelled anti-mouse IgG for 45-60 min. Following a further passage numbers 8th, 8th, 11th and 8th, respectively, were three washes in PBS, samples stained for immunofluorescence used. The cells were plated into 8-chamber tissue culture slides were mounted, using Flurostab (Eurodiagnostics) to prevent at low density. Three slides were used for each culture. The fading. Samples were examined using a Leitz microscope cells were fixed as described above at different times when the equipped with filters for fluorescein and rhodamine fluor- cultures were pre-confluent, confluent and post-confluent by escence. Peroxidase-labelled material was incubated with 3- microscopic examination. The cultures were stained for cyto- amino-9-ethylcarbazole to give a distinctive red reaction prod- keratin 18 using the immunoperoxidase method described uct. Following development of the peroxidase, the samples above. A haematoxylin counter-stain was employed to facilitate were counter-stained with haematoxylin and mounted. The counting of cytokeratin 18 negative cells. Nuclei and cytokeratin immuno-gold-stained material was silver-enhanced in accord- 18 positive cells were counted, using a X20 objective on a Leitz ance with the manufacturer's instructions and then counter- Orthoplan microscope. Nuclei and positive cells within the stained with light green (cell cultures) or haematoxylin (sec- 2 limits of the 35 mm frame marks (300x450 (an or 0-135 mm ) tioned material). Immuno-gold silver specimens were exam- were recorded for one field picked at random within each well. ined using a Reichert-Jung Polyvar microscope equipped for In this manner eight fields were counted from each culture at epi-polarization. each stage of confluence. In cases of double immunofluorescence microscopy both antibodies, a monoclonal and an antiserum raised in rabbit, were applied simultaneously at the appropriate dilution, as were the species-specific FITC and TRITC conjugates. Results The following controls were used to determine the specificity of the immunostaining. (1) The primary antibody was omitted. Immunoblotting (2) The primary antibody was substituted with normal sera The specificities of the antibodies to the various cytokera- from the appropriate species or an irrelevant monoclonal tins and the monoclonal antibody to vimentin were antibody. (3) To test for cross-species reactivity of the second- ary antibodies in the double immunofluorescence, one of the confirmed by immunoblotting following SDS-PAGE of pair of primaries was omitted. Following incubation with the RPE cell (Fig. 1). The antisera to epider- secondary antibodies, staining was assessed. Procedures (1) and mal keratins and bovine muzzle keratins showed similar (2) produced negative staining. Procedure (3) resulted in patterns of reactivity. Both stained bands of MT 56 to 58 positive staining when examined with one filter combination (X103) (probably cytokeratins 5, 6) and another group of 3 and negative or faint background staining with the other. bands at MT 50 to 54 (X 10 ) (probably cytokeratins 7, 8,

RPE cytoskeletons 305 Fig. 2. Immuno-gold/ of cryostat sections of RPE and choroid from a 25-year-old male. As the RPE is both pigmented and autofluorescent (it contains numerous lipofuscin granules), immunostaining was detected using epi-polarization. Bright-field and epi-polarization are shown in each case. A,B. Monoclonal antibody to vimentin. Both the RPE and the cellular elements of the choroid are stained. C,D. KG8.13. Although there is some background staining the RPE is intensely stained. E,F. RGE 53. This antibody produces light but specific staining of the RPE. G,H. Negative control showing only scattered background staining. Bar, 50 j.im.

14, 15). Both also stained a band at 40X103 (probably corresponding to cytokeratin 18 (45XlO3). NCL 5D3 cytokeratin 19). Unidentified bands in the region of stained bands at 40, 45 and 52-5 (XlO3) in agreement 3 3 80X10 were also stained. KG8.13 stained bands that with its reactivity with cytokeratins 19, 18 and 8 (X 10 ). would correspond with cytokeratins 5 and 6 (58 and 56 A fainter band in the region of 56X103 was also present. (X103)), cytokeratins 7 and 8 (54 and 52-5 (XlO3)) and RSK 102 detected a band at 52-5 (XlO3), which would cytokeratin 19 (40xl03). RGE 53 faintly stained a band correspond with its reactivity with cytokeratin 8. Two

306 N. M. McKechnie et al. Fig. 3. Low-power double immunofluorescence micrographs demonstrating the presence of: A, vimentin, using a rabbit antiserum; and B, cytokeratins, using an antiserum to human epidermal keratins and goat anti-rabbit IgG FITC conjugate as the second layer. C,D, the areas seen in A and B, respectively, but with the rhodamine filters in position to demonstrate the staining obtained with RGE 53 (specific for cytokeratin 18) and goat anti-mouse TRITC conjugate as the second layer. It can be seen that all cells stain with the antisera to vimentin and human epithelial keratins but only a proportion of the cells stain with the monoclonal RGE 53. Some corresponding cells are arrowed. Bar, 100 fim. other faint bands of lower molecular weight were also approximately 47X103 in addition to its expected reac- present. RSK 102 is reported to react with cytokeratin 5 tivity with a band at 54x 103. but this was not observed. AE1 stained bands in accord- ance with its reactivity with cytokeratins 14, 15 and 19 Immunostaining (50, 50 and 45 (X103)). A band corresponding to In situ RPE. Immunostaining of this material was cytokeratin 16 (48X103) may also have been present. AE3 complicated by pigmentation and autofluorescence. To stained a large number of bands in accordance with its overcome these problems we utilized the immuno-gold/ broad reactivity. Bands of molecular weight less than silver technique. Silver reaction product was detected 52-5 X103 were also stained. This is not in agreement with using epi-polarization illumination. Results indicated the its known lack of reactivity with the low molecular weight presence of both vimentin and cytokeratins. The antisera acidic cytokeratins. The monoclonal antibody to vimen- were not used on this material. The RPE stained posi- tin stained a very prominent band in the region 50-57 tively with the following monoclonal antibodies to cyto- 3 (X 10 ) in addition to some small bands of lower molecu- keratin. KG8.13, RGE 53, RCK 102, stained faintly with 3 lar weight. KG4.62 intensely stained a band of 40X10 , NCL 5D3, and was negative with KG8.12 and KG4.62 which corresponds with its reactivity with cytokeratin 19. (Fig. 2). KC8.12 (cytokeratin 13 and 16) did not stain any bands. Freshly isolated RPE. Immunofluorescent staining of RSK 105 recognized a band at 54xl03 (cytokeratin 7) in this material was difficult to interpret due to high levels of agreement with its expected reactivity and another band autofluorescence. This was less of a problem with ma- at approximately 48X103. In general the reactivities of terial from the 16-year-old donor, which was considered the antibodies was in accord with their published or to be vimentin positive. Using the monoclonal antibody stated (by the supplier) specificities. The exceptions to cytokeratin 18, very faint perinuclear staining was being the reaction of NCL 5D3 with a band correspond- observed in about 50% of the cells examined, but this ing to cytokeratins 5 and 6. The failure of RSK 102 to was so faint it proved impossible to photograph. recognize a band corresponding to cytokeratin 8. The The full battery of antibodies could not be used due to reaction of the vimentin monoclonal with some low the shortage of this material. molecular weight bands, these may be breakdown prod- Cultured RPE. By immunofluorescence, all cultured ucts. Finally, the reaction of RSK 105 with a band at RPE cells stained intensely for vimentin (Fig. 3A). All

RPE cytoskeletons 307 Fig. 4. Immunofluorescence micrographs illustrating: A, vimentin staining; and B, cytokeratin staining, using monoclonal antibodies to vimentin and cytokeratin 18 (RGE 53). The filament arrays are prominent. C,D, immuno-gold/silver staining of similar cultures; C, vimentin; D, cytokeratin 18 (RGE 53). The immuno-gold/silver technique has the advantage of allowing counter-staining and visualization of the immunochemically negative cells. Bars, 100 ^m. cells stained positively, to varying degrees, using the nostaining, using the immuno-gold/silver technique broadly cross-reacting antiserum to bovine muzzle epi- (IGS), is shown in Fig. 4C,D. Unlike the immunofluor- dermal keratins and the antiserum to human epidermal escence preparations, where negative cells cannot be cytokeratins. These two antisera produced similar pat- seen, the IGS technique permits counter-staining and terns of staining. It was particularly noticeable that, visualization of the negative cells. while some cells were lightly and diffusely stained, others Primary cultures of RPE. Primary cultures were im- were intensely stained and showed pronounced filaments munostained with RGE 53 to determine at what point and whorls within their cytoplasm (Fig. 3B). An essen- cultured cells started to show intense labelling with this tially similar pattern of immunofluorescent staining was antibody. All cells in primary culture stained positively obtained with the monoclonal antibodies KG8.13, RSK for vimentin (Fig. 5A). In areas where there were large 102 and NCL 503. KG8.12 showed no reactivity with cells containing conspicuous lipofuscin, cytokeratin 18 cultured human RPE. Positive immunofluorescence was positive cells were absent. Where cells were smaller, and obtained using the monoclonal antibody to cytokeratin 18 more densely packed, cytokeratin 18 positive cells were (RGE 53). However, this antibody only reacted with a identifiable (Fig. 5B,C). limited proportion of cultured RPE cells (Fig. 3C,D). The monoclonal antibody KG4.62 (to cytokeratin 19) produced a similar pattern of intracellular staining to that Cytokeratin 18 in pre-confluent, confluent and post- obtained with RGE 53 but considerably fewer cells were confluent RPE cultures stained. Fig. 3A,C show the same area of cells double As intense cytokeratin 18 immunoreactivity appeared to stained for vimentin (A) and cytokeratin 18 (C). be associated with the establishment of RPE cells in Fig. 3B,D show the same area of cells double stained culture, the effect of cell proliferation and cell density on using the antiserum to human epidermal keratins (B) and the expression of cytokeratin 18 was investigated by the monoclonal antibody to cytokeratin 18 (D). It can be establishing the number of cytokeratin 18 positive and seen that the cultured RPE cells show co-expression of negative cells in cultures differing only in the degree of vimentin and cytokeratins but that there is a variation in confluence they were allowed to reach before being fixed the pattern of cytokeratin expression from cell to cell. and stained for cytokeratin 18. The results (Fig. 6) The filamentous nature of the vimentin and cytokeratin indicated that three of the cultures showed a decrease in 18 staining can be better appreciated at higher magnifi- cation (Fig. 4A,B). Vimentin and cytokeratin 18 immu- the ratio of cytokeratin 18 positive cells to negative cells (cultures A, B, D). One culture (C) maintained a fairly

308 N. M. McKechnie et al. A 1r 90r-B 60 80 I 50 I 70 - 60 held 5 50

8 30 40 o •33 30 Z 20 - •35 ~ U -45 20 10 10 13 •26 11 0 1 0 I P c pc P pc 70 90 -D 1 80 60 - 70 5 - 2 ° 60 u |40 ; 50 ?30 40 "o 6 30 Z 20 20 •16 10 10 •12 J-16 •34 •09 n 0 i pc pc Fig. 6. Histograms of cytokeratin 18 positive cells (black area) and nuclear counts for RPE cell cultures at varying degrees of confluence (p, pre-confluent, c, confluent, pc, post-confluent). A. 11th passage, 66-year-old donor. B. 8th passage, 17-year-old donor. C. 8th passage, 85-year-old donor. D. 8th passage, 29-year-old donor. The ratio of cytokeratin 18 positive cells to cell nuclei falls slightly or Fig. 5. Immunofluorescence staining of a primary culture of remains constant (C), with increasing cell density. Bars show RPE from a 71-year-old female donor. All cells, regardless of the standard error. size, stain intensely with the monoclonal to vimentin (A). These cells also contain conspicuous lipofuscin granules. Using the monoclonal to cytokeratin 18 (RGE S3) large cells, monoclonal antibody to cytokeratin 18. Many cells within containing numerous lipofuscin granules, were negative (B). the epiretinal membranes contained autofluorescent ma- Small cells, with fewer inclusions, were faintly stained (/) or terial. This material fluoresced yellow/gold and was in some instances intensely (;') stained for cytokeratin 18 (C). easily distinguished from the apple green fluorescence of Bars, 100 ^m. B,C are at the same magnification. the FITC conjugate. For the purposes of black-and- white photography the specimens were also photo- constant ratio of cytokeratin positive to negative cells graphed with the rhodamine filters in position to demon- regardless of cell density. strate the autofluorescence in isolation from the specific With increasing cell density, in the confluent and post- staining (Fig. 8A,B). confluent cultures, the cytokeratin 18 positive cells were seen to be forming a second layer of cells lying over the cytokeratin 18 negative cells (Fig. 7). Discussion

Immunostaining: retinal scar tissue Throughout this investigation formaldehyde fixation was Immunofluorescent staining of the retinal scar tissue avoided because of its well-documented adverse effect on (epiretinal membranes) revealed the presence of cells the immunological identification of cytokeratins, particu- showing moderately strong immunoreactivity with the larly those of simple or glandular epithelia (Pinkus et al.

RPE cytoskeletons 309 t

Fig. 8. Immunofluorescence staining of a frozen section of an epiretinal membrane (massive periretinal proliferation specimen) stained for cytokeratin 18. Positively staining cells can be seen in A (arrowheads). Autofluorescence, with rhodamine filtration, is shown in B. Bar, 100(im.

vitro (Franke et al. 1979a), parietal endoderm (Lane et al. 1983), and HeLa cells (Franke et al. 19796). One of the purpose of this study was to determine if cytokeratin immunoreactivity can be used to identify RPE cells in vito and in vitro. Of the antisera and monoclonal antibodies used in this study the monoclonal antibodies KG8.13 and NCL 5D3 and RSK 102 were the Fig. 7. Representative micrographs of one of the cultures most reliable for this purpose as they positively stained all used in the generation of Fig. 6C (8th passage, 85-year-old RPE cells in situ and in culture. The two antisera to donor). A. Pre-confluent. B. Confluent. C. Post-confluent. epidermal keratins are useful but there is a greater danger The cells have been stained for cytokeratin 18 using RGE S3 of non-specific staining. Non-specific staining is difficult and an indirect peroxidase method. The cultures have been to assess when immunostaining cultured cells. Tissue counter-stained with haematoxylin to demonstrate the nuclei sections usually contain positively and negatively staining of the cells. With increasing cell density (B and C) cell types, allowing detection of false positives. cytokeratin 18 positive cells can be seen spreading over the All cells in all the cultures examined showed positive surface of the cytokeratin 18 negative cells (arrows). Bar, reactivity with the antisera to cytokeratins, the mono- 100 ^m. clonal and antisera to vimentin and the monoclonal antibodies KG8.13, RSK 102 and NCL 5D3, which recognizes several of the cytokeratin polypeptides. This 1985; Ramaekers et al. 1983c). We suspect that the use of finding confirms the purity of the RPE cultures as all of aldehyde fixation and the ensuing negative immunoreac- the possible contaminating cell types are cytokeratin tivity of RPE is responsible for the scarcity of reports of negative (glia, Muller cells, fibrocytes, melanocytes, cytokeratin immunoreactivity of RPE. endothelial cells and pericytes) (Lazarides, 1980). The RPE showed co-expression of vimentin and cyto- Taking into account the distribution of the various keratin both in situ and in culture. The co-expression of cytokeratins within epithelia as described by Moll et al. vimentin and cytokeratins has already been documented (1982) and the immunocreativity of the monoclonal in other cultured epithelia: mesothelium in vivo and in antibodies used in this study, the most likely candidates culture (LaRocca & Rheinvvald, 1984), in for the cytokeratins expressed by the RPE are nos. 5,6,7,

310 N. M. McKechnie et al. 8, 14, 15, 16, 17, 18 and 19. Bands corresponding to allowed to reach before being fixed and immunostained. cytokeratins 5, 6, 7 and 8 were detected by the two It was expected that as the cultures became confluent and antisera and the MAbs KG8.13 and AE3. Cytokeratins 14 division rates slowed the ratio of cytokeratin 18 positive to and 15 were probably detected by the two antisera and negative cells would decrease. The ratios did drop in the MAb AE3. Cytokeratin 16 may have been detected by three cultures with increasing cell density but remained the MAb. AE1, but MAb KG8.12 (reactive with cyto- fairly constant in one. A possible explanation of this keratins 13 and 16) did not confirm this result. Cytokera- pattern of cytokeratin 18 immunoreactivity is a transient tin 17 may be present but none of the antibodies used in unmasking of a determinant on the cytokeratin 18 pep- this study is thought to show specific reactivity with this tide, at some point during each cycle of division, similar cytokeratin. Cytokeratin 18 was detected by the MAbs to that already described for PtK.2 cells (Franke et al. KG8.13 and NCL 5D3. It is worth noting that both 1983). However, these observations were complicated by antisera failed to react with cytokeratin 18. This would be the fact that human RPE cells are not completely contact expected, as cytokeratin 18 does not occur in normal inhibited in culture and multi-layering does occur in epidermis. Cytokeratin 19 was detected by both antisera post-confluent cultures. Similar behaviour is often ob- and the MAbs NCL 5D3, AE1 and KG4.62. The served in pathological material (Grierson et al. 1987). In immunoreactivity of the RPE with antibodies to cyto- the cultures where the RPE cells had reached high- keratins suggests a pattern of cytokeratin expression density multi-layering was observed. In these areas similar to other simple or glandular epithelia, particularly cytokeratin 18 positive cells were often identified in the the immunoreactivity to cytokeratins 7, 8, 18 and 19 upper layer of cells. The underlying cells, presumably (Moll et al. 1982). The relatively faint immunoreactivity non-proliferating, were unstained. This suggests that of some RPE cells with antisera to epidermal cytokeratins cytokeratin 18 immunoreactivity may be associated with is also typical of simple epithelia (Sun et al. 1979). actively proliferating cells. Some weight is given to this In an attempt to reconcile our findings with those of hypothesis by the finding of cytokeratin 18 positive cells others studying cultured chick RPE (Docherty et al. in both of the epiretinal membranes investigated and in 1984) we have conducted immunostaining of cultured the primary cultures where cytokeratin 18 immunoreac- chick RPE and frozen sections of chick eyes (not reported tivity was associated with the appearance of proliferating here). To date we have been unable to demonstrate cells. cytokeratin staining in chick RPE in vitro or in vivo. In Taken together, the results of the cell confluence this respect chick RPE differs markedly from human experiments and the immunofluorescence staining of RPE. primary cultures suggest that cytokeratin 18 immuno- The pattern of reactivity with the monoclonal antibody reactivity may be useful as a marker of proliferating RPE. to cytokeratin 18 was rather complex. Using immuno- We suspect that RPE cells in late G2, M and Gj stain fluorescence methods, normal RPE was thought to be intensely for cytokeratin 18. In vivo, where RPE cells are cytokeratin 18 negative or faintly positive. Non-dividing presumably in Go, cytokeratin 18 staining, though de- cells in early primary culture, identifiable by their large tectable, is markedly less intense than that seen in culture size and numerous pigmented inclusions (Boulton & or in the pathological material. Marshall, 1985; Flood & Gouras, 1981), appeared to be In conclusion, we have demonstrated that in common cytokeratin 18 negative. In areas where cells appeared to with some other epithelia, the human RPE shows co- be proliferating, identifiable by their smaller size and loss expression of vimentin and cytokeratins both in vivo and of pigmented inclusions (Boulton & Marshall, 1985; in vitro and that cytokeratin immunoreactivity is a Flood & Gouras, 1981), cytokeratin 18 positive cells were reliable marker of human RPE cells. From the cytokera- identifiable. The finding of cytokeratin 18 positive cells in tin 18 positive immunoreactivity of cultured and patho- normal RPE by the immuno-gold/silver technique and logical RPE cells we suggest that the expression of not immunofluorescence may be due to increased sensi- cytokeratin 18 immunoreactivity in human RPE is associ- tivity of the method. ated with cell proliferation in vivo and in vitro. These Using the monoclonal antibody KG8.13, Franke et al. findings may be of use in histopathology and contribute (1983) have described positive staining of keratin fibres of to our understanding of the behaviour of pathological and mitotic PtK2 cells. Interphase cells were not stained, due cultured RPE. to masking of the antigenic determinant. In our hands K 8.13 stained all RPE cells to some degree, some cells The authors thank Pat Goodwin for secretarial assistance, G and also Meryl Bayly and Gursaren Sura for their excellent were intensely stained and had a similar filament pattern technical assistance. One of us (N.M.M.) thanks Drs M. Healy, to that obtained using RGE 53 (specific for cytokeratin B. Creamer and S. L. Grainger for their support and encour- 18). As the determinant recognized by KG8.13 occurs on agement throughout the course of this investigation. the cytokeratin 18 polypeptide, this similarity of staining M. Boulton and N. M. McKechnie are Fight for Sight is not surprising. It is conceivable that the association we Fellows. This work was supported by grants from Fight for have noted between cytokeratin 18 expression and pro- Sight, The Wellcome Trust, and the British Retinitis Pigmen- liferation may be due to a similar 'masking' to that already tosa Society. described in interphase PtK2 cells. To investigate cytokeratin 18 expression further in References proliferating cultures, RPE cultures were established, which differed only in the degree of confluence they were ANGUS, B., PARVIS, J., STOCK, D., WESTLY, B. R., SAMSON, A. C.

RPE cytoskeletons 311 R., ROUTLEDGE, E. G., CARPENTER, F. H. & HOME, C. H. W. 701-704. (1987). NCL-5D3: A new monoclonal antibody recognising low LAROCCA, P. J. & RHEINWALD, J. G. (1984). Coexpression of simple molecular weight cytokeratins effective for epithelial keratins and vimentin by human mesothelium and using fixed paraffin-embedded tissue. J. Path. 153, 377-384. mesothelioma in vivo and in culture. Cancer Res. 44, 2991-2999. BATTIFORA, H. & KOPINSKI, M. I. (1985). Distinction of LAZARIDES, E. (1980). Intermediate filaments as mechanical mesothelioma from adenocarcinoma. An Immunohistochemical integrators of cellular space. Nature, Loud. 283, 249-256. approach. Cancer 55, 1679-1685. MAKAI, M., TORIKATA, C, IRI, H., HANAOKA, H., KAWAI, T., BOULTON, M. & MARSHALL, J. (1985). Repigmentation of human YAKUMARU, K., SHIMODA, T., MIKATA, A. & KAGEYAMA, K. retinal pigment epithelial cells in vitro. Expl Eye Res. 41, 209-218. (1985). of epithelioid . An BOULTON, M. E., MARSHALL, J. & MELLERIO, J. (1983). Retinitis enzyme-histochemical, and immunohistochemical study. Am. J. pigmentosa: A preliminary report on tissue culture studies of Path. 119, 44-56. retinal pigment epithelial cells from eight affected human eyes. MOLL, R., FRANKE, W. W. & SCHILIER, D. L. (1982). The catalog of Expl Eye Res. 37, 307-314. human cytokeratins: Patterns of expression in normal epithelia, BROERS, J. L. V., CARNEY, D. N., ROT, M. K., SCHAART, G., tumours and cultured cells. Cell 31, 11-24. LANE, E. B., VERIJS, G. P. & RAMAEKERS, F. C. S. (1986). OSBORN, M. & WEBER, K. (1983). of disease. Tumor proteins in classic and varient type of small diagnosis by intermediate filament typing: A novel tool for surgical cell lung cell lines: A biochemical and immunochemical pathology. Lab. Invest. 48, 372-394. analysis using a panel of monoclonal and polyclonal antibodies. J. PINKUS, G. S., O'CONNER, E. M., ETHERIDGE, C. L. & CORSON, J. Cell Sci. 83, 37-60. M. (1985). Optimal immunoreactivity of keratin proteins in DOCHERTY, R. J., EDWARDS, J. G., GARROD, D. R. & MATTEY, D. I. formalin-fixed paraffin embedded tissue requires preliminary (1984). Chick embryonic piginented retina is one of the group of trypsinization. J'. Histochem. Cytochetn. 33, 465-473. epithelioid tissues that lack cytokeratins and desmosomes and have RAMAEKERS, F. C. S., PUTS, J. J. G., MOESKER, 0., KANT, A., intermediate filaments composed of vimentin. J. Cell Sci. 71, HUYSMANS, A., HAAG, D., JAP, P. H. K., HERMAN, C. J. & 61-74. Voous, G. P. (1983a). Antibodies to intermediate filament FLOOD, M. T. & GOURAS, P. (1981). The organization of human proteins in the immunohistochemical identification of human retinal pigment epithelium in vitro. Vision Res. 21, 119-126. tumours: an over view. Histochem. J. 15, 691-713. FRANKE, W. W., SCHMID, E. A., BREITKREUTZ, D., LUDER, M., RAMAEKERS, F., HUYSMANS, A., MOESKER, O., KANT, A., JAP, P., BOUKAMP, P., FuSENIG, N. E., OsBORN, M. & WEBER, K. HERMAN, C. & Voous, P. (19836). Monoclonal antibody to keratin (1979a). Simultaneous expression of two different types of filaments specific for glandular epithelia and their tumours: Use in intermediate sized filaments in mouse keratinocytes proliferating in surgical pathology. Lab. Invest. 49, 353-361. vitro. Differentiation 14, 35-50. RAMAEKERS, F., PUTS, J., MOESKER, O., KANT, A., JAP, P. & FRANKE, W. W., SCHMID, E., WEBER, K. & OSBORN, M. (19796). Voous, P. (1983c). Demonstration of keratin in human HeLa cells contain intermediate-sized filaments of the prekeratin adenocarcinomas. Am. J. Path. Ill, 213-223. type. Expl Cell Res. 118, 95-109. SCHLEGEL, R., BANKS-SCHLEGEL, S. & PINKUS, G. S. (1980). FRANKE, W. W., SCHMID, E., WELLSTEED, J., GRUND, C, GIGI, C, Immunohistochemical localization of keratin in normal human GIGO, O. & GEIGER, B. (1983). Change of cytokeratin filament tissues. Lab. Invest. 42, 91-96. organization during the cell cycle: selective masking of an STEINERT, P. M. (1975). The extraction and characterization of immunologic determinant in interphase PtK2 cells. J. Cell Biol. 97, bovine

312 N. M. McKechnie et al.