3025Journal of Cell Science 109, 3025-3034 (1996)Printed in Great Britain © The Company of Biologists Limited 1996JCS1290

The polarity of the plasma membrane protein RET-PE2 in retinal pigment

epithelium is developmentally regulated

Alan D. Marmorstein1,2, Vera L. Bonilha1,2,3, Silvia Chiflet4,5, James M. Neill6 and Enrique Rodriguez-Boulan1,2,*1Department of Cell Biology and Anatomy, Cornell University Medical College, 1300 York Avenue, New York, NY 10021, USA2Margaret M. Dyson Vision Research Institute, Department of Ophthalmology, Cornell University Medical College, 1300 YorkAvenue, New York, NY 10021, USA3Instituto De Biofisica Carlos Chagas Filho, Universidade Federal Do Rio De Janeiro, 21949-000, Ilma Do Fundao, Rio DeJaneiro, Brasil4Departmento de Bioquimica, Facultad De Medicina, Gral. Flores 2125, 11800 Montevideo, Uruguay5Seccion Biología Celular, Facultad De Ciencias, Tristán Narvaja 1674, 11200 Montevideo, Uruguay6Department of Anatomy and Cell Biology, SUNY Health Science Center at Brooklyn, 450 Clarkson Avenue, Brooklyn, NY 11203,USA

*Author for correspondence at address 2

The retinal pigment epithelium (RPE) differs from otherepithelia in that the apical surface is not free; instead, itinteracts with both photoreceptors and a specialized extra-cellular material, the interphotoreceptor matrix. Biochem-ical characterization of the apical and basolateral surfacesof RPE in adult rat eye cups, using a novel in situ biotiny-lation assay, revealed very different protein compositionsand identified a major surface antigen, RET-PE2, with apredominantly apical distribution (~74%). The apicalpolarity of RET-PE2 was confirmed by immunofluores-cence and laser scanning confocal microscopy. In strikingcontrast, RET-PE2 antigen was preferentially basolateralin primary cultures derived from adult rat RPE and in animmortalized RPE cell line (RPE-J). Under all conditions,RET-PE2 was highly soluble in Triton X-100 (>81% at

4°C), suggesting that its redistribution was not dependenton changes in cytoskeletal interactions. Analysis of thelocalization of RET-PE2 in normal rats at postnatal (PN)days 1, 7, and 14 indicated that RET-PE2 redistributesfrom predominantly basolateral to predominantly apicalduring that time. Since photoreceptors develop during thefirst two weeks after birth in the rat, our results suggestthat the apical redistribution of RET-PE2 is dependent onthe establishment of adult interactions between the RPEand the neural retina and/or the interphotoreceptormatrix, either via direct contacts or through alterations inthe intracellular sorting patterns of RPE cells.

Key words: Epithelium, Interphotoreceptor matrix, Biotin-avidin

SUMMARY

INTRODUCTION

The retinal pigment epithelium (RPE), a highly specializedderivative of the embryonic neural tube (Zinn and Marmor,1979), combines the barrier functions of a simple transportingepithelium with the support role of glia and the phagocyticactivity of macrophages. The RPE acts as a selective perme-ability barrier between the choroidal blood vessels and the sub-retinal space and performs a variety of essential supportfunctions for the neural retina, such as the uptake, transport,and processing of vitamin A and the phagocytosis anddigestion of photoreceptor outer segments (Bok, 1993). Theessential support functions of RPE require a characteristic dis-tribution of some surface proteins. N-CAM-140, an isoform ofthe calcium independent neural cell adhesion molecule thatlocalizes basolaterally in the kidney cell line MDCK (Powellet al., 1991), is observed on the apical surface of RPE in young(one-week-old) rats (Gundersen et al., 1993). Unlike most

other epithelia, RPE cells display Na+,K+-ATPase on theirapical surface (Bok, 1993; Caldwell and McLaughlin, 1984;Gundersen et al., 1991; Okami et al., 1990; Rizzolo, 1990).

It is likely that the reversed polarity of certain apical RPEproteins depends on their unique apical microenvironment.Indeed, N-CAM-140 changes its distribution to basolateral inprimary cultures of RPE (Gundersen et al., 1993). Furthermore,Na+,K+-ATPase in rats becomes nonpolar under the sameexperimental conditions (Marrs et al., 1995; Nabi et al., 1993).It has been shown that interaction with the basement (Bruch’s)membrane alone is insufficient to induce apical polarity ofNa+,K+-ATPase in cultured chicken RPE (Rizzolo, 1991). Thedistribution of Na+,K+-ATPase in a given cell may be deter-mined by two mechanisms acting alone or in combination:intracellular sorting into defined post-Golgi vesicles orselective stabilization by a domain-specific ankyrin-fodrin sub-membrane cytoskeleton (Caplan et al., 1986; Gottardi andCaplan, 1993; Hammerton et al., 1991; Zurzolo and

3026 A. D. Marmorstein and others

Rodriguez-Boulan, 1993). In MDCK cells, Na+,K+-ATPase isstabilized by a lateral ankyrin-fodrin cytoskeleton (Hammertonet al., 1991), probably induced by E-cadherin-mediated inter-cellular adhesion (Nelson, 1991; Marrs et al., 1995). Incontrast, the RPE in the eye expresses both ankyrin-fodrin andNa+,K+-ATPase on the apical surface, probably as a result ofthe absence of E-cadherin in RPE cells (Gundersen et al., 1991,1993; Rizzolo, 1990), which express cadherins of the N- or P-type (Gundersen et al., 1993; Lagunowich and Grunwald,1989; Marrs et al., 1995). Recent results suggest that ankyrinbut not fodrin codistribute with Na+,K+-ATPase in the apicalmicrovilli of avian RPE cells, implying that only ankyrindevelops functional interactions with the enzyme (Rizzolo andZhou, 1995).

In order to understand the mechanisms responsible for theunique polarity properties of the apical surface of RPE, it iscrucial to identify components of that membrane that mayinteract with the interphotoreceptor matrix (IPM) or the pho-toreceptor outer segments. Here, we use a novel adaptation ofthe biotin polarity assay previously developed in our labora-tory (Sargiacomo et al., 1989) to identify apical and basolat-eral surface proteins from RPE in situ. We show that the RET-PE2 antigen (Neill and Barnstable, 1990) is a major componentof the apical membrane of RPE in situ, but undergoes a reversalof its polarity from apical to basolateral in primary cell culture.Developmental studies indicate that this marker polarizes to theapical surface exactly at the time when photoreceptor outersegments and the IPM mature. The control of RET-PE2 dis-tribution by the maturation of photoreceptors constitutes anovel mechanism of apical polarization that depends on extra-cellular cues.

MATERIALS AND METHODS

ReagentsAll cell culture reagents were obtained from Gibco/BRL Life Tech-nologies (Grand Island, NY). All other reagents were obtained fromSigma Chemical Co. (St Louis, MO) unless otherwise indicated.

Animals and primary cell culturesMale and female Long-Evans rats of various ages were obtained fromCharles River (Wilmington, MA). Animals were kept on 12 hourlight/dark cycles and fed ad libitum.

Primary cultures of RPE were established from adult Long-Evansrats as follows: animals were sacrificed by CO2 asphyxiation, and theeyes enucleated and placed in 10 mM Hepes buffered Hank’sbalanced salt solution (HBSS) at 4°C in the dark for 4 hours. A cir-cumferential incission was made below the ora serrata and anteriorsegments and neural retinae were removed. After digestion for 1 hourwith trypsin (Difco, Detroit, MH) in HBSS (2 mg/ml) at 37°C, RPEwere teased from the eyecup with a 25 G needle. RPE cells werefurther dissociated in trypsin/EDTA, and plated on Matrigel® (Col-laborative Research, Bedford, MA) coated Transwell® filters(Corning-Costar Corp., Cambridge, MA) (2 eyes/1.2 cm diameterfilter) in DMEM containing 20% fetal bovine serum (FBS),glutamine, penicillin/streptomycin, and non-essential amino acids.Cells were kept at 37°C for 4-6 weeks without subculturing prior touse. Primary RPE cultures acquired transepithelial resistances of ~60-80 Ω × cm2.

RPE-J cell cultureThe SV-40 immortalized RPE-J cell line was maintained as previ-

ously described (Nabi et al., 1993) in Dulbecco’s minimum essentialmedium (DMEM) containing 4% Cellect Gold FBS (ICN, CostaMesa, CA), supplemented with glutamine, non-essential amino acidsand penicillin/streptomycin at 32°C. Cells were passaged withtrypsin/EDTA. For biochemical studies RPE-J cells were plated onMatrigel® coated Transwell filters® at a density of 3.5×105 cells/cm2.Cells were grown for 6-7 days at 32°C and then allowed to differen-tiate for 2 days at 39.5°C in medium supplemented with 10−8 Mretinoic acid. RPE-J cells grown in this fashion acquired transepi-thelial resistances in excess of 200 Ω × cm2.

Preparation of eyecups and isolation of RPEEyes were enucleated and a circumferential incision was made abovethe ora serrata. The cornea, lens, iris, and vitreous body wereremoved. To remove the neural retina without damaging the apicalsurface of the RPE we used a modification of the protocol of Wanget al. (1993). Eyecups were incubated in 10 ml of HBSS containing290 U/ml of bovine testicular hyaluronidase for 10-30 minutes at37°C. A second incision was then made below the ora serrata and theneural retina was carefully peeled away and cut at the optic nerve. Fordetergent solubility studies, RPE was isolated from the eyecups afterdigestion with hyaluronidase and collagenase, as described by Wanget al. (1993).

Immunofluorescence and confocal microscopyEyecups or RPE-J monolayers were fixed for 30 minutes in 2%paraformaldehyde in phosphate buffered saline containing 0.1 mMCaCl2, 1.0 mM MgCl2 (PBS/CM), and quenched with 50 mM NH4Clin PBS/CM. When necessary, RPE-J monolayers were permeabilizedin absolute methanol for 4 minutes at −20°C. For cryosectioning,eyecups were infiltrated successively with 10% and 20% sucrose, andthen with Tissuetek 4583 (Miles Inc., Elkhart, IN). When it wasdesirable to maintain the RPE-neural retina interaction, rats werekilled by CO2 asphyxiation, and subject to intracardiac perfusion withHBSS followed by 4% paraformaldehyde in PBS/CM. Perfusion fixedeyes were enucleated, the corneas incised, and then further immersionfixed overnight. Following immersion fixation, the eyes were infusedwith 30% sucrose in PBS/CM, and then with Tissuetek 4583. Afterfreezing in liquid nitrogen/isopentane, 10 µm cryosections were cutand the tissue allowed to adhere to SuperFrost slides (Fisher, Spring-field, NJ). After blocking in PBS/CM containing 0.2% BSA(PBS/CM/BSA), monolayers or sections were stained with a rabbitpolyclonal antisera raised against rat laminin (Gibco) and/or mousemonoclonal antibody RET-PE2 for 1 hour and then with FITC-con-jugated goat anti-mouse and/or Texas red-conjugated donkey anti-rabbit IgG secondary antibodies (Cappel, Durham, NC) inPBS/CM/BSA containing a 1:200 dilution of DNase free RNase I(Boehringer Mannheim, Indianapolis, IN) for 1 hour. The cells werewashed and in some cases stained with propidium iodide as describedpreviously (Hanzel et al., 1991). After a final wash, filters wereexcised and mounted in vectashield (Vector Labs, Burlingame, CA).Whole eyecup preparations were cut radially and mounted en face.Labeled cells were visualized with a dual channel laser scanningconfocal microscope (Sarastro/Molecular Dynamics, Sunnyvale, CA)as previously described (Hanzel et al., 1991).

Domain specific biotinylation of RPE-J and RPE in situThe steady state distribution of RET-PE2 antigen in RPE-J cells wasdetermined using a domain specific biotinylation assay as previouslydescribed (Hanzel et al., 1991; Zurzolo et al., 1993). To examine theRET-PE2 distribution in RPE in situ, we used a modification of apreviously described quenched biotinylation assay (Lisanti et al.,1990). The assay was performed as follows: adult Long-Evans ratswere euthanized by CO2 asphyxiation and the eyes enucleated. Theeyes were washed in ice-cold PBS/CM, eyecups prepared asdescribed above, washed in PBS/CM at 4°C, filled with a 2 mg/mlsolution of either NHS-LC-biotin (sulfosuccinimidyl-6-(bioti-

3027Regulated polarity of RET-PE2 antigen

bb

b

xxxxxx

x x x x x

ApicalLabeling

BasalLabeling

1

2

3

4

5Apical proteinslabeled with biotin

Basal proteinslabeled with biotin

x

xx x

b b b b b b

b b b b b b

Fig. 1. In situ biotinylation of RPE. Eyes are enucleated from adultLong-Evans rats (1) and the anterior segments and neural retinadissected away. An aliquot of either NHS-LC-biotin (b) or NHS-S-S-biotin (x) is placed within the eyecup (2) to either label (b) or quenchlabel (x) the apical membrane. After the biotinylation reaction,eyecups are incubated at 4°C (metabolic block) for 20-30 minutes inPBS containing 1 mM EDTA, and the RPE monolayer is dissectedout of the eyecup (3). The apically quenched RPE monolayer is thenbiotinylated in suspension with NHS-LC-biotin (4). After thereaction is completed the cells are lysed in detergent (5).

namido)-hexanoate; Pierce, Rockford, IL) or the disulfide cleavableNHS-S-S-biotin sulfosuccinimidyl-2-(biotinamido)-ethyl-1,3-dithio-proprionate; Pierce), and incubated for 20 minutes at 4°C. Thelabeling reaction was repeated 3 times, and the biotinylation reactionquenched in PBS containing 50 mM NH4Cl (10 minutes at 4°C).Eyecups were then incubated in PBS containing 1 mM EDTA for 30-60 minutes at 4°C, the RPE was dissected from the eyecup, placedin a 1.5 ml Eppendorf tube containing ice-cold PBS, pelleted for 10seconds in a microfuge, resuspended in 1 ml PBS containing 0.5mg/ml NHS-LC biotin and incubated at 4°C for 20-30 minutes. Thelabeling reaction was repeated, and the reaction was quenched for 10minutes with 1 ml PBS, 50 mM NH4Cl at 4°C for 10 minutes.Biotinylated RPE was lysed with 1% Triton X-100 in 50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 0.2% BSA, pH 7.4 (TBSE), con-taining protease inhibitors (1 mM PMSF, 0.5 mM aprotinin, 0.5 mMleupeptin, 0.5 mM antipain) for 1 hour at 4°C. The lysate was cen-trifuged at 10,000 g for 10 minutes, and the supernatant prepared forSDS-PAGE.

Immunoprecipitation, SDS-PAGE and streptavidin blottingAfter preclearing samples with Pansorbine (Calbiochem, San Diego,CA), RET-PE2 monoclonal antibody in the form of ascites fluid wasadded directly to cell lysates (1-3 µl/ml). After overnight incubation,the lysates were supplemented with 10 mg Protein-A Sepharoseprebound to 10 µg rabbit anti-mouse antibodies, incubated for 1 hourat 4°C, and the beads pelleted and washed as previously described (LeBivic et al., 1989). Immunoprecipitates were resuspended in 40 µl ofSDS-PAGE sample buffer containing 0.5 M 2-mercaptoethanol,heated to 95°C for 5 minutes, and analyzed by SDS-PAGE on dis-continuous 12% gels.

Gels were transferred to Immobilon-P membranes (Millipore,Bedford, MA) overnight, blocked in TBSE containing 10% dry milkfor 1 hour, and incubated with either [125I]streptavidin, or streptavidinconjugated to horseradish peroxidase, in 1% milk, in TBSE for 1 hour.After 3 washes in TBSE containing 0.1% Tween-20, the blots werevisualized by autoradiography or enhanced chemiluminescence (ECL,Amersham, Arlington Heights, IL) on Kodak X-OMAT AR film andan enhancing screen. To quantify results, gels were scanned and theintensity of bands determined using NIH Image 1.52 software.

Detergent solubilityFreshly isolated RPE or RPE-J cells were surface biotinylated andthen lysed in 1% Triton X-100 in 50 mM Tris-HCl, pH 7.4, con-taining 150 mM NaCl, 0.2% BSA and protease inhibitors, as above.In some samples, 5 mM EDTA was added to eliminate potentialeffects of divalent cations on interactions with the cytoskeleton. Cellswere lysed for 20 minutes at 4°C, and insoluble material wasremoved by centrifugation at 10,000 g for 10 minutes at 4°C.Insoluble pellets were heated for 5 minutes at 95°C in 1% SDS, 50mM Tris, pH 7.4, 150 mM NaCl, 0.2% BSA. The amount ofdetergent in each sample was equalized by the addition of SDS orTriton X-100 to final concentrations of 0.1, and 1%, respectively.RET-PE2 antigen was immunoprecipitated and the amount of antigendetermined from streptavidin blots of the immunoprecipitates asdescribed above.

Sorting of lipidsSorting of lipids was determined as previously described (van Meer,1993; van Meer and van ’t Hof, 1993). Differentiated monolayers ofRPE-J cells were allowed to incorporate C6-NBD-ceramide in HBSSfor 30 minutes at 4°C. Delivery to the cell surface of newly synthe-sized C6-NBD-glucosylceramide (C6-NBD-glcCer, a glycosphin-golipid) and C6-NBD-sphingomyelin (C6-NBD-SM, a phospholipid)were assayed by selective depletion from either the apical or basolat-eral surfaces with fatty acid free bovine serum albumin in HBSS. Flu-orescent lipids were separated by two-dimensional TLC, and quanti-fied by spectrofluorimetry.

RESULTS

Characterization of the apical surface of RPE in situGiven that two RPE proteins, Na+,K+-ATPase and N-CAM,have been shown to change their surface distribution uponremoval of RPE from the eye and culture in vitro (Gundersenet al., 1991; Rizzolo, 1990), our initial goal in this study wasto identify apical proteins in RPE in its normal eye localiza-tion. To this end, we adapted our biotinylation assay to the eyecup (Fig. 1). The RPE is conveniently fit for this type of studysince it lays as a continuous monolayer with its concave apicalsurface in apposition to the neural retina and the interphoto-receptor matrix. The neural retina can be easily peeled off afterdigestion with bovine testicular hyaluronidase, with minimaldamage to the RPE. To label only the apical surface of the RPE,the neural retina was removed, soluble components of the IPMwashed away, and the exposed apical surface labeled withNHS-LC biotin. Examination by laser scanning confocalmicroscopy (LSCM) of eye cup whole mounts exposed toFITC-streptavidin in the presence of detergent (Fig. 2A)showed intense fluorescence at the apical surface and no flu-orescence at the basolateral surface or the underlying choroid.For basolateral staining, apical sites were blocked with acleavable biotin derivative, NHS-S-S-biotin (Fig. 2B), baso-lateral sites were exposed by treatment with 1 mM EDTA, andlabeled with the uncleavable NHS-LC-biotin (Fig. 1). Afterremoval of RPE from the underying choroid, biotin is removed

3028 A. D. Marmorstein and others

Fig. 2. Biotinylation of intact eye cups labels only the apical surfaceof RPE. The apical surface of RPE in eyecups was reacted at 4°Cwith: (A) NHS-LC-biotin (4× 20 minutes); (B) NHS-S-S-biotin (4×20 minutes) or (C) with NHS-S-S-biotin (4× 20 minutes), followedby NHS-LC-biotin (2× 20 minutes) and reduction with 50 mM DTT.After fixation, eyecups were permeabilized and stained with FITC-labeled streptavidin (green) to reveal biotinylated sites, and nucleiwere stained with propidium iodide (red). Tissue was then examinedby LSCM in the x-z plane. Reaction with NHS-S-S-biotin labeled theapical surface (B) sufficient to block subsequent labelling with NHS-LC-biotin (C). Bars, 5 µm.

Fig. 3. Apically and basolaterally biotinylated proteins in RPE insitu. Apical and basolateral protein profiles of 2-month-old rat RPEwere generated by in situ domain specific biotinylation. Afterlabeling RPE from either the apical (Ap) or the basolateral (Bl) side,the cells were lysed, and the cell surface proteins resolved by SDS-PAGE, transfered to PVDF, and blotted with horseradish peroxidase-conjugated streptavidin. Labeled proteins were visualized by ECL.Note the different protein patterns on the apical and basolateralsurfaces (arrowheads indicate polarized proteins). At the apicalsurface, proteins of 36, 50, 60 and 100-120 kDa are particularlyprominent. Size standards (in kDa) are shown at the left.

from the apical proteins by reduction with the 2-mercap-toethanol in the SDS-PAGE sample buffer. After thisprocedure the only remaining biotinylated proteins are baso-lateral. Control eye cup experiments in which the apicalsurface was blocked with NHS-S-S-biotin, and subsequentlyexposed to NHS-LC-biotin, indicated that the apical surface isadequately blocked by this procedure (Fig. 2C).

Examination of [125I]streptavidin blots of RPE monolayerslabeled by this procedure indicated that the apical and baso-lateral protein patterns were quite different (Fig. 3). Proteinswith relative molecular masses of ~36, ~50, and ~60, and 100-120 kDa were apical, whereas proteins of ~55, and ~63 kDawere distinctly basolateral.

Identification of an apical protein as the RET-PE2antigenFew antibodies against RPE surface proteins are currentlyavailable. One of these is the antibody RET-PE2 which rec-ognizes a 50-55 kDa antigen that is found exclusively in ratRPE (Neill and Barnstable et al., 1990). Examination of strep-tavidin precipitated proteins by immunoblot with the RET-PE2antibody of RPE monolayers labeled according to the protocoldescribed in Fig. 1 detected a predominantly apical proteinband with the apparent molecular mass of RET-PE2, 50-55kDa (Fig. 4A). Quantitation of the polarized distribution of thisprotein by immunoprecipitation and [125I]streptavidin blotting(Fig. 4B) indicated a 3:1 apical to basolateral ratio. Examina-tion of 10 µm frozen sections of the retina, double stained for

RET-PE2 and and laminin (a component of the Bruch’sbasement membrane) (Fig. 5B,C; C was obtained by LSCM)confirmed that RET-PE2 is predominantly localized to theapical membrane.

RET-PE2 antigen is basolateral in primary RPEcultures and in RPE-JTo determine whether the apical polarity of RET-PE2 is main-tained or changed when RPE is removed from the eye microen-vironment, we examined the steady-state distribution of thisantigen in primary cultures of adult RPE and in RPE-J, animmortalized rat RPE cell line, using the biotin polarity assay(Fig. 6). Strikingly, the RET-PE2 antigen was found to haveredistributed to a preferentially basolateral distribution in bothcases (Fig. 6A). Quantitative analysis indicated that ~90% ofRET-PE2 in RPE-J was basolateral under these conditions(Fig. 6B). LSCM and immunofluorescence analysis confirmedthe basolateral distribution of RET-PE2 in RPE-J monolayers(Fig. 6C) and in primary RPE cultures (data not shown). Inter-estingly, the staining was confined to the lateral membrane ofunpermeabilized cells, with the basal surface excluded; part ofthe antigen was detected in structures with a vesicular appear-ance. These structures are presumably blebs in the lateralmembrane which are visible by TEM (unpublished observa-tion). The RET-PE2 staining was lost if the paraformaldehydefixed cells were permeabilized with either 0.075% saponin or0.1% Triton X-100 (data not shown). Methanol permeabiliza-tion at −20°C did not disrupt the RET-PE2 staining; therefore,

3029Regulated polarity of RET-PE2 antigen

Fig. 4. Apical distribution of RET-PE2 antigen determined by in situbiotinylation. The polarity of RET-PE2 antigen was determined by insitu domain specific biotinylation of eyecups prepared from 2-month-old Long-Evans rats followed by immunoprecipitation andblotting with [125I]streptavidin (A). The percentage of RET-PE2antigen present on the apical (Ap) and basolateral (Bl) surfaces isshown in B. Results are expressed as mean ± s.d. (n=3).

this fixative was used when nuclei were stained with propidiumiodide to provide a positional marker, or to facilitate difusionof the antibody to the basolateral side in whole mount prepa-rations of RPE.

The polarity of RET-PE2 antigen changes duringpost-natal developmentIn rats and mice, photoreceptors are without morphologicallydistinct inner or outer segments until after birth. Innersegments begin to form during the first week of postnataldevelopment. While the majority of outer segment develop-ment occurs during the second postnatal week, outer segmentdisks can be detected as early as PN5 (Olney, 1968). Similarlyapical processes are nearly absent on RPE cells until the secondweek of postnatal development, becoming very prominent byPN12 (Dowling and Gibbons, 1962; Micali et al., 1989).Previous work has indicated that the expression of RET-PE2antigen is widespread in the embryonic rat eye (Neill and Barn-

Fig. 5. Apical polarity of RET-PE2, determined by immunofluorescencewas analyzed by immunofluorescence of frozen sections from eyes fixedphase and fluorescence images in A and B were photographed with a fluseparation between laminin staining at the basement membrane (red, BMONL, outer nuclear layer; OS, outer segments; RPE, retinal pigment epi

stable, 1990) but becomes confined to the RPE as early as PN9(Neill et al., 1993). To determine whether the apical polarityof RET-PE2 was dependent on the maturation of the retina, weexamined its distribution by LSCM in 10 µm frozen sectionsand in whole mounts of RPE obtained from rats at PN1, PN7,PN14, and from 2- to 3-month-old rats (Figs 7 and 8).

Whole mounts stained with RET-PE2 and propidium iodide(to label nuclei) were examined using LSCM. Section serieswere generated from the first (apical most) point at which anystaining could be detected until both signals became weak(approx. 6 µm in depth). RET-PE2 staining in adult animals(Fig. 7A and C) was strongest in the first section and grew pro-gressively weaker while the propidium iodide stained nucleiincreased in strength before dropping off. This was evident inplots of the percentage of maximum mean pixel intensity forboth channels (Fig. 7C and D) and was in contrast to wholemounts of PN1 eyes (Fig. 7B and D). At PN1 RET-PE2staining took on a distinct honeycomb appearance similar tothat observed in monolayers of RPE-J cells. The RET-PE2staining became more intense as the plane of section movedtoward the basal surface, achieving a maximum intensity at thesame level as the propidium iodide (Fig. 7D). In frozensections (data not shown) obtained from PN1 rats, RET-PE2immunostaining was detected in both the RPE and underlyingchoroid, with the strongest signal along the lateral borders ofthe RPE and the Bruch’s membrane. A similar lateral stainingwas observed by examination of whole mounts (Figs 7B,D,8A,B); however, little staining was observed along the basalplasma membrane or in the choroid (Figs 7B, 8B). It is unclearwhether this is due to incomplete accessibility of the antibodiesto the basal side after methanol fixation, or to interference bythe abundant RPE pigment granules.

Examination of frozen sections, and x-y (Fig. 8C) and x-z(Fig. 8D) scans of whole mount eyecups at PN7 revealed anonpolar distribution of RET-PE2 staining. By PN14, however(Fig. 8E and F), the polarity of PE2 was apical in most cells.Biochemical analysis of polarity by in situ biotinylation ofeyecups was found to be technically impossible because oftheir small size at PN1-14. These results indicate that the polar-ization of RET-PE2 to the apical surface follows closely thedevelopment of rod outer segments, switching from basolateralat PN1 when outer segments are not yet formed, to fully apicalat PN14 when outer segments have reached maturity.

and confocal microscopy. The distribution of the RET-PE2 antigen by intracardiac perfusion with 4% paraformaldehyde (A,B,C). Theorescence microscope. C is a confocal x-y scan. Note in B and C the) and RET-PE2 staining (green). Ap, apical; BM, basal membrane;

thelium; Ch, choroid. Bars: 20 µm (A,B), 5 µm (C).

3030 A. D. Marmorstein and others

Fig. 6. Steady state distribution of RET-PE2 antigenin RPE-J monolayers. RPE-J monolayers cultured onTranswell filters were subjected to domain specificbiotinylation. Streptavidin blots (A) were scanned andthe percentage of RET-PE2 antigen present on theapical (Ap) and basolateral (Bl) surfaces weretabulated (B). Results in B are expressed as mean ±s.d. (n=8). RET-PE2 antigen was found to bepredominantly basolateral. The basolateral polarity ofRET-PE2 antigen in RPE-J cells was confirmed byimmunofluorescence and LSCM. Cell surface stainingin the x-y plane (C) had a honeycomb appearance andwas confined to the lateral borders of the cells. x-zscans (D) confirm the basolateral localization. Bars, 5µm.

Fig. 7. The polarity of RET-PE2 is similar tocultured RPE at PN1. The distribution of RET-PE2 was compared in whole mounts of adult (A)and PN1 (B) eyes stained with RET-PE2 (green,d) and propidium iodide (red, j). LSCM wasused to generate a section series (A,B). Sectionswere 0.1 µm in depth at 0.5 µm intervals,starting at the apical most point of staining.Mean pixel intensities for RET-PE2 andpropidium iodide were determined and thepercentage maxima graphed for both adult (C)and PN1 (D) eyes. The green and red lines in Cand D represent mean pixel intensities for RET-PE2 and propidium iodide staining, respectively.The numbers in the lower left hand corner of Aand B correspond to the sections shown. RET-PE2 staining is strongest in the first section anddecays relative to the peak of propidium iodidein the adult. In contrast at PN1 RET-PE2staining takes on a distinct honeycombappearance similar to that observed in RPE-Jcells and peaks along with the nuclear stain. Ap,apical; Ba, basal. Bars: 10 µm (A); 20 µm (B).

3031Regulated polarity of RET-PE2 antigen

Fig. 8. Change of polarity of RET-PE2 antigen during postnatal development. The distribution of RET-PE2 was studied using LSCM at variousstages of post-natal (PN) development in whole mount eyecups. (A,B) PN1; (C,D) PN7; (E,F) PN14. x-y scans of the apical most staining areshown in A,C,E, x-z scans are shown in B,D,F. At PN1, RET-PE2 immunostaining was preferentially associated with the basolateral surface ofthe RPE. At PN7, the staining was observed both laterally and apically. At PN14, the RET-PE2 staining was predominantly apical. Pigmentmay have obscured some basal staining that was apparant in x-y scans of frozen crossections (not shown). Bars, 20 µm.

Detergent solubility of RET-PE2 antigen As mentioned before, RET-PE2 was basolateral and highlyextractable by detergent in RPE-J cells, even afterparaformaldehyde fixation. Since the basolateral localizationof Na+,K+-ATPase in kidney cells (Nelson, 1991) and its apicallocalization in RPE cells (Gundersen et al., 1991) are thoughtto depend at least partially on interactions with the submem-brane cytoskeleton, we wished to test whether the polarityreversal of RET-PE2 from basolateral to apical in RPE in vivocorrelated with increased resistance to detergent extractability,suggestive of an interaction with an apical cytoskeleton. To testthis point, we extracted RPE or RPE-J cells with 1% Triton X-100 in the presence or absence of 5 mM EDTA for 20 minutesat 4°C and sedimented insoluble proteins by centrifugation for10 minutes at 10,000 g. RET-PE2 in both Triton X-100 solubleand insoluble fractions was solubilized in SDS and immuno-precipitated after dilution of the SDS. As shown in Fig. 9, lessthan 15% of RET-PE2 antigen was insoluble in detergent inthe presence of EDTA, both in RPE-J cells and in RPE mono-layers obtained from adult rats.

Sorting and delivery of lipidsIt has been suggested that some proteins may be sorted to theapical surface of epithelial cells in detergent insoluble glycol-ipid rafts (van Meer and Simons, 1988), and that exclusionfrom these rafts could be a potential sorting mechanism forNa+,K+-ATPase (Mays et al., 1995). We wished to determineif, in RPE-J cells, there is preferential delivery of lipids to oneplasma membrane domain which would suggest a higher rateof membrane turnover in that surface and a potential sortingmechanism in RPE cells. Lipid sorting assays were performedon RPE-J cells by first preloading with C6-NBD-ceramide, and

then selectively depleting newly synthesized NBD-labelledlipids from either the apical or basal membranes with albumin.This assay provides a measure of lipid flow to the apical vsbasolateral surfaces, as well as a determination of lipid polarityfrom the ratio of a glycosphingolipid (C6-NBD-glucosyl-ceramide) to a phospholipid (C6-NBD-sphingomyelin) trans-ported to either surface (van Meer, 1993, van Meer and van ’tHof, 1993). In RPE-J cells, lipids were not sorted (polarity ratioof 1.34) and both C6-NBD-glucosylceramide and C6-NBD-sphingomyelin were delivered in ~2-fold greater quantities tothe apical plasma membrane of RPE-J cells (2.56±3.9 and1.92±0.3, respectively) indicating that protein-sorting in thesecells is most likely independent of lipid sorting.

DISCUSSION

In this study we set out to identify apical proteins of RPE invivo that shift their polarity when the cells are placed inculture. To accomplish this objective, we designed a domainspecific biotinylation assay that allows the quantitative deter-mination of protein polarity in the RPE in the eye. BiotinylatedRPE eyecups stained with FITC-streptavidin demonstrated thatlabeling is restricted to the apical surface, and that basolaterallabeling can be obtained by first blocking apical sites withNHS-S-S-biotin, followed by cleavage of the S-S bonds. Theeffectiveness of the in situ assay is verified by our finding thatRET-PE2 antigen is also seen apically polarized by LSCM inthe RPE of adult rats (see Fig. 5). Due to a variable extent ofcontamination with the underlying choroid, this assay worksbest for proteins that, like the RET-PE2 antigen, are restrictedto or are much more abundant in the RPE than in the choroid.

3032 A. D. Marmorstein and others

Fig. 9. RET-PE2 is soluble in detergent in RPE in situ and in RPE-Jcells. RPE monolayers from 2-month-old rats or RPE-J monolayerswere surface biotinylated and then lysed at 4°C in the presence orabsence of 5 mM EDTA for 20 minutes. Lysates were centrifuged at10,000 g for 10 minutes, the supernatant (soluble, S) was decantedand the pellet (insoluble, P) resuspended by boiling for 5 minutes in1% SDS. After equalizing the concentrations of SDS and Triton X-100 in S and P fractions, RET-PE2 antigen was immunoprecipitatedand the cell surface antigen visualized by SDS-PAGE andstreptavidin blotting (A). The percentage of soluble RET-PE2antigen was determined by densitometry (B). In all cases >81% ofRET-PE2 antigen was soluble.

The monoclonal antibody RET-PE2 recognizes a ~50-55kDa protein originally described as a rat RPE-specific marker.RET-PE2 is an integral membrane protein that associates withthe detergent phase upon extraction with Triton X-114 and isnot anchored by glycosylphosphatidylinositol (unpublishedobservations). RET-PE2 displayed a predominantly apical dis-tribution in adult RPE in situ but was basolaterally distributedin early postnatal RPE cells, primary cultures of RPE, and inthe RPE-J cell line. A morphometric comparison of apical andbasolateral membrane areas has shown that the apical area inadult RPE cells exceeds the basolateral one by a factor of ~3.8.(Okami et al., 1990) Thus, the apical polarity shift that we haveobserved may be due to the elongation of the apical microvilliduring the second postnatal week, and the ensuing change inthe relative proportions of apical and basolateral membranes.In the case of Na+,K+-ATPase, the enzyme was found to beonly slightly enriched in the apical membrane (0.658 vs 0.440gold particles per linear micron of membrane; Okami et al.,1990). While this observation suggests that extension ofmicrovilli could account for the apical polarity of Na+,K+-ATPase, and potentially RET-PE2, its static nature does noteliminate the possibility of additional mechanisms responsiblefor their polarity. Differential rates of membrane turnover ineach surface, intracellular sorting or trapping at the cell surfaceby the cytoskeleton or interaction with extracellular compo-nents, probably do contribute to the observed steady statepolarities of these proteins.

The redistribution of RET-PE2 from basolateral to apicalduring development of RPE might involve an alteration in theintracellular targeting patterns of these cells. Studies withcultured epithelial cell lines have demonstrated a remarkableflexibility of the epithelial targeting pathways (Rodriguez-Boulan and Powell, 1992). In the thyroid epithelial cell lineFRT, the apical protein DPPIV is delivered to the apicalsurface from the basolateral by transcytosis during the first twodays after plating, but is directly delivered from the Golgi tothe apical surface as the monolayer matures (Zurzolo et al.,1992). Similar changes have been demonstrated for severalsurface proteins in MDCK cells (Wang et al., 1990a,b). DoRPE cells actively sort membrane proteins? Cultured humanRPE cells possess the ability to segregate influenza hemaglut-tinin and VSV G proteins into apical and basolateral plasmamembrane domains (Bok et al., 1992), and the same seems tobe true for rat RPE in vitro and in situ (V. L. Bonilha et al.,unpublished observations). Furthermore another apical markerprotein, p75-NTR, which has been shown to be sorted in theTGN of MDCK cells is apically polarized both in RPE-J cellsand in rat RPE in situ (A. D. Marmorstein et al., unpublishedobservations). The demonstration of different targetingpathways in situ versus in culture requires targeting assays tomeasure polarized surface delivery in the eye. We are currentlyattempting to develop an in situ biotin targeting assay,modelled after the quenched biotinylation assay described inFig. 1.

RET-PE2 could be stabilized at the apical surface by associa-tion with the subapical ankyrin-fodrin cytoskeleton, as has beenproposed for Na+,K+-ATPase (Gundersen et al., 1991) thusresulting in a longer half-life than on the opposite surface, fromwhere it would be rapidly removed and degraded. However,RET-PE2 remains detergent soluble throughout normal develop-ment of the retina and in the adult animal (Fig. 9), which appar-

ently excludes this mechanism. Furthermore, although recentstudies have shown an increased detergent insolubility ofNa+,K+-ATPase enzyme as it redistributes from non polar toapical during embryonic development of RPE (Gundersen et al.,1991; Rizzolo and Heiges, 1991; Huotari et al., 1995; Rizzoloand Zhou, 1995), they have also shown that fodrin appears to beexcluded from microvilli. This would limit the interaction of theenzyme to ankyrin, which is present in microvilli. Lastly, Mayset al. (1995) have presented data indicating that exclusion fromglycosphingolipid rafts in the Golgi apparatus might account forthe basolateral sorting of Na+,K+-ATPase in certain types ofMDCK cells. Na+,K+-ATPase is not polarized in RPE-J cells(Nabi et al., 1993; Marrs et al., 1995). We examined the sortingof lipids in RPE-J cells and found no evidence to suggest thatglycosphingolipids are sorted during delivery to the cell surface.This may account for the nonpolar distribution of Na+,K+-ATPase in RPE-J cells, however, it is unlikely that glycolipidexclusion could account for the polarized distribution of RET-PE2 in vivo or in vitro since this protein is basolaterally polarizedin RPE-J cells, even in the absence of glycolipid sorting.

Alternatively, RET-PE2 could be stabilized apically in adultRPE cells by external interaction with the photoreceptors orthe interphotoreceptor matrix. In this scenario, RET-PE2 mightbe an adhesive molecule with affinity for the IPM or for acomponent of the photoreceptor plasma membrane. Thismechanism is supported by our finding that RET-PE2 redis-

3033Regulated polarity of RET-PE2 antigen

tributes apically during the second postnatal week, which is thetime when photoreceptors complete their maturation in the rat(Olney, 1968), and that RET-PE2 expression in the retinabecomes RPE specific by PN9 (Neill et al., 1993), a time whenphotoreceptor differentiation is morphologically apparent. Thisis also the time when the IPM, composed of extracellularmatrix, is likely to be deposited between the outer segmentsand RPE; it is not known to what extent this material isproduced by the photoreceptors or by the RPE. Indeed, it hasbeen shown that addition of collagen to the apical surface ofMDCK cells results in apical redistribution of integrins withina few hours (Ojakian and Schwimmer, 1994) and that the local-ization of the anion transporter in intercalated epithelial cellsis determined by the extracellular matrix (van Adelsburg et al.,1993, 1994). Molecular and biochemical characterization ofthe RET-PE2 antigen is necessary to determine whetherinvolvement in adhesive interactions is plausible. It should bepointed out that, as mentioned above, such mechanism wouldnot be exclusive of additional polarity mechanisms; in fact,interactions with the neural retina might promote the develop-ment of the extremely elongated microvilli typical of RPE invivo.

Previous work (Neill and Barnstable, 1990; Neill et al.,1993) has shown that RET-PE2 immunoreactivity is broadlydistributed in the eye during embryonic development but that,postnatally it becomes restricted to the RPE. The develop-mental changes in RET-PE2 polarity described in this papersuggest a functional role of this protein in the maintenance ofphotoreceptors, or a dependence of the distribution of thisprotein on some still undefined component(s) of the neuralretina or the IPM. The characterization of RET-PE2 structureand function after cloning of its cDNA may yield excitinginsights on mechanisms of RPE adherence to the neural retinaand its role in RPE polarity.

We thank Drs Wouter van ’t Hof, and Anthony Scotto for sharingtheir expertise in the analysis of the fluorescent lipids, Drs SilviaFinnemann and Charles Yeaman for helpful discussion, and Dr GeriGurland for critical reading of the manuscript. Ms Leona Cohen-Gould and Ms Dena Almeida provided excellent technical assistance.Supported by NIH EY08538 to E.R.B., an NRSA award to A.D.M.,a CNPq fellowship to V.L.B., and the DYSON Foundation.

REFERENCES

Brown, D. A. and Rose, J. K. (1992). Sorting of GPI-anchored proteins toglycolipid enriched membrane subdomains during transport to the apical cellsurface. Cell 68, 533-544.

Bok, D., O’Day, W. and Rodriguez-Boulan, E. (1992). Polarized budding ofvesicular stomatitis and influenza virus from cultured human and bovineretinal pigment epithelium. Exp. Eye Res. 55, 853-860.

Bok, D. (1993). The retinal pigment epithelium, a versatile partner in vision. J.Cell Sci. Suppl. 17, 189-195.

Caldwell, R. B. and McLaughlin, B. J. (1984). Redistribution of Na-K-ATPase in the dystrophic rat retinal pigment epithelium. J. Neurocytol. 13,895-910.

Caplan, M. J., Anderson, H. C., Palade, G. E. and Jamieson, J. D. (1986).Intracellular sorting and polarized cell surface delivery of (Na+,K+) ATPase,an endogenous component of MDCK cell basolateral plasma membranes.Cell 46, 623-631.

Dowling, J. E. and Gibbons, I. R. (1962). The fine structure of the pigmentepithelium in the albino rat. J. Cell Biol. 14, 459-474.

Gottardi, C. J. and Caplan, M. J. (1993). Delivery of Na,K-ATPase inpolarized epithelial cells. Science 260, 550-552.

Gundersen, D., Orlowski, J. and Rodriguez-Boulan, E. (1991). Apicalpolarity of Na,K-ATPase in retinal pigment epithelium is linked to a reversalof the ankyrin-fodrin submembrane cytoskeleton. J. Cell Biol. 112, 863-872.

Gundersen, D., Powell, S. K. and Rodriguez-Boulan, E. (1993). Apicalpolarization of N-CAM in retinal pigment epithelium is dependent on contactwith the neural retina. J. Cell Biol. 121, 335-343.

Hammerton, R. W., Krzeminski, K. A., Mays, R. W., Ryan, T. A., Wollner,D. A. and Nelson, W. J. (1991). Mechanism for regulating cell surfacedistribution of Na+,K+-ATPase in polarized epithelial cells. Science 254,847-850.

Hanzel, D., Nabi, I. R., Zurzolo, C., Powell, S. K. and Rodriguez-Boulan, E.(1991). New techniques lead to advances in epithelial cell polarity. Semin.Cell Biol. 2, 341-353.

Huotari, V., Sormunen, R., Lehto, V. P. and Eskelinen, S. (1995). Thepolarity of the membrane skeleton in retinal pigment epithelial cells ofdeveloping chicken embryos in primary culture. Differentiation 58, 205-215.

Lagunowich, L. A. and Grunwald, G. B. (1989). Expression of calcium-dependent cell adhesion during ocular development: A biochemical,histochemical and functional analysis. Dev. Biol. 135, 158-171.

Le Bivic, A., Real, F. X. and Rodriguez-Boulan, E. (1989). Vectorialtargeting of apical and basolateral plasma membrane proteins in a humanadenocarcinoma epithelial cell line. Proc. Nat. Acad. Sci. USA 86, 9313-9317.

Lisanti, M. P., Caras, I. W., Gilbert, T., Hanzel, D. and Rodriguez-Boulan,E. (1990). Vectorial apical delivery and slow endocytosis of a glycolipid-anchored fusion protein in transfected MDCK cells. Proc. Nat. Acad. Sci.USA 87, 7419-7423.

Marrs, J., Andersone-Fissone, C., Jeong, M., Cohen-Gould, L., Zurzolo, C.Nabi, I., Rodriguez-Boulan, E. and Nelson, W. (1995). Plasticity inepithelial phenotype: Modulation by expression of different cadherin celladhesion molecules. J. Cell Biol. 129, 507-520.

Mays, R. W., Stemers, K. A., Fritz, B. A., Lowe, A. W. and van Meer, G.(1995). Hierarchy of mechanisms involved in generating Na/K-ATPasepolarity in MDCK epithelial cells. J. Cell Biol. 130, 1105-1115.

Micali, A., Parducci, F., La Fauci, M. A., Urbani, P., Santoro, G. andPizzolo, D. (1989). Maturazione post-natale della ratto albino:1 L’epiteliopigmentato. Arch. Ital. Anat. Embryol. 94, 405-424.

Nabi, I. R., Mathews, A. P., Cohen-Gould, L., Gundersen, D. andRodriguez-Boulan, E. (1993). Immortalization of polarized rat retinalpigment epithelium. J. Cell Sci. 104, 37-49.

Neill, J. and Barnstable, C. (1990). Expression of the cell surface antigensRET-PE2 and NCAM by rat retinal pigment epithelial cells duringdevelopment and in tissue culture. Exp. Eye Res. 51, 573-583.

Neill, J. M., Thornquist, S. C., Raymond, M. C., Thompson, J. T. andBarnstable, C. J. (1993). RET-PE10: A 61 kD polypeptide epitopeexpressed late during vertebrate RPE maturation. Invest. Ophthalmol. Vis.Sci. 34, 453-462.

Nelson, W. J. (1991). Cytoskeleton functions in membrane traffic in polarizedepithelial cells. Semin. Cell Biol. 2, 375-385.

Ojakian, G. K. and Schwimmer, R. (1994). Regulation of epithelial cellsurface polarity reversal by β1 integrins. J. Cell Sci. 107, 561-576.

Okami, T., Yamamoto, A., Omori, K., Takada, T., Uyama, M. and Tashiro,Y. (1990). Immunocytochemical localization of Na,K-ATPase in rat retinalpigment epithelial cells. J. Histochem. Cytochem. 38, 1267-1275.

Olney, J. W. (1968). An electron microscopic study of synapse formation,receptor outer segment development and other aspects of developing mouseretina. Invest. Ophthalmol. Vis. Sci. 7, 250-268.

Powell, S. K., Cunningham, B. A., Edelman, G. M. and Rodriguez-Boulan,E. (1991). Transmembrane and GPI anchored forms of NCAM are targetedto opposite domains of a polarized epithelial cell. Nature 353, 76-77.

Rizzolo, L. J. (1990). The distribution of Na+,K+-ATPase in the retinalpigmented epithelium from chicken embryo is polarized in vivo but not inprimary cell culture. Exp. Eye Res. 51, 435-446.

Rizzolo, L. J. (1991). Basement membrane stimulates the polarizeddistribution of integrins but not the Na,K-ATPase in the retinal pigmentepithelium. Cell Regul. 2, 939-949.

Rizzolo, L. J. and Heiges, M. (1991). The polarity of the retinal pigmentepithelium is developmentally regulated. Exp. Eye Res. 53, 549-553.

Rizzolo, L. J. and Zhou, S. (1995). The distribution of Na+,K+-ATPase and5A11 antigen in apical microvilli of the retinal pigment epithelium isunrelated to α-spectrin. J. Cell Sci. 108, 3623-3633.

Rodriguez-Boulan, E. and Powell, S. K. (1992). Polarity of epithelial andneuronal cells. Annu. Rev. Cell Biol. 8, 395-427.

Sargiacomo, M., Lisanti, M., Graeve, L., Le Bivic, A. and Rodriguez-

3034 A. D. Marmorstein and others

Boulan, E. (1989). Integral and peripheral protein compositions of the apicaland basolateral membrane domains in MDCK cells. J. Membr. Biol. 107,277-286.

van Adelsburg, J. S., Edwards, J. C., al-Awqati, Q. (1993). The apicalCl/HCO3 exchanger of beta intercalated cells. J. Biol. Chem. 268, 11283-11289.

van Adelsburg, J., Edwards, J. C., Takito, J., Kiss, B., al-Aqwati, Q. (1994).An induced extracellular matrix protein reverses the polarity of band 3 inintercalated epithelial cells. Cell 76, 1053-1061.

van Meer, G. and Simons, K. (1988). Lipid polarity and sorting in epithelialcells. J. Cell Biochem. 36, 51-58

van Meer, G. (1993). The transport and sorting of membrane lipids. Curr.Opin. Cell Biol. 5, 661-673.

van Meer, G. and van ’t Hof, W. (1993). Epithelial sphingolipid sorting isinsensitive to reorganization of the Golgi by nocodazole, but is abolished bymonensin in MDCK cells and by brefeldin A in Caco-2 cells. J. Cell Sci. 104,833-842.

Wang, A. Z., Ojakian, G. K. and Nelson, W. J. (1990a). Steps in themorphogenesis of a polarized epithelium. I. Uncoupling the roles of cell-celland cell substratum contact in establishing plasma membrane polarity inmulticellular epithelial (MDCK) cysts. J. Cell Sci. 95, 137-151.

Wang, A., Ojakian, G. K. and Nelson, W. J. (1990b). Steps in themorphogenesis of a polarized epithelium. 2. Disassembly and assembly ofplasma membrane domains during reversal of epithelial polarity inmulticellular epithelial (MDCK) cysts. J. Cell Sci. 95, 153-165.

Wang, N., Koutz, C. A. and Anderson, R. E. (1993). A method for theisolation of retinal pigment epithelial cells from adult rats. Invest.Ophthamol. Vis. Sci. 34, 101-107.

Zinn, K. M. and Marmor, M. F. (1979). The retinal pigment epithelium.Harvard University Press, Cambridge, MA.

Zurzolo, C., Le Bivic, A., Quaroni, A., Nitsch, L. and Rodriguez-Boulan, E.(1992). Modulation of transcytotic and direct targeting pathways in apolarized thyroid cell line. EMBO J. 11, 2337-2344.

Zurzolo, C., Le Bivic, A. and Rodriguez-Boulan, E. (1993). Cell surfacebiotinylation techniques. In Cell Biology, a Laboratory Handbook, vol. 3 (ed.J. Celis), pp. 185-192. Academic Press, NY.

Zurzolo, C. and Rodriguez-Boulan, E. (1993). Delivery of Na+,K+-ATPasein polarized epithelial cells. Science 260, 550-552.

(Received 12 July 1996 – Accepted 22 September 1996)