180 European Journal of Cell Biology 66, 180-191 (1995, February) © Wissenschaftliche Verlagsgesellschaft . Stuttgart

Cell adhesion to the apical pole of epithelium:a function of ce 11 polarity

Michael Thie1)a, Bärbel Harrach-Ruprechtb, Heinrich Sauere, Petra Fuchsa, Anja Albersa,Hans-Werner Denkera

a Institute of Anatomy, University of Essen, Medical School, Essen/Germanyb Institute für Arteriosclerosis Research, University of Münster, Münster/GermanyC Max-Planck-Institute for Molecular Physiology, Dortmund/Germany

ReceivedJuly 6, 1994Accepted September 16, 1994

Uterine epithelium - epithelial phenotype - polarization ­adhesiveness

Human uterine epithelinm displays a distinct polarized organizationwith apical, lateral, and basal plasma membrane domains. Althoughnon-adhesive throughout most of the menstrual cyde, epithelial cellsallow attachment of trophoblast cells to their apical pole duringembryo implantation. Arecent hypothesis postulates that epithelialcells turn off genes for apical-basal polarity and turn on genes for amore mesenchyme-Iike phenotype allowing cell-cell interaction withtrophoblast.

Using an in vitro assay human uterine celllines (RL95-2, HEC-I-A,AN3-CA) were selected on the basis of adhesiveness for trophoblast­type cells (JAR). Subsequently, uterine cells were examined forepithelium-specific ultrastructure using transmission electron micros­copy, and for the expression of E-cadherin, «6-, ßl-, ß4-integrin sub­units and cytokeratin using immunocytochemistry, confocallaser scan­ning microscopy, and surface replication technique. HEC-I-A mono­layers are non-adhesive for JAR cells and appear highly polarizedexpressing E-cadherin, «6-, ßl-, ß4-integrin subunits, and cytokeratin.Both, integrins and E-cadherin, are present at the lateral membrane.RL95-2 monolayers which are adhesive for JAR cells appear non­polarized. Like HEC-I-A cells, RL95-2 cells express E-cadherin, «6-,ßl-, and ß4-integrin subunits, and cytokeratin. In contrast to HEC-I-Acells, integrins and E-cadherin are distributed at the entire cell surface.AN3-CA monolayers are non-adhesive for JAR cells and appear non­polarized. Cells lack epithelial-specific markers such as keratin and E­cadherin. They show only low expression of «6-, ßl-integrin subunitsand lack ß4-integrin subunit. Conversely, they express vimentin.

Thus, modulation of the epithelial phenotype of uterine cells, i. e.loss of apical-basal polarity, might prepare the apical cell pole for cell­cell interaction with trophoblast. However, loss of cell polarity wouldnot lead to enhancement of adhesiveness for trophoblast if accom­panied by a loss of epithelium-specific adhesion molecules.

1) Dr. Michael Thie, Institut für Anatomie, Universitätsklinikum,Hufelandstr. 55, D-45122Essen/Germany.

Introduction

The epithelial cells lining the uterine cavity are structurallyand functionally polarized cells with distinct basal, lateral, andapical membrane domains. As is typical für simple epithelia,the apical surface of uterine epithelial cells (UECs) is free andnon-adhesive for opposing UECs or embryonic cells such astrophoblast. Nevertheless, the uterine epithelium is not a pas­sive surface, and UECs can be functionally reprogrammed tocontribute actively to trophoblast adhesion. When appropri­ately conditioned with steroid hormones, these epithelial cellsenter astate of so-called receptivity and switch from a non­adhesive state to a potentially adhesive state. When exhibitingthe adhesive state, the apical membrane domain of UECsallows the attachment of trophoblast (für review, see [22]).The processes involved in modulating adhesiveness of UECsfür trophoblast, however, have not been identified so far.

A mechanism to achieve UEC adhesiveness für trophoblastis recently being discussed postulating that UECs modulatetheir apical-basal polarity [6, 7]. This plasticity in the pheno­type of adult UECs may involve some of the elementary pro­cesses that playa rale in embryology during transformation ofepithelium to mesenchyme. A characteristic of that latter pro­cess is, likewise, that apical-basal polarity is lost, and adhesionmolecules are redistributed and/or newly acquired [12, 13]. Inanalogy, in UECs, part of the master gene pragram for theepithelial phenotype including genes for apical-basal polaritymay be turned off and, vice versa, certain genes for themesenchymal program may be turned on thus enhancingadhesiveness of UECs for trophoblast. The activation of themesenchymal pro gram in definitive epithelia occurs not onlyduring development in vivo [13] but also in vitro, e. g. in lensepithelium [11] and Madin-Darby canine kidney cells [40]. Sig­nals such as tumor-promoting phorbol esters [21], oncoprote­ins [31], growth factors [26], and/or signals generated by cell­cell and cell-matrix interactions [30] appear to contral geneactivation in these systems.

In this study we characterize parameters of the epithelialphenotype of certain human endometrial ceIl lines (RL95-2,HEC-I-A, AN3-CA) and correlate these with adhesive ornon-adhesive behavior for trophoblast-type ceIls (JAR) in anattempt to gain insight into the pro gram underlying UECadhesiveness. We examine ultrastructural features as weIl as

expression of markers associated to the epithelial phenotype,i. e. E-cadherin [35], a6-, ßl-, ß4-integrin subunits [33, 34],and keratin intermediate filaments [24].

On the basis of our data we postulate that modulation of theepithelial phenotype of UECs, specificaIly loss of apical-basalpolarity, prepares the apical ceIl pole for ceIl-ceIl interactionwith trophoblast. Loss of ceIl polarity, however,' would notlead to enhancement of UEC adhesiveness for trophoblast ifaccompanied by a loss of epithelium-specific adhesion mole­cules. This suggests that specific modulation of polarity­related parameters rather than down-regulation of the entireepithelial pro gram is a key event in this type of adhesive epi­thelial interactions .

Materials and methods

Routine cell cultureHuman endometrial carcinoma cell lines were purchased from theAmerican Type Culture Collection (ATCC), Rockville, MDIUSA, i. e.RL95-2 cells (CRL 1671; [37]), HEC-I-A cells (HTB 112; [19]), andAN3-CA cells (HTB 111; [5]). For routine culture, cell lines weregrown in plastic flasks in 5 % C02"95 % air at 37°C. In brief, RL95-2cells were seeded out in a 1+1 mixture of Dulbecco's modification ofEagle's medium and Ham's F12 (Gibco-Life Technologies, Eggenstein/Germany) supplemented with 10% fetal calf serum (Gibco), 10 mMHEPES (Gibco), and 0.5 flg/ml insulin (Gibco), HEC-I-A cells inMcCoy's 5A medium (Gibco) supplemented with 10 % fetal calfserum, and AN3-CA cells in Eagle's minimum essential medium withEarle's salts and non-essential amino acids (Gibco) supplemented with10 % fetal calf serum. All media were additionally supplemented withpenicillin (100 lU/mi; Gibco) and streptomycin (100 flg/ml; Gibco).The growth medium was changed every 2 to 3 days, and cells were sub­cultured by trypsinization (trypsin-EDTA solution; Gibco) when theybecame confluent.

Cell culture on coverslipsCells were harvested by trypsinization from confluent cultures,counted, and adjusted to the desired concentration, i. e. RL95-2700000 cells, HEC-I-A 200000 cells, and AN3-CA 300000 cells eachin 2.0 ml of their respective culture medium. Subsequently, cell sus­pension was poured out on POlY-D-lysine-coated glass coverslips (12mm in diameter) situated in 4 cm2 Falcon multiwells. Cells were grawnin medium to confluent monolayers and used for experiments within 3days after start of cultures.

Attachment assayAdhesiveness of endometrial cell monolayers for human JAR chorio­carcinoma cells (ATCC: HTB 144) was measured using a centrifugalforce-based adhesion assay. The establishment of the adhesion assayhas been described previously [17]. Briefly, JAR cell suspension(100000 cells per mJ RPMI 1640 medium (Gibco) suppJemented with10 % fetal calf serum and penicillin-streptomycin) was incubated on agyratory shaker at 110 x rpm obtaining multicelluJar spheroids 72 hafter initiation of culture. Subsequently, JAR spheroids were har­vested, counted, and gently deJivered onto a confluent monoJayer ofhuman endometrial celliines grawn on coversJips. Confrantation cul­tures were grown in RPMI 1640 medium with or without 10 % fetal calfserum supplementation in a humidified 5 % C02"95 % air incubator at37°C. For confrontation cultures in serum-free medium, monolayers

Loss of polarized epithelial phenotype 181

and spheroids were washed two times with RPMI 1640 medium priorto the experiment. After 1 h, spheraid adhesion to the endometrialmonolayers was quantified by cenrifuging coverslips with cell-spheroidsurface facing down at 12g for 5 min. Attached spheroids were countedand expressed as the percentage of the nu mber of spheroids seeded.

ImmunofluorescenceCells grown on gJass coversJips were rinsed twice in phosphate­buffered saline (PBS), fixed and permeabilized by incubation in 96 %methanol-water for 10 min at -20°C. After severaJ washings with PBSand a finaJ wash in PBS supplemented with 0.5 % bovine serum albu­min (BSA) , cells were incubated for 1 h at room temperature with theprimary antibody. Thereafter, cells were rinsed in PBS/0.5 % BSA (4 x10 min) and incubated with the corresponding fluorescein iso­thiocyanate-conjugated secondary antibody for 1 h at room temper­ature. In controJ experiments the primary antibody was omitted. Afterrinsing with PBS, specimens were mounted with 90 % gJycerol-PBS,supplemented with 1.0 % p-phenyJenediamine as an antiquenchingagent and examined with a Zeiss Axiophot microscope equipped withepiillumination (450-490 nm excitation; fiJterset 487909). Photographswere taken on Neopan 1600 film (Fuji, Tokyo/Japan).

AntibodiesRat monocJonal antibody to a6-integrin subunit (GoH3; [32]) was pro­vided by Dr. A. Sonnenberg (The Netherlands Cancer Institute, Divi­sion of Cell Biology, Amsterdam/The Netherlands), and diJuted 1:3with PBS/0.5 % BSA before use. Rat monocJonal antibody to ßl­integrin subunit (AllB2; [38]) was pravided by Dr. C. Damsky(Department of Anatomy, University of California, San Francisco/USA), and was diJuted 1:3 with PBS/0.5 % BSA before use. MousemonocJonal antibody to ß4-integrin subunit (3El) was purchased framBiomoJ, Hamburg/Germany, and was diluted 1: 100 with PBS/0.5 %BSA before use. Mouse monocJonal antibody to E-cadherin (6F9; [8])was donated by Dr. J. Behrens (Max-DeJbrück-Centrum, Berlin/Ger­many), and was diJuted 1:5 with PBS/0.5% BSA before use. MousemonocJonal antibody to cytokeratin No 8 (4.1.18) was purchased framBoehringer Biochemica, Mannheim/Germany, and was diJuted 1:20with PBS/0.5 % BSA before use. Mouse monocJonal antibody tovimentin (V9) was obtained fram Sigma-Aldrich, Deisenhofen/Ger­many, and was diJuted 1:40 with PBS/0.5 % BSA before use.

Fluorescein isothiocyanate (FITC)-conjugated rabbit anti-mousesecondary antibodies (F232) and FITC-conjugated rabbit anti-rat sec­ondary antibodies (F234) were obtained from Dako Diagnostika,HamburgiGermany. Gold (15 nm)-conjugated goat anti-mouse sec­ondary antibodies and goJd (15 nm)-conjugated goat anti-rat second­ary antibodies were obtained from Biotrend, KäJn/Germany. Anti­bodies were diJuted with PBS/0.5 % BSA according to manufacturersinstructions.

Confocallaser scanning microscopyConfocal microscopy was carried out using a confocaJ Jaser scanningmicrascope (MRC 600, BioRad, Heme! Hampstead/UK) equippedwith an argon ion Jaser (Ion Laser TechnoJogy, Salt Lake City, UT/USA) for 488 nm excitation. The confocaJ Jaser scanning unit wascoupled to a standard micrascope (Diaphot, Nikon, Düsseldorf/Ger­many). For the experiments a 60-fold oiJ immersion objective (Nikon,DüsseJdorf/Germany) with a numericaJ aperture of 1.4 was chosen.The theoretical value for the z-resolution of the microscape was calcu­lated to be 0.4 flm [39]. Images of horizontal optical sections wererecorded in the x-y plane with 256 Jines/image. For vertical optical sec­tions scans were performed in the x-z plane with 200 Jines/image. Pho­tographs were taken with a Ilford APX-I00 fiJm (Mobberley, Cheshire/UK) from a black and white monitor.

Surface replication techniqueFor surface replication cells grown on coverslips were fixed and incu­bated with primary and goJd-conjugated secondary antibodies asdescribed above. Thereafter, cells were washed, postfixed with 2.5 %gJutaraJdehyde in PBS for 2 h at raom temperature, dehydrated with

182 M. Thie, B. Harrach-Ruprecht, H. Sauer, P. Fuchs et 01.

graded ethanol and air dried. Platinum-carbon surface replicas of cellswere made in a Balzers BA 300 apparatus (Balzers/Liechtenstein)equipped with an electron gun evaporator and a quartz crystal thick­ness monitor. Replicas were obtained by shadowing the cell surfacewith platinum-carbon at an angle of 38°, followed by carbon at 90°.The replicas were cleaned overnight in sodium hypo chloride (12 %active chloride; Hedinger, Stuttgart/Germany) and washed in distilledwater. They were picked up on 200 mesh copper grids and examined ina Philips EM 410 at 60 kV

Electron microscopyCells were grown as monolayers on thermanox coverslips (13 mm indiameter; Nunc, Naperville, IL/USA) as described above. For subse­quent electron microscopy, samples were rinsed twice in PBS and fixedin 2.5 % glutaraldehyde in 0.1 M cacodylate buffer, pH 7.4, for 30 minat room temperature. After repeated washings in cacodylate buffer,sampIes were postfixed with 1 % OS04 in cacodylate buffer, dehy­drated with graded ethanol and propylene oxide, and embedded inepoxy resin [4]. The embedded monolayers were separated from thethermanox coverslip by snap freezing in liquid nitrogen. Ultrathin sec­tions were mounted on 200-mesh copper grids, double-stained withuranyl acetate and lead citrate, and examined with a Philips EM 400 at80 kV

Results

Adhesiveness of monolayersThe adhesiveness of RL95-2, HEC-1-A, and AN3-CA mono­layers for JAR cells was measured in the presence of serumand the absence of serum to consider mediating effects ofserum molecules in cell-cell interaction (Fig. 1).

100

~ ""0 I182.4C :J0.D";!2.

0~(/)

50(/) <D I• 137.3

140.2c<D>'(jjQ)..c""0cU

0 RL

HECANCO

Fig. 1. Adhesiveness of human endometrial celliines (RL = RL95-2;HEC = HEC-1-A; AN = AN3-CA) for human JAR choriocarcinomaspheroids as determined in the centrifugal force-based spheroid adhe­sion assay. Spheroids were delivered onto endometrial monolayers,and confrontation cultures were grown in medium supplemented witheither 10% fetal calf serum (black bars) or without fetal calf serum(hatched bars) for 1 h. Adhesiveness is expressed as the percentage ofthe number of spheroids seeded. For comparison, adhesiveness ofpolY-D-lysine-coated glass coverslips (CO) for spheroids is shown. Thevalues are mean ± SEM of 25 confrontation pairs counted, from deter­minations from five separate cultures.

JAR cells attached with high efficiency to RL95-2 cells.Attachment was more effective in confrontation cultures sup­plemented with 10% fetal calf serum (FCS) (RL: 82.4 % ±16.0 attachment) compared with cultures lacking FCS (RL:37.3 % ± 29.5 attachment). Values of JAR cell attachment toRL95-2 cells were significantly high er than values of JAR cellattachment to POlY-D-lysine-coated glass coverslips both in thepresence (CO: 36.4 % ± 29.6 attachment) and absence ofserum (CO: 0 % attachment) pointing to a degree of specific­ity of attachment between RL95-2 cells and JAR cells.

Compared to RL95-2 cells, JAR cells attached with low effi­ciency to HEC-1-A cells. As above, confrontation cultures inthe presence of FCS showed high er values of attachment(HEC: 40.2 % ± 23.4 attachment) than cultures in theabsence of FCS (HEC: 6.5 % ± 4.4 attachment). However,this was comparable to JAR cell attachment to POlY-D-lysinecoated glass coverslips either in the presence (CO: 36.4 % ±29.6 attachment) or absence of serum (CO: 0% attachment).This suggests that JAR cell attachment to HEC-1-A cells wasdue to serum and JAR cell-specific molecules rather than spe­cific for HEC-1- A cells. According to this, HEC-1- A cells canbe considered non-adhesive for JAR cells unless serum­derived bridging molecules are provided.

JAR cells did not attach to AN3-CA neither in the presenceof FCS (AN: 1.8 % ± 4.1 attachment) nor in the absence ofFCS (AN: 0 % attachment). Thus, AN3-CA cells are intrinsi­cally non-adhesive for JAR cells and lack bin ding sites forserum constituents that may act as bridging molecules.

Ultrastructural featuresTo determine whether differences in the adhesiveness of celllines were associated with differences in the epithelial mor­phology, we analyzed RL95-2, HEC-1-A, and AN3-CA cellsby transmission electron microscopy (Figs. 2-4).

RL95-2 cells grew predominantly as monolayers but cellsshowed a tendency of piling up. Cells were varying in size.Typically, a single cell had a roundish shape (Fig. 2a). The cellnuc1ei were predominantly located in the center of the cells.Cell organelles tended to pile up perinuc1early. The upper sur­face of the cell appeared dome-like and was free of microvilli .Thus, cells were lacking a distinct apical pole. RL95-2 cellsadhered to the substrate via large cytoplasmic extensionsexc1uding broad cell-matrix contact of the basal membrane.

Fig. 2. Photomicrographs of ultrathin sections of RL95-2 cells culti­vated on a POlY-D-lysine-coated coverslip. - a. Cells grow as a mono­layer but show a lack of structural polarization, no regular microvilliand no subplasmalemnal filament network at the apical plasma mem­brane. - b. Only primitive focal adherens junctions are seen at thelateral plasma membranes (arrows). - cl Cell 1. - c2 Cell 2. - csCoverslip. - me Growth medium. - N Nucleus. - Bars 2 [-Lm(a), 0.25[lm (b).

Fig. 3. Photomicrographs of ultrathin sections of HEC-1-A cells cul­tivated on a polY-D-lysine-coated coverslip. - a. Cells grow as anordered monolayer exhibiting a highly polarized epithelial phenotypewith numerous microvilli at the apical cell pole (arrowheads). - b.Adjacent cells show closely apposed plasma membranes with properformation of tight junctions (Zarge arrows), adherens junctions (smallarrows) and desmosomes (asterisks). - cl Cell 1. - c2 Cell 2. - csCoverslip. - me Growth medium. - N Nucleus. - Bars 2 [lm (a), 0.25[lm (b).

EJCB Loss of polarized epithelial phenotype 183

I

184 M. Thie, B. Harrach-Ruprecht, H. Sauer, P. Fuchs et al.

Fig. 4. Photomicrographs of ultrathin sections of AN3-CA cells cul­tivated on a polY-D-lysine-coatcd coverslip. - a. Cells grow as a mono­laycr but show a lack of structural polarization. - h. Adjacent cells

Cells formed small primitive adherens junctions. Regions ofinteracting plasma membran es were alternating with regionsof large intercellular spaces (Fig. 2b). Occasionally, desmo­somes could be observed (data not shown). Tight junctionswere lacking as proved by freeze fracture electron microscopy(data not shown).

HEC-I-A cells formed ordered monolayers in which thesingle cells were more or less uniform in size and shape grow­ing in close contact to one another and to the substrate (Fig.3a). Cells had a cylindrical shape, nuclei were predominantlysituated at the base of the cells. Mitochondria, Golgi appara­tus, and endoplasmic reticulum were mostly located at thesupranuclear region of the cel!. The apical surface was coveredwith numerous microvilli which were relatively short. The cellsshowed closely apposed plasma membranes at their lateralfaces with tight junctions in the subapical region and adherensjunctions and desmosomes scattered along the basolateralmembranes (Fig. 3b). The tight junctions bctwecn adjaccntcells were verified using freeze fracture electron microscopy(data not shown).

AN3-CA cclls formed a regular monolayer in which thesingle cells werc round to cylindrical in shape (Fig. 4a). Ccllsflattcned against the substratum. The cell nuclei were irregu­larly positioncd in thc cclls. Cell organelles piled up prcdom­inantly in thc vicinity of the nucleus. The apical surfaceappeared dome-like lacking microvilli. The cclls showed only

EJCB

'c ""'...•

]

show only primitive focal membrane contacts. Organized intercellularjunctions are lacking. - cl Cell 1. - c2 Cell 2. - cs Coverslip. - meGrowth medium. - N Nucleus. - Bars 2 flm (a), 0.25 [.lm(b).

loose attachment to one anothcr. The lateral membranes of

adjacent cells ran parallel without forming organized intercel­lular junctions (Fig. 4b).

Thus, RL95-2 cells which are adhesive for trophoblast-typecells showed morphological features indicating lack of struc­tural polarization while I-lEC-l-A cells being non-adhesivedisplayed in highly polarized phenotype with respect to thedistribution of organelles and to membrane organization.AN3-CA cells appeared non-polarized, althollgh being non­adhesive.

Epithelial marker expressionTo determine whether differences in epithelial morphology ofcell lines were associated with loss of epithelium-associatedmolecllles, we analyzed the expression of E-cadherin, 0.6-,ßl-, ß4-integrin sllbllnits, and cytokeratin (Figs. 5,6). The cell

Fig. 5. lmmunostaining of endometrial monolayers (RL = RL95-2;HEC = HEC-l-A; AN = AN3-CA) with monoclonal antibodies to E­cadherin (a-c), cytokeratin No 8 (d-f), and vimentin (g-i). Note thatRL95-2 cells and HEC-l-A cells but not AN3-CA cells are positive forE-cadherin and cytokeratin. AN3-CA cells but not RL95-2 cells andJ-IEC-l-A cclls are positive for vimentin. - Ecd E-cadherin. - kerCytokeratin intermecliate filaments. - vim Vimentin intermecliate fila­ments. - Bar 20 [.lm.

EJCB Loss of polorized epithelial phenotype 185

I

186 M. Thie, B. Harrach-Ruprecht, H. Sauer, P. Fuchs et al. EJCB

EJCB

lines were stained in parallel and the photographic processingwas identical thus allowing direct comparison of staining pat­terns.

RL95-2 cells and I-lEC-I-A cells were clearly positive for E­cadherin (Figs. 5a, b) and a6-, ~-l-, and r34-integrin subunits(Figs. 6a, b, d, e, g, h), showing characteristic membranc­bound staining. With all these markers, the overall intensity ofimmunostaining of RL95-2 cells was comparable with that ofHEC-I-A cells. AN3-CA cells, on the other hand, were com­

pletely negative for E-cadherin (Fig. 5c) and ~4-integrin sub­units (Fig. 6i), and only weakly positive for a6-, and ~1­integrin subunits (Figs. 6c, f).

When RL95-2 cells and HEC-I-A cells were labeled with an

antibody to cytokeratin No 8, strong intracellular stainingappeared (Figs. 5d, e). In RL95-2 cells, staining was mostintense in the perinuclear region indicating disorganization ofthe intermediate filament network, while HEC-I-A cellsshowed a well-defined network of filaments often associated

with regions of ccll-cell contacts. In contrast, cytokeratin No 8was not expressed by the AN3-CA cclls (Fig. 5f).

Whcn RL95-2 cells, HEC-I-A cells and AN3-CA cells werestained for vimentin, AN3-CA cells reacted prominently (Fig.5i), whereas RL95-2 cells and HEC-I-A cells were devoid ofimmunoreactivity (Figs. 5g, h).

In conclusion, the expression of E-cadherin, a6-, ~1-, ~4­

integrin subunits and cytokeratin indicated the epithelial phe­no type of J-lEC-I-A cells and of RL95-2 cells although the lat­tcr exhibited a lack of structural polarization. In contrast,AN3-CA cells showed non-polar morphology correspondingto the lack of E-cadherin, ~4-integrin, and cytokeratin.According to the expression of vimentin, these cells appearedto be of mesenchymal phenotype.

Apico-basolateral distribution of E-cadherin andintegrin subunits (u.6, ß1,ß4)In order to visualize the domain-specific localization of E­cadherin and integrins, monolayers were examined by confo­callaser scanning microscopy (Figs. 7, 8). Images of verticaloptical sections were individually processed to visualize pro­tein distribution in cells. Thus, intensity of staining differsfrom the original data given in Figures 5 and 6.

RL95-2 cells were labeled by the anti-E-cadherin antibodyalong thc cntire plasma membrane (Fig. 7a). However, fluo­rescence was not evenly distributed over the whole cell mem­brane as staining was often pronounced at regions of cell-cellcontacts. In I-IEC-I-A cells (Fig. 7b), in contrast, fluorescentstaining was largely confined to the sites of cell-cell contactsand was absent from the apical surface. As anticipated, theplasma membrane was not labeled at all by the antibodyagainst E-cadherin in AN3-CA cells (Fig. 7c).

A similar staining pattern was observed with antibodies toa6-, r31-, r34-integrin subunits. Again, RL95-2 cells were

Fig. 6. Immunostaining of endometrial monolayers (RL = RL95-2;HEC = HEC-I-A; AN = AN3-CA) with monoclonal antibodies to a6­integrin subllnit (a-c), ßl-integrin subllnit (d-C), ancl ß4-integrin sub­unit (g-i). RL95-2 cclls and l-IEC-I-A cclls are positive for a6-, ßl-,r34-integrin subunits. AN3-CA eells are only weakly positive for a6-,ßl-integrin sllbllnits and eompletely negative for ß4-integrin subunits.- Bar 20 fun.

Lass af palorized epithelial phenatype 187

Fig. 7. Confocal images of endometrial monolayers (RL = RL95-2;HEC = HEC-I-A; AN = AN3-CA) after staining with monoclonalantibody to E-eadherin. Vertieal seetions reveal that RL95-2 eells arelabeled along the entire plasma membrane (a). HEC-I-A eells arelabeled laterally (b), while AN3-CA eells are not labeled at all; noteautofluoreseenee of nllclells (e). Arrows mark the position of eell-eellcontaets in monolayers. - es Coverslip. - Bar 10 [.lm.

labeled rather uniformly at the entire plasma membrane (Figs.8a, d, g), while HEC-I-A cells were labeled mainly at sites ofcell-cell contacts (Figs. 8b, e, h). AN3-CA cells showed stain­ing at sites of cell-cell contacts using anti-a6-, ~ l-integrin anti­bodies (Figs. 8c, f), and no staining using anti-~4-integrin anti­bodies (Fig. 8i).

In summary, the random distribution of E-cadherin andintegrin subunits along the entire plasma membrane in RL95-2cells corroborated the morphological observation that ccllslacked a defined polarization. In contrast, the lateral distribu­tion of proteins in HEC-I-A cells confirmed the epithelialpolarization of these cells.

Antigen expression at the apical cell surfaceThe light microscopical analysis of RL95-2 cells showed a spe­cific fIuoroescence of E-cadherin and integrins at the apicalcell surface. However, confocallight microscopy did not allowto gain more details about distribution within the plane of themembrane or about membrane confinement. Therefore, we

searched for microdomains using whole-mount preparationsand a surface replication technique (Fig. 9).

Antibodies for E-cadherin as weIl as for the integrin sub­units showed an identical staining pattern. Single immunogoldparticles indicating reactivity were evenly distributed over the

I

188 M. Thie, B. Harrach-Ruprecht, H. Sauer, P. Fuchs et 01.

wh oie cell surface at low density. Occasionally, gold particleswere associated in sm all clusters. I-lowever, clusters revealed

no ordered distribution within the plasma membrane. In cor­responding experiments using HEC-l-A cells, virtually noimmunogold particles were detectable (data not shown).

EJCB

Fig. 8. Confoeal images of endometrial monolayers (RL = RL95-2;I-lEC = I-lEC-I-A; AN = AN3-CA) after staining with monoclonalantiboe!y to a6-inlegrin subunil (a-e), ßl-integrin subunit (d-f), anclr-\4-inlegrin subunit (g-i). Verlieal seetions reveal that RL95-2 cells arelabel cd at the entire plasma membrane (a, d, g), while I-IEC-I-A eellsare labelee! al siles of cell-cell conlacls (b, e, h)_ AN3-CA eells showstaining at cell-cell contacts using anti-a6-, ßl-integrin subunits (e, f)and no staining using anti-ß4-integrin subunits; note autofluorescence

of nucleus (i). Arrows mark the position of eell-eell eontaets in mono­layers. - es Coverslip. - Bar 10 [.tm.

Discussion

In the present communication, attachment and invasion­competent chüriocarcinüma cells were used as a probe fürattachment-permissiveness and non-permissiveness of theinvestigated endometrial cell lines. Three main conclusionsare suggested by our results. (i) Expression of apolar pheno­type (= f-IEC-l-A cells) prevents adhesion of invasive cells tothe apical plasma membrane. (ii) Lack of cell polarity (=AN3-CA cells) is not sufficient to allow adhesion. (iii) Expres­sion of adhesion moleeules of the types investigated here is aprecondition of attachment, but these mülecules need to bepresent at the exposed (apical) membrane (= RL95-2 cells),i. e., the adhesion-receptive phenotype is characterized by theexpression of the appropriate molecules in the membrane in anon-polar distribution. By extrapolation, one might postulatethat, in vivo, modulation of the epithelial phenotype of uter­ine cells, i. e. loss üf apical-basal polarity, might prepare theapical cell pole for cell-cell interaction with throphoblast.However, loss of polarity would not lead to receptivity ifaccompanied by a loss of expression of appropriate ac\hesionmolecules.

EJCB Lass af palarized epithelial phenatype 189

Fig. 9. Surfaee replieation of RL95-2 eells after staining with mono­

clonal antibody to aG-integrin subunits. Note single immunogold par­tielcs cvenly distributed ovcr thc wholc ccll surfaee. Oeeasionally, par-

These data extend findings by others who have pointed outthe importance of polar organization of UECs for receptivity/non-receptivity phenomena [10] and who have observedchanges in the apical, lateral and basal plasma membranedomains and in the organization of the cytoskeleton related tothe adhesive behavior of uterine epithelial cells in vivo [6, 22].Moreover, the data reported herc put those findings into anew context. The possibility of modulation of the epithelialphenotype, i. e. loss of apical-basal polarity by random distri­bution of cell-adhesion molendes, like E-cadherin and a6-,

ß1-, ß4-integrin subunits as shown in RL95-2 cells, renders theendometrial cell lines an interesting tool to study control ofadhesiveness in this respect.

As yet, we do not know whether E-cadherin and integrinsinserted into the exposed membrane of RL95-2 cells are

ticlcs are associated in small clusters (arrowheads). - es Coverslip. ­pm Exposcd plasma membrane. - Bar 0.5 [Am.

involved in adhesiveness for trophoblast-type cells as heparansulfate proteoglycans and dermatan sulfate-containing proteo­glycans as well as their corresponding binding sites might par­ticipate in the adhesiveness of RL95-2 cells [29]. E-cadherinwould mediate contact by calcium-dependent homotypic intcr­action 19], whereas integrins mediate both homotypic or het­erotypic interactions and often require the prescncc of plasma/matrix protein [14]. Binding of JAR cells to RL95-2 monolay­crs in serum-free medium could be due to cither homotypic orheterotypic integrin-integrin. interaction, while enhancementof JAR cell attachment to RL95-2 cells by serum suggestsaclhesion via cross-bridging molecules thus pointing to a possi­blc involvcment of integrins. JAR cell attachment to HEC-I­A cells in the presence of serum should be due to J AR cellreceptor molecules binding to serum malendes which in turn

I

190 M. Thie, B. Harrach-Ruprecht, H. Sauer, P. Fuchs et al.

are bound to the HEC-l- A cell surface again via some type ofbin ding site. Failure or proper association of serum moleculeswith the apical surface of AN3-CA cells might lead to a failureof JAR cell attachment to AN3-CA cells. Wh ether an inverse

correlation between cadherin expression and integrin expres­sion which occurs during terminal differentiation of keratino­cytes [16] also exists in RL95-2 cells will have to be investi­gated. In a study of the expression of various integrin subunitsin epithelial cells of human endometrium in vivo [20], a6-, ß4­integrin subunits were reported to be located at the basolater­al surface of the cells in the samples investigated. We havepreliminary data that a6-integrin subunit shows a redistribu­tion at the lateral plasma membrane, i. e. along the apical­basal axis, at transition from the proliferative to the secretoryphase in vivo [36]. Thus, further studies using human samplesshould give new insights into the distribution of integrins andE-cadherin throughout the menstrual cycle.

Current views suggest a hierarchial scheme for the develop­ment of epithelial cell polarity. Extracellular signals generatedby adhesion of the cell to the substratum and/or to neighbor­ing cells initiate bin ding of cell adhesion molecules as weil asjunction formation leading to reorganization and stabilizationof the cytoskeleton. This in turn leads to the repositioning ofcell organelles and the establishment of distinct membranetraffic patterns to nascent plasma membrane domains. Finally,a fully polarized epithelium is formed [28]. Conversely, epi­thelium might undergo transformation to mesenchyme uponinactivation of the epithelial pro gram and activation of mastergene(s) that turn on a catalogue of genes ultimately resultingin the expression of a mesenchymal phenotype [12]. The fea­tures described for RL95-2 cells do not indicate, however,complete epithelium to mesenchyme transformation sinceepithelium-specific cell adhesion molecules are still expressed.On the other hand, these molecules are evenly distributed inall portions of the plasma membrane, i. e. the main anomalyof these cells is the loss of polar membrane organization. Asimilar phenomenon was reported for cultured mammaryepithelial cells which were shown to regulate in a reversibleway distribution of apical and basolateral marker proteinswithout loosing their general epithelial phenotype by short­term activation of c-Fos estrogen receptor fusion proteins [27].

The generation of a restricted pro tein distribution at cellsurface domains in polarized epithelial cells might be regu­lated by both selective targeting and selective retention path­ways [23]. Thus, random distribution of E-cadherin and inte­grins might be due to distinct changes in the regulation of pro­tein sorting and intracellular trafficking, changes in recogni­tion and insertion of proteins in the targeting patch, and/orchanges of selective protein retention. Disassembly of thecomplexes formed between E-cadherin and integrins and theactin filament network might support random distribution ofthese molecules. Cadherins are complexed via their cyto­plasmic domains with intracellular proteins, named a-, ß-, andy-catenin, mediating the connection to the actin filaments[25]. As discussed for the regulation of the cytoplasmicanchorage during development and cell differentiation [18], asimilar mode of modulating the E-cadherin-catenin complexcould also lead to random distribution of E-cadherin in RL95­

2 cells. In analogy, disassembly of the complex of integrinswith actin filaments via talin, vinculin and other cytoskeletal­associated molecules [3] might lead to random distribution ofintegrins. However, altered interactions between the cyto­plasmic domains of trans membrane proteins and intracellular

proteins might impair adhesion function of proteins as shownfor E-cadherin [1, 2, 15]. Whether even distribution of E­cadherin and integrins on the free surface membrane of RL95­2 cells demonstrated by surface replication technique might bedue to disassembled interactions between transmembranemolecules and cytoskeletal proteins and/or selective targetingand retention of molecules remains to be seen.

Acknowledgements. The skillful technical assistance of Birgit Nowakand Dorothea Schünke is gratefully acknowledged. We also wish tothank Dr. J. Behrens, Dr. C. Damsky, and Dr. A. Sonnenberg for thegenerous gifts of antibodies. We are also obliged to Prof. Dr. R. K. H.Kinne (Director of the Department of Epithelial Physiology of theMax-Planck-Institute for Molecular Physiology, Dortmund) for mak­ing available to us the confocal microscopy unit and his interest in thiswork.

References

[1] Behrens, J., L. Vakaet, R. Friis, E. Winterhager, F. van Roy, M. M.Mareei, W. Birchmeier: Lass of epithelial differentiation and gain ofinvasiveness correlates with tyrosine phosphorylation of the E­cadherin/ß-catenin complex in cells transformed with a temperature­sensitive v-SRC gene. J. Cell Bio!, 120,757-766 (1993).[2] Bradley, R. S., P. Cowin, A. M. C. Brown: Expression ofWnt-1 inPC12 cells resu/ts in moduJation of plakoglobin and E-cadherin andincreased celluJar adhesion. J. Cell Bio!, 123, 1857-1865 (1993).[3] Burridge, K., K. Fath, T. Kelly, B. Nuckolis, C. Turner: Focaladhesions: transmembrane junctions between the extracellular matrixand the cytoskeleton. Annu. Rev. Cell Bio!, 4, 487-525 (1988).[4] Cross, R. H. M.: A reliable epoxy resin mixture and its applicationin routine biologicaJ transmission electron microscopy. Micron Mi­crasc. Acta 20,1-7 (1989).[5] Dawe, C. J., W G. Banfield, W D. Morgan, M. S. SJatick, H. O.Curth: Growth in continaus culture, and in hamsters, of cells from aneoplasm associated with acanthosis nigricans. J. Nat. Cancer Inst. 33,441-456 (1964).[6] Denker, H.-W.: TrophobJast-endometriaJ interactions at embryoimplantation: a cell biologicaJ paradox. In: H.-W Denker, J. D. Aplin(eds.): Trophoblast Invasion and Endometrial Receptivity. NoveJAspects of the Cell BioJogy of Embryo ImpJantation. TrophobJastResearch Va!. 4. pp. 3-29. Plenum Medical Book Camp. New York,London 1990.

[7] Denker, H.-W: Implantation: a cell biological paradox. J. Exp.Zoo!. 266, 541-558 (1993).[8] Frixen, U. H., J. Behrens, M. Sachs, G. Eberle, B. Voss, A.Warda, D. Löchner, W. Birchmeier: E-cadherin-mediated cell-celladhesion prevents invasiveness of human carcinoma cells. J. Cell Bio!,113,173-185 (1991).[9] Geiger, B., O.AyaJon: Cadherins. Annu. Rev. Cell Bio!, 8,307-332 (1992).[10] GJasser, S., J. MuJholland: Receptivity is a polarity dependentspecial function of hormonally regulated uterine epithelial cells. Mi­crasc. Res. Techn. 25, 106-120 (1993).[11] Greenburg, G., E. D. Hay: Cytodifferentiation and tissue pheno­type change during transformation of embryonic lens epithelium tomesenchyme-Jike cells in vitro. Dev. Bio!, 115, 363-379 (1986).[12] Hay, E. D.: Extracellular matrix, cell polarity and epitheJial­mesenchym al transformation. In: G. M. Edelman (ed.): MoJecularDeterminants of Animal Form. pp. 293-318. Alan R. Liss, Inc. NewYork 1985.

[13] Hay, E. D.: EpitheJial-mesenchymal transitions. Semin. Dev.Bio!, 1, 347-356 (1990).[14] Hemler, M. E.: Structures and functions of VLA pro teins andrelated integrins. In: J. A. McDonald, R. P.Mecham (eds.): Receptorsfor Extracellular Matrix. pp. 255-300. Academic Press. San Diego1991.

[15] Hinck, L., W J. Nelson, J. Papkoff: Wnt-1 modulates cell-celladhesion in mammalian cells by stabilizing ß-catenin binding to the celladhesion protein cadherin. J. Cell Bio!' 124,729-741 (1994).[16] Hodivala, K. J., F. M. Watt: Evidence that cadherins playa role inthe downregulation of integrin expression that occurs during keratino­cyte terminal differentiation. J. Cell Bio!' 124, 589-600 (1994).[17] John, N. J., M. Linke, H.-W Denker: Quantitation of humanchoriocarcinoma spheroid attachment to uterine epithelial cell mono­layers. In Vitro Cel!. Dev. Bio!' 29A, 461-468 (1993).[18] Kemler, R.: From cadherins to catenins: cytoplasmic pro teininteractions and regulation of cell adhesion. Trends Genet. 9,317-321(1993).[19] Kuramoto, H., S. Tamura, Y.Notake: Establishment of a celilineof human endometrial adenocarcinoma in vitro. Am. J. Obstet. Gyne­co!. ll4, 1012-1019 (1972).[20] Lessey, B. A., L. Damjanovich, C. Coutifaris, A. Castelbaum,S. M. Albelda, C. A. Buck: Integrin adhesion molecules in the humanendometrium. J. Clin. Invest. 90, 188-195 (1992).[21] Mullin, J. M., M. T. McGinn: Effects of diacylglycerols on LLC­PK1 renal epithelia: similarity to phorbol ester tumor promoters. J.Cello Physio!. 134,357-366 (1988).[22] Murphy, C. R.: The plasma membrane of uterine epithelial cells:structure and histochemistry. Progr. Histochem. Cytochem. 27, 1-66(1993).[23] Nelson, W. J.: Regulation of cell surface polarity from bacteria tomammals. Science 258, 948-955 (1992).[24] Osborn, M., K. Weber: Biology of disease. Tumor diagnosis byintermediate filament typing: a novel tool for surgical pathology. Lab.Invest. 48, 372-394 (1983).[25] Ozawa, M., M. Ringwald, R. Kemler: Uvomorulin-catenin com­plex formation is regulated by a specific domain in the cytoplasmicregion of the cell adhesion molecule. ProC. Nat!. Acad. Sci. USA 87,4246-4250 (1990).[26] Potts, J. D., E. B. Vincent, R. B. Runyan, D. L. Weeks: Senseand anti sense TGFß3 mRNA levels correlate with cardiac valve induc­tion. Dev. Dyn. 193, 340-345 (1992).[27] Reichmann, E., H. Schwarz, E. M. Deiner, I. Leitner, M. Eilers,J. Berger, M. Busslinger, H. Beug: Activation of an inducible c-FosERfusion pro tein causes loss of epithelial polarity and triggers epithelial­fibroblastoid cell conversion. Cell 71,1103-1116 (1992).

Lass af palarized epithelial phenatype 191

[28] Rodriguez-Boulan, E., W. J. Nelson: Morphogenesis of the polar­ized epithelial phenotype. Science 245, 718-724 (1989).[29] Rohde, L. H., D. D. Carson: Heparin-like glycosaminoglycansparticipate in binding of a human trophoblastic celiline (JAR) to ahuman uterine epithelial cell line (RL95). J. Cello Physio!. 155,185-196 (1993).[30] Schoenenberger, C. A., K. S. Matlin: Cell polarity and epithelialoncogenesis. Trends Cell Bio!' 1, 87-92 (1991).[31] Schoenenberger, C. A., A. Zuk, D. Kendall, K. S. Mat!in: Multi­layering and loss of apical polarity in Madin-Darby canine kidney cellstransformed with viral K-ras. J. Cell Bio!' 112, 873-889 (1991).[32] Sonnenberg, A., H. J anssen, F. Hogervorst, J. Calafat, J. Hilgers:A complex of platelet glycoproteins Ic and Ha identified by a ratmonoclonal antibody. J. Bio!' Chem. 262, 10376-10383 (1987).[33] Sonnenberg, A., C. J. T. Linders, J. H. Daams, S. J. Kennel: Thea6ß1 (VLA-6) and a6ß4 protein complexes: tissue distribution andbiochemical properties. J. Cell Sci. 96,207-217 (1990).[34] Stephens, L. E., J. E. Sonne, M. L. Fitzgerald, C. H. Damsky:Targeted deletion of ß1 integrins in F9 embryonal carcinoma cellsaffects morphological differentiation but not tissue-specific geneexpression. J. Cell Bio!' 123, 1607-1620 (1993).[35] Takeichi, M.: The cadherins: cell-cell adhesion molecules control­ling animal morphogenesis. Development 102,639-655 (1988).[36] Thie, M., A. Albers, H.-W Denker: Endometrial receptivity: cellbiological aspects. In: T. Mori, T. Aono, T. Tominaga, H. Hiroi (eds.):Perspectives on Assisted Reproduction. Serono Symposia Series Fron­tiers in Endocrinology Vo!. 4. pp. 331-337. Ares-Serono Symposia,Christengraf S. r. !. Rome 1994.[37] Way, D. L., D. S. Grosso, J. R. Davis, E. A. Surwit, C. D. Chris­tian: Characterization of a new human endometrial carcinoma (RL95­2) established in tissue culture. In Vitro 19, 147-158 (1983).[38] Werb, Z., P.M. Tremble, O. Behrendtsen, E. Crowley, C. H.Damsky: Signal transduction through the fibronectin receptor inducescollagenase and stromelysin gene expression. J. Cell Bio!' 109,877-889 (1989).[39] Xiao, G. Q., G. S. Kino: Areal-time confocal scanning opticalmicroscope. Proc. SPIE - Int. Soc. Opt. Eng. 809, 107-113 (1987).[40] Zuk, A., K. S. Matlin, E. D. Hay: Type I collagen gel inducesMadin-Darby canine kidney cells to become fusiform in shape and loseapical-basal polarity. J. Cell Bio!' 108, 903-919 (1989).

I