The effects of dimethyl sulfoxide and retinoic acid on the cell growth and

the phenotype of ovarian cancer cells

T. W. GRUNT1-*, C. SOMAY1, M. PAVELKA2, A. ELLINGER2, E. DITTRICH1 and Ch. DITTRICH1

^Laboratory for Applied and Experimental Tumor Cell Biology, Division of Oncology, Department of Internal Medicine I, and^Institute for Micromorphology and Electron Microscopy, University of Vienna, Vienna, Austria

* Author for correspondence

Summary

We have compared the in vitro effects of thedifferentiation inducers dimethyl sulfoxide (DMSO)and retinoic acid (RA) on a polyclonal humanovarian cancer cell line (HOC-7). Density gradientfractionation of untreated cells reveals that a pro-portion of rapidly growing, polygonal cells withmedium density is capable of spontaneous reversioninto a slowly growing low-density phenotype withflattened morphology similar to non-transformedhuman ovarian surface epithelial cells. Clonal expan-sion of these low-density cells proves that theobserved characteristics are stable for prolongedculture periods. Exposure of HOC-7 cells to DMSOand RA or removal of the serum from the medium iseffective in enhancing the proportion of these low-density cells. Application of DMSO causes the cells tobecome flattened and elongated, and to develop rod-like protrusions. In these cytoplasmic extensionsthick filament bundles are dominant. Immunofluor-escence studies demonstrate that both untreatedlow-density subclones and DMSO-treated polyclonalcells are much more reactive for cytokeratin than

medium-density subclones or untreated parentalcells. Furthermore, immunocytochemistry and fixed-cell ELISA reveal 2- to 5-fold greater amounts ofdesmoplakins I and II and of fibronectin in low-density subclones and in DMSO-treated cells ascompared to medium-density subclones and controlcultures. RA exerts weaker effects on the phenotypeof the cells. Both inducers reduce DNA synthesis andinhibit the anchorage-dependent and the anchorage-independent cell growth in a dose- and time-depen-dent manner. The restoration of the original mor-phology and growth rate after removal of thedifferentiation-inducing agents proves that the ob-served changes are reversible; this indicates that thecells do not become terminally differentiated. Thedescribed phenotypic and growth kinetic reactionsshould prove useful for the future recognition ofinduced maturation of ovarian cancer cells.

Key words: dimethyl sulfoxide, retinoic acid, ovarian cancercells.

Introduction

Malignant growth may be considered to be a result ofgenetic alterations enabling the cells to escape from thestringent control of proliferation and differentiation.Recent reports describing the induced differentiation ofvarious tumor cell lines indicate, however, that malignantcells may have retained at least some genes that controlproliferation and differentiation and that the transformedphenotype does not necessarily represent an irreversiblestate (Dexter et al. 1979; Hoosein et al. 1988; Langdon et al.1988; Lotan, 1979; Marks et al. 1986, 1989; Marth et al.1989; Maurer and Echarti, 1988; Meyskens and Salmon,1979; Shapiro and Poon, 1979). In such cases induction ofdifferentiation causes reversion of the malignant cells to amore benign phenotype, indicating that induced differen-tiation may become an alternative approach for themanagement of cancer patients (Lotan et al. 1990; Sachs,1987). This is supported, for instance, by the evidence thatpatients with ovarian carcinoma of high differentiationexhibit improved survival compared to patients bearingtumors of lower differentiation (Baak et al. 1986; McGuireet al. 1990; Schild et al. 1989).

Journal of Cell Science 100, 657-666 (1991)Printed in Great Britain © The Company of Biologists Limited 1991

Having previously demonstrated that the proportion of(undifferentiated) clonogenic cells in ovarian cancer cellsuspensions correlates with colony formation in soft agarand with the clinical prognosis of the patients (Dittrich etal. 1990, 1991), we subsequently isolated such cells fromdifferentiated cell subpopulations using discontinuousdensity gradient centrifugation of the HOC-7 cell line(Grunt et al. 1991). Differentiated large cells with lowdensity and with slow growth on solid substrata and in softagar were separated from undifferentiated cells withsmaller volume and higher density. The latter cells grewfaster and had higher clonogenicity in soft agar. Subclonesestablished from both cell populations were then analysedfor the formation of intermediate filaments (cytokeratin),intercellular contact structures (desmosomal proteins)and extracellular matrix elements (fibronectin). Sub-sequently, we speculated whether treatment of HOC-7cells with dimethyl sulfoxide (DMSO) or retinoic acid (RA)- two well-known inducers of cell differentiation - couldinduce cell phenotypes, which were already found with lowfrequency in the untreated cell line, and which have to beattributed to residual spontaneous differentiation pro-cesses. The phenotype of the treated cells was therefore

657

compared with that of the subclones isolated from theuntreated parental line. Special emphasis was placed onthe comparison of the induced phenotypes with data fromthe literature describing in vitro phenotypes of normalhuman ovarian surface epithelial cells, which representthe natural source of all common ovarian carcinomas(Auersperg et al. 1984; Siemens and Auersperg, 1988).

The phenotypic criteria established by such investi-gations may provide a basis for the future recognition ofdifferentiation-inducing effects during in vitro screeningfor new effective antitumor drugs. Such studies may,therefore, establish new guidelines and reference modelsfor further investigations dealing with possible growthinhibitors and/or differentiation inducers in ovariancarcinoma.

We demonstrate that DMSO and RA reduce monolayergrowth of HOC-7 cells reversibly and in a dose- and time-dependent manner. Colony growth in soft agar and DNAsynthesis are similarly inhibited. DMSO-treated cells andspontaneously differentiated subclones reveal a similarflattened morphology and develop cytoplasmic extensions.Both populations are characterised by increased amountsof cytokeratin, desmoplakin and fibronectin, indicatingthat the observed phenomena might be related to elevatedepithelial differentiation.

Materials and methods

Cell culture and treatmentThe well-differentiated, polyclonal ovarian adenocarcinoma cellline HOC-7 was a generous gift from Dr R. N. Buick (OntarioCancer Institute, Toronto, Canada) (Buick et al. 1985). The cellsmaintained in a-MEM (Gibco, Scotland) with 10% heat-inacti-vated fetal calf serum (FCSHI, Gibco) were incubated in ahumidified 5% CO2 atmosphere at 37 °C. They were refed after 4days and passaged weekly for a maximum period of 30 weeks.Tests for mycoplasma contamination were negative (DAPI,Boehringer Mannheim).

Dimethyl sulfoxide (Sigma, FRG) was diluted directly withgrowth medium, whereas stock solutions of fi-&\\-trans RA(Sigma) were prepared in 95% ethanol and stored at — 70 °C inlight-protected envelopes. For culture these stock solutions werediluted at a ratio of 1:1000 with medium containing the cellsuspension. Dose-response curves were established by plating2xlO4 cells in triplicate in 35 mm plastic Petri dishes (Greiner,Austria) in 2 ml standard medium containing varying concen-trations of inducers. Positive controls comprised cultures grown instandard medium without any additives or in medium containingthe solvent only (0.1 %, v/v, of 95% ethanol). Four days later thecells were trypsinised (0.5 mg trypsin, 0.2 mg EDTA, Gibco) andcounted in a Neubauer hemocytometer.

Trypan blue exclusion revealed that concentrations of up to1.2% DMSO (v/v) and 10 /IM RA did not compromise the cellviability significantly (Table 1). Most of the subsequent exper-

Table 1. Percentage of viable cells determined by trypanblue exclusion before and after 7 days of exposure of the

cells to the inducers

Substance

DMSO (%)

RA(/IM)

Medium

Dose

0.51.01.2

110

-

Viable cells (%)(mean±s.E.)

94±888±589±4

95±288±1

92±2

Statistical significance was determined by Student's

P value

>0.05>0.05>0.05

>0.05>0.05

-

(-test (ra=3).

iments were therefore performed with =Sl% DMSO andRA, which was markedly below the doses used in otherinvestigations dealing with induction of ovarian cancer celldifferentiation (Langdon et al. 1988). Determination of the proteincontent, performed according to the method of Lowry et al. (1951)revealed that both treated and untreated cells contain similaramounts of cellular protein (approximately 6;<g/lxlO4 cells).

Fractionation of untreated HOC-7 cells in adiscontinuous density gradientThe procedure for cell fractionation of untreated HOC-7 cells wasperformed as described previously (Grunt et al. 1991). Briefly,nine different Percoll density solutions (Pharmacia, Sweden)were layered carefully (1 ml of each) in a centrifuge tube. One mlof the cell suspension (1.5 xlO7 cells) was then centrifuged in thisgradient and 10 distinct cell fractions, formed at the interphasesof the density media, were recovered.

Clonogenic assay in soft agarThe tumor cells were cultured in a modified single-layer soft agarsystem. Four thousand cells were resuspended in lml cv-MEMcontaining 0.3% Bacto-Agar (Difco, MI, USA), 15% FCSHI,200i.u.ml~1 penicillin, 200/igml"1 streptomycin, lmM L-gluta-mine and various concentrations of DMSO or RA and were platedin triplicate in 35 mm bacteriological plastic Petri dishes(Greiner). Positive controls comprised cultures grown in standardmedium without DMSO or with the RA solvent only. On the day ofcultivation the plates were examined for the presence ofaggregates prior to the incubation in a humidified 5% CO2atmosphere at 37 °C for 3 weeks. For evaluation of the assays cellaggregates ^60 [an in diameter were defined as colonies.

In vitro morphologyCellular morphology and growth pattern were examined aftertreating the cells for 7 days. The media (± inducers) were renewedafter 4 days of culture. Using a Leitz Fluovert microscope wecompared the induced cell phenotypes with untreated controlcultures and with growth-arrested cells (after 48 h of serumdeprivation) or with subclones. The subclones were derived fromcell fractions that were originally obtained by discontinuousdensity gradient centrifugation of untreated HOC-7 cells followedby limiting dilution culture of these cells, i.e. fractionateduntreated cells were plated at a statistical density of 0.5 cell perwell in 96-well plates (Grunt et al. 1991).

For electron microscopy, the cells were seeded on glasscoverslips and incubated in standard medium as well as inmedium supplemented with 1 % DMSO or 10 ,UM RA. The mediumwas renewed after 4 days of culture. On day 7, the cells were fixedin situ by exposure to 2.5% phosphate-buffered glutaraldehydefor 2 h at 4°C. After an overnight rinse in 0.2 M sodium phosphatebuffer, the cells were postfixed in 1 % veronal acetate-bufferedOsO4 for 2h at 4°C, dehydrated in a graded series of ethanol andembedded in Epon. Thin sections were stained in ethanolic uranylacetate and alkaline lead citrate and examined in a Philips EM400 transmission electron microscope. For scanning electronmicroscopy the postfixed cells were dehydrated in an ascendingethanol series. The cell monolayer was dried from liquid CO2 bythe critical point drying procedure and sputtered with gold.Photomicrographs were taken with a Cambridge Stereoscan 90scanning electron microscope.

Growth kineticsCell growth curves were established by plating in triplicate1x10 cells in 35 mm plastic Petri dishes (Greiner) in 2 mlstandard medium without additives or with 1 % DMSO or with10 ,UM RA. Media were renewed on days 3, 7 and 10. The cellnumbers were determined on days 1, 2, 3, 4, 7, 10 and 14 and thedoubling times were calculated.

Parallel cultures treated for 5 days were returned to standardmedium without inducers and then cultured for an additional 10days. The cell numbers of these cultures were then compared withthose obtained after continuous exposure to the inducers for 15days.

658 T. W. Grunt et al.

For determination of DNA synthesis, 1 x 104 cells suspended in0.1 ml of standard medium were seeded into each well of a 96-welltissue culture plate (Costar, MA, USA). Twelve hours later thecells were washed and incubated for an additional 12 h in a-MEMwithout nucleosides (Gibco) supplemented with 10% FCSHI,lOOi.u.ml"1 penicillin, 100 fig ml"1 streptomycin, 10mM Hepes,2 mM L-glutamine, various concentrations of DMSO or RA and0.5 ;*Ci/ well [me%Z-3H]thymidine (specific activity: 74 Cimmol"1, Amersham, UK). Subsequently the cells were washedtwice with Dulbecco's phosphate-buffered saline (Gibco) andextracted 3 times (lxlOmin, 2x5min) with ice-cold 10%trichloracetic acid. The acid-insoluble material was then solu-bilised with 0.3 M NaOH/2% sodium dodecyl sulfate. A samplewas transferred into the scintillant (Aquassure, New EnglandNuclear, UK) and counted in a Packard Tri-Carb A3255 liquidscintillation counter. Parallel wells of each experimental groupwere trypsinised and counted and the estimated activities werenormalised to lxlO4 cells. Under these conditions untreatedcultures gave a mean activity (±S.E.) of 65838±3144ctsmin~1

per 104 cells, which was defined as 100±5% of thymidineincorporation.

ImmunocytochemistryAll antibodies were diluted in Dulbecco's phosphate-bufferedsaline (Gibco) containing 2.2% bovine serum albumin (Behring,FRG). Mouse monoclonal antibody to desmoplakin I and II (cloneDP 1&2-2.15, Boehringer Mannheim, FRG) was used in a dilutionof 1:20 (1/igmr1). Purified rabbit polyclonal anti-human fibro-nectin was purchased from Dakopatts (Denmark) and applied in adilution of 1:300 (33 j /gmr1) . Staining was done with a 2- or 3-step immunoperoxidase technique using peroxidase-conjugatedrabbit anti-mouse IgG (1:100, Dakopatts) or goat anti-rabbit IgG(1:100, Jackson Immunoresearch Lab., PA, USA) and peroxidase-conjugated swine anti-rabbit IgG (1:200, Dakopatts). Acetone-fixed cells (10min, +4°C) were incubated with each of theantibodies for 30 min in a humidified atmosphere at roomtemperature. Rinsing between the incubations was performedwith Dulbecco's phosphate-buffered saline (3x10 min). The colorwas then developed with 0.06% diaminobenzidine (Fluka,Switzerland) and 0.03% H2O2. Finally, the cells were counter-stained with hematoxylin. Cytokeratin was detected by immuno-fluorescence using purified ascites from the pan-reactive KL1monoclonal antibody (Immunotech S.A., France) in a dilution of1:40 (5/igmr1). After labelling with FITC-conjugated rabbitanti-mouse IgG (1:200, Jackson Immunoresearch Lab.) the cellswere embedded in Mowiol (Hoechst, Austria) and viewed in aLeitz Fluovert microscope. For negative controls the firstantibodies were substituted by non-immune IgG (JacksonImmunoresearch Lab.) diluted to the same concentrations as thespecific antibodies.

Fixed-cell ELISAAll antibodies were diluted in Dulbecco's phosphate-bufferedsaline containing 2.2% bovine serum albumin. For detection ofdesmoplakins I and II the cells were grown for 8 days in standardmedium or in medium supplemented with 0.8 % or 1 % DMSO orlOjtM RA. On day 7 the cells were transferred to 96-well plates(0.5xl04 to 4xl04 cells per well) and allowed to adhere for 24h.For detection of fibronectin the cells were seeded in 96-well platesat a density of 0.5 x 104 to 5x 104 per well and incubated for 4 daysin standard medium or in medium containing the inducers.Several wells of each experimental group were trypsinised andcounted in a hemocytometer.

Wells that were intended for the ELISA were fixed for 7 minwith ice-cold methanol, dried for 20 min in a desiccator andincubated for 2h at 37 °C in Dulbecco's phosphate-buffered salinesupplemented with 5 % bovine serum albumin in order to saturateprotein binding sites on the plastic. The cells were then incubatedfor 1 h at room temperature with non-immune rabbit serum (1:10,Dakopatts), followed immediately by mouse monoclonal anti-desmoplakins I and II diluted 1:80 (0.25 ̂ gml"1; 'specific wells')or by non-immune mouse IgG diluted to the same IgGconcentration ('background wells'). After washing with Dulbec-co's phosphate-buffered saline containing 0.1% Tween 20 the

wells were incubated with peroxidase-labelled rabbit anti-mouseIgG (1:1000). For fibronectin assays the incubation in non-immune rabbit serum was omitted. The background wells werefirst incubated with an excess of unlabelled rabbit anti-humanfibronectin (1:60, 167/(gml"1, Dakopatts). Subsequently towashing, these wells were exposed to the same antibody (1:2500,4/igml"1, Dakopatts), which was now conjugated to peroxidase.The specific wells were incubated with the labelled antibody only(1:2500). All incubations with antibodies were performed for30 min at room temperature in a humidified atmosphere.

Bound conjugate was then detected with O^mgml"1

O-phenylenediamine (Dakopatts), 0.05% hydrogen peroxide in50 mM citrate buffer. The enzyme reaction was terminated after30 min by addition of 50//I 1 M H2SO4. The absorbance wasdetermined in an Anthos ELISA reader 2001 at 492 nm. Meanvalues (±S.E.) of triplicate estimations were calculated. Thespecific net absorbance was determined by subtracting the meanoptical density of the background wells from that of the specificwells. These values were then normalised to lxlO4 cells. Eachexperiment was performed in quadruplicate.

Results

MorphologyHOC-7 cells grown on solid substrata in culture mediumsupplemented with 10 % FCSHI exhibit a regular, polyg-onal shape (Fig. 1A). Treatment with 1 % DMSO causescellular flattening. The cells become elongated anddevelop long rod-like extensions (Fig. IB). Transmissionelectron microscopy reveals thick filament bundles pro-truding into these cellular extensions (Fig. 1G). Scanningelectron micrographs (Fig. 1H) confirm the DMSO-induced alterations in cell shape and demonstrate thecomplicated growth pattern. Similar reactions, althoughto a much lower degree, are observed after treatment with10 ,UM RA or under serum-free culture conditions(Fig. 1C-D). Using limiting dilution culture we estab-lished several distinct cell subclones from cell fractionsobtained after density gradient centrifugation ofuntreated HOC-7 cells. Spontaneously differentiatedclones originating from low-density fractions are formedby slowly growing, large and flattened cells, which exhibitreduced saturation density, while medium density frac-tions give rise to undifferentiated subclones consisting ofsmaller, regular polygonal cells with high growth rates(Fig. 1E-F) (Grunt et al. 1991). Immunofluorescencerevealed only weak cytokeratin-specific staining inuntreated parental cells and in undifferentiated sub-clones. DMSO-treated cells and spontaneously differen-tiated subclones, however, demonstrate strong reactivityfor cytokeratin (Fig. 1I-L).

During the first 72 h of exposure the morphologicalalterations proceed until they reach a certain state, andthen remain stable. However, treatment of the cells for 5days followed by transfer back to standard medium resultsin a gradual reversion to the original morphology, which iscompleted after 72 h. These cellular responses occur inidentical fashion in unfractionated, polyclonal HOC-7cells and in undifferentiated clones from the mediumdensity fraction of the original cell line.

Growth kineticsDose-response relationships and cellular growth curvesreveal that DMSO and RA inhibit the anchorage-dependent and the anchorage-independent cell growth in aconcentration- and time-dependent manner. The ED5Qvalues (concentrations inducing half-maximal effects) forgrowth reduction on solid substrata are 0.24 % DMSO and

Induced differentiation in ovarian cancer 659

1 * - * » * *

660 T. W. Grunt et al.

Fig. 1. Cell morphology of monolayer cultures of HOC-7 cells. (A-F) Phase-contrast micrographs. Bars, 50/an. (A) Polyclonalculture grown for 7 days in standard medium. (B,C) Polyclonal culture treated for 7 days with medium containing 1 % DMSO (B)or 10 im. RA (C). (D) Polyclonal culture grown for 4 days in standard medium. After rinsing the cells were cultivated for 2 days instandard medium lacking serum. (E) Undifferentiated clone established from the medium-density fraction. (F) Differentiated clonefrom the low-density fraction of the gradient. (G) Transmission electron micrograph demonstrating thick filament bundles, whichprotrude into a rod-shaped extension of a cell treated for 7 days with 1 % DMSO. Bar, 1 /an. (H) Scanning electron micrograph ofcells treated for 7 days with 1% DMSO. Note the complicated growth pattern and the multiple cellular protrusions. Bar, 20 fan.(I-L) Immunofluorescent micrographs revealing cytokeratin expression. Bars, 20 /an. (I) Polyclonal culture, (J) undifferentiatedclone and (K) differentiated clone grown for 7 days in standard medium. (L) Polyclonal culture treated for 7 days with 0.8 %DMSO.

21 fiM RA, and for inhibition of clonogenic growth in softagar are 1.4% DMSO and 0.1 / M RA (Fig. 2A-C). Cellgrowth curves established from monolayer cultures(Fig. 3) were used to calculate the cell doubling time afterexposure to DMSO or RA. DMSO (at 1 %), for instance,causes an extension from 14.5 h in the control culture to33.6 h, while 10/<M RA reveals weaker effects prolongingthe doubling time to just 17.7 h. These values, obtainedbetween day 1 and day 4 of treatment, are in goodagreement with the doubling times of the untreatedsubclones, which range between 14-20 h for the undiffer-entiated and 37-38 h for the spontaneously differentiatedclones. Continuous exposure of HOC-7 cells to 2=0.5%DMSO or 2=7 ,UM RA induces constant growth reductionand lowers the saturation density. Removal of the agentsafter 5 days of treatment followed by 10 days of cultivationin standard medium, however, demonstrates that theinhibitory effects are completely reversible (Fig. 4).

Radioactive thymidine labelling reveals that DMSOcauses a significant reduction of DNA synthesis, while theeffects of RA do not reach statistical significance (Fig. 5).

ImmunophenotypeImmunocytochemistry for the detection of the desmosomalproteins desmoplakins I and II demonstrates a delicatestaining of the plasma membrane of untreated and RA-treated cells. Exposing the cells to DMSO, however, causesa striking increase in the membrane labelling(Fig. 6A-C).

Untreated cultures incubated with an antibody againstthe matrix glycoprotein fibronectin reveal a moderatecytoplasmic staining reaction and fine granular extra-cellular depositions. A similar pattern was found aftertreatment of the cells with RA. Incubation of the cells with

DMSO, however, resulted in a pronounced cytoplasmicreaction and in extensive extracellular depositions(Fig. 6D-F).

For semi-quantitative evaluation we developed a fixed-cell ELISA system and found that DMSO-treated cells andspontaneously differentiated clones contain 2- to 5-foldhigher levels of desmoplakin and 2-fold greater amounts offibronectin compared to untreated parental cells andundifferentiated subclones (Table 2). Exposure of the cellsto RA, however, did not significantly change the signalsfor desmoplakin or for fibronectin (P>0.05) (Figs 7 and 8).

Discussion

Previous work from our group (Grunt et al. 1991) wasfocused on the phenotypic heterogeneity seen in thepolyclonal HOC-7 cell line, which is derived from a well-differentiated adenocarcinoma of the ovary (Buick et al.1985). Centrifugation of the cells in a density gradient

Table 2. Semi-quantitative estimation of the levels ofdesmoplakins I and II and fibronectin by fixed-cell

ELISA

CellsDesmoplakins I+II Fibronectin(mean absorbance±s.E. per 104 cells)

ABCD

0.035+0.0040.178±0.0210.035±0.0020.181±0.013

0.190±0.0070.534±0.0610.140±0.0090.347±0.028

Parental untreated (A) and treated (0.8 % DMSO, B) polyclonal HOC-7 cells and untreated HOC-7 subclones (undifferentiated, C;differentiated, D) were seeded at various densities in 96-well plates. Themean (±s.E.) absorbance at 492 nm was normalised to lxlO4 cells.

Induced differentiation in ovarian cancer 661

0.001 0.01 0.1 1 10Concentration

•DMSO(%) -

100 1000 1 1.5

Concentration (%)

-e-DMSO

0.001 0.01 0.1 1 10Concentration

-<y- R A

1000

Fig. 2. Dose-response curves for the DMSO- and RA-inducedinhibition of the anchorage-dependent (A) and the anchorage-independent growth (B,C) of HOC-7 cells. (A) 2xlO4 cells were platedin 35 mm dishes and grown for 4 days in the presence of varying dosesof DMSO or RA. Controls reached mean cell counts (±S.E.) of42±6xlO4 (mean value equivalent to 100%). (B,C) 4xl0 3 cells wereplated in soft agar in 35 mm dishes and grown for 3 weeks in thepresence of varying doses of DMSO (B) or RA (C). Controls reachedmean colony counts (±S.E.) of 250±28 (mean value equivalent to100 %). Each point represents the mean value of triplicate estimations.The coefficient of variation was always less than 15 %. Data from 1 outof 3 separate experiments are given.

separates low-density fractions with flattened mor-phology, slow growth on plastic and low clonogenicity insoft agar from medium-density fractions with polygonalcell shape, rapid growth on plastic and high clonogenicpotential in agar. Expansion of subclones on solidsubstrata proves the clonal stability of the low-densityphenotypes, while the medium-density clones undergosome degrees of phenotypic diversification. We attributethis clonal heterogeneity to a residual capacity of themedium-density cells to differentiate spontaneously, lead-ing thereby to a low-density phenotype. Support for thishypothesis comes from the evidence that exposure ofunseparated HOC-7 cells to differentiating agents likeDMSO induces a similar low-density phenotype. Sinceundifferentiated medium-density clones respond to DMSO

in the same way as unseparated, polyclonal HOC-7 cells(including reversibility), we suggest that the observedeffects do not occur because of clonal selection caused byeradication of undifferentiated cells. The data presented inthis paper indicate that the low-density cell phenotype isindeed associated with a more differentiated state of thecells.

Dimethyl sulfoxide is a well-known differentiationinducer with polar-planar structure. Lyman et at. (1976)found that these compounds cause membrane stabilisationleading to decreased fluidity and increased microviscosityof the membrane (Tapiero et al. 1980). The ultimatemechanism of polar-planar inducers is still unknown.They are probably not targeted just to one specificmolecular structure. Retinoic acid, on the other hand,

662 T. W. Grunt et al.

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

U

-*-Untreated - o - 1% DMSO --*- 10/ZM RA

Fig. 3. Time course of the DMSO- and RA-induced inhibitionof the anchorage-dependent growth of HOC-7 cells. 1 x 104 cellswere plated in 35 mm dishes and grown for various times instandard medium or in medium containing 1 % DMSO or 10 f(MRA. Each point represents the mean value of triplicateestimations. The coefficient of variation was always less than15 %. Data from 1 out of 3 separate experiments are given.

140-

130-

120-

1.10-

100-

90-

80-

70-

60-

50-

40-

30-

20-

10-

0-0 15 5 15 5

Days of incubation with inducer

M Untreated I^Sl % D M S O ^ 10;/M RA CZ1 S.E.

Fig. 4. Reversibility of the DMSO- and RA-induced inhibitionof the anchorage-dependent growth of HOC-7 cells. lxlO4 cellswere continuously exposed to the inducers for 15 days.Alternatively, the cells were refed after 5 days of treatmentwith standard medium and cultured for an additional 10 days.100% represents the mean (±S.E.) cell number of the controlcultures (0 days) after 15 days of cultivation (360(+32)xl04).Statistical analysis was performed by the one-tailed U-test forunpaired samples according to Mann-Whitney-Wilcoxon (n&4;*P<0.025).

0.1 0.2 1 1 10Concentration

^Untreated OTDMSO(%)^RA (^M)CDS.E.

Fig. 5. Dose dependence of the DMSO- and RA-inducedinhibition of DNA synthesis. Labelling of the cells withtritiated thymidine is expressed as % mean ctsmin"1 (±S.E.)per 104 cells compared to 104 control cells (100±5%).Statistical analysis was performed by the one-tailed U-test forunpaired samples according to Mann-Whitney-Wilcoxon*P<0.025).

binds to the nuclear RA receptor. The reported half-life ofRA in culture is 3.5 h and its clearance from the medium iscompleted after 24h (Ng et al. 1989). The demonstratedresponses to RA (e.g. reduced clonogenicity in soft agar)may, therefore, be caused by distinct long-term effects ofRA. In our system RA was found to reduce the cell growth,but without altering the immunophenotype of the cells.

Reports on the cultivation of normal human ovariansurface epithelial cells are rather sparse. Available data(Auersperg et al. 1984; Siemens and Auersperg, 1988),however, indicate a close similarity between the in vitromorphology and the growth pattern of nonmalignantovarian cells and the differentiated malignant cellsdescribed in this investigation, supporting the hypothesisthat the induced phenotype does indeed represent a morebenign cell type.

The cells could not be triggered, however, to reach aterminal state of differentiation, which is demonstrated bythe reversibility of the observed responses.

In a further step we looked for possible markers of(ovarian) cell differentiation. Immunophenotyping revealsthat spontaneously differentiated clones and DMSO-induced cells contain greater amounts of cytokeratin,desmoplakin and fibronectin than undifferentiated cells.Electron microscopy demonstrates that the cytoplasmicextensions, which are characteristic of differentiated cells,contain thick filament bundles. Furthermore, semi-quan-titative estimations using a fixed-cell ELISA systemreveal that differentiated clones and DMSO-treated cellscontain 2- to 5-fold more desmoplakin and fibronectin thanundifferentiated clones and untreated parental cells. Allthese effects are independent of the serum lot. Futureinvestigations should clarify whether ovarian cancer celldifferentiation correlates with a specific pattern of distinctcytokeratin-polypeptides.

Induced differentiation in ovarian cancer 663

Fig. 6. Immunocytochemical localisation of desmoplakins I and II (A—C) and fibronectin (D—F) in cultures of HOC-7 cells grown instandard medium (A,D) or in medium containing 10 ; M RA (B,E) or 1 % DMSO (C,F). Bars, 20 fan.

664 T. W. Grunt et al.

0.30.6!

0 •>

Number of plated cells (x 10 )

* Untreated o I % DMSO x 10/IM RA— ,-=0.998 —,-=0.967 --- ,=0.901

Fig. 7. Semi-quantitative estimation of the cellular levels ofdesmoplakins I and II by fixed-cell ELISA. Various numbers ofuntreated cells or of cells exposed to DMSO or RA wereallowed to adhere to the bottom of 96-well plates and werelabelled with antibodies. Each point represents the meanabsorbance at 492 nm of triplicate estimations. The coefficientof variation was always less than 15 %. Data from 1 out of 4separate experiments are given.

Experimental evidence indicates that the cell shape,which is modulated in vitro by the extracellular matrix(Watt, 1986), has crucial importance in the regulation ofcell proliferation (Folkman and Moscona, 1978), differen-tiation and gene expression (Ben Ze'ev, 1985a). Accordingto results from Boyd and co-workers (1988) undifferen-tiated colon cancer cells can be stimulated to differentiate,if they are plated on extracellular matrix depositions,which originate from well-differentiated colon carcinomacells. It is known that fibronectin promotes cell attach-ment and spreading (Hynes and Yamada, 1982; Ruoslahti,1984). Stimulation of extracellular matrix elements ispresumed, therefore, to be involved in inhibition of growthand in induction of maturation processes (Spiegelman andGintz, 1983; Terranova et al. 1984).

According to Ben Ze'ev (19856), slowly dividing cells canelaborate increased cell-cell contacts by the formation ofdesmosome-like intercellular junctions, which may stimu-late cytokeratin synthesis. Ultrastructural studies fromPapadaki and Beilby (1971) demonstrate that the lateralborders of normal human ovarian surface epithelial cellsare connected by a multitude of well-developed desmo-somes. These data strongly suggest that the expression ofdesmosomal proteins is related to epithelial cell differen-tiation. However, we cannot discern whether the increasesin cytokeratin, desmoplakin and fibronectin exert specificfunctions in differentiation or simply represent a manifes-tation of this process. Nevertheless, the phenomenadescribed in this paper elucidate differentiation-associ-ated processes occurring in ovarian cancer cells and mayprovide a basis for future in vitro screening assays, whichshould be designed for the detection of differentiation-inducing effects of novel antitumor therapeutics.

^Unt rea ted 10/IM RA CD s.E.

Fig. 8. Semi-quantitative estimation of the levels offibronectin by fixed-cell ELISA. The cells were seeded atvarious densities in 96-well plates and incubated for 4 days instandard medium or in medium containing the inducers. Themean (+S.E.) absorbance at 492 nm was normalised to lxlO4

cells. Statistical analysis was performed by the one-tailed Li-test for unpaired samples according to Mann-Whitney-Wilcoxon (n»4; **P<0.005).

This work was supported by grants from the Jubilaeumsfondsder Oesterreichischen Nationalbank, the Medizinisch-Wissen-schaftlicher Fonds des Buergermeisters der BundeshauptstadtWien, the Kamillo Eisner Stiftung, the Oesterreichische Gesell-schaft fuer Chemotherapie, and from the Anton-DreherGedaechtnisschenkung fuer Medizinische Forschung.

References

AUERSPERG, N., SIEMENS, C. H. AND MYRDAL, S. E. (1984). Humanovarian surface epithelium in primary culture. In Vitro 20, 743-755.

BAAK, J. P. A., WISSE-BREHELMANS, E. C. M, LANGLEY, F. A.,TALERMAN, A. AND DELEMARRE, J. F. M. (1986). Morphometric data toFIGO stage and histological type and grade for prognosis of ovariantumours. J. din. Path. 39, 1340-1346.

BEN ZE'EV, A. (1985a). Cell shape, the complex cellular networks, andgene expression: cytoskeletal protein genes as a model system. CellMuscle Motil. 16, 23-53.

BEN ZE'EV, A. (19856). Cell-cell interaction and cell configurationrelated control of cytokeratin and vimentin expression in epithelialcells and fibroblasts. In Intermediate Filaments: Annals of the NewYork Academy of Science, vol. 455 (ed. E. Wang, D. Fischman, R. K.H. Liem and T.-T. Sun), pp. 597-613. New York: New York Academyof Science.

BOYD, D., FLORENT, G., CHAKRABARTY, S., BRATTAIN, D. AND BRATTAIN,M. G. (1988). Alterations of the biological characteristics of a coloncarcinoma cell line by colon derived substrata material. Cancer Res.48, 2825-2831.

BUICK, R. N., PULLANO, R. AND TRENT, J. M. (1985). Comparativeproperties of five human ovarian adenocarcinoma cell lines. CancerRes. 45, 3668-3676.

DEXTER, D. L., BARBOSA, J. A. AND CALABRESI, P. (1979). N,N-dimethylformamide-induced alteration of cell culture characteristicsand loss of tumorigenicity in cultured human colon carcinoma cells.Cancer Res. 39, 1020-1025.

DITTRICH, CH., DITTRICH, E., SEVELDA, P., HUDEC, M., GRUNT, TH.,SALZEB, H. AND ELIASON, J. (1991). Clonogenic growth in vitro - a newprognostic parameter for ovarian carcinoma. J. clin. Oncol. 9,381-388.

DITTRICH, CH., DITTRICH, E., WRBA, F., HUDEC, M., GRUNT, TH., SALZER,H., SEVELDA, P. AND ELIASON, J. (1990). Analysis of factorsinfluencing clonogenic growth in vitro of cells from ovarian carcinomapatients. Cancer Lett. 50, 183-189.

Induced differentiation in ovarian cancer 665

FOLKMAN, J. AND MOSOONA, A. (1978). Role of cell shape in growthcontrol. Nature 273, 345-349.

GRUNT, TH. W., DITTRICH, E., SOMAY, C, WAGNER, TH. AND DITTRICH,CH. (1991). Separation of clonogenic and differentiated cell phenotypesof ovarian cancer cells (HOC-7) by discontinuous density gradientcentrifugation. Cancer Lett. 58, 7-16.

HOOSEIN, N. M., BRATTAIN, D. E., MCKNIGHT, M. K., CHILDRESS, K. E.,CHAKRABARTY, S. AND BRATTAIN, M. G. (1988). Comparison of theantiproliferative effects of transforming growth factor-/?, N,N-dimethylformamide and retinoic acid on a human colon carcinoma cellline. Cancer Lett. 40, 219-232.

HYNES, R. 0. AND YAMADA, K. M. (1982). Fibronectins: multifunctionalmodular glycoproteins. J. Cell Biol. 95, 369-377.

LANCDON, S. P., HAWKES, M. M, HAY, F. G., LAWRIE, S. S., SCHOL, D.J., HILGERS, J., LEONARD, R. C. F. AND SMYTH, J. F. (1988). Effect ofsodium butyrate and other differentiation inducers on poorlydifferentiated human ovarian adenocarcinoma cell lines. Cancer Res.48, 6161-6165.

LOTAN, R. (1979). Different susceptibilities of human melanoma andbreast carcinoma cell lines to retinoic acid-induced growth inhibition.Cancer Res. 39, 1014-1019.

LOTAN, R., FRANCIS, G. E., FREEMAN, C. S. AND WAXMAN, S. (1990).Differentiation therapy (meeting report). Cancer Res. 50, 3453-3464.

LOWRY, O. N., ROSEBROUGH, N. J., FARR, A. L. AND RANDALL, R. J.(1951). Protein measurement with the Folin phenol reagent. J. biol.Chem. 193, 265-275.

LYMAN, G. H., PREISLER, H. D. AND PAPAHADJOPOULOS, D. (1976).Membrane action of DMSO and other chemical inducers of Friendleukaemic cell differentiation. Nature 262, 360-363.

MARKS, M. E., ZIOBER, B. L. AND BRATTAIN, M. G. (1986). N,N-Dimethylformamide-induced synthesis of an anti-fibronectin reactiveprotein in cultured human colon carcinoma cells. Cancer Res. 46,5248-5258.

MARKS, P. A., BRESLOW, R., RIFKIND, R. A., NGO, L. AND SINGH, R.(1989). Polar/apolar chemical inducers of differentiation oftransformed cells: Strategies to improve therapeutic potential. Proc.natn. Acad. Sci. U.S.A. 86, 6358-6362.

MARTH, CH., HELMBERG, M., MAYER, I., FUITH, L. C, DAXENBICHLER, G.AND DAPUNT, O. (1989). Effects of biological response modifiers onovarian carcinoma cell lines. Anticancer Res. 9, 461-468.

MAURER, H. R. AND ECHARTI, C. (1988). In vitro cytostatic drugscreening of leukemic cells using clonogenic and differentiationassays. In Drugs, Cells and Cancer: Annual report of the E.O.R.T.C.

Clonogenic Assay Screening Study Group (ed. Ch. Dittrich and M. S.Aapro), p. 16. Vienna: G. Welley.

MCGUIRE, W., TANDON, A. K., ALLRED, D. C, CHAMNESS, G. C. ANDCLARK, G. M. (1990). Commentaries: How to use prognostic factors inaxillary node-negative breast cancer patients. J. natn. Cancer Inst. 82,1006-1015.

MEYSKENS, F. L., JR AND SALMON, S. E. (1979). Inhibition of humanmelanoma colony formation by retinoids. Cancer Res. 39, 4055-4057.

NG, K. W., HUDSON, P. J., POWER, B. E., MANJI, S. S., GUMMER, P. R.AND MARTIN, T. J. (1989). Retinoic acid and tumour necrosis factor-tvact in concert to control the level of alkaline phosphatase mRNA. J.molec. Endocrin. 3, 57-64.

PAPADAKI, L. AND BEILBY, J. O. W. (1971). The fine structure of thesurface epithelium of the human ovary. J. Cell Sci. 8, 445-465.

RUOSLAHTI, E. (1984). Fibronectin in cell adhesion and invasion. CancerMetast. Rev. 3, 43-51.

SACHS, L. (1987). Cell differentiation and bypassing of genetic defects inthe suppression of malignancy. Cancer Res. 47, 1981-1986.

SCHILD, S. E., MARTENSON, J. A., JR, GUNDERSON, L. L., ILSTRUP, D. M.,BERG, K. K., O'CONNELL, M. J. AND WEILAND, L. H. (1989).Postoperative adjuvant therapy of rectal cancer: an analysis of diseasecontrol, survival and prognostic factors. Int. J. Radiat. Oncol. Biol.Phys. 17, 55-62.

SHAPIRO, S. S. AND POON, J. P. (1979). Retinoic acid-induced alterationsof growth and morphology in an established epithelial line. Expl CellRes. 119, 349-357.

SIEMENS, C. H. AND AUERSPERG, N. (1988). Serial propagation of humanovarian surface epithelium in tissue culture. J. cell. Physiol. 134,347-356.

SPIEGELMAN, B. M. AND GINTZ, C. A. (1983). Fibronectin modulation ofcell shape and lipogenic gene expression in 3T3-adipocytes. Cell 35,657-666.

TAPIERO, H., FUORCADE, A. AND BILLARD, C. (1980). Membrane dynamicsof Friend leukaemic cells. II. Changes associated with celldifferentiation. Cell Differ. 9, 211-218.

TERRANOVA, V. P., WILLIAMS, J. E., LIOTTA, L. A. AND MARTIN, G. R.(1984). Modulation of the metastatic activity of melanoma cells bylaminin and fibronectin. Science 226, 982-985.

WATT, F. M. (1986). The extracellular matrix and cell shape. Trendsbiochem. Sci. 11, 482-485.

(Received 7 May 1991 - Accepted, in revised form, 28 July 1991)

666 T. W. Grunt et al.