Vol. 5, 1253-1261, November 1994 Cell Growth & Differentiation 1253

Retinoids Suppress Proliferation, Induce Cell Spreading, andUp-Regulate Connexin43 Expression Only in Postconfluent1 OT��/2 Cells: Implications for the Role of Gap Junctional

S #{149}Communication

Mohammad Z. Hossain2 and John S. Bertram3

Molecular Oncology Program, Cancer Research Center of Hawaii,

Honolulu, Hawaii 96813

Abstract

The antiprohiferative actions of retinoids in the C3H/1 0T#{189}cell system are exhibited as a decrease inproliferation rate and a decreased cell saturation densityat confluence. These actions correlate with up-regulatedgap junctional communication (GJC) driven by theretinoid-induced increased expression of the gapjunctional protein connexin43 (Cx43). Here we examinewhich actions of retinoids occur only in cells makingextensive intercellular contacts, and thus may bemediated through GJC, and which are exhibited in theabsence of extensive intercellular contacts and thus maybe independent of GJC. In confluent cultures, thesynthetic retinoid tetrahydrotetramethylnapthalenyl-propenylbenzoic acid (TTNPB) increased GJC, reducedthe already low [3H]thymidine-labehing index of G1growth-arrested confluent cells from 4.2 to <0.1 %, andincreased the area occupied by each cell by 42%. Incontrast, none of these parameters was altered inlogarithmic growth phase cells with very limitedintercellular contacts. In order to separate cell-cellcontact from cell cycle-related phenomena, non-contacting cells were arrested in early G1 by lovastatin.In this situation, Cx43 expression was low and inducibleby retinoids, as in G1/G0 growth-arrested confluentcells; however, no cell spreading was induced byTTNPB. In contrast, in non-contacting cycling cells or incells arrested by aphidicohin, Cx43 expression washigher than in confluent cells. In this situation, TTNPBdid not induce Cx43 and did not induce spreading.These data demonstrate the cell cycle phase dependenceof connexin43 expression and of retinoid action.Because retinoids only reduce proliferation rates andinduce cell spreading in cells with extensive intercellularcontacts, these data support the involvement of gapjunctional communication in the antiprohiferative effectsof retinoids.

Introduction

The retinoids, natural and synthetic derivatives of retinoicacid, exert a variety of biological effects. Retinoic acid acts

Received 4/i 8/94; revised 8/i 1/94; accepted 8/30/94.1 This work was supported by USPHS Grant CA 39947.2 Present address: Department of Physiology, University of Manitoba,

Winnipeg, Manitoba, Canada R3E 0W3.3 To whom correspondence should be addressed at Molecular OncologyProgram, Cancer Research Center of Hawaii, 1236 Lauhala Street, Universityof Hawaii, Honolulu, HI 96813.

as a morphogen in developmental regulation (1 ), is requiredfor normal growth and cellular differentiation (2), and isproving of clinical value as a cancer preventive agent (3),confirming results in experimental animals and cell cultures(4). In the C3H/1 OTth cell in vitro model of carcinogenesis(5, 6), retinoids are potent inhibitors of neoplastic transfor-mation, reversibly suppressing the development of neoplas-tically transformed foci when applied during the postinitia-tion phase of carcinogenesis (7, 8). The demonstration thatinitiated cells may also be reversibly inhibited from under-going neoplastic transformation when in intimate contactwith normal cells (9) suggested that cell-cell communica-tion may be involved in retinoid action. This hypothesis hasbeen supported by the recent discovery that retinoids dra-matically enhance GJC,4 an action which correlates statis-tically with the ability of tested retinoids to inhibit transfor-mation and enhance growth control (1 0, 1 1). In thishypothesis, retinoids act by stabilizing carcinogen-initiatedcells, through interactions with surrounding normal cells,thus preventing their conversion to fully transformed cells(1 1). The retinoid-enhanced GJC is caused by the up-regu-lated expression of Cx43 at the mRNA and protein level(1 2). Cx43 is the only gap junctional gene known to beexpressed in 1 OT’/2 cells (1 2). These data provide a molec-ular basis for the action of retinoids in 1OT’h cells as cancerpreventive agents.

Gap junctions are water-filled channels which connectadjacent cells of most cell types and serve as conduits forthe transport of small molecules and ions (1 3, 1 4). Evidenceis growing that junctional communication serves manyphysiological functions, among these, the control of growthand differentiation are highly relevant to the process ofcarcinogenesis (1 4). By enhancing junctional communica-tion, retinoids could thus facilitatethe passage of putativegrowth regulatory signals from normal cells to initiatedcells, thereby enhancing growth control and inhibitingtransformation (1 1).

In many cell systems, retinoids are antiproliferative, slow-ing the transit time of cells through the cell cycle (1 5). In the1OTth cell system, however, retinoids do not reduce theproliferation rate of logarithmically growing cells but doreduce the final saturation density of both normal and car-cinogen-transformed cells in a dose-dependent manner(1 6). Since these cells form a confluent stable monolayer atconfluence, the reduction in saturation density is accom-panied by an increase in area occupied by each cell.Thisreduction correlated statistically with the degree of induc-tion of junctional communication (1 7). We equate this ac-tion on saturation density and cell spreading with enhanced

4 The abbreviations used are: GJC, gap junctional communication; Cx43,connexin43; TTNPB, tetrahydrotetramethylnapthalenylpropenylbenzoicacid; C3H/iOT’h, 1OT’h.

1254 Cell Cycle Dependence of Retinoid Action

growth control because both are opposite to the phenotypicchanges that occur during the neoplastic transformation of1OT’/2 cells (6).

These results do not define whether the action of retin-oids on growth control are cell-cell contact dependent orindependent; while effects were seen only in confluentcells, these cells differ from logarithmically cycling cells inboth cell cycle and in cell contact parameters. If theseactions were cell contact dependent, it would implicatejunctional communication as a primary mediator of retinoidaction on growth control. Alternatively, enhanced cellspreading in the absence of intercellular contact wouldindicate that effects on junctional communication and finalsaturation density may only be a secondary consequence ofretinoid action on cell-substrate or cell-cell interactions.Such effects of retinoids have been documented (18). Be-cause of the increasing interest in the use of retinoids ascancer chemopreventive and chemotherapeutic agents (19)and the increasing evidence for the role of junctional com-munication in regulating many physiological processes(20), we have examined in more depth the associationbetween growth control, junctional communication, andretinoid action.

ResultsEffect of Retinoids on Cell Spreading

Retinoid treatment of lOT1h cells decreased the cell satu-ration density at confluence; as expected, this was of ne-cessity accompanied by an increased area occupied byeach cell (cell spreading) so as to maintain a confluentmonolayer of cells. To determine if this increased spreadingoccurred prior to confluence, and was thus responsible fordriving the observed decreases in saturation density, weperformed the following experiments. 1 OT1h cells wereseeded at 10” and 4 x 10� cells per 60-mm dish (i.e., about2 and 80% of the normal saturation density) and weretreated for 72 h with various doses of retinol, retinoic acid,and the synthetic benzoidal retinoid, TTNPB. After thistime, these cultures were logarithmically proliferating orformed a confluent monolayer, respectively. Microscopicexamination of proliferating control cultures showed that,in a population of 400 randomly chosen cells, 60% werewithout direct contact with other cells, 1 6% had contacts inthe form of long thin cytoplasmic processes, usually withon.� other cell, while 22% had contacts, usually with one ortwo other cells, in the form of single or double pseudopo-dia, which occupied only a small fraction of the peripheralcell membrane. Only 2% ofcells had extensive intercellularcontacts (>20% membrane occupied; Fig. 1 , A and B).TTNPB-treated cultures were equivalent. Confluent cultureswere exactly that, with each cell completely surrounded byother such cells (Fig. 1 C). Fig. 1 D shows TIN PB-treatedcultures, where the decrease in cell density is apparent(compare Fig. 1C with 1 D). Note that in both cases (Fig.1,C and 0), cells covered the entire surface of the culturedish.

As shown in Fig. 2A, in sparsely seeded 1OTth cells,retinoids caused no increase in cell area; on the contrary, atendency for a reduction in cell area was observed at lowconcentrations of retinoic acid and retinol. These differ-ences were not, however, statistically significant. The largevariations in cell area observed in both control and treatedlogarithmic cultures were presumably due to size differ-ences within cells distributed throughout the cell cycle of

these proliferating cells. In contrast, treatment of confluentcultures caused large increases in cell area. The highlypotent retinoid TTNPB increased cell area by 41 % at 1O�M, which was the maximum response seen, while bothretinol and retinoic acid caused similar increases but re-quired 1 00-1 000 fold higher concentrations than TTNPB(Fig.2B). Itisinterestingto note that, in previous studies, inwhich inhibition of neoplastic transformation (8) and stim-ulation of junctional communication (21 ) were measured,the retinoids exhibited comparable structure/activityrelationships.

The increase in cell area induced by retinoid treatment ofconfluent cells was accompanied by a large reduction inthe proliferation rate of these cells measured as a decreasein the thymidine labeling index (solvent control, 4.2 ±

0.4%; TTNPB 1 �_8 M, <0.1 %). In contrast, the same treat-ment of proliferating, sparsely seeded cells resulted in noreduction in proliferation rate (solvent control, 19.3 ± 2.8%;TTNPB 1 0_8 M, 18.0 ± 1 .2%; Table 1 ). These data demon-strate that increased cell spreading and reduction in prolif-eration rate are not primary actions of retinoids but requiregrowth-arrested confluent cells. Since these cells also ex-hibit a high degree of cell-cell contact and up-regulate theirexpression of Cx43 in response to retinoids, our resultsindicate that gap junctional communication may be re-quired for the antiproliferative action of retinoids and theirability to cause cell spreading.

The Molecular Action of Retinoids on Induction of Cx43Expression Also Requires Confluent Cells

Our previous studies of retinoid action had only usedconfluent 1 OTth cells because it is in these cells that retin-oids exert their chemopreventive activity (7). In this situa-tion, retinoids elevate junctional communication and en-hance expression of Cx43 (1 2, 21 ). To determine if theproliferation status of 1 OT1h cells influences this molecularresponse to retinoids, we analyzed Cx43 mRNA levels inTTNPB-treated cells when proliferating and when conflu-ent. TTNPB was used because of its high potency and lowtoxicity.As shown in Fig.3, both the basal level of expres-sion and the induction of Cx43 mRNA by TTNPB are de-pendent on the proliferative status of the treated cells. Insparsely seeded, actively proliferating cells, Cx43 expres-sion was higher than that in confluent cells, where Cx43mRNA was below detection levels. Retinoid treatment, asexpected, resulted in a large induction of Cx43 mRNA inconfluent cells but surprisingly did not alter its expression inproliferating cells (Fig. 3). It is important to note that inTTNPB-induced confluent cells levels of Cx43 mRNA, as aproportion of total cellular RNA, were greater than thoseseen in proliferating cells, indicating that the inability ofTTNPB to increase Cx43 mRNA in proliferating cells wasnot due to the Cx43 gene being transcribed at its maximumlevel. To determine whether Cx43 protein exhibited similardifferential regulation as observed for Cx43 mRNA, sparseand dense cultures of 1 OTth cells were treated with TTNPB(1 0_8 M) for 3 days. Cx43 levels were then quantitated byWestern blotting. As expected, proliferating cells expressedhigher basal levels of Cx43 protein than did confluent cells,and this basal level was unchanged by TTNPB treatment(Fig. 4, Lanes 1 and 2). In contrast, basal levels ofCx43 werevery low in confluent 1 OT1h cells, and TTNPB treatmentcaused the expected dramatic increase (Fig. 4, Lanes 3 and4) to levels of Cx43 level approximating those seen in

‘ � ‘f a

4�

Btl�’

‘� V ,, �-. .4.

� I

� �

V 4

Fig. 1. Photomicrographs of 10T1/2 cells. A and B, proliferating cells 72 h after plating sparsely at itT’ cellsl60-mm dish, acetone control; A, low magnification;B, magnification as in (C) and (0); A, a long cytoplasmic process connecting two cells; �, a “pseudopod” connecting two cells; Panels C, D, confluent cells 72hours after plating heavily at 4 x i0� cells/60-mm dish. C, acetone control; 0, after treatment for 3 days with TTNPB i0� si. A, bar, 0.5 mm; B, C, and 0, bar,

0_i mm.

.,. ‘,

A .�

� �

1�. � ,,�. . a,, .

� \i �

.,,-i_ I,.’� ‘ I �. % � �

p. �#{149}#{149}

�, � (fl�.

,-� � �� -“C

‘f �i’ . �, 54

.� � ‘�

�

� � � .

/*

4,

I,

Cell Growth & Differentiation 1255

. . ‘ . . . 4)#{216}./.�

.v� .� , . . --

r � �

p.: � �, .� � �� � � , r � � #{149}�

� , ��t-#{231},�o..‘. � ..,.�g � � .,�, .�-..v. - . � - �

�;:�r�c�4

� , �‘ , ._.w

� .‘ - . ..�. ..t. �- �...,t..

. !‘

�

. .., . . � �

.� ..

proliferating cells. That Cx43 mRNA levels were so differentin the two situations suggests either that translation effi-

ciency was higher in proliferating cells or that Cx43 proteinis more stable in the former situation.

The location of Cx43 was probed by indirect immuno-fluorescence using a polyclonal antibody to the COOH-

�L’

--..:‘� � ‘i’;; %... ‘.-.‘. �:�#{149}r’��5 .

�- . .... � . I “ �‘ �

- ‘�‘ ,, *. .. �i�’‘ ‘. � .. ‘ ‘. . i!’� ‘;-, -.. “ . . . � s.’.. ,

�%. , ‘..‘. - ., . t�1�b:� �

� S � I 3. � � - .

� .., � . .

�::. . � , . .-

� i� �

� .� � .

� S �: ‘ � 54

� . #{149}

terminal region of Cx43. As shown in Fig. 5, in both ace-tone-treated controls (Fig. 5A) and in those treated withTTNPB as above (Fig. 58), cells exhibited a fine punctatepattern of immunostaining over the cytoplasm and as a haloaround the nucleus. Fluorescence over the nucleus was alsoseen; this was in a focal plane above that seen in the highly

[Retinoid]

1256 Cell Cycle Dependence of Retinoid Action

0I-

C00

4-

0

0a)L

a)

0

10�6

Fig. 2. Retinoid effects on cell spreading in proliferating versus confluentcells. iOT’/2 cells were seeded at low and high density (i04 and 4 xcells/60-mm dish, respectively) and treated with various doses ofTTNPB (0),retinol (#{149}),or retinoic acid (�) for 3 days. Cell area was quantitated asdescribed in �Materials and Methods.” A, sparsely seeded proliferating cells;B, confluent cells.

flattened cytoplasm and is thought to represent sites on thenuclear periphery contributing to the halo described above.In these cells, which were seeded sparsely as in the studiesdescribed above, cell-cell contacts were few, and junc-tional plaques of a size similar to those observed previouslyin confluent cells were not seen at sites of contact. Becausethese fibroblasts when at low cell density are highly motile,as exemplified by the formation of widely scattered diffusecolonies of cells (Fig. 1), it seemed probable that cell-cellcontacts were only transient and not conducive to junc-tional assembly. To address this question, other cultureswere seeded at xlO the cell number (i.e., 2.5 x iO� cells/dish) and treated with acetone or TTNPB as before. In thesecultures, cell-cell contacts were more extensive, and atthese sites, the expected junctional plaques were observed.Fig. 5C shows two acetone-treated contacting cells with anumber of junctional plaques. No difference in the forma-tion of junctional plaques was seen in acetone-treated con-trol cells versus TTNPB-treated cells (data not shown),which contrasts greatly with the situation described previ-ously in confluent cultures where plaques are infrequent inconfluent control cultures. Cells labeled with preimmuneserum exhibited a less intense, but nevertheless similar,pattern of cytoplasmic and nuclear fluorescence (Fig. 5D).The major qualitative difference in staining pattern was that,even in densely seeded cultures, junctional plaques werenever observed. This is demonstrated in Fig. 5D, whichshows a region of contact between adjacent cells. Becauseof the need to use an antifade mounting medium for theseslides, phase contrast images were of poor quality and arenot shown.

The presence of a Mr 45,000 protein band was alsoobserved after TTNPB treatment of confluent cells (Fig. 4,Lane 4), but not after treatment of proliferating cells. Wehave previously shown this band to represent a phosphor-ylated form of Cx43 (1 2), suggested to be important in theassembly of Cx43 into junctional plaques (22). Our datasupport this conclusion since plaque formation is extremelylimited in these sparsely seeded cells with few junctions.These molecular data, furthermore, demonstrate that, sim-ilar to the enlargement of cell area (Figs. 1 and 2b), induc-tion of Cx43 expression by TTNPB is observed only inconfluent cells.

Expression and Induction of Cx43 Is Cell CycleDependent

To determine if Cx43 down-regulation is controlled bycell cycle-dependent events or, alternatively, by intercellu-lar contact-dependent events, proliferating cells were ar-rested at various phases of the cell cycle, and Cx43 levelswere examined. To eliminate intercellular contact as a van-able, only sparsely seeded cells were used.

Cell Cycle Arrest by Aphidicolin. Aphidicolin, a fungalmycotoxin (23), inhibits DNA polymenase B and arrestsseveral cell lines at the G1-S boundary ofthe cell cycle (24).To contnol for potential toxicity, we assessed the reversibil-ity of cell cycle arrest after aphidicolin treatment of 1 OT1/2cells. Confluent growth-arrested cells were reseeded andtreated with 2 mM aphidicolin for 24 h. This reduced the[3H]dThD labeling index from 70 to 7.6%. Twenty-four hafter drug withdrawal, the labeling index was 80%, dem-onstnati ng complete reversibility. H ighen concentrations or48-h treatment times caused irreversible arrest (data notshown). To avoid these toxic effects of aphidicolin, 1 0T#{189}cultures were treated with TTNPB (1 0_8 M) for 3 days,followed by 2 mt�.i aphidicolin for the final 24 h beforeharvesting and Western blotting. We would expect thesecells to be arrested in S phase and at the G1-S boundary. Inthese cells comparable Cx43 levels to those seen in activelyproliferating cells were detected and, similar to proliferatingcells, aphidicolin-arrested cells did not exhibit any signifi-cant increase in Cx43 after TTNPB treatment (Fig. 6). Toensure that the Cx43 bands seen in this study were synthe-sized after aphidicolin treatment, additional cultures weretreated with 10 mM cycloheximide for the last 1 0 h of theexperiment. This almost completely abolished Cx43 expres-sion in acetone control or TTNPB-treated cultures (Fig. 6,Lanes S and 6), confirming that cells were still capable ofCx43 synthesis during aphidicolin treatment and demon-strating the rapid turnover of this protein, as has beenreported by others (25, 26). Thus, noncycling, aphidicolin-arrested cells resemble cycling cells in their high level ofexpression and lack of inducibility of Cx43 by TTNPB.

G1 Arrest by Lovastatin. Confluent 1 OTth cells are an-rested in G1-G0 ofthe cell cycle (27) and additionally differfrom cycling cells in their ext�nsive intercellularcontacts.To separate these two vaniables�requires the growth arrest ofsparsely seeded cells in early G1 . Unfortunately, culturingin 1% serum did not result in an acceptable level of cellcycle arrest; we thus used lovastatin, an inhibitor of thecholesterol biosynthetic pathway, which has been shown tocause early G1 arrest in many cell types (24). When sparselyseeded proliferating cells were treated with lovastatin for 24h, the labeling index fell from 1 9.3 ± 2.8% in controls to 0%in treated cells. In lovastatin-arrested cells, the basal level of

Cell Growth & Differentiation 1257

Table 1 Effect of TTNPB on cell area in proliferating versus growth-arrested confluen t cells

Treatment Growth phase[‘HIdThD labeling index

,,� Cell area (pm I

Acetone

TTNPB (i0�� M)

Logarithmic

Logarithmic

19.3±2.8

18.0 ± 1.2

3725±215’

3433 ± 176’

Acetone

TTNPB (1 08 si)

Confluent

Confluent

4.2 ± 0.4

<0.i �.d

4787 ± 1 52�’6789 ± 1 1 3”

Acetone

TTNPB (10�� M)

G,-arrested by lovastatin

G,-arrested by lovastatin

<0.V’

<0.P’

5082 ± 481”

5276 ± 286’

a Sparse 1OT/2 cultures were treated as in Fig. 7. Cell areas were determined by digital image analysis. Values represent means ± SE of about 40 cells chosento be without intercellular contacts.1� Cell area values were taken from Fig. 1.

� Significantly different (P < 0.05) from appropriate acetone controls.d Labeled cells present, but none in the i 000 cells counted.e No labeled cells counted or observed.

1234

-4.4

:;�� � �

Fig. 3. Effect of TTNPB on Cx43 mRNA levels in proliferating versus con-fluent iOT1/2 cells. Sparsely seeded (Lanes 1 and 2) and densely seeded(Lanes 3 and 4( 1OT’/2 cultures were treated with TTNPB (i0’� si) or acetone

as solvent control for 3 days and then harvested. Total RNA (20 pg( wasloaded per lane. Cx43 mRNA was identified with 32P-labeled Cx43 com-plementary DNA. Lanes 1 and 3, acetone control; Lanes 2 and 4, TTNPB

(i0_8 M). Molecular standards are expressed in kilobases.

1234

-�-49.J----

-32

Fig. 4. Induction of Cx43 by TTNPB in confluent but not in proliferating

cells. Cells were cultured and treated as in Fig. 2. Cellular proteins wereelectrophoretically separated on a 10% sodium dodecyl sulfate-polyacryl-

amide gel and subjected to Western blotting using a rabbit anti-Cx43-antibody (1 2). Proliferating cells: acetone (Lane 1); TTNPB 1 0� si (Lane 2);Confluent cells: acetone (Lane 3), TTNPB 1 08 54 (Lane 4). Molecular weightstandards are expressed in kilodaltons.

Cx43 expression was markedly decreased to a level com-parable to confluent G1/G0-arrested cells (Fig. 7). Further-more, TTNPB treatment increased Cx43 expression as inconfluent cells. Although lovastatin-arrested and confluent-arrested cells were similar in respect to Cx43 regulation,they differed when cell spreading was measured. Nospreading was induced by TTNPB in sparsely seeded, by-astatin-arrested cultures where cell-cell contact was denied(Table 1 ), whereas in contrast, as previously demonstrated,TTNPB treatment of confluent cells results in extensive cell

spreading. Furthermore, only under these latter conditionsdid TTNPB act as an antiproliferative agent (Table 1 ). Themajor variable in these two situations is that extensiveintercellular contacts and junctional communication canonly occur in confluent cultures.

Discussion

In the present study, we demonstrate that in C3H/1 OT’/2

cells, certain actions of retinoids, both at the cellular andmolecular level, are cell cycle dependent, while others aredependent on intercellular contact. The expression of Cx43and its induction by TTNPB is clearly cell cycle dependent.In confluent G,-G() growth-arrested cells and in cells ar-rested by bovastatin in early G,, Cx43 expression is low butis strongly inducible by TTNPB (Fig. 7). In contrast, cyclingcells and those growth arrested by aphidicolin express highlevels of Cx43, but this expression cannot be further in-creased by TTNPB (Fig. 6). Since the extent of Cx43 mRNAexpression is lower than that observed in TTNPB-inducedconfluent cells, this lack of response does not appear due toCx43 being transcribed at its maximum level (Fig. 3). Incontrast to these molecular events, the behavioral responsesto retinoids examined appear to require extensive intercel-lular contacts; both the antiproliferative action of retinoidsand their ability to cause enhanced cell spreading wereonly observed in cells allowed to reach confluence and, bydefinition, to form extensive intercellular contacts (Table 1).Here, under conditions of retinoid stimulation, we havepreviously shown by indirect immunofluorescence using ananti-Cx43 antibody that all C3H1 OT1h cells are surroundedby large numbers of junctional plaques (12). We wouldsuggest that these contact-dependent responses are causallyassociated with the ability of retinoids to stimulate gapjunctional communication, which can only occur to itsmaximal extent at confluent cell densities.

The immunofluorescent studies presented in Fig. 5 mdi-cate that the mere possession of Cx43 by cells does notensure the formation of junctional plaques. We did notdetect such plaques in cells seeded at low densities, pre-sumably because of the brief nature of contacts in thesemotile cells, and retinoid treatment did not cause plaqueformation. Only when cells were seeded at higher densitieswere junctional plaques seen of a size previously recordedin confluent cultures (1 2). Here, too, retinoid treatment didnot affect the formation of plaques, suggesting that retinoidsdo not function to influence connexin assembly in this

Fig. 5. Immunofluorescent localization of connexin43. Sparsely seeded cultures of 10T’/2 cells (25,000 cells/dish( were labeled with anti-Cx43 antibody (A andB) or with preimmune rabbit serum ID), then with goat anti-rabbit fluorescein isothiocyanate-conjugated IgG. C, a culture seeded heavily at 250,000 cells/dishand labeled with anti-Cx43 antibody then fluorescein isothiocyanate-conjugated IgG. Arrows, fluorescent junctional plaques. Bar, 10 pm.

1258 Cell Cycle Dependence of Retinoid Action

system . Unfortunately the immunofluorescent micrographsgave little definitive information regarding the localizationof Cx43 when incorporated into junctional plaques. Thiswas due partly to nonspecific cytoplasmic fluorescenceobserved in cells labeled with preimmune serum, andpartly, we would suggest, to difficulties in detection of small

amounts of signal distributed throughout the plasma mem-brane as opposed to being localized into junctional plaquescontaining possibly thousands of connexons (1 3).

This proposed role of gap junctional communication inmediating contact-dependent responses of retinoids stemsfrom our previous observations that modulation of junc-

.49

Cell Growth & Differentiation 1259

I 23456I S

� S

Fig. 6. Effect of cell cycle arrest by aphidicolin on the expression of Cx43.Sparsely seeded 1 0Th cells were treated with TTNPB (10� si) or acetone for3 days. These cells were also treated with aphidicolin or with cycloheximide

during the final 24 or i 0 h of the experiment, respectively. Cx43 levels weredetermined by Western blotting as in Fig. 4. Lane 1, acetone control; Lane 2,TTNPB; Lane 3, acetone + aphidicolin; Lane 4, TTNPB i- aphidicolin; Lane5, acetone + cycloheximide; Lane 6, TTNPB #{247}cycloheximide.

1234

-.49,�

.�-. 32

Fig. 7. Effect of early G, arrest on the expression of Cx43. Sparsely seeded

cells were treated with acetone or TTNPB (10� MI for 3 days and were G1arrested for the final 24 h with lovastatin. Cx43 levels were determined as inFig. 4. Lane 1, acetone control; Lane 2, TTNPB; Lane 3, acetone -s- lovastatin;

Lane 4, TTNPB + lovastatin.

tional communication by retinoids strongly correlates withtheir effects on growth control and neoplastic transforma-tion (10, 21). These observations thus support the originalhypothesis of Loewenstein (28) that gap junctions transfergrowth regulatory signals. The chemical or physiologicalnature of these signals are as yet unknown. However, sinceestablishment or enhancement of junctional communica-tion strongly correlates with decreased cell proliferation(10), while conversely inhibition of communication by tu-mon promoters and certain oncogenes results in decreasedgrowth control (29),we are tempted to speculate that theseputative signals are growth inhibitory in nature. If so, and ifthese signals play a physiological role in growth control,then cells responding to these signals would be expected tobe arrested in G1-G0, as would cells transmitting thesesignals. Experimental evidence in favor of this comes fromthe observation that transformed cells are inhibited incl-Go of the cell cycle when in junctional communication

with confluent (1 7) but not when cocultured with prolifer-ating 1OT’/2 cells (30).

Recently, several groups of investigators have confirmedthese correlative studies by transfecting connexin genes intoneoplastic cells. Stable integration of Cx43 into transformed10T1/2 cells led to enhanced growth control (31) and tosuppression of tumonigenicity in nude mice. Interestingly,the two tumors that did develop in these mice had lost thetransfected gene (32). In other cells, too, transfection ofCx43 has been shown to increase growth control (33, 34).Transfection of Cx32 also reduced tumonigenicity (35).Functional expression of gap junctions thus has many of thecharacteristics expected of tumor suppressor genes; indeed,

a search for such genes in human mammary carcinomacells identified Cx26 (36).

It is known that gap junctional communication can bemodulated by transcriptional and by posttranslational mod-ifications. The present results confirm our previous reportthat retinoids up-regulate expression (1 2). In myometrium,estrogen will also up-regulate expression of Cx43 mRNA,while progesterone antagonizes this increase (37). The mo-lecular basis for this inhibition is not understood; however,the steroids and retinoic acid activate the same family ofreceptors. Sequencing of the genomic up-stream non-cod-ing region of Cx43 has revealed putative estrogen andretinoic acid responsive elements as well as APi and AP2elements (37). The former would explain estrogen and ret-inoid responsiveness; the latter may explain the observedcell cycle regulation of Cx43 expression. APi , a cell cycle-dependent transcriptional complex (38), has been shown toantagonize transcriptional activation of retinoid and glu-cocorticoid responsive genes (39, 40). Interactions betweenAPi and retinoic acid receptors can be expected to bedynamic in cycling cells and could explain much of the cellcycle dependency of Cx43 expression and retinoid respon-siveness. The physiological reason for down-regulation ofCx43 expression in cells growth-arrested in G1-G0 is notapparent at this time. It may be speculated that cells pro-duce an excess of Cx43 while in logarithmic growth phasein order to maximize the potential for establishing GJC uponmaking the initial intercellular contacts; once cells haveachieved extensive communication, this need decreasesand synthesis is reduced. Because this phase of growthcorresponds to confluence and the need for growth control,Cx43 expression may have become regulated by growtharrest in G,-G0. We also must consider that the observeddown-regulation is a consequence of the original criteriaused to select the 1 0T1/2 cells; not only were these cellsrequired to be highly growth inhibited at confluence, theyalso were selected because of their capacity to exhibitneoplastic transformation after treatment with carcinogens(6). If, as others have shown in closely related BALB/3T3cell lines, GJC is not down-regulated at confluence, thentransformation is latent because of the suppressing effects ofsurrounding normal cells (41 , 42). When this down-regu-lation is prevented in 1 OTt/2 cells by retinoid treatment or,alternatively, when carcinogen-initiated cells are placed inintimate contact with normal cells, transformation is alsolatent (9).

We do not know how commonly down-regulation ofCx43 expression occurs; in human dermal fibroblasts, Cx43is expressed at high levels at confluence and is retinoidinducible (43), whereas in human mammary epithelial

cells, Cx43 is expressed at low levels throughout the cellcycle. In these cells, Cx26 is differentially regulated within

the cell cycle but inversely to the regulation of Cx43 re-ported here; expression is low in cells arrested in early G,by lovastatin but increases greatly in S-phase cells (44).Additionally, retinoic acid failed to induce Cx43 in thesehuman mammary epithelial cells (44), which contrasts withits effects in human and mouse fibroblasts. Thus, the regu-lation of connexin genes appears highly dependent uponthe cell type examined.

Posttranslational control of connexin function is believedto occur via extensive phosphorylation of Cx43. This can beobserved on Western blots as multiple phosphoforms of im-munoreactive protein migrating in the Mr 45,00046,000region of the gel (Ref. 22; Fig. 2, Lane 4). In confluent 1OTth

1260 Cell Cycle Dependence of Retinoid Action

cultures, three bands can usually be seen; the most distinct aretwo phosphorylated forms [notated P1 and P2 by Musil andGoodenough (22)], together with an indistinct nonphospho-rylated form in the Mr 42,00043,000 region (1 2). Phos-

phatase treatment of retinoid-treated 1 OT’h cell extracts priorto electrophoresis results in a single band in the predicted M�

42,000-43,000 region (12). It is of interest that only whenTINPB-treated 1OT’h cells were allowed to form extensivecontacts and functional junctions were higher molecularweight forms of Cx43 observed (Fig. 2, Lane 4). Induction ofCx43 by TTNPB in lovastatin-treated, G1-arrested, sparselyseeded cells did not result in the presence of the Mr 45,000Cx43 (presumably phosphorylated) band (Fig. 7, Lane 4).These results imply that phosphorylation ofCx43 is dependenton cell-cell contact which, in communication-competentcells, results in assembly into junctional plaques. Our studiesdid not address the question of the location, membrane orcytoplasmic, of Cx43 in sparsely seeded cells. These observa-tions confirm and extend the reports ofMusil eta!. (22, 25) thatphosphorylation of Cx43 is associated with, and may be func-tionally necessary for, the assembly of Cx43 into functionalgap junctions.

Materials and Methods

Chemicals. Lucifer Yellow CH, retinoic acid, retinol, andaphidicolin were purchased from Sigma Chemical Co.(St. Louis, MO). TTNPB was a gift from Hoffmann-LaRoche(Nutley, NJ). Lovastatin was kindly provided by Dr. A. W.Alberts (Merck, Sharpe and Dohme, NJ).

Cell Culture and Drug Treatment. C3H/1 OT’h cells werecultured in basal Eagle’s medium with 5% fetal bovineserum and gentamicin (25 pg/mI). For studies in sparsecultures, cells were seeded at 2.5% of their final saturationdensity (i.e., 25,000 cells/i 00-mm dish) and treated withretinoids for 3 days, after which cell density was about 10%offinal saturation density (1 00,000 cells/dish). Dense 1 OT’hcultures were obtained by seeding at 80-90% of the finalcell saturation density (i.e., 8-9 x iO� cells/100 mm dish).Retinoids and aphidicolin were dissolved in acetone and

absolute ethanol, respectively, and added directly to theculture medium in a dose volume of 20 pl/l 0 ml of culturemedium. Lovastatin stock solution (1 0 mM) was prepared asdescribed (24). In studies to induce cell cycle arrest withaphidicolin, cells were seeded from a confluent growth-arrested culture and treated for 24 h. This induced a para-synchronous wave of DNA synthesis upon drug with-drawal. The studies with lovastatin used asynchronous cellsplated from logarithmic growth phase cultures.

Quantitation of Cell Area. Digitized images of sparse1 011/2 cells were used for the determination of cell area.The boundary of each cell was identified, and cell area wasquantitated by using software from Loats Associated, Inc.(Westminster, MD). For each treatment group, 25-35 cellswere measured. Since in confluent 1OT1/2 cells, cell bound-aries were not well defined, average cell area in these cellswas determined by dividing the culture dish area by the cellsaturation density measured by Coulter counter. Phase-contrast microscopy revealed that, in confluent cultures,cells occupied the entire surface area (Fig. 1 C).

E3HlThymidine Incorporation Assay. Cells were incu-bated with [3H]thymidine (1-5 mCi/mi) for 2 h (aphidicolin)or for 1 h in studies described in Table 1 (lovastatin) andthen fixed with methanol:acetic acid (3:1, v/v) for 30 mm.Dishes were air dried after washing with methanol, coated

with photographic emulsion (Kodak), and developed after5-7 days of exposure. Only heavily labeled cells (above 50grains/cell) were scored as positive.

Immunofluorescent Analysis. 1OTth cells were seededat a density of 25,000 or 250,000 cells into dishes contain-ing glass slides. They were then treated as described above.Cells were fixed overnight in -20#{176}methanol and then im-mersed for 2 mm in acetone. Labeling with a rabbit poly-clonal anti-connexin43, prepared against the COOH-termi-nal residues 368-382 ofthe predicted sequence of rat Cx43(1 3) at a 1 :40 dilution, followed by goat-anti-rabbit fluores-cein isothiocyanate-labeled IgG (1 :40; Sigma), was as de-scnibed (1 2). All photomicrographs shown in Fig. 5 wereprepared under identical conditions of exposure, develop-ment, and printing.

Analysis of Cx43 mRNA. 1 OT’h cells were harvested incold phosphate-buffered saline/lO mt�t EDTA, and total cel-lular RNA was isolated using a commercially available kit(Cinna-Biotecx Laboratories, Inc., Houston, TX). Equalamounts of total RNA were added to each lane, electro-phoresed on a formaldehyde-agarose gel, and transferredonto a nylon membrane (Schleicher & Schuell, Inc., Keene,NH). Connexin43 message levels were examined by hy-bnidizing with 32P-labeled Cx43 cDNA obtained from theG2A clone kindly provided by Dr. E. Beyer (WashingtonUniversity, St. Louis, MO) as described previously (12).

Analysis of Cx43 Protein. Separation of cellular proteinsby sodium dodecyl sulfate gel electrophoresis, Westernblotting, and identification of Cx43 by a polyclonal anti-Cx43 antibody raised against the COOH-terminal domain(residues 368-382) were performed as described previously(1 2). Equal amounts of protein were added to each lane.

References1 . Brockes, J. P. Retinoids, homeobox genes, and limb morphogenesis.Neuron, 2: 1285-1294, 1989.

2. Mendelsohn, C., Ruberte, E., leMeur, M., Morriss-Kay, G., and Chambon,P. Developmental analysis ofthe retinoic acid-inducible RAR-b2 promoter intransgenic animals. Development, 1 13: 723-734, 1991.

3. Hong, W. K., lippman, S. M., Itri, I. M., Karp, D. D., lee, J. S., Byers,R. M., Schantz, S. P., Kramer, A., Lotan, R., Peters, I. J., Dimery, I. W.,Brown, B. W., and Goepfert, H. Prevention of second primary tumors withisotretinoin in squamous-cell carcinoma of the head and neck. N. EngI.J. Med., 323: 795-800, 1990.

4. Moon, R. C. Comparative aspects of carotenoids and retinoids as chemo-

preventive agents for cancer. J. Nutr., 1 19: 1 27-i 34, 1989.

5. Bertram, J. S. Neoplastic transformation in cell cultures: in vitro,/in-vivo

correlations. IARC Sci. PubI., 67: 77-91, 1985.

6. Reznikoff, C. A., Bertram, J. S., Brankow, D. W., and Heidelberger, C.Quantitative and qualitative studies of chemical transformation of clonedC3H mouse embryo cells sensitive to postconfluence inhibition of celldivision. Cancer Res., 33: 2339-2349, 1973.

7. Merriman, R., and Bertram, J. S. Reversible inhibition by retinoids of3-methylcholanthrene-induced neoplastic transformation in C3H1 OT’/2cells. Cancer Res., 39: 1661-1666, 1979.

8. Bertram, I. S. Structure activity relationships among various retinoids andtheir ability to inhibit neoplastic transformation and to increase cell adhesionin C3H/iOT’/2 Cl8 cell line. Cancer Res., 40: 3141-3146, 1980.

9. Mordan, I. J., Martner, J. E., and Bertram, J. S. Quantitative neoplastictransformation of C3H/1OT’/2 fibroblasts: dependence upon the size of theinitiated cell colony at confluence. Cancer Res., 43: 4062-4067, 1983.

1 0. Mehta, P. P., Bertram, J. S., and loewenstein, W. R. The actions ofretinoids on cellular growth correlate with their actions on gap junctionalcommunication. J. Cell Biol., 108: 1053-1065, 1989.

1 1 . Hossain, M. Z., Zhang, L., and Bertram, J. S. Retinoids and carotenoids

upregulate gap junctional communication: correlation with enhancedgrowth control and cancer prevention. In: J. E. Hall and G. A. Zampighi(eds.), Gap Junctions, Amsterdam: Elsevier Science Publishers By, i �

Cell Growth & Differentiation 1261

12. Rogers, M., Berestecky, J. M., Hossain, M. Z., Guo, H. M., Kadle, R.,Nicholson, B. J., and Bertram, J. S. Retinoid-enhanced gap junctional com-munication is achieved by increased levels of connexin43 mRNA and pro-tein. Mol. Carcinog., 3: 335-343, 1990.

13. Beyer, E. C., Paul, D. I., and Goodenough, D. A. Connexin family ofgap junction proteins. J. Membr. Biol., 116: 187-194, 1990.

14. loewenstein, W. R. The cell-to-cell channel of gap junctions. Cell, 48:725-726, 1987.

1 5. Fontana, J. A., Hobbs, P. D., and Dawson, M. I. Inhibition of mammarycarcinoma growth by retinoidal benzoic acid derivatives. Exp. Cell Biol., 56:254-263, 1988.

1 6. Mordan, I. J., and Bertram, J. S. Retinoid effects on cell-cell interactionsand growth characteristics of normal and carcinogen-treated C3H/1 OT’/acells. cancer Res., 43: 567-571 , 1983.

1 7. Mehta, P. P., Bertram, J. S., and loewenstein, W. R. Growth inhibition of

transformed cells correlates with their junctional communication withnormal cells. Cell, 44: 187-196, 1986.

1 8. Cai, D., Webber, M. M., and De luca, 1. M. Retinoids enhance lectinbinding to gpi3o, a glycoprotein of NIH-3T3 cells: correlation with cellgrowth and adhesion. Exp. Cell Res., 192: 366-372, 1991.

19. lippmann, S. M., Kessler, I. F., and Meyskens, F. I., Jr. Retinoids aspreventive and therapeutic anticancer agents (Part I). Cancer Treat. Rep., 71:391-405, 1987.

20. S#{225}ez,J. C., Berthoud, V. M., Moreno, A. P., and Spray, D. C. Gap junc-tions: multiplicity of controls in differentiated and undifferentiated cells and

possible functional implications. Adv. Second Messenger PhosphoproteinRes., 27: 163-198, 1993.

21 . Hossain, M. Z., Wilkens, I. R., Mehta, P. P., loewenstein, W., andBertram, I. S. Enhancement of gap junctional communication by retinoidscorrelates with their ability to inhibit neoplastic transformation. Carcinogen-esis (lond.), 10: 1 743-1 748, 1989.

22. Musil, I. S., and Goodenough, D. A. Biochemical analysis of con-nexin43 intracellular transport, phosphorylation, and assembly into gapjunctional plaques. J. Cell Biol., 1 15: 1 357-i 374, 1991.

23. Spadari, S., Sala, F., and Pedrali-Noy, G. Aphidicolin: a specific inhibitorof nuclear DNA replication in eucaryotes. Trends Biochem. Sci., 1: 29-32,1982.

24. Keyomarsi, K., Sandoval, I., Band, v., and Pardee, A. B. Synchroniza-tion of tumor and normal cells from G, to multiple cell cycles by lovastatin.CancerRes., 51:3602-3609, 1991.

25. Musil, I. S., Cunningham, B. A., Edelman, G. M., and Goodenough,D. A. Differential phosphorylation ofthe gap junction protein connexin43 in

junctional communication-competent and -deficient cell lines. J. Cell Biol.,1 1 1: 2077-2088, 1990.

26. laird, D. W., Puranam, K. I., and Revel, I. P. Turnover and phosphor-ylation dynamics of connexin43 gap junction protein in cultured cardiacmyocytes. Biochemistry, 273: 67-72, 1991.

27. Bertram, J. S., and Heidelberger, C. Cell cycle dependency of oncogenlctransformation induced by N-methyl-N’-nitro-N-nitrosoguanidine in culture.Cancer Res., 34: 526-537, i974.

28. loewenstein, W. R. Junctional intercellular communication and thecontrol of growth. Biochim. Biophys. Acta, 5�0: 1-65, 1979.

29. Yamasaki, H. Gap junctional intercellular communication and carcino-genesis. Carcinogenesis (Lond.), 1 1: 1051-1058, 1990.

30. Bertram, I. S., and Faletto, M. B. Requirements for and kinetics of growtharrest of neoplastic cells by confluent 1QT’/a fibroblasts induced by a specific

inhibitor of cyclic adenosine 3’ :5’-phosphodiesterase. Cancer Res., 45:1946-1952, 1985.

31 . Mehta, P. P., Hot.z Wagenblatt, A., Rose, B., shalloway, D., andloewenstein, W. R. Incorporation of the gene for a cell-cell channel proteininto transformed cells leads to normalization of growth. J. Membr. Biol.,124: 207-225, 1991.

32. Rose, B., Mehta, P. P., and loewenstein, W. R. Gap-junction proteingene suppresses tumorigenicity. Carcinogenesis (lond.), 14: 1073-1075,1993.

33. Zhu, D., Caveney, S., Kidder, G. M., and Naus, C. C. G. Transfection ofC6 glioma cells with connexin43 cDNA: analysis ofexpression, intercellularcoupling, and cell proliferation. Proc. NatI. Acad. Sci. USA, 88: 1883-i 887,1991.

34. Zhu, D., Kidder, G. M., Caveney, S., and Naus, C. C. G. Growth retar-dation in glioma cells cocultured with cells overexpressing a gap junctionprotein. Proc. NatI. Acad. Sci. USA, 89: 10218-10221, 1992.

35. Eghbali, B., Kessler, J. A., Reid, L. M., Roy, C., and Spray, D. C. Involve-ment of gap junctions in tumorigenesis: transfection of tumor cells withconnexin32 cDNA retards growth in vivo. Proc. NatI. Acad. Sci. USA, 88:10701-10705, 1991.

36. Lee, S. W., Tomasetto, C., and Sager, R. Positive selection of candidatetumor-suppressor genes by subtractive hybridization. Proc. NatI. Acad. Sci.

USA, 88:2825-2829, 1991.

37. lye, S. J., Nicholson, B. J., Mascarenhas, M., and Petrocelli, T. Increasedexpression of connexin-43 in the rat myometrium during labor is associatedwith an increase in the plasma estrogen:progesterone ratio. Endocrinology,

132: 2380-2386, 1993.

38. Yu, W., DahI, G., and Werner, R. The connexin43 gene is responsive tooestrogen. Proc. R. Soc. Lond. B. Biol., 255: 125-132, 1994.

39. Rauscher, F. J., Ill, Voulalas, P. 1., Franza, B. R., Jr., and Curran, T. Fosand Jun bind cooperatively to the AP-1 site: reconstitution in vitro. GenesDev., 2: 1687-i699, 1988.

40. Nicholson, R. C., Mader, S., Nagpal, S., leid, M., Rochette-Egly, C., andChambon, P. Negative regulation of the rat stromelysin gene promoter byretinoic acid is mediated by an APi binding site. EMBO J., 9: 4443-4454,1990.

41 . SchUle, R., Rangarajan, P., Yang, N., Kliewer, S., Ransone, L. J., Bolado,J., Verma. I. M., and Evans, R. M. Retinoic acid is a negative regulator of AP-1responsive genes. Proc. NatI. Acad. Sci. USA, 88: 6092-6096, 1991.

42. Katoh, F., and Yamasaki, H. Regulation of gap-junctional intercellularcommunication in transformation-sensitive and transformation-resistantBAIB/c 3T3 cell variants. Carcinogenesis (lond.), 12: 1923-1926, 1991.

43. Yamasaki, H., Enomoto, T., Shiba, Y., Kanno, Y., and Kakunaga, T.Intercellular communication capacity as a possible determinant of transfor-mation sensitivity of BAIB/c 3T3 clonal cells. Cancer Res., 45: 637-641,

1985.

44. Guo, H., Acevedo, P., Parsa, D, F., and Bertram, J. S. The gap-junctionalprotein connexin43 is expressed in dermis and epidermis of human skin:differential modulation by retinolds. J. Invest. Dermatol., 99: 460-467,1992.

45. lee, S. W., Tomasetto, C., Paul, D., Keyomarsi, K., and Sager, R. Tran-

scriptional downregulation of gap-junction proteins blocks junctional com-munication in human mammary tumor cell lines. J. Cell Biol., 1 18:1213-1221, 1992.