SUPPRESSED GAP JUNCTIONAL INTERCELLULAR COMMUNICATION INCARCINOGENESIS OF ENDOMETRIUMTsuyoshi SAITO

1* , Makoto NISHIMURA1, Ryuichi KUDO

1 and Hiroshi YAMASAKI2,3

1Department of Obstetrics and Gynecology, Sapporo Medical University, Sapporo, Japan2Unit of Multistage Carcinogenesis, International Agency for Research on Cancer, Lyon, France3School of Science, Kwansei Gakuin University, Uegahara, Nishinomiya, Japan

To examine whether and at which stage of endometrialcarcinogenesis decreased connexin expression occurs, we in-vestigated changes in the expression of the gap junctionproteins, connexin 26 (Cx26), Cx32 and Cx43, in humanendometrial hyperplasia and cancer samples. Forty-eight en-dometrial tissue samples (15 endometrial hyperplasias and33 endometrial cancers) were subjected to immunofluores-cence and RT-PCR analysis. In endometrial hyperplasia, Cx26was aberrantly expressed in all samples as revealed immu-nohistochemically. There was weak or negative expression in12 samples (80.0%) and diffuse expression in cytoplasm in 3samples (20.0%). Cx32 expression in those samples was sim-ilar to that of Cx26; there was weak or negative expression in11 samples (73.3%) and diffuse expression in 4 samples(26.7%). In endometrial cancer, Cx26 was expressed weaklyor negatively in 25 samples (75.8%), diffusely in 6 samples(18.2%) and normally in 2 samples (6.1%), while Cx32 wasexpressed weakly or negatively in 26 samples (78.8%), dif-fusely in 5 samples (15.2%) and normally in 2 samples (6.1%).It was confirmed that weak staining of Cx26 and Cx32 wasdue to poor expression of their mRNA. All samples showedweak Cx43 protein expression as revealed by immunohisto-chemical analysis. In the majority of samples, concomitantexpression levels of Cx26 and Cx32 protein were observed,confirming our long-term hypothesis that Cx26 and Cx32 areboth abnormally regulated in a coordinated fashion in theendometrium. Our results indicate that during endometrialcarcinogenesis, loss of gap junctional intercellular communi-cation (GJIC) may occur due to the suppressed expressionand the aberrant localization of connexin at relatively earlystages.© 2001 Wiley-Liss, Inc.

Key words: ovarian steroid hormone; immunohistochemistry; con-nexin; cell adhesion molecule; HRT

Endometrial cancer is the most frequently diagnosed gyneco-logical malignancy, and the incidence is increasing in industrialnations. Endometrial hyperplasia is apathologic condition of theendometrium known for reflecting hyperestrogenism, as well asbeing a precursor of endometrial carcinoma.1 Though it is widelyaccepted that endogenous and exogenous sources of unopposedestrogen increase the risk of endometrial adenocarcinoma, andseveral molecular alterations have been identified, the molecularpathogenesis of endometrial cancer remains poorly understood.

Gap junctionsare intercellular channels that directly connect thecytoplasm of the neighboring cells and allow exchanges of lowmolecular weight (less than about 1,000 dalton) metabolites, inor-ganic ionsand other small hydrophilic moleculesbetween thecellsin contact. Second messengers in signal transduction, such ascyclic AMP, Ca21 and inositol trisphosphate, can pass throughgap junction channels. Therefore, gap junctional intercellular com-munication (GJIC) is considered to play an important role in thecontrol of cell growth, differentiation, the maintenance of ho-meostasisand morphogenesis. Thegap junction channelsarecom-posed of hexagonal arrangements of oligomeric proteins calledconnexins (Cxs). It has been demonstrated that GJIC can beregulated by different factors such as growth factors, oncogenes,Ca21, pH and hormones.2

Since carcinogenesis involves adisturbance of homeostasis andcancer cellsshow uncontrolled growth, it is considered that alteredGJIC plays an important role in carcinogenesis. Several lines of

evidence suggest that a disturbance of GJIC facilitates the clonalgrowth of potential cancer cells and Cx genes may act as tumorsuppressors.3–6 Several reports have demonstrated that connexinexpression isdecreased in precancerous lesions.7–9For example, inthe rat liver, chemically induced preneoplastic and neoplasticregions, such as enzyme-altered foci, hyperplastic nodules andhepatocellular carcinomas, which are known to express variousaltered phenotypes, show obvious decreases in levels of Cx32mRNA and immunohistochemically demonstrated protein.8 In an-other study, Temme et al.9 reported that male and female 1-year-old mice deficient for Cx32 had 25-fold more and 8-fold morespontaneous liver tumors than wild-type mice, respectively, andincorporation of bromodeoxyuridine (BrdU) into the liver washigher for Cx32-deficient mice than for wild-type mice, conclud-ing that loss of Cx32 protein from hepatic gap junctions is there-fore likely to cause enhanced clonal survival and expansion ofmutated (initiated) cells, which results in a higher susceptibility tohepatic tumors.9 However, knowledge about possible change inexpression of connexin and their role in human carcinogenesis isstill limited.2,10–15

Estrogen causes cell proliferation in vitro and enhances tumorformation in animals exposed to carcinogens, and the doses com-monly used for estrogen replacement therapy are sufficient tocausehyperplasiaand carcinomaof theendometrium.16 Numerousepidemiological studieshaveshown that estrogens, in combinationwith progestins, clearly protect against endometrial cancer,16,17

however, littl e is known about how progestins act biologically toreduce the cancer risk. In a previous study, we showed thatconnexin 26 and 32 were expressed in endometrial epithelium andthat their expression is up-regulated in the presence of estradiolplus progesterone but down-regulated when only estradiol ispresent, suggesting that Cx gene expression acts as a tumor sup-pressor for endometrial carcinogenesis.18 In this study, we ana-lyzed the expression of connexin 26, 32 and 43 in 16 cases ofendometrial hyperplasia and 32 cases of endometrial cancer byimmunohistochemistry and RT-PCR assay in order to examine thecorrelation between connexin expression and endometrial carcino-genesis.

Abbreviations: Cx, connexin; GJIC, gap junctional intercellular commu-nication; HRT, hormone-replacement therapy.

Grant sponsor: Scientific Research from the Ministry of Education ofJapan; Grant numbers: 09470363, 11671638.

*Correspondence to: Department of Obstetrics and Gynecology, Sap-poro Medical University, S-1, W-16, Chuo-ku, Sapporo 060-0061, Japan.Fax: 181-11-614-0860. E-mail: tsaito@sapmed.ac.jp

Received 16 October 2000; Revised 8December 2000, 31 January 2001;Accepted 2 February 2001

Published online 17 May 2001

Int. J. Cancer: 93, 317–323 (2001)© 2001 Wiley-Liss, Inc.

Publicationof the InternationalUnionAgainstCancer

MATERIAL AND METHODS

Patients and samplesSamples of endometrial tissues were obtained from 48 women

who had undergone hysterectomy or curettage at the SapporoMedical University Hospital. Biopsy samples were obtained ac-cording to institutional guidelines (university hospital), and in-formed consent was obtained from patients. Endometrial hyper-plasia tissues (n515) were taken from the endometrial curettageand diagnosed according to the system of the World Health Or-ganization. As a result, 6 cases were found to be simple hyperpla-sia, 5 were endometrial hyperplasia complex and 4 were atypicalendometrial hyperplasia. Endometrioid adenocarcinoma tissues(n533) were taken from modified radical hysterectomy, salpingo-oophorectomy or selective pelvic lymphadenectomy with para-aortic lymphadenectomy. The endometrial carcinomas were alsograded according to the system of the World Health Organization.This resulted in 17 cases of tumor-grade G1 (well-differentiatedadenocarcinoma), 10 cases of G2 (moderately differentiated ade-nocarcinoma) and 4 cases of G3 (poorly differentiated adenocar-cinoma). The clinical features of the endometrial cancer samplesare summarized in Table I. The samples were fixed in 10%buffered formalin for immunohistochemistry and some of themwere frozen and kept at280°C until the analysis.

Two monoclonal and 2 polyclonal antibodies against syntheticpeptides were used: 1 monoclonal anti-Cx26, 1 monoclonal antiCx32, 1 polyclonal anti-Cx32 and polyclonal anti-Cx43 (1:500dilution; Zymed, San Francisco, CA). The frozen tissues were cutinto 6mm thick slices, mounted on albumin-coated slides and fixedwith the cold acetone. The fixed tissue sections were preincubatedwith a blocking solution (PBS containing 5% skimmed milk) for30 min at room temperature, incubated with the first antibodies for2 hr and washed in PBS. FITC-conjugated anti-mouse immuno-globlin diluted 1:200 in PBS was then added for Cx26 and Cx32,and anti-rabbit immunogloblin (Dakopatts, Copenhagen, Den-mark) diluted 1:200 in PBS was added for Cx43 polyclonal anti-

bodies; all were incubated for 1 hr. For the double staining,specimens were incubated with the first antibodies, monoclonalanti-Cx26 and polyclonal anti-Cx32, and then with the secondantibodies, FITC-conjugated goat anti-rabbit immunogloblin andrhodamine-conjugated goat anti-mouse immunogloblin. After in-cubation with secondary antibodies, the slides were washed inPBS, mounted by fluorescent mounting medium (Dakopatts) andexamined by immunofluorescent microscopy (Nikon, Tokyo, Ja-pan). As a positive control, normal endometrium in the secretoryphase was used because the endometrium is differentiated, thegrowth is arrested in the secretory phase during the menstrualcycle, and we have shown that Cx26 and Cx32 are well expressedand localized on cell-cell border.18 For each tissue sample, theresults of immunostaining were grouped into “weak/negative” and“positive” based on the intensity at the cell-cell border or “diffuse”when the staining was seen mainly in cytoplasm.

RNA isolation and RT-PCR analysisTo verify the presence of specific mRNAs of Cx26, Cx32 and

Cx43, we amplified them by RT-PCR and the GAPDH gene wasamplified as a control. Total RNA of the tissues was extracted bya single-step technique with TRIzol Reagent (GIBCO-BRL, Rock-ville, MD) according to the manufacturer’s protocol. The qualityof the RNA samples was determined by electrophoresis throughagarose gels and staining with ethidium bromide, and 18S and 28SRNA bands were visualized under UV light. Five micrograms oftotal RNA was denatured at 65°C for 10 min and incubated at 36°Cfor 60 min in RT buffer containing random primers, deoxynucle-otide triphosphates (dNTPs), RNAase inhibitors and avian myelo-blastosis virus RT (Takara, Tokyo, Japan) in a final volume of 20ml, followed by boiling for 5 min. One microliter of each RTreaction mixture was applied to 25ml of PCR mixture, containing2.5U AmpliTaq DNA polymerase (Takara), 1.5 mmol/L MgCl2,13 Taq buffer and 0.2 mmol/L each of 4 dNTPs. The specificprimers of Cx26, Cx32, Cx43 and GAPDH used for PCR areshown in Table II. Thirty-eight cycles of PCR were carried outwith a program of 30 sec at 94°C, 1 min at 58°C and 1 min at 72°C.Aliquots of the PCR products were electrophoresed on 2.5%agarose gel. mRNA from the normal endometrium of the secretoryphase was used as a positive control.

RESULTS

Expression and localization of connexins in endometrialhyperplasia

The expression and localization of Cx26, Cx32 and Cx43 wereanalyzed using frozen sections of 15 endometrial hyperplasiasamples, 6 cases of simple hyperplasia, 5 cases of endometrialhyperplasia complex and 4 cases of atypical endometrial hyper-plasia. Though Cx26 was weakly detected in the proliferativephase (Fig. 1a), it was detected as small fluorescent spots on thelateral membrane of the endometrial epithelium in normal endo-metrial epithelium in the secretory phase (Fig. 1b). Cx32 wascoordinately expressed and localized in the endometrial epithelium(data not shown). In contrast, as shown in Table III, in the 15hyperplasia samples, for example, Cx26 was expressed aberrantlyin all samples, weakly or negatively in 12 samples (80.0%) and

TABLE I – BACKGROUND OF THE ENDOMETRIAL CANCER SAMPLES

No. Age Clinical stage Histological grade Association with hyperplasia

1 49 Ia G1 Complex2 53 IIIc G1 No3 60 Ia G1 Complex4 51 Ib G1 Complex5 54 Ic G3 Simple6 46 Ib G2 Atypical7 58 IIb G1 No8 60 Ia G1 No9 66 Ic G2 No10 45 Ib G1 No11 59 Ic G1 No12 46 IIIa G1 No13 57 IIIc G2 No14 48 Ib G1 Complex15 61 Ia G1 No16 61 IIIc G2 No17 71 Ia G3 No18 63 Ia G1 No19 58 Ic G1 No20 63 Ib G2 No21 70 IIIc G2 No22 82 Ia G3 No23 62 IIIc G2 No24 56 IIIa G2 No25 47 Ib G1 Simple26 50 Ia G2 Unknown27 48 IV G3 No28 56 Ib G1 Complex29 62 IIb G2 No30 56 Ib G1 Complex31 56 Ic G2 No32 48 Ib G1 No33 48 Ib G2 Atypical

TABLE II – THE NUCLEOTIDE SEQUENCE OF THE SPECIFIC PRIMERS FORCx26, Cx32, Cx43 AND GAPD

Name Sequence (59-39)

Cx26 Forward: 59-GCTGCAAGAACGTGTGCTAC-39Reverse: 59-TGGGTTTTGATCTCCTCGAT-39

Cx32 Forward: 59-ACCAATTCTTCCCCATCTCC-39Reverse: 59-AAGACGGCCTCAAACAACAG-39

Cx43 Forward: 59-AGGAGTTCAATCACTTGGCG-39Reverse: 59-GCAGGATTCGGAAAATGAAA-39

GAPDH Forward: 59-GAGTCAACGGATTTCGTCGT-39Reverse: 59-GGTGCCATGGAATTTGCCAT-39

318 SAITO ET AL.

diffusely in cytoplasm for 3 samples (20.0%) (Fig. 2). The immu-noreactivity to Cx32 in these samples was similar to that observedfor Cx26: weak or negative for 11 samples (73.3%), diffuse for 4samples (26.7%). Cx43 was detected in fibroblasts but not in theepithelial cells as observed in the normal endometrium (Fig. 1c andresults not shown).

In our study, we analyzed the localization and expression ofconnexins in endometrial hyperplasia and cancer by immunofluo-rescence. In the case of normal localization, it is sufficient to

evaluate the intensity of expression by immunofluorescence be-cause the normally localized connexin shows a clear dot shape. Onthe other hand, when there is aberrant localization, especiallydiffuse localization in the cytoplasm, sometimes it is difficult toevaluate the expression by the immunohistochemical method.Therefore, to support the immunohistochemical findings, we ana-lyzed the connexin expression by RT-PCR. RT-PCR analysisshowed that mRNA expression levels were in accordance with theimmunofluorescence staining results (Fig. 3). Cx26 was only

FIGURE 2 – Immunostaining of Cxs in endometrial hyperplasia. In most cases, Cx26 was expressed weakly(a) and in other cases, they weredetected diffusely in cytoplasm in some parts(b, arrowheads) and negatively in other parts(b, arrows). Magnification 4003.

FIGURE 1 – Cx expression in normal endometrium. Cx26 was rarely detected in endometrial epithelium of the proliferative phase, but wedetected it intensively in the mid-secretory phase. The localization of Cx32 was coordinated to Cx26 (data not shown). Cx43 was detected inendometrial stromal cells throughout the menstrual cycle.(a) Cx26 in mid-proliferative phase;(b) Cx26 in mid-secretory phase;(c) Cx43 inmid-proliferative phase. Magnification 4003.

TABLE III – IMMUNOHISTOCHEMICAL FINDING OF CONNEXIN IN ENDOMETRIAL HYPERPLASIA

HistologyCx26 Cx32 Cx43

Weak Normal Diffuse Weak Normal Diffuse Weak Normal Diffuse

Simple 6 0 0 6 0 0 6 0 0Complex 4 0 1 3 0 2 5 0 0Atypical 2 0 2 2 0 2 4 0 0Total 12 0 3 11 0 4 15 0 0

Weak, weak/negative; normal, positive on cell-cell border; diffuse, diffuse in cytoplasm.

319CONNEXINS IN ENDOMETRIAL CARCINOGENESIS

weakly detected in 12 samples compared with the normal endo-metrium of the secretory phase. Cx32 was rarely detected com-pared with the positive control. Though Cx43 was detected in allendometrial hyperplasia samples at a level similar to that in normalsamples, it may have been derived from the fibroblasts (data notshown).

Expression and localization of connexins in endometrial cancerThe expression and localization of connexins were analyzed for

33 endometrial cancer samples, consisting of 17 cases of tumor-grade G1 (well-differentiated adenocarcinoma), 12 cases of G2(moderately differentiated adenocarcinoma) and 4 cases of G3(poorly differentiated adenocarcinoma). As shown in Table IV,Cx26 was detected weakly or negatively (Fig. 4a) in 25 of the 33samples (75.8%), diffusely (Fig. 4b) in 6 samples (18.2%) andnormally in 2 samples (6.1%). The 2 samples that showed normalexpression of Cx26 were G1 carcinomas. Cx32 was detectedweakly or negatively for 26 of the 33 samples (78.8%), diffuselyfor 5 samples (15.2%) and normally for 2 samples (6.1%). The 2samples that showed normal expression of Cx32 were also G1carcinomas. Though Cx43 was not detected in normal endometrialepithelium, it was weakly detected in 5 cancer samples in G1 and2 samples in G2 (Fig. 4c). However, the correlation between theimmunoreactivities of Cx26 and Cx32 and the histological differ-entiation was unclear. Of the 33 endometrial cancer samples, 23were in stage I, 2 in stage II, 7 in stage III and 1 in stage IV.However, we could not find any correlation between the clinicalstage and connexin expression.

As in the case of endometrial hyperplasias, the cancer samplesthat showed weak or negative expression of Cx26 or Cx32 byimmunofluorescent analysis showed faint expression of theirmRNAs in RT-PCR analysis (Fig. 5). The samples that showed

diffuse immunofluorescent staining expressed Cx26 and Cx32 atlevels similar to those in the normal endometrium by RT-PCRassay. Though Cx43 was detected in all endometrial cancer sam-ples, most of it may have been derived from fibroblasts (data notshown).

Correlation of the immunohistochemical findings for Cx26 andCx32

The correlations of the immunohistochemical findings for Cx26and Cx32 are shown in Table V. In the normal endometrium, thelocalization of Cx26 and Cx32 was identical (Fig. 6a,b). In endo-metrial hyperplasia, 14 of 15 samples showed identical findingsand in endometrial cancer, 30 of 33 samples showed identicalstaining (Fig. 6c,d) and the other 3 samples showed differentfindings, Cx26-weak vs. Cx32-diffuse for 1 sample (Fig. 6e,f) andCx26-diffuse vs. Cx32-weak for 2 samples. The 2 samples thatshowed normal expression of Cx32 also showed normal expres-sion of Cx26.

DISCUSSION

In a previous study, we investigated changes in the expression ofthe gap junction proteins Cx26 and Cx32 in human endometrialglandular epithelium during the reproductive cycle and the influ-ence of hormone-replacement therapy. It was concluded that theywere up-regulated in the presence of progesterone plus estrogenbut suppressed in the presence of estrogen alone. This suggestedthat the loss of connexin gene expression contributed to the accel-eration of endometrial carcinogenesis.18 However, there has beenno report on connexin expression in endometrial hyperplasia andcancer. In this study, we demonstrated changes of the connexinexpression and localization during the process of endometrialcarcinogenesis by immunohistochemical analysis.

Unopposed estrogen is the strongest risk factor associated withendometrial cancer. Although little is known about the molecularevents involved, a close relationship has been observed betweenestrogenic stimulation of the endometrium and the appearance ofendometrial hyperplasia.19 Endometrial hyperplasias constitute aspectrum from simple to complex to atypical regions. It is esti-mated that the incidences of progression to malignancy fromsimple, complex and atypical regions are from 1% to 5%, 5% to10% and 20% to 30%, respectively.1,20Our results indicated that inall endometrial hyperplasias, Cx26 and Cx32 were poorly ex-pressed or aberrantly localized. It has been revealed that in the caseof the loss of expression or the aberrant localization of connexins,GJIC is suppressed.21,22Therefore, it is likely that in all 15 hyper-plasia samples GJIC was impaired. There are several lines ofevidence suggesting that connexin expression is suppressed and/oraberrantly localized in precancerous regions in several organs,10–15

and many, if not all, tumor-promoting agents have been shown toinhibit GJIC of cultured cells as well as thosein vivo,2 suggestingthat the loss of GJIC enhances clonal dispersion, causing loss ofthe growth-suppressing signals from the surrounding cells. Forendometrial carcinogenesis, it may be concluded that the loss ofGJIC caused by the suppressed expression and the aberrant local-ization of connexin supports the clonal evolution of endometrialcancer cells originating in the hyperplasia cells. Therefore, estro-gen, which suppresses connexin expression of endometrial epithe-

FIGURE 3 – RT-PCR of Cx26 and Cx32 for endometrial hyperplasia.Lanes 1–6, simple hyperplasia; lanes 7–11, endometrial hyperplasiacomplex; lanes 12–15, atypical endometrial hyperplasia; PC, positivecontrol. Except for sample no. 1, all the samples showed little expres-sion.

TABLE IV – IMMUNOHISTOCHEMICAL FINDING OF CONNEXIN IN ENDOMETRIAL HYPERPLASIA

HistologyCx26 Cx32 Cx43

Weak Normal Diffuse Weak Normal Diffuse Negative Weak Normal Diffuse

G1 12 2 3 14 2 1 12 5 0 0G2 10 0 2 10 0 2 10 2 0 0G3 3 0 1 2 0 2 4 0 0 0Total 25 2 6 26 2 5 26 7 0 0

Weak, weak/negative; normal, positive on cell-cell border; diffuse, diffuse in cytoplasm.

320 SAITO ET AL.

lium and accretes cell proliferation,18 may act as a tumor-promot-ing agent for endometrium.

Most endometrial cancers showed aberrant expression or local-ization of Cx26 and/or Cx32. These results suggested that GJICwas disturbed in 31 of 33 (93.9%) of the endometrial cancers, andit is not sure that the other 2 samples, which showed normal

expression and localization of Cxs, show normal GJIC. It has beenhypothesized that GJIC controls cell growth by transmission ofgrowth-regulatory signals and that the loss of growth control incancer cells may be due to decreased communication capacity.23

Most cancer cells have no or decreased gap junctional structureand function, and aberrant expression and function of connexinsare associated with tumor progression.3,8,24–26Furthermore, recentstudies have shown that transfection of connexin cDNA intocommunication-deficient cancer cells can restore communicationand suppress cell growth and tumor growth in nude mice.4,6 Ourresults from endometrial cancer reinforce these previous hypoth-eses.

In this study, all 15 endometrial hyperplasia samples showedweak or diffuse staining of Cx26 and Cx32. However, in 2 of the33 (6.1%) endometrial cancer samples, normal expression andlocalization were observed. The results were similar to these in ourprevious article on telomerase activity in endometrial hyperplasiaand cancer.27 In that study, all of the endometrial hyperplasiasamples had telomerase activity, whereas in 6.7% of the endome-

FIGURE 4 – Immunostaining of Cxs endometrial cancer. Most sam-ples showed weak (arrows) or negative staining(a) and cytoplasmiclocalization(b). Cx43 was detected (arrowheads) in the cancer cellsin some cases(c). (a,b)Cx26; (c) Cx43. Magnification 4003.

FIGURE 5 – RT-PCR of Cx26 and Cx32 for endometrial cancer.Lanes 1 and 2, samples weakly or negatively stained by immunoflu-orescence; lanes 3 and 4, normally stained; lanes 5 and 6, diffuselystained in cytoplasm; PC, positive control.

TABLE V – CORRELATION OF IMMUNOHISTOCHEMICAL FINDINGBETWEEN Cx26 AND Cx32 IN ENDOMETRIAL HYPERPLASIA AND CANCER

Hyperplasia

Cx32Weak Normal Diffuse

Cx26 Weak 11 0 1Normal 0 0 0Diffuse 0 0 3

Cancer samplesCx26 Weak 24 0 1

Normal 0 2 0Diffuse 2 0 4

321CONNEXINS IN ENDOMETRIAL CARCINOGENESIS

trial cancer samples the telomerase activity was negative. Theseresults imply that the cell growth of most endometrial hyperplasiasis promoted; on the other hand, it was suppressed in some endo-metrial cancers. Furthermore, a study reported the possibility thatCx26 plays a role in intravasation and extravasation of tumor cellsthrough heterologous gap junction formation with endothelialcells.28 It is thus proposed that reduced GJIC is important for thegenesis and maintenance of endometrial cancers, but recovered

connexin expression and GJIC in a small population of carcinomamay help to acquire invasive ability.

In our previous study18 as well as in the present study, we foundthat Cx26 and Cx32 are coordinately expressed and co-localized inthe normal endometrium. Although in endometrial hyperplasia andcancer, the localization and expression of the 2 connexins wereaberrant and low in most cases, behavior of Cx26 and Cx32 werestill coordinated. These results suggest that the gene expressionand the subcellular localization of Cx26 and Cx32 were regulatedby the same mechanism even in the carcinogenic endometrium.However, such a coordinated expression has not been observedduring human and rat carcinogenesis.8,29 Therefore, this coordi-nated expression of 2 Cx genes may be due to physiologicalconditions inherent to the endometrium; one such mechanism ishormonal regulation.

Including endometrial cancers, most tumor cells have a reducedability to communicate among themselves and/or with surroundingnormal cells, confirming the importance of intact GJIC in growthcontrol.2 When connexin genes are transfected into such cells,normal cell growth control is often recovered.4 Certain dominant-negative mutant connexin genes can reverse such tumor suppres-sion.30 While these results suggest that connexin genes form afamily of tumor-suppressor genes, so far we have found no con-nexin gene mutations in human tumors. Only 2 connexin genemutations were found in chemically induced rat tumors.30,31 Onthe other hand, recent studies suggest that connexin genes may beinactivated by hypermethylation of their promoter regions, sug-gesting that epigenetic inactivation of connexin genes may be amechanism of GJIC disturbance in certain tumors.32 However, inmany tumor cells, connexins are normally expressed but aberrantlylocalized and the mechanisms of aberrant localization of connexinsinclude lack of an appropriate cell-cell recognition apparatus andaberrant phosphorylation of connexins.21,33We have reported thatb-catenin, which is one of the key proteins of the adherens junctionin epithelial cells, is localized to the nucleus with/without itsmutation in endometrial hyperplasia and cancer.34 These resultssuggest that GJIC disorders may occur not only because of aber-rant expression of connexin genes themselves but also as a resultof disruption of various control mechanisms of the protein func-tions.2

ACKNOWLEDGEMENTS

We thank for Mr. M. Kim Barrymore for editing the manuscript.

REFERENCES

1. Kurman RJ, Kaminski PF, Norris HJ. The behavior of endometrialhyperplasia. A long-term study of untreated hyperplasia in 170 pa-tients. Cancer 1985;56:403–12.

2. Yamasaki H, Omori Y, Krutovskikh V, Zhu W, Mironov N, Yamak-age K, et al. Connexins in tumour suppression and cancer therapy.Novartis Found Symp 1999;219:241–54.

3. Lee SW, Tomasetto C, Sager R. Positive selection of candidatetumor-suppressor genes by subtractive hybridization. Proc Natl AcadSci USA 1991;88:2825–9.

4. Mesnil M, Krutovskikh V, Piccori C, Elfgang C, Traub O, WilleckeK, et al. Negative growth control of HeLa cells by connexin genes:connexin species specificity. Cancer Res 1995;55:629–39.

5. Yotti LP, Chang CC, Trosko JE. Elimination of metabolic cooperationin Chinese hamster cells by a tumor promoter. Science 1979;206:1089–91.

6. Zhu D, Kidder GM, Caveney S, Naus CC. Growth retardation inglioma cells cocultured with cells overexpressing a gap junctionprotein. Proc Natl Acad Sci USA 1992;89:10218–21.

7. Kamibayashi Y, Oyamada Y, Mori M, Oyamada M. Aberrant expres-sion of gap junction proteins (connexins) is associated with tumorprogression during multistage mouse skin carcinogenesis in vivo.Carcinogenesis 1995;16:1287–97.

8. Krutovskikh VA, Oyamada M, Yamasaki H. Sequential changes ofgap-junctional intercellular communications during multistage ratliver carcinogenesis: direct measurement of communication in vivo.Carcinogenesis 1991;12:1701–6.

9. Temme A, Buchmann A, Gabriel H, Nelles E, Schwarz M, Willecke

K. High incidence of spontaneous and chemically induced liver tu-mors in mice deficient for connexin32. Curr Biol 1997;7:713–6.

10. Hanna EA, Umhauer S, Roshong SL, Piechocki MP, Fernstrom MJ,Fanning JD, et al. Gap junctional intercellular communication andconnexin43 expression in human ovarian surface epithelial cells andovarian carcinomas in vivo and in vitro. Carcinogenesis 1999;20:1369–73.

11. Huang RP, Hossain MZ, Sehgal A, Boynton AL. Reduced connexin43expression in high-grade human brain glioma cells. J Surg Oncol1999;70:21–4.

12. Krutovskikh V, Yamasaki H. The role of gap junctional intercellularcommunication (GJIC) disorders in experimental and human carcino-genesis. Histol Histopathol 1997;12:761–8.

13. Laird DW, Fistouris P, Batist G, Alpert L, Huynh HT, Carystinos GD,et al. Deficiency of connexin43 gap junctions is an independentmarker for breast tumors. Cancer Res 1999;59:4104–10.

14. Locke D. Gap junctions in normal and neoplastic mammary gland.J Pathol 1998;186:343–9.

15. Tsai H, Werber J, Davia MO, Edelman M, Tanaka KE, Melman A, etal. Reduced connexin 43 expression in high grade, human prostaticadenocarcinoma cells. Biochem Biophys Res Commun 1996;227:64–9.

16. Hulka BS. Links between hormone replacement therapy and neopla-sia. Fertil Steril 1994;62:168S–175S.

17. Bertram JS. Cancer prevention by carotenoids. Mechanistic studies incultured cells. Ann NY Acad Sci 1993;691:177–91.

18. Saito T, Oyamada M, Yamasaki H, Mori M, Kudo R. Co-ordinated

FIGURE 6 – Double staining of Cx26 and Cx32 in normal endome-trium and endometrial cancer. In normal endometrium, Cx26(a) andCx32(b) were coordinately expressed and localized in the endometrialepithelium. In endometrial cancer, they were also coexpressed in mostcases(c,d).However, in some cases, they were differently expressed:(e) weak, (f) diffuse. (a,b) normal endometrium;(c–f) endometrialcancer.(a,c,e)Cx26 (FITC);(b,d,f)Cx32 (rhodamine). Magnification1,0003.

322 SAITO ET AL.

expression of connexins 26 and 32 in human endometrial glandularepithelium during the reproductive cycle and the influence of hormonereplacement therapy. Int J Cancer 1997;73:479–85.

19. Kohler MF, Nishii H, Humphrey PA, Saski H, Marks J, Bast RC, etal. Mutation of the p53 tumor-suppressor gene is not a feature ofendometrial hyperplasias. Am J Obstet Gynecol 1993;169:690–4.

20. Wentz WB. Progestin therapy in endometrial hyperplasia. GynecolOncol 1974;2:362–7.

21. Saito T, Schlegel R, Andresson T, Yuge L, Yamamoto M, YamasakiH. Induction of cell transformation by mutated 16K vacuolar H1-atpase (ductin) is accompanied by down-regulation of gap junctionalintercellular communication and translocation of connexin 43 inNIH3T3 cells. Oncogene 1998;17:1673–80.

22. Yamasaki H, Naus CC. Role of connexin genes in growth control.Carcinogenesis 1996;17:1199–213.

23. Yamasaki H. Role of disrupted gap junctional intercellular commu-nication in detection and characterization of carcinogens. Mutat Res1996;365:91–105.

24. Grossman HB, Liebert M, Lee IW, Lee SW. Decreased connexinexpression and intercellular communication in human bladder cancercells. Cancer Res 1994;54:3062–5.

25. Loewenstein W. Junctional intercellular communication and the con-trol of growth. Biochem-Biophys Acta 1979;560:1–65.

26. Trosko J, Chang C. Potential role of intercellular communication inthe rate-limiting step in carcinogenesis. J Am Cell Toxicol 1983;2:5–22.

27. Saito T, Schneider A, Martel N, Mizumoto H, Bulgay-Moerschel M,

Kudo R, et al. Proliferation-associated regulation of telomerase activ-ity in human endometrium and its potential implication in early cancerdiagnosis. Biochem Biophys Res Commun 1997;231:610–4.

28. Ito A, Katoh F, Kataoka TR, Okada M, Tsubota N, Asada H, et al. Arole for heterologous gap junctions between melanoma and endothe-lial cells in metastasis. J Clin Invest 2000;105:1189–97.

29. Oyamada M, Krutovskikh VA, Mesnil M, Partensky C, Berger F,Yamasaki H. Aberrant expression of gap junction gene in primaryhuman hepatocellular carcinomas: increased expression of cardiac-type gap junction gene connexin 43. Mol Carcinog 1990;3:273–8.

30. Omori Y, Mesnil M, Yamasaki H. Connexin 32 mutations fromX-linked Charcot-Marie-Tooth disease patients: functional defectsand dominant negative effects. Mol Biol Cell 1996;7:907–16.

31. Saito T, Barbin A, Omori Y, Yamasaki H. Connexin 37 mutations inrat hepatic angiosarcomas induced by vinyl chloride. Cancer Res1997;57:375–7.

32. Piechocki MP, Burk RD, Ruch RJ. Regulation of connexin32 andconnexin43 gene expression by DNA methylation in rat liver cells.Carcinogenesis 1999;20:401–6.

33. Jongen WM, Fitzgerald DJ, Asamoto M, Piccoli C, Slaga TJ, Gros D,et al. Regulation of connexin 43-mediated gap junctional intercellularcommunication by Ca21 in mouse epidermal cells is controlled byE-cadherin. J Cell Biol 1991;114:545–55.

34. Nei H, Saito T, Yamasaki H, Mizumoto H, Ito E, Kudo R. Nuclearlocalization of beta-catenin in normal and carcinogenic endometrium.Mol Carcinog 1999;25:207–18.

323CONNEXINS IN ENDOMETRIAL CARCINOGENESIS