Carcinogenesis vol.23 no.11 pp.1791–1796, 2002

Changes in expression of the human homologue of the Drosophiladiscs large tumour suppressor protein in high-grade premalignantcervical neoplasias

Richard A.Watson1, Terry P.Rollason2, GaryM.Reynolds3, Paul G.Murray3, Lawrence Banks4 andSally Roberts1,5

1Cancer Research UK Institute for Cancer Studies, The Medical School,University of Birmingham, Birmingham B15 2TT, 2Department ofPathology, Birmingham Women’s Hospital, Birmingham, 3Department ofPathology, The Medical School, University of Birmingham, BirminghamB15 2TT, UK, 4International Centre for Genetic Engineering andBiotechnology, Padriciano 99, I-34012 Trieste, Italy5To whom correspondence should be addressedEmail: s.roberts@bham.ac.uk

The Drosophila tumour suppressor discs large (Dlg) is acell-junction localized protein that is required for themaintenance of epithelial cyto-architecture and the negativecontrol of cell proliferation. The mammalian homologue islikely to have a similar mode of action, and thereforefunctional perturbation of this protein may be linked tothe development of epithelial-derived cancers. The findingthat several unrelated viral oncoproteins, including the E6protein of oncogenic human papillomaviruses, bind to thehuman homologue of Dlg (hDlg) supports this proposition.Employing immunohistochemistry, we show that in uterinecervical squamous epithelia, prominent localization of hDlgat sites of intercellular contact occurs in cells that have leftthe proliferating basal cell layers and begun maturation.The presence of hDlg at sites of cell:cell contact diminishes,whilst intracellular cytoplasmic levels increase significantlyin high-grade, but not low-grade, cervical neoplasias. Ininvasive squamous cell carcinomas, total cellular hDlglevels are greatly reduced. Our data suggest that loss ofhDlg at sites of intercellular contact may be an importantstep in the development of epithelial cancers.

Introduction

Recent studies in Drosophila have shown that significant cyto-architectural changes of epithelial tissue, such as loss of cellpolarity, aberrant cell adhesion and changes in cell shape,occur when the structure of the discs large (Dlg) gene isperturbed (1,2). These changes are accompanied by increasedepithelial cell proliferation and neoplastic growth, indicatingthat Dlg functions in the maintenance of epithelial cell structureand as a negative regulator of cell growth; thus it is nominatedas a tumour suppressor (2,3). Dlg is the founding member of themembrane-associated guanylate kinase (MAGUK) superfamilywhose members are defined by a basic core of three differentprotein-interaction domains: a PDZ domain, an SH3 domainand a domain with homology to the enzyme guanylate kinase.MAGUKs are thought to act as scaffold proteins linkingcomponents of signalling pathways and structural proteins at

Abbreviations: APC, adenomatous polyposis coli; CIN, cervical intraepithelialneoplasia; Dlg, Drosophila tumour suppressor discs large; hDlg, humanhomologue of Dlg; HPV, human papilloma virus; MAGUK, membrane-associated guanylate kinase; SIL, squamous intraepithelial lesions.

© Oxford University Press 1791

specific domains at peripheral membranes of cells (4). InDrosophila imaginal disc epithelial cells, Dlg localizes tothe cytoplasmic face of septate junctions (1) (equivalent tovertebrate tight junctions) whilst in Caenorhabditis elegans itresides at adherens junctions (5,6). Less is known about thehuman homologue of Dlg (hDlg), although it also localizes atintercellular contact sites in epithelial cells and has been shownto bind to members of the protein 4.1/ezrin, radixin, moesin(ERM) family; interactions that may be important for targetingof hDlg to the cell periphery (7–10). HDlg binds to severalcellular factors that have roles in cell growth and proliferation(11,12). For example, it interacts with the tumour suppressoradenomatous polyposis coli (APC) protein and formation ofan APC:hDlg complex is important in inducing a cell cyclesuppression signal (11,13). It would seem likely therefore, thatthe human homologue of Dlg has similar functions as itsDrosophila counterpart, regulating the structure and growth ofepithelial tissue in response to cell:cell contact. Interestingly,hDlg has recently been described as a target of several viraloncoproteins; the E6 protein of high-risk types of humanpapillomaviruses (HPVs), E4-ORF1 of adenovirus 9 and Taxof human T-cell leukaemia virus (14–16). E6 has also beenshown to target hDlg for ubiquitin-mediated degradation viathe proteasome (17). The physiological effects of binding hDlgby these viral oncoproteins is not clear, however it has beendemonstrated that Tax is able to suppress the growth inhibitoryproperties of hDlg and that E6 mutants unable to bind hDlgalso lose their transforming activity in rodent cells (15,16).Therefore, given that binding to hDlg is a function conservedbetween unrelated viral oncoproteins, and the potential role ofhDlg as a tumour suppressor, it has been proposed thatinactivation of this MAGUK may contribute to the developmentof virus-associated cancers (14,15).

Therefore, in this study we have examined hDlg expressionin the normal ectocervical and endocervical epithelium of theuterine cervix – a common site of infection of carcinogenicHPV types. Development of invasive cervical cancer is pre-ceded by cervical intraepithelial neoplasia [CIN, also termedsquamous intraepithelial lesions (SIL), low and high-grade]that is graded by reference to the level of epithelial involvementby dedifferentiated cells, together with several additionalmorphological features. Therefore, to determine whether hDlgexpression is altered during the transformation of normalcervical epithelium to neoplastic epithelium, immunohisto-chemistry was used to examine the distribution of hDlg inarchival paraffin-embedded tissues of normal cervix, low andhigh-grade CINs and invasive squamous carcinoma.

Materials and methods

Tissue samples

Archival formalin-fixed, paraffin-embedded cervical tissue sections wereobtained from Good Hope Hospital, Sutton Coldfield, Birmingham, UK. Atotal of 52 samples were analyzed. Six samples had no squamous epitheliumand were removed from the study. The remaining 46 samples ranged fromhistologically normal through all grades of CIN. A majority of biopsies

by guest on June 21, 2015http://carcin.oxfordjournals.org/

Dow

nloaded from

R.A.Watson et al.

also contained endocervical epithelium and, a few cases, squamous maturemetaplasia and immature metaplasia, as well as regions showing morphologicalevidence of HPV infection in non-CIN areas. The numbers for each lesiontype were normal cervical epithelium, n � 14; CIN1, n � 11 CIN2, n � 10;CIN3, n � 11. Because some of the high-grade lesions also contained regionsof a lower grade the numbers for each lesion type were therefore CIN1 �13,CIN2 � 15 and CIN3 � 11. In addition, four cases of invasive cervicalcancer were studied, although only two of these had adjacent haematoxylin/eosin (H&E)-stained sections, kindly provided by Dr G.Teal, Good HopeHospital, Birmingham, UK. For all cases, both the H&E-stained and hDlg-stained sections were re-reviewed by a single pathologist (T.P.R.). Sectionswere photographed using a Leica DC200 camera or an Olympus digitalcamera (invasive carcinomas) and images were assembled in Adobe Photoshop,San Jose, CA.

Antibodies, cell culture, western blotting and immunofluorescence staining

Monoclonal antibodies (mAbs) against hDlg (cl. 2D11) and E-cadherin (cl.36) were purchased from Santa Cruz Biotechnology, California, CA andTransduction Laboratories Lexington, KY respectively. Epithelial linesexamined were Madine-Darby canine kidney (MDCK), an SV40-immortalizedhuman skin keratinocyte line, and two spontaneously immortalized keratinocytelines, HaCaT and SCC-12F. All except HaCaTs, were grown in DME mediumsupplemented with 10% foetal calf serum, 4 mM glutamine and antibiotics.HaCaT cells were grown in the media described with 0.4 µg/ml hydrocortisoneadded. Cells were lysed in 8 M urea, 50 mM Tris pH 8.0, 0.15 M β-mercaptoethanol and equal amounts of protein electrophoresed on 10%SDS-polyacrylamide gels, western-blotted using 2D11 (1 : 400 dilution) anddeveloped using ECL (Amersham International). Immunofluorescence micro-scopy was performed on confluent cultures of cells as described previously(18). 2D11 was used at 1 in 30 and immune-complexes were detected usinggoat anti-mouse IgG-Alexa 488 antibody (Molecular Probes, Eugene, OR).

Immunohistochemistry

Sections (4 µm) were cut, mounted on Vectabond (Vector Laboratories,Peterborough, UK)-treated Superfrost Plus slides and air-dried overnight at37°C. One section of each case was stained with H&E. Sections to be stainedwith antibodies were de-paraffinized and re-hydrated by first immersing inxylene, followed by absolute ethanol and then water. To block endogenousperoxidase activity, sections were immersed in 3% hydrogen peroxide inmethanol for 10 min. Antigen retrieval was performed using a modified low-temperature method (19). Briefly, sections were incubated in EDTA buffer(1 mM EDTA, pH 8.0, 0.01% Tween 20) at 65°C for 16 h with stirring(600 r.p.m.). When cool, sections were incubated in 0.2 M Tris-buffered saline(TBS), pH 7.6 for 30 min at 20°C. Primary antibody (2D11 diluted 1 in 25;E-cadherin mAb diluted 1 in 200) was applied for 60 min at room temperatureand subsequently slides were washed in TBS containing 0.1% Tween 20.Negative controls were sections in which the primary antibody had beenomitted. Primary antibody was localized by using a streptavidin–biotin–peroxidase kit (Universal ABC kit, The Binding Site, Birmingham, UK)according to the manufacturer’s instructions. The chromogenic reaction wasperformed using NovaRED substrate (Vector Laboratories, Peterborough, UK)for 10 min at 20°C. After counterstaining in Mayer’s haematoxylin, sectionswere dehydrated by immersion in absolute ethanol followed by xylene andthen mounted in DPX. There was no evidence that endogenous biotin interferedwith the staining patterns of hDlg or E-cadherin as identical staining patternswere obtained in sections in which endogenous biotin was blocked priorto staining.

Results

Testing the specificity of hDlg monoclonal antibody 2D11The commercially available hDlg mAb 2D11 was raised againstthe N-terminus (amino acids 1–229) of hDlg, a region that isunique to hDlg between MAGUK family members. To confirmthat 2D11 was specific for hDlg, various epithelial cell lineswere examined by western blotting and immunofluorescencemicroscopy. On western blots, 2D11 identified between twoand three bands around 120 kDa and no other bands weredetected (Figure 1, top panel). A similar molecular mass andmultiple band pattern has been described for hDlg (predictedmolecular mass is 103 kDa) and its rat homologue, synapse-associated protein 97 (SAP97) (7,10,20). In confluent culturesof the different epithelial cells stained with 2D11, hDlg wasimmunolabelled at the cell periphery at sites of cell:cell contact

1792

Fig. 1. Confirmation of mAb 2D11 specificity for hDlg. Top: western blotof whole-cell protein lysates of epithelial cell lines showing specificlabelling of hDlg by the anti-Dlg mAb 2D11. Lane 1, HaCaT keratinocytes;lane 2, SV40-immortalized human keratinocytes; lane 3, SCC-12Fkeratinocytes; lane 4, MDCK cells. Bottom: Immunofluorescent labelling ofDlg by 2D11 in confluent MDCK cells. Identical distribution of hDlg wasfound in human keratinocyte cell lines stained with 2D11.

(Figure 1, bottom panel and data not shown). A similarlocalization of hDlg in epithelial cells has been previouslyreported (7,10). Also, 2D11 labelled small clusters of cyto-plasmic granules near the cell boundary. These structures mayrepresent vesicular structures in which hDlg is thought toreside prior to recruitment to the cell periphery (10). Weconclude that 2D11 specifically recognizes hDlg.

hDlg protein expression in normal cervixThe distribution and cellular localization of hDlg was examinedin the different types of epithelium found in the uterine cervix.Twelve cases (12/14) contained both normal endocervical(glandular) and ectocervical epithelia, and hDlg staining wasmost often stronger in cells of the endocervix than in those ofectocervical epithelium. Furthermore, it was noticeable thatintensity of hDlg staining in the ectocervix became significantly

by guest on June 21, 2015http://carcin.oxfordjournals.org/

Dow

nloaded from

hDlg expression in cervical lesions

Fig. 2. Immunohistochemical analysis of hDlg expression in uterine cervical epithelium. (A) and (B) anti-hDlg (mAb 2D11) staining and E-cadherinexpression (C) in normal ectocervix (sections from same tissue sample). Note the change in hDlg expression and localization as cells move up through thestratifying layers. Although hDlg primarily localizes at intercellular borders, a cytoplasmic and nuclear location is also evident. B, inset: Enlargement ofnucleus identified by arrow to show hDlg nuclear staining. (D) 2D11-staining in the endocervix. (E) 2D11-staining in immature squamous metaplasia. Notethat the proliferating compartments of immature squamous metaplasia and ectocervix show significant levels of cytoplasmic hDlg. Bar, 50 µm.

weaker at sites farthest from the transformation zone (data notshown). In all cases of normal ectocervical squamous epithelia,expression of hDlg was limited to the basal, parabasal andintermediate suprabasal layers, with the superficial layers beinglargely negative (Figure 2A). This distribution pattern wassimilar to the adherens junction protein E-cadherin (Figure 2C).A diffuse cytoplasmic hDlg distribution was observed in basalcells and the immediate two to three layers of parabasalcells (Figure 2B). However, variations in basal hDlg stainingoccurred, with some samples showing a more intense cyto-plasmic labelling than those of the parabasal compartment(data not shown). In addition, weak staining at cell:cell bordersand cell junctions was observed in both compartments. In theintermediate suprabasal layers, hDlg levels increased at thecell periphery, particularly at structures resembling desmo-somes, and decreased significantly in the cytoplasm(Figure 2B). In some cells of basal and suprabasal compart-ments, hDlg was detectable at the nuclear periphery (Figure 2Band inset). In columnar cells of the endocervix, hDlg waspresent at the baso-lateral cell borders and in many of thesecells cytoplasmic staining was also detected (Figure 2D). Sub-columnar reserve cells showed intense cytoplasmic staining(data not shown). Significant cytoplasmic hDlg was evidentin cells of immature squamous metaplasia although some cellsalso showed hDlg at cell borders (Figure 2E). The stainingpattern in mature squamous metaplasia was similar to that in

1793

ectocervical tissue (data not shown). No staining was seenin controls.

hDlg in CIN and invasive carcinomas

No substantial differences were found with respect to hDlgexpression or localization in epithelia showing morphologicalevidence of HPV (wart) infection (data not shown). Similarly,the pattern of hDlg localization and level of protein expressionin low-grade dysplasias (12/13, CIN1) was not significantlydifferent from that found in normal tissue (Figure 3A). How-ever, one case (1/13) showed strong hDlg staining at theperiphery of cells in the mid-epithelial layers, but none (orvery little) in cells of lower layers (Figure 3B).

In contrast, hDlg expression was strikingly different in high-grade (CIN2 and CIN3) dysplastic lesions (Figure 3C–F). Thetypical pattern of hDlg distribution (22/26 cases) was a strongdiffuse cytoplasmic staining that extended through almost theentire thickness of epithelium (Figure 3C–E). The presence ofhDlg at cell:cell boundaries was evident in some cases, but theintensity varied and was restricted either to the uppermost layersof the epithelium or to small foci of cells (Figure 3C and E). Inmore differentiated lesions, no hDlg staining was evident at thecell boundaries or in the cytoplasm in the most superficial celllayers (Figure 3D). As was noted in normal epithelium, therewas some variation in the level of basal staining; in six out ofthe 22 typical cases basal hDlg levels were higher then elsewhere

by guest on June 21, 2015http://carcin.oxfordjournals.org/

Dow

nloaded from

R.A.Watson et al.

Fig. 3. Immunohistochemical analysis of hDlg expression in CIN and invasive squamous carcinoma. (A and B) anti-hDlg (mAb 2D11) staining in CIN1 withco-existent HPV-associated changes (low-grade SIL). Distribution of hDlg in low-grade samples was similar to that in normal tissue, except in one case (B)that showed a decrease in hDlg levels in lower epithelial cell layers. HDlg expression in high-grade CIN (C–F, CIN2 and 3, high-grade SIL). Note that inhigh-grade lesions there is an increase in cytoplasmic hDlg levels whilst levels at the cell periphery diminish. (G–I) hDlg in invasive squamous cellcarcinomas. Note that (G) shows hDlg-staining in a region of the tumour sample that is high-grade CIN. Contrast intensity of staining in this region with themarked decrease in hDlg levels in the invasive regions of the carcinomas (H and I), particularly at the leading edge of the tumour (indicated by an arrow inI). Bar, 50 µm (A–E); 100 µm (F), 200 µm (G and I), 50 µm (H).

in the epithelium (data not shown). Although in general cyto-plasmic hDlg levels were high, small areas or patches of weakstaining were evident in a minority of lesions (data not shown).In the four cases showing atypical hDlg staining, three (twoCIN2 and one CIN3) had relatively weak levels of hDlg in both

1794

the cytoplasm and at cell boundaries compared with areas ofnormal squamous and glandular epithelium, and to the levels inseparate cases of high-grade CIN (Figure 3F). The other case,graded CIN3, had weak staining of hDlg both within the lesionand in normal epithelia (data not shown).

by guest on June 21, 2015http://carcin.oxfordjournals.org/

Dow

nloaded from

hDlg expression in cervical lesions

Four invasive squamous carcinomas were stained with 2D11(Figure 3G–I). Two cases also contained high-grade (CIN3)dysplasia and in these regions hDlg protein expression washigh (Figure 3G), in agreement with hDlg expression in high-grade CIN samples. In the invasive tumour islands, however,hDlg protein levels were very weak (Figure 3H) and, signific-antly, staining was absent at the leading edge of the invadingtumours (Figure 3I).

Discussion

The human homologue of Drosophila tumour suppressor Dlgis expressed in all the different types of epithelia present inthe uterine cervix (Figure 2). In glandular epithelium, itlocalizes primarily to the basal and lateral membranes ofcolumnar cells, similar to hDlg distribution in cells of humancolonic tissue (20). Peripheral hDlg staining is also prominentin suprabasal cell layers of ectocervical epithelium. Thesefindings are consistent with a role for hDlg at sites ofintercellular contacts. However, hDlg is also located in thecytoplasm and nucleus. Cytoplasmic hDlg staining was mostpronounced in the proliferating compartments of the ectocervixand immature squamous metaplasia that have either no, orlittle, peripheral hDlg. This sub-cellular distribution of hDlgis reminiscent of that of SAP97 (the rat homologue ofhDlg) in epithelial cells on intestinal villi (21). Cells of theproliferating compartment of the intestinal crypts have lowerlevels of peripheral SAP97 compared with the intense labellingof the lateral membranes of more mature cells in the distalsegments of the villi (21). Localization of hDlg to the peripherymay therefore be important for the negative regulation ofproliferation. Nuclear hDlg was present in basal and suprabasalcompartments of the ectocervix. In cultured epithelial cells, anuclear form of hDlg is detected by an antibody that recognizesalternatively spliced forms of hDlg containing the I2 insertion,suggesting that nuclear localization may be regulated by post-transcriptional modification of hDlg transcripts (22). It remainsto be seen however, whether the different sub-cellular localiz-ations of hDlg reflect alternative functions of the protein.

In invertebrates, Dlg is required for the maintenance ofepithelial cell:cell junctions (1,2,5,6,23), and in Drosophilathis is important for the organization of epithelial tissues andthe prevention of their neoplastic overgrowth (1,2,23). Thechanges in sub-cellular expression of hDlg as cells move upthrough the stratifying layers of uterine ectocervical epitheliumcould therefore indicate that hDlg may have a role in theproper growth and formation of differentiated epithelium. Thisrole could be dependent on hDlg’s localization to the cellperiphery, where it may participate in aspects of assemblyand/or function of intercellular junctions. It is therefore ofinterest that the normal pattern of hDlg expression is signific-antly disturbed in high-grade dysplasias of cervical epithelium(see Figure 3). In these lesions, there was a loss of hDlglocalization at the cell periphery concomitant with an increasein cytoplasmic staining. Also, in contrast to normal epitheliaand low-grade lesions, cytoplasmic hDlg was expressed inalmost all layers of the epithelium. This is consistent with thereplacement of differentiated layers with more ‘basal-like’proliferating cells. Furthermore, in invasive tumours totalcellular hDlg levels were low and even absent in some regions(Figure 3). Thus, loss of hDlg from the cell periphery maynegate its function as an inhibitor of neoplastic transformation.This is supported by recent work in cervical tumour cell lines

1795

that shows that cells that exhibit a more fully transformedphenotype have lower levels of hDlg protein at sites of cell:cellcontact compared to less transformed cells (24; R.W. and S.R.,unpublished observations). Our results therefore suggest thatloss of hDlg at sites of intercellular contact may be a crucialstep in the progression of cervical neoplasia.

HPV is a recognized human carcinogen and it is wellestablished that HPV is present in nearly all cases of CIN andcarcinomas of the cervix (25). The HPV E6 oncogene mayprimarily contribute to the late stages of progression tomalignancy (26). Although inactivation of p53 is important inE6’s biological function, p53-independent activities may alsobe necessary. For instance, E6-induction of epithelia hyperpla-sia in transgenic mice is p53-independent (27). It is temptingto speculate that a loss of a functional hDlg through E6-mediated degradation may contribute to such changes inepithelial growth and that the changes in hDlg expressionobserved in high-grade dysplasias and cancers may be mediatedby the activity of an oncogenic E6 molecule. However, it hasbeen demonstrated that in cervical tumour cells a reduction inhDlg protein levels can occur in the absence of HPV genomessuggesting that regulation of hDlg expression in cervicaltissue could be controlled through HPV-independent cellularmechanisms (24). For example, in HPV-negative C33I cervicaltumour cells, the loss of hDlg expression is largely at the levelof transcription, and therefore this may be one of the meansby which hDlg is lost in the absence of E6 (24).

Studies of the invertebrate MAGUK proteins have shownthat they act in pathways important in maintaining growthcontrol and epithelial structure. It is proposed that the modesof action of these proteins are likely to be conserved betweeninvertebrates and vertebrates. Because the loss of cell growthcontrol and cyto-architecture disruption are hallmarks of onco-genic transformation, then perturbation of the function ofmammalian MAGUKs may contribute to tumourigenesis. Inter-estingly, cellular levels of the MAGUK protein ZO-1, acomponent of tight junctions, are significantly reduced inbreast cancers (28). We have shown that hDlg localization andexpression is severely disrupted during the gradual transforma-tion of cervical epithelia, indicating that like its Drosophilacounterpart, hDlg may function in cellular pathways that inhibitcell proliferation and maintain tissue structure.

Acknowledgements

We thank Steven Hay for help in the collection of tissuesamples and Jo Flavell for sectioning the paraffin blocks. RAWwas supported by a Ph.D. studentship provided by the E.B.Jones Bequest, University of Birmingham’s Medical School.L.B. is supported by the Associazione Italiana per la Ricercasul Cancro. This work was supported by Cancer Research UK.

References

1.Woods,D.F. and Bryant,P.J. (1991) The discs-large tumour suppressor geneof Drosophila encodes a guanylate kinase homolog localized at septatejunctions. Cell, 66, 451–464.

2.Woods,D.F., Hough,C., Peel,D., Callaini,G. and Bryant,P.J. (1996) Dlgprotein is required for junction structure, cell polarity, and proliferationcontrol in Drosophila epithelia. J. Cell Biol., 134, 1469–1482.

3.Woods,D.F. and Bryant,P.J. (1989) Molecular cloning of the Lethal(1)DiscsLarge-1 oncogene of Drosophila. Dev. Biol., 134, 222–235.

4.Dimitratos,S.D., Woods,D.F., Stathakis,D.G. and Bryant,P.J. (1999)Signalling pathways are focused at specialized regions of the plasmamembrane by scaffolding proteins of the MAGUK family. Bioessays, 21,912–921.

by guest on June 21, 2015http://carcin.oxfordjournals.org/

Dow

nloaded from

R.A.Watson et al.

5.Bossinger,O., Klebes,A., Segbert,C., Theres,C. and Knust,E. (2001) Zonulaadherens formation in Caenorhabditis elegans requires dlg-1, thehomologue of the Drosophila gene discs large. Dev. Biol., 230, 29–42.

6.Firestein,B.L. and Rongo,C. (2001) DLG-1 is a MAGUK similar to SAP97and is required for adherens junction formation. Mol. Biol. Cell, 123,465-475.

7.Lue,R.A., Marfatia,S.M., Branton,D. and Chishti,A.H. (1994) Cloning andcharacterization of hdlg: The human homologue of the Drosophila discslarge tumor suppressor binds to protein 4.1. Proc. Natl Acad. Sci. USA,91, 9818–9822.

8.Marfatia,S.M., Morais Cabral,J.H., Lin,L., Hough,C., Bryant,P.J., Stolz,L.and Chishti,A.H. (1996) Modular organization of the PDZ domains in thehuman discs-large protein suggests a mechanism for coupling PDZ domain-binding proteins to ATP and the membrane cytoskeleton. J. Cell Biol.,135, 753–766.

9.Lue,R.A., Brandin,E., Chan,E.P. and Branton,D. (1996) Two independentdomains of hDlg are sufficient for sub-cellular targeting: the PDZ1-2conformational unit and an alternatively spliced domain. J. Cell Biol.,135, 1125–1137.

10.Reuver,S.M. and Garner,C.C. (1998) E-cadherin mediated cell adhesionrecruits SAP97 into the cortical cytoskeleton. J. Cell Sci., 111, 1071–1080.

11.Matsumine,A., Ogai,A., Senda,T., et al. (1996) Binding of APC to thehuman homolog of the Drosophila discs large tumour suppressor protein.Science, 272, 1020–1023.

12.Gaudet,S., Branton,D. and Lue,R.A. (2000) Characterization of PDZ-binding kinase, a mitotic kinase. Proc. Natl Acad. Sci. USA, 97, 5167–5172.

13. Ishidate,T., Matsumine,A., Toyoshima,K. and Akiyama,T. (2000) TheAPC-hDlg complex negatively regulates cell cycle progression from G0/G1 to S phase. Oncogene, 19, 365–372.

14.Lee,S.S., Weiss,R.S. and Javier,R.T. (1997) Binding of human virusoncoproteins to hDlg/SAP97, a mammalian homolog of the Drosophiladiscs large tumor suppressor protein. Proc. Natl Acad. Sci. USA, 94,6670–6675.

15.Kiyono,T., Hiraiwa,A., Fujita,M., Hayashi,Y. and Akiyama,T. (1997)Binding of high-risk human papillomavirus E6 oncoproteins to the humanhomolog of the Drosophila discs large tumor suppressor protein. Proc.Natl Acad. Sci. USA, 94, 11612–11616.

16.Suzuki,T., Ohsugi,Y., Uchida-Toita,M., Akiyama,T. and Yoshida,M. (1999)Tax oncoprotein of HTLV-1 binds to the human homologue of Drosophiladiscs large tumour suppressor protein hDLG and perturbs its function incell growth control. Oncogene, 18, 5967–5972.

1796

17.Gardiol,D., Kuhne,C., Glaunsinger,B., Lee,S.S., Javier,R. and Banks,L.(1999) Oncogenic human papillomavirus E6 proteins target the discs largetumour suppressor for proteasome-mediated degradation. Oncogene, 18,5487–5496.

18.Roberts,S., Ashmole,I., Rookes,S.M. and Gallimore,P.H. (1997) Mutationalanalysis of the human papillomavirus type 16 E1 E4 protein shows thatthe C-terminus is dispensable for keratin cytoskeleton association but isinvolved in inducing disruption of the keratin filaments. J. Virol., 71,3554–3562.

19.Yao,Y., Minter,H.A., Chen,X., Reynolds,G.M., Bromley,M. and Arrand,J.R.(2000) Heterogeneity of HLA and EBER expression in Epstein–Barr virus-associated nasopharyngeal carcinoma. Int. J. Cancer, 88, 949–955.

20.Nix,S.L., Chishti,A.H., Anderson,J.M. and Walther,Z. (2000) hCASK andhDlg associate in epithelia, and their Src homology 3 and guanylate kinasedomains participate in both intramolecular and intermolecular interactions.J. Biol. Chem., 275, 41192–41200.

21.Muller,B.M., Kistner,U., Veh,R.W., Cases-Langhoff,C., Becker,B.,Gundelfinger,E.D. and Garner,C.C. (1995) Molecular characterization andspatial distribution of SAP97 and the Drosophila discs-large tumoursuppressor protein. J. Neurosci., 15, 2354–2366.

22.McLaughlin,M., Hale,R., Ellston,D., Gaudet,S., Lue,R.A. and Viel,A.(2002) The distribution and function of alternatively spliced insertions inhDlg. J. Biol. Chem., 277, 6406–6412.

23.Bilder,D., Li,M. and Perrimon,N. (2000) Cooperative regulation of cellpolarity and growth by Drosophila tumour suppressors. Science, 289,113–116.

24.Mantovani,F., Massimi,P. and Banks,L. (2001) Proteasome-mediatedregulation of the hDlg tumour suppressor protein. J. Cell Sci., 114,4285–4292.

25.Walboomers,J.M.M., Jacobs,M.V., Manos,M.M., et al. (1999) Humanpapillomavirus is a necessary cause of invasive cervical cancer worldwide.J. Pathol., 189, 12–19.

26.Song,S., Liem,A., Miller,J.A. and Lambert,P.F. (2000) Humanpapillomavirus type 16 E6 and E7 contribute differently to carcinogenesis.Virology, 267, 141–150.

27.Song,S., Pitot,H.C. and Lambert,P.F. (1999) The human papillomavirustype 16 E6 gene alone is sufficient to induce carcinomas in transgenicanimals. J. Virol., 73, 5887–5893.

28.Hoover,K.B., Liao,S-Y. and Bryant,P.J (1998) Loss of the tight junctionMAGUK Z0-1 in breast cancer. Am. J. Pathol., 153, 1767–1773.

Received April 26, 2001; revised June 25, 2002; accepted June 27, 2002

by guest on June 21, 2015http://carcin.oxfordjournals.org/

Dow

nloaded from