FULL PAPER British Journal of Cancer (2013) 108, 402–408 | doi: 10.1038/bjc.2012.570 Keywords: tumour hypoxia; E-cadherin; GLUT-1; CA IX; tumour invasiveness; uterine cervix cancers Lacking hypoxia-mediated downregulation of E-cadherin in cancers of the uterine cervix A Mayer*,1, M Ho¨ckel2, N Schlischewsky1, H Schmidberger1, L-C Horn3 and P Vaupel1 1DepartmentofRadiooncologyandRadiotherapy,UniversityMedicalCenter,Mainz,Germany;2DepartmentofGynaecologyand Obstetrics, University Medical Center, Leipzig, Germany and 3Department of Pathology, University Medical Center, Leipzig, Germany Background:Experimentalstudieshaveestablishedacausalconnectionbetweentumourhypoxia,hypoxia-associatedproteome changesanddownregulationofE-cadherin,thefinalcommonpathwayofepithelial-to-mesenchymaltransition(EMT).Ourstudy aimedat elucidating theinterrelationship of theseprocesses incancers of theuterine cervix invivo. Methods: Tumour oxygenation was assessed in 48 squamous cell carcinomas (SCC) of the uterine cervix using polarographic needle electrodes. The expression pattern of E-cadherin was investigated by immunohistochemistry and western blotting, and wascomparedwiththatofthehypoxia-inducibleproteinsglucosetransporter(GLUT)-1andcarbonicanhydrase(CA)IXinbiopsy specimens of theoxygenation measurementtracks. Results:Themajorityofcervicalcancers(52%)wereE-cadherinpositive,withacompleteabsenceoftheantigeninonly10%ofthe tumours. No correlation was found between the level of E-cadherin expression andthe oxygenation status (meanpO , median 2 pO andhypoxicfractions).InpatientsshowingpartialexpressionofE-cadherin(38%),stainingwasnotpreferentiallydiminishedin 2 GLUT-1- or CAIX-positive areas,and lossof E-cadherin occurredindependentlyof tumour cellscattering. Conclusion: Our data provide no evidence in favour of a hypoxia-induced EMT as a mechanistic basis of cervical cancer invasiveness. Direct measurements of the oxygenation status using microelec- regulated by oxygen availability (Wang et al, 1995). Under trodescarriedoutinpatientswithcancersoftheuterinecervix,the normoxic conditions, HIF-1a is modified by prolyl hydroxylase headandneckregionandinsofttissuesarcomashaveshownthat domain (PHD) enzymes and subsequently degraded via the criticallyloweredoxygenlevels(hypoxia)withinthetumourtissue proteasome,keepingitsproteinlevelsverylow.Hypoxiainterrupts are associated with a poor prognosis (Ho¨ckel et al, 1993, 1996, thisprocessandleadstoitsstabilisation,nucleartranslocationand 1998; Brizel et al, 1996; Nordsmark et al, 2005). In one study on bindingtoitsdimerisationpartnerHIF-1(cid:2)(Wenger,2002).HIF-1 cervical cancer, this correlation with outcome was independent of thenactivatesthetranscriptionofmorethan800genes(Semenza, the choice of treatment modality (Ho¨ckel et al, 1996), a finding 2012). Cellular processes driven by HIF-1 through this large-scale thathasindicatedthattumourhypoxiamaybeasignificantfactor shiftingeneexpressionareincreasinglyviewedaskeymediatorsof driving malignant progression (Ho¨ckel et al, 1996; Vaupel et al, hypoxia-associated malignant progression. This idea is supported 2004). With regard to the possible mechanisms underlying this by numerous studies that have shown clear correlations between phenomenon, hypoxia-induced proteome changes, which are elevated levels of hypoxia-inducible proteins and a poor patient coordinated primarily by the hypoxia-inducible factor (HIF)-1, prognosis(VaupelandMayer,2007).Hypoxia-induciblefactor-1- may be of pivotal importance. This transcription factor is a mediated mechanisms include the metabolic adaptation (‘repro- heterodimeric protein whose a-subunit (HIF-1a) is directly gramming’) to hypoxic, energy-deprived conditions, enabling *Correspondence:DrmedAMayer;E-mail:[email protected] ThismanuscriptpartlycontainsresultsfromtheDrmed.thesisofNadjaSchlischewsky. Received15October2012;accepted27November2012; publishedonline15January2013 &2013CancerResearchUK.Allrightsreserved0007–0920/13 402 www.bjcancer.com|DOI:10.1038/bjc.2012.570 E-cadherininhypoxic cervicalcancer BRITISH JOURNALOF CANCER cancer cells to survive in hypoxic niches, for example by the Table1.Patientandtumourcharacteristicsatthetimeof upregulation of glucose influx via glucose transporter (GLUT)-1 pretherapeuticpO measurements 2 andprotoneffluxbyamechanisminvolvingmembranouscarbonic anhydrase (CA) IX. Owing to their longer half-lives and greater No.ofpatients % Range protein abundance compared with HIF-1a, both proteins have FIGOstage been used as surrogate markers for the activation of the hypoxic IB1 5 10.4 response in numerous studies. IB2 4 8.3 Hypoxia-associatedproteomechangeshavealsorepeatedlybeen IIA 2 4.2 connected to increased cancer cell invasion and metastasis – core IIB 8 16.7 traits of malignancy – by means of the induction of epithelial-to- IIIB 19 39.6 mesenchymal transitions (EMTs). As the term implies, this IVA 7 14.6 complexphenotypictransformationcomprisesthelossofepithelial IVB 3 6.3 and acquisition of mesenchymal characteristics, including an Grade enhanced motility and local invasiveness associated with single- 1 4 8.3 cell ‘scattering’. Abolition of the tight cohesion found in normal 2 33 68.8 epitheliaabsolutelyrequiresthedissolution ofmechanicallystable 3 11 22.9 physiologicalcell-to-cellcontacts,inwhichtheproteinE-cadherin hasakeyrole.Severalinvivoandinvitrostudiesoftumoursand Largesttumourdiameter(mm)a 20–120 tumour cells of different origins indicate that HIF-1a may be o60 22 45.8 directly involved in the downregulation of E-cadherin. Mecha- X60 26 54.2 nisms include upregulation of the EMT-inducing transcription Patientage(yr)a 29–73 factorsSNAIL(Imaietal,2003)andTWIST(Yangetal,2008),as o50 21 43.8 wellastheinduction ofLOX(Erleretal,2006).Sofar,numerous X50 27 56.3 studies have already described a downregulation of E-cadherin in cervicalcancer(Vesseyetal,1995;Jeffersetal,1997;Narayanetal, cHb(g/dl)a 8.2–16.1 2003;Faleiro-RodriguesandLopes,2004;vandePutteetal,2004; o13.5 24 50 Rodriguez-Sastreetal,2005;Dursunetal,2007;Cabergetal,2008; X13.5 24 50 Lee et al, 2008). In fact, a progressive downregulation of this aThemedianvaluewasusedasthecut-offpointbetweenthetwogroups.Sumsofrounded proteinhasevenbeenobservedincervicaldysplasia(deBoeretal, percentagesmaynotaddtoexactly100percent. 1999). As cancers of the uterine cervix are the tumour entity for which the best evidence for hypoxia-associated malignant progression exists, this study has examined the question of whether correlations between the directly measured oxygenation status (Eppendorf pO histography), the expression of hypoxia- 2 associatedmarkerproteins(GLUT-1,CAIX)andtheexpressionof from those tumour areas from which pO readings had been 2 E-cadherin can be identified. obtained. Intravaginal temperature, arterial blood pressure, heart rate, haemoglobin concentration, haematocrit, and arterial oxy- haemoglobin saturation were monitored at the time when pO 2 MATERIALS AND METHODS readings were taken. The pretherapeutic pO measurements were 2 recorded 1–5 days before oncological treatment. The oxygenation Patients. Allofthepatientsinthisstudywerepartofaprospective status was described quantitatively by the parameters mean pO2, median pO and the percentage of pO values below 2.5, 5 and clinical trial for the evaluation of the significance of tumour 2 2 10mmHg (i.e., ‘hypoxic fractions’, HF 2.5, HF 5 and HF 10, oxygenationinprimary,locallyadvancedcarcinomasoftheuterine respectively). cervix at the Department of Obstetrics and Gynaecology, University Medical Centre, Leipzig, Germany. The study design Immunohistochemistry and morphometry. Histologic slides wasapprovedbythelocalMedicalEthicsCommittee,withpatients were prepared from the paraffin blocks and dried overnight at giving informed written consent before being enrolled. All 48 371C. On the next day, specimens were dewaxed in xylene and patients with squamous cell carcinomas (SCCs) for whom biopsy then rehydrated in a descending alcohol series. Retrieval of specimens of the oxygen measurement tracks were available were antigenic binding sites was performed by heating specimens in included in this study (for patient and tumour characteristics see 10/1mM Tris/EDTA buffer (pH 9.0) in a steamer (Braun FS 10, Table 1). Braun,Kronberg,Germany)for40min.Primaryantibodiesagainst Tumour oxygen tension measurements. Tumour pO was E-cadherin, GLUT-1 and CD34 were obtained from DAKO 2 measured pretherapeutically with the computerised Eppendorf (DAKO, Hamburg, Germany, cat.-no. M3612, A3536 and pO histographysystem(Eppendorf,Hamburg,Germany),usinga M7165, respectively), and antibodies against CA IX and Ki67 2 protocol that has been described in detail (Ho¨ckel et al, 1991). were obtained from Abcam (Abcam, Cambridge, UK, cat.-no. Briefly, pO readings were performed in conscious patients ab15086 and ab16667). Detection of the first antigen, CD34, was 2 along linear electrode tracks, first in the subcutaneous fat of the carried out using overnight incubation with the primary antibody mons pubis followed by cervical measurements at the 12 and at 41C,appropriate biotinylated secondaryantibodies followedby 6 o’clock sites in macroscopically vital tumour tissue. Within a standard streptavidin/biotin/horseradish peroxidase reagent and the tumour tissue, up to 35 pO measurements were made along colour development using DAB (brown reaction product; DAKO 2 eachelectrodetrack(70readingsintotal),startingatatissuedepth Duet kit). Detection of the second target protein (E-cadherin, of approx. 5mm. The individual pO measurement points were GLUT-1 or CA IX) was preceded by a second round of antigen 2 situated 0.7mm apart, resulting in an overall measurement retrieval to prevent cross-reactions with the first antigen and to track length of 25mm. Immediately after pO measurements, enhance unmasking efficiency. After primary antibody incubation 2 needle core biopsies (obtained using Biopty; Radioplast, Uppsala, for1hat371,themicropolymer-based,biotin-freeVectorImpress Sweden) of 2mm in diameter and 20mm in length were taken detection system was applied and Vector VIP (purple reaction www.bjcancer.com|DOI:10.1038/bjc.2012.570 403 BRITISHJOURNAL OF CANCER E-cadherininhypoxic cervical cancer product) was used as the peroxidase substrate (Vector Labora- RESULTS tories, Burlingame, CA, USA). Negative control specimens were incubated in PBS without the primary antibody under the same conditions.Atumourspecimenwithaknownstrongexpressionof Prevalence of E-cadherin in SCC of the uterine cervix. On each antigen was run as a positive control with every staining performing immunohistochemistry (IHC), the majority (52%) of batch. Slides were dehydrated in an ascending alcohol series, and 48 squamous cell carcinomas of the uterine cervix showed strong coveredwithacoverslipusingEukittmountingmedium(Riedel-de membranousE-cadherinexpressioninalltumourcells.Inall,38% Haen, Seelze, Germany). Digital images of the specimens were of SCCs showed a partial loss (ranging from no more than focal acquired with a Zeiss AxioImager microscope and AxioCam HRc areas of absence of the antigen to its expression in only selected using the software AxioVision (Zeiss, Oberkochen, Germany). groups of cells). In the remaining 10%, E-cadherin was not Diffusiondistancesweremeasuredin2to15areasineachbiopsy detectableatall.Conversely,9of37(24%)specimensavailablefor using measure tools in AxioVision. western blot (WB) analyses of E-cadherin expression were negative, whereas the remaining showed 120-kDa bands of Quantitative assessment of antigen expression. Only immuno- different intensities (Supplementary Figure S1). A significant staining compatible with the known biological function and positive correlation between both assays was found (P¼0.01), corresponding subcellular localisation of each antigen was butindividualcasesshowedsignificantdisagreement;forexample, considered as being evaluable as marker expression: membranous one tumour showed E-cadherin expression in all cells in IHC but staininginthecaseofE-cadherin,GLUT-1andCAIX,andnuclear wastestednegativeusingWB.Owingtothelowersensitivityofthe stainingforKi67.Asemiquantitativescoringsystemwasused.For WB assay, all further results refer to the IHC data. E-cadherin,score0,negative;score1,partiallynegative;score2,all cells positive. For GLUT-1, CA IX and Ki67, score 0, no staining E-cadherin and pO2-histography data. The Kruskal–Wallis Test (‘absent’); score 1, o10% positive (‘weak’); score 2, 11% to 50% yielded no significant differences in the values of any of the positive (‘moderate’); and score 3, 450% positive (‘strong’). oxygenation status parameters (mean pO2, median pO2, HF 2.5, Because of limited amounts of tissue, only 45 (of 48) biopsies HF 5, and HF 10) between the three categories of E-Cadherin stained for CA IX and 47 (of 48) stained for Ki67 could be expressionintheIHCanalysis(i.e.,thepresenceoftheantigenin analysed. no, some or all cancer cells within the tumour biopsy). Boxplots depicting these data are presented in Figure 1. Protein extraction, western blotting and band densitometry. E-cadherinandlevelsofGLUT-1,CAIXandKi67. Alltumours Whole-cell protein extracts (including membrane proteins) were showedavaryingdegreeofGLUT-1expression,whereasone-third generated using a FastPrep-24 tissue homogeniser (MP Biomedi- werenegativeforCAIX(seeTable2).HigherlevelsofE-cadherin cals, Solon, OH, USA). Cryospecimens were placed in 2.0ml of impact-resistanttubescontaining250mlofRIPAbuffer(C-C-pro, were correlated with higher expression scores of the endogenous Oberdorla, Germany) supplemented with 1mM fresh DTT and hypoxia-relatedmarkersGLUT-1(r¼0.51,Po0.001)andCAIX (r¼0.33,P¼0.03).ExpressionlevelsofGLUT-1andCAIXwere Lysing Matrix D (i.e., 1.4mm ceramic spheres, MP Biomedicals), also correlated with each other (r¼0.48, P¼0.001). Correlations homogenised for 10–20sec, incubated on ice for 30min and centrifuged at 15000 U/min for 30min at 41C. An aliquot of the with proliferation (Ki67) were not found. supernatant was used for the determination of the protein Expression patterns and interrelationships of E-cadherin, concentration using Bradford’s reagent (Roti-Quant, Roth, GLUT-1, CA IX and Ki67. Focal downregulation of E-cadherin Karlsruhe, Germany), BSA as a protein standard (Sigma-Aldrich, inGLUT-1-orCAIX-positiveareas,asexpectedfromtheconcept StLouis,MO,USA)andaTecanF200microplatereaderequipped ofhypoxia-induced EMT,wasnot observed. Insubregions ofmost witha595-nmfilter(Tecan, Crailsheim,Germany).Ameasure of biopsies (41 of 48), expression of GLUT-1 was found to increase 50mg of protein was separated by SDS-PAGE using a Bio-Rad beyond an average distance of 76mm from the nearest microvessel Mini-Protean3electrophoresischamberandtransferredtoPVDF- membranes using a Bio-Rad transblot SD transfer cell (Bio-Rad, Hercules, CA, USA). The primary antibody against human 35 E-cadherin (DAKO M3612), identical to the one used for IHC, was incubated overnight at 41C and detected using an HRP- 30 conjugated secondary antibody (NXA931, GE Healthcare, Waukesha, WI, USA). To control for equal protein loading, g) 25 H membraneswerestrippedandreprobedwithantibodiesagainst(cid:2)- m m aacntdina(pspc-r1o6p1ri6art,eSaannttia-rCabrubiztBsieoctoencdhanroyloagnyt,ibSoanditeasC(rsucz-2,0C5A4,,USaSnAta) O (2 20 p Cruz). Bands were developed using ECL reagent (GE Healthcare) n 15 andimageswerecapturedwithaCCDcamera(FujifilmLAS-3000, dia Fujifilm,Tokyo,Japan).Molecular-weightdeterminationwascarried Me 10 out using the Fujifilm Multigauge software and the Bio-Rad PrecisionPlusDualColourStandards,whichwererunoneachgel. 5 Statisticalanalysis. Allofthestatisticaltestswereperformedusing SPSS(Version20,IBM,Armonk,NY,USA).Thesignificancelevel 0 was setata¼5%forallcomparisons.Linearcorrelations between 0 1 2 two parameters were described by Spearman’s rank correlation E-cadherin score coefficient (r). Two-sided Mann–Whitney U tests and Kruskal– Wallistests were usedfor comparisonofcategorised variables. Figure1.NocorrelationbetweenE-cadherinexpressionandthe Oxygenation status was correlated with FIGO stages and with oxygenationstatusofcervixcancers.Boxplotsshowequal tumour size, which was dichotomized using the median tumour distributionsofmedianpO valuesinthethreecategoriesof 2 diameter (60mm) as the cut-off value. Diagrams were generated E-cadherinexpression.Similardistributionswerefoundformean using SigmaPlot 11. pO valuesandhypoxicfractions(seetextfordetails). 2 404 www.bjcancer.com|DOI:10.1038/bjc.2012.570 E-cadherininhypoxic cervicalcancer BRITISH JOURNALOF CANCER (CD34expression),asevaluatedbymultiplemanualmeasurements were tested as continuous variables. When tumour size was ineachdouble-stainedsection(seeFigure2Aforanexample).Of30 dichotomizedusingthemedianvalueasthecut-off,largertumours CAIX-positivebiopsies,23showedincreasedantigenexpressionata showed a trend towards a slightly poorer oxygenation status, but mean distance of 83mm from the nearest microvessel. In the only regarding the mean pO (P¼0.054). Higher FIGO stages, 2 remaining seven positive tumours, a relationship between CA IX however, did not show a poorer oxygenation status. expression and diffusion distances could not be identified. Conversely,nospatialrelationshipbetweenfocalareasofE-cadherin downregulationandmicrovesselswasobservedinanyofthepositive DISCUSSION biopsies(Figure2B).E-cadherinexpressionwasalsoindependentof the local growth pattern. To illustrate this observation, Figure 3A BeginningwiththeexperimentsofComan(1944),whowasableto showsanexampleofstronglyE-cadherin-positivecellsthatexhibita showalmost70yearsagothatscrapingsobtainedfromsquamous reticular pattern of infiltration. Conversely, Figure 3B shows cellcarcinomasoftheuterinecervixexhibitareducedintercellular compact and coherent tumour cell aggregates that completely lack cohesion, numerous studies have investigated the importance of theexpressionofE-cadherin. loosened cell–cell contacts for the invasiveness of cancers of this Other findings. No correlation was found between clinical and other tumour types. It had been recognised even earlier that tumour size and the oxygenation status when both parameters extracellular Ca2þ levels are important for cell-to-cell coherence, anditisnowestablishedthatthe(calcium-dependent) homotypic interaction of the extracellular domains of E-cadherin molecules, Table2.ExpressionscoresofGLUT-1andCAIXinsquamouscell locatedinadherensjunctions,iscrucialforthemechanicalstability carcinomasoftheuterinecervix of intercellular contacts (Perez-Moreno et al, 2003). Possible GLUT-1 CAIX mechanisms mediating a specific downregulation of these structures in hypoxic tumour areas are of great interest, as Score numberofpatients % numberofpatients % numerous studies have demonstrated a significant prognostic 0 0 0 15 33.3 importance of hypoxia and hypoxia-associated proteome changes. 1 11 22.9 10 22.2 Our study is the first to investigate this question in SCCs of the 2 13 27.1 16 35.6 uterine cervix. Our results clearly demonstrate that hypoxic tumours or tumours withahigh expression of hypoxia-associated 3 24 50.0 4 8.9 proteins do not show downregulation of E-cadherin or other Total 48 100.0 45 100.0 characteristic signs of EMT. On the contrary, we have even Figure2.ExpressionpatternofGLUT-1,butnotE-cadherin,isrelatedtodiffusiongradientsincancersoftheuterinecervix.GLUT-1 expression(purple)showsagradualincreasewithincreasingdistancefromthemicrovessels(brown)inpanel(A).CircularareasoflowGLUT-1 expression(markedbywhitedashedlines)canbeidentifiedaroundselectedmicrovessels.NosuchpatternisevidentforE-cadherin(purple)in panel(B).200xmagnification,scalebarinB¼100mm. Figure3.ExpressionofE-cadherinisnotrelatedtothemicroregionalgrowthpatternofcancersoftheuterinecervix.Areticular,highly branchedpatternofinvasion,includingapparentlyscatteredsinglecells(arrowheads),canbeobservedintumourswithstrongmembranous E-cadherinexpression(A).Conversely,compact,spheroid-likeconfigurationsoftumourcellclusters(demarcatedbyawhitedashedline)canbe observedintumourscompletelynegativeforE-cadherin(B).200xmagnification,scalebarinB¼100mm. www.bjcancer.com|DOI:10.1038/bjc.2012.570 405 BRITISHJOURNAL OF CANCER E-cadherininhypoxic cervical cancer observedsignificantpositivecorrelationsbetweentheexpressionof vast majority of tumours. However, in our hands, western blots the hypoxia-associated proteins (GLUT-1 and CA IX) and turnedouttobesignificantlylesssensitivethanIHC,whichiswhy E-cadherin. Such a correlation would seem plausible if a reduced wereliedondatafromthelatterassayforallfurtheranalyses.The E-cadherin expression were to be accompanied by increased cell inability to check for the amount of stromal tissue present in the scattering, which, in turn, would lead to the ‘reoxygenation’ of portion of the biopsy lysed for protein extraction is a significant tumour cells upon their infiltration into the microvessel-carrying drawback of this method and may well be decisive for the and thus ‘oxygen-providing’ tumour stroma. This possibility was, quantitative differences between the results of these two methods. however, excluded in this study, as in terms of E-cadherin We are aware of the fact that the finding of preserved E-cadherin expression no systematic differences between highly fragmented expression in the cell membrane of the majority of cancer cells tumourcellclustersandisolatedtumourcellsontheonehandand investigated in our study does not necessarily imply that these coherent tumour cell aggregates on the other could be identified. molecules are fully functional. However, this aspect was not As shown in Figure 3, examples both of tumour areas in the first investigated in our work, as the hypothesis explicitly implied a category (scattered phenotype), which consistently showed strong downregulationofproteinabundance.Thistopicmay,however,be expression of E-cadherin, and areas in the second category an interesting subject for further studies. E-Cadherin expression (cohesive phenotype), which completely lacked the expression of levels were also not correlated with the proliferation rate, but this antigen, have been found. These observations confirm earlier knowledgeoftheexpressionpatternofthisantigeninconsecutive findings in cancers of the uterine cervix (Jeffers et al, 1997) and sectionsenabledustoascertainthatouranalyseswerecarriedout breast cancer (Colpaert et al, 2003), which also did not show an in proliferating, and hence vital, areas of the tumours. impact of E-cadherin expression on the invasion pattern. With In keeping with the results of our prior studies in the same regard to the positive correlations between E-cadherin and tumour entity (Mayer et al, 2005a, b), we again did not observe hypoxia-related markers, we currently have no biologically correlationsbetweenendogenoushypoxia-relatedmarkersGLUT- plausible explanation for this finding, but would assume that it is 1 and CA IX and any parameter of the oxygenation status in this unlikelythatthesefindingsaretheresultoftechnicalshortcomings completely different and more recent cohort of patients. As our of our immunohistochemical protocols, as these were already protocol specified biopsy acquisition from tumour microregions validated extensively in prior studies (e.g., Mayer et al (2008)). identical to the areas in which oxygenation measurements had Jeffersetal(1997)alsoreportedE-cadherinexpressioninall20 been made, the important influence of tumour heterogeneity on tumours investigated in their study, i.e., even more than the 90% the expression of hypoxia-related markers has already been positive tumours identified in our work (in the IHC assay). minimisedbyourapproach.However,tumourheterogeneityalone Nevertheless, both studies stand out among the literature, which isprobablynotsufficienttoexplainthesediscrepancies,aseventhe commonly reported much lower E-cadherin levels. For example, elaborate IHC analysis protocol of systematically investigating the largest study to date, the work of van de Putte et al (2004), 3levelsoftwodifferentbiopsiesusedinthestudyofIakovlevetal which comprised 219 patients, reported a high expression of (2007) did not result in the identification of correlations between E-Cadherin(inmorethan50%ofthetumourcells)beingrestricted the oxygenation status and the expression of CA IX. Correlations toonly10%ofthepatientsanalysed,whereasourownresultsshow between HIF-dependent markers and oxygen microsensor mea- 100%positivetumourcellsinmorethanhalfofourpatientcohort. surements may also be distorted by the occurrence of tumour Furthermore, as opposed to our study, which found only 10% of necroses, which are severely hypoxic using needle electrode patientstohavecompletelynegativetumours,relativelyhighrates measurements (Lyng et al, 1997; Jenkins et al, 2000) but do not of negative tumours (34 to 100%) have been reported in the necessarily go hand-in-hand with extensive expression of HIF- majority of prior studies (e.g., Ancuta et al (2009); Faleiro- relatedmarkersinthesamespecimens.Theinfluenceofthisfactor Rodrigues and Lopes (2004); Lee et al (2008); Rodriguez-Sastre is also expected to be rather modest in our study, as pO 2 etal(2005)).Thesedifferencesmaypartlybeexplainedbytheuse measurements and corresponding biopsies were excluded from of different tissue fixation protocols, primary antibody clones and analyses in cases in which histopathological examination showed detectionsystemsoracombinationofthesefactors.Indeed,theuse widespread tumour necrosis. Conversely, it has long been known of the purple peroxidase substrate VIP for the detection of that the expression levels of both HIF-1a and its target genes E-cadherin - which in our study was performed to enable double GLUT-1andCAIX,asinvestigatedinthisstudy,areinfluencedby staining with CD34 - resulted in much stronger (yet decisively avarietyoffactorsotherthanhypoxia,suchasglucosedeprivation specific)stainingresultsthanDAB,astestedduringthepilotphase (Boado and Pardridge, 2002; Kwon and Lee, 2005), acidosis of our study (data not shown). Particular incongruities exist (Mekhailetal,2004),oncogenes(e.g.,c-MYC(Osthusetal,2000)) between our study and the report from Lee et al (2008), who and oncogene-associated signalling pathways (e.g., PI3K/AKT studied SCC of the uterine cervix using immunofluorescence (Kaluz et al, 2002)). These significant hypoxia-independent detection of E-cadherin and vimentin in specimens, which, influences have been summarised in a review (Mayer et al, according to the interpretation of these authors, represented an 2006). In addition, recent research has demonstrated a role for ascendingorderoflocalspread(i.e.,normalsquamousepithelium, reactive oxygen species and NO in stabilizing HIF-1a (Fong and superficial tumour tissue, tumour cell nests in the parametrium Takeda, 2008; Dewhirst, 2009). andtumourcellsinpelviclymphnodes).Theyfoundaprogressive Beyond mere quantitative correlations between E-cadherin, decrease of E-cadherin and an opposed induction of vimentin, GLUT-1 and CA IX expression levels, we were able to carry out suggesting bona fide EMT in 10 out of 10 patients analysed. systematic computer-aided classifications of antigen expression Interestingly, their results agree with our data in one important patternswithregardtothelocaldiffusiongeometry(bymeasuring aspect: E-cadherin downregulation obviously did not necessarily distancesfrommarker-positivecellstothenearestmicrovessel),as lead to single-cell scattering, as is evident from the E-cadherin- sectionsofallthreeantigenshadeachbeendouble-stainedforthe negative, but still compact, tumour cell nodule found in the markerCD34(microvascularendothelium).Resultsofthisanalysis parametrium in the study of Lee et al (see Figure 1A in Lee et al showed that most of the (antigen-positive) tumours exhibited the (2008)compared to Figure 3Bofthisstudy).Nevertheless, weare expectedpatternofexpressionofGLUTandCAIXbeyondamean currently unable to explain the substantial differences between distance of approx. 80mm (76mm for GLUT-1 and 83mm for CA these two studies. IX)fromthenearestmicrovesselsinsubregionsofthebiopsies.In DetectionofE-Cadherinusingwesternblottingoftissueextracts contrast, such a constellation could not be identified for confirmedthefindingofmaintainedE-cadherinexpressioninthe E-cadherin, which was typically expressed both in the immediate 406 www.bjcancer.com|DOI:10.1038/bjc.2012.570 E-cadherininhypoxic cervicalcancer BRITISH JOURNALOF CANCER vicinity of blood vessels and areas further down the oxygen DewhirstMW(2009)Relationshipsbetweencyclinghypoxia,HIF-1, diffusiongradients.Theabsenceofanysystematicspatialrelation- angiogenesisandoxidativestress.RadiatRes172:653–665. ship between E-cadherin and the hypoxia-associated proteins DursunP,YuceK,UsubutunA,AyhanA(2007)Lossofepitheliumcadherin clearly contradicts the model of a hypoxia-mediated downregula- expressionisassociatedwithreducedoverallsurvivalanddisease-free tion of E-cadherin in cancers of the uterine cervix. survivalinearly-stagesquamouscellcervicalcarcinoma.IntJGynecol Cancer17:843–850. Theresults obtained inthisstudyindicate thatdownregulation ErlerJT,BennewithKL,NicolauM,Dornho¨ferN,KongC,LeQT,ChiJT, of E-cadherin in cancers of the uterine cervix does not show JeffreySS,GiacciaAJ(2006)Lysyloxidaseisessentialforhypoxia-induced associations with the ‘physical’ oxygenation status of the tissue or metastasis.Nature440:1222–1226. withthelocoregionalexpressionlevelsofhypoxia-relatedmarkers. Faleiro-RodriguesC,LopesC(2004)E-cadherin,CD44andCD44v6in Hence, other mechanisms leading to E-cadherin downregulation squamousintraepitheliallesionsandinvasivecarcinomasoftheuterine may predominate in this tumour entity. However, the possibility cervix:animmunohistochemicalstudy.Pathobiology71:329–336. thatthegeneralimportanceofE-cadherindownregulation(andof FongGH,TakedaK(2008)Roleandregulationofprolylhydroxylasedomain theassociatedEMT)maybesmallerinSCCsoftheuterinecervix proteins.CellDeathDiffer15:635–641. than in other tumour entities should also be considered. It is well FriedlP,WolfK(2003)Tumour-cellinvasionandmigration:diversityand known that alternative patterns of tumour invasion exist (Friedl escapemechanisms.NatRevCancer3:362–374. andWolf,2003)and,accordingtoonestudy(Wickietal,2006),a GaggioliC,HooperS,Hidalgo-CarcedoC,GrosseR,MarshallJF,Harrington K,SahaiE(2007)Fibroblast-ledcollectiveinvasionofcarcinomacellswith collective migration pattern may in fact predominate in SCCs of differingrolesforRhoGTPasesinleadingandfollowingcells.NatCellBiol the uterine cervix, consistent with our finding of widespread 9:1392–1400. E-cadherinexpressioninthisentity.Inaddition,inarecentstudy, Ho¨ckelM,KnoopC,SchlengerK,VorndranB,BaussmannE, wehave identifiedrecurrent histologicalsignsindicating anactive MitzeM,KnapsteinPG,VaupelP(1993)IntratumoralpO 2 roleofthetumourstromainmediatinginvasivenessintheformof predictssurvivalinadvancedcanceroftheuterinecervix.Radiother cell collectives in this tumour entity (Mayer et al, 2011). Indeed, Oncol26:45–50. priorexperimentalfindingshavedirectlyvisualisedhowfibroblasts Ho¨ckelM,SchlengerK,AralB,MitzeM,Scha¨fferU,VaupelP(1996) may enable the invasion of cancer cells derived from SCCs (e.g., Associationbetweentumorhypoxiaandmalignantprogressionin Gaggioli et al, (2007)). Future work is needed to clarify to what advancedcanceroftheuterinecervix.CancerRes56:4509–4515. extent hypoxia-mediated mechanisms may have a role in the Ho¨ckelM,SchlengerK,Ho¨ckelS,AralB,Scha¨fferU,VaupelP(1998) context of these alternative types of cancer invasion. Tumorhypoxiainpelvicrecurrencesofcervicalcancer.IntJCancer79: 365–369. Ho¨ckelM,SchlengerK,KnoopC,VaupelP(1991)Oxygenationof carcinomasoftheuterinecervix:evaluationbycomputerizedO tension 2 measurements.CancerRes51:6098–6102. ACKNOWLEDGEMENTS IakovlevVV,PintilieM,MorrisonA,FylesAW,HillRP,HedleyDW(2007) Effectofdistributionalheterogeneityontheanalysisoftumorhypoxia The authors thank Dr. Debra Kelleher for her valuable editorial basedoncarbonicanhydraseIX.LabInvest87:1206–1217. help during preparation of this manuscript. This work was ImaiT,HoriuchiA,WangC,OkaK,OhiraS,NikaidoT, supported by a grant from the Deutsche Krebshilfe (no. 106758). KonishiI(2003)HypoxiaattenuatestheexpressionofE-cadherin viaup-regulationofSNAILinovariancarcinomacells.AmJPathol163: 1437–1447. JeffersMD,PaxtonJ,BolgerB,RichmondJA,KennedyJH,McNicolAM CONFLICT OF INTEREST (1997)E-cadherinandintegrincelladhesionmoleculeexpressionin invasiveandinsitucarcinomaofthecervix.GynecolOncol64:481–486. The authors declare no conflict of interest. JenkinsWT,EvansSM,KochCJ(2000)Hypoxiaandnecrosisinrat9L gliomaandMorris7777hepatomatumors:comparativemeasurements usingEF5bindingandtheEppendorfneedleelectrode.IntJRadiatOncol BiolPhys46:1005–1017. REFERENCES KaluzS,KaluzovaM,ChrastinaA,OlivePL,PastorekovaS,PastorekJ, LermanMI,StanbridgeEJ(2002)Loweredoxygentensioninduces AncutaE,AncutaC,CozmaLG,IordacheC,Anghelache-LupascuI,AntonE, expressionofthehypoxiamarkerMN/carbonicanhydraseIXinthe CaraseviciE,ChirieacR(2009)Tumorbiomarkersincervicalcancer: absenceofhypoxia-induciblefactor1astabilization:arolefor focusonKi-67proliferationfactorandE-cadherinexpression.RomJ phosphatidylinositol3’-kinase.CancerRes62:4469–4477. MorpholEmbryol50:413–418. KwonSJ,LeeYJ(2005)Effectoflowglutamine/glucoseonhypoxia-induced BoadoRJ,PardridgeWM(2002)Glucosedeprivationandhypoxiaincrease elevationofhypoxia-induciblefactor-1ainhumanpancreaticcancer theexpressionoftheGLUT1glucosetransporterviaaspecificmRNAcis- MiaPaCa-2andhumanprostaticcancerDU-145cells.ClinCancerRes11: actingregulatoryelement.JNeurochem80:552–554. 4694–4670. BrizelDM,ScullySP,HarrelsonJM,LayfieldLJ,BeanJM,ProsnitzLR, LeeMY,ChouCY,TangMJ,ShenMR(2008)Epithelial-mesenchymal DewhirstMW(1996)Tumoroxygenationpredictsforthelikelihoodof transitionincervicalcancer:correlationwithtumorprogression, distantmetastasesinhumansofttissuesarcoma.CancerRes56:941–943. epidermalgrowthfactorreceptoroverexpression,andsnailup-regulation. CabergJH,HubertPM,BegonDY,HerfsMF,RoncaratiPJ,BoniverJJ, ClinCancerRes14:4743–4750. DelvennePO(2008)SilencingofE7oncogenerestoresfunctional LyngH,SundforK,RofstadEK(1997)Oxygentensioninhumantumours E-cadherinexpressioninhumanpapillomavirus16-transformed measuredwithpolarographicneedleelectrodesanditsrelationshipto keratinocytes.Carcinogenesis29:1441–1447. vasculardensity,necrosisandhypoxia.RadiotherOncol44:163–169. ColpaertCG,VermeulenPB,vanBeestP,SoubryA,GoovaertsG,DirixLY, MayerA,Ho¨ckelM,HornLC,SchmidbergerH,VaupelP(2011)GLUT-1 HarrisAL,vanMarckEA(2003)Cutaneousbreastcancerdepositsshow stainingofsquamouscellcarcinomasoftheuterinecervixidentifiesa distinctgrowthpatternswithdifferentdegreesofangiogenesis,hypoxia novelelementofinvasion.IntJOncol38:145–150. andfibrindeposition.Histopathology42:530–540. MayerA,Ho¨ckelM,VaupelP(2005a)CarbonicanhydraseIXexpression ComanDR(1944)Decreasedmutualadhesiveness,apropertyofcellsfrom andtumoroxygenationstatusdonotcorrelateatthemicroregionallevel squamouscellcarcinomas.CancerRes4:625–629. inlocallyadvancedcancersoftheuterinecervix.ClinCancerRes11: deBoerCJ,vanDorstE,vanKriekenH,Jansen-vanRhijnCM,WarnaarSO, 7220–7225. FleurenGJ,LitvinovSV(1999)Changingrolesofcadherinsandcatenins MayerA,Ho¨ckelM,VaupelP(2006)Endogenoushypoxiamarkersinlocally duringprogressionofsquamousintraepitheliallesionsintheuterine advancedcancersoftheuterinecervix:realityorwishfulthinking? cervix.AmJPathol155:505–515. StrahlentherOnkol182:501–510. www.bjcancer.com|DOI:10.1038/bjc.2012.570 407 BRITISHJOURNAL OF CANCER E-cadherininhypoxic cervical cancer MayerA,Ho¨ckelM,WreeA,LeoC,HornLC,VaupelP(2008)Lackof vandePutteG,KristensenGB,BaekelandtM,LieAK,HolmR(2004) hypoxicresponseinuterineleiomyomasdespiteseveretissuehypoxia. E-cadherinandcateninsinearlysquamouscervicalcarcinoma.Gynecol CancerRes68:4719–4726. Oncol94:521–527. MayerA,Ho¨ckelM,WreeA,VaupelP(2005b)Microregionalexpressionof VaupelP,MayerA(2007)Hypoxiaincancer:significanceandimpacton glucosetransporter-1andoxygenationstatus:lackofcorrelationinlocally clinicaloutcome.CancerMetastasisRev26:225–239. advancedcervicalcancers.ClinCancerRes11:2768–2773. VaupelP,MayerA,Ho¨ckelM(2004)Tumorhypoxiaandmalignant MekhailK,GunaratnamL,BonicalziME,LeeS(2004)HIFactivationby progression.MethodsEnzymol381:335–354. pH-dependentnucleolarsequestrationofVHL.NatCellBiol6: VesseyCJ,WildingJ,FolarinN,HiranoS,TakeichiM,SoutterP,StampGW, 642–647. PignatelliM(1995)AlteredexpressionandfunctionofE-cadherinin NarayanG,Arias-PulidoH,KoulS,VargasH,ZhangFF,VillellaJ, cervicalintraepithelialneoplasiaandinvasivesquamouscellcarcinoma. SchneiderA,TerryMB,MansukhaniM,MurtyVV(2003)Frequent JPathol176:151–159. promotermethylationofCDH1,DAPK,RARB,andHIC1genesin WangGL,JiangBH,RueEA,SemenzaGL(1995)Hypoxia-induciblefactor1 carcinomaofcervixuteri:itsrelationshiptoclinicaloutcome.MolCancer isabasic-helix-loop-helix-PASheterodimerregulatedbycellularO 2 2:24. tension.ProcNatlAcadSciUSA92:5510–5514. NordsmarkM,BentzenSM,RudatV,BrizelD,LartigauE,StadlerP,Becker WengerRH(2002)Cellularadaptationtohypoxia:O-sensingprotein 2 A,AdamM,MollsM,DunstJ,TerrisDJ,OvergaardJ(2005)Prognostic hydroxylases,hypoxia-inducibletranscriptionfactors,andO-regulated 2 valueoftumoroxygenationin397headandnecktumorsafterprimary geneexpression.FASEBJ16:1151–1162. radiationtherapy.Aninternationalmulti-centerstudy.RadiotherOncol WickiA,LehembreF,WickN,HantuschB,KerjaschkiD,ChristoforiG 77:18–24. (2006)Tumorinvasionintheabsenceofepithelial-mesenchymal OsthusRC,ShimH,KimS,LiQ,ReddyR,MukherjeeM,XuY,WonseyD, transition:podoplanin-mediatedremodelingoftheactincytoskeleton. LeeLA,DangCV(2000)Deregulationofglucosetransporter1and CancerCell9:261–272. glycolyticgeneexpressionbyc-Myc.JBiolChem275:21797–21800. YangMH,WuMZ,ChiouSH,ChenPM,ChangSY,LiuCJ,TengSC,WuKJ Perez-MorenoM,JamoraC,FuchsE(2003)Stickybusiness:orchestrating (2008)DirectregulationofTWISTbyHIF-1alphapromotesmetastasis. cellularsignalsatadherensjunctions.Cell112:535–548. NatCellBiol10:295–305. Rodriguez-SastreMA,Gonzalez-MayaL,DelgadoR,LizanoM,TsubakiG, MoharA,Garcia-CarrancaA(2005)AbnormaldistributionofE-cadherin This work is licensed under the Creative Commons and(cid:2)-cateninindifferenthistologictypesofcanceroftheuterinecervix. Attribution-NonCommercial-ShareAlike3.0Unported GynecolOncol97:330–336. License.Toviewacopyofthislicense,visithttp://creativecommons. SemenzaGL(2012)Hypoxia-induciblefactors:mediatorsofcancerprogression andtargetsforcancertherapy.TrendsPharmacolSci33:207–214. org/licenses/by-nc-sa/3.0/ Supplementary Information accompanies this paper on British Journal of Cancer website (http://www.nature.com/bjc) 408 www.bjcancer.com|DOI:10.1038/bjc.2012.570