Role of E-Cadherins in Development of LymphaticTumor Emboli

Anita Gupta, M.D.1

Charuhas G. Deshpande, M.B.B.S., M.D.1

Sunil Badve, M.B.B.S., M.D.1,2

1 Department of Pathology, Northwestern Univer-sity, Chicago, Illinois.

2 Department of Pathology, Indiana UniversitySchool of Medicine, Indianapolis, Indiana.

Supported by National Institutes of Health SPOREGrant 1P 50 CA89018-01 and by funds from theAvon Products Foundation Breast Cancer Research& Care Program.

The authors thank Dr. Susan Perkins for help withthe statistics section.

Address for reprints: Sunil Badve, M.D., Division ofSurgical Pathology, University Hospital, Room3465, 550 North University Boulevard, Indianapo-lis, IN 46202; Fax: (317) 274-4356; E-mail:sbadve@iupui.edu

Received April 11, 2002; revision received Decem-ber 31, 2002; accepted January 9, 2003.

BACKGROUND. E-cadherin (E-cad) is a cell adhesion molecule that is expressed in

normal breast tissue. While loss of E-cad expression is a characteristic feature of

lobular carcinoma, it also is observed in infiltrating ductal carcinoma (IDC). The

presence of peritumoral intralymphatic emboli also is a poor prognostic feature in

IDC. Invasive lobular carcinoma rarely is associated with intralymphatic emboli. In

the current study, the authors assessed E-cad expression in cases of IDC with and

without intralymphatic tumor emboli to examine the potential role played by these

molecules in the development of lymphatic emboli.

METHODS. Fifty patients with high-grade invasive ductal carcinoma—25 with

prominent lymphatic invasion (LVI) and intralymphatic tumor emboli and 25

without LVI—were tested for expression of E-cad. For both groups, the intensity

and frequency of E-cad expression was evaluated in tumor cells and lymphatic

emboli; normal lobules were used as internal controls.

RESULTS. Membranous expression of E-cad was observed in normal lobules and

tumor cells in all patients, with the tumor cells exhibiting varying degrees of loss of

expression. In the 25 LVI-positive patients, the majority of tumor cells (including

intralymphatic emboli) expressed E-cad with an intensity and distribution similar

to what was seen in normal lobules. In the LVI-negative patients, the intensity and

the distribution of E-cad staining varied significantly. Tumor cells at the tumor-

stroma interface showed a greater frequency and intensity of E-cad expression

than did cells in the central region of the tumor.

CONCLUSIONS. Strong expression of E-cad was observed in LVI-positive patients

with high-grade IDC but not in LVI-negative patients. Emboli also exhibited high-

intensity expression. These findings, taken in conjunction with the knowledge that

intralymphatic tumor emboli in lobular carcinoma (which is E-cad-negative) are

rare, suggest that E-cad plays an important role in tumor development and growth

within the lymphatics. Cancer 2003;97:2341–7. © 2003 American Cancer Society.

DOI 10.1002/cncr.11332

KEYWORDS: breast cancer, E-cadherin, intralymphatic emboli.

E-cadherin (E-cad) is a 120-kilodalton transmembrane glycoproteinthat is distantly related to the immunoglobulin superfamily.1,2 It

mediates calcium-dependent, homotypic cell-cell adhesion, therebymaintaining the normal adhesive phenotype of epithelial cells.3 E-cadmolecules exist on cell membranes as dimers that, in the presence ofcalcium, form ‘zipperlike’ complexes4 with E-cad molecules fromadjacent cells.5,6 E-cad is believed by some7,8 to act as a tumorsuppressor that attenuates malignant behavior. However, E-cad-neg-ative tumors still are able to form tubules.9 Loss of E-cad expressionis thought to facilitate tumor cell detachment from primary tu-mor.10,11 Reduction or loss of expression of either E-cad or catenins(which, along with E-cad, form E-cad/catenin units) is associated with

2341

© 2003 American Cancer Society

invasion, metastasis, and poor prognosis in severaltypes of human malignancies,12,13 including breastcancer.7,14

Absence of E-cad due to a genetic mutation is acharacteristic feature of lobular carcinomas.7 The lossof expression is more variable in ductal carcinomas,with high-grade tumors exhibiting a greater tendencytoward reduction or loss of expression.14 As a result,expression of E-cad has been used to help distinguishbetween ductal and lobular carcinomas.15 Loss of nor-mal E-cad expression in invasive ductal carcinomas(IDC) is usually due to epigenetic mechanisms ratherthan genetic mutations.16,17 Re-expression of E-cadhas been noted in metastatic sites.18 The underlyingmechanisms of this phenomenon could be loss ofepigenetic inhibition or metastases arising fromclones of tumor cells that never lost E-cad expres-sion.18

Loss of normal E-cad expression in IDC is anindicator of increased invasiveness and dedifferentia-tion.14 Even small (T1a– b) or low– histologic grade(Grade I) breast tumors (per AJCC) were found to showreduction in E-cad expression; this finding suggeststhat reduction or loss of expression is an early event.19

Loss of expression was observed to be associated withhigh histologic grade, lymph node involvement, anddistant metastases in breast cancer.14,20 Reduced E-cad expression was found to be associated with short-ened disease-free survival even without lymph nodeinvasion at diagnosis.21–23

High-grade ductal carcinomas, which show vari-able loss of E-cad expression, frequently are associ-ated with intralymphatic tumor emboli. This featurerarely (if ever) is seen in lobular carcinomas, whichcharacteristically have E-cad mutations. Based onthese observations, we postulated that E-cad expres-sion might play an important role in the formation ofintralymphatic tumor emboli. In the current study, weanalyzed the expression of E-cad in fifty patients withhigh-grade ductal carcinoma. Twenty-five of these pa-tients exhibited prominent intralymphatic tumor em-boli (LVI), whereas the other 25 did not. Comparativeanalysis of LVI-positive and LVI-negative patients wasperformed to test for the presence of differences inE-cad expression patterns.

MATERIALS AND METHODSCases of high-grade IDC from the files of the Depart-ment of Pathology at Northwestern Memorial Hospitalwere reviewed. Twenty-five consecutive cases of pa-tients with LVI and 25 cases of patients without LVIwere selected. Because the presence of lymphatic in-vasion is strongly correlated with lymph node involve-ment, most patients without LVI did not have lymph

node involvement. For the purposes of the study, lym-phatic tumor emboli were defined as the presence ofemboli in lymphatics in the breast parenchyma adja-cent to or well beyond the margin of the invasivetumor. To prevent confusion with tumor retraction,intratumoral lymphatics were not considered for anal-ysis. Clinical data and pathology reports for all 50patients were analyzed to identify age, grade, presenceof lymphatic invasion, lymph node status, estrogenand progesterone receptors, and expression ofc-erb-B2 and p53. Histologic grading of tumors wasconfirmed by the modified Richardson and Bloommethod.24 One block from each patient was sectionedfor analysis of E-cad expression.

Because lymphatic emboli are best discerned atthe periphery of the tumor, in all cases a block thatwas taken from the peripheral zone of the tumor andcontained normal elements was selected for analysis.

ImmunohistochemistryFour-micron-thick sections of each block were cutonto silane-coated plus slides. The sections then weredeparaffinized and incubated in 3% H2O2 in methanolfor 30 minutes. Antigen retrieval was performed in amicrowave pressure cooker with 0.1 M pH 6.0 Citrabuffer (BioGenex, San Ramon, CA). Tissue sectionswere incubated with mouse monoclonal immuno-globulin G (IgG) anti-E-cad antibody (dilution, 1:700;Zymed Laboratories Inc., South San Francisco, CA) for2 hours at 37 °C. Bound primary antibody was visual-ized by the biotinyl-tyramide–streptavidin– horserad-ish peroxidase amplification technique (NEN LifeScience Products, Inc., Boston, MA). 3,3�-diaminoben-zamidine was used as a chromogen, and Harris hema-toxylin was used as a counterstain. Normal breasttissue sections incubated with phosphate buffer solu-tion (PBS) or isotypic mouse serum instead of theprimary antibody served as the appropriate negativecontrols.

To exclude false positive results due to tyramidesignal amplification (TSA), lobular carcinoma, whichlacks E-cad expression, was used as a negative control.Similar staining patterns were observed in controlsand test cases when staining was repeated for a smallnumber of samples using the Ventana Benchmark�

autostainer (Ventana Medical Systems, Inc., Tucson,AZ).

EvaluationThe distribution and the intensity of E-cad reactivitywere analyzed for each patient using a semiquantita-tive scoring system. The percentages of cells lining theadjacent normal terminal duct lobular units, tumorcells, and lymphatic tumor emboli that expressed E-

2342 CANCER May 1, 2003 / Volume 97 / Number 9

cad were measured. Staining was graded as 4� if� 75% of cells showed reactivity, 3� if 50 –74% showedreactivity, 2� if 26 – 49% showed reactivity, 1� if1–25% showed reactivity, and 0 if there was no evi-dence of staining. The intensity of staining (low orhigh) and pattern of E-cad expression in the tumor(peripheral or homogenous) also were recorded. Foreach patient, normal lobules served as a positive in-ternal control.

Clinicopathologic data were summarized and fre-quency of E-cad expression was tabulated by tumorgrade for both the LVI-positive and LVI-negativegroups. For the purposes of statistical analysis, sam-ples that showed weak intensity and/or low frequencyof expression were grouped together. The distribu-tions of intensity (high vs. low), frequency (high vs.low), and expression pattern (peripheral vs. homoge-neous) were compared between the LVI-positive andLVI-negative groups using the chi-square test or, whenthe assumptions of the chi-square test were not met,the Fisher exact test. Analyses were conducted withthe Prophet 5.0 statistics program (NIH Software, Be-thesda, MD), and P � 0.05 was considered significant.

RESULTSAll patients were women ranging in age from 30 to 92years. The expression of E-cad in normal ducts, inva-sive tumor, and intralymphatic tumor emboli was an-alyzed individually in each patient. Normal breast lob-ules adjacent to the tumor displayed a diffuse(reactivity, 4�), high-intensity, membranous stainingpattern (Fig. 1) and served as an internal control.Myoepithelial and stromal cells did not express E-cad.

Tumor cells, when positive, showed membranous ex-pression of E-cad, although significant variations inintensity and distribution of positive cells were noted.

LVI-Positive PatientsThe clinicopathologic data of the 25 LVI-positive pa-tients are shown in Table 1. Six patients had Richard-son and Bloom Grade II tumors, and 19 had Grade IIItumors. Estrogen and progesterone receptors werepositive in 14 of 25 and 9 of 25 tumors, respectively.Six tumors overexpressed c-erb-B2, and 7 were posi-tive for p53. Twenty-three of the 25 patients had me-tastasis to axillary nodes. A membranous pattern ofE-cad expression was seen throughout the tumor (Fig.2). The intensity of E-cad expression in the tumor washigh for all patients. The frequency of tumor cellsexpressing E-cad also was high (2�, 1 patient; 3�, 1patient; 4�, 23 patients). Expression within the lym-phatic emboli had a similar distribution that of themain tumor (Fig. 1). Positivity was 2� for 1 patient, 3�for 1 patient, and 4� for 23 patients, with 24 of 25patients exhibiting high-intensity reactions (Table 2).

LVI-Negative PatientsOf the 25 LVI-negative patients, 1 had Richardson andBloom Grade II disease, and 24 had Grade III disease(Table 1). No patients had nodal metastases; this find-ing confirmed the strong association of lymphatic in-vasion with nodal metastasis. Estrogen and progester-one receptors were positive in 11 and 6 tumors,respectively. Overexpression of c-erb-B2 was noted in5 tumors, and 13 tumors were positive for p53. LVI-negative patients also exhibited a membranous pat-

FIGURE 1. Low-magnification photomicrograph of a high-grade ductal car-

cinoma with prominent lymphatic emboli. Note the uniform E-cad expression in

normal ducts (arrows). The intralymphatic tumor emboli (*) show staining that

is similar in intensity and distribution to that of normal epithelium (arrows).

TABLE 1Clinicopathologic Data for the Patient Populationin the Current Study

LVI-Positive(no. of patients)

LVI-Negative(no. of patients)

Age (yrs) � 50 14 16� 50 11 9

Grade I n nII 6 1III 19 24

ER Positive 14 11Negative 11 14

PR Positive 9 6Negative 16 19

p53 Positive 7 13Negative 18 12

C-erb B2 Positive 6 5Negative 19 20

LVI: prominent lymphatic invasion; ER: estrogen receptor; PR: progesterone receptor.

E-Cadherin and Tumor Emboli/Gupta et al. 2343

tern of E-cad expression; however, expression wasobserved predominantly at sites where there was atumor-stroma interface (Fig. 3). Cells that were closerto the center of a tumor island exhibited weaker in-tensity (or no staining) relative to cells on the periph-ery (Figs. 3, 4). This peripheral pattern was discerniblein 80% of LVI-negative patients. The number of tumorcells expressing E-cad was lower (1�, 2 patients; 2�, 5patients; 3�, 9 patients; 4�, 9 patients) and intensitywas weaker (negative, 1 patient; low, 6 patients; high,18 patients) than in LVI-positive patients (Table 2).

Table 3 shows the intensity, frequency, and pat-tern of E-cad expression in the 50 patients who wereexamined. LVI-positive patients were more likely tohave a high intensity and frequency of expression (P� 0.001). In addition, LVI-positive patients were lesslikely to exhibit a peripheral expression pattern (P� 0.001). In fact, all LVI-positive patients displayed a

homogeneous pattern. These findings retained theirstatistical significance even when the analysis was re-stricted to patients with Grade III disease.

DISCUSSIONCadherins are believed to play a role in the regulationof tissue morphology.25,26 E-cad, an integral part ofadherens junction, maintains the integrity of epithelialcells by facilitating cell-cell adhesion.27 Disruption ofnormal cell-cell adhesion is believed to result in de-hiscence of tumor cells, which is an essential initialstep in the onset of invasion and metastasis.3,25

Although it is a characteristic feature of invasivelobular carcinoma,28 loss of E-cad expression is signif-icantly more variable in IDC. In addition, loss of E-cadexpression in IDC is attributable to epigenetic mech-anisms rather than mutations,16,17 and E-cad-positive

FIGURE 2. High-magnification photomicrograph of a high-grade ductal

carcinoma. Note the diffuse uniform membranous staining (arrows) of the

tumor cells.

TABLE 2Correlation of E-Cadherin Expression with Tumor Grade

LVI-Positive LVI-Negative Total

Frequency ofexpression 1 2 3 4 1 2 3 4

Grade I(no. ofpatients) N N N N N N N N N

Grade II(no. ofpatients) 0 0 0 6 0 0 1 0 7

Grade III(no. ofpatients) 1 0 1 17 2 5 8 9 43

LVI: prominent lymphatic invasion.

FIGURE 3. (A) Low-magnification and (B) high-magnification photomicro-

graphs of a high-grade ductal carcinoma without lymphatic emboli. Note the

peripheral pattern of staining, with prominent E-cad expression at the tumor-

stroma interface (arrows) as compared with expression in areas closer to the

center of the tumor (*).

2344 CANCER May 1, 2003 / Volume 97 / Number 9

nodal metastases can be associated with tumors thatdo not express E-cad.18 These observations lead to thefollowing hypotheses: 1) because lymph node metas-tases can occur in the presence or absence of E-cadexpression in the primary tumor, different mecha-nisms for tumor metastasis to lymph nodes must exist;2) because loss of E-cad expression is common ininvasive lobular carcinomas and high-grade IDC,which are tumors commonly associated with lymphnode metastases, loss of expression is an importantmechanism for the development of nodal metastases;3) the focal E-cad-positive component in an otherwiseE-cad-negative tumor may be the component that islikely to metastasize (clonal theory of metastasis);18 4)expression or re-expression of E-cad in the nodal de-posits could offer a survival advantage to the meta-static tumor deposit; 5) because lymphatic emboli arenot a feature of invasive lobular carcinoma, which ischaracteristically E-cad-negative, E-cad expression

probably is required for the development of intralym-phatic tumor emboli. The current study was designedto test this hypothesis.

In contrast with the results of previous studies, wedetected at least focal E-cad expression in almost allpatients. A major reason could be the use of the TSAsystem, which resulted in the vastly increased sensi-tivity of the assay. One reason for using the TSAmethod was to counteract the loss of antigenicity as-sociated with formalin fixation;14,25,26,28 –31 however, asdemonstrated by the negative controls, which in-cluded not only PBS controls but also isotypic mouseantibody controls, the increased sensitivity was notassociated with nonspecific reactivity. Biotin blockwas performed to exclude nonspecific staining. In ad-dition, as expected, the expression of E-cad was re-stricted to the epithelial cells.

In sharp contrast with the diffuse strong positivityobserved in LVI-positive patients, a marked variationin intensity and frequency of E-cad expression withinthe tumor was seen in the 25 LVI-negative patientswith high-grade IDC (Figs. 3, 4). Tumor cells at theleading edge of the carcinoma showed minimal loss ofexpression, whereas marked reduction or loss of ex-pression was noted in the center of the tumor (Fig. 3).This pattern was observed in 22 of 25 patients (88%).We were unable to find descriptions of this type ofstaining pattern in the existing literature. Nonetheless,the presence of this feature in approximately 90% ofpatients with high-grade IDC and no LVI or lymphnode involvement suggests that there is some etio-pathologic relation; however, the nature of this rela-tion is difficult to predict.

Reduced expression of E-cad in IDC was found tobe correlated with higher histologic tumor grade insome14,25,32 but not all studies.1,32,33 Reduced expres-sion of E-cad also was found to be correlated withlocally aggressive disease in the form of nodal involve-ment in several studies.14,19,20 In the current study, wefound that the degree of E-cad expression (both inten-sity and distribution) were correlated with the pres-ence of LVI. This correlation held true even when theanalysis was restricted to Grade III cases—i.e., statis-tical significance was maintained (P � 0.001) whenLVI-positive patients with Grade III IDC were com-pared with LVI-negative patients with Grade III IDC.High-grade tumors that exhibited reduced and pe-ripheral patterns of E-cad expression did not showobservable emboli. These results suggest that the for-mation of intravascular tumor emboli is related toE-cad expression.

The finding of diffuse strong E-cad expression inLVI-positive cases is not surprising. In their animalmodel of inflammatory breast carcinoma, Alpaugh et

FIGURE 4. Low-magnification photomicrograph of a high-grade ductal car-

cinoma without lymphatic emboli. Note the low-intensity (*) and high-intensity

(arrows) regions of E-cad expression in the tumor cells.

TABLE 3E-Cadherin Expression in LVI-Positive and LVI-Negative Patients

VariableLVI-Positive(no. of patients)

LVI-Negative(no. of patients) P

IntensityHigh 24 18 0.001Low 1 7

FrequencyHigh 24 18 0.001Low 1 7

Peripheral 0 20 � 0.001Homogeneous 25 5

LVI: prominent lymphatic invasion.

E-Cadherin and Tumor Emboli/Gupta et al. 2345

al.34 demonstrated similar expression in intralym-phatic emboli. Tomlinson et al.16 also demonstratedthat administration of anti-E-cad antibodies pre-vented the development of intravascular emboli andthe inflammatory breast phenotype in mice. More re-cently, similar high levels of E-cad expression in intra-lymphatic tumors have been demonstrated in humaninflammatory carcinoma.16,35

Because metastases can occur irrespective of E-cad expression levels, alterations in E-cad expressionlevels in tumor cells could play a role in invasion bymore than one mechanism. Reduction or loss of E-cadexpression would permit easier migration of individ-ual tumor cells into the vasculature. In contrast, per-sistence of E-cad in cells that invade the vasculaturecould lead to the development of intravascular nestsof tumor cells or intravascular emboli. The formermechanism may be operative in E-cad-negative IDCand in lobular carcinomas, whereas the latter mayoccur in E-cad-positive IDC.

REFERENCES1. Bukholm IK, Nesland JM, Karesen R, Jacobsen U, Borresen-

Dale AL. E-cadherin and alpha-, beta-, and gamma-cateninprotein expression in relation to metastasis in human breastcarcinoma. J Pathol. 1998;185:262–266.

2. Overduin M, Harvey TS, Bagby S, et al. Solution structure ofthe epithelial cadherin domain responsible for selective celladhesion [see comments]. Science. 1995;267:386 –389.

3. Takeichi M. Cadherin cell adhesion receptors as a morpho-genetic regulator. Science. 1991;251:1451–1455.

4. Shapiro L, Fannon AM, Kwong PD, et al. Structural basis ofcell-cell adhesion by cadherins [see comments]. Nature.1995;374:327–337.

5. Eidelman S, Damsky C, Wheelock M, Damjanov I. Expres-sion of the cell-cell adhesion glycoprotein Cell-CAM 120/80in normal human tissues and tumors. Am J Pathol. 1989;135:101–110.

6. Nagar B, Overduin M, Ikura M, Rini JM. Structural basis ofcalcium-induced E-cadherin rigidification and dimeriza-tion. Nature. 1996;380:360 –364.

7. Vos CB, Cleton-Jansen AM, Berx G, et al. E-cadherin inacti-vation in lobular carcinoma in situ of the breast: an earlyevent in tumorigenesis. Br J Cancer. 1997;76:1131–1133.

8. Velminckx K, Vakaet Jr L, Mareel M, Fiers W, van Roy F.Genetic manipulation of E-cadherin expression by epithelialtumor cells reveals an invasion suppressor role. Cell. 1991:107–119.

9. Tan DS, Potts HW, Leong AC, et al. The biological andprognostic significance of cell polarity and E-cadherin ingrade I infiltrating ductal carcinoma of the breast. J Pathol.1999;189:20 –27.

10. Frixen U, Behrens J, Sachs M, et al. E-cadherin-mediatedcell-cell adhesion prevents invasiveness of human carci-noma cells. J Cell Biol. 1991;113:173–185.

11. Sommers CL. The role of cadherin-mediated adhesion inbreast cancer. J Mammary Gland Biol Neoplasia. 1996;1:219 –229.

12. Shimoyama Y, Hirohashi S. Expression of E- and P-cadherinin gastric carcinomas. Cancer Res. 1991;51:2185–2192.

13. Umbas R, Isaacs WB, Bringuier PP, et al. Decreased E-cadherin expression is associated with poor prognosis inpatients with prostate cancer. Cancer Res. 1994;54:3929 –3933.

14. Siitonen SM, Kononen JT, Helin HJ, Rantala IS, Holli KA,Isola JJ. Reduced E-cadherin expression is associated withinvasiveness and unfavorable prognosis in breast cancer.Am J Clin Pathol. 1996;105:394 – 402.

15. Lehr HA, Folpe A, Yaziji H, Kommoss F, Gown AM. Cytoker-atin 8 immunostaining pattern and E-cadherin expressiondistinguish lobular from ductal breast carcinoma. Am J ClinPathol. 2000;114:190 –196.

16. Tomlinson JS, Alpaugh ML, Barsky SH. An intact overex-pressed E-cadherin/alpha,beta-catenin axis characterizesthe lymphovascular emboli of inflammatory breast carci-noma. Cancer Res. 2001;61:5231–5241.

17. Graff JR, Gabrielson E, Fujii H, Baylin SB, Herman JG. Meth-ylation patterns of the E-cadherin 5� CpG island are unsta-ble and reflect the dynamic, heterogeneous loss of E-cad-herin expression during metastatic progression. J Biol Chem.2000;275:2727–2732.

18. Bukholm IK, Nesland JM, Borresen-Dale AL. Re-expressionof E-cadherin, alpha-catenin and beta-catenin, but not ofgamma-catenin, in metastatic tissue from breast cancer pa-tients [see comments]. J Pathol. 2000;190:15–19.

19. Hunt NC, Douglas-Jones AG, Jasani B, Morgan JM, PignatelliM. Loss of E-cadherin expression associated with lymphnode metastases in small breast carcinomas. Virchows Arch.1997;430:285–289.

20. Jones JL, Royall JE, Walker RA. E-cadherin relates to EGFRexpression and lymph node metastasis in primary breastcarcinoma. Br J Cancer. 1996;74:1237–1241.

21. Charpin C, Garcia S, Bonnier P, et al. Reduced E-cadherinimmunohistochemical expression in node-negative breastcarcinomas correlates with 10-year survival. Am J ClinPathol. 1998;109:431– 438.

22. Asgeirsson KS, Jonasson JG, Tryggvadottir L, et al. Alteredexpression of E-cadherin in breast cancer. Patterns, mech-anisms and clinical significance. Eur J Cancer. 2000;36:1098 –1106.

23. Guriec N, Marcellin L, Gairard B, et al. E-cadherin mRNAexpression in breast carcinomas correlates with overall anddisease-free survival. Invasion Metastasis. 1996;16:19 –26.

24. Recommendations for the reporting of breast carcinoma.Association of Directors of Anatomic and Surgical Pathol-ogy. Mod Pathol. 1996;9:77– 81.

25. Charpin C, Garcia S, Bouvier C, et al. E-cadherin quantita-tive immunocytochemical assays in breast carcinomas.J Pathol. 1997;181:294 –300.

26. Oka H, Shiozaki H, Kobayashi K, et al. Expression of E-cadherin cell adhesion molecules in human breast cancertissues and its relationship to metastasis. Cancer Res. 1993;53:1696 –1701.

27. Nelson WJ, Hammerton RW, Wang AZ, Shore EM. Involve-ment of the membrane-cytoskeleton in development of ep-ithelial cell polarity. Semin Cell Biol. 1990;1:359 –371.

28. Moll R, Mitze M, Frixen UH, Birchmeier W. Differentialloss of E-cadherin expression in infiltrating ductal andlobular breast carcinomas. Am J Pathol. 1993;143:1731–1742.

29. Gamallo C, Palacios J, Suarez A, et al. Correlation of E-cadherin expression with differentiation grade and histolog-ical type in breast carcinoma. Am J Pathol. 1993;142:987–993.

2346 CANCER May 1, 2003 / Volume 97 / Number 9

30. Rasbridge SA, Gillett CE, Sampson SA, Walsh FS, Millis RR.Epithelial (E-) and placental (P-) cadherin cell adhesionmolecule expression in breast carcinoma. J Pathol. 1993;169:245–250.

31. Lipponen P, Saarelainen E, Ji H, Aaltomaa S, Syrjanen K.Expression of E-cadherin (E-CD) as related to other prog-nostic factors and survival in breast cancer. J Pathol. 1994;174:101–109.

32. Gonzalez MA, Pinder SE, Wencyk PM, et al. An immunohis-tochemical examination of the expression of E-cadherin,alpha- and beta/gamma-catenins, and alpha2- and beta1-

integrins in invasive breast cancer. J Pathol. 1999;187:523–529.

33. Shiozaki H, Tahara H, Oka H, et al. Expression of immuno-reactive E-cadherin adhesion molecules in human cancers.Am J Pathol. 1991;139:17–23.

34. Alpaugh ML, Tomlinson JS, Shao ZM, Barsky SH. A novelhuman xenograft model of inflammatory breast cancer.Cancer Res. 1999;59:5079 –5084.

35. Kleer CG, van Golen KL, Braun T, Merajver SD. PersistentE-cadherin expression in inflammatory breast cancer. ModPathol. 2001;14:458 – 464.

E-Cadherin and Tumor Emboli/Gupta et al. 2347