MOLECULAR SUBTYPES IN BREAST CANCER

UNIVERSITY OF MEDICINE AND PHARMACY GRIGORE T. POPA IAŞI FACULTY OF MEDICINE MOLECULAR SUBTYPES IN BREAST CANCER SUMMARY OF THESIS SCIENTIFIC LEADER IRINA-DRAGA CĂRUNTU, MD, PhD, Professor PhD Student CARMEN IONESCU (POPESCU) Investeşte în oameni! Proiect cofinanţat din Fondul Social European prin Programul Operaţional Sectorial Dezvoltarea Resurselor Umane 2007 2013 Axa prioritară Educatia si formarea profesională în sprijinul cresterii economice si dezvoltării societătii bazate pe cunoastere Domeniul major de intervenţie 1.5 Programe doctorale si post-doctorale în sprijinul cercetării Titlul proiectului: Burse doctorale pentru cresterea competitivitatii in domeniul medical si farmaceutic Numărul de identificare al contractului: POSDRU/88/1.5/S/58965 Beneficiar : Universitatea de Mdicina si Farmacie Gr. T. Popa Iasi Partener : Universitatea de Medicina si Farmacie Iuliu Hatieganu Cluj Napoca 2013

TABLE OF CONTENTS OF THE THESIS TABLE OF CONTENTS i ABBREVIATIONS iii ACKNOWLEDGEMENTS iv STATE OF ART CHAPTER 1 MAMMARY GLAND IN NORMALITY STATUS CELLULARITY, 1 HISTOARCHITECTONY, HISTOPHYSIOLOGY 1.1. ELEMENTS OF HISTOLOGY 1 1.2. TRANSFORMATIONS DURING MORPHOGENESIS AND PHYSIOLOGICAL VARIATIONS 2 1.3. MAMMARY STEM CELS 6 CHAPTER 2 THE MORPHOLOGICAL PROFILE OF BREAST CANCER 8 2.1. IN SITU DUCTAL CARCINOMA 9 2.1.1. Low grade in situ ductal carcinoma 9 2.1.2. Intermediary grade in situ ductal carcinoma 11 2.1.3. High grade in situ ductal carcinoma 11 2.2. INVASIV DUCTAL CARCINOMA 12 2.3. IN SITU LOBULAR CARCINOMA 14 2.4. INVASIV LOBULAR CARCINOMA 16 CHAPTER 3 MOLECULAR CLASSIFICATION OF BREAST CANCER 3.1. THE GENETIC SIGNATURE OF BREAST CANCER THE ROLE OF STANFORD GROUP 3.2. GENETIC TESTING FOR BREAST CANCER: CONCEPTION, CHARACTERISTICS 20 3.2.1. MammaPrint The 70-Gene Assay 20 3.2.2. The 76-Gene Assay 20 3.2.3. The HOXB13:IL17BR Assay 20 3.2.4. Oncotype DX The 21-Gene RT-PCR Assay 21 3.2.5. Other tests 21 3.2.6. Genetic tests between the benefits and the limitations of the results 21 3.3. MOLECULAR SUBTYPES 22 3.3.1. Luminal A subtype 23 3.3.2. Luminal B subtype 23 3.3.3. Luminal A subtype versus luminal B subtype 24 3.3.4. Luminal C subtype 24 3.3.5. HER2 positive subtype 24 3.3.6. Basal-like subtype 25 3.3.7. Normal-like subtype 26 3.3.8. Other molecular subtypes 27 CHAPTER 4 THE IMMUNOHISTOCHIMICAL SUROGATE MARKERS ALTERNATIVE SOLUTION FOR GENE ANALYSIS. DIAGNOSTIC AND PROGNOSTIC VALUE 4.1. ESTROGENIC RECEPTORS 28 4.2. PROGESTERONIC RECEPTOR 31 4.3. HER2/NEU 32 4.4. EGFR 34 4.5. CYTOKERATINS 34 4.6. OTHER MOLECULAR MARKERS 34 PERSONAL CONTRIBUTIONS CHAPTER 5 MOTIVATION AND OBJECTIVES OF THE THESIS 36 18 19 28 i

CHAPTER 6 TYPES AND SUBTYPES OF MOLECULAR DIAGNOSIS IDENTIFICATION, 38 MEANING 6.1. INTRODUCTION 38 6.2. MATERIAL AND METHOD 38 6.3. RESULTS 41 6.3.1. Clinico-pathological features 41 6.3.2. ER evaluation 42 6.3.3. PR evaluation 46 6.3.4. Her2/neu evaluation 50 6.3.5. CK5/6 evaluation 53 6.3.6. EGFR evaluation 56 6.3.7. Classification according to diagnostic criteria of molecular classification 58 6.3.8. Correlations between the molecular profile versus clinico-pathological features with prognostic value 63 6.4. DISCUTIONS 66 6.4.1. ER positive molecular subtype: luminal A subtype, luminal B subtype 66 6.4.2. ER negative molecular subtype: HER2 subtype, basal-like subtype, normal like subtype 67 CHAPTER 7 COMPLEMENTARY MARKERS IN ASSESSING MOLECULAR SUBTYPES 70 7.1. INTRODUCTION 70 7.2. MATERIAL AND METHOD 70 7.3. ANDROGENIC RECEPTOR PROFILE IN MOLECULAR SUBTYPES OF BREAST 71 CANCER 7.3.1. Argument 71 7.3.2. Features of quantification method 73 7.3.3. Results 73 7.3.4. Discussions 77 7.4. CLDN3 PROFILE IN MOLECULAR SUBTYPES OF BREAST CANCER 80 7.4.1. Argument 80 7.4.2. Features of quantification method 81 7.4.3 Results 81 7.4.4. Discutions 87 7.5. E-CADHERINEI EXPRESSION IN MOLECULAR SUBTYPES OF BREAST CANCER 90 7.5.1. Argument 90 7.5.2. Features of quantification method 91 7.5.3. Results 91 7.5.4. Discussions 97 7.6. p53 AND BCL-2 PROFILE IN MOLECULAR SUBTYPES OF BREAST CANCER 98 7.6.1. Argument 98 7.6.2. Features of quantification method 100 7.6.3. Results 100 7.6.4. Discussions 108 7.7. LYMPHANGIOGENESIS PROFILE IN MOLECULAR SUBTYPES OF BREAST 112 CANCER 7.7.1. Argument 112 7.7.2. Features of quantification method 113 7.7.3. Results 114 7.7.4. Discussions 127 CHAPTER 8 CONCLUSIONS 130 REFERENCES 132 ANNEX 1. PUBLISHED PAPERS DURING THE DOCTORAL TRAINING 149 ii

ABBREVIATIONS AIB-3 eng. Amplified in Breast cancer-3 protein AR androgenic receptor BCL-2 eng. B-Cell Leukemia/Lymphoma-2 BRCA1/2 eng. Breast Cancer Type I/Type II CD49f eng. Integrin Subunit α6 CDI invasive ductal carcinoma CDIS in situ ductal carcinoma CGH eng. Comparative Genomic Hybridization CIS in situ carcinoma CK cytokeratin CLDN eng. Claudin CLI invasive lobular carcinoma CLIS in situ lobular carcinoma CM breast cancer CNN1 eng. Calponina 1 DSF eng. Disease-Free Survival DDFS eng. Distant Disease-Free Survival EGFR eng. Epidermal Growth Factor Receptor ELISA eng. Enzyme-Linked Immunosorbent Assay EMA eng. Epithelial Membrane Antigen EpCAM eng. Epithelial Cell Adhesion Molecule ER estrogenic receptor FDA eng. Food and Drug Administration FISH eng. Fluorescence In Situ Hybridization FOXA1 eng. Forkhead box protein A1 GATA3 eng. Trans-acting T-cell-specific Transcription Factor GATA-3 GPER eng. G Protein-coupled Estrogen Receptor 1 GRB7 eng. Growth Factor Receptor-Bound Protein 7 hcs eng. Human Chorionic Somatomammotropin Hormone HER2 eng. Human Epidermal Growth Factor Receptor 2 IHC immunohistochemistry LOH eng. Loss of Heterozigozity NOS eng. Not Otherwise Specified NOTCH eng. Neurogenic Locus Notch Protein Homolog OS eng. Overall Survival PCR eng. Polymerase Chain Reaction RT PCR eng. Real Time Polymerase Chain Reaction PELP 1 eng. Leucine-rich Protein 1 PI3CA eng. PI3K Catalytic Subunit-α PI3K eng. Phosphatidylinositol 3-Kinases PI3K/AKT/mTOR eng. Phosphatidylinositide 3-Kinases PR progesteronic receptor RER rough endoplasmic reticulum SERM eng. Selective Estrogen Receptor Modulators STAT1 eng. Signal Transducers and Activators of Transcription UDLT terminal ducto lobular unit TGF-α/β eng. Transforming Growth Factor α/β upa eng. Urokinase Plasminogen Activator VEGF eng. Vascular Endothelial Growth Factor WHO eng. World Health Organization Doctoral thesis is illustrated in 85 figures and 57 tables. All research accomplished during the doctoral studies could be completed and due to my status, as fellow in the project Doctoral fellowships for competitive PhD students in the European research area, POSDRU/88/1.5/S/58965. iii

Special thanks for the decisive intervention chosen in my professional founding to, Mrs. Prof. M.D. PhD. Irina-Draga Căruntu University of Medicine and Pharmacy Grigore T. Popa Iaşi iv

CHAPTER 5 MOTIVATION AND OBJECTIVE OF THE DOCTORAL STUDY According to WHO, breast cancer classification reveals over 20 particular histological types, some of which are characterized by accurate clinical and prognostic highlights. Over the last decade, gene expression analysis has been extensively submitted in breast cancer research, aiming to elucidate the molecular bases underlying biological features (histological grade, metastasis ability), also finding the specific patterns associated with prognostic and therapy response. Moreover delineating the 5 subtypes (luminal A, luminal B, HER2, basal-like, normal like) resembled to gene expression pattern confirmed at molecular level the subsisting concept of breast cancer clinical and morphological heterogeneity. Over the last decade breast cancer research lead toward achieving the gene expression pattern as the gold standard for rating this pathology, translated into an ideal grading system, with beneficial impact on diagnostic and therapeutic pursue. Until now, however, the results are incompletely validated; there are many differences which can also be explained by the use of different intrinsic genes, limited methodology reproducibility or limited prevalence of special histological types. Hence, research assignment reveals new quest path in order to survey this issue as an accurate molecular hallmark providing a new angle for individualized therapeutic approach. However, at present, the taxonomy associated with molecular classification must be considered as a pattern for upgrading the assignment rank, and not as assured operational tool. Since the diagnostic grading according to molecular classification has just been aligned and correlations with 5 year survival lack, this study aims to hollow the molecular classes of breast cancer. The enlightenment innate of this scientific approach is presented, as a framework, by the lack of consent in the interpretation of each subtype, and as form, by methodology contrast related to the immunohistochemical approach which causes inconsistent reports concerning the five molecular subtypes. In this context, the PhD project aimed to evaluate the surrogate markers profile in breast cancer establish the diagnostic subtypes according to molecular classification investigate in the molecular subtypes of breast cancer androgen receptor profile CLDN4 and E-cadherin cell adhesion molecules profile pro- and antiapoptotic expression, displayed by p53 and bcl-2 respectively tumor lymphangiogenesis impact, performed by VEGF-C, VEGFR3 and D2-40. CHAPTER 6 TYPES AND SUBTYPES OF MOLECULAR DIAGNOSIS - IDENTIFICATION, SIGNIFICATION 6.1. INTRODUCTION In 2000, the study of gene expression profile in breast cancer has led to the introduction of molecular classification, relevant in determining prognostic and treatment options (Perou et al., 2000). The classification estimated as the gold standard assembles the five classes of breast cancer: 1) ER positive, luminal type A, 2) ER positive, luminal type B, 3) type HER2, 4) basal-like type, 5) unclassifiable - normal-like. Microarray profile includes only key genes (MammaPrint, Oncotype DX, PAM50), allowing a refined investigation of the gene profile, but the high associated costs make impossible their routine use. Consequently, research lines are focused on the identification and validation of alternative IHC markers, known as surrogate markers, which would yield this classification. Despite the adjusting efforts on this issue, conflicting reports on the five molecular classes lead to lack of consensus in the interpretation of each molecular subtype a fact explained by the variability between laboratory work methods, including by differences in establishing biomarkers. In this context, the doctoral study focused on: - investigating surrogate markers profile in breast cancer - identifying diagnostic subtypes according to molecular classification categories. 1

6.2. MATERIAL AND METHOD The study included 42 malignant breast tumors diagnosed between January 1 st 2006 and December 31 st, 2011 at the University Hospital of Obstetrics and Gynecology Cuza Vodă of Iași, respectively Elena Doamna Obstetrics and Gynecology Hospital of Iași. The study group was approved by the Ethics Committee of the University of Medicine and Pharmacy Grigore T. Popa of Iași based on the informed consent of patients for using the biological product and based on the doctoral study protocol, conducted at the completion of the first year of doctoral studies. The study group was subject to IHC analysis, using a panel of antibodies (ER, PR, Her2/neu, CK5/6, EGFR) for diagnostic categorization according to molecular classification. IHC investigation was conducted at the Angiogenesis Research Centre - University of Medicine and Pharmacy Victor Babeș of Timișoara and as part of the Interdisciplinary Platform for Molecular Medicine - University of Medicine and Pharmacy Grigore T. Popa of Iași. Immunohistochemical reaction for the markers used was semi quantitatively assessed, through appropriate score systems, based on the percentage of positive cells and enhancement of staining intensity reaction. The database was designed in order to achieve a dynamic assessment of the clinical and morphological profile of each patient. Statistical data processing was performed using the SPSS 13.0 software, which allowed the application of the Pearson test, the most widely used type of χ2 test of significance, based on associating columns and lines of a table with two entries, cross frequencies concerning discrete or disparaged variables. Yates correction (also known as the continuity corrected chi-square) was applied due to the relatively small dimension of the whole considered, in the situation where there were cells with less than 5 items. The statistical significance was interpreted in a standard manner for p <0.05. 6.3. RESULTS 6.3.1. Clinical and pathological characteristics Clinical and pathological characteristics of patients included in the studied group are summarized in Table 6.4. Table 6.4. Synopsis for the clinical and pathological characteristics of the studied lot CLINICO-PATHOLOGICAL CHARACTERISTICS Cases (n =42) N (42) % Varstă (ani) 55 14 33,33 >55 28 66,66 HISTOLOGICAL TYPE Invasiv ductal carcinoma 26 61,90 Invasiv lobular carcinoma 7 16,66 Other morphological subtypes 9 21,42% Medullary carcinoma 1 2,38 Tubular carcinoma 1 2,38 Mucinous carcinoma 1 2,38 Apocrine carcinoma 1 2,38 Mixte carcinoma 5 11,90 T STADE I 9 21,42 II 23 54,76 III 7 16,66 IV 3 7,14 N STADE Nx 1 9,52 N0 13 33,33 N1 16 40,47 N2+N3 12 16,66 M STADE Mx 39 92,85 M0 2 4,76 M1 1 2,38 DEGREE OF DIFFERENTIATION G1 (well differentiated) 12 28,57 G2 (moderately differentiated) 17 40,47 G3 (poorly differentiated) 13 30,95 2

Strictly relating to the pathological pattern, the 5 cases ranked as mixed carcinoma included the association between CDI and CDL 2 cases, cribriform carcinoma 2 cases, mucinous carcinoma 1 case. In order to determine the degree of differentiation, we used the criteria of Scarff-Bloom- Richardson grading (Elston, Ellis, 1991), in relation to the formation of tubules, nuclear grade, and mitotic status. 6.3.2. ER assessment According to the Allred scoring, ER semi quantitative nuclear expression assessment allowed the identification of 26 ER positive cases and 16 ER negative cases, respectively. The 26 ER positive cases pointed out score values between 3 and 8 therefore: score 3 2 cases, score 5 3 cases, score 6 3 cases, score 7 10 cases, score 8 8 cases. This assessment indicates that two thirds cases (18 of 26) had extremely strong ER staining, translated by a percentage of positive cells of over 75%, predominantly in the range 90-100%, with high intensity of the IHC reaction. According to complete absence of ER expression, 15 of the 16 ER negative cases were assessed as 0 score value; only one case was estimated as score 2, displaying 1% ER positive tumor cells percentage for weak staining. 6.3.3. PR assessment Semi quantitative assessment of PR nuclear expression, according to Allred scoring, allowed the identification of 27 PR positive cases and 15 PR negative cases, respectively. The 27 cases unveiled positive PR score values between 3 and 8 as follows: score 3 2 cases, score 4 3 cases, score 5 3 cases, score 6 2 cases, score 7 9 cases, score 8 8 cases. The score analysis reveals that approximately two thirds of all cases (17 of 27) were featured by a high PR expression reflected by more than 75% positive tumor cell and strong immunoreaction intensity. From all the 15 PR negative cases, 14 had 0 score value, indicating total absence of PR expression, meanwhile 1 case had 2 score value, reflecting a 1% weakly positive PR tumor cells. 6.3.4. Her2/neu assessment Semi quantitative assessment of Her2/neu expression, focalized at the membrane level, according to the FDA score, led to the assignment of 17 cases as HER2 positive and 25 cases as Her2 negative, respectively. All 17 Her2 positive cases had 3 score value assigned based on the presence of the uniform membrane reaction in more than 30% of tumor cells. Regarding the 25 Her2 negative cases, since 15 cases were assigned score 0, as the immunoreactivity was completely absent, the remaining 10 cases had the score 1 assigned on the basis of weak intensity immunoreactivity identified in less than 30% of tumor cells. 6.3.5. CK5/6 assessment Semi quantitative assessment of CK5/6 expression, localized at the membrane level, allowed the determination of the CK5/6 positive pattern for 28 cases and CK5/6 negative for 14 case, respectively. In the 28 CK5/6 positive cases, the positive cells percentage ranged between 10% and 100%, the reaction intensity being assessed in 3 standard degrees: weak, moderate and strong. 6.3.6. EGFR assessment Semi quantitative evaluation of EGFR expression, situated in the membrane, has led to the classification of 8 cases as EGFR positive, while the remaining 34 were considered negative. EGFR positive cases generally presented a reduced proportion of positive tumor cells (10% 1 case, 20% 2 cases, 40% 1 case, 50% 2 cases, 60% 1 case, 100% 1 case), the reaction intensity was weak and moderate, except for two cases in which it was assessed as strong (in actual situation, the cases with 60% and 100% EGFR positive tumor cells, respectively). 6.3.7. The disposal according to diagnostic categories of molecular classification Based on the semi quantitative assessment for each individual marker, and in accordance with the molecular classification criteria, the diagnostic framing was along these lines: - ER positive molecular type - A luminal subtype: 15 cases (35.71%); - B luminal subtype: 13 cases (30.95%); - ER negative molecular type 3

- HER2 subtype: 4 cases (9.52%); - basal like subtype: 8 cases (19.04); - unclassifiable subtype (normal like): 2 cases (4.76%). 6.3.8. Clinical and pathological features with prognostic significance similarities The statistical analysis correlation (by applying the Chi-square test) in order to identify the presence or absence of analogies between the diagnosed molecular subtypes and standard clinico morphological parameters: tumoral staging, differentiation grade, morphological subtype. We noted the absence of any statistically significant differences. 6.4. DISCUSSIONS BC types and subtypes introduced by the molecular classification are different from the standard defined entities (Elston, Ellis, 1991, Ellis et al., 1992) based on the features of the architectural, pathological and immunohistochemical pattern. 6.4.1. ER positive molecular type: Luminal A subtype, Luminal B subtype In this class of diagnosis, our study allowed the separation of the two different subtypes, called luminal A subtype (8 cases) and luminal B subtype (1 case). The markers used did not allow identifying the recently introduced third subtype luminal C (Hergueta Redondo et al., 2008). Significance of diagnostic classification in luminal A subtype versus luminal B subtype By reference to the histological classification of BC, luminal subtype (A and B) has a counterpart in most well-differentiated infiltrating ductal carcinoma (grade 1), and sometimes also moderately differentiated (grade 2) and undifferentiated (grade 3) forms expressing ER and/or PR, as well as in infiltrating lobular carcinoma, tubular carcinoma, cribriform and colloid carcinoma. In our study, immunohistochemical profile of the luminal A subtype - ER+/PR+/Her2 - was identified in 15 cases. These cases had as a morphological correspondence, in 2 cases, for the invasive lobular carcinoma type, in 11 cases, for the invasive ductal carcinoma type, and in 2 cases other histological types. Analyzing the profile of histopathological lesions associated to the pattern of invasive carcinoma, 1 case presented comedocarcinoma areas, 2 cases presented cribriform and micropapillary carcinoma areas and another case presented micropapillary carcinoma areas. The variety of histopathological elements established in the luminal A subtype assay is an additional argument for the remark according to which - as invasive ductal carcinomas included in luminal A subtype are extremely different in terms of tumor grade (from 1 to 3), hormone receptor level of expression and proliferation rate - it is considered that this molecular subtype should be subclassified in the future (Moinfar, 2008). In our study, the immunohistochemical profile of the luminal B subtype, ER+/PR+/HER2+ was revealed in 13 cases. Histopathologically, 8 cases were classified as invasive ductal type, 3 cases lobular invasive carcinoma type and 2 cases mixt type, based on the lobular extension areas of the intraductal solid or apocrine carcinoma, with areas of ductal carcinoma in situ type, as comedo, apocrine and cribriform. Classification under the subliminal B diagnosis subtype is extremely important because this subtype has a high histological grade, high proliferation rate and significantly more reserved prognostic than luminal A type tumors (Sorlie et al., 2001, Sorlie et al., 2003, Perou et al., 2000, Parker et al., 2009, Weigelt et al., 2010, Correa Geyer et al., 2009). 6.4.2. ER negative molecular type: HER2 subtype, basal-like subtype, unclassifiable subtype Significance of diagnostic classification beneath HER2 subtype HER2 positive subtype immunohistochemical profile is RE-/RP-/Her2+. In our study, this subtype was identified in two cases whose histopathological correspondent was, in one case, an invasive ductal carcinoma, where the morphological background allowed the affiliation with some ductal carcinoma in situ cribriform-type areas, while in the other case, invasive apocrine carcinoma with extensive fields of intraductal carcinoma micropapillary apocrine type or comedo-like; the characteristic aspects for the high grade are supported by nuclear hyperchromasia and pleomorphism, beyond the presence of atypical mitosis. Our data are consistent with the literature regarding the BC HER2+ histological classification which agnate predominantly with invasive ductal carcinoma moderately differentiated (grade 2) and undifferentiated (Grade 3) (Vanden Bempt et al., 2007), also with apocrine breast carcinomas (Varga et 4

al., 2004, Farmer et al., 2005). The alikeness between HER2 molecular subtype to the apocrine histological type described in our study supports the idea of existence of other possible molecular subtypes whose pattern are partially converging over the existing ones, and partly distinctive. A separate mention should be made for apocrine molecular tumor group whose immunohistochemical pattern is ER-/AR+/Her2+. Based on the potential alikeness to HER2 subtype, presence of AR genes, pathogenic range in which AR pathway signaling activation occurs and apocrine histological features, the studies consecrated to this group urge its individualization as a distinct entity within molecular classification (Farmer et al., 2005, Doane et al., 2006). Significance of diagnostic classification beneath basal-like subtype Our study described 8 cases as affiliated to the basal-like subtype, according to the specific immunohistochemical pattern, ER-/AR-/Her2-. We must emphasize that for this subtype, the IHC delineation is still demanding - given the fact that the generic term of triple negative refers exclusively to ER, PR and Her2 status (ER-/AR-/Her2-) and not underlying other markers (Reis-Filho and Tutt, 2008). The unveiling of EGFR and CK5/6 molecular markers is justified by the fact that the basal-like subtype proposed (Nielsen et al., 2004) is defined by the sequence of molecular markers ER-/Her2-/EGFR + and / or CK5/6+, common to 15-25% of BC. With definite quotation to histopathological resemblance, two of the seven cases were diagnosed as invasive lobular carcinoma, 4 cases as invasive ductal carcinoma, and one case as mixed invasive ductal and lobular carcinoma. Overall, all cases were characterized by high proliferation rate and elevated histological grade. Our data correspond only partially with reports in the literature, according to which, in relation to histological classification, basal-like BC includes undifferentiated invasive ductal carcinomas (16-37% of the total number), with grade 3 and significant morphological heterogeneity (tubular solid / atypical medullary type, or having an important central acellular zone), typical medullary carcinomas, tumors with medullary features, metaplastic carcinomas such as carcinosarcomas (Reis-Filho and Tutt, 2008, Turner et al., 2007 Livasy et al., 2006 Reis -Filho et al., 2006, Rodríguez-Pinilla et al., 2007 Sarrió et al., 2008, Sasaki et al., 2009). In our study group, no specific issues characteristic for medullary carcinoma, medullary-like or carcinosarcoma were identified. The single accurate aspect was the presence of osteoclastic areas in one cases of invasive ductal carcinoma. Diagnostic classification significance under the unclassifiable subtype The normal-like unclassifiable group, initially defined by the expression similar to normal breast tissue and benign tumors (Perou et al., 2000, Sorlie et al., 2001), is still poorly characterized (McCafferty et al., 2009, Correa Geyer et al., 2009, Pusztai et al., 2006), and some studies even dispute its existence (Gruvberger et al., 2001, Sorlie et al., 2001, Hergueta-Redondo et al., 2008). It is likely that this subtype is in fact due to a sampling artefact, by sample contamination with a significant amount of normal tissue (Peppercorn et al., 2008, Parker et al., 2009). In our case study, 2 cases based on immunohistochemical profile of ER negative/pr negative/her2 positive have been identified as belonging to this subtype. The histopathological correspondent of the two cases was the diagnosis of tubular carcinoma and invasive ductal carcinoma respectively. Based on the literature available so far, we believe that the affiliation to this subtype can be a starting point for the breast cancer molecular classification development. CHAPTER 7 COMPLEMENTARY MARKERS IN MOLECULAR SUBTYPES EVALUATION 7.1. INTRODUCTION Molecular classification of BC resulted in major changes in the approach of breast cancer pathology, from diagnosis to treatment. In the literature, a real avalanche of studies argues for revealing types and subtypes of tumor molecular markers by using surrogate markers. However, despite the numerous existing data, molecular classification is still a tool where there is room for improvement. This is based, on the one hand, on the heterogeneity of reports and publications mainstream, and on the other hand, on the constant concerns of research groups focused on BC to refine diagnostic categories by introducing additional molecular markers which could differentiate 5

individual biological features, with direct impact on a personalized therapy. In this context, based on the results obtained while investigating surrogate markers in breast cancer and profiling diagnostic subtypes according to the categories of molecular classification, the doctoral research aimed to investigate a panel of complementary markers to analyze the value of their association with the five operational molecular subtypes. Specifically, research directions were focused on: androgen receptor profile; CLDN4 and E cadherin molecular cell adhesion profile pro and anti-apoptotic activity profile, reflected by p53 and bcl 2 respectively - tumor lymphangiogenesis profile, translated by VEGF C, VEGFR3 and D2 40. 7.2. MATERIAL AND METHOD The study group presented in Chapter 6 Section 6.2 Material and method was subject to IHC test using a broad panel of antibodies, whose characteristics are shown in Table 7.1. The observance of the ethical research principles - as mentioned above in Chapter 6 Section 6.2. Material and method, was regulated by the informed consent of patients for the use of the biological product and the doctoral research protocol, conducted at the completion of the first year of doctoral studies. The working methodology followed the steps of the working protocol as outlined in Chapter 6 Section 6.2. Material and method and the IHC test was conducted at the Angiogenesis Research Centre University of Medicine and Pharmacy Victor Babeș of Timisoara and within the Interdisciplinary Platform for Molecular Medicine University of Medicine and Pharmacy Grigore T. Popa of Iași, respectively. The database built to achieve a dynamic assessment of each patient s clinical and morphological profile (Chapter 6 Section 6.2. Material and Method) was completed with the following information: diagnosis subtype, according to molecular classification based on IHC profile corresponding ER, PR, Her2/neu, CK5 / 6, EGFR; IHC expression of AR; IHC expression of E cadherin and CLDN3; IHC expression of p53 and bcl 2 expression of VEGF C, VEGFR3; D2 40 expression. Statistical analysis applied was similar to that described in Chapter 6 Section 6.2. Material and method, except that, as far as VEGF C VEGFR3 lymphatic vascular density clinical and morphological parameters comparative assessment is concerned, we used One Way Anova test (95% CI), also known as analysis of dispersion or analysis of variance. 7.3. ANDROGEN RECEPTOR PROFILE IN BREAST CANCER MOLECULAR SUBTYPES 7.3.1. Argument One of the potential markers that could provide additional information concerning the biological features of BC, including through predictive and prognostic capacity assessment is AR (Moinfar et al., 2003, Ogawa et al., 2008, Park et al., 2010, Yu et al., 2011). Member of the nuclear steroid receptor family, which includes over 100 molecules, AR presents structural, functional and topographical characteristics, relatively similar to those of ER and PR (Agrawal et al., 2008, Ogawa et al., 2008, Moe, Anderson, 2007). Studies of epidemiological retrospective and prospective type, reputed in literature significant association between increased serum levels of androgens and risk of BC in pre- and postmenopausal woman (Dorgan et al., 1996, Hankinson et al., 1998, Tworoger et al., 2006, Bryan et al., 1984), as well as the fact that androgens can directly stimulate growth of mammary tumor cell lines (Wong, Xie, 2001, Kaaks et al., 2005). AR expression is commonly reported in BC (in over 70% of cases), its positivity being equal to that of ER and PR, or even higher (70 90% versus 70 80% versus 50 70%) (Isola, 1993, Moinfar et al., 2003, Kuenen Boumeester et al., 1996, Nahleh, 2008, Lillie et al., 2003, Lea et al., 1989, Hall et al., 1996, Riva et al., 2005, de Mattos Lima Lin et al., 2012, Yu et al., 2011, Niemeier et al., 2010, Ogawa et al., 2008, Peters et al., 2009, Agoff et al., 2003, Micello et al., 2010). AR expression, however, is highly variable, due to large tumor heterogeneities (Micello et al., 2010). In light of histological profile, AR expression mainly characterizes lobular and apocrine carcinoma, being less expressed in mucinous, metaplastic, medullary type (Shim et al., 2006, Riva et al., 2005, Cho et al., 2008, Park et al., 2010, Bratthauer et al., 2002, Gatalica, 1997). 6

The entire data supports the role of AR in the molecular mechanisms of mammary carcinogenesis in relation to hormone dependency (Gonzales et al., 2008) and in the development of BC. However, AR is not yet validated as prognostic or predictive factor, the clinical significance of its expression is not clarified (Park et al., 2010), and its different effects in molecular subtypes of BC are difficult to understand (Nicolas et al., 2007, Luo et al., 2010). Based on available data in the literature, the objectives were: - the AR profile analysis in operational molecular subtypes in BC, in order to identify possible differences in expression; - establishing the existence or the absence of several statistical significant disparities between AR expression and some clinic pathological and molecular features. 7.3.2. Quantification method highlights Evaluation of AR expression was carried out using a semi quantitative score based on the percentage of positive cells (P) (0 1%, 1 2 25%, 2 26 50%, 3 for 51 75%, 4 > 75%) and staining intensity (I) of the IHC reaction (0 none, 1 weak, 2 moderate, 3 intense, 4 strong), total score value resulting by multiplying the score value P by the score of I (Luo et al., 2010). The score with greater or equal value with 2 corresponds to positive reaction. 7.3.3. Results Evaluation of AR profile, reflected by nuclear positivity, revealed, in the whole study group, positive expression in 32 cases and negative in 10 cases. The cases considered positive had score values between 2 and 12 as follows: score 2 5 cases, score 3 4 cases, score 4 9 cases, score 6 5 cases, score 8 6 cases, score 9 1 case, score 12 2 cases. From the point of view of IHC reaction intensity, this was predominantly moderate (17 cases), followed by decreased staining levels (9 cases) and strong level (6 cases). In relation to molecular subtypes, positive AR cases had the following distribution: luminal A subtype 12 cases, luminal B subtype 11 cases, HER2 subtype 2 cases, basal like subtype 7 cases. The 10 negative AR cases corresponded, in report of molecular classification, as follows: luminal A subtype 3 cases, luminal B subtype 2 cases, HER2 subtype 2 cases, basal like subtype 1 case, unclassifiable subtype 2 cases. Considering the basal like subtype and the unclassified one as classification in the negative category, and luminal A, luminal B and HER2 subtypes in non triple negative category, the distribution was as follows: non triple negative category, positive AR 25 cases; non triple negative category, negative AR 7 cases; triple negative category, positive AR 7 cases; triple negative category, negative AR 3 cases. Statistical analysis correlation between AR expression (positive and, respectively, negative) and a series of clinical and pathological (age, histological subtype, T stage, tumor grade) and molecular parameters (expression of RE, PR, Her2/neu, molecular subtype) revealed the absence of any significant differences, except the AR ER association (p = 0.0173). 7.3.4. Discussions Currently, the research directions on AR aim not only correlations with prognostic clinical and pathological aspects along with survival, but also the assessment of its role in molecular subtypes of BC. With reference to the types and subtypes of BC according to molecular classification, recent studies indicate that AR is stronger expressed in non triple negative molecular subtypes versus those triple negative (Park et al, 2010, de Mattos Lima Lim et al., 2012). There are, however, tumors with negative ER and PR hormonal profile, but Her2/neu positive, expressing AR (Micello et al., 2010), as well as for triple negative subtypes, a rate of up to 30% positive AR tumors (Gonzales et al., 2008, Park et al., 2010). With strict reference to molecular subtypes, our study revealed the following configuration for RA: in luminal A subtype 12 positive and 3 negative cases, in luminal B subtype 11 positive and 2 negative cases, in HER2 subtype an equal number of cases, 2 positive, respectively, negative in basal like subtype 7 positive cases, 1 negative case, in unclassifiable subtype 2 negative cases. It is noted the presence of predominantly positive AR in luminal A and B subtype, its expression being more or less aligned to ER and PR expression. It should be noted that the AR expression, positive in 76.19% of all analyzed cases (32 of 42) was higher than ER expression, positive in 61.90% of cases (26 of 42), and, respectively, with PR expression, positive in 64.28% of cases (27 of 42) data that 7

are overlapping on literature reports reviewed in section 7.3.1.Argument. A special accent should be placed on the expression achieved for AR in HER2 subtype, the 4 cases classified as HER2 being, as we mentioned above, equally AR positive and negative, result that deserves a remark, pending the published data. The pathogenesis involved in prostate carcinogenesis, in which occur the interconnection of AR and HER pathways (Ricciardelli et al., 2008), has been extrapolated in mammary carcinogenesis, for negative ER molecular type. Furthermore, recent data indicate that for the HER subtype, the AR association has a positive role in relation to the ER coexpression (de Mattos Lima Lin et al., 2012). Consequently, HER subtype could be further subdivided into other 3 subtypes: AR + / ER + AR + / ER, RA /RE, each with a distinct therapeutic approach. Among these 3 subtypes AR + / ER phenotype has, as morphological correspondent the apocrine histological type, and associate a high degree of malignancy. A study more or less aligned is focused on positive Her2/neu subtype with negative ER and PR expression (Micello et al., 2010), but positive AR expression, the results indicating a significant statistical correlation between HER and AR, with impact on the clinical and therapeutic management of negative ER/PR cases. From AR expression point of view, in non triple negative category versus triple negative, our results are, however, different from the data in the literature, because for the first category, AR was positive in 25 out of 32 cases (78%) than the second one, in which AR was positive in 7 out of 10 cases (70%). The increased percentage obtained of negative triple and positive AR cases (70% in our study group versus 30% literature) can/could be explained, at first, by the fact that our series included a relatively small number of cases and the number of cases included in the two diagnostic groups was not equal. The review of literature data reveals contradictory views on the relation between AR classical clinic pathological prognostic factors. There are studies indicating no statistical significant correlation between AR expression and tumor grade (Isola, 1993, Riva et al., 2005, Moinfar et al., 2003, Agoff et al., 2003, Bieche et al., 2001, Narita et al., 2006], lymph nodes invasion (Soreide et al., 1992, Schippinger et al., 2006), ER or PR status (Riva et al., 2005, Miller et al., 1985, Brentani et al., 1986, Isola, 1993, Moinfar et al., 2003), while other results indicate statistical significant correlations between AR and positive prognostic factors (small tumor sizes, menoreic status, absence of lymph nodes metastases, low tumor grade, low histological grade and ER expression), thus advocating for the AR value as a positive prognostic factor (Isola, 1993, Narita et al., 2006, Bryan et al, 1984, Riva et al., 2005, Moinfar et al., 2003, Agoff et al., 2003, Bieche et al., 2001, Gonzales et al., 2008). To complicate matters further, other authors associate AR with negative prognosis factor value, AR expression level (gene and protein) being correlated with lymph node metastasis (Soreide et al., 1992), in which AR is highly expressed (Lea et al, 1989, Bayer Garner et al., 2000). A further refinement is provided by the AR correlation with classical prognostic factors in negative non triple tumors versus negative triple. Discrepancies of the same type also characterize these studies. For negative non triple tumors, there are opinions both pro (Agrawal et al., 2008, Rakha et al., 2007, Ogawa et al., 2008, Hanley et al., 2008, de Mattos Lima Lim et al., 2012, Park et al., 2010) and contra (Luo et al., 2010), and for triple negative tumors, designations to the analogy concerning AR vascular invasion, sizes of primary tumor, tumor stage, lymph nodes metastases (Luo et al., 2010). Our study led to negative results anent / regarding/ about the correlation between the expression of AR and a series of clinic pathological and molecular parameters, with no statistical significant differences in relation to age, tumor stage, tumor grade and histological subtype, namely, Her2/neu, PR expression and non triple negative, respectively, triple negative molecular subtypes. However, statistical significant differences (p = 0.0173) were obtained for the AR ER correlation, a result that can be interpreted as additional evidence for the favorable prognostic factor value of complex hormonal profile. With strict reference to the relation between AR negative triple molecular subtype, the AR significance analysis was motivated by the fact that they have aggressive biological features and limited benefits from hormone therapy (given that they are negative ER), chemotherapy or targeted therapy being redundancy in adjuvant treatment (Park et al., 2010). The fact that in the analyzed group, there were no statistical significant differences obtained between AR expression and triple negative tumors, respectively, non triple negative, can be explained by the relatively small sizes of our study group. AR remains undoubtedly a marker whose manner should be further analyzed in terms of prognostic and therapeutic valences. 8

Clearly, in the triple negative subtypes, there can be identified, in relation to the existence, respectively, absence of AR, two distinct categories of evolution management: with better prognosis, and more reserved prognosis (Gonzales et al., 2008, Luo et al., 2010). In this perspective, our intention is to pursue in time the biological performance of the analyzed cases, in order to harness the molecular profile identified in terms of prognostic value. 7.4. CLDN3 PROFILE IN BREAST CANCER MOLECULAR TYPES 7.4.1. Argument Claudins family consists of 24 tissue specific transmembrane proteins, essential for the ocludin or tight junctions (eng. tight junctions, TJs) design in epithelial and endothelial cells (Furuse et al., 1998, Hewitt et al., 2006, Oliveira, Morgado-Díaz, 2007, Singh et al., 2010). CLDNs are essential for regulating cell proliferation, differentiation, polarization (Matter et al., 2005) and epithelial assignment. They achieve control of biochemical transfer through the epithelial layer, framing thus a structure of vesicular traffic. Literature data support the involvement of CLDNs in carcinogenesis, resulting in TJs structure and function changes in different cancer types (Oliveira, Morgado Díaz, 2007, Singh et al., 2010). TJ loss or alteration favors the purchase of a malignant cell phenotype, mainly through loss of cell adhesion, alteration of cell differentiation, proliferation control events that lead to local invasion and metastasis (Oliveira, Morgado Díaz, 2007, Singh et al., 2010). CLDNs level may either rise by stimulating gene activity or fall by its inhibition (Oliveira, Morgado Díaz, 2007). If their intervention in carcinogenesis is relatively easy to understand in terms of decreasing the expression level, resulting in structural and functional alterations of TJs, it is more difficult to explain how the increasing of CLDNs expression determines, as stage of the pathogenic mechanism, the initiation and development of malignant transformation (Singh et al., 2010). Based on available data in the literature, the objectives were: the CLDN3 profile analysis in operational molecular subtypes in BC, in order to identify expression differences; finding the existence or the absence of certain differences with statistical significance between the CLDN3 expression and some clinic pathological and molecular features. 7.4.2. Features of quantification method IHC reaction for CLDN3 was semi quantitatively assessed by a scoring system (Seo et al., 2010) based on the percentage (P) of positive cells (1 for 10%, 2 for 10 50%, 3 for 50 75%, 4 for 75%) and the intensity (I) of the coloration reaction (1 weak 2 moderate 3 strong, 4 very strong), the final value being given by the sum of P + I. According to this score, CLDN3 expression was found to be low for range from 1 4, and increased for values comprised between 5 8. 7.4.3 Results Evaluation of CLDN3 profile, reflected by membrane staining, revealed, in the whole study group, three different ways of expression, according to the achieved score values, that is increased positive expression in 25 cases, low positive expression in 12 cases, negative expression in 5 cases. Cases with increased positive expression showed score values between 5 and 8, as follows: score 5 7 cases, score 6 7 cases, score 7 8 cases, and score 8 3 cases. From the IHC strength reaction point of view, it was predominantly strong and very strong (18 cases out of 25). Cases with low positive expression showed score values between 1 and 4, as follows: score 2 6 cases, score 3 2 cases, score 4 4 cases, and score 1 not being certified. From the point of view of IHC reaction strength, this was predominantly low (8 cases out of 12). In relation to molecular subtypes, the distribution of cases was as follows: - high positive CLDN3 cases: luminal A subtype 7 cases, luminal B subtype 11 cases, HER2 subtype 2 cases, basal like subtype 5 cases. - weak positive CLDN cases: luminal A subtype 5 cases, luminal B subtype 2 cases, HER2 subtype 2 cases, basal like subtype 3 cases. - negative CLDN3 cases: luminal A subtype 3 cases, unclassifiable subtype 2 cases. Considering the basal like subtype and the unclassified one as inclusion in negative triple category and luminal A, luminal B and HER2 subtypes in non triple negative group, the distribution was as follows: non triple negative category, high positive CLDN3 20 cases; non triple negative 9

category, weak positive CLDN3 9 cases; non triple negative category, negative CLDN3 3 cases; triple negative category, high positive CLDN3 5 cases; triple negative category, weak positive CLDN3 3 cases; triple negative category, negative CLDN3 2 cases. The results obtained in the semi quantitative assessment of each case, as well as the correlation with molecular subtypes are detailed in table 7.4. Correlation statistical analysis between CLDN3 expression (high positive, weak positive and negative) and a series of clinical and pathological parameters (age, tumor stage, tumor grade, histological subtype) revealed the absence of any statistical significant differences. Given that we met the TI tumor stage with TII and TIII tumor stage with TIV, there were achieved statistical significant differences (p = 0.0466) (Table 7.5, Fig. 7.12). Also, statistical significant differences were obtained in relation to molecular subtypes (p = 0.0045). 7.4.4. Discussions Currently, claudins are not seen only as cell stabilizers through membership in junction structures, but also as mediators or regulators in various pathological pathways involved in carcinogenesis. Their significance in the tumor gynecologic field significantly grew because their expression was correlated with the ability of invasion and metastasis, including breast cancer research focus on their role as prognostic factors (Lanigan et al., 2009, Heinzelmann-Schwarz et al., 2004, Kleinberg et al., 2008, Dahiya et al., 2011, Boylan et al., 2011). The most studied claudins in breast cancer are CLDN1, CLDN3, CLDN4 and CLDN7 (Kominsky et al., 2003, Tokes et al., 2005, Hewitt et al., 2006, Kulka et al., 2009, Blanchard et al., 2009, Lanigan et al., 2009). CLDN3 and CLDN4 are generally overexpressed (Blanchard et al., 2009), while CLDN 1 and CLDN7 are under expressed or absent (Kominsky et al., 2003, Tőkés et al., 2005, Hewitt et al., 2006). With strict reference to CLDN3, a baseline study for breast cancer indicates a membrane localization of CLDN3 in epithelial cells, near the apical end or at laterobasal membrane level, but no clear positive expression in the cytoplasm (Tőkés et al., 2005). Our study was focused on the CLDN3 analysis in all molecular subtypes of breast cancer, considering that the changes in the CLDN3 expression can constitute clear evidence to illustrate the changes in cell adhesion. In luminal A and B subtypes, 18 cases out of those 28 showed an increased expression. Surprisingly, 3 cases were negative, and 7 cases were characterized by low expression these data paving the way for talks on the absence or reducing the CLDN level in relation to the luminal subtype. HER2 CLDN3 subtype relationship is, in our opinion, inconclusive given that in the analyzed group there were only 4 cases classified as HER2 subtype, two of them showing high positive CLDN3 expression, the other two weak positive expressions. CLND3 profile study revealed, for the basal like subtype, 5 CLDN3 cases with high positive expression and 3 cases with weak positive expression. Unclassifiable, normal like subtype was associated with complete absence of CLDN3 expression. Based on operational molecular diagnostic algorithm, we can consider the 3 cases with basal like subtype and low CLDN3 expression as essentially being claudin low subtype. Claudin low subtype (Herschowitz et al., 2007) classified in the triple negative category, together with basal like subtype constitute a potential new subtype in molecular classification of BC (Prat et al., 2010, Eroles et al., 2011), with molecular, clinical and histological features partially superimposed on those of the basal like subtype, but also distinct (Kulka et al., 2009, Blanchard et al., 2009, Perou, 2010). Our results revealed, in parallel, 5 negative triple cases, basal like that associated increased CLDN3 expression which are arguments to support the possible claudin high category recently proposed as a distinct entity in the reporting of cases with an expression extremely high of CLDN1 and CLDN4 (Myal et al., 2010). Although our study has, as major limit, a small number of cases investigated, the results point to the variability of CLDN3 expression, in each molecular subtype, as well as in the two associated diagnostic categories, non triple negative, or triple negative. Consequently, it is considered that the relationship of claudins with molecular subtypes of breast cancer is still incompletely defined. In the light of correlating the CLDN3 expression with classical clinical and morphological parameters, the statistical analysis correlation revealed the absence of any statistical significant differences in age, tumor stage, tumor grade and histological subtype. However, it should be noted that linking the TI tumor stage with TII, and TIII tumor stage with TIV, statistically analyzed in 10

relation to CLDN3, revealed statistical significant differences (p = 0.0466). This result deserves to be highlighted, as reports in the literature on CLDN3 as prognostic factor, are very limited. To our knowledge, only one study, extremely recent, reports the CLDN3 expression in 21.8% of breast invasive tumors, of high histological grade, with metastatic dissemination (Ricardo et al., 2012). CLDN3 overexpression, analyzed with CLDN4 may be associated with a free metastasis at distance survival, overall survival becoming statistically significant (Ricardo et al., 2012). CLDN value in the biological conduct of breast tumors, defined in terms of molecular features, was reinforced by the results of statistical analysis correlation between the CLDN3 expressions, assessed as high, low and respectively absent and the molecular subtypes of diagnostic corresponding to the analyzed cases. The obtained statistical significant differences (p = 0.0045), may be, in our view, arguments to advocate for the definition of new molecular, "claudin low" and "claudin high" subtypes to complete the operational diagnostic entities, currently, by implementing molecular classification. Based on the results obtained, we believe that the dynamics of CLDN3 expression in breast cancer provides important information on the status of maintaining cell adhesion, being potentially usable as a prognostic factor and therapeutic target. The claudin low subtype hallmark is still a difficult process that requires the introduction of additional parameters in molecular diagnostic algorithm. 7.5. E CADHERIN EXPRESSION IN BREAST CANCER MOLECULAR SUBTYPES 7.5.1. Argument E cadherins are, together with claudins, epithelial morphological stabilizers, acting as receiver for maintaining cellular, respectively, tissue structural integrity and mediator of functional interrelations (Andrews et al., 2012). E cadherins modulate invasive mechanisms either through individually participation, either beyond association with members of the cadherin family, interfering by their simple structural organization / distribution or involvement in various signaling pathways (E cadherin intervention in the mammary gland is notified since the embryonic period (involved in epithelial tubulogenesis), until the healing and tumor progression processes, thanks to E cadherins ability to change their phenotype from epithelial to mesenchymal features. With this potential, E cadherins are key elements in the epithelial mesenchymal transition mechanism (which provides mobility and invasion), respectively, mesenchymal epithelial (which provides extravasation and migration) (Chao et al., 2010). Based on available data in the literature, the objectives were: - E cadherin profile analysis into operational molecular subtypes in BC to identify expression differences; - identifying the existence or absence of several statistical significant differences between E cadherin expression and some clinical, pathological and molecular features. 7.5.2. Quantification method highlights IHC reaction to E cadherin was semi quantitatively evaluated, by a scoring system (Qureshi et al., 2006) which combines the percentage of positive cells and staining intensity as follows: 0 no staining or membrane staining in <10% of the tumor cells, 1 incomplete, weak membrane staining in > 10% of tumor cells, 2 complete, weak or moderate membrane in > 10% of tumor cells, 3 complete, strong membrane staining in > 10% of the tumor cells. According to this score, for the 0 and 1 values, the reaction is considered negative, for the score 2 weak positive, and for the score 3 high positive. 7.5.3. Results Evaluation of E cadherin profile, reflected by membrane positivity, revealed, in the whole study group, high positive expression in 26 cases and weak positive in 16 cases. There were no negative cases. Positive expression of E cadherin was characterized by high intensity of immune reaction, in a proportion of tumor cells with values betwen 90 100% in 20 cases, the other 6 cases showing the percentage values of 50%, 70%, 75%, and 80%. The cases with low positive expression showed weak and moderate intensity of IHC reaction, the percentage of positive tumor cells ranging between 20% and 80%. In relation to molecular subtypes, the distribution of cases was as follows: 11