Epithelial Mesenchymal Transition (EMT) in Metastatic Breast Cancer in Omani Women

Abstract

Breast cancer (BC) in Oman affects younger women and has a more aggressive course. Clinical and biological variables like age, pregnancy, tumor size, type, grade, receptor expression and proliferation predict disease aggression but there is no direct predictor of metastasis except lymphovascular invasion. Epithelial-mesenchymal transition (EMT) is characterized by epithelial cells losing epithelial and acquiring mesenchymal morpho-immunophenotypic characteristics. In tumors, EMT-like transitions may signify a metastatic phenotype and have features in common with cancer stem cells (CSC) which show resistance to chemotherapy. This study aimed to identify EMT and CSC phenotypes in metastatic and non-metastatic breast cancer in Omani women and their association with conventional clinico-pathological predictors of BC. In a retrospective study of ninety-six Omani women with breast cancer, the association of age, pregnancy/lactation, tumor size, type, grade, ductal carcinoma insitu (DCIS), lymphovascular invasion, hormone/ HER2 receptor expression and Ki67 proliferation index (Ki67 PI) was tested with EMT/ CSC phenotype and metastasis. Young age ≤ 40 years, lymphovascular invasion and EMT had a strong association with metastasis; CSC approached significance. Vimentin expression in tumor cells, fibronectin and MMP-11 in stroma were reliable markers of EMT; dual EMT and CSC phenotype (Vim+/ CD44+/ CD 24−/low) had a strong association with apocrine variant, basal-like tumors and triple negative cancers. EMT had a strong association with Ki67 proliferation index (PI) and CSC with HER2-like tumors and distant metastasis. These select markers may be useful in metastasis-prediction in pre-treatment biopsies.

Introduction

Epithelial-mesenchymal transition (EMT) is a phenomenon characterized by epithelial cells losing epithelial and acquiring mesenchymal morpho-immunophenotypic characteristics. This change is seen physiologically as EMT type 1 in embryological development and EMT type 2 in the body’s inflammatory response [1]. Type 3 EMT is seen in several solid malignancies including breast, lung etc. The influence of the tumor microenvironment (TME) on tumor initiation and progression or conversely, on tumor inhibition, originated in the ‘seed and soil’ theory of tumor growth. In the last century, individual components of the microenvironment have been intensively investigated including inflammation, immunologic responses, tumor microvasculature and the extracellular matrix [2].

EMT in tumor cells is typified by changes in cellular morphology, altered cell–cell and cell–matrix adhesion, and the development of migratory behavior and invasiveness [3]. Thus, EMT represents morphologic, phenotypic and molecular changes with several implications; to the pathologist it poses a diagnostic challenge for sub-classification of the tumor, to the biologist it translates into behavioral changes from epithelial-stromal cross-talk [4, 5] to locomotion, invasion and metastasis [6, 7]; to the oncologist it may connote altered response or resistance to therapy.

The pioneering work of Bissell saw the emergence of TME from the shadows of genetics- dominated determinants of tumor development [8] especially in relation to breast cancer. Simultaneously, Sager suggested that cancer-associated genetic mutations alone could not explain all the phenotypic expressions of malignancy especially in the later stages of tumor growth and spread when TME may influence tumor behavior [9].

It has been proposed that metastases are seeded by migrating cancer stem cells (CSCs) that undergo EMT, allowing them to disseminate [10, 11]. EMT-like transitions have many features in common with cancer stem cells (CSC) [12] and show general resistance to chemotherapy [13, 14]. While chronic activation of growth factors like EGFR and IGFR cause EMT, once achieved, the phenotype offers resistance to drugs targeted at these receptors. Molecular and epigenetic changes at inception and evolution of the EMT state have been explored [15]. It is ironical that anti-angiogenic therapy can actually promote formation of CSC’s, an undesirable outcome that may increase the metastatic potential of a tumor [16].

Despite accumulating experimental data there is a lack of convincing and adequate exploration into patient’s biopsies that demonstrate EMT as a predictor of a tumor’s metastatic phenotype [17, 18]. A recent report claims the utility of claudin2 immunohistochemistry as a negative biomarker for predicting liver metastasis [19]. Breast cancer cells surviving chemotherapy have been shown to demonstrate the cancer stem cell phenotype (CD44⁺ /CD24^−/low) indicating the possibility that they can be responsible for recurrence [20]. In a study on pre-chemotherapeutic biopsies, though CD 8+ tumor infiltrating lymphocytes showed a significant association with the cancer stem cell or EMT phenotype they conferred higher partial chemotherapeutic remission (pCR) [21].

Breast cancer is a leading malignancy in women in Oman. It constitutes one out of five cancers diagnosed in Oman and one-third of patients are less than 40 years in whom the cancer exhibits an aggressive behavior [22]. About 15–18% patients are Stage IV at presentation. Currently accepted prognostic and predictive factors estimating the biological behavior of breast cancer and its response to therapy include tumor size, type, grade, stage, hormonal receptor (HR) status (ER, PR), growth factor receptor activation (Her2neu expression) and proliferative potential eg. Ki67 proliferation index (PI). However, prediction of metastatic ability has remained uncertain. Response to chemotherapy is also unpredictable. Its affliction of younger women in Oman introduces the additional confounding co-existence with pregnancy, that itself induces epithelial-stromal changes within the breast.

The rationale of this study was to to identify the EMT and the CSC phenotype in the initial diagnostic breast biopsies in Omani women and establish it’s association both with conventional clinico-pathological predictors of BC and the presence or absence of metastasis. The hypothesis was that expression of the EMT and CSC morpho-immunophenotype may have a bearing on predicting subsets of breast cancer with metastatic potential. This, in turn, may alter therapeutic management strategies targeting more aggressive disease.

Material and Methods

Patient Selection and Retrieval of Archival Material

Ninety-six Omani females with biopsy-proven Invasive ductal carcinoma of the breast, who presented to the Breast Unit of Sultan Qaboos University Hospital (SQUH) with a breast lump or were found to have mammographic/ultrasound detected disease, formed the study group. Clinical records were accessed from the hospital information system (HIS) for clinical presentation and cTNM staging of disease. A retrospective retrieval of histopathologic archival material (archived slides and paraffin blocks) to review and document tumor characteristics was carried out. Initial (pre-treatment) biopsies of patients who presented with regional (nodal) or distant metastasis were compared with patients not showing metastasis at presentation.

Inclusion/Exclusion Criteria

1. All biopsies were from female patients

2. Core biopsies, wide local excision specimens and mastectomy specimens were included.

3. Biopsies/excisions after the administration of chemo- or radiotherapy were excluded to obviate observations as a result of reparative changes.

4. Invasive ductal carcinoma, no special type (NST) and variants were included. Invasive lobular carcinoma, Paget’s disease of breast and pure duct carcinoma in situ (DCIS) were excluded.

Clinical Variables

These included:

a) Age at presentation i) ≤ 40 years (young women with breast cancer who are considered to have more aggressive disease) and ii) >40 years using a conventional cut-off age criterion [23, 24] for young and older women with breast cancer respectively; b) concurrent pregnancy/lactation; c) tumor size and d) tumor stage (both defined by the TNM staging system) [25]; e) presence or absence of metastasis.

Pathological Variables (Morphologic and Biologic)

These included a) morphologic type as per the WHO classification [26]: tumors were grouped as i) Invasive ductal carcinoma NST ii) morphologic variants e.g. cribriform, mucinous, apocrine, neuroendocrine etc. b) ductal carcinoma in situ (DCIS) present/absent; tumor grade (Nottingham grading system) [27]; lymphovascular invasion present/absent (confirmed by CD34 immunostaining); pathological stage pTNM [25]; hormone receptor expression - estrogen and progesterone (ER/ PR) - determined by immunohistochemistry as positive/negative (Allred score) [28]; human epidermal growth factor (HER2) expression by immunohistochemistry positive/ negative by scoring as 0/1+ (negative), 2+ (equivocal, final confirmation by FISH) and 3+ (positive) [29]; Ki 67 proliferation index (PI) low/high (St. Gallen’s consensus criteria) [30].

Epithelial-Mesenchymal Transition (EMT) and Cancer Stem Cell (CSC) Phenotype

EMT phenotype was determined by diffuse expression of one or focal/patchy strong expression of at least two of Vimentin (cytoplasmic); E-cadherin and beta catenin (reduced/absent membrane expression compared to normal ducts; N-cadherin (membranous); smooth muscle actin, fibronectin and stromelysin- 3 (MMP11) (peritumoral stroma). CSC phenotype was marked by CD44⁺/CD24^−/low in the tumor cells. Antibody selection, antigen retrieval method and autostainers used for immunohistochemistry are detailed in Table 1. 3,3′-Diaminobenzidine (DAB) was used as chromogen and the end product visualized as a brown colour at the reaction site.

Table 1.

Antibody specifications and methodology of use

Antibody	Manufacturer	Clone	Dilution	Antigen Retrieval	Auto-stainers*
Fibronectin Rabbit Monoclonal	Abcam	F1	1:500	High pH PH 9.0 PT link	D
MMP-11 Mouse Monoclonal	Abcam	SL3.05	1:500	High pH PH 9.0 PT link	D
Vimentin Mouse Monoclonal	DAKO	VIM3B4	1:100	High pH PH 9.0 PT link	D
Smooth muscle actin Mouse Monoclonal	Ventana	1A4	Ready to use	CC1	V
Beta-Catenin Mouse Monoclonal	Ventana / Cell Marque	14	Ready to use	CC1	V
CD 44 Rabbit Monoclonal	Ventana	SP37	Ready to use	CC1	V
E.Cadherin Mouse Monoclonal	Ventana	36	Ready to use	CC1	V
CD 24 Mouse Monoclonal	Abcam	8.B.76	1:500	CC1	V
N-Cadherin Mouse Monoclonal	Life Technologies / Thermo Fisher	3B9	1:500	CC1	V

*Autostainers: DAKO Autostainer Link 48 (D) Ventana BenchMark Ultra (V)

Statistical Methods and Analysis

The data base of clinical and pathological variables detailed above was subjected to statistical analysis. Descriptive analysis of prevalence and frequencies was expressed as percentages. Statistical association between the clinical and pathological variables and prevalence of metastasis was carried out using Chi-square test. The odds ratios (ORs) and their 95% confidence intervals (CI) were calculated for degree of association between the clinico-pathological categories and metastasis status of the subjects as well as between the biological variables ER/PR, HER2, Ki 67 PI and EMT & CSC separately. The odds ratios were tested using Mantel-Haenszel Chi-square test. A p-value (two-tailed) of <0.05 was considered as statistical significance. All data was analysed using IBM SPSS Statistics version 23.

Ethical Approval

The study was approved by the Medical Research & Ethics Committee of College of Medicine & Health Sciences, Sultan Qaboos University, Oman vide MREC # 903A/2014.

Results

The study cohort comprised ninety-six Omani women with biopsy-diagnosed and staged invasive ductal carcinoma and its variants at SQUH. Fifty-five patients had metastasis, while forty-one patients were non-metastatic. Regional metastasis to axillary lymph nodes were present in 43 patients while 12 had distant metastasis. Table 2. illustrates the association of clinical and pathological variables of breast cancer with presence or absence of metastasis. Table 3. depicts the relative significance of these factors in predicting metastasis. Tables 4 and 5. display the association of conventional biologic prognosticators of clinical outcome and predictors of response to therapy in breast cancer with EMT and CSC phenotype.

Table 2.

Baseline characteristics and clinical variables associated with the development of metastasis among female breast cancer patients

Variables	Total	With Mets	Without Mets	p-value
n (%)	96	55 (57.3)	41 (42.7)
Age groups (years), n (%)
≤ 40	30	23(76.7)	7(23.3)	0.018**
> 40	66	32(48.5)	3451.5)
Preg/Lac
No	84	46(54.8)	38(45.2)	0.225
Yes	12	9(75.0)	3(25.0)
Tumor size
T1 + T2	49	22(44.9)	27(55.1)	0.021**
T3 + T4	47	33(70.2)	14(29.8)
Histological Grade
I	10	4(40.0)	6(60.0)	0.508
II	34	20(58.8)	14(41.2)
III	52	31(59.6)	21(40.4)
DCIS
Negative	66	38(57.6)	28(42.4)	0.934
Positive	30	17(56.7)	13(43.3)
LVI
Negative	78	41(52.6)	3747.4)	0.066*
Positive	18	14(77.8)	4(22.2)
HR
Negative	63	40(63.5)	23(36.5)	0.139
Positive	33	15(45.5)	18(54.5)
HER2
Negative	62	38(61.3)	24(38.7)	0.393
Positive	34	17(50.0)	17(50.0)
Ki67 PI
Low	22	13(59.1)	9(40.9)	0.846
High	74	42(56.8)	32(43.2)
EMT
Negative	63	31(49.2)	32(50.8)	0.046**
Positive	33	24(72.7)	9(27.3)
CSC
Negative	73	38(52.1)	35(47.9)	0.091*
Positive	23	17(73.9)	6(26.1)
Tumor type
Variant	24	11(45.8)	13(54.2)	0.284
IDC	72	44(61.1)	28(38.9)

** p < 0.05; *p < 0.10

Table 3.

Degree of association of clinico-pathological variables with metastasis among female BC patients

Variables	OR	95% CI	P-value
Age group (years)
> 40	1
≤ 40	3.49	1.32–9.25	0.012
Preg/Lac
No	1
Yes	2.48	0.63–9.80	0.196
Tumor size
T1 + T2	1
T3 + T4	2.89	1.25–6.71	0.013
Histological Grade
I	1
II	2.14	0.51–9.02	0.29
III	2.21	0.56–8.81	0.26
DCIS
Negative	1
Positive	0.96	0.40–2.30	0.933
LVI
Negative	1
Positive	3.16	0.95–10.45	0.060
HR
Positive	1
Negative	2.09	0.88–4.91	0.092
HER2
Positive	1
Negative	1.58	0.68–3.68	0.286
Ki67 PI
Low	1
High	0.90	0.34–2.38	0.846
EMT
Negative	1
Positive	2.75	1.11–6.84	0.029
CSC
Negative	1
Positive	2.61	0.92–7.36	0.070
Tumor Type
Variant	1
IDC	1.86	0.73–4.72	0.193

Table 4.

Biological variables associated with EMT and CSC phenotypes among female BC patients

Variables	N	EMT positive	EMT negative	p-value	CSC positive	CSC negative	p-value
HR(ER/PR)
Negative	63	20(31.7)	43(68.3)	0.601	15(23.8)	48(76.2)	0.962
Positive	33	13(39.4)	20(60.6)		8(24.2)	25(75.8)
HER2
Negative	62	25(40.3)	37(59.7)	0.152	19(30.6)	43(69.4)	0.068*
Positive	34	8(23.5)	26(76.5)		4(11.8)	30(88.2)
Ki67 PI
Low	22	4(18.2)	18(81.8)	0.079*	4(18.2)	18(81.8)	0.661
High	74	29(39.2)	45(60.8)		19(25.7)	55(74.3)

* p < 0.10

Table 5.

Degree of association of metastasis-related biological factors with EMT and CSC among female BC patients

Variables	EMT			CSC
	OR	95% CI	P-value	OR	95% CI	P-value
HR
Positive	1			1
Negative	1.40	0.58–3.36	0.454	1.02	0.38–2.74	0.962
HER2
Positive	1			1
Negative	2.20	0.86–5.62	0.101	3.31	1.02–10.72	0.046**
Ki67 PI
Low	1			1
High	2.9	0.89–9.43	0.077*	1.56	0.46–5.17	0.472

** p < 0.05; * p < 0.10

Clinical Variables

Age at Diagnosis

The study cohort comprised ninety-six women ranging in age from 23 to 82 years (mean 47.2 years). Thirty women ≤40 years had an age range from 23 to 40 years (mean 35.5 years) and sixty-six women were above 40 years with an age range of 41 to 82 years (mean 54.59 years). In the ≤40 years group, 23/30 women had cancers associated with metastasis compared to 32/66 in the >40 years group. Thus younger women had a higher metastatic potential compared to older women the difference being statistically significant (p = 0.018). (Tables 2and 3). Distant metastasis to supraclavicular nodes, contralateral breast and bones were seen in 11/96 women - 8 women ≤40 years vs. 3 women over 40.

Pregnancy/Lactation Status

The age of 12/96 women with pregnancy associated breast cancer (PABC) ranged from 23 to 34 years (mean 26.5 years). Though three-fourth (9/12) PABC had metastasis the association was not statistically significant (p = 0.225) probably due to the gross disparity of sample size between the non-pregnant and pregnant sub-groups (n = 84 vs.12).

Tumor Size

For estimating the influence of tumor size (category cT in the TNM classification) the earlier/smaller (T1 + T2) categories were compared to locally advanced disease (T3 + T4). Metastasis was significantly associated with the latter (p = 0.021).

Pathological Variables

Tumor Type

The majority of tumors (n = 72) were diagnosed as invasive ductal carcinoma, no special type (IDC-NST). The remaining twenty-four included IDC variants apocrine (n = 8), cribriform (n = 4), neuroendocrine (n = 3), mucinous (n = 1) and micropapillary (n = 1). Basal-like phenotype (n = 7) a biological sub-type (suspected on morphology, often triple-negative and confirmed through its expression of basal markers) was also analysed as a variant. There was no significant association with metastasis among the morphologic variant categories.

Histological Grade

More than half the patients had a Grade III cancer (n = 52). The association with metastasis among different grades was not statistically significant (p = 0.508) especially because there were few Grade I tumors (n = 10).

Ductal Carcinoma in Situ (DCIS)

Ductal carcinoma in situ was seen in association with the invasive tumor in 30/96 patients. There was no significant association with metastasis (p = 0.934). In the majority of cases DCIS was intermediate to high grade (26/30). Extensive DCIS was seen in 3/30 cases.

Lymphovascular Invasion (LVI)

This microscopic feature, which is the most direct evidence of metastatic potential was seen only in 18 patients. This is a disadvantage of core biopsies where the limited tissue available for evaluation may not reveal this feature. However, when present, it is a strong predictor of metastasis as is evident through its achieving statistical significance in the analysis (p = 0.066) (Tables 2 and 3).

Hormone Receptor (HR)

Estrogen/progesterone (ER/PR) receptor positivity was seen in 2/3rd patients (n = 63). Conventionally, this biologic marker confers good prognosis in terms of overall outcome and is a predictor for institution of endocrine therapy. However its association with metastasis prediction was not statistically significant (p = 0.139).

HER2 Receptor

Thirty-four patients had HER2 positive tumors. Conventionally this biologic marker confers poor prognosis in terms of overall outcome though it makes tumors amenable to targeted therapy. The association of HER2 with metastasis was not statistically significant (p = 0.393).

Ki 67 Proliferation Index (PI)

A majority of patients had a high Ki 67 PI (n = 74). This surrogate biomarker of cell proliferation was not significantly associated with metastasis (p = 0.846). This may again be explained by very few tumors in the low Ki67 PI category (n = 12).

Morphologic Observations on EMT and CSC Expression

1. The most common stromal expression of EMT was in the form of fibronectin and MMP- 11 expression. Smooth muscle actin expression was patchy and seen more often when the tumor had a desmoplastic stroma (Figs. 1 and 2).

2. The most consistent expression of EMT in the tumor cells was strong and diffuse vimentin expression. Less reliable was the weakened E-cadherin and β catenin expression which was useful when there was an internal control to compare with. N-cadherin was expressed in few tumor cells and was focal in distribution (Figs. 2 and 3).

3. Apocrine variants of IDC and the triple negative phenotype (ER⁻/PR⁻/Her2⁻) most often showed co-expression of EMT and CSC combined phenotype (Vim+/ CD44⁺/CD 24^−/low) (Figs. 4 and 5).

4. Basal-like tumors showed strong expression of vimentin (Fig. 5).

Fig. 1.

Metastatic invasive ductal carcinoma with EMT phenotype. a and b. Invasive ductal carcinoma surrounded by dense sclerosis (a ×100; b ×400). Immuno-profile – c-d (IHC with DAB). The stroma is rich in fibronectin (c ×100; d ×400)

Fig. 2.

An invasive ductal carcinoma (IDC) Grade III, hormone receptor (HR) positive with metastasis to bone expressing epithelial-mesenchymal transition (EMT) phenotype. a and b. Invasive ductal carcinoma Grade III (H&E. a ×100; b ×400). Immuno-profile (IHC with DAB) (c). MMP-11 rich stromal cells (×400). d and e. Smooth muscle actin-rich stromal cells (d ×100; e ×400). f. N-cadherin expression in tumor cells (f ×400)

Fig. 3.

Core biopsy of an Invasive ductal carcinoma with EMT and CSC (CD 44⁺/CD24⁻) phenotype. a and b. Invasive carcinoma - cohesive (arrow) and dyscohesive areas (arrow-head) (H&E a.×100; bx400). Immuno-profile (IHC with DAB) The tumor cells express Vimentin (c ×100; d ×400); and CD 44 (ex100; f ×400); are negative for CD 24 (g ×100; h ×400). E-cadherin expression is weaker in the dyscohesive component compared to the cohesive component (i ×100 vs. j ×400)

Fig. 4.

Core biopsy of an Invasive ductal carcinoma (IDC) with epithelial mesenchymal transition (EMT) and cancer stem cell (CSC) phenotype with perineural and perivascular invasion. A. Perineural spread of IDC (H&E ×400). B. Perivascular spread of IDC (H&E ×400). Immuno-profile (IHC with DAB). C. Perineural and D. perivascular invasive tumor cells express beta catenin (×400); perivascular invasive tumor cells also express fibronectin (E ×400) and CSC phenotype (CD44+) (F ×400). Abbreviations: n – nerve, v – blood vessel

Fig. 5.

Invasive ductal carcinoma, surrounded by dense lympho-plasmacytic infiltrates. The tumor was triple-negative, basal-like, with EMT and a CSC phenotype (CD 44⁺/CD24⁻) a and b. Infiltrating cords of tumor cells (arrow) with a prominent inflammatory response (arrow-head) (H&E ×100 & 400). Immuno-profile (IHC with DAB) – (c). Basal markers CK5/6 and (d). CK 14 are expressed in the tumor cells (×400); e. Ki67 PI is 50% (×400); f. CD 44 expression is seen both in tumor cells and the inflammatory cells (×400). g. N-cadherin positivity (×400). h. CD 24 expression is weak (×400)

Epithelial Mesenchymal Transition (EMT) and Cancer Stem Cell (CSC) Phenotype

The expression of EMT was seen in 33/96 patients and CSC in 23/96 patients. A dual (EMT + CSC) phenotype was seen in ten patients. EMT presence showed significant association with metastasis (p = 0.046) and CSC presence approached significance (P = 0.091) (Table 2).

On correlating EMT and CSC with the biological behaviour predictors HR receptor, HER2 and Ki 67 PI, there was a significant association between EMT and high Ki67 PI (p = 0.079) and between CSC and HER2 positivity (p = 0.068). (Tables 4 and 5).

When EMT and CSC were compared with respect to their association with luminal categories of breast cancer, EMT was seen more often in triple negative cancers while CSC was more often associated with triple positive and HER2-like cancers.

On comparing the metastatic sites (43 regional, 12 distant) in relation to EMT and CSC, the majority of EMT-only tumors were metastatic to regional lymph nodes (11/14). EMT + CSC phenotypes showed metastasis to supraclavicular lymph nodes, contralateral breast or bone in 4/10 cases. The CSC-only phenotype showed distant metastasis in 5/7 biopsies.

Bivariate Analysis

Clinical and Pathological Variables with Metastasis

The odds ratio among clinical variables showed the greatest influence of age ≤ 40 years in the development of metastasis (OR = 3.49; p = 0.012). Among pathological variables lymphovascular invasion was the most significant factor (OR = 3.16; p = 0.060) followed by EMT expression (OR = 2.75; p = 0.029) in prediction of metastasis (Table 3).

Biological Variables with EMT and CSC

The odds ratio for conventional biological variables in relation to EMT and CSC (HR/HER2 and Ki67 PI) showed the highest correlation of Ki67 PI with EMT (OR = 2.9; p = 0.077) and of HER2 with CSC (OR = 3.31; p = 0.046) compared to the other variables (Table 5).

Discussion

Extracellular Matrix (ECM) and the Biology of Metastasis

Tumor microenvironment constitutes a complex mileu of structural and matrix proteins which are remodified to interact with and influence the tumor. The salient constituents of the interacting components of the ECM include basement membranes (laminin and collagen IV interlinked by nidogen/perlecan) and glycosoaminoglycans link with proteoglycans (perlecan, decorin, agrin and syndecan). The matrix proteins form the ‘glue’ between the fibrous proteins collagen, laminin and fibronectin [31]. Advances in the knowledge of the biochemical composition of the ECM and ECM-mediated signaling in normal tissues, inflammation and tumors in the last three decades, has opened up an entire new field of exploration.

Pathologists have long recognized the peculiarities of the peri-tumoral stroma through morphologic observations of structural changes like sclerosis, myxoid degeneration and the inflammatory response. In tumors, ECM changes affect growth, differentiation and tumor spread to regional and distant sites identified by the surrogate demonstration of tumor emboli in lymphovascular channels. This has shifted attention from the tumor characteristics alone to the possibilities of developing newer therapies that affect the ECM.

Tumor proliferating stimuli of the ECM can induce or suppress growth. The induction of tumor dormancy is a two-edged sword. On the one hand, it may prevent tumor cells establishing themselves at metastatic sites; on the other, cells that are stalled at the G1/S phase would be resistant to chemotherapy [32].

The complex interactions between tumor cells and ECM are responsible for motility, invasion and migration of tumor cells. Expression of integrins on tumor cells and elaboration of matrix metalloproteinases (MMP) both contribute to tumor migration. These can form targets for inhibiting metastasis of single or clusters of tumor cells [33]. Heparan sulfate proteoglycans (HSPG’s) found both in cell membranes and in ECM (especially basement membranes) constitute the substrates for the action of heparanase. One of the HSPG’s are CD44 variants like the V-3 isoform. The HSPG’s in turn regulate the Wnt/β catenin pathway which itself has a role in cell migration [31]. Heparanase inhibitors are now under study to determine their efficacy in limiting or retarding the development of metastasis in experimental models [34].

Cell membrane proteins called integrins (made up of 18α and 8β subunits) form an important bridge between the ECM on the outside and cytoskeleton and intracellular signaling (RAS/MAPK and PI3 kinase/mTOR) within the cell. Thus integrins affect proliferation, adhesion and migration. Interestingly, the heterogeneity in subunit overexpression is tumor-specific hence therapeutic targeting cannot be generic [35]. This has led to the emergence of several sub-unit targeted inhibitors like cilengitide, S247, Abituzumab and Vedolizumab for metastatic melanoma, lung and breast cancers [36].

Exosomes, membrane-bound vesicles of exocytic origin, encapsulate miRNA’s and proteins and travel outside the cancer cells to transfect and transform neighbouring normal cells to the cancer phenotype. Their recognition has generated a new line of thinking in cancer progression, spread and metastatic lodging in distant organs. This understanding brings with it possibilities for use as biomarkers in liquid biopsies. A step further would be targeting these membrane bound products to complement therapeutic choices available currently [37].

Epithelial-Mesenchymal Transition (EMT) and Cancer Stem Cell (CSC) Phenotype

The term Epithelial Mesenchymal transition (EMT) was coined as a unifying concept to combine Epithelial Mesenchymal transformation and Epithelial Mesenchymal transdifferentiation in the first meeting of The International EMT Association (TEMTIA) in 2003. Epithelial-mesenchymal transition has been divided into three classes. Type I EMT occurs in embryogenesis and organogenesis with migrating cells enabling organ formation. In Type II EMT there is a reawakening of mesenchymal plasticity in mature tissues in response to inflammation to enable repair. Type III EMT involves the development of mesenchymal properties in epithelial tumor cells (influenced by matrix proteins) to enable tumors to travel to metastatic sites and establish themselves through the reverse process of mesenchymal-epithelial transition (MET) [12, 23, 38]. A host of upregulation and downregulation of epithelial and mesenchymal markers permit the recognition of this phenomenon [18].

EMT Biomarkers

Several biomarkers have been utilized to demonstrate EMT. Cell membrane proteins like E-cadherin can show reduced expression [39] or a switch for e.g. from E-cadherin to N-cadherin. Similarly, acquisition of specific integrins for e.g. high expression of β6 integrin in colon cancer can be an indicator of progression and the metastatic phenotype [40]. In the current study tumor expression of E-cadherin was weaker in the poorly differentiated areas. N-cadherin expression was seen focally and in few cases only.

Cytoskeletal mesenchymal proteins like vimentin finding aberrant co-expression in epithelial tumors have long been researched [41]. The expression of vimentin not only ‘marks’ the EMT stamp on a tumor cell but probably directs properties of change in shape, motility and reduced adhesiveness: all of which are pro-metastatic [42]. Vimentin expression was a consistent, reliable and easily identifiable marker of EMT in breast biopsies in the current study. Its usual absence in normal breast ducts gives it greater importance as an abnormal phenotypic change. Smooth muscle actin [43] has been found to be increased in the basal phenotype of breast cancers. In our study it corresponded to tumors with strong desmoplasia. The re-localisation of beta catenin from the membrane to the cytoplasm or nucleus may suggest disruption of its interaction with E-cadherin [44, 45]. We found reduced beta catenin expression in poorly differentiated areas but did not identify nuclear relocation of this marker.

Several stromal proteins are overexpressed in the tumor stroma in Type III EMT; examples include fibronectin [36, 46] and laminin 5 [47] (produced by fibroblasts and epithelial cells respectively) and responsible for epithelial-mesenchymal adhesion. Changes in these proteins are notably different between benign and malignant breast tissues and may contribute to tumor invasiveness [46]. Fibronectin binds to ECM and increases binding to integrins and is often found to be increased in association with more anaplastic tumor cells [36, 46]. In the current study, fibronectin increase was a reliable stromal indicator of EMT being expressed strongly and diffusely either alone or in conjunction with MMP11.

Transcription factors have been under study with specific focus on the snail family (which decrease the expression of E-cadherin, claudins, cytokeratins and increase mesenchymal expression) [48]; the twist transcriptional proteins independently do the same [49]. Finally the latest addition to influencers of EMT have been epigenetic events like modification of microRNAs for eg. downregulation of miR-200 in ductal and metaplastic breast cancers [50].

Cancer Stem Cell (CSC) Biomarkers

Tumor-stroma cross-talk influences the expression of EMT. The change confers enhanced survival through a greater migratory potential, cancer stem cell-like (CSC) properties and chemoresistance [51]. Breast cancer stem cells (BCSC) with CD44⁺/CD24^−/low phenotype were recognized by Mani et al. within the EMT state and have also been linked with increase in invasive capacity of the tumor cells [12, 52, 53]. In the current study, EMT was seen in 33/96 patients while CSC were identified in one-fourth of the cases and EMT-CSC co-expression was seen in ten. It is important to note that EMT-CSC co-expression was associated with higher rate of distant metastasis.

EMT/CSC Phenotype in Breast Cancer

Triple-negative breast cancers (ER/PR/HER2 negative) have been a challenge to oncologists because of lack of targets to therapy, aggressive behavior and a uniformly poor outcome [54]. Tumor biologists have demonstrated the transcriptional heterogeneity of the group. There has been revival of interest in sub-classifying these tumors into basal-like (BL), luminal androgen receptor type (LAR) and immunomodulatory (IM) and mesenchymal stem-like subtypes (MSL); the last two based on quantification of inflammatory response and tumor-associated stromal phenotype respectively [55].

Breast cancers that develop mesenchymal characteristics with acquisition of vimentin expression and decreased/low cytokeratin/claudin expression have been referred as claudin-low tumors [56]. This trans-differentiation occurs normally in development but is increasingly being observed in cancers [57].

The seminal work of Creighton et al. demonstrated that breast cancers are found to be enriched in EMT (in vivo in biopsies) and cancer stem cell phenotype (in vitro formation of mammospheres reflecting self-renewal capacity) after endocrine (letrozole) and chemotherapy (docetaxel) of breast cancer and in metastatic effusions [20]. These cells express the CSC profile CD44⁺/CD24^−/low (CSC) and/or Vimentin-cytokeratin co-expression with matrix metalloproteinase 2 (MMP2). The authors used flow cytometry and functional assays to establish this change in tumor phenotype. It is postulated that these cells may be tumor initiators; responsible for disease persistence and spread and cause refractoriness to chemotherapy [58, 59]. On the other hand, Chang et al. have shown that lapatinib treatment reduced the CD44⁺/CD24^−/low (CSC) population [42]. Increasing demonstration of overlapping characteristics of EMT and CSC recycling sub-populations opens the door for a unified targeted therapy that decimates these undifferentiated cells with resultant improvement in cancers that show significant expression of this phenotype [60].

Conventional Clinico-Pathological Predictors and EMT/CSC in Tumors

Young women with breast cancer have been reported to have more aggressive disease along with several other poor prognostic parameters that determine the biology of the disease [23, 24].

Pregnancy associated breast cancer (PABC) is well known to be aggressive with a poor outcome compared to non-pregnant women [61]. Several biological characteristics including younger age, ER negativity or triple negativity have been considered responsible. It is possible that EMT may be another contributing factor. In 12 PABC’s in our cohort, nine had metastasis but in comparison to the remaining 84 non-pregnant patients this did not achieve statistical significance.

ER/PR positivity and HER2 negativity conventionally confer good prognosis in breast cancer. However, in the present study they did not influence the metastasis prevalence. In addition, since a majority of our patients had a high Ki67 PI the influence of this biomarker did not find association with metastasis.

EMT biomarkers have been the subject of investigation in clinical biopsies in cancer in recent years. Studies on gastric and intestinal adenocarcinomas consistently show a relationship with aggressive behavior and poor outcome. Sung et al. studying EMT phenotype of esophageal squamous cell carcinoma using epithelial marker E-cadherin and mesenchymal markers vimentin, smooth muscle actin and fibronectin classified the tumors into wild-type (epithelial), complete (mesenchymal) and intermediate (hybrid) [62]. Their study in 2011, revealed the worst overall survival (OS) and disease-free survival (DFS) in the complete group and also a strong correlation with tumor size, differentiation and depth of invasion. Rhuy et al. [63] found loss of E-cadherin and acquisition of vimentin to be the strongest predictors of poor outcome in gastric adenocarcinoma. In a combination of E-cadherin, snail 1, vimentin and CD44, the strongest indicator of EMT was abnormal expression of 3 or more markers [63].

Bae et al. [64] explored EMT expression in breast carcinoma using only a combination of e-cadherin and fibronectin expression and using the nomenclature of Sung et al. found a complete EMT phenotype in 4.3% of samples from 1495 patients. The phenotype correlated strongly with young patients, higher pT and pN stages, grade, lymphovascular invasion and triple negative receptor profile. The results were limited by use of only 2 markers and the microarray platform limited the area of tumor tissue examined [64]. In a large study with microarrays of 617 clinical biopsies from breast cancer patients transcription factors that define ‘stemness’ were studied. There was stong nuclear expression of SLUG and loss of expression of SOX9 and SOX 10 seen in triple negative breast cancers compared to luminal A/B or Her-2 positive cancers [65] . In the current study, the higher EMT and CSC phenotypic expression compared to the previous studies may be reflective of the demographics of the cohort since women ≤ 40 years accounted for one-third of the patients. We demonstrated that young age and lymphovascular invasion showed significant correlation with metastasis. Thus EMT and CSC expression may contribute to disease aggression seen in women in the region. Lymphovascular invasion may not always be detected in the limited tissue sampled by core biopsies for initial diagnosis.

In the present study, strong and diffuse stromal expression of fibronectin and co-expression of fibronectin and MMP 11 were found to consistent and reliable indicators of EMT. In the tumor cells strong and diffuse vimentin expression was alone diagnostic of EMT. It has been noted that vimentin expression is an independent poor prognosticator in tumors [63].

In our study, among tumor sub-types, apocrine carcinoma and basal-like tumors showed EMT/CSC co-expression. Vimentin expression was especially strong with the basal-like phenotype. This is in consonance with other studies which found that basal-like breast cancers have an aggressive behavior, increased invasiveness and higher EMT expression than invasive ductal carcinoma NST [18]. Vimentin expression on tumor cells has been associated with higher grade and lower ER expression in breast cancer [66]. Sarrio et al., in a tissue microarray based study of 479 invasive breast carcinomas and 12 carcinosarcomas with 28 IHC-based biomarkers found a strong association of EMT with the basal subtype [43].

In the present study, both EMT and CSC expression achieved statistical significance in predicting metastasis in univariate analysis and followed young age and lymphovascular invasion in predictive value on bivariate analysis. EMT was associated more with triple negative tumors and higher Ki67 PI while CSC showed association with triple positive (ER+/PR+/Her2+) and HER2-like tumors.

Future Directions

Establishing ECM changes that predict metastatic potential serves to alert the clinician to prognosis and decision on a more aggressive initial approach to designing patient-specific therapy. There is now increasing understanding of the significant role ECM proteins play in tumor progression: both as scaffolding and by influencing tumor growth and spread. A few are already emerging as potential therapeutic targets [34]. Heparanase and protease that affect anchoring of tumor cells and subsets of integrins (αvβ3 and αvβ5) targeted by cilengitide may be amenable to therapy [36]. Upregulated MET oncogene directs invasiveness and constitutes a potential therapeutic target; the undesirable outcome of angiogenic therapy which showed an increase in metastasis can be overcome by muti-target tyrosine kinase inhibitors to inhibit the alternative metastatic pathways stimulated by anti-angiogenic therapy alone [42]. Reversion of EMT phenotype by manipulating expression of miRNA’s has been proposed [67] with research addressing natural compounds like cucuramin [68] or drugs targeting NF-kappaB or TGFβ-induced EMT. Several therapeutic drugs are currently under investigation, targeting signaling pathways activated in EMT [38]. Circulating tumor cells (CTC) that show an EMT signature could form an early metastatic alert [69] or serve to detect minimal residual disease after therapy [70]. In an in vitro study on luminal breast cancer MCF-7 cell lines developing EMT expression induced by transcription factors, Lapitinib administration was shown to downregulate EMT expression [71]. Additionally, newer therapies may complement or supplement existing modalities improving patient outcome especially in poor-prognosis subsets of breast cancer. There remains the possibility of tumor ECM changing its composition after chemo-radiotherapy to produce a tumor-protective and therapy-resistant environment.

Conclusion

1. This study’s significance lies in its application to clinical biopsies where surrogate markers of EMT and CSC would find use as predictors and prognosticators.

   1. There was a strong association of metastasis with younger women ≤40 years, lymphovascular invasion (which however is difficult to identify in small biopsies), and EMT expression. CSC expression approached significance and its presence maybe specifically predictive of distant metastasis.
   2. This study adds to the determination of an EMT signature that may signify sub-sets of breast cancer with a higher metastatic potential. Definitive markers useful for identification of EMT include Vimentin expression in tumor cells and fibronectin/MMP 11 in peritumoral stroma.
   3. CSC can be recognised by the CD44⁺/CD24^−/low phenotype.
2. EMT and CSC showed the following associations

   1. EMT has a strong association with Ki67 PI
   2. CSC has a strong association with HER2-like cancers.
   3. Apocrine carcinoma with dual EMT/CSC phenotype deserves further study with larger number of cases of this sub-type.
   4. Basal-like tumors known for their aggressive behavior frequently show EMT.

3. This study has identified subsets of breast cancer in which EMT/CSC expression is seen in relation to the conventional morpho-immunophenotypic classification used in diagnostic practice, for its predictive and prognostic value. It has suggested EMT expression as a surrogate marker for metastasis-prediction. Future work is planned to include a larger sample size of the identified subsets with EMT/CSC expression viz. young women with breast cancer, Her-2 positive cancers and those with the basal phenotype to validate these initial observations. To accrue the numbers for a good sample size it is planned to source the biopsies from multiple centers in the population under study. Further, in these studies correlating cohort expression of EMT/CSC markers to patient outcome may provide validation or otherwise. This was not the intended outcome or within the scope of the current study.