Clinical & Experimental Metastasis

, Volume 25, Issue 6, pp 629–642

Epithelial mesenchymal transition traits in human breast cancer cell lines

Authors

  • T. Blick
    • VBCRC Invasion and Metastasis UnitSt. Vincent’s Institute
  • E. Widodo
    • University of Melbourne, Department of Surgery, St. Vincent’s Hospital
    • Faculty of MedicineBrawijaya University
  • H. Hugo
    • Embryology LaboratoryMurdoch Children’s Research Institute, The Royal Children’s Hospital
  • M. Waltham
    • VBCRC Invasion and Metastasis UnitSt. Vincent’s Institute
  • M. E. Lenburg
    • Department of Genetics and GenomicsBoston University School of Medicine
    • Life Sciences DivisionLawrence Berkeley National Laboratory
  • R. M. Neve
    • Life Sciences DivisionLawrence Berkeley National Laboratory
    • VBCRC Invasion and Metastasis UnitSt. Vincent’s Institute
    • University of Melbourne, Department of Surgery, St. Vincent’s Hospital
Research Paper

DOI: 10.1007/s10585-008-9170-6

Cite this article as:
Blick, T., Widodo, E., Hugo, H. et al. Clin Exp Metastasis (2008) 25: 629. doi:10.1007/s10585-008-9170-6
  • 1.7k Views

Abstract

Epithelial mesenchymal transition (EMT) has long been associated with breast cancer cell invasiveness and evidence of EMT processes in clinical samples is growing rapidly. Genome-wide transcriptional profiling of increasingly larger numbers of human breast cancer (HBC) cell lines have confirmed the existence of a subgroup of cell lines (termed Basal B/Mesenchymal) with enhanced invasive properties and a predominantly mesenchymal gene expression signature, distinct from subgroups with predominantly luminal (termed Luminal) or mixed basal/luminal (termed Basal A) features (Neve et al Cancer Cell 2006). Studies providing molecular and cellular analyses of EMT features in these cell lines are summarised, and the expression levels of EMT-associated factors in these cell lines are analysed. Recent clinical studies supporting the presence of EMT-like changes in vivo are summarised. Human breast cancer cell lines with mesenchymal properties continue to hold out the promise of directing us towards key mechanisms at play in the metastatic dissemination of breast cancer.

Keywords

EMTBasalLuminalMesenchymalBreast cancerBreast cancer stem cells

Abbreviations

EMT

Epithelial mesenchymal transition

EGF

Epidermal growth factor

HBC

Human breast cancer

IGF-IR

Type I insulin-like growth factor receptor

MET

Mesenchymal epithelial transition

TNFα

Tumor necrosis factor alpha

The Epithelial Mesenchymal Transition (EMT) occurs during development to generate the primary mesenchyme, and then subsequently from the ectoderm in multiple scenarios resulting in muscle, bone, nerve and connective tissues [1, 2]. In many cases, the EMT occurs transiently, and is followed by the reverse transition (MET) at the destination, to result in structures such as the segmental plates, kidney, GI tract, lung and skin (reviewed in [1]). During EMT, otherwise sessile epithelial cells organised into a collective unit lose characteristic cell junctional proteins (such as epithelial cadherins/adherens junction proteins, tight junction and desmosomal proteins, and lateral integrins) and cytoskeletal elements (such as cytokeratins), and gain motility (reflected in altered actin organization from cortical bundles into stress fibres), mesenchymal cadherins (such as N-Cadherin and Cadherin-11), and vimentin-rich intermediate filaments [3]. Profiling studies have increasingly defined the molecular changes associated with EMT [4, 5].

Parallels between the developmental EMT and the metastatic process in malignancy have been proposed for many years [6, 7], and continue to flourish with an exponential increase in examples from many different carcinoma systems (reviewed in [5]). Traditionally rationalized on the basis of translocation during metastasis, additional aspects of EMT biology, such as resistance to anoikis [8], enhanced survival [9], genomic instability [10] and resistance to chemotherapies [11] have more recently emerged in relation to malignancy. Difficulties in definitively proving EMT in clinical subjects hindered early acceptance of EMT, leading to a stimulating debate [1214], although much evidence has accrued in recent times. This picture is complicated further by the emerging likelihood that the ultimate metastatic colonization requires a reverse transition (MET), such that evidence of a lack of EMT in metastases may be misleading [15, 16]. Also, current concepts suggest that a hybrid state coined metastable phenotype is seen after carcinoma EMT, rather than full mesenchymal conversion [3, 17].

The mesenchymal proteome is rich in extracellular matrix proteins and associated factors, such as collagens, laminins, fibronectin, SPARC, proteoglycans and their receptors. Accordingly, EMT has been implicated in several fibroses, including diabetic nephropathy [18, 19], hepatic [20] and cardiac [21]. In addition to EMT, endothelial–mesenchymal transition has emerged in vascular endothelial cells [22]. A number of factors which transcriptionally repress E-Cadherin have emerged as potent EMT drivers during development and cancer. These include the zinc finger Snail homologs (Snail1, Snail2/Slug, Snail3), and several basic helix-loop-helix- factors such as Twist, ZEB1/, ZEB2/SIP1 and TCF3/E47/E12 (reviewed in [23]).

Indications of EMT in breast cancer cell lines

Evidence of EMT in a breast cancer setting came from observations that invasiveness of human breast cancer cell lines in vitro, and metastatic potential in vivo, correlated better with the expression of the mesenchymal intermediate filament protein vimentin than with lack of oestrogen receptor ([24]; see also Table 1 for summary of this section). These cells lines were further found to exhibit reduced cytokeratin levels, and reduced or absent components of the various cell:cell adhesion complexes such as E-cadherin, desmoplakin and ZO-1 [2427]. N-cadherin expression was often found to be selectively expressed in cell lines showing reduced or absent E-cadherin [28], and N-cadherin transfection promoted breast cancer cell invasiveness independent of whether it caused loss of E-cadherin or not [28, 29]. Vimentin itself was thought initially to serve as a marker of the mesenchymal state, however recent data suggests that it also plays a functional role in the mesenchymal phenotype (reviewed in [30], see also [3133]).
Table 1

Summary of human breast cancer cell line characteristics with respect to EMT status, subgrouping, invasiveness and Matrigel morphology

Cell Line

Thompson

Nieman 1999 (4)

Zajchowski

2001 (5)

Perou et al. 2000 (6)

Charafe-Jauffret 2006 (8)

Lombaerts 2006 (9)

Neve 2006 (10)

Invasiveness (study number ref)

Matrigel morphology (study number ref)

Year (ref)

1992 (1)

Ross and Perou 2001 (7)

Kenny 2007 (11)

Year (ref)

1994 (2)

 

Year (ref)

1994 (3)

 

Delimiter

VIM

N-Cad

Score

Subtype

Subtype

Subtype

Subtype

 

 

184

VIM+ (3)

 

 

B (6)

 

 

 

Low (3)

Sph (3)

184-A1/Aa

VIM low (3)

 

 

B (6,7)

 

 

 

Med (3)

Sph (3,5)

184-B5

VIM− (3)

 

Low

B (6,7)

B

 

 

Low (3,5)

Sph (3)

48RS

 

 

Low

 

 

 

 

Low (5)

Sph (5)

600MPE

 

 

 

 

 

 

Lu (10)

 

 

A1N4/T/H/M

VIM+ (3)

 

 

 

 

 

 

Med (3)

Sph (3)

A1N4-TH/MH

VIM+ (3)

 

 

 

 

 

 

High (3)

Stellate (3)

AU565

 

 

 

 

 

 

Lu (10,11)

Low (10)

Grape (11)

BrCa-MZ-01

 

 

 

 

not assnd

 

 

 

 

BRF71T1

 

 

High

 

 

 

 

Med (5)

Stellate (5)

BT-20

 

N-Cad−

 

 

B

 

BaA (10)

Low (4)

 

BT-474

VIM− (2)

 

 

Lu (6,7)

Lu

Ep

Lu (10,11)

Low (2,10)

Sph (2,11)

BT-483

VIM− (2)

 

 

 

Lu

Ep

Lu (10,11)

Low (2)

Sph (2,11)

BT-549

VIM+ (1,2)

N-Cad+

High

M (6,7)

 

Fib

BaB (10,11)

High (1,2,10)

Stellate (1,2,11)

CAMA1

VIM− (2)

 

 

 

Lu

Ep

Lu (10,11)

Low (2,10)

Grape (2,11)

*DU-4475

 

 

 

 

 

Ep

Not assnd

 

 

HB2

 

 

 

Lu (7)

 

 

 

 

 

*HBL-100

 

 

High

 

 

Fib

BaB (10)

Med (5)/High (10)

Stellate (5)

HCC1007

 

 

 

 

 

 

Lu (10)

 

 

HCC1008

 

 

 

 

 

 

Not assnd

 

 

HCC1143

 

 

 

 

 

 

BaA (10)

Low (10)

 

HCC1187

 

 

 

 

 

 

BaA (10)

 

 

HCC1428

 

 

 

 

 

 

Lu (10)

 

 

HCC1500

 

 

 

 

Lu

 

BaB (10,11)

High (10)

Sph (11)

HCC1569

 

 

 

 

 

 

BaA (10,11)

Low (10)

Sph (11)

HCC1599

 

 

 

 

 

 

Not assnd

 

 

HCC1937

 

 

 

Lu (7)

B

 

BaA (10)

 

 

HCC1954

 

 

 

 

not assnd

 

BaA (10)

Med (10)

 

HCC202

 

 

 

 

 

 

Lu (10)

Low (10)

 

HCC2157

 

 

 

 

 

 

BaA (10)

 

 

HCC2185

 

 

 

 

 

 

Lu (10)

 

 

HCC3153

 

 

 

 

 

 

BaA (10)

 

 

HCC38

 

 

 

 

B

 

BaB (10)

 

 

HCC70

 

 

 

 

 

 

BaA (10,11)

 

Sph (11)

HME-1

 

 

 

 

B

 

 

 

HME31

 

 

 

B (7)

 

 

 

 

 

HMEC

 

 

 

B (6,7)

 

 

 

 

 

HMT3522-S1

 

 

 

 

 

 

BaB (11)

 

Sph (11)

HMT3522-T4-2

 

 

 

 

 

 

BaB (11)

 

Sph (11)

HS578T

VIM+ (1,2)

N-Cad+

High

M (6,7)

M

Fib

BaB (10,11)

High (1,2)/Med (5,10)

Stellate (1,2,11)

LY2

 

 

 

 

 

 

Lu (10)

 

 

MCF10A/F

 

 

Low

M (7)

B

Fib (10A/F)

BaB (10)

Low (5,10)

Sph (5)

MCF12A

 

 

 

B (7)

 

Fib

BaB (10,11)

Low (10)

Sph (11)

MCF-7

VIM− (1,2)

N-Cad−

Med

Lu (6,7)

Lu

Ep

Lu (10,11)

Low (1,2,4,5,10)

Sph (1,2,5,11)

*MCF-7-ADR

VIM+ (1,2)

 

 

 

 

 

 

Med (1,2)

Sph/Stellate (1,2)

MDA-MB-134

VIM− (2)

 

 

 

Lu

Ep

Lu (10)

Low (2)

Grape (2)

MDA-MB-157

 

 

 

 

M

 

BaB (10)

Med (10)

 

MDA-MB-175/VII

VIM− (2)

 

 

 

Lu

Ep

Lu (10)

Low (2)

Sph (2)

MDA-MB-231

VIM+ (1,2)

N-Cad−

High

M (6,7)

M

Fib

BaB (10,11)

High (1,2,5,10)

Stellate (1,2,5,11)

MDA-MB-330

 

 

 

 

 

Ep

 

 

 

MDA-MB-361

VIM− (2)

 

Weak

 

 

Ep

Lu (10,11)

Low (2,10)

Sph (2)/Grape (11)

MDA-MB-415

 

 

 

 

 

 

Lu (10,11)

Med (10)

Sph (11)

*MDA-MB-435

VIM+ (1,2)

N-Cad+

 

 

 

Fib

BaB (10)

Med (1,2)/High (4)

Stellate (1,2)

MDA-MB-435S

 

 

High

 

 

 

 

High (5)

 

MDA-MB-436

VIM+ (1,2)

N-Cad+

 

 

 

 

BaB (10,11)

Med (1,2)

Stellate (1,2,11)

MDA-MB-453

VIM− (2)

N-Cad−

Weak

 

Lu

Ep

Lu (10,11)

Low (2,10)

Grape (2,11)

MDA-MB-468

VIM− (1,2)

 

Med

 

 

 

BaA (10,11)

Low (1,2)/Med (5)

Sph (1,2)/Grape (11)

MPE600

 

 

 

 

 

Ep

Lu (11)

 

Sph (11)

OCUB-F

 

 

 

 

 

Ep

 

 

 

PMC42-ET

VIM+

N-Cad+

 

 

 

 

 

 

Sph

PMC42-LA

VIM low

N-Cad+

 

 

 

 

 

 

Grape

S68

 

 

 

 

Lu

 

 

 

 

SKBR3

VIM− (1,2)

N-Cad−

Med

Lu (6,7)

Lu

Ep

Lu (10,11)

Low (1,2,4,5,10)

Grape (1,2,5,10)

SKBR5

 

 

 

 

 

Ep

 

 

 

SKBB7

 

 

 

 

M

 

 

 

 

SUM102PT

 

 

Low

 

 

 

 

Low (5)

Sph (5)

SUM1315/mo2

 

N-Cad−

High (m02)

 

 

 

BaB (10)

Low (4)/Med (5)

Grape (5 (mo2))

SUM149PT

 

N-Cad−

 

 

B

 

BaB (10)

Low (4)/High (10)

 

SUM159PT

 

N-Cad+

High

 

 

 

BaB (10)

High (4,10)/Med (5)

Stellate (5)

SUM185PE

 

 

 

 

not assnd

Ep

Lu (10)

Low (10)

 

SUM190PT

 

 

 

 

not assnd

 

BaA (10)

 

 

SUM225CWN

 

 

 

 

B

 

BaA (10)

Low (10)

 

SUM44PE

 

 

Med

 

 

Ep

Lu (10)

Low (5)

Sph/Grape (5)

SUM52PE

 

 

Med

 

Lu

 

Lu (10)

Med (5)/Low (10)

Sph (5)

SW872

 

 

 

M (6)

 

 

 

 

 

T47D

VIM− (1,2)

 

Weak

Lu (6,7)

Lu

Ep

Lu (10,11)

Low (1,2)/Med (10)

Sph (1,2,11)

UACC812

 

 

 

 

Lu

 

Lu (10,11)

Low (10)

Grape (11)

ZR75-1

VIM− (1)

 

Med

 

Lu

Ep

Lu (10,11)

Low (1,5)/Med (10)

Sph (1)/Grape (11)

ZR75-30

 

 

 

 

Lu

 

Lu (10)

Low (10)

 

ZR75-B

VIM− (2)

 

 

 

 

 

Lu (10,11)

Low (2,10)

Sph (2)/Grape (11)

The results from each study are summarised with respect to the ‘delimiters’ as listed. Studies summarised include (1) Thompson et al. [24]; (2) Sommers et al. [27]; (3) Thompson et al. [40]; (4) Nieman et al. [28]; (5) Zajchowski et al. [58]; (6) Perou et al. [91]; (7) Ross and Perou [92]; (8) Charafe-Jauffret et al. [61]; (9) Lombaerts et al. [60]; (10) Neve et al. [62] and (11) Kenny et al. [93]. Studies with similar delimiters have been grouped together (i.e. 1–3, 6&7, 10&11). Invasiveness and Matrigel morphology are recorded from each study in which they were measured, with designations to the studies as numbered in the top row. Invasiveness levels have been estimated based on data in each paper. Matrigel morphologies such as spheroidal, fused, pseudo-acinar, round and mass have been grouped together as ‘Spheroidal (Sph)’. Spherical and grape-like are grouped as ‘Grape’. Branched and stellate are grouped as ‘Stellate’. Designation for subgroups are Luminal (Lu), Basal A (BaA), Basal B (BaB), Normal/Basal (B), Mesenchymal (M) or Fibroblastic (fib). VIM = Vimentin, N-Cad = N-Cadherin. ‘Not assnd” = not assigned in the study but listed there. Italics font = invasive/stellate. Stippled means reclassified as non breast-derived. DU-4475 are a colorectal carcinoma cell line based on expression profiling clusters and the presence of an APC mutation (personal communication, Mark Lackner, Genentech), HBL-100 carry the Y Chromosome [95], MCF-7-ADR are not MCF-7-derived [95], and MDA-MB-435 are M14 melanoma cells (see text, [95]). PMC42 parental cells (ET) and the LA sublines are added to the table (bold), however the have not yet been formally compared to other breast cancer cell lines for invasiveness. Vimentin expression is described in Ackland et al. [50] and Hugo et al. [94], and Matrigel morphology is from unpublished data

Several systems have emerged in which EMT can be induced in human mammary epithelial cells and/or human breast cancer cell lines (also reviewed in [30, 34]). The MCF10A cell line, derived by spontaneous transformation of cells isolated after reduction mammoplasty, appears poised to undergo EMT in response to a variety of factors and through a variety of mechanisms. Combined expression of H-ras and erbB2 caused a mesenchymal phenotype in 3D Matrigel cultures, although EMT markers were not examined [35]. MCF10A cells undergo EMT in response to monolayer wounding, and this is enhanced by epidermal growth factor (EGF) [26]. Overexpression of the type I insulin-like growth factor receptor (IGF-IR) also causes EMT in MCF10A cells, characterised by Snail-mediated downregulation of E-cadherin, and high levels of NF-κB [36]. Introduction of the constitutively active p65 subunit of NF-κB into MCF10A cells caused EMT associated with ZEB-1, and MCF10A cells chronically exposed to tumour necrosis factor alpha (TNFα), a potent NF-κB inducer, also exhibited the EMT-like phenotype and ZEB-1/ZEB-2 induction [37]. The retinoblastoma suppressor-associated protein 46 (RbAp46/RBBP7), which is a component of the histone-modifying and -remodelling complexes, also caused EMT in these cells [38].

Human mammary cell lines derived from normal tissue share a somewhat “basal” phenotype (see below and [39]), and appear as a group to be prone to EMT. Despite the most normal morphology in 3D Matrigel culture, they share a transcriptome characteristic of the most stellate cell lines – Basal B [93]. Co-expression of vimentin and E-cadherin is seen in the 184 cells, A1N4–A1 and A1N4-B5 transformed derivatives, and single oncogene transfectants, however double oncogene transfectants lose their E-cadherin and become stellate in Matrigel [40]. TGF-β stimulated EMT in human mammary epithelial cells was recently shown to involve an interplay between αvβ3 integrin and Src [41]. It is not only such “basal” models which can undergo EMT. MCF-7 cells are a well accepted model of ER-positive, “luminal” type breast cancer that have been shown recently to undergo EMT changes in response to oestrogen [42], and the cell adhesion molecule L1 [43].

The human breast PMC42 cell line, when first established, was reported to show stem-like capacity, producing 8 morphological subtypes in culture after cloning. It showed oestrogen and progesterone responsivity [4446] and expression of predominantly luminal cytokeratins 8 and 18 (reviewed in [47]). The PMC42-LA is an especially epithelial subline variant which develops acini-like structures in 3-dimensional Matrigel cultures, and these produce milk proteins in response to lactogenic hormones [48]. These cells are somewhat pluripotent, and can elaborate myoepithelial markers in peripheral cells when grown as 3-dimensional clusters [49]. Stimulation of PMC42-LA cells with EGF leads to EMT marker expression both in 2D monolayer culture [50] and 3D collagen cultures (unpublished data), and 3D Matrigel cultures of PMC42-LA show increased expression of these markers when treated with factors selectively secreted by carcinoma-associated fibroblasts over normal mammary fibroblasts [51]. The parental PMC42 cells (PMC42-ET) [4446] are 100% VIM-positive (Fig. 1) and also respond to EGF with increased VIM expression and a further reduction in their already low E-cadherin levels (not shown). Thus, the PMC42 system provides a spectrum of EMT progression stages, and could provide important leads into the identification of markers which either indicate EMT and/or a propensity for breast cancer cells to undergo an EMT. One observation in the PMC42 system is that the end result from the EGF-induced EMT is not fully mesenchymal [50], and there are a growing number of such observations of this so-called ‘metastable state’ in various EMT/MET systems in vitro and in vivo [17, 5255].
https://static-content.springer.com/image/art%3A10.1007%2Fs10585-008-9170-6/MediaObjects/10585_2008_9170_Fig1_HTML.gif
Fig. 1

Immunocytochemical staining of E-Cadherin and Vimentin in the parental PMC42-ET cells and the epithelial subline PMC42-LA, using methodology as described [50]. Note the relative absence of membranous staining of E-Cadherin in the parental cells, and the vimentin-positive subpopulation of vimentin expression (∼10–15%) in the LA subline

The advent of gene array technology has allowed increasingly larger analyses of human breast cancer cell lines (summarised also in Table 1). Gene array analysis of the NCI60 cell lines showed strong divergence between the more epithelial MCF-7 and T47D cells and invasive lines such as Hs578T and BT-549 [56]. MDA-MB-435 clustered into the melanoma group and ultimately was proven to be the M14 melanoma cell line [57]. Studies by Zajchowski confirmed an association between mesenchymal status and invasiveness [58]. Mesenchymal gene products dominated a 24 gene signature predicting invasiveness of human breast cancer cell lines, while high levels of cytokeratins 18 and 19 and plakoglobin, amongst other epithelial markers, predicted non-invasiveness. In compiling a comprehensive review on human breast cancer cell lines, Lacroix and Leclerq also summarised gene array data partitioning invasive/mesenchymal-like/Basal breast cancer cell lines from those with Luminal/epithelial properties [59]. Lombaerts et al. [60] showed that while cell lines in which E-Cadherin had been silenced through promoter methylation showed a mesenchymal phenotype, this was not seen when E-Cadherin was absent due to sporadic mutation, such that EMT arose through an orchestration of coordinated gene regulation.

Such studies have recently been expanded further, with good concordance and increased robustness (also summarised in Table 1). Charafe-Jaufrett et al. interrogated 31 breast cancer cell lines with Affymetrix U133 Plus 2.0 arrays and found that the cells clustered into Luminal, Basal and Mesenchymal subgroups [61]. Neve et al. profiled 51 human cell lines with Affymetrix HG-U133A chips, and also integrated CGH and proteome analysis [62]. Each of these provides an exceptional database of gene expression amongst cell lines, and each identified a subset of mesenchymal-like cell lines. Cell lines clustering to Basal B (as distinct from Basal A or Luminal) in the Neve study correspond well to the group designated by Charafe-Jaufrett et al. as Mesenchymal (compared to Luminal or Basal). Exceptions were HCC38, MCF10A, and SUM149PT which were classified as Basal (i.e. Basal A), and HCC1500 which was classified as Luminal. Further genotype analysis of the cell lines by Neve and coworkers found the cell line termed HCC1500 was in fact not that cell line, explaining the inconsistency with the Charafe-Jauffret et al [61]. Basal B/Mesenchymal cells were shown in both studies to resemble ‘basal/myoepithelial’ cells in the expression of mesenchymal gene products, but lacked a number of cytokeratins seen in myoepithelial cells. Consistencies amongst these data engender a degree of confidence in the reproducibility of these cell lines and the robustness and accuracy of gene array technology. We have explored further the Neve dataset [62] for aspects of EMT, as described below.

EMT markers in HBC cell lines

We examined expression of the most widely-used markers for EMT (Vimentin, E-Cadherin, N-Cadherin, fibronectin) across the cell lines (Fig. 2). Vimentin is absent in the majority of lines in the luminal cluster and clearly overexpressed in the Basal B, but also quite widely expressed also in Basal A cell lines which exhibit features of both Basal and Luminal tumours. E-Cadherin, on the other hand, is an important discriminator between Basal A (which all express levels close to the median) and Basal B which have reduced expression of it in most lines. Surprisingly, 8 of the 25 luminal cell lines also show reduced expression of E-Cadherin. N-Cadherin and fibronectin, classically used to monitor EMT, each showed relative enrichment in the Basal B subgroup, compared with the other subgroups, although fibronectin expression was relatively high in several Luminal lines. Several N-cadherin-negative lines have clustered as mesenchymal, most notably MDA-MB-231, but also SUM149 and SUM1315. These SUM cell lines appear poorly invasive in early studies [28], but more recently have been reported to be highly invasive in Matrigel assays [58, 62]. Further analysis of additional markers closely defining the Basal B subgroup may reveal important novel mediators of EMT in the breast cancer context. Examination of these in the PMC42 system, and other breast cancer EMT systems may provide support for their potential role in EMT, which then requires further examination at the functional level.
https://static-content.springer.com/image/art%3A10.1007%2Fs10585-008-9170-6/MediaObjects/10585_2008_9170_Fig2_HTML.gif
Fig. 2

Gene expression levels for common EMT markers in 51 breast cancer cell lines. Median centred mRNA expression levels for (a) Vimentin, (b) E-Cadherin, (c) N-Cadherin and (d) fibronectin are shown on a log2 scale. The 51 cell lines are organised by subclass defined in Neve et al. [62]. For genes represented by multiple probesets on the arrays, the probeset with the greatest standard deviation across samples was selected

Classical EMT drivers in HBC cell lines

Considerable literature is emerging on the relative expression of EMT drivers mentioned above in model systems and clinical specimens, on interrelationships among these factors, and on their potency when added ectopically. Since each of these is represented on the Affymetrix U133A chips, we assessed their relative expression across the breast cancer cell lines in the Neve data [62]. A reasonable concordance between expression of these EMT drivers in the Basal B/mesenchymal cell lines was seen on the whole (Fig. 3). Snail1 shows very little differential across the subgroups. Snail2 shows overexpression in quite a few Basal A lines as well as most cell lines in the Basal B group. Twist was expressed in most Basal B lines, but surprisingly, it was also expressed in several Luminal lines. The transcription factor found upregulated in Basal B with the highest discriminator score was ZEB1/TCF8. High relative expression of ZEB1 was limited to Basal B lines, with most Basal B cell lines expressing it. Although relative expression of TCF3 did not discriminate between the cell line clusters, it has previously been shown to interact with Twist [63], TCF4 [64] and the ETS-domain protein Elk-3 [65]. Higher relative expression of TCF4 is almost exclusive to Basal B and most members of Basal B express it. All members of Basal B have a high relative expression of ELK3, as do a few members of Basal A, but no Luminal cell lines have a high level of ELK3 expression. ELF3, also an ETS-domain protein, is markedly underexpressed by most Basal B cell lines, and almost exclusively so. Several of these factors have each been implicated in various breast cancer EMT systems, resulting in a rather complicated scenario, as described in Table 2. Recently the high mobility group A2 protein (HMGA2) induced by the Smad pathway during TGFβ-induced EMT was shown to integrate transcriptional input for the expression of Snail1, Snail2, Twist and inhibitor of differentiation 2 (ID2), each of which has been implicated in EMT. The complex interplay seen amongst EMT-driving transcription factors may be mediated in part by HMGA2.
https://static-content.springer.com/image/art%3A10.1007%2Fs10585-008-9170-6/MediaObjects/10585_2008_9170_Fig3_HTML.gif
Fig. 3

Gene expression levels for known EMT transcriptional drivers in 51 breast cancer cell lines. Median centred mRNA expression levels for (a) Snail1, (b) Snail2, (c) Twist, (d) ZEB1, (e) TCF3, (f) TCF4, (g) ELK3 and (h) ELF3 are shown on a log2 scale. The 51 cell lines are organised by subclass defined in Neve et al. [62]. For genes represented by multiple probesets on the arrays, the probeset with the greatest standard deviation across samples was selected

Table 2

Expression of various E-cadherin repressor genes and their correlation with gene products which are typically up or down in invasive breast cancer (e.g. up: vimentin, Her-2, aromatase, down: E-cadherin, Na/K-ATPase β1). All lines listed except for JIMT-1 have been examined further in this study

E-Cad Repressor

Cell Line(s)

Correlation

References

Zeb2, Snail1

MDA-MB-231,

Inverse correlation between Zeb2 and E-cad, stronger inverse correlation with Zeb2 and E-cad than Snail1 and E-cad eg. MDA-MB-231 do not express Snail1, however do express Zeb2 and are E-cad negative

Comijn et al. (2001) [96]

MDA-MB-435S,

MCF7/AZ,

HBL100

Zeb2

MDA-MB-231,

Positive correlation between Zeb2 and VIM in MDA-MB-231, BT549, HS578T

Bindels et al. (2006) [97]

BT549, HS578T

Snail1

MCF7, MDA435

Inverse correlation with Snail1 and E-cad, positive correlation with E-cad and regulator of ATP1B1. Snail directly represses this molecule

Espineda et al. (2004) [98]

Snail1

SK-BR-3, MCF7,

Snail1 expressed specifically in normal and stromal breast cell lines, not in cancer cell lines. Expression is inverse with aromatase, an estrogen synthetase which promotes breast cancer growth

Okubo et al. (2001) [99]

MDA-MB-231,

MCF10A, HBL100

Snail2

Her-2 pos line

Snail2 not Snail1 expressed. Her-2 is also amplified in MDA361 and SKBR3, likely to show a similar profile as JIMT-1

Rennstam et al. (2007) [100]

JIMT-1

Zeb1/dEF1

MDA-MB-231

Inverse correlation between Zeb1 and E-cad: Zeb1 knockdown restored expression of E-cad and other epithelial cell junction molecules, however Snail1 knockdown did not

Aigner et al. [15]

Various

27 breast cell lines

In "fibroblastic" or breast tumor cluster, Snail2 and Zeb2 were upregulated and responsible for E-cad downregulation whereas Twist and Snail1 were not. Zeb1 expression was not examined

Lombaerts et al. [60]

Relationships to clinical breast cancer

Mesenchymal derivatives of carcinoma cells show a number of attributes which would favour metastasis, such as survival after separation from the collective as individual cells, increased migratory and invasive potential, increased survival in suspension and resistance to apoptosis in response to hormone ablation and chemotherapy. Sustained expression of mesenchymal traits would assist in extravasation at the secondary site, and possibly also survival at the secondary site. Evidence of EMT in clinical samples has been somewhat slow to accrue, leading to ongoing controversy as to its existence and relevance [1214]. A number of factors have been advanced to explain the relative paucity of clinical data, including (i) the possibility of an incomplete EMT, resulting in a hybrid state; (ii) the likelihood that EMT in carcinoma would generate small numbers of cells at the tumour edge which would immediately divorce themselves from the tumour collective, (iii) the production of EMT molecules in other scenarios, such as Snail family members in relation to cell survival, vimentin in myoepithelial cells, EMT markers such as vimentin and N-cadherin in the tumour parenchyma of certain breast cancer subsets, and (iv) the apparent requirement for disseminated carcinoma cells to “re-epithelialise” in order to generate a viable metastasis [14, 17].

In the past year or so, however, considerable support for carcinoma EMT has emerged in reports of EMT-related factors in provocative locations, building on the observation of EMT in colon carcinoma cells at the invasive front [66]. Vimentin expression in the tumour parenchyma has long been associated with poorer prognostic features, particularly in breast cancer (reviewed in [67]), however vimentin expression is not itself proof of EMT, and a recent comprehensive study suggested that vimentin-expressing tumours may derive from breast progenitor cells with a bilinear (glandular, myoepithelial) differentiation potential [68]. However, a wide-scale analysis of bladder cancer in 572 patients showed that E-cadherin, beta-catenin, plakoglobin, and vimentin, but not N-cadherin, were associated with grade and stage, while plakoglobin only was associated with lymph node status. Low levels of both plakoglobin (P = 0.02) and beta-catenin (P = 0.02) significantly associated with survival. Although multivariate analysis showed no significant influence of the EMT biomarkers on survival, alterations associated with plakoglobin were identified as significant prognostic features in these tumours.

More compelling, however, have been the recent reports of the transcriptional drivers of EMT in breast carcinoma. Snail1 expression has been found in infiltrating ductal carcinomas associated with lymph node metastases [69, 70] and distant metastases including effusions [55, 71], and has been associated with recurrence of experimental breast carcinomas [72]. Snail2 expression is also associated with tumour effusions, metastasis and recurrence [71, 73], but is also associated with partially differentiated breast cancers [55], reflective of the role of Snail2 in the developing breast [74]. Twist expression appears to be specific for ductal breast carcinoma where it is associated not only with the invasive state but also the development of carcinoma angiogenesis [75], and as such is a prognostic factor for poor clinical outcome [73]. Thus, traditional EMT markers are found in breast cancers in scenarios other than EMT, possibly in a survival role [9]. ZEB1 expression was stringently coupled to cancer cell dedifferentiation in invasive ductal and lobular breast cancer [15] and upregulated in invasive cancer cells at the tumour-host interface, where it was accompanied by downregulation of plakophilin-3 expression levels [76]. Affymetrix analysis of normal ductal and lobular cells, IDC cells and ILC cells microdissected from cryosections revealed that 7 differentially expressed genes which are involved in epithelial–mesenchymal transition, TGF-beta and Wnt signalling (CDH1, EMP1, DDR1, DVL1, KRT5, KRT6, KRT17) could distinguish IDC and ILC [77]. Recently, array profiling of EpCAM-immunopurified breast cancer cells in malignant effusions revealed a subset enriched for mesenchymal markers such as vimentin, S100A4, uPAR and CXCR4 [78]. Evidence of EMT has also been observed downstream of HOXB7 in bone-marrow derived (laser catapulted) breast cancer cells [79].

Other indications have come from other tumour types. ZEB1 expression was found at the tumour-host interface in colorectal cancer [15], and Snail2, TCF3, ZEB2 and Snail1 were examined in gastric carcinoma where Snail2 upregulation was found to be associated with E-cadherin downregulation in diffuse and intestinal-type gastric carcinoma, and this effect could be complemented by the presence of other EMT regulators [80]. EMT signatures have been associated with metastatic progression in melanoma [81], and with high risk in head and neck squamous cell carcinoma (HNSCC; [82]). In prostate cancer specimens, nuclear Twist was associated with metastatic potential, and high levels of Twist correlated with altered E-cadherin expression [83]. Gene expression profiles of microscopically dissected intratumoural samples from central and invasive regions of invasive papillary thyroid carcinoma identified reduced levels of mRNAs encoding proteins involved in cell–cell adhesion and communication, indicative of EMT [33]. In oesophageal squamous cell carcinoma, tumours with reduced E-cadherin or increased Snail1 expression invaded deeper and had more lymph node metastasis and more lymphatic invasion than their counterparts, although Snail1 expression was not significantly correlated with reduced E-cadherin expression [84]. Patients with reduced E-cadherin expression or positive Snail1 expression had poor clinical outcomes.

One scenario for the regulation of EMT emerged recently in studies of the activation of IκB kinase α (IKKα) by the Receptor Activator of NF-κB (RANK/TNFRSF11A). Increased EMT and metastasis occurred though downregulation of the tumour suppressor Maspin. Active nuclear IKKα in mouse and human prostate cancer was found to correlate with metastatic progression, reduced Maspin expression and infiltration of prostate tumours with RANK ligand—expressing inflammatory cells. [85]. NF-κB was also implicated in HNSCC [82], and mediates EMT effects caused by the principal EBV oncoprotein, latent membrane protein 1 (LMP1) in nasopharyngeal carcinoma cells. LMP1 induced EMT via Twist, expression of Twist and LMP1 was directly correlated, and expression of Twist was associated with metastasis clinically [86]. Overexpression of Twist was also correlated with hepatocellular carcinoma metastasis and its expression was negatively correlated with E-Cadherin expression (P = 0.001, r = −0.443) by tissue microarray [87].

Another clinical manifestation associated with EMT is drug resistance. Although the mesenchymal subgroup were found to be selectively sensitive to the dual specific (c-abl and c-src) kinase inhibitor Dasatinib [88], this is in stark contrast to studies showing that mesenchymal/EMT status in lung carcinoma cell lines correlates with resistance to EGFR-tyrosine kinase inhibitors [11, 89]. Accordingly, selection for drug resistance has been shown previously to correlate with EMT-like changes [25]. Despite early indications coming from the mis-identified MCF7ADR derivative, other selected resistant cells exhibit increased vimentin expression, and recent observations of a similar scenario were reported in colorectal carcinoma cells [90].

Conclusions

The growing number of cell lines, and the extent and robustness of the transcriptional analysis now available for breast cancer research, have dramatically increased our capacity to determine and trust phenotype/genotype relationships. We have summarised here some of the many conclusions that can be made in relation to EMT in these systems, and have demonstrated the use of the EGF-inducible EMT in the PMC42 system to begin to validate these. Some expected associations have been confirmed in the dataset (such as the transcription factors Snail2, Twist and Zeb1), while others have been found surprisingly lacking (such as Snail1). Systematic analysis has revealed novel transcription factor associations (such as TCF4) which more tightly define the Basal B/mesenchymal subgroup. Considerable evidence of EMT in clinical carcinoma has now been documented, with considerable implications for progression and outcome. Associations with the invasive front, with disseminated cells, and with resistance to different drugs should place EMT squarely in the sights of those wishing to control breast cancer, and other carcinomas. The work described here in human breast cancer cell lines highlights their utility in discovering and understanding the molecular processes underpinning breast cancer EMT.

Acknowledgements

This research was funded in part by the U.S. Army Medical Research and Materiel Command (DAMD17-03-1-0416) to EWT. TB and EWT are supported in part by the Victorian Breast Cancer Research Consortium. EW is the recipient of an AUS Aid Scholarship. Parts of this work were also supported by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research (Contract DE-AC03-76SF00098) and the California Breast Cancer Research Program (CBCRP) grant # 7FB-0027.

Copyright information

© Springer Science+Business Media B.V. 2008