Basal/HER2 breast carcinomas

Abstract

High rates of inherent primary resistance to the humanized monoclonal antibody trastuzumab (Herceptin) are frequent among HER2 gene-amplified breast carcinomas in both metastatic and adjuvant settings. The clinical efficacy of trastuzumab is highly correlated with its ability to specifically and efficiently target HER2-driven populations of breast cancer stem cells (CSCs). Intriguingly, many of the possible mechanisms by which cancer cells escape trastuzumab involve many of the same biomarkers that have been implicated in the biology of CS-like tumor-initiating cells. In the traditional, one-way hierarchy of CSCs in which all cancer cells descend from special self-renewing CSCs, HER2-positive CSCs can occur solely by self-renewal. Therefore, by targeting CSC self-renewal and resistance, trastuzumab is expected to induce tumor shrinkage and further reduce breast cancer recurrence rates when used alongside traditional therapies. In a new, alternate model, more differentiated non-stem cancer cells can revert to trastuzumab-refractory, CS-like cells via the activation of intrinsic or microenvironmental paths-to-stemness, such as the epithelial-to-mesenchymal transition (EMT). Alternatively, stochastic transitions of trastuzumab-responsive CSCs might also give rise to non-CSC cellular states that lack major attributes of CSCs and, therefore, can remain “hidden” from trastuzumab activity. Here, we hypothesize that a better understanding of the CSC/non-CSC social structure within HER2-overexpressing breast carcinomas is critical for trastuzumab-based treatment decisions in the clinic. First, we decipher the biological significance of CSC features and the EMT on the molecular effects and efficacy of trastuzumab in HER2-positive breast cancer cells. Second, we reinterpret the genetic heterogeneity that differentiates trastuzumab-responders from non-responders in terms of CSC cellular states. Finally, we propose that novel predictive approaches aimed at better forecasting early tumor responses to trastuzumab should identify biological determinants that causally underlie the intrinsic flexibility of HER2-positive CSCs to “enter” into or “exit” from trastuzumab-sensitive states. An accurate integration of CSC cellular states and EMT-related biomarkers with the currently available breast cancer molecular taxonomy may significantly improve our ability to make a priori decisions about whether patients belonging to HER2 subtypes differentially enriched with a “mesenchymal transition signature” (e.g., luminal/HER2 vs. basal/HER2) would distinctly benefit from trastuzumab-based therapy ab initio.

Keywords: :

• basal-like
• cancer stem cells
• EMT
• HER2
• trastuzumab
• breast cancer
• reprogramming

Because they cannot benefit from hormonal therapy with HER2-targeted drugs,¹ there is no established systemic therapy other than chemotherapy for patients affected by the basal-like subtype of breast cancer, which is the most impacted breast cancer disease among the intrinsic spectrum of tumor entities originally identified by applying breast cancer molecular taxonomy in 2000 (i.e., luminal A, luminal B, HER2-enriched, basal-like and a normal breast-like group).² Conversely, the HER2-enriched breast cancer subtype, which originally carried a poor prognosis in landmark studies of breast cancer genomics,^3,4 has largely benefited from the era of HER2-targeting therapies. In 2012, an institutional-based review confirmed that women with HER2-positive breast cancer disease who received the anti-HER2 monoclonal antibody trastuzumab (Herceptin™, Roche Biomedical Laboratories) had an improved prognosis compared with women with HER2-negative breast cancer disease.⁵ Indeed, there is no doubt that pivotal trials showing the clinical benefit of trastuzumab in combination with chemotherapy have led to a new standard of care for women with metastatic and early-stage HER2-positive breast cancer.^6-9 Based upon the high-risk assessment of recurrence among HER2-positive patients with small and node-negative breast carcinomas (pT1a,b,N0,M0), it has been suggested that prospective studies and randomized clinical trials should clarify whether women with node-negative, sub-centimeter, HER2-positive tumors could also benefit from trastuzumab-based adjuvant systemic therapy.^10,11 However, it should be noted that while the addition of trastuzumab to adjuvant chemotherapy can impressively reduce the recurrence rate by almost 50% and improve survival in patients with early-stage HER2-positive breast cancer, approximately 15% of women diagnosed with early HER2-positive disease show de novo (i.e., primary or intrinsic) resistance to trastuzumab and relapse in spite of treatment with trastuzumab-based therapies.^12-18 In the same regard, the overall response rate of women diagnosed with HER2-positive metastatic breast cancer and treated with trastuzumab as a single, first-line agent is only 26%. Thus, > 70% of HER2-overexpressing metastatic breast carcinomas show resistance to trastuzumab ab initio.

Breast cancer heterogeneity, in which biologically aggressive sub-clonal cell populations occur spontaneously within an individual tumor, has been suggested to underlie both the metastatic recurrence capacity and the variable response of HER2-positive tumors to selective HER2-targeting inhibitors, including trastuzumab.^19,20 Unfortunately, the precise genomic and biological features of HER2-positive breast cancer with intrinsic aggressive behavior and de novo non-sensitivity to trastuzumab remain largely obscure. Estrogen receptor (ER)-negative breast carcinomas are of either the basal-like or HER2-enriched subtype. ER-positive, clinically defined HER2-positive tumors are generally enriched with the luminal gene cluster, and rather than belonging to the HER2-enriched subtype, these tumors fall into the luminal (i.e., luminal B) subtype of breast carcinomas. Therefore, at least two types of HER2 gene-amplified breast carcinomas exist (ER-positive and ER-negative). Because several risk factors for developing basal-like breast cancer have been identified (e.g., BRCA1 mutations carriers and young African-American women), and there are no known specific risk factors (including no apparent interaction with race or age) for the HER2-enriched subtype,^1,21,22 it appears reasonable to etiologically consider basal-like and HER2-enriched breast carcinomas as two different breast cancer diseases.²³ Moreover, the basal-like breast cancer subtype displays an obvious and strong common biology, whereas the HER2-enriched subtype solely describes the occurrence of a high but imperfect correlation with HER2 gene-amplification status. That is, not all tumors of this molecular subtype exhibit HER2 gene amplification/HER2 protein overexpression, and some, but not all, clinically defined HER2-positive breast tumors fall into the HER2-enriched subtype. Although we should acknowledge that ~25% of basal-like breast carcinomas are not equivalent in clinical terms to immunohistochemically defined (ER, PR and HER2) triple-negative breast cancer,¹ it should be noted that up to 45% of basal-like breast carcinomas, which by rigorous definition depends on the assessment of the mRNA expression status of approximately 500 genes, are not triple-negative,^24-26 because they do express ER, PR and/or HER2. In this complex scenario, we hypothesize that the occurrence of HER2 gene amplification/HER2 protein overexpression within basal-like gene signatures might naturally delineate a novel breast cancer sub-entity that is inherently resistant to trastuzumab.^27,28

Gene Signatures in HER2-Positive Breast Cancer: The Association of a Poor Prognosis with HER2-Negative Basal-Like Tumors

Although an ever-growing, deeper understanding of the signal network integration of breast cancer cells is rapidly providing promise for genotype/phenotype-based “personalized breast cancer medicine,”^29,30 accurate tools to improve our current understanding of the molecular therapeutic behavior of individual HER2-positive breast carcinomas are lacking.^31,32 When employing currently available genomic signatures, patients overexpressing the HER2 oncogene are classified as high-risk individuals that should benefit, in terms of improved survival, from treatment with HER2-targeted agents, including trastuzumab.^33-35 However, currently available gene expression signatures cannot identify HER2-positive subgroups of good-trastuzumab responders, who will not experience relapse ab initio, or bad-trastuzumab responders, who will not respond to trastuzumab-based therapies ab initio. By correlating the genetic heterogeneity of HER2-positive tumors with clinical outcome by molecular profiling, Staaf and colleagues³⁶ provided strong support for the notion that variability of clinical outcome and lack of uniformity in the beneficial effect of trastuzumab in HER2-positive breast cancer patients may be due to the inherent heterogeneity of HER2-positive breast tumors. The identification of a 158-gene prognostic predictor, termed the HER2-derived prognostic predictor (HDPP), significantly improved the stratification of good and poor prognosis for both overall survival and distant metastasis-free survival. This improved stratification was realized not only in their HER2 tumor set, but also in multiple breast cancer data sets, including the Netherlands Cancer Institute data set, in which the HDPP gene signature demonstrated superior prognostic stratification of HER2-positive tumors compared with the Mammaprint³³ (Agendia) and Oncotype DX³⁴ (Genomic Health) systems.

Notably, the HDPP signature described by Staaf and colleagues³⁶ performed well not only in the HER2-enriched molecular subtype, but also in the basal-like subtype, and was associated with survival in triple-negative breast carcinomas. By contrast, HDPP showed no prognostic value in luminal (A and B) or normal-like breast cancer subtypes. In agreement with earlier studies, these findings may merely suggest that different gene expression modules or signatures must be applied to predict the outcome of different subgroups of breast tumors. Because the HER2-positive and basal-like breast cancer subtypes have been shown to be associated with the worst prognosis, and the luminal A subtype has been associated with the best prognosis, these findings may only reveal that the ability of the HDPP signature to predict survival is largely attributable to the differential and selective expression of HDPP signature in well-recognized subsets of aggressive breast carcinomas. Furthermore, because proliferation is an important component of most prognostic gene signatures and has been reported to be associated with outcome solely in ER-positive/HER2-negative breast cancer,^37-39 the ability of the HDPP signature to add independent prognostic information for ER-negative, lymph node-negative or high-grade clinically aggressive tumors strongly suggests that the recognition of a biological phenomenon beyond HER2 gene amplification and high proliferation status may be crucial to improve the prediction of outcome and the selection of patients for molecularly targeted therapies. The notion that HER2-positive breast tumors do not form a homogenous biological entity, but rather a clinical subset of breast cancer patients, is clearly supported by recent studies addressing the biological homogeneity of HER2-positive tumors at the genomic level.^40,41 HER2-positive breast carcinomas are known to display a great diversity of anomalies, which affect most of the genome. Moreover, amplification of the HER2 locus at 17q yields two different patterns that may be related to different breast cancer subsets. Narrow amplicons centered around the HER2 gene that comprise six to 10 genes have been identified in more aggressive ER-negative tumors, whereas large amplicons that start centromeric to HER2 and extend telomeric to the TOP2A gene correlate with ER-positive tumors and better survival.^40,41

The above-mentioned findings confirm and further extend recent studies that failed to delineate a HER2-positive breast cancer subgroup exclusively composed of HER2-amplified breast tumors; rather, HER2-positive tumors appear to cluster in other molecular subgroups. In other words, although the majority of HER2-positive breast carcinomas have HER2-enriched gene expression profiles, other intrinsic breast cancer subtypes have HER2-positive tumors within them, suggesting that the HER2-positive clinical category is biologically heterogeneous. In this extremely complex scenario, at least two challenging questions remain to be answered in the context of the inherent breast cancer response to HER2-targeted therapies. The first question refers to the possibility that patients who are clinically HER2-negative within the HER2-enriched breast cancer subtype could benefit from anti-HER2 therapies, such as trastuzumab. Pre-clinical HER2-negative breast cancer models engineered to exhibit a constitutive activation of HER2 by overexpressing the HER3 ligand heregulin preliminarily defined a specific HER2-negative breast cancer patient population for which trastuzumab plus chemotherapy may be extremely efficient.^42,43 Accordingly, de Alava and colleagues⁴⁴ confirmed that the spectrum of breast cancer patients who may benefit from trastuzumab-based therapies should be widened to include patients with metastatic breast tumors lacking HER2 gene amplification but expressing high levels of EGFR ligands, which leads to HER2 transactivation. In this regard, women with HER2-negative breast cancers who do not meet the established criteria for HER2-positive disease on central review have been found to receive as much clinical benefit from adjuvant trastuzumab as women with HER2 gene-amplified breast cancer.⁴⁵ The second question refers to the possibility that patients who are clinically HER2-positive within the basal-like breast cancer subtype might delineate a naturally occurring subgroup of breast cancer patients with de novo (primary) resistance to trastuzumab-based therapies. In this regard, it is tempting to suggest that the use of signatures, such as the above-mentioned HDPP, should be part of the decision-making process for the therapeutic management of HER2-positive breast cancer disease. However, as is the case with most breast cancer gene signatures, the performance of the HDPP signature is based on retrospective data from patients who did not receive trastuzumab therapy.⁴⁶ Therefore, the prognostic value of HDPP may not necessarily be predictive of responsiveness to trastuzumab-based therapies. Indeed, a subgroup with favorable prognosis, as identified by HDPP signature, was demonstrated to display high activation of PI3K, a well-recognized determinant of trastuzumab resistance.⁴⁷ Interestingly, when applied to a small HER2-positive data set of patients preoperatively treated with trastuzumab and vinorelbine, the HDPP signature correlated with the occurrence of trastuzumab resistance. Although the sample size was insufficient to form any conclusion on the value of HDPP for predicting the response to targeted treatment modalities (e.g., trastuzumab) in HER2-positive patients, it should be noted that when formerly analyzing this data set, Harris and colleagues⁴⁸ auspiciously suggested that “HER2-overexpressing tumors with a basal-like phenotype” were more likely to be intrinsically resistant to pre-operative trastuzumab.

Breast Cancer Stem Cells and HER2-Targeted Therapies: From Biology to Therapeutics

Integrating breast cancer molecular taxonomy and mammary-stem cell biology could be the best approach to understanding de novo resistance to HER2-targeted therapies and the increased intrinsic aggressiveness of specific subgroups of HER2-positive breast cancers.⁴⁹ The HER2 gene appears to play a crucial role in mammary carcinogenesis by actively driving the pool of replenishing cancer-initiating cells or cancer-propagating cells, popularly known as cancer stem cells (CSCs), which have the unique ability to both promote self-renewal and give rise to phenotypically heterogeneous breast cancer progeny.^50-53 HER2 overexpression was originally described to actively enhance the cancer-initiating component that drives tumorigenesis, invasion and metastasis by significantly increasing the fraction of breast cancer cells positive for ALDH1 (aldehyde dehydrogenase),⁵⁴ a marker of normal and cancerous human mammary epithelial cells with stem/progenitor properties.⁵⁵ Accordingly, a landmark study by Li and colleagues⁵⁶ demonstrated that in contrast to chemotherapy treatment, the epidermal growth factor receptor (EGFR)/HER2 dual tyrosine kinase inhibitor (TKI) lapatinib (Tykerb™) decreases the subpopulation of chemotherapy-resistant cancer-initiating cells, which was measured by calculating the percentage of cells with the CD44⁺/CD24^-/low stem/progenitor immunophenotype that exhibited the ability to form mammospheres in vivo as an indication of self-renewal. Magnifico and colleagues⁵⁷ have further expanded this notion by demonstrating that HER2-positive breast CSCs, which appear to express the highest levels of HER2 protein regardless of HER2 gene amplification status, are exquisitely sensitive to trastuzumab. This crucial finding (i.e., that the HER2 oncoprotein may be selectively upregulated in breast CSCs but not in bulk breast cancer cell populations) might explain the results of the above-mentioned retrospective studies, which demonstrated that in breast tumors currently classified as HER2-negative [i.e., immunohistochemistry (IHC) 1/2+ or fluorescence in situ hybridization (FISH) negative], HER2-targeting agents may be effective when administered in adjuvant but not advanced settings.^45,58

In breast tumors clinically classified as bona fide HER2-positive (i.e., IHC 3+ or FISH positive), HER2 is overexpressed in both breast cancer stem cells and bulk tumor populations. Therefore, these tumors are expected to be sensitive to HER2-targeting agents in both advanced and adjuvant settings to promote tumor regression and reduce recurrence, respectively.⁵⁸ However, while accumulating evidence indicates that molecularly targeted agents directed against HER2, such as trastuzumab, could effectively target cancer-initiating cells in HER2-positive breast carcinomas, counterintuitively, recent observations have also suggested that clinical resistance to trastuzumab and other HER2-targeted therapies may be driven by breast CSCs. We already recognize that CS-like cells isolated from breast carcinomas intrinsically possess high resistance to hormonal therapy, chemotherapy and radiotherapy.^56,59-61 Indeed, CS-like cells are highly prevalent at the invasive edge of breast carcinomas following treatment with conventional therapies.^56,62,63 Correspondingly, it could be argued that overexpression of CSC-related biomarkers would promote resistance to trastuzumab by generally enhancing self-renewal/proliferation in progenitor cells and/or inducing stem cell properties, including the epithelial-to-mesenchymal transition (EMT) genetic program. Accordingly, trastuzumab treatment reduces the stem-cell population in trastuzumab-sensitive breast cancer cell lines but not in trastuzumab-unresponsive breast cancer cell lines.^49,54 Moreover, a variety of possible mechanisms of escape from trastuzumab, such as the overexpression of the stem cell marker CD44, which leads to the loss or blockade of the trastuzumab-binding site at the HER2 extracellular domain, and the upregulation of stem cell markers such as CXCR4, β1 integrin or Notch-1, which lead to the activation of alternative pathways circumventing HER2 signaling, appear to involve many of the same biomarkers that have been implicated in the biology of CS-like cells.⁵⁴

The suggestion that HER2 gene-amplified breast cancer stem-like cell populations are unresponsive to trastuzumab ab initio is the ultimate cause underlying the inherent resistance of HER2-positive breast carcinomas to trastuzumab largely lacks the observational and experimental evidence we normally expect of standard predictive factors for breast cancer treatment selection. Indeed, a major challenge now is to define gene signatures and/or gene products and specific breast CSC-like and/or EMT attributes that may specifically inform about the intrinsic sensitivity of HER2-positive breast carcinomas to HER2-targeted therapies. Apparent approaches to provide biomarkers of trastuzumab-refractory breast CSCs include microarray analyses-based detection of stemness-focused gene signatures, which are significantly enriched in ab initio trastuzumab-refractory HER2-positive breast carcinomas. We recently envisioned that the occurrence of SC-like features in trastuzumab-refractory HER2-positive breast carcinomas could be re-derived, not only from clinically well-defined patient cohorts, but also by identifying biological determinants of trastuzumab sensitivity in panels of cancer cell lines that are able to accurately capture the genetic heterogeneity of the HER2-enriched breast cancer subtype.

Stem Cell-Like Features are Overexpressed in Trastuzumab-Resistant HER2-Gene Amplified Breast Cancer Cells: A Genome-Wide Pathway-Focused Analysis

The cell line SKBR3 is a widely used preclinical model of HER2 gene-amplified breast cancer cells that is exquisitely sensitive to trastuzumab.⁶⁴ Another unique model of HER2-positive breast cancer disease that exhibits de novo cross-resistance to multiple HER2-targeting drugs is the JIMT1 cell line,⁶⁵ which is a ductal carcinoma pleural metastasis of a 62-y-old patient who did not respond to trastuzumab treatment, despite having a tumor with HER2 gene amplification. Taking advantage of the properties of these two cell lines, we recently performed a genome-wide pathway-focused analysis to evaluate whether the presence of tumor-initiating, breast CS-like genetic signatures might underlie de novo resistance to HER2-targeting agents in HER2 gene-amplified breast carcinomas (Fig. S1). Genome-wide analyses of 44K Agilent's whole human arrays were bioinformatically evaluated by Gene Set Enrichment Analysis (GSEA)-based screening of the KEGG pathway database. GSEA performs competitive analyses of predefined gene sets; these analyses are appropriate for examining relatively heterogeneous biological samples. Four enriched gene sets that could be used to functionally differentiate trastuzumab-responsive SKBR3 cells from trastuzumab-refractory JIMT1 cells were identified. Within the non-small cell lung cancer pathway, the JIMT1 transcriptome exhibited downregulation of genes such as ERBB2 (i.e., the trastuzumab target itself), PLCγ, PI3K and PKB (Akt) (Fig. S1, top). Within the cell cycle (Fig. S1, middle) and ribosome (Fig. S1, bottom) pathways, the trastuzumab-refractory JIMT1 transcriptome displayed co-downregulation of 15 ribosomal protein genes that have been described as haploinsufficient tumor suppressors,^66,67 and of numerous genes encoding cyclin-dependent kinases (CDKs). Importantly, trastuzumab-resistant JIMT1 cells exhibited a significant repression of the CDK inhibitor p16^INK4a, a condition common to stem cells and many tumor cells that can link sustained induction of EMT and the activation of DNA methylation of genes that are silenced in basal-like breast carcinomas.⁶⁸ p16^INK4a downregulation has been observed in some basal-like breast cancer cell lines and cells that are CD44⁺CD24^-/low, a phenotype associated with stem-like breast cancer cells that is more frequent in ER-negative/p16^INK4a-negative breast cancer cell lines than in ER-negative/p16^INK4-positive lines.⁶⁹ Given that p16^INK4a depletion has been suggested to reduce the response of ER-negative breast cancer cells to chemotherapy by increasing the percentage of CD44⁺CD24^-/low cells and enhancing the expression of embryonic stem-like genes (e.g., Nanog, Oct4 and Sox2), it is tempting to suggest that, by conferring CSC-like properties, downregulation of p16^INK4 expression could also underlie the de novo resistance to trastuzumab in HER2 gene-amplified JIMT1 cells. Within the extracellular matrix(ECM)-receptor interaction pathway (Fig. 1), the JIMT1 transcriptome exhibited upregulation of the CD44 transcript, which is positively associated with stem cell-like characteristics,⁷⁰ and the expression of gene transcripts encoding structural elements of basal epithelial cells, including the α6 and β1 integrin subunits (ITGA6 and ITGB1)^71-76 and the laminin γ2 subunit (LAMC2),⁷⁷ all of which represent putative stem cell biomarkers.

Figure 1. Trastuzumab-refractory JIMT1 cells are enriched with stem cell markers.

To confirm these findings, we profiled the expression of genes related to the identification, growth and differentiation of stem cells in trastuzumab-sensitive SKBR3 cells and in trastuzumab-refractory JIMT1 cells by quantitative real-time PCR (qRT-PCR) arrays (Fig. 2, left). These analyses not only confirmed the overexpression of the CD44 gene in trastuzumab-refractory JIMT1 cells, but also revealed significantly higher transcriptional activation of the following well-established stem cell-related genes: PPAR-gamma;^78,79 the ABCG2 transporter/breast cancer-resistance protein (BCRP), which is a characteristic of stem cells termed “side population cells” that may contribute to the intrinsic resistance and longevity of normal cells and CSCs^80-82 and can be identified by their ability to exclude the fluorescent dye Hoechst 33342; the FGFR1 component of the FGF signaling pathway;^83,84 and the membrane endopeptidase (MME; CD10) neprilysin, a zinc-dependent metalloprotease that normally regulates the growth of the ductal tree during mammary gland development, because its ability to cleave signaling factors, together with β1 integrin-mediated contact with the basement membrane, is required to maintain the progenitor and stem cell pools in the mammary lineage.^85-88 Perhaps more importantly, the two major breast cancer subtypes, based on their gene-expression profiles (i.e., luminal- and basal-type tumors), express markers corresponding to the major differentiation states of epithelial cells in the breast: luminal (EpCAM⁺) and basal/myoepithelial (CD10⁺). In this scenario, the fact that trastuzumab-refractory, HER2 gene-amplified JIMT1 cells are positive for the myoepithelial marker CD10, strongly suggests that HER2 gene amplification occurs in basal-like carcinomas that exhibit a certain degree of myoepithelial differentiation. In this regard, the expression of the epithelial marker E-cadherin (CDH1) in JIMT1 cells is consistent with a “mixed-lineage” nature of human basal-like breast cancers and, therefore, might reflect an aberrant/incomplete myoepithelial differentiation, but not necessarily the occurrence of global stem cell/progenitor features in mesenchymal-like phenotypes.⁹⁰ Notably, immunofluorescence studies have revealed that E-cadherin is associated with disorganized adhesive structures in trastuzumab-unresponsive JIMT1 cells (Fig. 2, right). We previously demonstrated that when grown under conventional culture conditions, JIMT1 cells exhibit a spiky-like appearance in which the cells are surrounded by numerous lamellipodial protrusions.²⁸ This morphology is relatively common in mesenchymal cells but not in cells expressing functional, E-cadherin-based cell-cell adhesion complexes, such as bona fide luminal epithelial-like cells (e.g., MCF-7; Fig. 2, right). In luminal MCF-7 cells, the localization of E-cadherin is restricted to the plasma membrane at areas of cell-cell contact. In “basoluminal” JIMT1 cells, E-cadherin was found across the entire surface of the plasma membrane and was notably distributed in various cytoplasmic locations (e.g., peri-nuclear areas). Indeed, E-cadherin staining was notably diffused at areas of cell-cell contact, and JIMT1 cells normally exhibited reduced and punctuated E-cadherin staining that was not restricted to areas of cell-cell contact but was also observed along cell edges without adjacent cells. These findings, together with the previously reported abundance of filopodia protrusions, strongly suggest that trastuzumab-refractory JIMT1 cells largely lack junctional organization and have disorganized adhesive structures containing E-cadherin. Although E-cadherin is usually assigned a tumor-suppressor role in epithelial cells, under some circumstances, E-cadherin is associated with increased tumorigenesis and tumor dispersion. Our data are consistent with the idea that deregulated E-cadherin signaling can be an important component of trastuzumab-refractory growth of HER2-positive breast carcinomas cells. Indeed, when JIMT1 cells were infected with a lentivirus expressing E-cadherin-targeted small hairpin RNAs, we failed to obtain surviving cell populations with a good depletion of the E-cadherin protein (data not shown) when the cells were exposed to antibiotic selection to produce a polyclonal population of JIMT1 cells with a stable knockdown of E-cadherin; this finding suggests a crucial role for E-cadherin in the biological properties of trastuzumab-unresponsive JIMT1 cells. Although our data on E-cadherin expression may suggest an inability of JIMT1 cells to induce a classic EMT cadherin-switching, it should be noted that the JIMT1 cells were also positive for N-cadherin (CDH2), a classic EMT marker. Therefore, E-cadherin expression appears to be an exception to the EMT rule, because many other aspects of the EMT are operational in trastuzumab-refractory JIMT1 cells (see below). Although forthcoming studies should definitely elucidate how deregulated E-cadherin function may facilitate primary unresponsiveness to trastuzumab in HER2-positive breast carcinomas, our findings provide a glimpse into the poorly recognized importance of E-cadherin in the biology of breast cancer cell populations enriched with SC-like features.

Figure 2. Left: Activation of EMT/CSC-related gene signatures in trastuzumab-refractory JIMT1 cells. Right: Aberrant E-cadherin localization in trastuzumab-refractory JIMT1 cells.

Basal/HER2-Positive Breast Carcinomas: Evidence for a Novel Stratification of Breast Cancer Taxonomy to Predict Primary Resistance to Trastuzumab

The data obtained from trastuzumab-refractory, HER2 gene-amplified JIMT1 cells suggests that when HER2 gene amplification, which generally occurs within differentiated luminal breast cancer phenotypes, occurs in a basal molecular background, it might result in a basal/HER2-positive subtype of breast carcinoma that naturally exhibits an inherent (i.e., primary) resistance to trastuzumab.²⁸ Mechanistically, an intrinsic tumor cell plasticity that is able to efficiently drive the emergence of a breast CSC-related CD44⁺CD24^-/low mesenchymal phenotype might account for the de novo resistance to trastuzumab in this distinct subgroup of basal/HER2-positive breast carcinomas.^27,89 Alternatively, EMT-related CD44⁺CD24^-/low mesenchymal subpopulations within basal/HER2-positive breast carcinomas might reflect the intrinsic activation of an inherent mechanism of escape from more epithelial, trastuzumab-responsive CS-like cellular states. Celià-Terrasa and colleagues⁹¹ recently proposed that highly metastatic tumor-initiating cells are enriched for a strong epithelial gene program, whereas tumor cell subpopulations with stable mesenchymal traits are notably depleted in CSCs. In their study, the constitutive overexpression of EMT transcription factors in epithelial, CSC-enriched cancer cell populations fully suppressed their self-renewal and metastatic phenotypes upon the activation of a mesenchymal gene program; conversely, knockdown of EMT factors in the mesenchymal subpopulation caused an increase in CSC-like properties, along with an enhancement of epithelial features. Accordingly, Sarrio and colleagues⁹⁰ recently demonstrated that epithelial and mesenchymal subpopulations within basal breast cell lines could exhibit distinct stem cell/progenitor properties. Intriguingly, stem cell/progenitor properties, such as regenerative potential, high aldehyde dehydrogenase 1 activity and the formation of three-dimensional acini-like structures, were found to predominantly reside within the epithelial subpopulation,⁹⁰ and the acquisition of global stem cell/progenitor features was lost when a mesenchymal-like phenotype was spontaneously generated through an EMT process in epithelial cells. In their study, well-recognized EMT inducers, such as SLUG/SNAIL2 and ZEB1, inhibited luminal differentiation, thus preventing the loss of epithelial progenitor and mesenchymal-like phenotypes within basal breast cell lines.

Taken together, these findings suggest that the occurrence of high levels of EMT-driving transcriptional regulators might allow basal/HER2-positive breast carcinoma cells to “enter” and “exit” trastuzumab-responsive CS-like epithelial states. To preliminarily test this hypothesis, we took advantage of earlier studies that aimed to summarize the molecular and cellular analyses of EMT features across breast cancer cell lines in the publically accessible “Neve breast cancer dataset” (Fig. 3).^92,93 Neve and colleagues⁹⁴ profiled 51 human breast cancer cell lines with the Affymetrix HG-U133A chip and integrated further comparative genome hybridization (CGH) and proteomic analyses. With some exceptions, cell lines that clustered to basal B (as distinct from basal A or luminal) in Neve’s study corresponded with the group designated as mesenchymal by Charafe-Jaufreet and colleagues,⁹⁵ who 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. More recently, Prat and colleagues^96,97 analyzed a data set of 52 breast cancer cell lines, including three cell lines that were not included in Neve’s study,⁹⁴ using a 50-gene supervised risk predictor of breast cancer based on intrinsic subtypes (i.e., luminal A, luminal B, HER2-enriched and basal-like). The major luminal, basal A and basal B subgroups originally identified by Neve and colleagues⁹⁴ were again evident, and most of the basal B/mesenchymal cell lines exhibited a gene expression pattern similar to that of the claudin-low tumor subtype [i.e., breast cancer cells with the lowest expression of genes involved in epithelial cell-cell adhesion (E-cadherin and claudins 3, 4 and 7), luminal differentiation (CD24 and EpCAM) and high CD44/CD24 and CD49f/EpCAM mRNA ratios].^96,97

Figure 3. Association of the clinical subtypes of breast cancer cell lines and the primary response to trastuzumab based on the relative EMT/CSC features’ enrichment.

When we organized the 51 cell lines by the subclasses defined in the public Lawrence Berkeley Laboratory (LBL) Breast Cancer Collection,⁹⁴ the expression status of the most widely used EMT markers (i.e., E-cadherin and vimentin) and EMT transcriptional drivers (i.e., SLUG/SNAI2, TWIST and ZEB1), as well as the CD44/CD24 ratios and most of the HER2 gene-amplified breast carcinomas cell lines (i.e., AU565, BT474, HCC202, MDA-MB-361, SKBR3, UACC812 and ZR7530), were found to belong to the luminal subclass of breast cancer cell lines. This finding was expected because luminal differentiation is a property shared by the majority of HER2-overexpressing breast carcinomas and by the cell lines derived from them. Another expected finding was that the entire subset of mesenchymal-like basal B cell lines lacked HER2 gene amplification. Interestingly, a few HER2 gene-amplified breast cancer cell lines matched not only the luminal subgroup, but also the basal A/basal subgroup of cell lines (i.e., HCC1569, HCC1954 and SUM190T). Moreover, when HER2 gene-amplified breast cancer cell lines were classified as trastuzumab-sensitive and trastuzumab-refractory based on data from the literature, we observed that trastuzumab sensitivity ab initio was notably restricted to the luminal/HER2-positive cell lines, whereas all the basal/HER2-positive cell lines exhibited inherent, primary resistance to trastuzumab. Given the predominance of UT Southwestern-derived HCC breast cancer cells among the basal A subgroup, it could be argued that the correlation between the basal/HER2-positive phenotype and the occurrence of primary resistance to trastuzumab might reflect certain biases in the source of cell lines and/or in the culturing methods. However, the non-HCC-related JIMT1 cells, which were not originally included in the Neve data and have been shown to exhibit a strong correlation with the HER2-positive and basal-like subtypes and an inverse correlation with the luminal A subtype of breast carcinomas,^98-100 also belonged to the basal/HER2-positive subgroup of trastuzumab-refractory breast cancer cell lines.

EMT-Like Features are Overexpressed in Trastuzumab-Resistant HER2 Gene-Amplified Breast Cancer Cells

The luminal epithelial marker E-cadherin, which is an important discriminator between basal A cells, which all express approximately median levels of E-cadherin and basal B/mesenchymal cells, which mostly exhibit reduced expression of E-cadherin, was not a good discriminator between trastuzumab-sensitive luminal/HER2-positive breast cancer cells and trastuzumab-resistant basal/HER2-positive cell lines. With the exception of SKBR3 cells, which have a homozygous deletion at 16q22.1 that includes the CDH1 (E-cadherin) gene,¹⁰⁰ all the HER2 gene-amplified breast carcinoma cells could be cataloged as E-cadherin-positive, regardless of their luminal or basal status. However, as has been noted for trastuzumab-resistant JIMT1 cells, the examination of E-cadherin expression alone does not inform about its aberrant localization in disorganized adhesive structures. Considering that many epithelial metastases retain E-cadherin expression (i.e., non-EMT cells can establish metastases or MET can occur at the metastatic site), it might be relevant to evaluate the potential relationship between deregulated E-cadherin function in HER2-positive breast carcinomas exhibiting a JIMT1-like “pseudo-epithelial” differentiation and the intrinsic aggressiveness of these tumors. Although vimentin (VIM) expression is not itself proof of EMT, it was absent in the entire subgroup of trastuzumab-responsive luminal/HER2-positive breast cancer cell lines, whereas it was notably expressed in all the trastuzumab-refractory breast cancer cell lines belonging to the basal/HER2 subgroup. Among the EMT transcriptional drivers, high relative expression of ZEB1 was limited to HER2-negative basal B/mesenchymal lines, and little differential expression was observed among the luminal/HER2-positive and basal/HER2-positive breast cancer cell lines. TWIST1 was also highly expressed in HER2-negative basal/mesenchymal cell lines and, unexpectedly, in a significant number of luminal cell lines. The mixed expression of TWIST1 among the basal A group of cell lines further precluded a discriminating role between trastuzumab responders and non-responders. Of note, the EMT transcription factor found to have the highest discriminator score between trastuzumab-sensitive luminal/HER2-positive and trastuzumab-resistant basal/HER2-positive breast cancer cell lines was SLUG/SNAIL2 (Fig. 3). Thus, all the trastuzumab-refractory basal/HER2-positive cells exhibited relatively high expression of SLUG/SNAIL2, whereas all the trastuzumab-sensitive luminal/HER2-positive cells exhibited relatively low expression of SLUG/SNAI2.

The CD44⁺CD24^-/low phenotype, as determined by mRNA expression, has been repeatedly used to define the mesenchymal immunophenotype of human breast CSCs^51,52 and was, as expected, found to be a hallmark of the HER2-negative basal B/mesenchymal subgroup of breast cancer cell lines. Although basal A breast cancer cell lines exhibited a certain degree of overexpression of mesenchymal markers, they notably lacked the CD44⁺CD24^-/low mesenchymal phenotype and were positive for both markers (i.e., they exhibited a double-positive CD44⁺CD24⁺ phenotype). Breast cancer cell lines maintaining a luminal epithelial phenotype exhibited strong positivity for CD24, in agreement with numerous studies addressing the fact that CD24⁺ cells are associated with more differentiated tissues or tumors.⁷⁰ Given the positive association between CD44 and stem cell-like characteristics and the correlation between CD24 expression and differentiated epithelial features,¹⁰¹ the expression status of CD44 can therefore be considered, in addition to the EMT transcriptional factor SLUG/SNAIL2, a good discriminator between trastuzumab-sensitive luminal/HER2-positive (SLUG^-/CD44^-) and trastuzumab-refractory basal/HER2-positive (SLUG⁺/CD44⁺) breast cancer cells. Indeed, our and other groups have previously described the importance of the CD44-hyalyronan pathways for the escape of breast cancer from trastuzumab-based therapies. Dhillon and colleagues¹⁰² observed that long-term exposure of luminal/HER2 BT474 cells to trastuzumab selected for a resistant subpopulation that was enriched for CD44.¹⁰² Conversely, small interfering RNA inhibition of CD44 restored trastuzumab sensitivity. We have demonstrated that long-term exposure of luminal/HER2-positive SKBR3 cells to trastuzumab similarly selects for a resistant subpopulation of CD44⁺ cells.¹⁰³ Therefore, it is intriguing to note that both inherent (primary) and acquired (secondary) resistance to trastuzumab converge in CD44⁺CD24⁺ breast cancer cells with a basal-epithelial phenotype (Fig. 3).

To definitively confirm the hypothesis that trastuzumab-refractory, basal/HER2-positive tumor cells may be intrinsically enriched with EMT features, we profiled the expression of 84 key genes that either exhibit changes in expression during the EMT process or regulate those EMT-related expression changes by using commercially available qRT-PCR arrays with trastuzumab-responsive SKBR3 cells (i.e., luminal/HER2-positive) and trastuzumab-refractory JIMT1 cells (i.e., basal/HER2-positive).¹⁰⁴ When a 3-fold or greater difference in mRNA levels was used to determine significant changes in the expression of genes involved in the EMT genetic program, trastuzumab-refractory JIMT1 cells were found to exhibit significantly enhanced expression of developmental EMT drivers that have been repeatedly demonstrated to play pivotal roles in EMT-driven metastatic breast cancer progression, including SNAIL2 (SLUG), TWIST1 and, to a lesser extent, ZEB1 (Fig. 2, left). Moreover, the expression of specific, invasion-promoting EMT target genes, such as plasminogen activator inhibitor type-1 (PAI-1; SERPINE1)¹⁰⁵ and the mesenchymal marker VIM, was significantly increased in trastuzumab-refractory JIMT1 cells. In addition to an upregulation of the expression of α(5)-integrin (ITGA5), a fibroblast marker and the natural receptor for fibronectin, whose upregulation generally occurs in parallel with an upregulation of α-smooth muscle actin,¹⁰⁶ and caldesmon (CALD1), an actin-linked regulatory protein found in smooth muscle and non-muscle cells that has numerous functions in cell motility exerted via the reorganization of the actin cytoskeleton,¹⁰⁷ compared with trastuzumab-responsive SKRB3 cells, trastuzumab-refractory JIMT1 cells were also found to exhibit increased levels of TIMP1, an endogenous inhibitor of matrix metalloproteinases (MMPs) that also functions as a signaling molecule that induces the expression of developmental EMT transcription factors, such as SNAIL2/SLUG, TWIST1 and ZEB,¹⁰⁸ and of secreted protein acidic and rich in cysteine (SPARC), a matricellular protein involved in extracellular matrix remodeling and invasion that is directly involved in the acquisition of mesenchymal traits that favor metastatic dissemination.^109-111

Suppression of EMT-Like Features Circumvents Primary Resistance to Trastuzumab in Basal/HER2-Positive Breast Cancer Cells

In the above-mentioned scenario, it might be tempting to suggest that primary resistance to trastuzumab can be explained in terms of trastuzumab-unresponsive, EMT-enriched CD44⁺ CSC-like populations. In other words, an intrinsic enrichment of EMT-like CSC cellular states might explain the intrinsic refractoriness of basal/HER2-positive breast cancer cells to trastuzumab. Although this suggestion might appear counterintuitive, given the widely accepted CD44⁺CD24^-/low mesenchymal immunophenotype of the so-called breast CSCs^51,52 and that trastuzumab-resistant basal/HER2-positive cells are positive for both marks (i.e., CD44⁺CD24⁺),²⁷ it should be noted that CD24 exhibits a dynamic regulation, as was recently demonstrated by Mayer and colleagues.¹¹² CD44⁺CD24⁺ basal cells can readily give rise to CD44⁺CD24^-/low mesenchymal cells and vice versa. Moreover, when we recently explored the spontaneous evolution of the CD44⁺CD24^-/low mesenchymal immunophenotype in trastuzumab-refractory basal/HER2-positive JIMT1 cells, we concluded that the dynamic expression of EMT-related markers was not limited to CD44/CD24, i.e., the number of cells bearing the CD44⁺CD24^-/low mesenchymal immunophenotype switched over time from 10% in early passages to 80% in late passages. This phenotypic switch occurred because the trastuzumab-unresponsive, basal/HER2-positive JIMT1 cell cultures enriched with CD44⁺CD24^-/low mesenchymal cells also exhibited a reduced expression of the HER2 protein and an increased secretion of pro-invasive/metastatic chemokines and metalloproteases.²⁷ Korkaya and colleagues¹¹³ have also reported that generation of trastuzumab resistance, either by knocking down PTEN expression or by long-term exposure to trastuzumab, is mediated in both cases by the expansion of CSC populations that display EMT-like phenotypes and oversecrete pro-invasive/metastatic chemokines (i.e., IL6). Taken together, these findings appear to confirm that either intrinsic or microenvironmental activation of paths to stemness, such as the EMT, directly regulate the responsiveness of CS-like cells to trastuzumab.

We have recently explored the causal relationship between EMT-driven tumor cell plasticity, which can drive the emergence of a CS-related CD44⁺CD24^-/low mesenchymal phenotype, and the maintenance of de novo resistance to trastuzumab in basal/HER2-positive breast cancer cells.¹¹⁴ Lentivirus-delivered small hairpin RNAs were employed to specifically and stably reduce the expression of EMT transcription factors in trastuzumab-refractory basal/HER2-positive cells. Then, cell proliferation assays and pre-clinical nude mice xenograft-based studies were performed to assess the contribution of specific EMT transcription factors to inherent trastuzumab resistance. The specific knockdown of SLUG/SNAIL2 suppressed the CD44⁺CD24^-/low mesenchymal immunophenotype, and the isolation of these cells by magnetic-activated cell sorting confirmed that their intrinsic unresponsiveness to trastuzumab was mediated by transcriptional upregulation of the luminal epithelial marker CD24 in basal/HER2-positive cells, which, in turn, gained sensitivity to the growth-inhibitory effects of trastuzumab following SLUG/SNAIL2 gene depletion. Accordingly, depletion of the SLUG/SNAIL2-driven CD44⁺CD24^-/low mesenchymal subpopulation reduced the tumorigenic potential of basal/HER2-positive JIMT1 cells and switched their trastuzumab-refractory phenotype to a sensitive phenotype when injected into nude mice. Therefore, aberrant expression of the EMT transcription factor SLUG/SNAIL2 appears to bias basal/HER2-positive epithelial cells toward a highly tumorigenic, CSC-like mesenchymal fate that is unresponsive to trastuzumab. In other words, the special proclivity of basal/HER2-positive breast cancer cells to undergo EMT reflects the intrinsic phenotypic plasticity of basal/HER2-positive breast cancer cells to rapidly acquire an invasive differentiation status of trastuzumab-refractory CS-like cells. Alternatively, the CSCs might not be considered distinct cancer cell entities (i.e., normal and tumor non-stem cells can spontaneously convert to a stem-like state and vice versa), because tumor cells can transiently acquire or lose SC-like properties as a consequence of stochastic state transitions, such as EMT regulation.^115-119 Trastuzumab-responsive CSCs could potentially give rise to non-CSC cellular states that, by lacking the major attributes of CSCs, could remain “hidden” from trastuzumab activity. In the latter scenario, the predominant CSC cellular state in trastuzumab-responsive luminal/HER2-positive breast carcinomas would be different from that of basal/HER2-positive cell types, which may be characterized as an intermediate CD44⁺CD24⁺ stage of EMT in epithelial stem cells¹²⁰ that, unlike luminal cell types, may be susceptible to SLUG/SNAIL2-driven acquisition of the bona fide CD44⁺CD24^-/low mesenchymal phenotype. The CSC/non-CSC social structure within HER2-enriched breast carcinomas may be critical for trastuzumab-based treatment decisions in the clinic, because it may primarily dictate both the intrinsic sensitivity to trastuzumab in luminal/HER2-positive breast cancer cells and the intrinsic refractoriness to trastuzumab in basal/HER2⁺ breast cancer cells. Thus, in addition to the relationship of the trastuzumab-refractory CD44⁺CD24^-/low mesenchymal phenotype of basal/HER2-positive breast carcinoma cells with the developmental origins of the intrinsic breast cancer subtypes (discussed below), it is relevant to note our recent discovery that EMT transcription factors, such as SLUG/SNAIL2, might induce an enhanced phenotypic plasticity that would allow basal/HER2-positive breast cancer cells to dynamically “enter” and “exit” trastuzumab-responsive SC-like states. It remains to be unambiguously defined whether SLUG/SNAIL2-driven CD44⁺CD24^-/low subpopulations reflect an enrichment of trastuzumab-refractory CSC mesenchymal cells or, alternatively, an intrinsic mechanism of escape from more epithelial, trastuzumab-responsive CS-like cellular states.

Suppression of EMT-Like Features Might Counterintuitively Enhance the Occurrence of CS-Like Pluripotent States in Basal/HER2-Positive Breast Cancer Cells

It was recently demonstrated that activation of the EMT genetic program could efficiently suppress major attributes of human epithelial CSCs.^90,91 In this regard, it should be noted that reprogramming of the normal cellular fate, such as activating a new pathological developmental program, in the case of cancer tissues, or generating induced pluripotent stem cells (iPSCs) from somatic cells, and cellular plasticity are tightly intertwined; thus, only some tumor or normal cells intrinsically possess the necessary plasticity for tumoral or somatic reprogramming to occur, respectively. In induced pluripotency, the reprogramming factors [the four Yamanaka transcription factors: Oct4, Sox2, Klf4 and c-Myc (OSKM)] are, in principle, all that is required to induce dedifferentiation. However, the extremely low efficiency and kinetics of the process clearly indicate that other molecular aspects of the target cell’s biology also play a crucial role. To generate a CSC cellular state, the capacity for self-renewal and asymmetric division must be re-acquired not only through oncogenic changes that destabilize the normal transcriptome of the target cell, but also through permissive/selective genetic, epigenetic and metabolic alterations that prevent cells from changing their original identity. Therefore, in both scenarios, the loss of a molecular feature that is essential for the maintenance of cell identity may act as a tumor- and iPSC-inducing alteration. One of the master regulators of cellular identity that must be eliminated to allow cells to become amenable to full reprogramming is the degree of EMT differentiation.¹²¹ The generation of iPSCs from normal adult fibroblasts requires the proper activation and functioning of the mesenchymal-to-epithelial transition (MET) program, which is orchestrated by suppressing pro-EMT signals from the extracellular milieu (e.g., the EMT activator TGFβ) and activating an epithelial gene program intracellularly.^122-127 In fact, the four canonical Yamanaka pluripotency factors transcriptionally impede the EMT process: Sox2/Oct4 suppress the EMT mediator SNAIL; c-Myc downregulates TGF-β1 and TGF-β receptor 2, and Klf4 induces epithelial genes, including E-cadherin.^122-124 Moreover, during cellular reprogramming of fibroblasts to iPSCs by OSKM, fully reprogrammed cells can be exclusively observed within cell populations that are positive for the epithelial marker E-cadherin, but they cannot be obtained in the absence of E-cadherin. E-cadherin is upregulated during somatic cell reprogramming, and the established iPSCs possess E-cadherin-mediated cell-cell contacts and, thus, are morphologically indistinguishable from embryonic stem (ES) cells, in which E-cadherin is highly expressed and interference with E-cadherin causes differentiation. Accordingly, it has been shown that (1) the expression of E-cadherin alone can facilitate the reprogramming of adult fibroblasts and the acquisition of pluripotency, even in the absence of exogenous Oct4, and that (2) chemicals reported to upregulate E-cadherin considerably increase reprogramming efficiency, whereas knockdown of endogenous E-cadherin or the abrogation of E-cadherin-mediated cell-cell contact compromises the generation of iPSCs.^125-127

We have demonstrated that the steady depletion of the CSC-related CD44⁺CD24^-/low mesenchymal immunophenotype in response to the specific knockdown of EMT drivers, such as SLUG/SNAIL2, efficiently circumvents the primary resistance of basal/HER2-positive cells to trastuzumab.¹¹⁴ However, genetic ablation of the EMT program failed to alter the ability of trastuzumab-sensitized basal/HER2-positive breast cancer cells to form bona fide mammospheres, which are non-adherent, spherical cell clusters enriched in mammary stem/progenitor cells. Our findings, along with the recently revealed requirement of epithelial cellular states to enable the activation of stemness programs in normal and cancer stem cells, strongly favor a functional heterogeneity in the breast CS-like compartment and suggest that, in addition to the widely recognized mesenchymal-like phenotypes of breast CSCs, tumor-initiating cells with an epithelia-like morphology and functionality should also be considered to gain a better understanding of the primary responses of HER2 gene-amplified breast cancer cell to trastuzumab. To preliminarily evaluate the hypothesis that, although trastuzumab effectively targets CSCs, the degree of trastuzumab responsiveness in a clinical setting may counterintuitively be driven by changes in the epithelial/mesenchymal cellular state of CSCs, we performed two sets of experiments. First, to directly explore the mechanisms by which differentiated cancer cells dynamically enter into the so-called cancer stem cell (CSC) cellular state, we investigated whether the expression status of the epithelial marker E-cadherin operates as one of the biological barriers to prevent the “differentiated cancer cell state” from being reversed to a “pluripotent SC-like cancer cell state” (even partially) (unpublished observations). Second, to establish a linkage between the EMT phenomenon and the reprogramming ability of basal/HER2-positive breast cancer cells, we determined whether the gain of epithelial features (e.g., E-cadherin expression) upon the silencing of EMT drivers (e.g., SLUG/SNAIL2, TWIST1) negatively or positively impacted the endogenous expression of pluripotent Yamanaka transcription factors.

The lentiviral introduction of the four canonical reprogramming transcription factors used in the Yamanaka cocktail to E-cadherin-positive MCF-7 breast cancer cells induced the appearance of rare colonies with characteristic embryonic SC-like morphologies (unpublished observations); the immature status of E-cadherin-positive MCF-7/iPSC clones was confirmed by their positivity for widely used pluripotency markers (e.g., Tra-1-60, Tra-1-81, Ssea-1, Ssea-4, Oct4, Sox2 and Nanog). Notably, when we introduced the OSKM transcription factors into E-cadherin knocked-down (KD) MCF-7 cells generated in our laboratory by short hairpin technology, we only observed changes in cell culture morphology that were somewhat analogous to “background colonies.” However, E-cadherin KD MCF-7 cells failed to generate bona fide iPSC clones, suggesting that epithelial breast cancer cells cannot be reprogrammed to pluripotent stem cell states (even partially) by silencing E-cadherin. Notably, HER2 expression, in the absence of HER2 gene amplification, was upregulated in E-cadherin-positive MCF-7/iPSC clones (unpublished observations). Indeed, HER2 expression was also upregulated when fibroblasts, a product of the EMT, were pushed to an E-cadherin-positive stem cell-like state via reprogramming with the OSKM transcription factors (unpublished observations). Furthermore, unpublished observations by Wichas’s group indicate that clinically HER2-negative primary luminal breast carcinomas contain a CS-like subpopulation of cells that express the HER2 protein on the invasive edge of the tumor.¹²⁸

To assess whether repression of mesenchymal features in basal/HER2-positive breast carcinomas cells significantly alters the expression of endogenously expressed reprogramming transcription factors, we monitored the expression of each individual reprogramming factor in JIMT1 parental cells and in JIMT1 cells engineered to ablate the expression of the EMT transcription factor SLUG/SNAIL2. When we performed real-time PCR analyses to assess Oct4, Sox2, Klf4 and c-Myc mRNA levels, we concluded that the gain of epithelial features (e.g., E-cadherin expression) in response to silencing of the CD44⁺CD24^-/low-driver SLUG/SNAIL2 in basal/HER2-positive cells was accompanied by a significant upregulation of the pluripotency gene Klf4 (Fig. 4). These findings confirm and extend recent studies about the joint link between SLUG, Klf4 and EMT in tumor formation and metastatic progression.^129,130 E-cadherin is a transcriptional target of Klf4, and, therefore, Klf4 is required to maintain E-cadherin expression and prevent the EMT in mammary epithelial cells. Indeed, forced expression of Klf4 in pure CD44⁺CD24^-/low mesenchymal breast cancer cells is sufficient to restore E-cadherin expression and suppress migration and invasion.¹³¹ The tumor- and metastasis-suppressive role of Klf4 in breast cancer has been recently confirmed by Yori and colleagues,¹³² who reported that Klf4 expression is lost in a mouse model of HER2-positive breast cancer, even in the context of the MMTV-Neu transgene, thus suggesting a negative selection for the sustained expression of Klf4 in HER2-positive breast cancer cells. Moreover, transient overexpression of Klf4 in a HER2-expressing 4T1 orthotopic mammary cancer model attenuated not only primary tumor growth, but also micrometastases to the lungs and liver via a reduction in the expression of SNAIL.¹³² However, Yu and colleagues¹³³ have reported that Klf4 is highly expressed in CSC-enriched populations in mouse primary mammary tumor and breast cancer cell lines. Knockdown of Klf4 significantly decreased the proportion of stem/progenitor cells, suppressing cell migration and invasion, reducing colony formation in vitro and inhibiting tumorigenesis in immunocompromised mice. In this regard, although our findings appear to confirm that the tumor-suppressing activity of the EMT driver SLUG in basal/HER2-positive breast cancer cells operates via a reciprocal regulation of SLUG and the reprogramming factor Klf4, as previously suggested by other groups,^129,130,132 our findings further complicate the disputed nature (i.e., tumor suppressive vs. oncogenic) of the EMT/Klf4 duo.

Figure 4. Top: Specific regulation of the reprogramming factor Klf4 in response to the genetic ablation of the EMT factor SLUG/SNAIL2 in trastuzumab-refractory JIMT1 cells. Bottom: Restoration of E-cadherin along the basolateral membrane in response to the genetic ablation of the EMT transcription factors SLUG/SNAIL2 and TWIST1.

These results delineate the following scenarios: (1) breast cancer cells require the expression of the epithelial (basal/luminal) marker E-cadherin for the de novo generation of pluripotent, HER2-expressing CS-like cellular states, and (2) suppression of the trastuzumab-unresponsive CSC-related CD44⁺CD24^-/low mesenchymal immunophenotype results in the endogenous upregulation of the pluripotent transcription factor Klf4. Therefore, by impeding the activation of the EMT program in basal/HER2-positive cells, the remaining HER2-positive cell population switches to a trastuzumab-responsive, more epithelial phenotype that is apparently enriched with the reprogramming factor Klf4. In response to knockdown of SLUG/SNAIL and the concomitant upregulation of Klf4, basal/HER2-positive JIMT1 cells gained a better junctional organization, and the adhesive structures containing E-cadherin became more structured (Fig. 4). Accordingly, the maturation of adherens junctions was concomitant with a cobblestone or cell-cell clustered morphology and the formation of cell contacts, which was not observed in basal/HER2-positive cells transduced with a control-shRNA. The morphological shift of trastuzumab-unresponsive basal/HER2-positive JIMT1 cells toward the epithelial end of the mesenchymal-epithelial spectrum occurred irrespective of the cell culture density; therefore, this shift was not affected when we limited the establishment of cell-cell connections. Thus, it is reasonable to suggest that outside-in signaling mediated by the upregulation of E-cadherin is not necessary for the morphological and functional change of basal/HER2-positive breast cancer cells. While it is well-accepted that E-cadherin expression is intimately connected to the degree of epitheliality of a cell with respect to its morphology and migratory/invasive abilities, it appears that restoring E-cadherin functionality is intimately connected with the ability of trastuzumab to induce growth-inhibitory and antitumor activity in HER2-positive breast carcinoma cells. This scenario suggests that enhancing the occurrence of trastuzumab-responsive, CS-like epithelial phenotypes could potentially circumvent the primary resistance to trastuzumab observed in basal/HER2-positive breast carcinomas. It is well-established that tumorigenesis and reprogramming are partially promoted by overlapping mechanisms, and research on reprogramming might help unravel the mechanisms of carcinogenesis and vice versa.^134-141 However, the scarce knowledge that is currently available regarding the induction of a pluripotent phenotype via the activation of Yamanaka transcription factors in cancer cells indicates, somewhat unexpectedly, that reprogramming decreases the tumorigenicity of cancer cells by promoting their terminal differentiation, thus preventing the expression of aggressive phenotypes (i.e., a loss of mesenchymal lineage markers).^142-144 Given that basal/HER2-positive breast cancer cells may express some reprogramming transcription factors at baseline, it might be plausible that by reversing a critical roadblock toward induced pluripotency (i.e., the EMT), e.g., preventing the EMT in epithelial cells cultured with serum can produce iPSCs with only two reprogramming factors,¹²² basal/HER2-positive breast cancer cells may stably acquire a more appropriate CS-like phenotype in terms of tumor cell “hibernation” and trastuzumab “responsiveness.”

Molecular Forecasting of HER2-Positive Breast Carcinomas Based on CSC States and the EMT

The currently available molecular taxonomy of breast carcinomas appears to be robust, reproducible and relatively easy to apply in the clinic. Conversely, our ability to a priori identify which patients from the major breast cancer subclasses, luminal, basal and HER2 subtypes, are suitable for specific therapies is rather limited. We have recently envisioned that new predictive approaches aimed at better forecasting early responses of HER2-positive breast cancer patients to trastuzumab can be re-derived not only from clinically well-defined patient cohorts, but also by identifying biological determinants of trastuzumab sensitivity in cancer cell lines that capture the genetic heterogeneity of HER2-positive breast carcinomas in a manner that can help us to differentiate at least two biologically distinct subgroups of trastuzumab-responsive and trastuzumab-unresponsive tumors ab initio. Our current approach is based on the molecular detail that the clinical efficacy of trastuzumab is mostly related to its ability to specifically and efficiently target HER2-driven populations of CSCs and that, somewhat counterintuitively, a variety of recently discovered mechanisms of escape from trastuzumab appear to involve many of the same well-accepted biomarkers implicated in the biology of CS-like, tumor-initiating cells. How can we reconcile these findings to improve our ability to a priori identify which patients of the HER2 subtype are suitable for trastuzumab-based therapy ab initio? In the traditional, one-way hierarchy of CSCs, in which all cancer cells should descend from special subpopulations of self-renewing CSCs, HER2-positive CSCs can occur solely by self-renewal; trastuzumab treatment, by impeding the self-renewal property of HER2-positive CSCs and their intrinsic resistance to conventional treatments, is expected to induce tumor shrinkage and further reduce the rate of breast cancer recurrence when used in combination with traditional anti-breast cancer therapies. In the new, alternate model that we present here, more differentiated, non-stem CS might revert to trastuzumab-refractory CS-like cells due to the intrinsic or microenvironmentally driven activation of paths-to-stemness, similar to the CS-related EMT phenomenon. However, this model can also operate in a reverse mode in which stochastic transitions of trastuzumab-responsive CSC-like cellular states might give rise to non-CSC cellular states that, by lacking major attributes of CSCs, remain molecularly obscured from trastuzumab activity.

Trastuzumab is an anti-CSC drug

If the effectiveness of HER2-targeted drugs, including trastuzumab, largely depends on their ability to target breast CSCs, we can no longer underestimate the cellular state and focus solely on cellular phenotype when defining HER2-positive CSCs. A better understanding of the CSC/non-CSC social structure within HER2-overexpressing breast carcinomas should be viewed as a critical aspect for trastuzumab-based treatment decisions in the clinic. When deciphering the biological significance of CSC features and EMT-related phenomena for reinterpreting the genetic heterogeneity that differentiates trastuzumab-responders from non-responders among HER2-positive breast carcinoma cell lines, we concluded that novel predictive approaches aimed at better forecasting early tumor responses to trastuzumab should incorporate biological determinants causally underlying the intrinsic flexibility of HER2-positive CSCs to “enter” or “exit” trastuzumab-sensitive states. An accurate integration of CSC cellular state- and EMT-related biomarkers could better refine the currently available breast cancer molecular taxonomy of HER2-enriched breast carcinomas to distinguish, at a minimum, two intrinsically distinct HER2-positive breast cancer subtypes (i.e., luminal/HER2-positive and basal/HER2-positive) based on their prognostic and/or trastuzumab-responsiveness features. Critical to this new sub-classification of the HER2-enriched breast cancer subtype, is the notion that the effectiveness of anticancer therapies, such as trastuzumab, which are able to target CSCs, can be greatly impacted by the de novo generation of either CSCs (because non-CSCs would regenerate CSCs after cessation of trastuzumab therapy and lead to renewed tumor growth) or well-camouflaged CSC-like stages that escape trastuzumab therapy. Therefore, to be effective, anti-HER2 breast carcinoma approaches will need to combine agents that are selectively toxic to HER2-positive CSCs (i.e., the trastuzumab antibody itself) by targeting the bulk, non-CSC populations within HER2-positive breast carcinomas (e.g., taxane- or anthracycline-based chemotherapy), as well as by inhibiting EMT-related transitions from non-CSC to CSC states and/or from a trastuzumab-responsive CSC (epithelial) state to a non-responsive (mesenchymal) state (e.g., salinomycin, mTOR inhibitors, CD44-targeting antibodies, polyamine metabolism inhibitors and AMPK agonists such as metformin).^{89,103,145-155}

CSCs “are made, not born”: Incorporating “mesenchymal transition signatures” to predict the anti-CSC efficacy of trastuzumab

Assuming that CSCs “are made, not born,” and that CSCs are flexible entities that can transform their trastuzumab-sensitivity-state, we must determine how to integrate CSC- and/or EMT-related biomarkers with currently available clinic-pathological variables to determine their prognostic value in HER2-positive breast carcinomas treated with trastuzumab-based therapy. While we acknowledge that an accurate development and validation of CSC/EMT-based trastuzumab-associated predictive signatures requires prospectively controlled clinical trials, it is worth noting that Liu and colleagues¹⁵⁶ have recently confirmed that a 17-gene signature from enriched HER2 mammary CSCs can accurately predict clinical outcome for human HER2-positive/ER-negative breast carcinomas. Of note, their 17-gene HER2-CSCs-enriched signature, which was generated on the basis of differentially expressed genes in CSC vs. non-CSC fractions and trained on one HER2-positive breast cancer cohort, revealed that patients with HER2-positive/ER-negative tumors enriched with CSC features were significantly resistant to chemotherapy but responded well to chemotherapy plus trastuzumab. Thus, it again appears that the intrinsic enrichment of CSC features is associated with an effective response to trastuzumab in HER2-positive breast carcinomas. Although Giordano and colleagues¹⁵⁷ recently revealed that gene transcripts of EMT-inducing transcription factors and CSC features are elevated in a significant percentage of HER2-positive metastatic breast cancer patients, they did not determine whether EMT and/or CSC features had prognostic value in patients treated with trastuzumab-based therapy. Therefore, additional studies are needed to unambiguously determine if the application of predictive CSCs/EMT-related tests can inform which HER2-positive tumors are suitable for trastuzumab-based regimens. A set of genes comprising an “EMT core signature” has been recently derived after triggering the EMT and observing the resulting shared changes in gene expression.¹⁵⁸ Another multi-cancer gene expression signature involving several EMT markers that are coordinately overexpressed in only a subset of malignant samples that have exceeded a particular cancer-staging threshold was also recently identified.¹⁵⁹ The latter signature, which includes 20 known EMT-associated genes, suggests a fibroblastic nature after passing through a SLUG/SNAIL2-based EMT, and, notably, E-cadherin is not downregulated, at least at the mRNA level. Anastassiou and colleagues¹⁶⁰ have confirmed that their precise multi-cancer EMT signature is present in numerous publicly available data sets and in human cancer cell xenografts in vivo but not in stromal cells. Because Anastassiou’s group acknowledges that their “cancer mesenchymal transition signature” may initiate from spontaneously initiated CSCs states,^117,160 it might be tempting to suggest that such a mesenchymal transition signature could be used as a predictive signature that merits validation both prospectively and in samples from completed trastuzumab-based clinical trials.¹⁶¹ Moreover, prospective trastuzumab therapy-based clinical trials might be designed with predictive EMT-related marker validation in mind, following rigorous design,¹⁶¹ because although it is plausible that not all EMT markers included in the mesenchymal transition signature will meet all surrogacy criteria, their use could provide crucial information on the occurrence of primary sensitivity/resistance to trastuzumab. In this regard, the neoadjuvant setting constitutes an ideal scenario for the validation of such markers because we can easily relate early-CSC/EMT biomarker response to trastuzumab outcome.¹⁶² However, a mesenchymal transition signature may likely reflect a reversible process in which the mesenchymal markers in CSC/non-CSC cancer cells may dynamically appear and disappear depending on intrinsic and/or microenvironmental signals. Consequently, the potential use of mesenchymal transition-like predictive signatures would have high selectivity but not sensitivity to predict the trastuzumab unresponsiveness of HER2-positive breast carcinomas ab initio (e.g., the presence of the signature would likely imply inherent resistance to trastuzumab, but the lack of the signature in a particular extracted sample from a HER2-positive breast carcinoma might not imply a trastuzumab-sensitive tumor state, because the signature may have appeared earlier).

Apart from their direct implication in the EMT phenomenon, EMT transcriptional drivers (i.e., SNAIL factors) can also regulate other cellular processes related to cell proliferation and survival, thus contributing to resistance to pro-apoptotic stresses.^163-165 In epithelial cells, promoting cell survival can be a primary function of EMT factors such as SLUG/SNAIL2 rather than being only acquired concomitantly with the EMT.¹⁶⁶ It is noteworthy that, in processes that do not require a complete EMT and in which cells retain some cell-cell junctions and do not acquire full mesenchymal characteristics at any point, as appears to occur in mixed basal/HER2-positive breast carcinoma cells, E-cadherin is redistributed in a manner that is no longer restricted to adherens junctions. Moreover, the redistribution of E-cadherin is not accompanied by a reduction in the levels of E-cadherin protein, even in cells with high concentrations of nuclear SLUG/SNAIL2. In this scenario, SLUG/SNAIL2 does not function as a repressor of E-cadherin but rather rescues the cells from apoptosis by repressing pro-apoptotic factors that otherwise would promote cell death during a partial EMT.^163-166 Although it has been suggested that members of the SNAIL family could act primarily as survival factors and inducers of cell movement rather than inducers of the EMT,^167-169 a more likely scenario is that SLUG/SNAIL2’s survival function can precede its role in inducing a full EMT.¹⁶⁶ The latter notion is supported by our recent findings that the SLUG/SNAIL2-mediated pro-survival/resistance function is associated, at least in part, with EMT and stemness in basal/HER2-positive breast cancer cells inherently resistant to trastuzumab.¹¹⁴ Although we acknowledge that further studies in large tumor series using highly specific anti-SLUG/SNAIL2 antibodies are required to achieve a better understanding of the clinical relevance of EMT/CSC-related factors in the primary response of HER2-positive breast carcinomas to HER2-targeted therapies, the potential use of SLUG/SNAIL2-based predictive signatures would certainly have high selectivity and sensitivity for predicting the trastuzumab unresponsiveness of HER2-positive breast carcinomas, because these signatures can function as surrogates of two HER2-positive breast cancer subtypes (e.g., basal/HER2 vs. luminal/HER2), which distinctly benefit from trastuzumab-based therapies ab initio.

Corollary: It is Time to Refine HER2-Positive Breast Cancer Molecular Taxonomy

Candidate predictive classifiers for trastuzumab responsiveness ab initio need to be re-derived from well-characterized patient cohorts defined by virtue of clinical and biological variables. The best method and most accurate biomarkers for the identification of CSC states within distinct molecular subtypes need to be better explored, because we urgently need to translate the CSC concept to clinical practice, particularly when using drugs with anti-CSC activity, such as the anti-HER2 monoclonal antibody trastuzumab. In this regard, it must be taken into consideration that whenever breast cancer CSC-associated markers (CD44⁺CD24^-/low) are present in tumors, they most likely identify the intrinsic ability of the tumor cell of origin to dynamically originate CSC/EMT cellular states (Fig. 5). Previous studies have already demonstrated an association between basal-like carcinomas and the widely accepted breast CSC phenotype, CD44⁺CD24^-/low.^{52,70,170-172} Within basal-like tumors, Ricardo and colleagues¹⁷² have demonstrated a tendency toward reduced patient survival [disease-free survival (DFS) and overall survival (OS)] when carcinomas exhibit a predominant CD44⁺CD24^-/low CSC-like phenotype. Although HER2-overexpressing tumors, as a group, are unlikely to exhibit increased expression of CD44⁺CD24^-/low markers, it would be of interest to evaluate whether rapidly interchangeable CD44⁺CD24⁺/CD44⁺CD24^-/low phenotypes could help to distinguish different HER2 subtypes in terms of trastuzumab responsiveness ab initio.

Figure 5. Refining HER2-positive breast cancer molecular taxonomy based on EMT/CSC features: From trastuzumab-responsive luminal/HER2 breast carcinomas to trastuzumab-refractory basal/HER2 breast carcinomas.

Gene expression microarrays have almost exclusively been able to discriminate HER2 oncogene-amplified tumors and basal-like tumors as separate clusters.^1,3,4 The inverse association is not complete, because basal cytokeratin expression and HER2 gene amplification have been observed concomitantly in tumors^173,174 and in the trastuzumab-unresponsive, basal/HER2-positive JIMT1 cell line.⁹⁸ Indeed, the presence of a basal/HER2-positive phenotype enriched with an intrinsic plasticity to “enter” may certainly explain the pioneering finding by Harris and colleagues⁴⁸ that a specific HER2-positive tumor phenotype that was more likely to express genes associated with the basal-like phenotype, including higher expression of basal keratins (CK5, CK14, CK15 and CK17), was more frequent in the non-responding group of patients receiving pre-operative trastuzumab than in the responding group. Supporting the notion that the “basal phenotype” based on basal cytokeratins 5 and 14 immunostaining patterns may comprise more than one breast cancer biological entity, Laakso and colleagues¹⁷³ demonstrated that the amplification of HER2 can be found almost exclusively in the so-called “basoluminal” group, which exhibited shorter relapse-free survival than the basal tumor group. Notably, when the basoluminal tumor group was sub-stratified by HER2 status, the authors confirmed that the difference between basal and basoluminal tumors was not due to more frequent HER2 amplification in the basoluminal group.¹⁷³ Although there was a trend toward faster progression of basoluminal/HER2-positive breast carcinomas, it became obvious that basal and basoluminal tumors differed in their natural biological aggressiveness. Can we propose a basal-cytokeratin surrogate definition of HER2-positive, breast carcinomas to predict the response to trastuzumab therapy? Liu and colleagues,¹⁷⁵ when initially characterizing the prognostic features of basal-like breast carcinomas without the influence of hormone receptor status in a series of hormone receptor-negative breast tumors, confirmed the existence of a small group of hormone-receptor-negative tumors that express HER2 and basal markers simultaneously. Moreover, and in contrast to previous studies, the subtype with the poorest outcome was not basal-like or HER2-positive but rather a basal/HER2-positive phenotype.¹⁷⁵ Bagaria and colleagues¹⁷⁶ have recently followed a similar strategy to characterize the prognostic features of three different immunohistochemically defined subgroups of HER2-positive breast carcinomas. When “luminal-HER2⁺” (ER-positive and basal CK-negative), “HER2⁺” (ER-negative and basal CK-negative) and “basal/HER2⁺” (ER-negative and basal CK-positive) were correlated with clinicopathological features and overall survival, the authors concluded that the basal/HER2⁺ subtype was associated with the worst prognosis after adjusting for age, tumor size, lymph node status and adjuvant treatment. Although the authors acknowledged that the use of adjuvant trastuzumab and chemotherapy did not greatly differ between patients with HER2⁺ and basal/HER2⁺ tumors, the fact that patients with basal/HER2⁺ tumors displayed worse survival appears to suggest that the basal/HER2⁺ subtype perhaps may not only have prognostic value but may also have utility as a predictive marker of trastuzumab resistance. Our own laboratory has shown that a subgroup of HER2-overexpressing breast carcinomas overlaps in the histological features of the basal-like phenotype.²⁸ We observed that in some cases, HER2-positive breast carcinomas stained positively for basal epithelium cytokeratin markers (i.e., CK5/6), often displaying checkerboard-type intratumoral heterogeneity, as previously described by Laakso and colleagues in “basoluminal” HER2⁺ breast carcinomas,¹⁷³ and sometimes displaying a uniform or almost uniform CK5/6-positive staining that mimicked the pattern of pure basal-like breast carcinomas. Indeed, the subgroup of basal/HER2-positive breast carcinomas displayed histological features of basal-like tumors (i.e., poor differentiation, high-grade, geographic necrosis, pushing margins of invasion, syncytial arrangement of tumor cells, ribbon- or festoon-like architecture, squamous metaplasia, stromal lymphocytic infiltrates, high mitotic index and p53 positivity). Preliminary analyses in our own series of “luminal-HER2⁺,” “HER2⁺” and “basal/HER2” appear to definitely corroborate that the occurrence of the basal-like phenotype in HER2-overexpressing breast carcinomas, as defined by the immunohistochemical expression of basal CKs, is associated with an aggressive phenotype that significantly shortens the overall survival of HER2-positive breast cancer patients (unpublished observations).

Given that predictive studies are a specific form of prognostic study in which the outcome of interest is the response to treatment, and the application of the test is to guide the suitability of patients for specific therapy, these limitations make the identification of predictive signatures more difficult than the identification of prognostic factors and require prospectively controlled clinical trials.¹⁶¹ Beyond confirming that the basal-like phenotype in HER2-overexpressing breast cancer can predict worse overall survival, we are currently evaluating whether the simple immunohistochemical identification of the basal/HER2-positive subtype alone may be sufficient to predict primary resistance to trastuzumab, which may have crucial implications for the current use of targeted therapies in patients originally identified as suitable for trastuzumab-based therapies based on their HER2 gene-amplified phenotype. We have already determined that the SNAIL factors can negatively regulate the expression of cytokeratins associated with luminal epithelial cells;^177,178 therefore, we can identify a significant overlap between mesenchymal transition-like signatures and the occurrence of the basal-like immunophenotype in HER2-positive breast carcinomas. If forthcoming studies definitely confirm that an accurate integration of CSC cellular states and EMT-related biomarkers with the currently available breast cancer molecular taxonomy may significantly improve our ability to a priori inform whether patients belonging to HER2 subtypes that are differentially enriched with a “mesenchymal transition signature” (i.e., basal/HER2-positive vs. luminal/HER2-positive) would distinctly benefit from trastuzumab-based therapy ab initio, larger retrospective studies that include hundreds of samples from each of the major subtypes, as defined by nodal status and a small number of markers (ER/PR/HER2 and basal cytokeratins), should be conducted to generate both a more definitive HER2-positive breast cancer taxonomy and also better predictive molecular forecasting signatures to be validated both prospectively and in samples from trastuzumab-based clinical trials.

Acknowledgments

This work was financially supported through funding from the Instituto de Salud Carlos III [Ministerio de Sanidad y Consumo, Fondo de Investigación Sanitaria (FIS), Spain, Grants CP05–00090 and PI06–0778 and RD06–0020–0028]; the Fundación Científica de la Asociación Española Contra el Cáncer (AECC, Spain); and the Ministerio de Ciencia e Innovación (SAF2009–11579, Plan Nacional de I+D+ I, MICINN, Spain). A.V.-M. received the Sara Borrell post-doctoral contract (CD08/00283, Ministerio de Sanidad y Consumo, FIS, Spain). S.C. received a Research Fellowship (Formación de Personal Investigador, FPI) from the Ministerio de Ciencia e Innovación (MICINN, Spain).