Epithelial–Mesenchymal Transition and Stem Cell Markers in Patients with HER2-Positive Metastatic Breast Cancer

The presence of circulating tumor cells (CTC) in the blood of patients with metastatic breast cancer (MBC) is an independent prognostic and predictive factor (1–9). Several techniques have been used to isolate, enumerate, and characterize CTCs, all of which rely on procedures to enrich for epithelial cells that express the epithelial cell adhesion molecule, EpCAM/CD326. The U.S. Food and Drug Administration–cleared CellSearch system defines CTCs as EpCAM/CD326⁺ nucleated cells that co-express cytoplasmic cytokeratins (CK8, CK18, and CK19) and not the common leukocyte antigen, CD45 (1). Recently, we found that CTCs detected by CellSearch are heterogeneous, with different prognostic values among breast cancer subtypes (10). In particular, CTCs lack predictive value in patients with HER2-amplified (HER2⁺) tumors treated with targeted therapy (11, 12). One possible explanation for this observation is the loss of the expressions of EpCAM (13) on CTCs that may be undergoing epithelial–mesenchymal transition (EMT-CTCs) and the high efficacy of treatment with HER2-targeted therapy eliminating HER2⁺ CTCs (11, 14).

Earlier, we reported that a majority of disseminated tumor cells express a cancer stem cell (CSC) phenotype (15) and others have suggested that a subset of disseminated tumor cells can be induced to become CSCs (8, 16). To initiate metastases, disseminated tumor cells with characteristics of CSCs must acquire the ability to intravasate into and survive in the peripheral circulation. To acquire the properties of extravasation, migration, invasion, and metastasis seeding, epithelial cancer cells must undergo multiple biochemical and structural changes that enable them to acquire a mesenchymal cell phenotype. Through oncogenic EMT type 3, cancer cells shed many of their epithelial characteristics, detach from epithelial sheets, and undergo drastic morphologic and phenotypic alterations (17–19). Loss of functional E-cadherin adhesion protein expression, leading to loss of cell polarity, is considered the hallmark of EMT. Members of the E-cadherin repressor/EMT inducers of the zinc-finger transcription factor SNAIL, Slug, ZEB1, ZEB2, FoxC2, and TWIST have been extensively evaluated (18). In addition to loss of epithelial characteristics, EMT frequently coincides with the acquisition of motility, invasiveness, cytoskeletal protein changes (vimentin and α-smooth muscle actin expression), altered adhesion receptor expression (switching from E- to N-cadherin), and proteinase secretion (e.g., matrix metalloproteinases; ref. 20). These findings suggest that determining the expressions of epithelial cell surface markers, such as EpCAM (13), and cytoskeletal proteins, such as CKs (21), is not the ideal strategy for evaluating and characterizing heterogeneous CTCs with a mesenchymal phenotype.

Independent of these findings, studies of neoplastic tissues have shown that the presence of self-renewing, stem-like cells also known as CSCs (22) constitute a small minority of neoplastic cells in tumors and are defined operationally by their ability to seed new tumors (23). During the tumor metastasis process, which is often enabled by EMT (24), disseminated cancer cells seem to acquire a self-renewal capability, similar to that exhibited by stem cells, to spawn macroscopic metastases. This observation raises the possibility that the EMT process, which enables cancer cell dissemination, imparts a self-renewal capability to disseminating cancer cells. Therefore, researchers have addressed the possible association between EMT and stem cell formation. In many respects, cells that have undergone EMT behave similar to stem cells isolated from normal or neoplastic cell populations (25). Moreover, evidence shows that HER2 plays a role in mammary carcinogenesis by regulating the stem cell population (26) and that HER2 overexpression in multiple breast cancer cell lines increases the fraction of ALDH1⁺ cells, which has a greater ability to invade and form tumors when transplanted into immunodeficient mice (27). The correlation between breast CSCs and EMT has been shown in vitro and here we will try to investigate how the CTCs transit from primary tumor to enter peripheral blood and traverse to distant organs.

In this prospective translational study, we isolated epithelial and nonepithelial cells from blood of patients with HER2⁺ MBCs and characterized their epithelial differentiation, transcript levels of transcription factors known to induce EMT (EMT-TF), and stem cell features to better understand CTC seeding.

This is a prospective translational study (LAB08-0079) conducted in the Department of Hematopathology, in collaboration with the Department of Breast Medical Oncology, The University of Texas MD Anderson Cancer Center (Houston, TX), and was approved by the Institutional Review Board (IRB).

All patients with MBCs with immunohistochemical analysis (IHC) or FISH HER2⁺ disease, regardless of previous treatment history were eligible for this study. Between August 2009 and March 2010, we recruited 28 patients with MBCs with HER2⁺ tumors before starting a new line of therapy and 2 with HER2⁻ tumors and HER2⁺ CTCs (28). At the same time of patient enrollment, 20 healthy women older than 18 years old (healthy donors) were recruited as control subjects; all patients with MBCs and healthy donors gave informed consent according to the IRB-approved protocol.

All tumor specimens were assigned a study identification number that was distinct from the patient's medical record number. The histologic type and grade of invasive disease were coded according to the World Health Organization classification system (29) and modified Black nuclear grading system, respectively (30). Tumor specimens were analyzed in a Clinical Laboratory Improvement Act–certified clinical pathology laboratory for estrogen and progesterone receptor status by IHC staining. Patients with at least one positive hormonal receptor were considered hormone receptor–positive. HER-2/neu status was determined using IHC analysis or the PathVysion FISH Kit (Abbott Molecular; ref. 31) For FISH, 30 nuclei were scored per sample, and the number of HER2 and CEP17 signals was recorded. A ratio of HER2 to CEP17 >2.2 was defined as gene amplification; polysomy 17 was defined as ≥3 CEP17 signals per nucleus (average for 30 cells). Specimens with no evidence of staining on IHC analysis or no gene amplification by FISH were considered negative for HER-2/neu. Specimens that had a 3+ stain intensity on IHC analysis or gene amplification by FISH were considered HER2+.

Patients were required to have clinical and radiological evidence of MBCs, with measurable or evaluable disease, before initiating therapy. All patients underwent imaging studies, laboratory evaluations, and treatment planning at MD Anderson Cancer Center. Disease status was evaluated every 6 to 12 weeks using the same imaging techniques, depending on the treatment schedule, until loss to follow-up or death. Response to therapy was evaluated according to Response Evaluation Criteria in Solid Tumors (RECIST) v1.1 (32).

Patients provided 30 mL of peripheral blood in heparin anticoagulant before initiating a new treatment to isolate tumor cells. The blood samples were processed immediately or no later than 6 hours after phlebotomy. Contemporaneously, 7.5 mL of peripheral blood was collected in a Veridex CellSave tube for CTC isolation and enumeration using CellSearch (Veridex, LLC).

The CellSearch system was used to detect CTCs in 7.5 mL of peripheral blood, as previously described (1). In brief, peripheral blood samples were subjected to EpCAM⁺ cell enrichment with anti-EpCAM–coated ferrous particles. Thereafter, EpCAM-enriched cells were stained with 4′,6-diamidino-2-phenylindole (DAPI) and reacted with anti-CD45 antibody to identify leukocytes and with a cocktail of antibodies against CK8, CK18, or CK19 (33). CTCs were defined as EpCAM⁺-nucleated cells (DAPI-stained nucleus) that lacked surface expression of CD45 but expressed cytoplasmic CK (33). Samples were considered positive if they had ≥5 CTCs per 7.5 mL of peripheral blood.

Peripheral blood mononuclear cells (PBMC) were isolated from 30 mL of peripheral blood using a Ficoll-Hypaque density gradient, washed twice with sterile PBS, and counted. Patient samples with more than 30 × 10⁶ PBMCs were enriched for epithelial cells (EpCAM/CD326⁺, step 1) using magnetic bead separation methods, and the residual PBMCs were further depleted of leukocytes (CD326⁻CD45⁻, step 2). Patient samples with less than 30 × 10⁶ PBMCs and healthy donor blood were directly separated into leukocyte-enriched (CD45⁺) and -depleted (CD45⁻) cell fractions (step 2). The reason for not conducting step 1 separation in samples with less than 30 × 10⁶ PBMCs was the lower yield of cells of interest from each of the depletion steps than in samples that contained more than 30 × 10⁶ PBMCs. The inability to collect sufficient numbers of cells of interest after each depletion step compromised our ability to extract sufficient concentration of RNA for downstream analysis by reverse transcription PCR (RT-PCR). In step 1, the PBMCs were adjusted to a concentration of 10⁷ PBMCs per mL and incubated with 100 μL of magnetic beads coated with anti-CD326 antibody (Miltenyi-Biotec) and 100 μL of FcR blocking reagent (Miltenyi-Biotec) for 30 minutes at 4°C. In step 2, the CD326-depleted PBMCs were incubated with 20 μL of magnetic beads coated with anti-CD45 antibody (Miltenyi-Biotec) for 15 minutes at 4°C. Thereafter, the PBMCs were passed through a magnet-filled column on an AutoMACSPro Cell Separator system (Miltenyi-Biotec) using the positive selection protocol (POSSELD program) to enrich for EpCAM-positive cells and the negative selection protocol (DEPLETE program) to enrich for CD45-depleted CTCs. The CD45-depleted (CD45⁻) fraction underwent an additional run through the magnet-filled column using the MACS DEPLETES program to prevent any residual contamination of CD45⁺ cells. The mean purity of CD45⁻ fraction was 96.6% (range, 89.4%–99.9%). The efficacy of the separation process is already shown in a previous article (13).

Total RNA was extracted using the miRNeasy Mini Kit (Qiagen), according to the manufacturer's instructions, and purified using the fully automated QIAcube system (Qiagen) to standardize the RNA isolation procedure. All RNA preparation and handling steps were conducted under RNase-free conditions in a laminar flow hood. RNA concentration was determined using a NanoDrop 2000 spectrophotometer (Thermo Scientific), and RNA quality was assessed by evaluating the RNA integrity number using RNA 6000 Pico Kit and Agilent's 2100 Bioanalyzer (Agilent Technologies), according to the manufacturer's instructions. RNA was converted in cDNA using the high-capacity cDNA Reverse Transcription Kit (Applied Biosystems Inc.).

The MCF-7, BT-474, SKBR-3, and MDA-MB-231 breast cancer cells were obtained from the American Type Culture Collection and grown in Dulbecco's Modified Eagle's Medium (DMEM):F-12 culture medium supplemented with 10% FBS (Tissue Culture Biologicals), 1% antibiotics penicillin and streptomycin. The immortalized human mammary epithelial cells HMLE and HMLE-TWIST were kindly provided by S.A. Mani (co-author) and were cultured in a 1:1 mixture of MEGM (Lonza) and DMEM:F-12 supplemented with 5% horse serum, 1% antibiotics penicillin and streptomycin, 10 μg/mL insulin, 10 ng/mL EGF, and 0.5 μg/mL hydrocortisone. RNA extracted from 1 × 10⁶ cells was used as control for EMT-TFs (Supplementary Fig. S1; ref. 34).

Gene expression was quantified with the TaqMan gene expression assay and RT-PCR on a 7900HT Fast Real-Time PCR System (ABI Applied Biosystems), following the manufacturer's instructions. The glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene was used to normalize gene expression levels, and the relative expression of each gene was calculated using the equation |$2^{- \Delta C_{\rm t}}$|⁠, where |${\rm \Delta}C_{\rm t} = {\rm mean}\;C_{{\rm t{-gene}}} - C_{{\rm t{- GAPDH}}}$|⁠, or |$2^{- \Delta \Delta C_{\rm t}}$|⁠, where |${\rm \Delta \Delta}C_{\rm t} \, = \, (C_{{\rm t{-gene}}} \, - \, C_{{\rm t{-GAPDH}}}) \ {\rm patient} \ {\rm sample} - (C_{{\rm t{-gene}}}$||$C_{{\rm t{-GAPDH}}}){\rm HD}\;{\rm sample}$|⁠. The TaqMan assays ZEB1 (Hs00232783_m1), SNAIL1 (Hs00195591_m1), TWIST1 (Hs00361186_m1), TG2 (Hs00190278_m1), GAPDH (Hs02758991_g1), EpCAM (Hs00158980_m1), KRT19 (Hs00761767_s1), and HER2 (Hs00170433_m1) were purchased from ABI.

PBMCs were evaluated for ALDH activity using the ALDEFLUOR assay, according to the manufacturer's protocol (StemCell Technologies), as previously described (15). In brief, 4 × 10⁶ PBMCs were suspended in ALDEFLUOR buffer containing a proprietary ATP-binding cassette transport inhibitor. One third of the cells were reacted with 5 μL of the ALDH inhibitor, diethylamino benzaldehyde, to serve as a negative control. Both the test reaction and negative control were incubated for 35 minutes at 37°C in a 5% CO₂ atmosphere. Purified anti-CD44 monoclonal antibody (BD Pharmingen) was conjugated with Alexa700 using the Zenon antibody labeling Kit (Invitrogen) before being reacted with the ALDEFLUOR-labeled cells. In addition, pre-conjugated antibodies to anti-CD24 (PE) and anti-CD45 (PE-Cy7), both from BD Pharmingen, and anti-CD326 and anti-CD133 (APC, Miltenyi-Biotec) were used to label cells at ambient temperature, in the dark, for 30 minutes. An additional tube of ALDEFLUOR-labeled cells was stained with the appropriate immunoglobulin isotype-matched controls. The stained cells were washed twice with PBS, and the cell pellet was suspended in 200 μL of PBS before being analyzed on a LSR-II flow cytometer capable of discriminating 6-color fluorescence emission signals (BD Biosciences). Cellular debris was excluded from the analysis on the basis of low-forward light scatter. In the ALDEFLUOR⁺ epithelial cell population, a subset of CSCs was defined as cells with a CD44⁺CD24^lo and CD133⁺ phenotype.

Thirty patients with MBCs were enrolled in this pilot study that was designed to determine the feasibility of isolating RNA from CTCs for the detection of EMT-TF genes. We excluded the 2 patients with HER2⁻ tumor from this analysis. Patient characteristics were summarized using the median (range) for continuous variables and frequency (percentage) for categorical variables. The Fisher exact test was used to assess the association between overexpression of EMT-TFs and other patient characteristics. Although the expression and the data are unlikely to yield sufficient data to relate to clinical outcome, we analyzed overall survival (OS) rate at median follow-up in the 17 patients according to the CTC gene expression profiling. OS duration was defined as the time from baseline blood draw to the date of death. All living patients were censored at the last follow-up date. P < 0.05 was deemed statistically significant. All statistical analyses were conducted using SPSS 19 software (IBM).

To determine whether CTC samples overexpressed gene transcripts, we established the following cutoff points: in the CD326⁺ cell fraction, because the EpCAM/CD326⁺ cell fraction was not available in healthy donors, EMT-TF and HER2 genes were considered elevated if the |$2^{- \Delta C_{\rm t}}$| value of each gene was higher than the mean |$2^{- \Delta C_{\rm t}}$| + 3 SD in the CD45⁺ cell fraction. In the CD45⁻ cell fraction, an EMT-TF was considered elevated if |$2^{- \Delta \Delta C_{\rm t}}$| value was equal to or higher than 2. EMT-TF gene expression medians in the CD326⁺ cell fraction were compared using the Mann–Whitney test.

Twenty-eight patients and 20 healthy donors were included in the study. Patient characteristics are shown in Table 1. Ten patients received trastuzumab plus chemotherapy (docetaxel or paclitaxel) as first-line treatment for metastatic disease, before study enrollment. Among the 28 patients included in this study, 17 had more than 30 × 10⁶ PBMCs and underwent separation (step 1) plus depletion (step 2) and 11 patients had fewer (only step 2).

In the EpCAM/CD326⁺ cell fractions, 9 of 17 (52.9%) patients presented with low expression of EpCAM or KRT19 in the CD326⁺ cell fraction. We combined the EpCAM and KRT19 expressions and found that 7 (41.2%) patients had low epithelial gene expression in their CD326⁺ cell fractions. At least one EMT-TF was elevated in 15 of 17 (88.2%) patients. TWIST1, SNAIL1, and TG2 were elevated in 13, 11, and 11 patients, respectively. ZEB1 was not elevated in any patients. HER2 was elevated in 16 of 17 (94.1%) patients. The median expression levels (⁠|$2^{- \Delta C_{\rm t}}$|⁠) of TWIST1, SNAIL1, TG2, and HER2 in the EpCAM/CD326⁺ cell fractions were higher than those in the CD45⁺ fractions (P < 0.05). Figure 1 summarizes the mRNA expression levels of the analyzed genes in the CD326⁺ and CD45⁺ cell fractions.

For TWIST1, SNAIL1, and ZEB1 gene expression, samples were available from 28 patients, respectively. For TG2, and HER2/neu gene expression, samples were available from 26 patients. The TWIST1 gene was not elevated (⁠|$2^{- \Delta \Delta C_{\rm t}}$|< 2) in any of the patients; SNAIL1 was elevated in 5 of 28 (17.9%) patients; ZEB1 was elevated in 7 of 28 (25%) patients; and TG2 was elevated in 8 of 26 (30.8%) patients. At least one EMT-TF gene was elevated in 17 of 28 (60.7%) patients. HER2/neu was elevated in 7 of 26 (26.9%) patients. Figure 2 summarizes the results for the CD45⁻ cell fractions.

CTC counts, as determined using CellSearch, were available for 22 of the 28 patients. Nineteen (86.4%) evaluable patients had less than 5 CTCs per 7.5 mL of peripheral blood, and only 3 (13.6%) had ≥5 CTCs. Two of the 3 patients with ≥5 CTCs were about to receive first-line trastuzumab at the time of study enrollment. The CellSearch CTC numbers were correlated with EMT-TF expression levels in the CD326⁺ (n = 10) and CD45⁻ (n = 22) cell fractions (Table 2). In 19 patients with less than 5 CTCs per 7.5 mL of peripheral blood, 7 of 9 (77.7%) presented with an elevated EMT-TF transcript in the CD326⁺ cell fraction and 9 of 19 (47.4%) in the CD45⁻ fraction. In particular, even in patients with no CTCs (n = 12), we detected at least one EMT-TF transcript that was elevated in 66.7% of CD326⁺ cell fractions and in 50% of CD45⁻ cell fractions. All patients with ≥5 CTCs per 7.5 mL of peripheral blood had elevated EMT-TF transcripts in their CD45⁻ cell fractions.

Ten patients (group 1) underwent baseline assessments before starting first-line therapy with trastuzumab plus chemotherapy for metastatic disease. The remaining 18 patients (group 2) had been treated with at least one line of trastuzumab-based therapy plus chemotherapy. As shown in Table 2, we detected at least one elevated EMT-TF transcript in the CD326⁺ cell fractions of groups 1 and 2 (83.3% vs. 90.9% of patients, respectively, P = 0.596). On the other hand, we found at least one elevated EMT-TF transcript in the CD45⁻ cell fraction in 8 of 9 (88.9%) patients in group 1 but only in 8 of the 18 (44.4%) patients in group 2 (P = 0.033).

Twenty-eight patients were evaluated for CSC markers by multiparameter flow cytometry. We identified a CSC population (ALDEFLUOR⁺ CD44⁺CD24^lo) in the EpCAM⁺ (ALDEFLUOR⁺ CD44⁺CD24^lo/CD326⁺CD45⁻) and EpCAM⁻/CD45⁻ cell fractions (ALDEFLUOR⁺ CD44⁺CD24^lo/CD326⁻CD45⁻; Fig. 3A). A mean of 0.1% of PBMCs were ALDEFLUOR⁺ (range, 0.008%–0.57%); among these, 4.8% (range, 0.3%–17.2%) were CD326⁺CD45⁻ and 59% (range, 43%–82%) were CD326⁻CD45⁻. A mean of 3% (range, 0.6%–13.5%) of ALDEFLUOR⁺ cells was CD44⁺CD24^lo/CD326⁺CD45⁻ and 48.7% (range, 37.2%–71.9%) of these cells were CD44⁺CD24^lo/CD326⁻ CD45⁻. After combining the CD44 and CD24 stem cell markers with EMT-TFs and the 2 groups of treatment, we found no particular association among stem cell markers (Fig. 3B). Interestingly, there was an association between CD133 expression and EMT-TFs elevated in CD45⁻ cell fractions. Patients with elevated EMT-TFs (SNAIL1 and ZEB1) in the CD45⁻ cell fraction had a higher expression of CD45⁻ ALDEFLUOR⁺CD133⁺ cells in the blood (P = 0.038). Moreover, HER2⁺ trastuzumab-naive patients had a higher ratio of CD133⁺ stem cells in the blood than did those who were previously treated with trastuzumab (P = 0.042; Fig. 3B).

After a median follow-up duration of 19 months, 20 had experienced disease progression and 8 of 28 patients had expired. Among the 17 who underwent gene expression profiling in the CD326⁺ cell fraction, 9 patients with low EpCAM transcript expression had lower OS and progression-free survival (PFS) rates than those of the 8 patients with high expression (44.4% of low-expression patients died vs. 0% of high-expression patients; P = 0.041; for PFS, 87% vs. 75%, respectively, P = 0.343; Fig. 4).

In this pivotal translational trial, we characterized CTCs from patients with HER2⁺ MBCs, specifically looking for EMT-TFs and CSC features. We successfully isolated CTCs from the peripheral blood of patients with MBCs; these CTCs differed in epithelial differentiation and expressed EMT-TFs. Moreover, we showed that CTCs with EMT features (EMT-CTCs) can be detected even in patients with undetectable CTCs using CellSearch. Specifically, we found that TWIST1 and SNAIL1 transcripts were elevated in the CD326⁺ cell fractions and SNAIL1 and ZEB1 transcripts were elevated in the CD45⁻ cell fractions. These data are in agreement with published reports in which they indicated that a subset of patients with primary breast cancer showed a subtype of CTCs that had undergone EMT could not be detected using currently available detection methods for CTCs (35). Other researchers have found, independently of HER2⁺ subtype, positive expression of the EMT-related markers vimentin and fibronectin in EpCAM⁺/CK^dim/CD45⁻ cells, which do not meet the established definition of a CTC by CellSearch (25, 36). The existence of hybrid CTCs with an epithelial–mesenchymal phenotype was observed also in patients with non–small cell lung cancer (37).

Recent studies (25, 38) have shown an association between EMT and acquisition of CSC characteristics, leading to the hypothesis that cancer cells that undergo EMT are capable of metastasizing through their acquired self-renewal potential, thereby enabling them to spawn the large cell populations that constitute macroscopic metastases. We found that most cells with a CSC phenotype were in the CD45⁻ cell fraction. However, no EMT-TF genes were associated with the putative CSC markers (ALDH⁺ and CD44⁺/CD24^lo/−). This inconsistency could be explained by the fact that some of patients were heavily pretreated and the heterogeneity of patient population and the small number of samples could explain the absence of significant association between stem cell markers and EMT-TFs. Interestingly, we found an association between ALDH⁺CD133⁺ cells and SNAIL1 and ZEB1 transcript expression in the CD45⁻ cell fractions. Moreover, the ability of trastuzumab to reduce the percentage of CSCs was shown in a previous report (27). These findings are consistent with those of Armstrong and colleagues (39) that showed in addition to cells expressing both epithelial and mesenchymal markers, an unknown number of CTCs that are more mesenchymal-like and thus are EpCAM⁻. All these observations lend support to the new metastatic cascade theory postulated by Chaffer and colleagues in which CTCs are plastic and undergo bidirectional interconversions between stem and non-stem compartments (40). In our study, CTCs in peripheral blood of patients with HER2⁺ MBCs are composed of a heterogeneous population of differentiated epithelial cells that express EpCAM that are readily detected by CellSearch, to a CTC with an EMT/stem cell–like phenotype that does not express EpCAM and, hence, cannot be detected by CellSearch. We are cognizant that our study included a small number of patients with only 22 of 28 women analyzed with CellSearch and more studies are needed to understand the biologic characteristics of the metastatic cascade and the possible effect of CSC targeted therapy on clinical outcomes. Another caveat of our study is related to the small quantity of gene material analyzed from a limited number of CTCs. All isolated CTCs were used for the EMT-TF expression and stem cell markers analysis, and no additional assays for protein and DNA assessments were possible.

In conclusion, we showed, for the first time, that a cluster of CTCs with elevated EMT-TFs can be isolated from the peripheral blood of patients with HER2⁺ MBCs and that this cluster of CTCs is not always detectable using the current EpCAM enrichment enumeration methods. We found that patients with EMT-TFs in CTCs had more ALDH⁺CD133⁺ CSC-like cells. Considering the small number of patients enrolled in this study, additional studies are needed to determine whether EMT events and CSCs have prognostic value in patients with HER2⁺ MBCs treated with targeted therapy.

M. Cristofanilli is a consultant/advisory board member of Alere. No potential conflicts of interest were disclosed by the other authors.

Conception and design: V. Valero, S. De Placido, J.M. Reuben

Development of methodology: A. Giordano, H. Gao, E. Cohen, M. Mego, B.-N. Lee, S.A. Mani, J.M. Reuben

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Giordano, H. Gao, S. Anfossi, E. Cohen, S. Tin, C.A. Parker, R.H. Alvarez, V. Valero, F.J. Esteva, J.M. Reuben

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Giordano, S. Anfossi, B.-N. Lee, V. Valero, N.T. Ueno, M. Cristofanilli, J.M. Reuben

Writing, review, and/or revision of the manuscript: A. Giordano, H. Gao, S. Anfossi, E. Cohen, M. Mego, M. De Laurentiis, V. Valero, N.T. Ueno, S. De Placido, S.A. Mani, F.J. Esteva, M. Cristofanilli, J.M. Reuben

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F.J. Esteva

Study supervision: A. Giordano, J.M. Reuben

The authors thank Ann M. Sutton from the Department of Scientific Publications at The University of Texas MD Anderson Cancer Center (Houston, TX) for editing the manuscript.

The study was supported by Human Breast Cancer Stem Cell Surrogates, CA138239-02, NIH/NCI (S.A. Mani, M. Cristofanilli, and J.M. Reuben); State of Texas Rare and Aggressive Breast Cancer Research Program, The Morgan Welch Inflammatory Breast Cancer Research Program and Clinic (H. Gao, S. Tin, R.H. Alvarez, V. Valero, N.T. Ueno, M. Cristofanilli, and J.M. Reuben); Assessment of Circulating Breast Cancer Stem Cells To Predict Recurrent Disease, W81XWH-09-1-0031 01, DOD (E. Cohen, and J.M. Reuben); and University of Naples Federico II and Second University of Naples, PhD Program in Medical and Surgical Oncology and Clinical Immunology (A. Giordano).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.