During cancer progression, malignant cells undergo epithelial-mesenchymal transitions (EMT) and mesenchymal-epithelial transitions (MET) as part of a broad invasion and metastasis program. We previously observed MET events among lung metastases in a preclinical model of prostate adenocarcinoma that suggested a relationship between epithelial plasticity and metastatic spread. We thus sought to translate these findings into clinical evidence by examining the existence of EMT in circulating tumor cells (CTC) from patients with progressive metastatic solid tumors, with a focus on men with castration-resistant prostate cancer (CRPC) and women with metastatic breast cancer. We showed that the majority (>80%) of these CTCs in patients with metastatic CRPC coexpress epithelial proteins such as epithelial cell adhesion molecule (EpCAM), cytokeratins (CK), and E-cadherin, with mesenchymal proteins including vimentin, N-cadherin and O-cadherin, and the stem cell marker CD133. Equally, we found that more than 75% of CTCs from women with metastatic breast cancer coexpress CK, vimentin, and N-cadherin. The existence and high frequency of these CTCs coexpressing epithelial, mesenchymal, and stem cell markers in patients with progressive metastases has important implications for the application and interpretation of approved methods to detect CTCs. Mol Cancer Res; 9(8); 997–1007. ©2011 AACR.

This article is featured in Highlights of This Issue, p. 977

Most metazoan cells can be classified as either epithelial or mesenchymal based on morphology, behavior, and molecular signatures. In adult animals, epithelial and mesenchymal cells usually remain in one phenotypic state; that is, epithelial cells do not change their properties and become mesenchymal. During development, however, epithelial cells of the early embryo give rise to all 3 embryonal layers (endoderm, mesoderm, and ectoderm), which include mesenchymal cells (1). Therefore, these early embryonal cells have the ability to transition between epithelial and mesenchymal states, a property we define as epithelial plasticity (for a slightly different definition see ref. 2). Indeed, observations in embryos showed epithelial-mesenchymal transitions (EMT) and mesenchymal-epithelial transitions (MET; ref. 3), which may be viewed, perhaps naively, as forward and reverse directions of the same reaction mechanism (for review, see refs. 2, 4).

Although it is convenient to present EMT/MET as a reversible reaction between binary states, there are suggestions that intermediate states exist and that these may play important roles. The importance of EMT and MET in cancer progression is now widely, albeit not unanimously, accepted (see recent reviews refs. 4–8). EMT is observed in human cancer cells in vitro and in xenografts, and in the leading edge of invasive carcinomas in vivo (5, 6). In human prostate carcinoma, loss of E-cadherin expression and overexpression of N-cadherin, which indicates the presence of an EMT, independently correlates with high Gleason score and systemic and metastatic recurrence after surgery, linking EMT to more aggressive clinical behavior (9–12). In addition, recent studies have shown the importance of N-cadherin expression during castration-resistant metastatic progression in preclinical models of prostate cancer and in human metastases. These translational studies have suggested a link between loss of epithelial markers, gain of mesenchymal markers, and the induction of signaling pathways that promote survival and androgen-receptor independent growth (13). In breast cancer, a similar link has been established between EMT markers in primary and disseminated bone marrow tumor cells and aggressive clinical behavior (14–18). Likewise, evidence for MET was obtained from microscopic analysis of colorectal carcinoma metastases, which adopted epithelial characteristics of the noninvasive regions of the primary tumor (19). In prostate cancer, attachment of metastatic cells to bone cells correlates with expression of E-cadherin (20). These and many other studies describe the existence of these transitions during carcinogenesis and raise questions about their functional importance. There is strong evidence that EMT is important for metastatic behavior and chemoresistance (18, 21); however, the importance of MET has been more difficult to ascertain. Previously, we found that the preponderance of MET events among lung metastases in rats bearing AT3 rat prostate adenocarcinoma tumors suggested an important functional relationship between the capacity to revert to a more epithelial state and metastatic growth in the lung parenchyma (22, 23).

A strict view of epithelial plasticity in cancer posits that a mesenchymal-like state reached post-EMT is the driver of malignant fitness. Indeed, there is strong evidence that the mesenchymal properties of invasiveness and motility are required for metastases (see above) and that EMT leads to expression of cancer stem cell markers, including CD44 (24). Nonetheless, observations above suggest that mesenchymal properties per se are not sufficient for optimal malignant behavior (19, 22, 23, 25). A broader interpretation suggests that the ability to easily transition between epithelial-like and mesenchymal-like states, which we define as phenotypic plasticity, may be linked to stem cell–like properties and is a more important determinant of aggressive metastatic behavior than the properties of the end states. In a preclinical model examining the importance of stem cells and cancer growth, subcutaneous injection of CD133-positive cells, which are predicted to be stem cell like, into immunodeficient mice causes tumor growth, whereas injection of CD133-negative cells do not (26). A recent clinical study that measured mRNAs coding for stem-like markers in the bloodstream of patients with resected colorectal cancer found that the expression of CD133, CEA, and cytokeratin (CK) RNAs was associated with recurrent disease and an overall poor prognosis (27).

Therefore, we posit that the most plastic cells will be those that inhabit transitional or intermediate states with properties of both epithelium and mesenchyme, and that these transitional cells will be particularly malignant and stem-like. To test this proposal in human disease, we sought evidence for markers of both mesenchymal and stemness phenotypes in circulating tumor cells (CTC) from patients with metastatic castration-resistant prostate cancer (CRPC) and metastatic breast cancer.

Analysis of human CTCs

Patients eligible for the CTC biomarker protocols included (i) men with mCRPC, with metastatic progression by prostate-specific antigen (PSA; 2 consecutive rises over nadir separated by more than 1 week) or radiologic criteria (Response Evaluation Criteria in Solid Tumors or new bone scan lesions), a PSA 5 or more, age 18 years or older; or (ii) women with metastatic breast cancer with disease progression or with initiation of a new systemic therapy, who were older than 18 years of age, and who were at least 7 days from treatment with an anthracycline-containing regimen. All subjects provided informed consent as part of an Institutional Review Board-approved prospective clinical protocol. Blood (15 mL) was collected from patients and processed within 48 hours at the Duke University, Department of Molecular Pathology and Clinical Pathology Laboratory using the Cell Search System (Veridex, Raritan). Veridex profile kits, which isolate epithelial cell adhesion molecule (EpCAM)-positive cells by using a ferromagnetic immunoabsorbtion assay without additional staining, were used to collect CTCs. An additional tube was collected and processed in parallel for CTC enumeration by using the Veridex CellSearch method using the standard test kit. Following profile kit processing, the isolated cells were either processed immediately or stored overnight in 4% paraformaldehyde (PFA) and processed the next day. An initial wash using a bench top magnet to enrich the EpCAM-bound cells was conducted to further isolate CTCs, with resuspension of the cell pellet after magnet release into 100 μL PBS. Immunostaining was done on teflon-coated slides. Briefly, cells were pipetted into the wells of the slides and left to settle for ∼30 minutes followed by standard immunostaining procedures with careful aspiration to minimize cell loss at room temperature. Following 4% PFA fixation, permeabilization with PBT (PBS with 0.2% v/v Triton), and blocking with 10% goat serum for 30 minutes, triple immunostaining was done using CD45 antibody labeled with Alexa 647, CK labeled with Alexa 555 and either Vimentin (BD Biosciences), N-cadherin (BD Biosciences), O-cadherin (Invitrogen), or CD133 (Novus Biologics) labeled with Alexa 488. Supplementary Table S1 provides specifics on the control cells, dilutions, and products used. Nuclear staining with 4′,6-diamidino-2-phenylindole (DAPI) was then done. A CTC was defined as an intact cell, containing a nucleus and expressing CK but lacking CD45 expression. Control cells were evaluated in parallel with each patient sample for immunofluorescent staining intensity and scoring (see Supplementary Table S1 for controls used and Supplementary Fig. S1 for images of these real-time controls). Human peripheral blood mononuclear cells (PBMC), obtained by Ficoll purification of buffy coats from normal American Red Cross donors, were kindly provided by Micah Luftig (Duke) and used as positive control cells for CD45 expression and negative controls for CK and cadherin proteins. In addition, the coexpression of E-cadherin (BD Biosciences) with N-cadherin on EpCAM-captured, CD45-negative DAPI-positive cells, irrespective of their CK expression, was examined. In this situation, all CD45-negative cells were manually enumerated and the proportion of CD45-negative nucleated cells that expressed E- or N-cadherin were scored. To ensure that the antibodies against E- and N-cadherin (Supplementary Table S1) did not cross-react with their respective antigens, a series of control cells with known E- and N-cadherin expression levels were examined by Western blot analysis against E- and N-cadherin and actin as a loading control.

The slides were mounted with gel/mount media (Biomeda). The slides were analyzed with an Olympus IX 71 epifluorescence microscope, and images were acquired using an Olympus DP70 digital camera. Image processing was done with DP Controller software (Olympus). All fields on each slide were analyzed sequentially, with each CK-positive–nucleated cell that was CD45 negative being counted as a CTC. Standardized exposure times optimized for each antibody were used consistently throughout the analysis for case and control cells.

Immunohistochemical analysis of metastases

Under the same informed consent protocol, men undergoing CTC collection additionally consented to have a radiologic-guided metastatic biopsy for analysis of biomarker expression by immunohistochemistry (IHC). Samples were obtained through core needle biopsies during light sedation, and immediately formalin fixed and paraffin embedded. For analysis, slides were deparaffinized, rehydrated, and endogenous peroxidase was inactivated for 30 minutes in 0.3% H2O2 (hydrogen peroxide) in methanol. Specific antigen retrieval steps were done for individual antigens. Three markers were evaluated by IHC: vimentin (M7020; Dako, 1:150; antigen retrieval with pepsin treatment at 37°C for 15 minutes), CK cocktail (18-0132; Invitrogen, 1:50 and 349205; BD Biosciences 1:50, antigen retrieval with pepsin treatment at 37°C for 15 minutes) and CD45 (M0701; Dako, 1:200; antigen retrieval with sodium citrate 10 mmol/L, pH 6.0 at 100°C for 30 minutes). Primary antibody was incubated for 60 minutes at room temperature. Dako Envision horseradish peroxidase secondary antibody was used for 30 minutes at room temperature and the signal was detected with 3,3′-diaminobenzidine reagent (Vector kit SK 4100). Slides were counter stained with hematoxylin and eosin and assessed by a trained pathologist for expression using appropriate positive (localized prostate tissue microarray sections) and negative controls (mock antibody) for each marker.

Statistical analyses

We used simple descriptive statistics to estimate the prevalence of mesenchymal and CD133 antigen coexpression on EpCAM-captured CTCs, summarizing these findings on both an intraindividual level and within group (CRPC and breast cancer) level. To compare CTC count (standard CellSearch method) against the proportion of CTCs that coexpress vimentin, N- and O-cadherin, or CD133, linear regression analysis was done. Goodness of fit was tested by ANOVA.

To examine the coexpression of mesenchymal and/or stemness antigens on CTCs from patients with metastatic disease, we took advantage of the existing Food and Drug Administration (FDA)-approved capture method to initially capture and isolate cells from whole blood. CTCs have both independent prognostic and predictive significance in multiple epithelial malignancies, including mCRPC and metastatic breast cancer (28, 29), and can be collected, isolated, and analyzed for a variety of biomarkers relevant to cancer biology (30–32). The approved technology for CTC capture relies on the expression of EpCAM on the surface of epithelial cells, and thus currently measured CTCs must be epithelial like. Yet, these cells have escaped from the primary tumor, possibly as a result of an EMT/invasiveness program, and may have mesenchymal properties.

To test for the existence of mesenchymal-like CTCs, blood was collected from 41 men with mCRPC and 16 women with metastatic breast cancer (see baseline characteristics for the patients in Table 1 and Supplementary Table S2) and CTCs were processed by using the CellSearch EpCAM-based immunocapture method and profiled for expression of CD45 (PTPRC) (a leukocyte marker), CK, and E-cadherin (CDH1; epithelial markers), vimentin (VIM), N-cadherin (CDH2), and O-cadherin (CDH11; mesenchymal markers), and CD133 (a stem cell marker) by immunofluorescence (IF; refs. 10, 33; Supplementary Table S1). Leukocytes were defined as nucleated (DAPI positive), CD45-positive and CK-negative cells (Fig. 1A), whereas CTCs were defined as nucleated (DAPI positive), CD45-negative and CK-positive cells (Fig. 1B–G). Among CTCs, we identified subsets of cells that additionally expressed vimentin (Fig. 1C) or N-cadherin (Fig. 1D and E) or O-cadherin (Fig. 1F and G). Real-time positive and negative control cells were used at the time of individual patient CTC analysis to determine positive or negative IF expression using identical exposure, antibody concentration, and camera settings without alteration. Real-time control cell images for each corresponding antigen and subject depicted in Figures 1, 3, and 4 are provided in Supplementary Figure S1.

Table 1.

Baseline demographic and clinical characteristics of the men with metastatic CPRC in this study (n = 41)

Demographics n = 41 
Median age, y (range) 71 (50–89) 
Race, ethnicity  
White, non-Hispanic 73% 
Black, non-Hispanic 27% 
Baseline disease history  
Median Gleason score (range) 8 (5–10) 
Median baseline PSA (ng/dL, range) 248.9 (14.0–13,419.5) 
Median baseline pain (range)a 0 (0–7) 
Median Karnofsky performance status (range) 90 (60–100) 
Median number of prior hormonal therapies (range) 2.5 (0–5) 
Prior chemotherapy 68% 
Prior bisphosphonates 73% 
Sites of metastatic disease  
Visceral (lung + liver) 54% 
Lymph node only 0% 
Metastatic to bone:  
Metastatic to bone with lymph nodes (no visceral metastases) 24% 
Metastatic to bone without lymph nodes (no visceral metastases) 22% 
Demographics n = 41 
Median age, y (range) 71 (50–89) 
Race, ethnicity  
White, non-Hispanic 73% 
Black, non-Hispanic 27% 
Baseline disease history  
Median Gleason score (range) 8 (5–10) 
Median baseline PSA (ng/dL, range) 248.9 (14.0–13,419.5) 
Median baseline pain (range)a 0 (0–7) 
Median Karnofsky performance status (range) 90 (60–100) 
Median number of prior hormonal therapies (range) 2.5 (0–5) 
Prior chemotherapy 68% 
Prior bisphosphonates 73% 
Sites of metastatic disease  
Visceral (lung + liver) 54% 
Lymph node only 0% 
Metastatic to bone:  
Metastatic to bone with lymph nodes (no visceral metastases) 24% 
Metastatic to bone without lymph nodes (no visceral metastases) 22% 

aPain is scored as a linear analog scale (0–10 range).

Abbreviations: CRPC, castration-resistant prostate cancer; PSA, prostate specific antigen.

Figure 1.

Coexpression of epithelial and mesenchymal proteins in CTCs from men with metastatic castration-resistant prostate cancer (mCRPC). All panels represent merged images derived from phase/DAPI, CD45/DAPI, CK/DAPI, and either vimentin (Vim)/DAPI, N-cadherin (N-cad)/DAPI expression, or O-cadherin (O-cad)/DAPI as indicated. Shown are examples of (A) a leukocyte with CD45 expression, (B) a CTC with no vimentin expression, (C) a CTC with vimentin expression, (D) a CTC with N-cadherin expression, (E) 3 CTCs, 2 with N-cadherin expression (arrowheads), (F) a CTC with O-cadherin expression and a nearby leukocyte, and (G) an additional CTC with O-cadherin expression. Scale bars represent 20 μm and were added from an image taken at identical magnification and resolution. Control cells were assayed in parallel at the same time of CTC collection and analysis with each set of patient samples and are shown in Supplementary Figure S1.

Figure 1.

Coexpression of epithelial and mesenchymal proteins in CTCs from men with metastatic castration-resistant prostate cancer (mCRPC). All panels represent merged images derived from phase/DAPI, CD45/DAPI, CK/DAPI, and either vimentin (Vim)/DAPI, N-cadherin (N-cad)/DAPI expression, or O-cadherin (O-cad)/DAPI as indicated. Shown are examples of (A) a leukocyte with CD45 expression, (B) a CTC with no vimentin expression, (C) a CTC with vimentin expression, (D) a CTC with N-cadherin expression, (E) 3 CTCs, 2 with N-cadherin expression (arrowheads), (F) a CTC with O-cadherin expression and a nearby leukocyte, and (G) an additional CTC with O-cadherin expression. Scale bars represent 20 μm and were added from an image taken at identical magnification and resolution. Control cells were assayed in parallel at the same time of CTC collection and analysis with each set of patient samples and are shown in Supplementary Figure S1.

Close modal
Figure 2.

Expression of vimentin and CK in prostate cancer metastases. Images are taken from a CT-guided targeted bone metastasis biopsy at the same time as CTCs were collected and evaluated for vimentin coexpression by immunofluorescence as described. Images are from patient 6 (top) and patient 7 (bottom), with CK (left), and vimentin (right) expression assayed by IHC (20× magnification). Scale bars in the CK panels represent 50 μm and were added from an image taken at identical magnification and resolution.

Figure 2.

Expression of vimentin and CK in prostate cancer metastases. Images are taken from a CT-guided targeted bone metastasis biopsy at the same time as CTCs were collected and evaluated for vimentin coexpression by immunofluorescence as described. Images are from patient 6 (top) and patient 7 (bottom), with CK (left), and vimentin (right) expression assayed by IHC (20× magnification). Scale bars in the CK panels represent 50 μm and were added from an image taken at identical magnification and resolution.

Close modal
Figure 3.

E-cadherin and N-cadherin coexpression among CD45-negative–nucleated cells from men with metastatic CRPC. All panels represent merged immunofluorescent images derived from phase/DAPI, CD45/DAPI, E-cadherin (E-cad)/DAPI, and N-cadherin (N-cad)/DAPI expression as indicated. A, a leukocyte with CD45 expression, (B) a CD45-negative–nucleated cell, with E-cadherin and N-cadherin coexpression, (C) a CD45-negative–nucleated cell with N-cadherin expression and no E-cadherin expression, (D) a CD45-negative–nucleated cell with E-cadherin expression and no N-cadherin expression, and (E) a nucleated cell lacking expression of CD45, E-cadherin, and N-cadherin. Scale bars represent 20 μm and were added from an image taken at identical magnification and resolution. Control cells were assayed in parallel at the same time of CTC collection and analysis with each set of patient samples and are shown in Supplementary Figure S1.

Figure 3.

E-cadherin and N-cadherin coexpression among CD45-negative–nucleated cells from men with metastatic CRPC. All panels represent merged immunofluorescent images derived from phase/DAPI, CD45/DAPI, E-cadherin (E-cad)/DAPI, and N-cadherin (N-cad)/DAPI expression as indicated. A, a leukocyte with CD45 expression, (B) a CD45-negative–nucleated cell, with E-cadherin and N-cadherin coexpression, (C) a CD45-negative–nucleated cell with N-cadherin expression and no E-cadherin expression, (D) a CD45-negative–nucleated cell with E-cadherin expression and no N-cadherin expression, and (E) a nucleated cell lacking expression of CD45, E-cadherin, and N-cadherin. Scale bars represent 20 μm and were added from an image taken at identical magnification and resolution. Control cells were assayed in parallel at the same time of CTC collection and analysis with each set of patient samples and are shown in Supplementary Figure S1.

Close modal
Figure 4.

Expression of the stem-like cell marker CD133 by immunofluorescence in CTCs from men with mCRPC. All panels represent merged images derived from phase/DAPI, CD45/DAPI, CK/DAPI, and CD133/DAPI expression as indicated. Shown are examples of (A) a leukocyte with CD45 expression, (B) a CD133-negative CTC, (C) a CD133-positive CTC, and (D) an additional example of a CD133-positive CTC. Scale bars represent 20 μm and were added from an image taken at identical magnification and resolution. Control cells were assayed in parallel at the same time of CTC collection and analysis with each set of patient samples and are shown in Supplementary Figure S1.

Figure 4.

Expression of the stem-like cell marker CD133 by immunofluorescence in CTCs from men with mCRPC. All panels represent merged images derived from phase/DAPI, CD45/DAPI, CK/DAPI, and CD133/DAPI expression as indicated. Shown are examples of (A) a leukocyte with CD45 expression, (B) a CD133-negative CTC, (C) a CD133-positive CTC, and (D) an additional example of a CD133-positive CTC. Scale bars represent 20 μm and were added from an image taken at identical magnification and resolution. Control cells were assayed in parallel at the same time of CTC collection and analysis with each set of patient samples and are shown in Supplementary Figure S1.

Close modal

Among men with mCRPC, we found that CTCs coexpressed vimentin and CK in 10/10 (100%) patients, and by this criterion 108/126 (86%) of enumerated CTCs co-expressed these markers (Fig. 1B–C, Table 2 and Supplementary Table S3, and also see Supplementary Fig. S1 for control cells and Supplementary Fig. S2 for additional examples). Biopsies of bony metastases done within 1 week of initial CTC collection in 2 of these patients (patients 6 and 7 in Table 1) revealed no vimentin expression in the CK-positive tumor foci, but strong vimentin expression in the surrounding bone stroma, which lacks CK expression (Fig. 2). These same patients had CTCs taken at the same time as the CT-guided tumor biopsy that commonly expressed coexpressed CK and vimentin. These findings are consistent with invasion and metastasis by CTCs that subsequently undergo EMT/MET or exist in a transitional state; an alternative explanation may be that vimentin coexpression may be heterogeneous in metastases, similar to CTC expression.

Table 2.

Prevalence of EMT and stemness antigen marker expression by immunofluorescence in CTCs (intact, nucleated CK+, CD45, DAPI + cells) from the circulation of men with metastatic CRPC and women with metastatic breast cancer

AntigennMarker-positive CTCs (%)Patients with marker-positive CTCs (%)
Vimentin (CRPC) 10 108/126 (86%) 10/10 (100%) 
Vimentin (BC) 10 65/97 (65%) 7/10 (70%) 
N-Cadherin (CRPC) 11 205/244 (84%) 11/11 (100%) 
N-Cadherin (BC) 78/95 (82%) 4/6 (67%) 
O-Cadherin (CRPC) 107/120 (89%) 6/6 (100%) 
CD133 (CRPC) 11 127/153 (83%) 9/11 (82%) 
AntigennMarker-positive CTCs (%)Patients with marker-positive CTCs (%)
Vimentin (CRPC) 10 108/126 (86%) 10/10 (100%) 
Vimentin (BC) 10 65/97 (65%) 7/10 (70%) 
N-Cadherin (CRPC) 11 205/244 (84%) 11/11 (100%) 
N-Cadherin (BC) 78/95 (82%) 4/6 (67%) 
O-Cadherin (CRPC) 107/120 (89%) 6/6 (100%) 
CD133 (CRPC) 11 127/153 (83%) 9/11 (82%) 

NOTE: Columns on the right indicate the number of manually scored CTCs scoring positive for each marker and the number of patients in each group who have at least 1 CTC that stains positive for a given marker.

Abbreviations: CRPC, castration-resistant prostate cancer; BC, breast cancer.

Among the next cohort of 11 men with mCRPC, we found CTCs coexpressing N-cadherin and CK in 11/11 (100%) patients, and by this criterion 205/244 (84%) of CTCs were identified as coexpressing these markers (Fig. 1D and E, Table 2 and also see Supplementary Table S3, Supplementary Fig. S1 for real-time control cells and Supplementary Fig. S3 for additional examples). The expression of N-cadherin among CTCs varied from undetectable, determined by the real-time negative controls described in Supplementary Table 1, to very strong (Fig. 1D). This variation was observed among CTCs from individual patients as can be seen by examination of 3 CTCs from patient 11 (Fig. 1E). Although we noted heterogenous vimentin expression among patient and control leukocytes, we did not observe N-cadherin expression among patient or control leukocytes.

Among 10 women with metastatic breast cancer, 9 had detectable CTCs and of these, we found evidence of vimentin coexpression in 7 (78%) patients, and 55/88 CTCs overall (63%) coexpressed vimentin (Supplementary Fig. S2, Table 2 and Supplementary Table 4, and controls in Supplementary Fig. S1). Among another 6 women with detectable CTCs and metastatic breast cancer, 4 had evidence of CK and N-cadherin coexpression, and overall 78/95 CTCs (82%) had N-cadherin expression, with significant heterogeneity in expression in a given individual (Table 2, Supplementary Fig. S3 and controls in Supplementary Fig. S1). These data indicate that the majority of CTCs in patients with metastatic breast cancer and mCRPC coexpress epithelial (EpCAM and CK) and mesenchymal (vimentin, N-cadherin) markers, and thus exist in a transitional phenotypic state, similar to that observed in our preclinical models.

Given these findings, we next sought to better characterize the dual expression of the epithelial E-cadherin with the more mesenchymal N-cadherin on circulating CD45-negative–nucleated cells. Western blot analysis of control cells (T47D, BT549, PC-3, and PBMC cells) indicated a lack of cross-reactivity of our E-, N-, or O-cadherin antibodies against their respective antigens (data not shown), providing assurance that these antibodies are able to measure the intended target antigen. In an additional cohort of 6 men with progressive metastatic CRPC, we characterized all CD45-negative cells that were initially captured by using the standard CellSearch method. Given the limited number of proteins that may be evaluated using IF in an individual cell, we examined the presence of both E- and N-cadherin expression on these CD45-negative cells but were not able to additionally measure CK expression and therefore cannot define a CTC as described above. Interestingly, as shown in Table 3, Figure 3, and Supplementary Figure 1, we found clear evidence for 4 distinct subtypes of CD45-negative cells, including those cells that coexpressed both E- and N-cadherin (20%–71% of cells examined in a given subject), cells that expressed E- but not N-cadherin (2%–22% of cells), cells that expressed N- but not E-cadherin (0%–45% of cells), and cells that lacked both cadherin family members (25%–48% of cells). The proportion of CD45-negative cells that expressed N-cadherin was similar to the proportion that expressed E-cadherin (61% vs. 53%), although this varied greatly in individual men (48%–72% vs. 23%–75%, respectively). Finally, we noted that the subtype of E-cadherin–negative/N-cadherin–positive cells was relatively rare among these EpCAM-captured cells except in 2 men (subjects 39 and 40) who had progressive disease on their most recent treatment with the mTOR inhibitor temsirolimus as part of an experimental protocol. Although these analyses are limited to those cells captured based on EpCAM expression, the results indicate that CTCs coexpress epithelial and mesenchymal proteins and suggest the existence of currently undetected CTCs with reduced or absent epithelial markers. Methods that evaluate whole blood for non-leukocyte, nucleated cellular populations without initial EpCAM-based capture may thus be able to identify these additional cells.

Table 3.

Prevalence of E-cadherin and N-cadherin expression on CD45-negative–nucleated cells isolated from the circulation using an EpCAM-based ferrofluid (CellSearch method) among men with progressive metastatic CRPC

Subject numberClinical CTCCD45-negative cellsE + N + cells (%)E − N − cells (%)E + N − cells (%)E − N + cells (%)
35 24 17 (71%) 6 (25%) 1 (4%) 0 (0%) 
36 55 23 11 (48%) 8 (35%) 4 (17%) 0 (0%) 
37 112 122 82 (67%) 30 (25%) 4 (3%) 6 (5%) 
38 1000 65 27 (42%) 22 (34%) 14 (22%) 2 (3%) 
39a 45 44 9 (30%) 21 (48%) 1 (2%) 13 (30%) 
40a 16 83 18 (22%) 26 (31%) 2 (22%) 37 (45%) 
Summary 1235 361 164 (45%) 113 (31%) 26 (7%) 58 (16%) 
Subject numberClinical CTCCD45-negative cellsE + N + cells (%)E − N − cells (%)E + N − cells (%)E − N + cells (%)
35 24 17 (71%) 6 (25%) 1 (4%) 0 (0%) 
36 55 23 11 (48%) 8 (35%) 4 (17%) 0 (0%) 
37 112 122 82 (67%) 30 (25%) 4 (3%) 6 (5%) 
38 1000 65 27 (42%) 22 (34%) 14 (22%) 2 (3%) 
39a 45 44 9 (30%) 21 (48%) 1 (2%) 13 (30%) 
40a 16 83 18 (22%) 26 (31%) 2 (22%) 37 (45%) 
Summary 1235 361 164 (45%) 113 (31%) 26 (7%) 58 (16%) 

NOTE: Clinical CTC = enumeration by using the FDA-approved method. Each column represents the number and percentage of CD45-nucleated cells that were identified based on dual marker expression.

aIndicates 2 patients with metastatic CRPC who were progressing on therapy with the mTOR inhibitor temsirolimus.

Given that prostate cancer has a tendency to metastasize to bone, we next hypothesized that adhesion molecules that favor an osteoblastic tumor microenvironment may be visualized on CTCs from men with CRPC. O-cadherin has been recently linked to metastasis to bone in preclinical models of prostate cancer (34, 35), and we thus sought to examine its coexpression with CK in CTCs. Indeed, we found coexpression in 6 of 6 (100%) of men with CRPC and detectable CTCs by EpCAM-based ferromagnetic capture, and O-cadherin was expressed in 64%–96% of detectable CTCs (e.g., shown in Fig. 1F and G, summarized in Table 2, and Supplementary Table S3 with controls in Supplementary Fig. S1). These findings suggest that prostate cancer CTCs coexpress adhesion molecules that may promote homing to bone and homotypic binding to osteoblastic cells.

Given the expression of the stem cell–associated antigen CD133 in putative prostate cancer stem cells and other cancer stem cell populations (27, 33, 36), we investigated CD133 expression in CTCs from men with mCRPC. We found CD133 to be expressed in 11/11 (100%) men with CTCs, and in 127/153 (83%) of CTCs from these men (Fig. 4, Table 2, Supplementary Fig. S4 and controls in Supplementary Fig. 1). These data suggest that CTCs from patients with common epithelial malignancies coexpress both epithelial and mesenchymal markers, suggesting that EMT/MET transitions may be contributing to metastatic progression. In addition, in men with metastatic CRPC, the coexpression of the stemness antigen CD133 in the majority of CTCs suggests that these cells may have acquired properties of stemness during their migration into the bloodstream (21, 24).

In these studies, we have identified biomarkers suggestive of epithelial plasticity and stemness in CTCs from patients with common metastatic epithelial malignancies, including breast and prostate cancer. The identification of both epithelial and mesenchymal phenotypes among CTCs in a significant subset of patient samples offers several important clinical opportunities. These data suggest that CTCs may undergo phenotypic changes from epithelial to more mesenchymal transitional states during metastatic transit, whereas metastases themselves may be more epithelial in phenotype and marker expression. Our findings also suggest that in addition to cells expressing both epithelial and mesenchymal markers, there may be an unknown number of CTCs that are more mesenchymal-like and thus are EpCAM negative. These cells will be missed by the FDA-approved CellSearch method, the Adna Test (AdnaGen AG) system, and current microfluidic technologies, which enrich for CTCs by immunoabsorbtion of cells expressing MUC1 or EpCAM (37). Indeed, recent studies in breast cancer have suggested that “normal” type breast cancer cell lines that overexpress EMT and stem cell antigens (CD44+, CD24) may lack EpCAM and are thus not detectable by currently approved CTC detection systems (38). Therefore, it is possible that the number of CTCs in patients with metastatic cancer is much higher than currently appreciated.

It is well appreciated that cells induced to undergo EMT activate stem cell pathways (24). Indeed, a recent study showed a striking relationship between prostate cancer–associated fibroblast chemokine expression of interleukin-6 or TGF-β and the acquisition of tumor invasiveness, metastatic propensity, EMT antigen expression, and stemness characteristics (39). Our findings suggest that CTCs captured by using an epithelial-based capture assay, the current FDA standard method, express the stemness marker CD133. These findings are consistent with the results of Mani and colleagues (24) showing a relationship between EMT and stemness states in breast cancer models. It is tempting to speculate that these CTCs may represent transitional cells with both epithelial and mesenchymal phenotypes, and this heterogeneity or plasticity may also extend to stem cell markers and to treatment-induced effects. For example, we observed an increased predominance of N-cadherin expression in CTCs from 2 men who were progressing on their most recent treatment with an mTOR inhibitor, suggesting that certain systemic agents may alter CTC phenotype. An alternative explanation to this transitional state theory is that tumor heterogeneity for these markers exist, without the need to transition phenotypically. Further preclinical and clinical studies of metastatic tumors in which the EMT/MET process is experimentally disrupted or induced may shed further light on these observations.

Our working model of plasticity also predicts that cells with maximal stem cell character, which by definition will be highly malignant, should display both epithelial and mesenchymal traits, because they inhabit intermediate states in the epithelial-mesenchymal axis. Here, we show that patients with metastatic breast and prostate carcinomas have CTCs that exist in phenotypic states that may be intermediate to epithelial and mesenchymal states. Although the enumeration of EpCAM-captured CTCs or CD133-positive CTCs correlates with disease progression, there is great heterogeneity to the number of CTCs isolated from individual patients with metastatic carcinomas, and thus a further refinement in the methods of CTC detection by using EMT antigen–based capture methods may result in improved prognostication CTC identification (27–29). Aktas and colleagues recently showed that a population of cells enriched in CTCs expressed RNAs encoding mesenchymal markers; however, this study could not prove coexpression of epithelial and mesenchymal markers in the same cell (40). On the basis of the results presented here, further studies to explore methods to capture cells based on these markers in addition to EpCAM are warranted to investigate their relevance to metastatic progression and chemoresistance.

There are several inherent limitations to this work. First, we have not shown true stemness among CTCs, which would require further experimental evidence of serial clonal passage and transplantation, or prolonged culture of CTCs from patients. This is not technically feasible with current methods. Our markers of stemness are correlative in nature only, and may be associated with properties other than stemness. We also acknowledge that coexpression of epithelial and mesenchymal markers does not, in itself, represent plasticity, as we are unable to observe this dynamic in vivo in patients. Our clinical observations suggest plasticity based on the coexpression in real time on CTCs during the process of metastasis, and the lack of expression of vimentin in paired metastases from the same patients. The importance of this plasticity to highly aggressive metastatic behavior can only be tested through experimental manipulation of preclinical systems in which either EMT/MET is prevented; future experiments will need to address this issue. Finally, these studies have not correlated coexpression of EMT factors on CTCs with clinical outcomes; these prognostic studies require large appropriately powered studies and patients with long-term follow-up, such as has been recently reported with CD133-positive colorectal CTCs and postoperative outcomes (27). Our findings, however, suggest that the measurement of CTCs collected through both EpCAM-enriched and EMT antigen–enriched methods may complement each other in providing prognostic or predictive information during systemic therapy that should be prospectively evaluated.

Finally, CTCs expressing mesenchymal or stem like markers expression, which comprise the majority of cells isolated in this study, and additional cells that may go undetected due to EpCAM loss, represent a therapeutic problem. It has been well documented that EMT alters drug sensitivity (21, 41, 42) and it has been challenging to direct therapy to cancer cells with stem cell–like properties, perhaps because of their recalcitrance to undergo apoptosis (43). Although recent studies suggest both a screening method and actual compounds (e.g., salinomycin) that can selectively target cancer stem cells (21), these aggressive cells still represent a formidable challenge. Our findings suggest that these cell types may be highly prevalent among patients with metastatic epithelial tumors, and suggest methods for the improved detection of these cells in vivo to assist in developing novel therapeutic strategies.

A.J. Armstrong, S. Oltean, D. George, and M.A. Garcia-Blanco are listed as inventors in a related patent application (application number PCT/US10/50233) filed on September 24, 2010.

A.J. Armstrong, S. Oltean, D. George, and M.A. Garcia-Blanco conceived of the original study. A.J. Armstrong and M.A. Garcia-Blanco directed the research. G. Kemeny, S. Oltean, R. Bitting, and M.S. Marengo carried out laboratory experiments. A.J. Armstrong, J. Turnbull, C.I. Herold, P.K. Marcom, and D. George carried out clinical procedures and recruitment. Although all coauthors contributed to the writing of the manuscript, A.J. Armstrong, M.S. Marengo, and M.A. Garcia-Blanco did the majority of writing and editing.

We thank Drs. M. Dewhirst, P. Febbo, J Somarelli, and J. Pearson for important discussions and Dr. M. Luftig for providing buffy coats from normal donors.

NIGMS grant R01 GM63090 (M.A. Garcia-Blanco); National Cancer Institute grant R01 CA127727, (M.A. Garcia-Blanco). A.J. Armstrong was supported by Robert B. Goergen Prostate Cancer Foundation Young Investigator Award, the Duke Cancer Institute K12 program (5K12-CA-100639-05, PI; H.K. Lyerly), Department of Defense Physician Research Training Award (W81XWH-10-1-0483), the American Cancer Society Pilot Grant program, and the H.L. Kirkpatrick Foundation. M.S. Marengo was supported by National Research Service Awards T32-CA059365 and F32 CA142095.

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.

1.
Hay
ED
,
Zuk
A
. 
Transformations between epithelium and mesenchyme: normal, pathological, and experimentally induced
.
Am J Kidney Dis
1995
;
26
:
678
90
.
2.
Zeisberg
M
,
Neilson
EG
. 
Biomarkers for epithelial-mesenchymal transitions
.
J Clin Invest
2009
;
119
:
1429
37
.
3.
Acloque
H
,
Adams
MS
,
Fishwick
K
,
Bronner-Fraser
M
,
Nieto
MA
. 
Epithelial-mesenchymal transitions: the importance of changing cell state in development and disease
.
J Clin Invest
2009
;
119
:
1438
49
.
4.
Kalluri
R
,
Weinberg
RA
. 
The basics of epithelial-mesenchymal transition
.
J Clin Invest
2009
;
119
:
1420
8
.
5.
Blick
T
,
Widodo
E
,
Hugo
H
,
Waltham
M
,
Lenburg
ME
,
Neve
RM
, et al
Epithelial mesenchymal transition traits in human breast cancer cell lines
.
Clin Exp Metastasis
2008
;
25
:
629
42
.
6.
Hugo
H
,
Ackland
ML
,
Blick
T
,
Lawrence
MG
,
Clements
JA
,
Williams
ED
, et al
Epithelial–mesenchymal and mesenchymal–epithelial transitions in carcinoma progression
.
J Cell Physiol
2007
;
213
:
374
83
.
7.
Klymkowsky
MW
,
Savagner
P
. 
Epithelial-mesenchymal transition: a cancer researcher's conceptual friend and foe
.
Am J Pathol
2009
;
174
:
1588
93
.
8.
Weinberg
RA
. 
Twisted epithelial-mesenchymal transition blocks senescence
.
Nat Cell Biol
2008
;
10
:
1021
3
.
9.
Contreras
HR
,
Ledezma
RA
,
Vergara
J
,
Cifuentes
F
,
Barra
C
,
Cabello
P
, et al
The expression of syndecan-1 and -2 is associated with Gleason score and epithelial-mesenchymal transition markers, E-cadherin and beta-catenin, in prostate cancer
.
Urol Oncol
2010
;
28
:
534
40
.
10.
Gravdal
K
,
Halvorsen
OJ
,
Haukaas
SA
,
Akslen
LA
. 
A switch from E-cadherin to N-cadherin expression indicates epithelial to mesenchymal transition and is of strong and independent importance for the progress of prostate cancer
.
Clin Cancer Res
2007
;
13
:
7003
11
.
11.
Tomita
K
,
van Bokhoven
A
,
van Leenders
GJ
,
Ruijter
ET
,
Jansen
CF
,
Bussemakers
MJ
, et al
Cadherin switching in human prostate cancer progression
.
Cancer Res
2000
;
60
:
3650
4
.
12.
Zhau
HE
,
Odero-Marah
V
,
Lue
HW
,
Nomura
T
,
Wang
R
,
Chu
G
, et al
Epithelial to mesenchymal transition (EMT) in human prostate cancer: lessons learned from ARCaP model
.
Clin Exp Metastasis
2008
;
25
:
601
10
.
13.
Tanaka
H
,
Kono
E
,
Tran
CP
,
Miyazaki
H
,
Yamashiro
J
,
Shimomura
T
, et al
Monoclonal antibody targeting of N-cadherin inhibits prostate cancer growth, metastasis and castration resistance
.
Nat Med
2010
;
16
:
1414
20
.
14.
Blanco
MJ
,
Moreno-Bueno
G
,
Sarrio
D
,
Locascio
A
,
Cano
A
,
Palacios
J
, et al
Correlation of Snail expression with histological grade and lymph node status in breast carcinomas
.
Oncogene
2002
;
21
:
3241
6
.
15.
Dumont
N
,
Wilson
MB
,
Crawford
YG
,
Reynolds
PA
,
Sigaroudinia
M
,
Tlsty
TD
. 
Sustained induction of epithelial to mesenchymal transition activates DNA methylation of genes silenced in basal-like breast cancers
.
Proc Natl Acad Sci U S A
2008
;
105
:
14867
72
.
16.
Sarrio
D
,
Rodriguez-Pinilla
SM
,
Hardisson
D
,
Cano
A
,
Moreno-Bueno
G
,
Palacios
J
. 
Epithelial-mesenchymal transition in breast cancer relates to the basal-like phenotype
.
Cancer Res
2008
;
68
:
989
97
.
17.
Willipinski-Stapelfeldt
B
,
Riethdorf
S
,
Assmann
V
,
Woelfle
U
,
Rau
T
,
Sauter
G
, et al
Changes in cytoskeletal protein composition indicative of an epithelial-mesenchymal transition in human micrometastatic and primary breast carcinoma cells
.
Clin Cancer Res
2005
;
11
:
8006
14
.
18.
Yang
J
,
Mani
SA
,
Donaher
JL
,
Ramaswamy
S
,
Itzykson
RA
,
Come
C
, et al
Twist, a master regulator of morphogenesis, plays an essential role in tumor metastasis
.
Cell
2004
;
117
:
927
39
.
19.
Brabletz
T
,
Jung
A
,
Reu
S
,
Porzner
M
,
Hlubek
F
,
Kunz-Schughart
LA
, et al
Variable beta-catenin expression in colorectal cancers indicates tumor progression driven by the tumor environment
.
Proc Natl Acad Sci U S A
2001
;
98
:
10356
61
.
20.
Saha
B
,
Arase
A
,
Imam
SS
,
Tsao-Wei
D
,
Naritoku
WY
,
Groshen
S
, et al
Overexpression of E-cadherin and beta-catenin proteins in metastatic prostate cancer cells in bone
.
Prostate
2008
;
68
:
78
84
.
21.
Gupta
PB
,
Onder
TT
,
Jiang
G
,
Tao
K
,
Kuperwasser
C
,
Weinberg
RA
, et al
Identification of selective inhibitors of cancer stem cells by high-throughput screening
.
Cell
2009
;
138
:
645
59
.
22.
Oltean
S
,
Febbo
PG
,
Garcia-Blanco
MA
. 
Dunning rat prostate adenocarcinomas and alternative splicing reporters: powerful tools to study epithelial plasticity in prostate tumors in vivo
.
Clin Exp Metastasis
2008
;
25
:
611
9
.
23.
Oltean
S
,
Sorg
BS
,
Albrecht
T
,
Bonano
VI
,
Brazas
RM
,
Dewhirst
MW
, et al
Alternative inclusion of fibroblast growth factor receptor 2 exon IIIc in Dunning prostate tumors reveals unexpected epithelial mesenchymal plasticity
.
Proc Natl Acad Sci U S A
2006
;
103
:
14116
21
.
24.
Mani
SA
,
Guo
W
,
Liao
MJ
,
Eaton
EN
,
Ayyanan
A
,
Zhou
AY
, et al
The epithelial-mesenchymal transition generates cells with properties of stem cells
.
Cell
2008
;
133
:
704
15
.
25.
Brabletz
T
,
Jung
A
,
Spaderna
S
,
Hlubek
F
,
Kirchner
T
. 
Opinion: migrating cancer stem cells: an integrated concept of malignant tumour progression
.
Nat Rev Cancer
2005
;
5
:
744
9
.
26.
Ricci-Vitiani
L
,
Lombardi
DG
,
Pilozzi
E
,
Biffoni
M
,
Todaro
M
,
Peschle
C
, et al
Identification and expansion of human colon-cancer-initiating cells
.
Nature
2007
;
445
:
111
5
.
27.
Iinuma
H
,
Watanabe
T
,
Mimori
K
,
Adachi
M
,
Hayashi
N
,
Tamura
J
, et al
Clinical significance of circulating tumor cells, including cancer stem-like cells, in peripheral blood for recurrence and prognosis in patients with Dukes' stage B and C colorectal cancer
.
J Clin Oncol
2011
;
29
:
1547
55
.
28.
Cristofanilli
M
,
Budd
GT
,
Ellis
MJ
,
Stopeck
A
,
Matera
J
,
Miller
MC
, et al
Circulating tumor cells, disease progression, and survival in metastatic breast cancer
.
N Engl J Med
2004
;
351
:
781
91
.
29.
de Bono
JS
,
Scher
HI
,
Montgomery
RB
,
Parker
C
,
Miller
MC
,
Tissing
H
, et al
Circulating tumor cells predict survival benefit from treatment in metastatic castration-resistant prostate cancer
.
Clin Cancer Res
2008
;
14
:
6302
9
.
30.
Attard
G
,
Swennenhuis
JF
,
Olmos
D
,
Reid
AH
,
Vickers
E
,
A'Hern
R
, et al
Characterization of ERG, AR and PTEN gene status in circulating tumor cells from patients with castration-resistant prostate cancer
.
Cancer Res
2009
;
69
:
2912
8
.
31.
Scher
HI
,
Jia
X
,
de Bono
JS
,
Fleisher
M
,
Pienta
KJ
,
Raghavan
D
, et al
Circulating tumour cells as prognostic markers in progressive, castration-resistant prostate cancer: a reanalysis of IMMC38 trial data
.
Lancet Oncol
2009
;
10
:
233
9
.
32.
Wülfing
P
,
Borchard
J
,
Buerger
H
,
Heidl
S
,
Zänker
KS
,
Kiesel
L
, et al
HER2-positive circulating tumor cells indicate poor clinical outcome in stage I to III breast cancer patients
.
Clin Cancer Res
2006
;
12
:
1715
20
.
33.
Zhu
L
,
Gibson
P
,
Currle
DS
,
Tong
Y
,
Richardson
RJ
,
Bayazitov
IT
, et al
Prominin 1 marks intestinal stem cells that are susceptible to neoplastic transformation
.
Nature
2009
;
457
:
603
7
.
34.
Chu
K
,
Cheng
CJ
,
Ye
X
,
Lee
YC
,
Zurita
AJ
,
Chen
DT
, et al
Cadherin-11 promotes the metastasis of prostate cancer cells to bone
.
Mol Cancer Res
2008
;
6
:
1259
67
.
35.
Huang
CF
,
Lira
C
,
Chu
K
,
Bilen
MA
,
Lee
YC
,
Ye
X
, et al
Cadherin-11 increases migration and invasion of prostate cancer cells and enhances their interaction with osteoblasts
.
Cancer Res
2010
;
70
:
4580
9
.
36.
Vander
Griend DJ
,
Karthaus
WL
,
Dalrymple
S
,
Meeker
A
,
DeMarzo
AM
,
Isaacs
JT
. 
The role of CD133 in normal human prostate stem cells and malignant cancer-initiating cells
.
Cancer Res
2008
;
68
:
9703
11
.
37.
Nagrath
S
,
Sequist
LV
,
Maheswaran
S
,
Bell
DW
,
Irimia
D
,
Ulkus
L
, et al
Isolation of rare circulating tumour cells in cancer patients by microchip technology
.
Nature
2007
;
450
:
1235
9
.
38.
Sieuwerts
AM
,
Kraan
J
,
Bolt-de Vries
J
,
van der Spoel
P
,
Mostert
B
,
Martens
JW
, et al
Molecular characterization of circulating tumor cells in large quantities of contaminating leukocytes by a multiplex real-time PCR
.
Breast Cancer Res Treat
2009
;
118
:
455
68
.
39.
Giannoni
E
,
Bianchini
F
,
Masieri
L
,
Serni
S
,
Torre
E
,
Calorini
L
, et al
Reciprocal activation of prostate cancer cells and cancer-associated fibroblasts stimulates epithelial-mesenchymal transition and cancer stemness
.
Cancer Res
2010
;
70
:
6945
56
.
40.
Aktas
B
,
Tewes
M
,
Fehm
T
,
Hauch
S
,
Kimmig
R
,
Kasimir-Bauer
S
. 
Stem cell and epithelial-mesenchymal transition markers are frequently overexpressed in circulating tumor cells of metastatic breast cancer patients
.
Breast Cancer Res
2009
;
11
:
R46
.
41.
Thomson
S
,
Buck
E
,
Petti
F
,
Griffin
G
,
Brown
E
,
Ramnarine
N
, et al
Epithelial to mesenchymal transition is a determinant of sensitivity of non-small-cell lung carcinoma cell lines and xenografts to epidermal growth factor receptor inhibition
.
Cancer Res
2005
;
65
:
9455
62
.
42.
Yauch
RL
,
Januario
T
,
Eberhard
DA
,
Cavet
G
,
Zhu
W
,
Fu
L
, et al
Epithelial versus mesenchymal phenotype determines in vitro sensitivity and predicts clinical activity of erlotinib in lung cancer patients
.
Clin Cancer Res
2005
;
11
:
8686
98
.
43.
Li
HZ
,
Yi
TB
,
Wu
ZY
. 
Suspension culture combined with chemotherapeutic agents for sorting of breast cancer stem cells
.
BMC Cancer
2008
;
8
:
135
.

Supplementary data